Impact of age and frailty on the effect of a multimodal individualized long-term intervention after stroke: the LAST-long trial

Clinical Question (PICO)

In adults with residual impairments approximately three months after stroke, does adding an individualized, community-based stroke coordinator with monthly meetings for 18 months to standard care improve functional outcome? Does response differ between patients aged ≥80 versus <80 years, or between those with Frailty Index ≥0.2 versus <0.2?

This is an exploratory secondary analysis, published October 26, 2026, of the LAST-long randomized trial—not a new trial independently randomized within these subgroups. The paper describes age and frailty as exploratory variables; its findings should not be presented as confirmatory subgroup evidence.

Bottom Line

Age, but not frailty, modified the association between the assigned intervention and 18-month functional outcome. Among participants aged ≥80 years, model-adjusted probabilities of an excellent outcome, modified Rankin Scale (mRS) 0–1, were 23.4% with intervention versus 57.0% with control; the reported within-older-group OR was 0.18 (95% CI 0.05–0.57), P=0.004.

The parent trial was neutral overall. This sub-study raises a hypothesis about age-specific response; it does not establish that stroke coordinators generally harm older adults, justify withdrawing rehabilitation, or demonstrate that frailty identifies a treatment-responsive subgroup.

Design

Trial type: Exploratory sub-study of a multicentre, assessor-blinded, parallel-group randomized controlled trial. Patients and coordinators could not be blinded to monthly meetings.
N randomized in the parent trial: 301.
- Coordinator plus standard care: 152.
- Standard care alone: 149.
N in this analysis: 268 participants with an available 18-month mRS, including deaths scored as mRS 6.
- Intervention: 136.
- Control: 132.
Randomization: 1:1, web-based, variable block sizes. The parent report describes stratification by age (>80 years), dependency (mRS <3), and hospital. Note that this sub-study defines older age as ≥80 years.
Setting: Four Norwegian hospitals—two university and two regional hospitals—with primary-care services in six municipalities.
Enrollment: The sub-study describes recruitment during 2019–2022; follow-up ended August 28, 2024.
Follow-up: Scheduled 18 months after enrollment, not 18 months after the acute stroke. This is the planned observation interval rather than a reported mean follow-up.
Analysis: Available-case analysis derived from the randomized intention-to-treat population; 33 randomized participants lacked the primary endpoint. Logistic models included treatment, age or frailty, and their interaction, adjusted for sex, admission NIHSS, hospital, and pre-stroke mRS. Secondary continuous outcomes used linear regression; an ordinal mRS sensitivity analysis was also reported.
Primary outcome: Parent-trial mRS at 18 months; this sub-study principally examines excellent outcome, mRS 0–1, and effect modification. Secondary-outcome interaction tests underwent Benjamini–Hochberg false-discovery-rate correction.

Population

Inclusion Criteria

Age ≥18 years; outpatient enrollment approximately 2–4 months after stroke, after initial rehabilitation.
mRS <5 and life expectancy exceeding one year.
Evidence of impairment on at least one screening measure: SPPB <10, MoCA <26, FSS-7 >27, HADS anxiety or depression subscale >7, or impaired advanced hand activity on the specified Motor Assessment Scale item.

Exclusion Criteria

Severe comorbidity judged likely to interfere with participation, including severe neurological disease, dementia, or substance abuse.
Failure to meet the functional, survival, or screening criteria above.

Baseline Characteristics (268-person sub-study)

Mean age 71.1 years (SD 12.4; range 33–92).
Women: 123/268, or 45.9% (calculated from Table IV).
Age ≥80: 66 participants, including 35 intervention and 31 control participants.
Frail: 89 participants, including 44 intervention and 45 control participants; 35 older participants were also frail. These overlapping categories are not interchangeable.
Most strokes were mild: admission NIHSS <8 in 89.4% of the older group and 85.2% of the younger group.
Both ischemic and hemorrhagic stroke were represented; ischemic stroke accounted for 92.4% of the older group and 91.1% of the younger group.
Frailty was measured after stroke, at study entry. It can therefore incorporate residual stroke-related disability rather than isolate pre-stroke biological vulnerability.

Interventions

Coordinator plus standard care: Monthly meetings for 18 months. A checklist assessed health/lifestyle, physical function, cognition/mood, and social function. The participant and coordinator developed personal goals and an individualized plan, revised at subsequent meetings, using existing health and community services.
Standard care: Stroke-unit care and rehabilitation according to need, followed by municipal services and general-practitioner care; specialist stroke follow-up was typically offered around three months. The comparator lacked the study's additional fixed coordinator meetings, not all stroke care.
Important delivery limitation: Complete systematic records of the specific treatments actually received and adherence to individualized goals were not collected. A coordinator meeting is not a standardized dose of rehabilitation.

Outcomes

Primary Outcome — excellent functional outcome, mRS 0–1 at 18 months:

All subgroup comparisons below are exploratory. Percentages are model-adjusted estimated probabilities, not raw event rates.

Subgroup	Intervention	Control	Inferential result reported in the paper
Age ≥80 years	23.4%	57.0%	Within-group OR 0.18, 95% CI 0.05–0.57; P=0.004
Age <80 years	53.8%	53.1%	No treatment effect detected; P=0.69; a within-group OR/CI is not supplied in the main-text results
Frail, FI ≥0.2	36.2%	31.4%	No significant frailty–treatment interaction; P=0.145
Non-frail, FI <0.2	51.0%	64.5%	Same interaction test, P=0.145; not evidence that these two subgroup contrasts differ reliably

The age–treatment interaction OR was 0.17 (95% CI 0.05–0.63), P=0.008. This interaction estimate is distinct from the treatment OR within the older subgroup. The older-group probability difference is −33.6 percentage points, calculated from the adjusted estimates; it is not an unadjusted risk difference or a reliable basis for a number needed to harm. Frailty-specific treatment ORs, confidence intervals, and within-group P values are not supplied in the main-text results and are not reconstructed here.

The ordinal sensitivity analysis also reported an age–treatment interaction of 0.17, P=0.008, using the paper's “RRR” terminology. It does not create an independent replication.

For context, the parent LAST-long trial reported an overall between-group mRS difference of 0.03 points (95% CI −0.16 to 0.22), P=0.79 at 18 months. That continuous-scale estimate is not the same estimand as the sub-study's dichotomized subgroup outcome.

Secondary Outcomes:

Executive function, TMT-B: In those aged <80, intervention minus control was −32.2 seconds (95% CI −58.0 to −6.4), P=0.015. Among older participants, the difference was +43.9 seconds (95% CI −14.1 to 101.8), P=0.137; lower completion time is better.
The TMT-B age interaction was nominally significant, P=0.019, but not after multiplicity correction: FDR-adjusted P=0.247. The abstract's cognitive signal should not be read without this qualification.
No significant age interactions were demonstrated for mortality, MoCA, SPPB, fatigue, EQ-5D-5L, or extended activities of daily living.
No significant frailty interactions were demonstrated across the reported outcomes. Frailty predicted poorer prognosis, which is different from predicting treatment response.

Adverse Events / Safety:

Eight deaths occurred in the analyzed cohort: five among older participants and three among younger participants. Mortality interaction analyses did not establish an age-specific treatment effect.
The parent trial reported serious adverse events in 32/152 (21.1%) intervention participants and 33/149 (22.1%) controls, encompassing death and hospitalization for cardiovascular or cerebrovascular events or falls with or without fracture.
The poorer older-group mRS result is a functional-outcome signal, not proof of increased mortality or a demonstrated mechanism of treatment-related injury. Symptomatic intracranial hemorrhage was not a dedicated outcome of this follow-up intervention analysis.

Figures

Overview of the LAST-long individualized stroke-coordinator intervention — *Figure 1.* Overview of the study intervention. This figure describes the care model; it is not a plot of the subgroup treatment effect.

Criticisms

Subgroup rescue is not confirmation: Only 66 participants were aged ≥80. Randomization of the parent trial does not turn an exploratory interaction into a separately powered trial of older adults.
Multiplicity changes the cognitive story: TMT-B's interaction P value becomes 0.247 after false-discovery-rate correction. A positive within-younger-group P value does not establish a robust subgroup benefit.
Missingness matters: Primary-endpoint availability was 268/301 (89.0%); secondary outcomes had additional missing data, disproportionately among older and frailer participants. This is not the same analysis as the parent trial's longitudinal mixed-model intention-to-treat estimate.
The endpoint may be poorly aligned with need: An mRS 0–1 threshold can miss worthwhile maintenance of function, participation, or caregiver benefit in an older person with pre-existing limitations.
Frailty is partly entangled with stroke disability: Assessment occurred approximately three months after stroke and includes cognition, mobility, and daily activities.
Mechanism remains unknown: Treatment burden, fatigue, insufficient targeting, and random variation are competing explanations. The study did not measure enough of the delivered care to establish which explanation is correct.
Transportability is limited: This mostly mild-stroke cohort received care in a well-developed Norwegian healthcare system. The incremental effect of coordination could differ where baseline access is poor.

Funding

Research Council of Norway, grant 273487; the Joint Research Committee between St. Olavs Hospital and the Faculty of Medicine and Health Sciences at NTNU (FFU); and Foundation Dam. The authors state that funders had no role in this sub-study and did not influence its results or outcomes. No conflicts of interest were declared. Registration: NCT03859063.

The paper

Authors. Håkon Ihle-Hansen et al.
Title. Impact of age and frailty on the effect of a multimodal individualized long-term intervention after stroke: the LAST-long trial.
Journal. Journal of Rehabilitation Medicine, volume 58, article 46250.
Year. 2026; published October 26.
DOI. 10.2340/jrm.v58.46250
PMCID. PMC13717427

Deep Dive — click to expand

What this is

LAST-long tests an attractive service-design idea: identify unmet needs, agree on personal goals, and provide a regular coordinator rather than leaving recovery to fragmented follow-up. This October 2026 sub-study asks whether an overall-neutral trial hid a benefit in older or frailer patients. It did not find that rescue. Instead, it found an age interaction pointing toward worse excellent-outcome probabilities in older intervention participants, alongside a cognitive signal that did not survive multiplicity correction. The following frameworks are editorial interpretation of those results, not additional trial findings or a statement of current guidelines.

1. Shadow Audit

The headline is age; the quieter result is that frailty is prognostic without being demonstrably predictive. Frail participants had poorer outcomes, yet the frailty–treatment interaction was P=0.145. A high-risk patient does not automatically gain more from this particular additional service.

A second shadow sits in the abstract's executive-function result. The younger-group TMT-B difference is −32.2 seconds, but the interaction's adjusted P value is 0.247. The sensible reading is an exploratory lead, not a dependable reason to preferentially fund the programme for younger survivors. The parent trial's overall mRS result—0.03 points, P=0.79—remains the anchor.

2. Inversion Engine

For the opposite conclusion—this programme reliably improves excellent outcomes among older adults—to hold, a replication would have to reverse the direction of the reported older-group treatment OR from 0.18 to above 1, with uncertainty supporting benefit rather than harm. On the adjusted-probability scale, merely erasing this study's contrast would require closing a 33.6-percentage-point gap. These are descriptions of the reported signal, not sample-size targets or forecasts of a replicating trial.

The easier inversion is interpretive: the apparent disadvantage could diminish under a more suitable endpoint, better missing-data handling, or chance-balanced subgroups. That possibility is why a large-looking subgroup contrast is a reason to investigate, not a license to stop care.

3. Second-Order Catalyst

The first people who should change are the teams designing or evaluating coordinator programmes, not emergency reperfusion teams. A rehabilitation service considering a fixed 18-month package should ask whether visits actually connect patients to useful treatment and whether the workload fits the person's capacity.

An immediate, low-risk response is to add explicit checks for fatigue, practical burden, goal attainment, and access barriers when reviewing an older survivor's plan. That is an editorial implementation suggestion, not an intervention proven by LAST-long. Reducing contact solely because someone turns 80 would go beyond the evidence and could remove valuable support.

4. Asymmetric Leverage

The leverage is in distinguishing contact delivered from care accomplished. Monthly meetings can consume staff and patient effort even when an actionable therapy, social intervention, or treatment adjustment never follows. This study cannot quantify that mechanism because detailed treatment delivery and goal adherence were not systematically captured.

A small addition to programme evaluation—a record of the identified problem, completed action, patient burden, and meaningful outcome—could expose whether coordination is generating benefit or simply activity. The payoff would be better targeting of an existing resource, not a claim that this trial proved a cheaper or more effective replacement model.

5. Paradigm Destroyer

The reflex this paper challenges is: “More follow-up must help, especially in the most vulnerable.” It does not challenge the value of secondary prevention, needs-based rehabilitation, or humane continuity of care; it challenges assuming that an intuitively attractive delivery package has established incremental benefit.

A two-sentence protocol update for service evaluation would be: “When offering long-term coordinator follow-up, document the patient's goals, completed care actions, and treatment burden rather than using visit count as the success measure. Review older and frailer patients individually, without using this exploratory subgroup analysis as an age-based exclusion rule.”

MVP — Minimum Viable Proof

The next practice-changing proof would be an adequately powered, prospectively specified comparison demonstrating a patient-important benefit in older survivors, with credible missing-data handling, measured delivery fidelity, and no unacceptable burden. LAST-long already supplies the minimum reason to pause automatic expansion of this exact package: no overall mRS benefit and no convincing frailty-defined rescue. It does not supply the minimum proof needed to withdraw existing support or to label the programme causally harmful.

Best Combination

Read this sub-study beside the primary LAST-long report. The primary trial answers whether the package improved average functional outcome; this analysis asks whether age or frailty changes that answer. Combining those two reports supports a narrower conclusion than either “coordination works” or “coordination is harmful”: this specific add-on programme did not establish an overall advantage in a mostly mild-stroke Norwegian cohort, and its older-group signal merits prospective scrutiny.

The LAST-long protocol provides the original trial framework. Neither a general policy preference for continuity of care nor observational evidence that frailty predicts poor prognosis can substitute for demonstrating that this particular service improves outcomes. Different models, settings, and populations remain separate questions.

Overvalue Warning

Do not convert an exploratory subgroup into an age cutoff for care. The 23.4% versus 57.0% figures are adjusted estimates from a small subgroup, not a universally transportable harm rate. The paper does not establish that withholding coordination improves recovery.
Do not confuse statistical labels with positive clinical evidence. “No frailty interaction” is not proof of equal effects, and a nominal cognitive P value is not robust after the reported correction. Age, frailty, disability, and unmet need should inform assessment—not be treated as interchangeable selectors for a proven treatment effect.

Educational use notice. The Time Is Brain Review is dedicated to medical education. Figures from published articles are embedded for teaching purposes, with attribution to the original publication. We do not claim copyright over any embedded figure; all rights remain with the original publisher and authors under the article's published license. If you are a rights holder and would like a figure removed, please contact the editorial team. See our editorial notice for the full policy.