What the component evidence supports, and what a 240-patient pragmatic trial can and cannot answer
Prepared: 2026-11-02 Subject trial: STRONG — Study on blueberries, protein and exercise for improving frailty and cardiovascular disease — protocol published in BMJ Open (2026), registered as [1]. Status of the subject trial: protocol only. No outcome data exist. Every statement in this review about STRONG describes its design, its registered endpoints, or the arithmetic consequences of its sample size. None describes a result. Evidence base: 74 PubMed-indexed papers, the ClinicalTrials.gov record for the subject trial, and current guideline documents. Every PMID was verified against NCBI E-utilities esummary on the preparation date.
Abstract
Frailty and cardiovascular disease are not two problems that happen to co-occur in old age. They are a single clinical problem viewed from two angles: each independently predicts the other, each accelerates the other, and both converge on the same final common pathway of lost physiological reserve, disability, hospitalisation and death. A three-component package — supervised exercise, supplemental dietary protein, and anthocyanin-rich berries — is now being tested in a 240-patient, 12-month pragmatic randomised superiority trial in Nova Scotia, with a second year of follow-up [1]. The trial’s rationale is that all three components are cheap, already recommended in some form, and plausibly act on overlapping mechanisms: anabolic resistance and sarcopenia, endothelial nitric oxide bioavailability and arterial stiffness, and chronic low-grade inflammation with cellular senescence.
This review takes the trial’s premise seriously and examines each component on its own evidence, then asks what the package can realistically demonstrate. The exercise literature is the strongest of the three: in 1,635 mobility-limited older adults, structured physical activity reduced major mobility disability from 35.5% to 30.1% (HR 0.82, 95% CI 0.69–0.98) [2]; in 1,519 adults with physical frailty and sarcopenia, a multicomponent programme reduced mobility disability from 52.7% to 46.8% in the subgroup with SPPB 3–7 (HR 0.78, 0.67–0.92) [3]. Pooled effects on frailty scores are larger but far less stable (SMD −0.57, I² = 88%, prediction interval crossing the null) [4]. Protein is a modifier rather than a treatment: protein supplementation augments resistance-training gains in lean mass and strength, with the benefit shrinking with age and no further gain above about 1.62 g/kg/day [5], and protein combined with exercise improves muscle mass and grip strength in sarcopenic older adults on very low-certainty evidence [6]. Blueberries produce a reproducible, small improvement in brachial artery flow-mediated dilation — about 1.1–1.5 percentage points across meta-analyses — with inconsistent effects on blood pressure and no evidence of effect on clinical events [7] [8] [9].
The multidomain trial record is sobering. FINGER produced a small cognitive benefit (between-group difference 0.022 Z-score units per year) [10]; MAPT, preDIVA, SPRINT MIND and Look AHEAD were null on their primary endpoints [11] [12] [13] [14]. The pattern is consistent: multidomain interventions move functional and risk-factor surrogates, sometimes move composite cognitive scores by amounts of uncertain clinical importance, and have never been shown to reduce hard cardiovascular events.
Arithmetic, not cynicism, sets the ceiling for STRONG. At 120 participants per arm, a two-sided α of 0.05 and 80% power, the trial can detect a standardised difference of about 0.36 SD, rising to about 0.42 SD with 25% attrition. It has seven registered primary outcome measures, no stated multiplicity strategy, and a single-masked design in which participants — the source of several of those measures — are unblinded. It cannot detect a plausible effect on cardiovascular events: at a 10% annual event rate it would observe roughly 24 events, whereas detecting a 25% relative reduction at 80% power requires about 379. What STRONG can do is establish feasibility, adherence, and the direction and rough magnitude of effects on physical function, frailty instruments and risk-factor surrogates in a population that is under-represented in trials. That is worth doing. It is not the same as showing that blueberries, protein and exercise prevent cardiovascular disease, and clinicians should not let the framing of the question do work the design cannot support.
1. Introduction: one clinical problem, three inexpensive components, and an evidence base that does not line up
1.1 The problem the trial is aimed at
An 82-year-old with heart failure, a slow gait and three chronic conditions is the modal patient in Canadian internal medicine, and the one least likely to be enrolled in the trials that generate the guidelines applied to her. Frailty — however it is operationalised — is the state that makes her fragile to everything: a diuretic, a urinary tract infection, a hip fracture, an angiogram. Cardiovascular disease is simultaneously the most common reason she presents and the most common reason she deteriorates.
The epidemiological association is not subtle. A meta-analysis of 26 cohort studies with more than four million participants found cardiovascular mortality roughly doubled in people with frailty (HR 2.11, 95% CI 1.86–2.40) and substantially elevated even in pre-frailty (HR 1.80, 1.46–2.23), with heterogeneity above 80% and no single study-level factor explaining it [15]. In a meta-analysis of 30 randomised cardiovascular trials pooling 87,711 participants, frail patients had higher rates of major adverse cardiac events (HR 2.33, 1.87–2.91), all-cause mortality (HR 2.34, 1.80–3.05), cardiovascular mortality (HR 1.76, 1.60–1.93) and hospitalisation (HR 2.38, 1.65–3.43) than non-frail patients [16]. The same analysis contains the most clinically important finding in this literature: frailty did not blunt treatment efficacy. Among the frailest participants, trial interventions reduced MACE (HR 0.81, 0.74–0.88) and hospitalisation (HR 0.81, 0.72–0.90), with no reduction in mortality, and improved physical function and quality of life by small amounts (SMD 0.15) without moving frailty scores [16]. Frailty, in other words, is a marker of risk and a reason to treat — not a reason for therapeutic nihilism. That single sentence is the strongest justification for a trial like STRONG, and it comes from cardiology’s own trial database rather than from geriatrics.
1.2 Why these three components
Exercise, protein and anthocyanin-rich food are attractive because they are cheap, familiar, and each has a plausible mechanism that does not require a patent. Exercise addresses the muscle and the vasculature simultaneously; protein addresses the substrate limitation that makes exercise less anabolic in old age; blueberries address the endothelial and inflammatory milieu in which both operate. The mechanistic story is coherent enough to be believable, which is exactly why it needs testing rather than assertion.
The three literatures, however, have almost nothing in common methodologically. Exercise trials measure physical function, mobility disability and falls. Protein trials measure lean mass, strength and, less often, function. Blueberry trials measure flow-mediated dilation, augmentation index and blood pressure — surrogates that sit at a considerable distance from the outcomes patients care about. Almost no trial has combined all three in a frail population with cardiovascular measurement, and the one that is doing so is the subject of this review.
1.3 What STRONG is, and what it is not
STRONG is a pragmatic, parallel-group, randomised controlled superiority trial in 240 adults aged 65 and older (120 women, 120 men) recruited in Nova Scotia, Canada. The intervention is 1 cup (150 g) of blueberries per day, 30 g per day of protein powder (whey isolate, with pea, pumpkin, casein or rice alternatives for participants who do not consume dairy), and three 60-minute multimodal exercise sessions per week supervised by an exercise physiologist. The comparator is usual care. Intervention lasts one year, with follow-up to 24 months. The protocol reports CONSORT and SPIRIT compliance, ethics approval through the Nova Scotia Health Research Ethics Board, and registration as [1]; these design facts are stated in the published protocol and in the registry record [17] [1].
The protocol abstract describes measurement across frailty (frailty scores, balance, falls, life-space, social engagement), cardiovascular risk factors (blood pressure, lipid profile, Framingham risk score), cardiovascular health (6-minute walk, cardiac function), inflammatory markers, physical fitness, diet quality and health-related quality of life, at baseline and 12 months with intermediate time points. It names no primary outcome and reports no effect sizes or power calculation. Where a primary outcome is relevant to this review, it is taken from the registry and labelled as registry-derived — and the registry is more specific, and more revealing, than the abstract.
The ClinicalTrials.gov record (retrieved 2026-11-02; status RECRUITING; study start 2025-12-15 actual; estimated primary completion 2028-06-30; estimated enrolment 240) lists seven primary outcome measures [1]:
- Cardiovascular risk by echocardiography — ejection fraction by M-mode and global longitudinal strain, at baseline and 12 months.
- Inflammatory markers — a multiplex panel of cytokines and related proteins (APRIL/TNFSF13, BAFF, sCD30, sCD163, chitinase-3-like, gp130, interferons, interleukins including IL-6 and IL-10, MMP-1/2/3, osteocalcin, osteopontin and others) at baseline and 12 months.
- Frailty by a deficit-accumulation Frailty Index, at baseline and 3, 6, 9 and 12 months.
- Cardiovascular risk by resting blood pressure, at baseline and 3, 6, 9 and 12 months.
- Cardiovascular risk by lipid profile, with Framingham and Reynolds risk scores calculated from the results, at baseline and 12 months.
- Frailty by the Clinical Frailty Scale, at baseline and 3, 6, 9 and 12 months.
- Frailty by the Pictorial Fit-Frail Scale (PFFS), a self-completed pictorial instrument scored out of 43, at baseline and 3, 6, 9 and 12 months.
Nine secondary outcomes cover lower- and upper-body strength, functional fitness, balance, lower- and upper-body mobility, overall health, a household food-work assessment and the ASA24 dietary assessment. The design is registered as single-masked, with the outcomes assessor masked. The registry records no data monitoring committee.
Those seven co-primary outcomes are the single most important design fact for anyone reading the eventual results, and Section 10 returns to them. For now the point is narrower: the trial is not a test of a drug with one pre-specified endpoint. It is a test of a package against a panel of correlated but non-identical measurements, several of which are self-reported by participants who cannot be blinded.
1.4 Disclosure
The protocol’s competing-interest statement, reproduced here because it is material to how the review should be read, records that KR is co-founder of Ardea Outcomes (DGI Clinical until 2021), which has had contracts with Danone, Hollister, INmune, Novartis, Takeda and the LuMind IDSC Down Syndrome Foundation; that KR has received frailty-lecture honoraria, chairs a DSMB for EIP Pharma, and has sat on NIA-funded and Wake Forest advisory boards; and that the rest of the team reports no conflicts [17]. The intervention is food- and supplement-based, and one investigator has food-industry contracts. That does not invalidate the trial. It does mean that the eventual reporting of a food-industry-adjacent investigator team should be read with the same scrutiny applied to any industry-linked study, and that independent replication matters more than usual.
1.5 What this review does
Sections 2 and 3 deal with the frailty–cardiovascular interface and with the measurement problem that makes frailty trials so hard to interpret. Sections 4 to 6 examine exercise, protein and blueberry polyphenols in turn, with the numbers and the heterogeneity on display. Section 7 asks where the three mechanisms actually converge, and where the convergence is wishful. Section 8 reviews the multidomain trial record honestly, including its failures. Sections 9 to 11 deal with design consequences, endpoints, detectability and safety. Sections 12 and 13 cover delivery, cost and guidelines. Section 14 gives the bottom line for clinicians and the limitations of this review.
2. The frailty–cardiovascular interface: shared biology, shared prognosis, and two fields that measure each other badly
2.1 The association runs in both directions, and both directions are large
The forward direction — frailty predicting cardiac events — is the one cardiologists know. In the 30-trial meta-analysis, frailty more than doubled MACE and all-cause mortality and increased cardiovascular mortality by roughly three-quarters, with essentially no heterogeneity for the cardiovascular mortality estimate (I² = 0%) [16]. In a national cohort of 931,133 patients hospitalised with acute myocardial infarction in England and Wales between 2005 and 2019, 13% had severe frailty, 22% moderate and 36% mild by the SCARF administrative index. Compared with fit patients, severe frailty was associated with a three-fold one-year mortality (adjusted HR 3.01, 2.93–3.10), a 2.6-fold MACE risk, a 3.7-fold heart-failure readmission risk and a 1.85-fold major bleeding risk, with graded associations across mild and moderate frailty [18]. The same cohort documents the therapeutic consequence: severely frail patients were less likely to receive coronary angiography (44.8% vs 69.3%), dual antiplatelet therapy (75.5% vs 93.4%) or referral to cardiac rehabilitation (71.8% vs 89.7%) [18]. Frailty is therefore not only a prognostic marker; it is a gatekeeper that determines whether evidence-based treatment is offered.
The reverse direction — cardiac disease predicting incident frailty — is less familiar and more interesting mechanistically. In a 20-year cohort of 3,998 community-dwelling adults aged 65 and older, baseline heart disease (heart failure, myocardial infarction or angina) independently predicted incident frailty on the FRAIL scale after adjustment for age, diabetes, hypertension, smoking and body-mass index: adjusted HR 1.77 (1.26–2.48) in men and 1.34 (1.00–1.80) in women, over 62,012 person-years [19]. The effect is modest in women and larger in men, and the sex difference is one of the few pieces of evidence the STRONG protocol’s rationale explicitly anticipates.
2.2 Prevalence depends on where you look and what you use
Frailty prevalence in cardiac populations ranges so widely that the number is almost meaningless without its instrument attached. In a systematic review of 158 studies across acute myocardial infarction, chronic coronary disease, heart failure, diabetes, TAVI, emergency departments, intensive care and hip fracture, prevalence ranged from 4.4% to 90% depending on the tool and setting; in acute MI the Clinical Frailty Scale, Fried phenotype and Frailty Index were the most-used instruments, with CFS-associated mortality hazard ratios of 1.90 to 3.32 across studies [20]. In 2,147 elderly patients with severe valvular heart disease, frailty prevalence ranged from 16.8% to 39.3% and pre-frailty from 50.9% to 64.5% depending on which of four instruments was applied, and only the simplified four-item score was consistently associated with one-year death (adjusted HR 1.50, 1.25–1.81) [21]. In heart failure, a review reports frailty in 15–76% of chronic heart failure patients depending on setting and tool, approaching 90% in hospitalised patients with preserved ejection fraction [22]. In acute coronary syndrome patients undergoing PCI, a laboratory-based frailty index classified 25.2% as frail, and frailty predicted readmission (HR 1.95, 1.64–2.32) and all-cause mortality (HR 3.68, 2.68–5.04) after propensity matching, but not recurrent myocardial infarction [23].
Two conclusions follow. First, “frailty prevalence” without an instrument is not a quantity. Second, the choice of instrument is not neutral: it determines who is labelled frail, who is offered what, and — in a trial — who is counted as improved.
2.3 The mechanistic overlap is real but not yet a treatment target
The shared biology is well described and largely plausible: chronic low-grade inflammation, neurohormonal activation, skeletal muscle hypoperfusion and sarcopenia, mitochondrial dysfunction, cellular senescence, and malnutrition with its own inflammatory component. A 2026 integrated review of heart failure and frailty lists these pathways and notes that frailty independently predicts incident heart failure and is associated with progressive deterioration in cardiac structure and function, while heart failure accelerates frailty through inflammation, muscle hypoperfusion, sarcopenia, neurohormonal activation, malnutrition and hospital-associated deconditioning [22]. The same review notes that no trial has yet shown that targeting frailty modifies heart-failure outcomes, and proposes a heart-failure-specific frailty score for validation.
That is the honest state of the mechanistic argument: a plausible, mutually reinforcing loop, with each arrow supported by observational data and none yet shown to be interruptible in a way that changes outcomes. The clinical implication is not that the loop is wrong. It is that mechanism alone cannot justify a treatment recommendation, and a trial that measures inflammatory markers is measuring the loop rather than the outcome.
2.4 How cardiology and geriatrics measure each other
Cardiology has largely adopted short, administrable tools: the Clinical Frailty Scale, the Essential Frailty Toolset, simple four-item scores, and administrative or laboratory-derived indices that can be computed from data already collected. Geriatrics and gerontology have largely retained the Fried phenotype and the deficit-accumulation Frailty Index, both of which require dedicated data collection and both of which have far better psychometric documentation. The two traditions produce different prevalences in the same population — as the valvular heart disease study shows — and different sensitivities to change.
This matters for a trial because the choice of instrument determines what counts as success. A package that improves gait speed and SPPB may leave the Clinical Frailty Scale unchanged, because the CFS is a global clinical judgement anchored on function and comorbidity rather than on measured performance. A package that improves the CFS may leave the Frailty Index unchanged, because the index counts deficits across systems and is dominated by comorbidity that a lifestyle intervention cannot reverse. STRONG measures all three constructs — a deficit index, the CFS, and a self-rated pictorial scale — which is scientifically sensible and analytically dangerous, because three instruments can disagree and the protocol does not say which one wins.
3. Measuring frailty: instruments, responsiveness, and the discordance problem
3.1 The Fried phenotype
The phenotype that defined the field was derived in the Cardiovascular Health Study: 5,317 men and women aged 65 and older, with frailty defined as three or more of unintentional weight loss (10 lb in the past year), self-reported exhaustion, weakness by grip strength, slow walking speed and low physical activity [24]. Prevalence was 6.9% in that community-dwelling cohort, four-year incidence 7.2%, and frailty was independently predictive over three years of incident falls, worsening mobility or ADL disability, hospitalisation and death, with unadjusted hazard ratios from 1.82 to 4.46 [24]. The phenotype’s strengths are its biological coherence and its explicit performance measurement; its weaknesses are that it requires grip strength and gait speed, that it treats the five criteria as equally weighted and dichotomous, and that it captures a narrow slice of the deficit space — a person can be profoundly comorbid and not “frail” by Fried.
3.2 The Clinical Frailty Scale
The CFS was developed as a seven-point clinical judgement tool and validated in 2,305 participants of the Canadian Study of Health and Aging, where it correlated strongly with the Frailty Index (r = 0.80), and each one-category increment was associated with a 21.2% increase in medium-term mortality (95% CI 12.5–30.6%) and a 23.9% increase in institutionalisation (8.8–41.2%) after adjustment; the area under the curve was 0.77 for 18-month and 0.70 for 70-month mortality [25]. It has since been expanded to nine points and is now the default bedside tool in Canadian hospitals and in much of cardiology. Its strength is that a clinician can use it in thirty seconds; its weakness is that it is a judgement, and judgements vary. In 531 older adults undergoing elective noncardiac surgery, agreement between patient self-report and trained-assessor rating was only moderate (weighted kappa 0.58, 95% CI 0.53–0.63), with assessors rating patients systematically 0.45 points higher than patients rated themselves [26]. Both ratings predicted death or new disability at 90 days similarly, which is reassuring for prognosis and troubling for trials: an unblinded participant’s self-rating carries prognostic information and can move with expectation.
3.3 The Frailty Index and the interchangeability assumption
The deficit-accumulation Frailty Index treats frailty as the proportion of accumulated health deficits and assumes, for practical purposes, that deficits are interchangeable. A 2026 analysis of 3,669 adults with cardiovascular disease in NHANES tested that assumption with random survival forests using a 46-item index. Models using the 46 individual items outperformed composite index approaches for all-cause and cardiovascular mortality (C-index 0.71 vs 0.64–0.67), and randomly constructed indices converged with importance-ranked composites only at about 35 items for all-cause mortality and 20 for cardiovascular mortality [27]. In other words, when enough items are included, which items you choose matters little — but a short index chosen for convenience loses prognostic information, and a short index chosen because it predicts a particular outcome is overfitted to that outcome. The same lesson appears in a 17-country comparison of frailty distribution and dynamics, where frailty levels, mortality and rates of deficit accumulation varied widely between countries and “zero-state” analyses captured population differences that did not generalise to people who already had deficits [28].
For a lifestyle trial, the Frailty Index has one further property that matters: it is dominated by comorbidity items that a 12-month exercise and diet programme cannot plausibly reverse. An index that counts arthritis, hypertension, hearing loss and cancer history will move slowly and incompletely in response to improved muscle function. That is a feature for population epidemiology and a liability for intervention trials.
3.4 The Pictorial Fit-Frail Scale
STRONG’s third frailty instrument, the PFFS, is a self-completed pictorial questionnaire with 14 sets of images, scored out of 43, designed to be usable by people with limited literacy or cognition. A Turkish validation in 156 community-dwelling adults aged 65 and older reported high internal consistency (Cronbach α 0.838), near-perfect inter-rater and test-retest reliability (ICC 0.995 and 0.996) and a strong correlation with the CFS (Spearman r = 0.760) [29]. Those psychometrics are encouraging for a self-report instrument. They do not establish responsiveness, and responsiveness is what a trial needs.
3.5 Responsiveness, and why the minimal clinically important difference matters
A statistically significant change in an instrument is not a clinically meaningful one, and the frailty field has few agreed anchors. The clearest numbers exist for physical performance. In moderately frail older adults with fear of falling, the anchor-based minimal clinically important difference for the 6-minute walk distance was 17.8 m, with a distribution-based estimate of 12.9 m; participants who perceived improvement walked 37.3 m further on average, versus 9.3 m in those who did not [30]. For gait speed, the pooled effect of exercise in frail older adults was 0.09 m/s (95% CI 0.02–0.15) over short-term follow-up, with very low certainty [31] — a number that sits at or below most published thresholds for meaningful change, which is one reason frailty exercise trials can be positive on a composite and unimpressive on a single measure. For the SPPB, the SPRINTT trial’s between-group difference of 0.8 points (0.5–1.1) was treated as supportive of a clinically relevant effect [3]; 0.8 points is within the range usually described as small but perceptible.
For the Frailty Index and the CFS there is no widely accepted minimal clinically important difference that this review could retrieve. That is not a trivial gap. It means a trial using these instruments as primary outcomes must either pre-specify an anchor or accept that a statistically significant difference of uncertain size will be reported as a win.
3.6 The discordance problem, stated plainly
Put the instruments side by side and the problem becomes obvious.
| Instrument | Construct | Data source | Strengths | Vulnerabilities in a lifestyle trial |
|---|---|---|---|---|
| Fried phenotype | Five physical criteria | Measured performance + self-report | Biological coherence; strong outcome prediction (HR 1.82–4.46) [24] | Requires grip and gait; dichotomous; ignores comorbidity |
| Clinical Frailty Scale | Global clinical judgement, 1–9 | Clinician rating | Fast; validated (r = 0.80 with FI; AUC 0.77 at 18 months) [25] | Judgement variance; assessor must be masked; ceiling effects in robust volunteers |
| Frailty Index | Proportion of deficits | Item count from history, labs, function | Precise risk stratification; within-phenotype discrimination [32] | Dominated by irreversible comorbidity; short forms lose information [27] |
| PFFS | Self-rated pictorial fitness/frailty, /43 | Participant self-report | Usable with low literacy; good reliability and CFS correlation [29] | Self-report in an unblinded participant; responsiveness unproven |
| 6-minute walk | Functional exercise capacity | Measured | MCID estimated at 12.9–17.8 m [30] | Effort-dependent; unblinded participant; learning effect |
| SPPB | Lower-extremity function | Measured | Responsive; trial-validated [3] | Ceiling effects; short-term gains may not persist |
| Gait speed | Usual walking speed | Measured | Strong survival predictor; simple | Small effects (0.09 m/s) at the edge of meaning [31] |
The direct comparison of Fried and the Frailty Index in the Canadian Study of Health and Aging is the cleanest demonstration of discordance: the two measures correlated only moderately (R = 0.65), median index scores rose from 0.12 in the robust to 0.30 in the pre-frail and 0.44 in the frail, and within each phenotypic class there were marked differences in outcome — among people classified as robust, median five-year survival was 85% for those with lower index scores and 55% for those with higher scores [32]. Two instruments, both defensible, disagreeing about who is frail and about who will die.
STRONG’s designers have responded to this problem the way most trialists do: measure several instruments and hope they agree. The risk is that they will not, and that the trial will report a mosaic of small effects across seven primary outcomes, some positive and some null, with no pre-specified rule for adjudicating between them. Section 10 quantifies how likely that mosaic is to include at least one false-positive result.
4. Exercise in frailty: what the randomised evidence shows, and how much of it is supervision
4.1 What exercise is being asked to do
Exercise is the only component of the STRONG package with randomised evidence of an effect on a hard, patient-centred outcome in a population selected for vulnerability. That sentence needs to be read precisely. The evidence shows that structured physical activity reduces the incidence of major mobility disability — the inability to walk 400 metres — in older adults with physical limitations. It does not show that exercise prevents myocardial infarction, stroke or death, and the trials that come closest to those outcomes were not designed to test them.
Three mechanisms are usually invoked. First, exercise improves muscle mass, strength and power, which are the proximate determinants of gait, balance and the ability to recover from a stumble. Second, it improves aerobic capacity and endothelial function, which affects exercise tolerance and possibly blood pressure. Third, it has anti-inflammatory and metabolic effects that may slow the frailty cascade. Only the first two are measured well in trials; the third is inferred.
4.2 LIFE-P: the trial that set the benchmark
The Lifestyle Interventions and Independence for Elders (LIFE) study randomised 1,635 sedentary men and women aged 70 to 89 with a Short Physical Performance Battery score of 9 or below but the ability to walk 400 m, across eight US centres, to a structured moderate-intensity physical activity programme (n = 818) or a health education programme with upper-extremity stretching (n = 817) [2]. The intervention was delivered in a centre twice weekly and at home three to four times weekly, and included aerobic, resistance and flexibility training; average participation was 2.6 years and outcome assessors were blinded.
Incident major mobility disability occurred in 30.1% of the physical activity group and 35.5% of the health education group (HR 0.82, 95% CI 0.69–0.98; P = 0.03), and persistent mobility disability in 14.7% versus 19.8% [2]. A 5.4-percentage-point absolute reduction in a disabling outcome, sustained over two and a half years, in a population that was sedentary but not severely frail, is a genuinely important result. It is also smaller than most people remember, and it was obtained against an active control that received workshops and stretching rather than nothing.
The cognitive secondary outcomes were null. At 24 months, Digit Symbol Coding scores were 46.26 in the physical activity group and 46.28 in the health education group (mean difference −0.01 points, 95% CI −0.80 to 0.77; P = 0.97), with no difference in delayed recall [33]. This is worth remembering when reading claims that exercise protects the ageing brain: the same intervention that reduced mobility disability did not move cognition in the same trial.
4.3 SPRINTT: the trial closest to STRONG’s population
The SPRINTT trial randomised 1,519 community-dwelling adults aged 70 and older with physical frailty and sarcopenia — defined as SPPB 3–9, low appendicular lean mass, and ability to walk 400 m — across 16 sites in 11 European countries, to a multicomponent intervention (moderate-intensity physical activity twice weekly at a centre and up to four times weekly at home, tailored with actimetry, plus personalised nutritional counselling) or to monthly education on healthy ageing [3]. Follow-up averaged 26.4 months and outcome assessment was blinded.
Among participants with SPPB scores of 3–7, mobility disability occurred in 46.8% of the intervention group and 52.7% of controls (HR 0.78, 95% CI 0.67–0.92; P = 0.005). Persistent mobility disability occurred in 21.0% versus 25.0% (HR 0.79, 0.62–1.01; P = 0.06), and the between-group difference in SPPB was 0.8 points (0.5–1.1) [3]. The trial is the best existing model for what STRONG is attempting: a frail population, a centre-plus-home multicomponent programme, a nutritional co-intervention, and a functional primary outcome. It is also instructive in what it did not show. The persistent-disability result did not reach significance. The nutritional counselling was a co-intervention that cannot be separated from the exercise. And the primary analysis was restricted to the subgroup with lower baseline SPPB, which is a defensible pre-specified choice but also a reminder that the overall population result was less impressive.
4.4 The meta-analytic picture: large pooled effects, fragile precision
A 2026 meta-analysis of unimodal exercise modalities in older adults pooled 13 exercise–control comparisons with 1,361 participants and found an overall effect favouring exercise on frailty scores (SMD −0.57, 95% CI −0.91 to −0.24), with substantial heterogeneity (I² = 88%) and a 95% prediction interval that crossed the null (−1.73 to 0.57) [4]. Resistance exercise alone showed a significant pooled association (SMD −0.78, −1.22 to −0.35), whereas walking and Tai Chi estimates were imprecise and not significant [4]. The prediction interval is the important number: it says that in a new setting, the true effect could plausibly be large or zero.
A broader meta-analysis of 37 randomised trials with 3,409 pre-frail or frail participants found that exercise reduced frailty scores (SMD −0.95, 95% CI −1.38 to −0.51) and improved physical and cognitive function, with the largest benefits in people with established frailty, while concluding that multicomponent training appears promising but is not established as superior to single-modality exercise [34]. A third review, restricted to older adults with hypertension, found that only one of 30 included studies actually measured frailty, and concluded that there is no evidence on whether exercise ameliorates physical frailty in that population [35]. The contrast between the first two reviews and the third is a useful warning: the pooled effect depends heavily on which trials are eligible, and “exercise improves frailty” is a claim about a literature, not a biological law.
4.5 Falls: the outcome that matters most to patients, and the weakest evidence
Exercise is widely believed to prevent falls, and the belief is only partly supported in frail populations. A 2026 meta-analysis of 14 trials with 2,719 frail older adults found that exercise improved gait speed (mean difference 0.09 m/s, 0.02–0.15; I² = 94%), Timed Up and Go (−3.73 s, −5.92 to −1.55; I² = 95%) and five-times sit-to-stand (−2.66 s, −4.19 to −1.14; I² = 77%) over short-term follow-up, with low-to-very-low certainty; balance improved with very low certainty, and data on fall incidence were available from only two trials [31]. The authors’ conclusion is carefully worded: exercise training enhances physical function in frail older adults, but its impact on fear of falling and fall incidence remains uncertain.
The Cochrane review of multifactorial and multiple-component falls prevention is the most rigorous summary available and is sobering in a different way. Across 110 randomised trials with 48,919 community-dwelling older people, multifactorial interventions may reduce the falls rate (rate ratio 0.76, 95% CI 0.67–0.87; I² = 92%; low-certainty evidence) and the risk of recurrent falls (RR 0.87, 0.78–0.98), with little or no effect on the risk of falling (RR 0.95, 0.90–1.01). Multiple-component interventions — the same package given to everyone, which is what STRONG is — may have little or no effect on falls rate compared with usual care (rate ratio 0.91, 0.80–1.04; I² = 72%) and may reduce the risk of falling (RR 0.88, 0.80–0.97), with uncertain effects on recurrent falls [36]. In other words, the individualised, multifactorial model performs better than the uniform package, and the uniform package’s effect on the outcome patients care about most is small and uncertain. STRONG is a uniform package.
4.6 The dose question: how much of the benefit is the prescription, and how much is the supervision?
The trials that succeeded were not home-leaflet interventions. LIFE delivered two centre sessions and three to four home sessions per week for 2.6 years [2]. SPRINTT delivered two centre sessions and up to four home sessions per week for up to 36 months, with actimetry to tailor the programme [3]. STRONG proposes three 60-minute supervised sessions per week for 12 months — more centre-based contact than either landmark trial, and therefore more expensive and less scalable, but also more likely to produce adherence and a measurable physiological dose.
That trade-off is the central unresolved question in the field and it is not answerable from the existing literature. No trial in this review randomised supervision itself. The inference that supervision matters rests on three observations: the trials with the largest functional effects had substantial supervised contact; adherence barriers in community-dwelling older adults are dominated by access, transport, social support and monitoring, all of which supervision addresses [37]; and home-based programmes with minimal contact have produced smaller and less consistent effects. All three are consistent with a supervision effect, and none proves one. A 16-week multicomponent programme in 44 older adults improved SPPB (10.1 vs 9.1), grip strength (18.3 vs 16.9 kg), gait speed (0.85 vs 0.78 m/s) and Fried criteria (1.6 vs 2.1) [38], which shows that a short supervised programme can move the measures — not that the supervision is the active ingredient.
4.7 Table: the exercise evidence at a glance
| Trial or review | Population | Intervention | Primary result | Certainty / caveat |
|---|---|---|---|---|
| LIFE [2] | 1,635 sedentary adults 70–89, SPPB ≤9 | Centre 2×/wk + home 3–4×/wk, aerobic + resistance + flexibility, 2.6 y | Major mobility disability 30.1% vs 35.5%; HR 0.82 (0.69–0.98) | Blinded assessors; active control; cognition null [33] |
| SPRINTT [3] | 1,519 adults ≥70 with physical frailty + sarcopenia | Centre 2×/wk + home up to 4×/wk + nutritional counselling, up to 36 mo | Mobility disability 46.8% vs 52.7% (SPPB 3–7); HR 0.78 (0.67–0.92) | Persistent disability HR 0.79 (0.62–1.01), NS; nutrition co-intervention |
| Modality meta-analysis [4] | 1,361 participants, 13 comparisons | Resistance, walking or Tai Chi | SMD −0.57 (−0.91 to −0.24); prediction interval −1.73 to 0.57 | I² = 88%; resistance −0.78; walking/Tai Chi NS |
| Pre-frail/frail meta-analysis [34] | 37 RCTs, 3,409 participants | Mixed exercise | SMD −0.95 (−1.38 to −0.51) | Substantial heterogeneity; multicomponent not superior |
| Falls in frailty [31] | 14 RCTs, 2,719 frail adults | Multicomponent or resistance | Gait speed +0.09 m/s; TUG −3.73 s; STS −2.66 s | Very low certainty; falls incidence in only 2 trials |
| Cochrane multiple-component falls [36] | 110 RCTs, 48,919 older people | Uniform package vs usual care | Falls rate RaR 0.91 (0.80–1.04) | Little or no effect on falls rate; risk of falling RR 0.88 |
4.8 What a clinician should take from the exercise literature
Multicomponent exercise, delivered with supervision and sustained for a year or more, improves physical function and reduces mobility disability in older adults who are sedentary or mildly frail. The effect is real, modest in absolute terms, and best documented in populations less frail than the ones who present to hospital. The effect on falls is smaller than the effect on function, and the effect on hard cardiovascular events is unestablished. The STRONG exercise prescription — three supervised sessions per week — is at the intensive end of what has been tested, which maximises the chance of a physiological signal and minimises the chance that the model is deliverable at scale.
5. Protein and anabolic resistance: dose, thresholds, the exercise dependency, and the kidney question
5.1 Anabolic resistance is multifactorial, which is why single fixes disappoint
Older muscle responds less to the same amount of protein and to the same bout of exercise. The phenomenon is called anabolic resistance, and a 2026 systems-modelling study of leucine-mediated signalling and protein metabolism in human muscle reached a conclusion that should temper most claims in this field: no single dysregulated mechanism reproduced the age-related reduction in muscle protein synthesis, and anabolic resistance emerged only when multiple impairments operated together, so restoring synthesis required coordinated, multitarget strategies [39]. That is a model, not a clinical trial, but it explains a consistent empirical pattern — single-nutrient interventions in older adults produce small and inconsistent effects, because the defect is distributed.
It also explains why exercise is not optional in this story. Protein provides substrate; resistance exercise provides the anabolic signal that makes the substrate useful. Remove the signal and the substrate is largely wasted, which is precisely what the trial literature shows.
5.2 The per-meal threshold and the distribution problem
The recommendation to increase total daily protein rests partly on the observation that older adults eat most of their protein at dinner and very little at breakfast. In a cross-sectional study of 78 adults over 60, protein intake exceeded current recommendations on average but was unevenly distributed, with under-consumption at breakfast and dinner and few participants reaching 25 g at one, two or three meals [40]. The study is small and cross-sectional, and its conclusions should be held lightly, but the pattern has been reproduced often enough that per-meal distribution is now a standard part of the rationale for supplementation.
This is directly relevant to STRONG’s 30 g of protein powder per day. If the powder is consumed as a breakfast supplement, it addresses the meal at which intake is lowest and may therefore have a larger effect on 24-hour muscle protein synthesis than the same 30 g added to dinner. If it is consumed at random, the effect may be smaller. The registry describes the supplement as a daily 30 g dose but does not, in the fields retrieved, specify timing. That is a design detail with a plausible mechanistic consequence and it is not reported.
5.3 What the guidelines say
The PROT-AGE position paper recommends an average daily intake of at least 1.0–1.2 g/kg body weight per day for people over 65 to maintain and regain lean mass and function, with 1.2–1.5 g/kg/day for those with acute or chronic disease, and notes that people with severe kidney disease (eGFR < 30 mL/min/1.73 m²) who are not on dialysis are an exception and may need to limit protein [41]. The ESPEN expert group recommendations are concordant: at least 1.0–1.2 g/kg/day for healthy older people, 1.2–1.5 g/kg/day for those malnourished or at risk because of illness, with daily physical activity or exercise for all older people for as long as possible [42]. Both documents emphasise that protein and exercise are complementary rather than interchangeable.
Neither guideline recommends protein powder. Both recommend food first, with supplementation as a practical route when intake is inadequate. That distinction matters for how the STRONG intervention should be described to patients: 30 g of whey isolate is a means of reaching a protein target, not a therapeutic dose of a drug.
5.4 The dose-response evidence: a ceiling around 1.6 g/kg/day, and a declining benefit with age
The most-cited meta-analysis of protein supplementation with resistance training pooled 49 studies with 1,863 participants and found that protein supplementation significantly augmented gains in muscle mass and strength during prolonged resistance training, that the effect on fat-free mass was reduced with increasing age, and that protein intakes above about 1.62 g/kg/day produced no further gains in fat-free mass [5]. Two features of that analysis deserve emphasis. The age interaction is not a footnote: it means the population most in need of the intervention is the population in which it works least well. And the 1.62 g/kg/day ceiling is well above the 1.0–1.2 g/kg/day recommended by PROT-AGE, which implies that for most older adults the guideline target is not the limiting factor in gaining muscle — the exercise stimulus and total energy intake are.
5.5 Protein plus exercise versus protein alone
A 2025 meta-analysis of 13 randomised trials with 1,057 participants aged 65 and older with sarcopenia or physical frailty found that combined protein and exercise interventions improved skeletal muscle index (mean difference 0.89 kg/m², 95% CI 0.45–1.33) and handgrip strength (+2.64 kg, 0.75–4.53) compared with exercise alone, whereas protein supplementation alone showed only modest benefits in muscle strength and limited effects on physical performance; no serious adverse events were reported and the certainty of evidence was rated very low [6]. A smaller meta-analysis of seven randomised trials and one quasi-experimental study (854 participants, 10–24 weeks) reached a similar conclusion for muscle mass (SMD 0.95, 0.13–1.78), while explicitly noting that the limited number of trials restricts robustness [43].
The pattern is consistent across both: protein adds to exercise, protein alone does little for function, and the certainty is low because the trials are small, short and heterogeneous. A clinician should read “protein plus exercise works better than exercise alone” as a reasonable inference from weak evidence rather than as an established fact.
5.6 The observational counterpoint: modest association, and a null
Two cohort analyses illustrate how easily observational protein data can be over-read. In 27,128 UK Biobank participants aged 50 and older followed for a mean of five years, higher protein intake was associated with a moderately reduced risk of frailty incidence and a greater likelihood of recovery from pre-frailty in a dose-dependent manner [44]. In 130 home-dwelling Norwegian adults aged 70 and older followed for eight years, the Frailty Index rose from 0.12 to 0.20 and the rise was not prevented by dietary protein intake; the authors concluded that frailty accelerates with age independent of protein intake in their population [45]. The two studies differ in size, exposure measurement and outcome, and neither is a trial. Together they suggest that protein intake is probably a modest determinant of frailty trajectories at population level and is unlikely to be a strong lever in an individual who is already frail.
5.7 Leucine, timing and the limits of the supplement logic
Leucine is the amino acid most often proposed as the anabolic trigger. A 2026 randomised double-blind trial in 24 young and 24 older adults undergoing three days of unilateral leg immobilisation found that leucine supplementation (5 g three times daily with meals) did not attenuate the decline in daily muscle protein synthesis or preserve leg muscle mass in either age group; muscle protein synthesis fell by about 15% in the young and 23% in the older immobilised leg [46]. This is a disuse model rather than a training model, and three days is short, but it is a direct test of the “leucine fixes anabolic resistance” hypothesis and it failed.
A separate 2026 exploratory trial in 24 older adults undergoing knee arthroplasty measured protein synthesis rates across muscle, bone, cartilage, tendon, ligament, meniscus and periarticular soft tissue with or without 40 g/day of whey protein for 14 days. Muscle turned over at about 1.17%/day, synovium at 0.79%/day and fibrous tissues and bone at roughly 0.12–0.46%/day [47]. The relevance to a 12-month trial is straightforward: tissues with turnover rates of 0.2% per day will not be substantially remodelled by a one-year intervention, so any expectation that protein supplementation changes cartilage, tendon or bone composition over 12 months is optimistic. Muscle is the tissue that can respond within the trial’s time frame; that is where the effect, if any, will be.
5.8 The 30 g dose in context
Thirty grams of protein powder per day is a meaningful increment for someone eating 50–60 g/day and a marginal one for someone already eating 80 g/day. For an 80 kg adult, the PROT-AGE target of 1.0–1.2 g/kg/day is 80–96 g/day; 30 g of powder supplies roughly a third of that target. Whether it closes a gap depends entirely on baseline intake, which the trial measures with the ASA24 dietary assessment at baseline and follow-up [1]. That is the right design choice: without baseline intake, a fixed supplement dose cannot be interpreted. But it also means the trial’s protein effect will be heterogeneous by construction — large in the undernourished, negligible in the well-nourished — and an overall null result could mask a benefit in the subgroup that needed it. No subgroup analysis of that kind is described in the registry record.
5.9 The kidney question
The concern that higher protein intake harms kidneys rests largely on the physiology of hyperfiltration and on observational associations in people with established chronic kidney disease. A 2026 meta-analysis of 22 randomised trials of high-protein diets (typically 25–35% of energy or ≥2.0 g/kg/day) in adults without chronic kidney disease found that high-protein diets increased estimated glomerular filtration rate (SMD 0.39, 95% CI 0.09–0.69; I² = 74.9%) without a significant change in serum creatinine (SMD 0.08, −0.09 to 0.25; 19 trials, 1,044 participants), and the authors cautioned that the eGFR finding may reflect short-term haemodynamic adaptation rather than injury [48]. That is reassuring for people with normal kidneys and uninformative for people with unrecognised CKD, who are the ones at risk.
The practical issue for a trial like STRONG is ascertainment. The registry eligibility criteria, as retrieved, list age ≥65 years, willingness to participate in a low-to-moderate exercise programme, and exclusion of medical or cognitive conditions that prevent participation and of allergy or dislike of blueberries or protein powder [1]. No eGFR threshold or CKD exclusion is listed. A 30 g/day protein supplement is unlikely to harm someone with an eGFR of 45, and PROT-AGE’s restriction applies to eGFR < 30 [41], but a trial that does not measure kidney function cannot report on the safety question it will inevitably be asked about. That is a modest design gap with a cheap fix.
5.10 Table: the protein evidence at a glance
| Source | Design | Population | Key result | Caveat |
|---|---|---|---|---|
| PROT-AGE [41] | Position paper | Adults >65 | 1.0–1.2 g/kg/d; 1.2–1.5 in illness; eGFR <30 exception | Expert consensus, not trial evidence |
| ESPEN [42] | Expert recommendations | Older adults | 1.0–1.2 g/kg/d healthy; 1.2–1.5 if malnourished/ill | Concordant with PROT-AGE |
| Morton et al. [5] | Meta-analysis, 49 trials, 1,863 participants | Adults in resistance training | Protein augments lean mass/strength; no further gain >1.62 g/kg/d; effect falls with age | Healthy training populations, not frail |
| Protein ± exercise in sarcopenia/frailty [6] | Meta-analysis, 13 RCTs, 1,057 participants | ≥65 with sarcopenia or frailty | SMI +0.89 kg/m²; grip +2.64 kg vs exercise alone; protein alone limited | Very low certainty; no serious AEs |
| Community sarcopenia [43] | Meta-analysis, 7 RCTs + 1 quasi | ≥60 with sarcopenia | Muscle mass SMD 0.95 (0.13–1.78) | 10–24 weeks; small |
| High-protein diets and kidney [48] | Meta-analysis, 22 RCTs | Adults without CKD | eGFR SMD +0.39; creatinine unchanged | Haemodynamic, short-term; not CKD |
| UK Biobank [44] | Cohort, 27,128 | ≥50 | Higher protein modestly reduces frailty incidence | Observational |
| Norway [45] | Cohort, 130, 8 y | ≥70 | Frailty Index rose regardless of protein intake | Small, null |
6. Blueberry polyphenols and vascular health: bioavailability, doses, and the gap between acute feeding and chronic outcomes
6.1 What is being measured, and why it is not the same as a health outcome
Blueberry trials measure endothelial function, usually as brachial artery flow-mediated dilation (FMD), arterial stiffness as augmentation index or pulse wave velocity, and office or ambulatory blood pressure. Each is a surrogate. FMD reflects nitric-oxide-mediated conduit artery dilator capacity; augmentation index reflects wave reflection and arterial stiffness; blood pressure is itself a surrogate for cardiovascular events but a much better validated one. None of these is a clinical outcome, and the distance between a 1.5-percentage-point improvement in FMD and a reduction in myocardial infarction is considerable.
That does not make the measurements worthless. It means the correct claim is “blueberries improve a marker of endothelial function”, not “blueberries protect the heart”.
6.2 Bioavailability: the constraint that shapes the whole literature
Anthocyanins are unstable — sensitive to temperature, light, oxygen, pH and enzymatic activity — and have limited bioavailability, which is why a large fraction of ingested anthocyanin appears in plasma and urine as phenolic acid metabolites rather than as intact glycosides [49]. The 6-month blueberry trial documented exactly this: total concentrations of anthocyanin-derived phenolic acid metabolites increased dose-dependently in serum and 24-hour urine, which is evidence of absorption and metabolism rather than of intact compound reaching the endothelium [9]. The mechanistic implication is that any vascular effect is likely mediated by metabolites and by downstream signalling rather than by the parent anthocyanin, and the practical implication is that dose and food matrix cannot be assumed to be interchangeable.
This is why the literature is so heterogeneous. A freeze-dried powder, a fresh fruit dose, a juice drink and an extract capsule deliver different amounts of anthocyanins, different matrices, and different metabolite profiles, and the meta-analyses pool them.
6.3 The 6-month, 1-cup trial: the best single piece of dose-response evidence
The longest-duration randomised trial in this literature enrolled 115 participants with metabolic syndrome (mean age 63, 68% male, mean BMI 31.2) and compared ½ cup and 1 cup of blueberries per day (75 g and 150 g) with a matched placebo over six months, double-blind [9]. A daily intake of 1 cup improved flow-mediated dilation by 1.45% (95% CI 0.83–2.10; P = 0.003), reduced augmentation index by 2.24% (−3.97 to −0.61; P = 0.04), and attenuated cyclic guanosine monophosphate concentrations. In statin non-users (n = 71), 1 cup/day also raised HDL cholesterol (+0.08 mmol/L; P = 0.03), HDL particle density and apolipoprotein A-I. Treatment compliance was 94.1% by wrapper returns [9].
Three features make this the anchor trial. The dose is dietarily achievable — one cup of fruit, not a concentrate. The duration is long enough to be a chronic intervention rather than an acute feeding study. And the primary endpoint was insulin resistance, which was not the endpoint that moved; the vascular effects were secondary. That last point matters: the trial is often cited as a positive vascular trial, and it is, but its pre-specified primary outcome did not drive the result.
6.4 The meta-analyses: a consistent FMD signal, inconsistent everything else
A 2024 meta-analysis of 16 studies (11 pooled) found that blueberry consumption improved FMD by 1.50% (95% CI 0.81–2.20) with substantial heterogeneity (I² = 87%), improved the reactive hyperemia index by 0.26 (0.09–0.42), and reduced diastolic blood pressure by 2.20 mmHg (−4.13 to −0.27; I² = 11%). Systolic blood pressure did not change overall (−1.43 mmHg, −3.11 to 0.26), nor did HDL or LDL cholesterol; in the smoking subgroup, systolic pressure fell by 3.92 mmHg (−6.88 to −0.97) and diastolic by 2.20 mmHg [7].
A 2026 meta-analysis restricted to healthy individuals pooled 14 randomised trials with 632 participants and found a smaller but more homogeneous FMD improvement of 1.11% (0.50–1.72; I² = 23%), with no effect on systolic pressure (−0.11 mmHg, −2.82 to 2.60), diastolic pressure (−0.46 mmHg, −2.80 to 1.87) or cognitive outcomes (executive function SMD 0.21, −0.05 to 0.46) [8]. The contrast between the two meta-analyses is informative: the FMD effect is more homogeneous in healthy populations and more heterogeneous in mixed populations, and the blood-pressure effect appears in some subgroups and not others.
6.5 Individual trials: dose, duration and population all matter
A 12-week double-blind trial in 61 healthy older adults aged 65–80 compared 26 g/day of freeze-dried wild blueberry powder (302 mg anthocyanins, equivalent to 178 g fresh fruit) with placebo. FMD increased by 0.86% (95% CI 0.56–1.17) and 24-hour ambulatory systolic blood pressure fell, with improvements in cognition reported [50]. A 12-week trial in 43 postmenopausal women with above-normal blood pressure found that absolute FMD normalised to shear-rate area under the curve was 96% higher after 22 g/day of freeze-dried highbush blueberry powder than at baseline, and that the improvement was attenuated by supraphysiological ascorbic acid infusion, consistent with an oxidative-stress mechanism [51].
Against these, a double-blind crossover trial in healthy young adults given 300 mL of a wild blueberry drink containing 740 mg total polyphenols and 460 mg anthocyanins found no effect on vascular reactivity, augmentation index or adhesion molecules two hours after consumption [52]. The population was young and healthy, the exposure was acute, and the result is a useful reminder that an acute feeding study in people with normal endothelial function is not a test of whether blueberries improve vascular health in people who have lost it.
Cognitive outcomes deserve separate mention because they are the most over-claimed. In the 6-month metabolic syndrome trial, secondary cognitive analyses found no effect of 6-month blueberry intake on attention, working memory, episodic memory, executive function or mood, with the single exception of improved postprandial self-rated calmness (11.6% improvement vs baseline for the 1-cup group) [53]. The healthy-population meta-analysis likewise found no cognitive effect [8]. The wild blueberry trial that reported cognitive improvement [50] is a single 61-person study, and the balance of evidence does not support a cognitive claim.
6.6 Table: the blueberry evidence at a glance
| Study | Design | Dose / duration | Result | Caveat |
|---|---|---|---|---|
| Curtis et al. [9] | Double-blind RCT, n = 115, metabolic syndrome | 1 cup (150 g)/d, 6 months | FMD +1.45% (0.83–2.10); AI −2.24% (−3.97 to −0.61); HDL +0.08 mmol/L in statin non-users | Primary endpoint was insulin resistance, which did not drive the result; compliance 94.1% |
| Endothelial meta-analysis [7] | 16 studies, 11 pooled | Mixed doses/durations | FMD +1.50% (0.81–2.20); RHI +0.26; DBP −2.20 mmHg; SBP NS | I² = 87% for FMD; lipids null |
| Healthy-adult meta-analysis [8] | 14 RCTs, 632 participants | Mixed | FMD +1.11% (0.50–1.72); BP and cognition null | I² = 23%; short durations |
| Wild blueberry [50] | Double-blind RCT, n = 61, 65–80 y | 26 g powder (302 mg anthocyanins) ≈178 g fresh, 12 weeks | FMD +0.86% (0.56–1.17); 24-h SBP reduced; cognition improved | Single trial; multiple outcomes |
| Postmenopausal women [51] | Double-blind RCT, n = 43 | 22 g powder/d, 12 weeks | FMD/SR AUC 96% higher; ascorbate attenuated effect | Small; surrogate outcome |
| Acute crossover [52] | Double-blind crossover, healthy young adults | 460 mg anthocyanins, 2 h | No effect on RHI, AIx or adhesion molecules | Acute; healthy population |
6.7 What the blueberry literature can and cannot support
It can support a claim that regular consumption of a dietarily achievable dose of blueberries — roughly one cup a day for six months — improves brachial artery flow-mediated dilation by about 1 to 1.5 percentage points, and may lower diastolic blood pressure by about 2 mmHg in some populations. It cannot support a claim that blueberries reduce cardiovascular events, improve cognition reliably, or substitute for statin therapy. The 1-cup dose in STRONG is well chosen against this literature: it is the dose that produced the largest chronic effect, it is food rather than extract, and it is achievable. Whether 12 months of it changes anything a patient can feel is unknown, and the trial’s inflammatory and echocardiographic endpoints are exploratory in all but name.
7. Converging mechanisms: inflammaging, senescence, endothelium and muscle
7.1 The case for convergence, and its limits
The intellectual appeal of a three-component package is that the components are supposed to meet in the middle. Exercise improves muscle anabolism and endothelial function; protein supplies substrate for the muscle; anthocyanins reduce oxidative stress and improve endothelial function. If all three act on the same pathway — chronic inflammation, mitochondrial dysfunction, endothelial nitric oxide bioavailability — then combining them should be more than additive, or at least should produce a larger signal than any one alone.
That is a hypothesis, and it is worth being explicit about what would be required to test it. A synergy claim needs either a factorial design (exercise alone, protein alone, blueberries alone, combinations) or a mediation analysis with measured mediators. STRONG has neither. It compares the package with usual care. Any effect it observes is attributable to the package as a whole, and any mechanistic interpretation is post hoc. That is a legitimate design for a pragmatic trial — pragmatic trials are supposed to test packages as they would be delivered — but it means the mechanistic narrative in the protocol is a rationale, not an analysable hypothesis.
7.2 Inflammaging: a robust risk marker with a weak causal record
Interleukin-6 and C-reactive protein are the best-studied inflammatory markers in this field, and their association with functional decline is consistent. In a pooled analysis of older adults with slow gait speed, each unit increase in log IL-6 was associated with a 26% higher risk of incident self-reported major mobility disability (HR 1.26, 95% CI 1.13–1.41); IL-6 above 2.5 pg/mL carried an HR of 1.31 (1.12–1.54), and CRP above 3.0 mg/L an HR of 1.38 (1.10–1.74). The CRP effect was larger in participants with elevated IL-6 (HR 1.62, 1.12–2.33) than in those without (HR 1.19, 0.85–1.66) [54]. The pattern — an inflammatory marker predicting disability, with the effect concentrated in people who have both markers elevated — is what a shared inflammatory pathway would produce.
Whether reducing inflammation reduces disability is a different question, and the exercise literature gives a nuanced answer. In a post hoc analysis of the LIFE study (1,300 participants), the effect of physical activity on 400-metre gait speed was significant only among participants whose yearly IL-6 change fell within a narrow window of −1 to +2 pg/mL (difference 0.041 m/s, 95% CI 0.008–0.074; Cohen’s d 0.26), with no effect when IL-6 changed more widely [55]. That is a hypothesis-generating interaction in a post hoc subgroup, and it should be read as such. But it does suggest that the inflammatory response to exercise may modify the functional response, which is a more interesting proposition than the simple claim that exercise is anti-inflammatory.
7.3 Cellular senescence: predictive, but not obviously modifiable by exercise
Senescent cells secrete a characteristic cocktail of proteins, and several of them predict functional decline. In 1,377 LIFE participants, a panel of 27 senescence-associated proteins was measured longitudinally. VEGFA, TNFR1 and MMP7 were most distinctly associated with the onset of major and persistent mobility disability, and VEGFA, GDF15, osteopontin and others were associated with declines in SPPB score [56]. That is strong evidence that senescence biology is part of the frailty story.
The intervention result is the part that gets left out of reviews. The change in senescence biomarkers did not differ between participants randomised to physical activity and those randomised to health education. What did differ was that, across the whole cohort, higher levels of physical activity were associated with greater reductions in ten senescence-related proteins at 12 and/or 24 months [56]. In other words, the randomised comparison was null and the observational association within the cohort was positive. That is a familiar pattern: the biomarker moves with behaviour, but the trial did not show that the intervention moved the biomarker. Anyone citing “exercise reduces senescence” as established should cite this trial’s randomised result, not its observational one.
7.4 Endothelium and nitric oxide: the most measurable, least clinical pathway
The endothelial story is the most coherent of the three mechanisms and the easiest to measure. Anthocyanin metabolites increase nitric oxide-related signalling, improve flow-mediated dilation and reduce augmentation index in several trials [9] [7], and the postmenopausal-women trial showed that the FMD improvement was attenuated by supraphysiological ascorbate, implicating oxidative stress [51]. Exercise independently improves endothelial function and reduces arterial stiffness, and the two interventions plausibly act on the same final pathway.
But the clinical distance is large. Flow-mediated dilation is a surrogate whose relationship with cardiovascular events is observational and modest; a 1.5-percentage-point improvement is statistically robust and clinically unquantified. The STRONG protocol’s decision to measure global longitudinal strain and ejection fraction by echocardiography is a step toward a more meaningful cardiac endpoint, but GLS is itself a surrogate, validated mainly in heart failure and cardiotoxicity populations rather than in general older adults. Neither FMD nor GLS nor the cytokine panel is an established surrogate endpoint for cardiovascular events in the sense that a regulator would accept. They are risk markers and mechanism probes. That is a legitimate thing for a trial to measure; it is not a substitute for events.
7.5 Anabolic resistance and mTOR: the pathway the package is best designed to hit
The one place where the three components genuinely converge is muscle. Exercise provides the anabolic signal through mTORC1; protein provides amino acid substrate, with leucine acting as a signalling molecule as well as a substrate; and chronic inflammation and oxidative stress impair both. Anabolic resistance is the age-related failure of this system, and it is multifactorial: the systems model described earlier found that no single impairment reproduced it and that restoring synthesis required coordinated, multitarget intervention [39]. A package of exercise plus protein is, in that sense, mechanistically well matched to the defect — which is exactly what the protein-plus-exercise trials show, with modest gains in muscle mass and strength on low-certainty evidence [6].
Blueberries contribute to this pathway only indirectly, through oxidative stress and endothelial function. There is no strong evidence that anthocyanins increase muscle protein synthesis or augment resistance-training gains. The plausible route is vascular: better perfusion, better exercise tolerance, more activity. That is a hypothesis without a trial behind it, and the protocol’s rationale — that blueberries “can improve vascular health and protect against cellular senescence” — is a statement about plausibility, not about demonstrated effect in the trial population.
7.6 What the measured biomarkers can and cannot establish
STRONG’s inflammatory panel is broad — more than thirty analytes, including interleukins, interferons, MMPs, osteocalcin and osteopontin [1]. A panel that broad has two problems. The first is multiplicity: with thirty analytes tested at α = 0.05, roughly one and a half false positives are expected by chance alone, and the trial has no stated strategy for handling that. The second is interpretation: if IL-6 falls and IL-10 rises and MMP-3 is unchanged, what is the conclusion? Without a pre-specified primary analyte and a pre-specified direction, the panel generates hypotheses rather than answers.
The biomarkers also have a validation problem that is worth stating plainly, because it is often glossed. A risk marker becomes a surrogate endpoint only when an intervention that changes the marker reliably changes the outcome. For inflammatory markers in frailty, that has not been demonstrated. Trials that lowered IL-6 have not consistently improved function, and trials that improved function have not consistently lowered IL-6 [56] [55]. The markers are useful for understanding mechanism and for generating hypotheses; they cannot carry the weight of a primary endpoint.
8. The multidomain lifestyle trial record: what has worked, what has been null, and why
8.1 FINGER: a small effect, honestly reported
The Finnish Geriatric Intervention Study to Prevent Cognitive Impairment and Disability (FINGER) randomised 1,260 adults aged 60–77 with a CAIDE dementia risk score of at least 6 and cognition at or slightly below expected for age to a two-year multidomain intervention (diet, exercise, cognitive training, vascular risk monitoring) or general health advice [10]. The estimated mean change in the neuropsychological test battery total Z score at two years was 0.20 in the intervention group and 0.16 in the control group, a between-group difference in the change per year of 0.022 (95% CI 0.002–0.042; P = 0.030) [10].
FINGER is the positive multidomain trial, and it is worth being precise about the size of the win. A difference of 0.022 Z-score units per year is small in absolute terms; whether it corresponds to a clinically detectable difference in cognitive trajectory is debated. The trial also reported a safety signal that matters for anyone planning a similar intervention: adverse events occurred in 46 (7%) intervention participants versus 6 (1%) controls, with musculoskeletal pain the most common (32 participants, 5%, versus none) [10]. A multidomain lifestyle intervention is not a no-risk intervention, and the risk is concentrated in the exercise component.
A 2025 secondary analysis asked whether baseline frailty modified the FINGER effect. It did not: the frailty × intervention × time interaction was not significant for global cognition, and within-group analyses suggested benefit in pre-frail participants, with only 15 frail participants available [57]. The authors’ conclusion — that prefrailty may be the optimal window for prevention — is a reasonable reading of a subgroup analysis and should be treated as hypothesis-generating.
8.2 MAPT: null, with narrow confidence intervals
The Multidomain Alzheimer Preventive Trial randomised 1,680 non-demented adults aged 70 and older with memory complaints, limitation in one instrumental activity of daily living, or slow gait speed, to a three-year multidomain intervention plus omega-3, multidomain plus placebo, omega-3 alone, or placebo [11]. In the modified intention-to-treat population of 1,525, there were no significant differences in three-year cognitive decline between any intervention group and placebo: the between-group difference was 0.093 (95% CI 0.001–0.184; adjusted P = 0.142) for the combined group, 0.079 (−0.012 to 0.170; P = 0.179) for multidomain plus placebo, and 0.011 (−0.081 to 0.103; P = 0.812) for omega-3 alone [11]. Serious adverse events occurred in 32–36% of participants across groups, with no difference between arms, and there were four treatment-related deaths (two in the multidomain-plus-placebo group and two in the placebo group) [11].
MAPT is important because it is not an underpowered null. The confidence intervals exclude anything but a small benefit, and the trial had a three-year duration and a plausible intervention. When a well-conducted multidomain trial with a 1,680-person sample produces a confidence interval that tight around zero, the burden of proof shifts to those who claim a large effect.
8.3 preDIVA: a six-year null in primary care
The Prevention of Dementia by Intensive Vascular Care (preDIVA) trial cluster-randomised 116 general practices with 3,526 participants aged 70–78 to a six-year nurse-led multidomain cardiovascular intervention or usual care [12]. Dementia developed in 7% of the intervention group and 7% of controls (HR 0.92, 95% CI 0.71–1.19; P = 0.54), and disability scores did not differ (adjusted mean difference −0.02, −0.38 to 0.42; P = 0.93) [12]. Sixteen per cent of the intervention group died during follow-up.
preDIVA is the most pragmatic of the multidomain trials — delivered through existing primary care, over six years, in a general older population — and it is null. The lesson is not that vascular risk factor management is useless; it is that a nurse-led, protocolised, multidomain intervention delivered at population scale did not produce a detectable cognitive benefit. Intensity and adherence are the usual explanations, and they are plausible, but they are also unfalsifiable without a trial that manipulates them.
8.4 SPRINT MIND: outcome definition determines the answer
SPRINT MIND randomised 9,361 adults aged 50 and older with hypertension to a systolic target below 120 mmHg or below 140 mmHg and followed cognitive outcomes for a median of 5.11 years [13]. Adjudicated probable dementia occurred in 149 intensive-treatment participants versus 176 standard-treatment participants (7.2 vs 8.6 cases per 1,000 person-years; HR 0.83, 95% CI 0.67–1.04) — not significant. Intensive control did significantly reduce mild cognitive impairment (14.6 vs 18.3 per 1,000 person-years; HR 0.81, 0.69–0.95) and the composite of MCI or probable dementia (20.2 vs 24.1 per 1,000 person-years; HR 0.85, 0.74–0.97) [13].
The trial is often described as positive, and on the composite it is. On its primary endpoint it is not. That is not a criticism of the investigators; it is an illustration of how much the choice of primary outcome determines the headline. A blood-pressure intervention is also the closest thing in this literature to a hard-outcome win, and even it produced a borderline result on the outcome that matters most.
8.5 Look AHEAD: the canonical warning about surrogates
Look AHEAD randomised 5,145 overweight or obese adults with type 2 diabetes to an intensive lifestyle intervention promoting weight loss through reduced calories and increased activity, or to diabetes support and education, and followed them for a median of 9.6 years before the trial was stopped for futility [14]. Weight loss was greater in the intervention group throughout (8.6% vs 0.7% at one year; 6.0% vs 3.5% at study end), and the intervention produced greater reductions in glycated haemoglobin and greater initial improvements in fitness and all cardiovascular risk factors except LDL cholesterol. The primary composite of cardiovascular death, non-fatal myocardial infarction, non-fatal stroke or hospitalisation for angina occurred in 403 intervention and 418 control participants (1.83 vs 1.92 events per 100 person-years; HR 0.95, 95% CI 0.83–1.09; P = 0.51) [14].
Look AHEAD is the single most important trial for anyone designing a lifestyle intervention with cardiovascular aspirations. It moved almost every surrogate in the right direction, for almost a decade, in more than five thousand patients — and did not reduce events. The reasons are debated: the control group also improved, the event rate was lower than expected, the weight loss was not sustained, and the follow-up may have been too short. All of those explanations are plausible. None of them changes the conclusion that surrogate improvement is not event reduction.
8.6 DO-HEALTH: supplements and simple exercise in the healthy old
The DO-HEALTH trial randomised 2,157 community-dwelling adults aged 70 and older in a 2 × 2 × 2 factorial design to 2,000 IU/day of vitamin D3, 1 g/day of marine omega-3 fatty acids, a simple home exercise programme, or placebo/control, over three years [58]. Among generally healthy, active and vitamin D-replete participants, a daily high dose of vitamin D or the home exercise programme had no benefit on falls, while omega-3 supplementation may have had a modest benefit on total falls [58]. The trial is a useful counterweight to the assumption that anything labelled “exercise” or “supplement” in an older population must help. The population was healthy and active, the exercise was home-based and simple, and the outcome was falls. Under those conditions, the intervention was null.
8.7 The exception that proves the rule: PipELINe
The PipELINe trial randomised 512 patients aged 65 and older recovering from myocardial infarction to a multidomain cardiac rehabilitation programme or usual care in a 2:1 ratio, and a prespecified subgroup analysis examined frailty [59]. Overall, 68.4% were pre-frail or frail. Among pre-frail or frail patients, assignment to multidomain cardiac rehabilitation was associated with a lower risk of the composite of cardiovascular death or unplanned cardiovascular hospitalisation at one year (HR 0.57, 95% CI 0.34–0.94; P = 0.028), with no significant treatment-by-frailty interaction [59].
This is the most encouraging result in the multidomain literature for a frailty-focused audience, and it differs from the null trials in three ways: the population had established cardiovascular disease, the outcome was a clinical composite rather than a cognitive score, and the intervention was delivered in the context of cardiac rehabilitation, which already has infrastructure, reimbursement and a culture of participation. It is a single trial with a subgroup analysis and a wide confidence interval, and it should not be over-read. But it is a proof of concept that a multidomain intervention can reduce clinical events in a frail cardiac population when the outcome is chosen well and the delivery system exists.
8.8 Table: the multidomain record
| Trial | Population | Intervention / duration | Primary outcome | Result |
|---|---|---|---|---|
| FINGER [10] | 1,260 adults 60–77 at dementia risk | Diet, exercise, cognitive training, vascular monitoring; 2 y | NTB Z-score change | +0.022 Z/year (0.002–0.042); AEs 7% vs 1% |
| MAPT [11] | 1,680 adults ≥70 with memory complaint | Multidomain ± omega-3; 3 y | Cognitive composite | Null; combined 0.093 (0.001–0.184), adjusted P = 0.142 |
| preDIVA [12] | 3,526 adults 70–78, primary care | Nurse-led vascular care; 6 y | Dementia; disability | 7% vs 7%; HR 0.92 (0.71–1.19) |
| SPRINT MIND [13] | 9,361 adults ≥50 with hypertension | BP <120 vs <140; 5.1 y | Probable dementia | HR 0.83 (0.67–1.04); MCI HR 0.81 (0.69–0.95) |
| Look AHEAD [14] | 5,145 adults with T2D and overweight | Weight loss via diet + activity; 9.6 y | CV composite | HR 0.95 (0.83–1.09); stopped for futility |
| DO-HEALTH [58] | 2,157 healthy adults ≥70 | Vitamin D, omega-3, home exercise; 3 y | Falls | Vitamin D and exercise null; omega-3 modest |
| PipELINe [59] | 512 post-MI adults ≥65 | Multidomain cardiac rehab; 1 y | CV death or CV hospitalisation | Pre-frail/frail HR 0.57 (0.34–0.94) |
8.9 What distinguishes the successes from the failures
Four patterns emerge, and none of them is about the number of components.
First, outcome choice dominates. FINGER won on a composite cognitive Z-score; MAPT lost on a similar composite; SPRINT MIND won on MCI but not dementia; Look AHEAD lost on events despite winning on risk factors. The same class of intervention produces different verdicts depending on which endpoint is primary.
Second, population risk matters, but not in the direction intuition suggests. FINGER enrolled people at risk but cognitively intact, and won. MAPT enrolled people with memory complaints, and lost. preDIVA enrolled a general older population, and lost. PipELINe enrolled post-MI patients, and won on a clinical composite. The common thread is not baseline risk but the presence of a modifiable process: cognitive reserve in FINGER, post-infarct remodelling and deconditioning in PipELINe.
Third, delivery infrastructure matters. FINGER and PipELINe were delivered through systems that already existed (research clinics with dedicated staff; cardiac rehabilitation). preDIVA was delivered through primary care, at scale, with nurses who had many other duties.
Fourth, adherence decays and the trials do not measure it well. None of these trials reports a clean dose-response between adherence and outcome, which is the analysis that would tell us whether the intervention failed or the implementation failed.
For STRONG, the implication is uncomfortable but clear. It has a package rather than a single intervention, a pragmatic design, a usual-care comparator, and seven primary outcomes spanning function, risk factors, inflammation and cardiac structure. That is a design optimised for breadth of description, not for a decisive answer on any one question.
9. Design consequences: pragmatic superiority, usual care, and an unblinded behavioural intervention
9.1 Where STRONG sits on the pragmatic–explanatory continuum
The PRECIS-2 tool describes trial design across nine domains, from eligibility and recruitment through setting, organisation, delivery flexibility, adherence flexibility, follow-up, primary outcome and primary analysis, with each domain scored from explanatory to pragmatic [60]. The purpose is not to classify trials as good or bad but to make explicit the trade-offs a design has made.
STRONG is pragmatic on several domains. Eligibility is broad (age ≥65, willingness to exercise, exclusion only of conditions preventing participation or food allergy) [1]. The comparator is usual care rather than a placebo or attention control. The setting is a provincial health system rather than a specialised research clinic. The primary purpose is prevention, and the outcomes include measures that a health system might actually collect.
It is explanatory on one important domain: delivery. Three 60-minute sessions per week supervised by an exercise physiologist is not how exercise is delivered in the community. It is how exercise is delivered in a trial. That is a reasonable choice for a first test — it maximises the chance of demonstrating a biological effect — but it means a positive result will not tell a health system whether the programme is deliverable, and a null result will leave open the possibility that a less intensive, more scalable version might have worked.
A recent commentary on pragmatic trials makes the same point in general terms: pragmatism is not a single axis, and trials that are pragmatic in some domains and explanatory in others require careful interpretation of what their results generalise to [61]. That is the correct frame for STRONG.
9.2 The usual-care comparator is not a no-treatment comparator
“Usual care” in a Canadian province in 2026 includes family physicians who counsel activity, cardiac rehabilitation for eligible patients, dietitian referrals, diabetes education programmes, and whatever exercise the participant was already doing. It also includes the possibility that participation in a trial — being measured, being asked about exercise, being weighed — changes behaviour in the control group. That is the Hawthorne effect, and in unblinded behavioural trials it is not a nuisance; it is a threat to the estimated effect.
The Look AHEAD experience is instructive: the control group received diabetes support and education and also improved some risk factors over time [14]. The between-group difference is what the trial measures, and it will be smaller than the within-group change in the intervention arm. A trial that reports “the intervention group improved” without reporting the control group’s trajectory has not reported an effect.
Contamination can also run the other way. If control participants learn about the intervention, or if their physicians do, some will adopt parts of it. In a small province with a tightly connected clinical community, that is likely. The result is bias toward the null, which is conservative for a superiority trial but makes a null result harder to interpret.
9.3 Blinding is impossible; masking is partial
The registry records the trial as single-masked, with the outcomes assessor masked [1]. That is the maximum achievable for a diet-and-exercise intervention, and it is worth acknowledging that it is a genuine limitation rather than a design failure.
The problem is that several of the registered outcomes are participant-reported. The Pictorial Fit-Frail Scale is self-completed [1] [29]. Life-space and social engagement, mentioned in the protocol abstract as frailty domains, are self-reported. Dietary assessment by ASA24 is self-reported. A participant who has been randomised to three supervised exercise sessions a week, free blueberries and free protein powder knows exactly what arm they are in, and their self-assessment of fitness and diet is not independent of that knowledge. The CFS and the Frailty Index are assessor-derived and therefore maskable; the PFFS is not.
This is not a reason to discard self-reported outcomes. It is a reason to weight them differently. A treatment effect that appears only on self-reported measures, in an unblinded behavioural trial, is weak evidence. A treatment effect that appears on measured performance, blinded assessment and objective biomarkers is stronger. The trial’s seven co-primary outcomes mix both categories, and the protocol does not pre-specify how they will be weighted.
9.4 Adherence: the outcome that is not listed
Among the outcomes retrieved from the registry, adherence is not listed as a primary or secondary outcome. The nine secondary outcomes cover strength, functional fitness, balance, mobility, overall health, food-work assessment and dietary assessment [1]. Exercise session attendance and supplement consumption will presumably be recorded as process measures, but they are not part of the registered outcome set.
That is a gap, because adherence is the variable that determines how the results should be interpreted. A trial with 90% attendance and a null result tells a different story from a trial with 50% attendance and the same null result. The literature on adherence in community-dwelling older adults identifies 54 factors and 38 behaviour-change techniques across eight strategy areas — tailored programmes, appropriate environments, social support, monitoring and feedback, emotional management, education, self-efficacy and autonomy [37] — which is a way of saying that adherence is multifactorial and not simply a matter of motivation. STRONG’s supervised, centre-based design addresses several of those factors by construction; whether it achieves adherence is an empirical question the trial should report in detail.
Attrition is the other half of the problem. FINGER lost 12% of participants over two years [10]. A 12-month intervention with a further 12 months of follow-up in a frail older population should expect at least that, and possibly more. Every participant lost reduces the detectable effect size (Section 10).
9.5 Intention-to-treat, per-protocol, and what “superiority” means when adherence decays
The protocol describes a superiority trial. Superiority is normally judged on the intention-to-treat population, which estimates the effect of assigning the intervention rather than the effect of receiving it. In a trial with imperfect adherence, the intention-to-treat estimate is diluted toward the null by the proportion of participants who did not receive the intervention as intended.
This creates a predictable interpretive problem. If STRONG is null on intention-to-treat but positive on per-protocol, the correct conclusion is not “the intervention works in people who take it” — that is a different, weaker claim, and per-protocol analyses are vulnerable to confounding by the characteristics that predict adherence. The correct conclusion is that the intervention did not work well enough, in the population studied, to be recommended as delivered. If STRONG is positive on intention-to-treat, the effect is a lower bound on efficacy and the per-protocol analysis can characterise the dose-response.
The protocol’s description of the primary analysis is not available in the fields retrieved, so it is not possible to say whether the intention-to-treat approach is pre-specified. It should be, and it should be reported with the per-protocol analysis alongside it.
9.6 The 24-month horizon
The design includes 12 months of intervention and a further 12 months of follow-up, to 24 months total [1]. That is a sensible choice and it creates two distinct questions. The first is whether the intervention produces effects at 12 months, while it is being delivered. The second is whether any effects persist after it stops, which is the question of whether the intervention changed the trajectory rather than the state.
Both questions are worth asking, and they have different answers in the literature. Exercise effects on physical function attenuate after supervised programmes stop, and the LIFE study’s benefits were observed during a programme that continued for 2.6 years rather than after it ended [2]. A 12-month intervention followed by 12 months of observation is likely to show attenuation, and attenuation should not be interpreted as failure — it is what happens when a stimulus is removed. But it does mean that the 24-month data will be more informative about sustainability than about efficacy.
10. Endpoints and detectability: surrogates versus hard outcomes, Framingham-score insensitivity, and what 240 participants can find
10.1 The endpoint hierarchy, and where each STRONG outcome sits
It is useful to sort the registered outcomes by how much a positive result would actually mean.
At the top are clinical events: cardiovascular death, myocardial infarction, stroke, heart-failure hospitalisation, and death from any cause. STRONG does not register any of these as a primary or secondary outcome [1]. That is not a criticism — a 240-patient, 12-month trial cannot study events — but it defines the ceiling of what the trial can conclude.
One rung down are validated functional outcomes with established prognostic value: the 6-minute walk test, gait speed, the Short Physical Performance Battery, and mobility disability. These predict events and death, and they matter to patients. STRONG measures 6-minute walk and functional fitness, balance and mobility as secondary outcomes; the 400-metre mobility disability used by LIFE and SPRINTT is not among the registered outcomes [1].
Below those are risk factors and risk scores: blood pressure, lipids, Framingham and Reynolds risk scores. These are validated predictors at population level and weak proxies for individual benefit over 12 months.
Below those are mechanistic biomarkers: the cytokine panel, echocardiographic strain, and the frailty instruments. They tell us whether the intervention engaged the biology. They do not tell us whether the patient will live longer or better.
A review that treats all of these as equivalent — “the trial will show whether the intervention improves cardiovascular health” — is doing the trial a disservice and the reader a bigger one.
10.2 Framingham risk score as an outcome is a category error waiting to happen
The Framingham risk score is a ten-year absolute risk function derived from a community cohort, calibrated to a population and an era, and intended to guide the decision to start preventive therapy. Using its change over 12 months as an outcome in a lifestyle trial is problematic for three reasons.
First, a ten-year risk estimate moves slowly. A 12-month change in systolic pressure of a few mmHg, or in HDL cholesterol of 0.08 mmol/L, produces a small absolute change in a ten-year risk percentage. The intervention could be biologically active and the risk score could barely move. Second, the score is dominated by age and sex, which do not change; the modifiable components contribute a minority of the total. Third, the score was not developed or validated as a measure of treatment effect, and we did not identify a trial that has validated change in Framingham risk as a surrogate for events.
That does not make the measurement useless. A change in the risk score is a transparent way to summarise changes in several risk factors at once, and it is inexpensive because the inputs are already being collected. But a null result on Framingham risk would not show that the intervention failed, and a positive result would not show that it worked. It is a descriptive summary, not a verdict.
10.3 What effect sizes are realistic for blood pressure and lipids?
The best available anchors are modest. The blueberry meta-analysis found a diastolic reduction of 2.20 mmHg (95% CI −4.13 to −0.27) and no overall systolic effect (−1.43 mmHg, −3.11 to 0.26), with a larger systolic effect in smokers [7]. The healthy-population meta-analysis found no blood-pressure effect at all [8]. Exercise meta-analyses in hypertensive populations show larger effects when the prescription adheres to ACSM recommendations (standardised systolic effect −1.20 vs −0.75 for lower-adherence programmes), but these are standardised effects across heterogeneous trials, not millimetres of mercury that can be added to a patient’s chart [62].
For lipids, the blueberry literature is null: the meta-analysis found no effect on HDL or LDL cholesterol overall [7], and the 6-month trial’s HDL effect was confined to statin non-users and was small (+0.08 mmol/L) [9]. Diet-induced LDL changes are possible with larger dietary change than a blueberry supplement, but the STRONG package does not include a structured lipid-lowering diet. Against a background of high statin use in this population — the 2021 Canadian Cardiovascular Society guidelines recommend statins for most patients with clinical atherosclerosis, diabetes or CKD [63] — the incremental lipid effect of the package is likely to be small.
Put together, a realistic 12-month effect on the risk-factor outcomes is a systolic change in the low single digits and a diastolic change of about 2 mmHg, with little or no lipid change. That is not nothing at population level. It is not a cardiovascular event reduction at individual level.
10.4 The arithmetic of n = 240
The trial plans 240 participants, 120 per arm, with 12 months of intervention and 24 months of follow-up [1]. The following calculations are arithmetic, not protocol claims; the protocol reports no power calculation, so these are the consequences of its sample size under conventional assumptions.
For a continuous outcome analysed with a two-sample comparison, two-sided α = 0.05:
| Attrition | n per arm | Detectable standardised difference, 80% power | 90% power |
|---|---|---|---|
| 0% | 120 | 0.36 SD | 0.42 SD |
| 15% | 102 | 0.39 SD | 0.45 SD |
| 25% | 90 | 0.42 SD | 0.48 SD |
A standardised difference of 0.36–0.42 SD is a moderate effect. It is achievable for physical function outcomes in a frail population — the pooled exercise effect on frailty scores was SMD −0.57, and on gait speed 0.09 m/s — but it is not achievable for outcomes that move slowly or for effects that are diluted by imperfect adherence.
For the 6-minute walk, the detectable difference depends entirely on the standard deviation of the change score, which this review could not retrieve for this population. If that standard deviation is 40 m, the trial can detect about 14 m; if 50 m, about 18 m; if 60 m, about 22 m. The anchor-based minimal clinically important difference in moderately frail older adults was 17.8 m [30]. The trial is therefore roughly at the threshold of detecting a clinically important change on this outcome, with no margin for attrition. That is a narrow window, and it is the single most important number for interpreting a null 6-minute walk result.
For a binary outcome, the power is worse than most clinicians expect. With 120 per arm, a reduction from 30% to 18% — a 40% relative reduction, which would be a large effect for a lifestyle package — has about 59% power. A reduction from 30% to 20% has about 44%. A reduction from 15% to 9% has about 30%. With 25% attrition, those fall to roughly 48%, 34% and 24% respectively. In other words, the trial is underpowered for anything but a large absolute reduction in a common binary outcome.
For clinical events, the arithmetic is decisive. Detecting a hazard ratio of 0.75 at 80% power with two-sided α = 0.05 requires about 379 events; an HR of 0.70 requires about 247, and an HR of 0.60 about 120. At a 10% annual event rate, 240 participants followed for one year would be expected to produce about 24 events, and at 15% about 36. The trial would need roughly ten to fifteen times its sample, or a much longer follow-up, to test a plausible event effect. No analysis of STRONG’s event data can be informative, and the protocol does not propose one.
10.5 Multiplicity: seven primary outcomes and no stated rule
Seven co-primary outcomes, each tested at α = 0.05, give a probability of at least one false-positive result of about 30% if the outcomes were independent (1 − 0.95⁷ = 0.302). They are not independent — frailty instruments correlate, blood pressure and Framingham risk correlate — so the true family-wise error rate is lower than 30% but still substantially above 5%. The registry record, as retrieved, does not describe an alpha-spending plan, a hierarchical testing order, or a composite primary endpoint [1].
This is the most likely route by which STRONG produces a misleading headline. A trial with seven outcomes will almost always produce a mixture of significant and non-significant results, and the temptation to lead with the significant one is considerable. The defence is pre-registration of a testing hierarchy, and the registry entry does not contain one. Readers of the eventual paper should look for it, and its absence should be treated as a limitation of the report rather than a detail.
10.6 Sex as a pre-specified dimension
The trial’s 120:120 sex balance is a genuine strength, and the protocol’s rationale explicitly raises the possibility that frailty–cardiovascular relationships are sex-specific [1] [19]. But a 120:120 design does not give the trial power to detect sex-by-treatment interactions. Detecting an interaction of moderate size typically requires about four times the sample of detecting a main effect, which would put the requirement in the high hundreds. A subgroup analysis by sex will therefore be descriptive, and any apparent difference between men and women should be reported as a hypothesis.
That does not make the balance pointless. Equal recruitment by sex ensures that the main estimates are not dominated by one sex, and it makes the descriptive subgroup analyses interpretable. It does not make them confirmatory.
10.7 What n = 240 buys, in one paragraph
It buys a well-characterised cohort, a feasible intervention, a description of adherence and attrition, estimates of effect on several functional and risk-factor outcomes with confidence intervals wide enough to exclude only large effects, and a rich mechanistic dataset that will generate hypotheses for a decade. It does not buy a definitive answer on frailty reversal, and it certainly does not buy an answer on cardiovascular events. The trial’s value will depend on whether it reports its estimates honestly, with the confidence intervals and the pre-specified hierarchy visible, rather than on whether it produces a significant p-value.
11. Safety, tolerability and interaction profile of each component
11.1 Exercise: the benefits are real and so are the musculoskeletal costs
The most direct safety data come from FINGER, where adverse events occurred in 7% of the multidomain intervention group versus 1% of controls, with musculoskeletal pain accounting for most of the excess (5% versus 0%) [10]. That is a modest absolute risk in a largely healthy population of 60–77 year-olds, and it is the price of a programme that includes supervised exercise. In a frailer population, the risk of musculoskeletal injury and of falls during exercise is higher, and the trials that enrolled frail participants screened for eligibility and delivered the intervention under supervision [3] [2].
Falls are the safety outcome that matters most, and the evidence is not uniformly reassuring. Exercise improves gait speed, Timed Up and Go and sit-to-stand performance in frail older adults with low-to-very-low certainty [31], but the Cochrane review found that multiple-component interventions may have little or no effect on falls rate (rate ratio 0.91, 0.80–1.04) and may reduce the risk of falling (RR 0.88, 0.80–0.97) [36]. A programme that increases activity in a frail person can increase exposure to fall risk even as it improves the capacity to avoid falling. The net effect in an individual is not predictable from the trial averages.
The practical implication for a clinician referring a patient to a programme like STRONG’s is straightforward: the programme should screen for falls risk and musculoskeletal disease, start at a tolerable intensity, and progress under supervision. STRONG’s eligibility criteria exclude medical conditions that prevent participation [1], which is a reasonable screen but not a falls assessment.
11.2 Protein: the kidney is the question, and the trial does not measure it
The safety debate about protein in older adults is almost entirely about the kidney. The evidence summarised in Section 5 suggests that high-protein diets in people without CKD raise eGFR without raising creatinine, consistent with haemodynamic adaptation rather than injury [48]. The guideline position is that people with eGFR below 30 mL/min/1.73 m² who are not on dialysis may need to limit protein [41], and that people with acute or chronic illness need more protein, not less [42].
The gap is ascertainment. A 30 g/day supplement is unlikely to harm a person with an eGFR of 45, and it is unlikely to harm a person with an eGFR of 25 over 12 months either, but nobody knows, because the trials exclude the latter. STRONG’s registry eligibility does not list CKD or an eGFR threshold [1], and kidney function is not among the registered outcomes. A physician whose patient has CKD stage 4 should not assume the trial’s safety data will apply, and a trial report that does not describe baseline kidney function will not be able to reassure them.
There is a second, less discussed issue: satiety and displacement. A 30 g protein supplement taken before or with a meal can reduce intake of other foods, and in an older adult at risk of malnutrition the net effect on energy and micronutrient intake could be negative. This is an inference rather than a documented harm, and the trial’s ASA24 dietary assessment should be able to detect it. It is worth watching.
11.3 Blueberries: a low-risk food with a specific, narrow interaction concern
The interaction concern with berries and warfarin is vitamin K. A systematic review of food and oral anticoagulant interactions identified green leafy vegetables supplying more than 250 µg/day of vitamin K as the dietary exposure that should be cautioned in warfarin patients, and advised maintaining a highly constant intake of green leafy vegetables to ensure stable anticoagulation; it also identified ginger, mango, St John’s wort and grapefruit as interactions of concern, and found cranberry juice up to 240 mL/day probably safe [64]. Berries are not among the foods flagged.
That does not mean no interaction exists; it means the interaction literature does not identify one, and the mechanism that would produce one — a large vitamin K load — is not a feature of a one-cup blueberry serving. The practical advice for a warfarin patient is the standard one: keep intake consistent rather than avoid the food, and monitor the INR if a large dietary change is made. A patient who starts eating a cup of blueberries every day is making a dietary change, and consistency is the mitigation.
Anthocyanins are also substrates and possible modulators of drug transporters and metabolising enzymes, and the bioavailability literature notes extensive metabolism and instability [49]. No clinically important interaction with commonly prescribed cardiovascular drugs has been established in the retrieved literature, and the trial’s eligibility criteria exclude only allergy or dislike [1]. The realistic risk is low. The realistic concern is a patient on warfarin whose INR becomes labile after a dietary change, which is a monitoring problem rather than a contraindication.
11.4 Supplement quality
Protein powders are dietary supplements, and dietary supplements are a category with documented adulteration risk. A 2025 review of laboratory approaches to supplement adulteration notes that manufacturers selling in the United States must comply with good manufacturing practice, including identity, purity, strength and composition testing and limits on contaminants, and that certification programmes, supplier quality agreements and third-party testing are the tools that mitigate risk [65]. That is a general statement about the category, not a finding about any particular product, and it does not imply that a trial-grade whey isolate is adulterated. It does mean that the identity and purity of the supplement should be documented in the trial report, and that a clinician recommending a protein powder should recommend a product with third-party testing rather than an arbitrary one.
11.5 The safety profile in one paragraph
Exercise carries a real but modest risk of musculoskeletal pain and injury, and an uncertain effect on falls that depends on the population and the programme. Protein carries a theoretical renal risk concentrated in people with unrecognised or advanced CKD, and a possible risk of dietary displacement in malnourished patients. Blueberries carry a low interaction risk, with warfarin the only plausible concern and dietary consistency the mitigation. None of these is a reason not to test the package; all of them are reasons for the trial to measure kidney function, falls, musculoskeletal adverse events and dietary intake, and to report them.
12. Delivery, adherence, workforce and cost — including the Atlantic Canadian context
12.1 The delivery burden, counted rather than estimated
Three 60-minute supervised sessions per week for 52 weeks is 156 sessions per participant. For 120 intervention participants, that is 18,720 supervised sessions, plus the space, equipment, scheduling and documentation that go with them [1]. This review did not retrieve a cost per session for exercise physiology in Nova Scotia, and it does not attempt a dollar estimate; the point of the arithmetic is that the delivery burden is an order of magnitude larger than a prescription, and it is the binding constraint on scalability.
The alternative is home-based or group-based delivery with less individual supervision. The trials that used that model — LIFE’s home component, SPRINTT’s actimetry-tailored home programme, DO-HEALTH’s simple home exercise — produced mixed results, and the home-based programme in DO-HEALTH had no effect on falls in healthy older adults [58]. That does not prove supervision is necessary, but it is consistent with the proposition that contact and structure matter.
12.2 Adherence in the real world
The barriers and facilitators literature is large and consistent: 64 studies, 54 factors and 38 behaviour-change techniques, organised into eight strategy areas — tailored exercise programmes, appropriate environments, multidimensional social support, monitoring and feedback, managing emotional experiences, education, self-efficacy and autonomy [37]. The trial’s design addresses several of these by construction: supervision provides monitoring and feedback, the group setting provides social support, and the individualised prescription provides tailoring.
Whether that translates into adherence over 12 months in a Nova Scotian population is unknown, and it is the single most important process outcome the trial could report. The blueberry trial achieved 94.1% compliance by wrapper returns over six months [9], which shows that a provided food can be consumed reliably in a trial; it does not show that a supervised exercise programme can be attended reliably for a year.
12.3 Exercise referral and food-as-medicine: what the delivery literature shows
Exercise referral schemes — primary care referral to a supervised programme — have been studied extensively. An umbrella review of 12 reviews covering 110 individual studies and 62,815 participants found that earlier reviews observed no change in physical activity or fitness, while more recent, larger studies observed low-to-moderate improvements; recent reviews reported reductions in blood pressure, body mass index and depression, and cost-effectiveness findings were conflicting [66]. That is a familiar pattern in implementation research: early small studies are null, later larger studies are modestly positive, and the economic case is unresolved.
Produce prescription programmes are the closest analogue to the blueberry component. A three-arm randomised community trial of monthly electronic benefits of $40, $80 or $110 for six months in 242 adults found a dose-response: compared with $40, the $110 dose significantly improved food insecurity and fruit and vegetable intake at three months and benefit redemption at six months, with redemption rates of 60.2%, 67.6% and 70.4% across the three arms [67]. The population was young (mean age 35) and predominantly female and Hispanic, so the results do not transfer directly to older Nova Scotians. But the design lesson does: dose matters, and a benefit that is too small to change purchasing behaviour will not change intake.
12.4 The Nova Scotia context
The most directly relevant local data come from a linked cohort of 944 older adults who received comprehensive geriatric assessment after emergency department referral to internal medicine in Nova Scotia. Frailty was measured with the Clinical Frailty Scale; 41.6% died within one year. Physician service costs were concentrated in the months after the emergency presentation, were about $300–$400 higher for mildly and moderately frail patients than for well patients (estimated imprecisely), and were $704 lower for severely frail patients over one year, primarily because they died sooner [68].
Three implications follow. First, the population that a frailty trial wants to enrol is a population with high one-year mortality, which means competing risks and missing data will be substantial. Second, the cost signal is non-linear and survivorship-adjusted, so a naïve “frailty costs more” claim is wrong in this setting. Third, the health-system case for an intervention is not just about averting costs — in a population where 41.6% die within a year, the case is about function, dignity and days at home.
Atlantic Canada also brings geography. A three-times-weekly centre-based programme is feasible in Halifax or Sydney and much harder in a rural county where the nearest exercise physiologist may be an hour away. The trial’s single-province design will not answer the rural-access question unless it deliberately samples rural participants, and the registry does not describe a rural sampling frame [1].
12.5 Equity and food security
The intervention supplies the blueberries and the protein powder for 12 months. That removes cost as a barrier during the trial and does not remove it afterwards. A cup of fresh blueberries a day is an expensive habit in Atlantic Canada for much of the year, and frozen fruit is cheaper but not free. A protein powder is cheaper per gram of protein than most whole foods, which is one argument for supplementation in food-insecure populations, but it does not address the broader dietary pattern.
The trial measures diet quality with the ASA24 and household food-work with the FIA-Q10 [1], which suggests the investigators are aware of the issue. Whether the trial reports equity-relevant subgroup analyses — income, rurality, food security — is not described in the registry record, and it should.
12.6 What implementation would actually look like
If the package worked, implementation would require four things that the Canadian system does not routinely provide together: an exercise physiologist or trained kinesiologist available three times a week; a supply of fruit and supplement; a dietitian to assess protein intake and kidney function; and a primary care clinician to coordinate. Cardiac rehabilitation provides a partial template — it has infrastructure, reimbursement and a culture of participation, and the PipELINe result suggests it can deliver a multidomain programme that reduces events in frail post-MI patients [59]. Extending that model to community-dwelling frail older adults without a recent cardiac event would require a different funding model.
The honest conclusion is that the delivery model being tested is a research model, and the trial will tell us whether the intervention can work under favourable conditions, not whether it can be delivered at scale.
13. Guidelines today: physical activity, protein, and the absence of a polyphenol recommendation
13.1 Physical activity
The WHO 2020 guidelines recommend that all adults undertake 150–300 minutes of moderate-intensity or 75–150 minutes of vigorous-intensity aerobic physical activity per week, or an equivalent combination, and that regular muscle-strengthening activity be undertaken by all age groups; they also recommend reducing sedentary behaviour, while noting that the evidence was insufficient to quantify a threshold [69]. The guidelines apply to older adults and to people living with chronic conditions, and they emphasise that some activity is better than none.
The Canadian 24-Hour Movement Guidelines for Adults aged 18–64 and 65 and older integrate physical activity, sedentary behaviour and sleep into a single 24-hour recommendation, developed through an AGREE II process with two de novo systematic reviews and four overviews of reviews, and they explicitly address the older adult population [70]. The integration matters for frailty because sleep and sedentary time are part of the same behavioural system as exercise, and an intervention that increases exercise while displacing sleep may not improve health.
The International Exercise Recommendations in Older Adults (ICFSR) go further and treat exercise as a prescription with a dose, an indication and a titration, arguing that exercise programmes improve the hallmarks of frailty and cognition and should be individualised and adjusted like any other therapeutic agent [71].
13.2 Protein
The PROT-AGE and ESPEN recommendations are summarised in Section 5: at least 1.0–1.2 g/kg/day for healthy older adults, 1.2–1.5 g/kg/day in acute or chronic illness, with an exception for severe kidney disease (eGFR < 30) not on dialysis [41] [42]. Both emphasise food first and exercise alongside.
The Fisterra clinical practice guideline for frailty in primary care recommends active opportunistic case finding in people aged 70 and older using simple performance-based tools, and multicomponent interventions integrating physical exercise, nutritional support, medication optimisation and psychosocial strategies as effective in preventing or reversing frailty [72]. It is a primary care guideline, and it is the closest thing in the retrieved literature to a recommendation for a package rather than a component.
13.3 Cardiovascular prevention
The 2021 Canadian Cardiovascular Society dyslipidemia guidelines recommend statin therapy for patients with clinical atherosclerosis, abdominal aortic aneurysm, most patients with diabetes or chronic kidney disease, and those with LDL cholesterol ≥5 mmol/L; for others they emphasise risk assessment linked to lipid evaluation, and they describe health behaviour modification — regular exercise and a heart-healthy diet — as the cornerstone of prevention [63]. The guideline’s structure is instructive: behaviour is the cornerstone, statins are the evidence-based therapy, and the two are not presented as alternatives.
That is the correct frame for a clinician thinking about STRONG. The trial is testing an adjunct to standard prevention, not a substitute for it. A patient who stops a statin in order to eat blueberries and exercise has misunderstood the evidence, and the trial’s reporting should not encourage that misunderstanding.
13.4 The polyphenol gap
None of the guidelines retrieved for this review recommends polyphenol supplementation, and none recommends a specific berry intake for cardiovascular prevention. The WHO guidelines address physical activity and sedentary behaviour, not diet. The Canadian 24-Hour Movement Guidelines address movement, not food. PROT-AGE and ESPEN address protein. The CCS guideline addresses lipids and behaviour in general terms. The Fisterra guideline addresses multicomponent frailty management, including nutritional support, without naming polyphenols [69] [70] [41] [42] [63] [72].
This is not an oversight. It reflects the state of the evidence: a reproducible effect on a surrogate (flow-mediated dilation) without evidence of effect on clinical outcomes is not enough to justify a population-level recommendation. A clinician can reasonably tell a patient that blueberries are a healthy food and that a cup a day may improve a marker of vascular function. A clinician cannot reasonably tell a patient that blueberries prevent heart attacks, and no guideline does.
13.5 Adjacent and competing interventions
A clinician deciding what to do for a frail 78-year-old with cardiovascular disease is choosing among several strategies, not just deciding whether to recommend blueberries.
Deprescribing and medication optimisation. The Fisterra frailty guideline includes medication optimisation as one of the components of effective multicomponent care [72], and the heart-failure–frailty review lists structured medication review alongside exercise rehabilitation and nutritional optimisation [22]. For a patient on twelve medications, stopping three may produce more functional benefit than adding a protein shake, and it carries less delivery burden. The two are not mutually exclusive, but a trial that tests only the additive package cannot estimate the opportunity cost.
Vitamin D and omega-3. DO-HEALTH tested both in a 2 × 2 × 2 factorial design in 2,157 healthy adults aged 70 and older and found no benefit of high-dose vitamin D or a home exercise programme on falls, with a possible modest benefit of omega-3 on total falls [58]. These are the two supplements most commonly taken by older adults for musculoskeletal and cardiovascular reasons, and the trial is a useful corrective: in a vitamin D-replete population, more vitamin D did not help. A patient who is vitamin D deficient should still be treated; a patient who is replete should not expect a falls benefit.
Incretin therapy and the muscle problem. Glucagon-like peptide-1 receptor agonists and dual incretin agonists are now central to cardiovascular prevention in people with obesity and established cardiovascular disease, and they cause substantial lean mass loss — estimated at 15–45% of total weight reduction in one review [73]. Whether that lean mass loss matters functionally is contested. One 2026 analysis argues that much of the DXA-measured lean mass decline represents physiological adaptation to reduced mechanical loading rather than pathological muscle loss, and proposes testable predictions to distinguish the two [74]. The clinical question is real: a 78-year-old who loses 12 kg on semaglutide and 4 kg of it as lean mass may be lighter, metabolically better and functionally worse.
The SELECT post hoc analysis is the best available evidence on whether frailty modifies incretin benefit. Among 17,604 participants with established cardiovascular disease and overweight or obesity, the primary composite outcome became more common as baseline frailty index rose, and the benefit of semaglutide appeared consistent across frailty categories (HR 0.84, 95% CI 0.65–1.07 for the least frail; 0.70, 0.59–0.82 for the middle category; 0.92, 0.76–1.10 for the most frail; P for interaction 0.09) [75]. The confidence intervals overlap and the interaction is not significant; the honest reading is that there is no clear evidence of effect modification, not that semaglutide works best in the middle frailty category.
The relevance to STRONG is conceptual rather than competitive. If incretin therapy is going to be widely prescribed to older adults with cardiovascular disease, then a cheap intervention that preserves muscle mass and function during weight loss becomes more valuable, not less. Protein plus resistance exercise is the obvious candidate for that role, and the component evidence supports a modest effect on lean mass and strength [6] [5]. A future trial that randomised protein-plus-exercise against usual care in patients starting an incretin agonist would answer a more contemporary question than STRONG’s, and would probably recruit faster.
13.6 Where the guidelines leave the clinician
The guidelines recommend, separately and with different strengths of evidence: exercise (strong, population-level), protein at 1.0–1.2 g/kg/day (consensus, moderate), a heart-healthy dietary pattern (strong, general), and statins for secondary prevention and high-risk primary prevention (strong). They do not recommend a combined package, they do not recommend blueberries, and they do not address the question of whether adding protein and polyphenols to exercise produces more benefit than exercise alone.
That is exactly the gap STRONG is trying to fill, and it is a legitimate gap. The trial’s contribution will be to describe what happens when the three are combined in a frail population, not to establish a new guideline.
14. Bottom line for clinicians
14.1 What is defensible to recommend today
Exercise. Recommend multicomponent exercise — aerobic, resistance and balance — to older adults with frailty or pre-frailty, at a volume consistent with the WHO and Canadian guidelines, and with supervision at the start for people who are frail or at risk of falls. The evidence for an effect on physical function and mobility disability is the strongest in this review: a 5.4-percentage-point absolute reduction in major mobility disability in LIFE (HR 0.82, 0.69–0.98) [2] and a 5.9-point reduction in SPRINTT’s lower-SPPB subgroup (HR 0.78, 0.67–0.92) [3]. The evidence for falls prevention is weaker than commonly assumed (Cochrane multiple-component rate ratio 0.91, 0.80–1.04) [36]. The evidence for hard cardiovascular event reduction does not exist.
Protein. Recommend a total intake of 1.0–1.2 g/kg/day for healthy older adults and 1.2–1.5 g/kg/day during illness, food first, with attention to distribution across meals and to total energy intake [41] [42]. Check kidney function before pushing protein in someone with known CKD, and remember the eGFR <30 exception. Protein powder is a means of reaching the target, not a therapy in itself, and protein without exercise does little for function [6].
Blueberries. Recommend them as food — a palatable, nutrient-dense fruit — and be honest about the evidence: a cup a day for six months improved flow-mediated dilation by about 1.45 percentage points and reduced augmentation index by about 2.24% in adults with metabolic syndrome [9], with meta-analytic effects of 1.1–1.5 percentage points on FMD and inconsistent blood-pressure effects [7] [8]. There is no evidence that blueberries reduce cardiovascular events, and no guideline recommends them for that purpose. For patients on warfarin, consistency of intake matters more than avoidance [64].
Everything else. Continue statins and other evidence-based secondary prevention. Do not substitute the package for them. The CCS guideline’s framing — behaviour as the cornerstone, statins as the evidence-based therapy — is the right one [63].
14.2 What STRONG can and cannot settle
It can settle whether a 12-month supervised package is feasible and acceptable in older Nova Scotians, what adherence and attrition look like, and whether the package produces effects on frailty instruments, physical function, blood pressure, lipids, inflammatory markers and echocardiographic measures that are large enough to justify a larger trial. That is a real contribution, and it is more than most lifestyle trials deliver.
It cannot settle whether the package prevents cardiovascular events, whether it reverses frailty in a clinically meaningful way, whether any one component is responsible for the effect, or whether the model is deliverable outside a research setting. Those questions require a different design, a larger sample, a longer horizon, and a delivery model that looks like the health system rather than a trial unit.
14.3 Signals to watch when the results land
Five things will distinguish a meaningful result from a noisy one.
-
Consistency across instruments. If the Frailty Index, the Clinical Frailty Scale and the Pictorial Fit-Frail Scale all move in the same direction, the result is more credible than if only one moves. If only the self-reported PFFS moves, the result is compatible with unblinded response bias.
-
Effect size relative to the minimal clinically important difference. A 6-minute walk improvement of 18 m or more is at the MCID threshold [30]; a gait speed improvement of 0.09 m/s is small [31]; an SPPB improvement of 0.8 points is modest [3]. A statistically significant result below these thresholds is not a clinical win.
-
Adherence and dose-response. If the effect is concentrated in participants who attended most sessions, that supports a causal interpretation. If it is uniform across adherence levels, it suggests a non-specific effect.
-
Blinded versus unblinded outcomes. Effects on measured performance and blinded assessment are more informative than effects on self-report. A trial that reports only the self-reported outcomes should be read with caution.
-
Safety. Musculoskeletal adverse events, falls and kidney function. FINGER’s 7% versus 1% adverse-event rate is the benchmark [10], and a frailer population should be expected to do worse.
14.4 What would actually change practice
A single 240-patient trial will not. Practice would change if a larger trial, in a population similar to the one clinicians actually see, showed a consistent effect on a validated functional outcome plus a signal on a clinical endpoint; if the effect persisted after the intervention stopped; if the delivery model were feasible in a non-research setting; and if the cost per quality-adjusted life-year were within the range that payers accept. None of those conditions is met by any trial in this literature, including the positive ones.
What would change practice faster is a modest, reproducible effect on a measure that clinicians already use — gait speed, 6-minute walk, SPPB — combined with a delivery model that primary care can actually prescribe. That is a lower bar than event reduction, and it is the bar STRONG has a realistic chance of clearing.
Limitations
The seed is a protocol with no results. Every statement about STRONG in this review describes its design, its registered endpoints, or the arithmetic consequences of its sample size. The protocol abstract does not name a primary outcome, report a power calculation, or report any effect size [17]. The seven primary outcomes, the single-masked design, the absence of a data monitoring committee and the eligibility criteria are taken from the ClinicalTrials.gov record and are labelled as registry-derived [1]. No sentence in this review should be read as reporting a STRONG result, because none exists.
The evidence is abstract-level. No full texts were downloaded for the cited biomedical papers. All numbers attributed to a trial or meta-analysis come from its abstract, and methods details — randomisation procedures, adherence definitions, adverse-event ascertainment — are therefore unavailable. Where a number is reported only in an abstract, the review says so; where a mechanism or interpretation is inferred, the review marks it as an inference. The consequence is that some claims are less precise than a full-text review would allow, and some design details of the component trials cannot be verified.
Selection was purposive, not systematic. There was no dual screening, no PRISMA flow, no risk-of-bias instrument, and no formal grey-literature search. Approximately 120 queries were run across PubMed, Europe PMC and ClinicalTrials.gov, and roughly 85 abstracts were read; the citations were selected for relevance to the sixteen review axes, not by an exhaustive or reproducible protocol. A systematic review on any one of these axes would reach more sources and might reach different conclusions. The counts reported in the provenance file describe what was done, not a systematic sample of the literature.
The component literatures are heterogeneous, and the review says so but cannot resolve it. Exercise effects on frailty scores range from SMD −0.57 to −0.95 across meta-analyses with I² of 88% and above, with prediction intervals that cross the null [4] [34]. Blueberry effects on FMD range from 1.11 to 1.50 percentage points depending on the population and the pooling [7] [8]. Protein effects are consistently small and consistently rated low or very low certainty [6]. Pooling heterogeneous trials produces averages that may describe no actual population, and the review’s statements should be read as ranges rather than point estimates.
The surrogate-endpoint problem is not solvable within this literature. Flow-mediated dilation, augmentation index, inflammatory cytokines and global longitudinal strain are risk markers and mechanism probes, not validated surrogates for cardiovascular events. The review treats them as such. A trial that improves them has not shown that it prevents heart attacks, and a guideline that recommends an intervention on the basis of them would be overreaching.
Costs are not estimated. The review counts sessions and describes the delivery burden but does not convert it to dollars, because no reliable per-session cost for exercise physiology in Nova Scotia was retrieved. The Nova Scotia cost data that are cited describe physician service costs in a frail cohort, not intervention costs [68]. A health-economic evaluation of the package would require a costing study that this review did not attempt.
Registry searching has known limits. ClinicalTrials.gov’s query.term engine matches across fields and produces false positives. The reported counts — 182 for “blueberry”, 940 for “exercise frailty older adults”, 3 for “blueberry frailty” — are raw hit counts, and the three “blueberry frailty” hits include two studies that are not blueberry–frailty trials at all [76]. WHO ICTRP and national registries were not searched, so “the only registered trial of this combination” is a claim about ClinicalTrials.gov only, and the review makes the weaker claim that only one trial in this registry combines the three components.
Competing interests are disclosed, not resolved. The seed protocol’s competing-interest statement records food-industry contracts for one investigator [17]. That disclosure is reproduced in Section 1.4. It does not establish bias, and it does not disqualify the trial. It does mean that independent replication and transparent reporting of the primary outcome hierarchy matter more than usual.
The review was prepared on a single date with a single reviewer. The searches, the screening decisions and the synthesis were performed by one agent on 2026-11-02. No second reviewer checked the extraction of numbers. Errors of transcription are possible, and the reference list is generated mechanically from the retrieved records rather than checked by hand against the source. The provenance file records exactly which checks were performed and which were not.
References
- ClinicalTrials.gov. NCT06693271: registered study record. National Library of Medicine. https://clinicaltrials.gov/study/NCT06693271
- Pahor M, Guralnik JM, Ambrosius WT, Blair S, Bonds DE, Church TS, et al. Effect of structured physical activity on prevention of major mobility disability in older adults: the LIFE study randomized clinical trial. JAMA. 2014;311(23):2387-96. doi:10.1001/jama.2014.5616. PMID: 24866862.
- Bernabei R, Landi F, Calvani R, Cesari M, Del Signore S, Anker SD, et al. Multicomponent intervention to prevent mobility disability in frail older adults: randomised controlled trial (SPRINTT project). BMJ. 2022;377:e068788. doi:10.1136/bmj-2021-068788. PMID: 35545258.
- Li J, Yang H, Chen L. A systematic review and meta-analysis of the effects of different exercise modalities on frailty in older adults. Front Public Health. 2026;14:1931299. doi:10.3389/fpubh.2026.1931299. PMID: 42891303.
- Morton RW, Murphy KT, McKellar SR, Schoenfeld BJ, Henselmans M, Helms E, et al. A systematic review, meta-analysis and meta-regression of the effect of protein supplementation on resistance training-induced gains in muscle mass and strength in healthy adults. Br J Sports Med. 2018;52(6):376-384. doi:10.1136/bjsports-2017-097608. PMID: 28698222.
- Yoshimura Y, Matsumoto A, Inoue T, Okamura M, Kuzuya M. Protein supplementation alone or combined with exercise for sarcopenia and physical frailty: A systematic review and meta-analysis of randomized controlled trials. Arch Gerontol Geriatr. 2025;131:105783. doi:10.1016/j.archger.2025.105783. PMID: 39955964.
- Deng B, Lei Y, Zhou R, Ruan T, Lu W, Ying J, et al. Effect of blueberry intervention on endothelial function: a systematic review and meta-analysis. Front Physiol. 2024;15:1368892. doi:10.3389/fphys.2024.1368892. PMID: 38887319.
- He Y, Yang J, Zhang S, Mai J, Li M. Blueberry consumption enhances vascular function but not cognitive abilities in healthy individuals: a systematic review and meta-analysis. Nutr Res. 2026;149:162-177. doi:10.1016/j.nutres.2026.03.007. PMID: 41974068.
- Curtis PJ, van der Velpen V, Berends L, Jennings A, Feelisch M, Umpleby AM, et al. Blueberries improve biomarkers of cardiometabolic function in participants with metabolic syndrome-results from a 6-month, double-blind, randomized controlled trial. Am J Clin Nutr. 2019;109(6):1535-1545. doi:10.1093/ajcn/nqy380. PMID: 31136659.
- Ngandu T, Lehtisalo J, Solomon A, Levälahti E, Ahtiluoto S, Antikainen R, et al. A 2 year multidomain intervention of diet, exercise, cognitive training, and vascular risk monitoring versus control to prevent cognitive decline in at-risk elderly people (FINGER): a randomised controlled trial. Lancet. 2015;385(9984):2255-63. doi:10.1016/S0140-6736(15)60461-5. PMID: 25771249.
- Andrieu S, Guyonnet S, Coley N, Cantet C, Bonnefoy M, Bordes S, et al. Effect of long-term omega 3 polyunsaturated fatty acid supplementation with or without multidomain intervention on cognitive function in elderly adults with memory complaints (MAPT): a randomised, placebo-controlled trial. Lancet Neurol. 2017;16(5):377-389. doi:10.1016/S1474-4422(17)30040-6. PMID: 28359749.
- Moll van Charante EP, Richard E, Eurelings LS, van Dalen JW, Ligthart SA, van Bussel EF, et al. Effectiveness of a 6-year multidomain vascular care intervention to prevent dementia (preDIVA): a cluster-randomised controlled trial. Lancet. 2016;388(10046):797-805. doi:10.1016/S0140-6736(16)30950-3. PMID: 27474376.
- Williamson JD, Pajewski NM, Auchus AP, Bryan RN, Chelune G, Cheung AK, et al. Effect of Intensive vs Standard Blood Pressure Control on Probable Dementia: A Randomized Clinical Trial. JAMA. 2019;321(6):553-561. doi:10.1001/jama.2018.21442. PMID: 30688979.
- Wing RR, Bolin P, Brancati FL, Bray GA, Clark JM, Coday M, et al. Cardiovascular effects of intensive lifestyle intervention in type 2 diabetes. N Engl J Med. 2013;369(2):145-54. doi:10.1056/NEJMoa1212914. PMID: 23796131.
- Zhao Y, Wu Y, Liu Z, Zhu A. Association of frailty and pre-frailty with cardiovascular mortality: a meta-analysis of 26 cohort studies. Front Public Health. 2025;13:1688014. doi:10.3389/fpubh.2025.1688014. PMID: 41323622.
- Jiang H, Wong JJ, Tan RS, Gao F, Teo LL, Strom JB, et al. Effect of Frailty on Cardiovascular Clinical Trials: A Systematic Review and Meta-Analysis. JACC Adv. 2025;4(7):101889. doi:10.1016/j.jacadv.2025.101889. PMID: 40555009.
- Cahill LE, Grandy SA, Gottheil C, Mercer O, Wright D, Consmueller C, et al. Study on blueberries, protein and exercise for improving frailty and cardiovascular disease (STRONG): protocol for a parallel-group randomised controlled superiority trial in older Nova Scotians. BMJ Open. 2026;16(9):e114286. doi:10.1136/bmjopen-2025-114286. PMID: 42790922.
- Antoun I, Mohiaddin H, Mccann GP, Lawson C, Khan HR, Ahmed R, et al. Evolving burden and consequences of frailty in patients with acute myocardial infarction: evidence from a nationwide cohort. Age Ageing. 2026;55(2). doi:10.1093/ageing/afag033. PMID: 41729759.
- Auyeung TW, Lee JSW, Fung E, Leung J, Kwok T, Woo J. Heart Diseases Can Predict Incident Frailty: A 20-Year Cohort Study in 3,998 Older Adults. Heart Lung Circ. 2026;35(6):815-820. doi:10.1016/j.hlc.2026.04.001. PMID: 42128756.
- Bermejo Boixareu C, Martínez Velilla N, Neira Álvarez M, Abizanda Soler P, Alonso Bouzón C, Casas Herrero A, et al. Frailty prevalence and clinical outcomes according to assessment tools: A systematic review. Rev Esp Geriatr Gerontol. 2026;61(4):101805. doi:10.1016/j.regg.2026.101805. PMID: 42214810.
- Jiang C, Li Z, Ye Y, Li Z, Zhang B, Feng D, et al. Prevalence and Prognostic Significance of Frailty in Elderly Patients With Severe Valvular Heart Diseases. JACC Adv. 2026;5(10 Pt 2):103193. doi:10.1016/j.jacadv.2026.103193. PMID: 42664542.
- Uchmanowicz I, Vellone E, Lee CS, Gobbens RJJ, Hage C, Lambrinou E, et al. Frailty and Heart Failure: An Integrated Review of a Bidirectional Relationship. ESC Heart Fail. 2026. doi:10.1093/eschf/xvag178. PMID: 42364248.
- Gan S, Seo EJ, Rhee SY, Seo WW, Kim WJ, Park HA, et al. Frailty Status and Outcomes in Patients with Acute Coronary Syndrome Undergoing Percutaneous Coronary Intervention: a multicentre retrospective cohort study. Eur J Cardiovasc Nurs. 2026. doi:10.1093/eurjcn/zvag162. PMID: 42411588.
- Fried LP, Tangen CM, Walston J, Newman AB, Hirsch C, Gottdiener J, et al. Frailty in older adults: evidence for a phenotype. J Gerontol A Biol Sci Med Sci. 2001;56(3):M146-56. doi:10.1093/gerona/56.3.m146. PMID: 11253156.
- Rockwood K, Song X, MacKnight C, Bergman H, Hogan DB, McDowell I, et al. A global clinical measure of fitness and frailty in elderly people. CMAJ. 2005;173(5):489-95. doi:10.1503/cmaj.050051. PMID: 16129869.
- Alkadri J, Hladkowicz E, DesAutels K, Taljaard M, Amado L, Bryson GL, et al. Agreement between patient self-reported and assessor-rated frailty using the Clinical Frailty Scale. Br J Anaesth. 2026;137(3):732-739. doi:10.1016/j.bja.2026.05.029. PMID: 42486705.
- Quach J, Theou O, Rockwood K, Kehler S, Blodgett J. Frailty index deficit interchangeability: an empirical test using random survival forests in adults with cardiovascular disease. Age Ageing. 2026;55(9). doi:10.1093/ageing/afag262. PMID: 42696642.
- Drijver-Headley S, Godin J, Rockwood K, Hanlon P. Comparing frailty index distribution, mortality and frailty dynamics across multiple countries: applying the ‘zero-state’ principle to harmonised ageing data. Age Ageing. 2026;55(8). doi:10.1093/ageing/afag242. PMID: 42607277.
- Boz S, Dikmeer A, Tuna Dogrul R, Erdogan K, Gozukan G, Nevin Dagdemir A, et al. Recognizing Frailty Through Pictures: Turkish Validation of the Pictorial Fit-Frail Scale in Older Adults. Medicina (Kaunas). 2026;62(7). doi:10.3390/medicina62071330. PMID: 42512872.
- Kwok BC, Pua YH, Mamun K, Wong WP. The minimal clinically important difference of six-minute walk in Asian older adults. BMC Geriatr. 2013;13:23. doi:10.1186/1471-2318-13-23. PMID: 23510291.
- Mehta P, Costa EC, Galliano LM, Vlietstra L, Saragiotto BT, Cadore EL, et al. Exercise to prevent falls in older adults with physical frailty: A systematic review and meta-analysis. J Frailty Aging. 2026;15(4):100187. doi:10.1016/j.tjfa.2026.100187. PMID: 42611710.
- Rockwood K, Andrew M, Mitnitski A. A comparison of two approaches to measuring frailty in elderly people. J Gerontol A Biol Sci Med Sci. 2007;62(7):738-43. doi:10.1093/gerona/62.7.738. PMID: 17634321.
- Sink KM, Espeland MA, Castro CM, Church T, Cohen R, Dodson JA, et al. Effect of a 24-Month Physical Activity Intervention vs Health Education on Cognitive Outcomes in Sedentary Older Adults: The LIFE Randomized Trial. JAMA. 2015;314(8):781-90. doi:10.1001/jama.2015.9617. PMID: 26305648.
- Luo D, Qin X, Yan J, Yan Y, Liang J, Hu Q, et al. The effectiveness of exercise on physical and cognitive function in older adults with pre-frailty or frailty: A systematic review and meta-analysis. Arch Gerontol Geriatr. 2026;152:106440. doi:10.1016/j.archger.2026.106440. PMID: 42858639.
- Vlietstra L, Fordyce AM, Costa EC, Coffey S, Walker XJ, Whalley GA, et al. Exercise interventions to improve physical frailty and physical frailty components in older adults with hypertension: A systematic review. Ageing Res Rev. 2025;107:102714. doi:10.1016/j.arr.2025.102714. PMID: 40049452.
- Swarnalatha C, Wood L, Newell P, Clark CE, Adedire O, Clemson L, et al. Multifactorial and multiple component interventions for preventing falls in older people living in the community. Cochrane Database Syst Rev. 2026;9(9):CD012221. doi:10.1002/14651858.CD012221.pub3. PMID: 42764178.
- Yang Y, Gao Y, An R, Wan Q. Barriers and facilitators to exercise adherence in community-dwelling older adults: A mixed-methods systematic review using the COM-B model and Theoretical Domains Framework. Int J Nurs Stud. 2024;157:104808. doi:10.1016/j.ijnurstu.2024.104808. PMID: 38823146.
- Marcos-Perez D, Cruces-Salguero S, Saeteros M, Molina RG, Alcantud-Córcoles R, Alonso-Bouzon C, et al. Supervised Exercise Intervention Restores Physical Activity and MultiOMIC-Derived Frailty Biomarkers in Older Adults. J Cachexia Sarcopenia Muscle. 2026;17(4):e70362. doi:10.1002/jcsm.70362. PMID: 42571782.
- McColl TJ, Moore DR, Emberly E, Church DD, Clarke DC. Multifactorial nature of anabolic resistance in ageing skeletal muscle: A systems modelling study. J Physiol. 2026;604(15):6405-6439. doi:10.1113/JP290799. PMID: 42464764.
- Valenzuela RE, Ponce JA, Morales-Figueroa GG, Muro KA, Carreón VR, Alemán-Mateo H. Insufficient amounts and inadequate distribution of dietary protein intake in apparently healthy older adults in a developing country: implications for dietary strategies to prevent sarcopenia. Clin Interv Aging. 2013;8:1143-8. doi:10.2147/CIA.S49810. PMID: 24039411.
- Bauer J, Biolo G, Cederholm T, Cesari M, Cruz-Jentoft AJ, Morley JE, et al. Evidence-based recommendations for optimal dietary protein intake in older people: a position paper from the PROT-AGE Study Group. J Am Med Dir Assoc. 2013;14(8):542-59. doi:10.1016/j.jamda.2013.05.021. PMID: 23867520.
- Deutz NE, Bauer JM, Barazzoni R, Biolo G, Boirie Y, Bosy-Westphal A, et al. Protein intake and exercise for optimal muscle function with aging: recommendations from the ESPEN Expert Group. Clin Nutr. 2014;33(6):929-36. doi:10.1016/j.clnu.2014.04.007. PMID: 24814383.
- Whaikid P, Piaseu N. The effectiveness of protein supplementation combined with resistance exercise programs among community-dwelling older adults with sarcopenia: a systematic review and meta-analysis. Epidemiol Health. 2024;46:e2024030. doi:10.4178/epih.e2024030. PMID: 38374703.
- Mendonça N, Rodrigues AM, Henriques AR, Canhão H, Raimundo A, Celis-Morales C. Protein intake and transitions between frailty states in middle-aged and older adults: a multi-state transition model in the UK Biobank. BMC Med. 2025;23(1):621. doi:10.1186/s12916-025-04397-0. PMID: 41204398.
- Bålsrud P, Ulven SM, Hjartåker A, Ottestad I, Retterstøl K, Schwab U, et al. Association between dietary protein intake and frailty index among home-dwelling older adults; an 8-year follow-up study. J Nutr Health Aging. 2026;30(6):100850. doi:10.1016/j.jnha.2026.100850. PMID: 41999683.
- Churchward-Venne TA, Pinckaers PJ, Smeets JS, Marzuca-Nassr GN, Gorissen SH, Fuchs CJ, et al. Leucine supplementation does not attenuate the decline in daily muscle protein synthesis rates or preserve leg muscle mass during leg immobilization in young or older adults: a double-blind randomized trial. Am J Clin Nutr. 2026;123(4):101205. doi:10.1016/j.ajcnut.2026.101205. PMID: 41580240.
- Houtvast DC, Hendriks FK, Weijzen ME, Zorenc AH, Smeets JS, Emans PJ, et al. Assessing the responsiveness of musculoskeletal tissues to protein supplementation in vivo in older adults: an exploratory randomized controlled trial. Am J Clin Nutr. 2026;124(1):101349. doi:10.1016/j.ajcnut.2026.101349. PMID: 42142739.
- Teixeira Silva PR, Santos P, de Miranda Silva Nogueira PA, Nogueira IDB, Vieira RP, Nascimento MF, et al. Effects of High-Protein Diets on Renal Function and Body Composition in Adults Without Chronic Kidney Disease: A Systematic Review and Meta-Analysis of Randomised Trials. Diabetes Obes Metab. 2026;28(9):7672-7683. doi:10.1111/dom.70996. PMID: 42289790.
- Ramzan R, Muhammad Z, Amjad A, Sharif HR, Zhang G, Chen A. Enhanced Bioavailability and Health Benefits of Blueberry Anthocyanins: An Updated Review on Mechanisms and Approaches. Molecules. 2026;31(5). doi:10.3390/molecules31050793. PMID: 41828782.
- Wood E, Hein S, Mesnage R, Fernandes F, Abhayaratne N, Xu Y, et al. Wild blueberry (poly)phenols can improve vascular function and cognitive performance in healthy older individuals: a double-blind randomized controlled trial. Am J Clin Nutr. 2023;117(6):1306-1319. doi:10.1016/j.ajcnut.2023.03.017. PMID: 36972800.
- Woolf EK, Terwoord JD, Litwin NS, Vazquez AR, Lee SY, Ghanem N, et al. Daily blueberry consumption for 12 weeks improves endothelial function in postmenopausal women with above-normal blood pressure through reductions in oxidative stress: a randomized controlled trial. Food Funct. 2023;14(6):2621-2641. doi:10.1039/d3fo00157a. PMID: 36847333.
- Rendine M, Venturi S, Marino M, Martini D, Riso P, Battezzati A, et al. Short-Term Effect of a Wild Blueberry (Vaccinium angustifolium) Drink on Vascular Reactivity and Circulating Adhesion Molecules in Healthy Adults: A Double-Blind, Randomized, Placebo-Controlled, Crossover Trial. Nutrients. 2026;18(14). doi:10.3390/nu18142376. PMID: 42514445.
- Curtis PJ, van der Velpen V, Berends L, Jennings A, Haag L, Minihane AM, et al. Chronic and postprandial effect of blueberries on cognitive function, alertness, and mood in participants with metabolic syndrome – results from a six-month, double-blind, randomized controlled trial. Am J Clin Nutr. 2024;119(3):658-668. doi:10.1016/j.ajcnut.2023.12.006. PMID: 38432713.
- Beavers DP, Kritchevsky SB, Gill TM, Ambrosius WT, Anton SD, Fielding RA, et al. Elevated IL-6 and CRP Levels Are Associated With Incident Self-Reported Major Mobility Disability: A Pooled Analysis of Older Adults With Slow Gait Speed. J Gerontol A Biol Sci Med Sci. 2021;76(12):2293-2299. doi:10.1093/gerona/glab093. PMID: 33822946.
- Custodero C, Pahor M, Mazzoccoli C, Manini TM, Anton SD, Mazzocca A, et al. Effect of change of interleukin-6 over time on gait speed response: Results from the lifestyle interventions and independence for elders study. Mech Ageing Dev. 2023;210:111763. doi:10.1016/j.mad.2022.111763. PMID: 36526102.
- Fielding RA, Atkinson EJ, Aversa Z, White TA, Heeren AA, Mielke MM, et al. Biomarkers of Cellular Senescence Predict the Onset of Mobility Disability and Are Reduced by Physical Activity in Older Adults. J Gerontol A Biol Sci Med Sci. 2024;79(3). doi:10.1093/gerona/glad257. PMID: 37948612.
- Pöyhönen J, Lehtisalo J, Roitto HM, Levälahti E, Strandberg T, Kivipelto M, et al. Impact of frailty status on the effect of a multidomain lifestyle intervention on cognition. Age Ageing. 2025;54(2). doi:10.1093/ageing/afaf041. PMID: 40036319.
- Bischoff-Ferrari HA, Freystätter G, Vellas B, Dawson-Hughes B, Kressig RW, Kanis JA, et al. Effects of vitamin D, omega-3 fatty acids, and a simple home strength exercise program on fall prevention: the DO-HEALTH randomized clinical trial. Am J Clin Nutr. 2022;115(5):1311-1321. doi:10.1093/ajcn/nqac022. PMID: 35136915.
- Dal Passo B, Tonet E, Raisi A, Zagnoni S, Chiaranda G, Pasanisi G, et al. Effectiveness of Multidomain Cardiac Rehabilitation After Myocardial Infarction by Patient Frailty: Prespecified Subgroup Analysis of the PIpELINe Trial. Circ Popul Health Outcomes. 2026;19(9):e013410. doi:10.1161/CIRCOUTCOMES.126.013410. PMID: 42605666.
- Loudon K, Treweek S, Sullivan F, Donnan P, Thorpe KE, Zwarenstein M. The PRECIS-2 tool: designing trials that are fit for purpose. BMJ. 2015;350:h2147. doi:10.1136/bmj.h2147. PMID: 25956159.
- Naylor MD, Hirschman KB. Pragmatic Trials: The Importance of Foundational Research, Fidelity, Implementation, and Context. Ann Intern Med. 2026;179(7):1051-1052. doi:10.7326/ANNALS-26-01922. PMID: 42330501.
- Jiu Chen, Cui W, Xie J. Effect of exercise based on ACSM recommendations on blood pressure and heart rate in hypertensive patients: A systematic review and meta-analysis of randomized controlled trials. PLOS Glob Public Health. 2024;4(12):e0003743. doi:10.1371/journal.pgph.0003743. PMID: 39680548.
- Pearson GJ, Thanassoulis G, Anderson TJ, Barry AR, Couture P, Dayan N, et al. 2021 Canadian Cardiovascular Society Guidelines for the Management of Dyslipidemia for the Prevention of Cardiovascular Disease in Adults. Can J Cardiol. 2021;37(8):1129-1150. doi:10.1016/j.cjca.2021.03.016. PMID: 33781847.
- Talasaz AH, McGonagle B, HajiQasemi M, Ghelichkhan ZA, Sadeghipour P, Rashedi S, et al. Pharmacokinetic and Pharmacodynamic Interactions between Food or Herbal Products and Oral Anticoagulants: Evidence Review, Practical Recommendations, and Knowledge Gaps. Semin Thromb Hemost. 2025;51(5):560-571. doi:10.1055/s-0044-1790258. PMID: 39288907.
- Ma C, Monagas M, Bronstein L, Cadwallader A, Goldman V. Dietary Supplement Adulteration: Laboratory Approaches to Risk Mitigation. J Nat Prod. 2025;88(8):2029-2040. doi:10.1021/acs.jnatprod.5c00456. PMID: 40700739.
- O’Brien MW, Shivgulam M, Liu H, Courish M, Wu Y, Fowles J, et al. The effectiveness of exercise referral schemes on patient health and their cost: an umbrella review. Appl Physiol Nutr Metab. 2025;50:1-12. doi:10.1139/apnm-2024-0185. PMID: 39693607.
- Kane RM, Perez-Velazco X, Levi R, Goldberg A, Kuhn M, Mehri N, et al. Produce Prescription Dose Response: A Randomized Community Trial. J Gen Intern Med. 2026. doi:10.1007/s11606-026-10773-0. PMID: 42745148.
- Pulok MH, van der Valk AM, Mondor L, Theou O, Rockwood K. Frailty severity, survival and physician service costs among older adults: a linked cohort study from Nova Scotia, Canada. Age Ageing. 2026;55(10). doi:10.1093/ageing/afag286. PMID: 42910912.
- Bull FC, Al-Ansari SS, Biddle S, Borodulin K, Buman MP, Cardon G, et al. World Health Organization 2020 guidelines on physical activity and sedentary behaviour. Br J Sports Med. 2020;54(24):1451-1462. doi:10.1136/bjsports-2020-102955. PMID: 33239350.
- Ross R, Chaput JP, Giangregorio LM, Janssen I, Saunders TJ, Kho ME, et al. Canadian 24-Hour Movement Guidelines for Adults aged 18-64 years and Adults aged 65 years or older: an integration of physical activity, sedentary behaviour, and sleep. Appl Physiol Nutr Metab. 2020;45(10 (Suppl. 2)):S57-S102. doi:10.1139/apnm-2020-0467. PMID: 33054332.
- Izquierdo M, Merchant RA, Morley JE, Anker SD, Aprahamian I, Arai H, et al. International Exercise Recommendations in Older Adults (ICFSR): Expert Consensus Guidelines. J Nutr Health Aging. 2021;25(7):824-853. doi:10.1007/s12603-021-1665-8. PMID: 34409961.
- Martin-Lesende I, Gorroñogoitia-Iturbe A, Goñi-Ruiz N, Abizanda-Soler P. Frailty in older adults: detection and management in primary care. Fisterra clinical practice guideline. Eur Geriatr Med. 2026. doi:10.1007/s41999-026-01541-8. PMID: 42560630.
- Haner Wasserstein D, Whitford T, Whiteson HZ, Frishman WH. Lean Mass Loss in Glucagon-Like Peptide-1/GIP Therapy: Clinical Implications for Obesity and Cardiovascular Care. Cardiol Rev. 2026. doi:10.1097/CRD.0000000000001178. PMID: 41636548.
- Moscucci F, Tocci G, Nardoianni G, Savoia C, Contino V, Leo G, et al. Lean Mass Reduction During Glucagon-like Peptide-1 Receptor Agonists-Induced Weight Loss: Rethinking Its Clinical Significance. High Blood Press Cardiovasc Prev. 2026. doi:10.1007/s40292-026-00819-z. PMID: 42701941.
- Ostrominski JW, Plutzky J, Scirica BM, Hovingh GK, Jeppesen OK, Lincoff AM, et al. Efficacy and Safety of Semaglutide According to Frailty Status: A Post Hoc Analysis of the SELECT Randomized Clinical Trial. JAMA Cardiol. 2026. doi:10.1001/jamacardio.2026.3566. PMID: 42747817.
- ClinicalTrials.gov (U.S. National Library of Medicine). Keyword query counts for “blueberry”, “blueberry frailty”, “exercise frailty older adults” and related terms, queried 2 November 2026. https://clinicaltrials.gov/search?term=blueberry%20frailty