Exercise Effects on Brain Health and Learning from Minutes to Months: the Brain EXTEND trial

Age-related cognitive decline is a growing public health challenge. Animal models robustly show aerobic exercise protects hippocampal structure and function and improves learning and memory. Human trials have yielded mixed cognitive results despite documented exercise-related benefits to brain regions vulnerable to aging (e.g., hippocampus, prefrontal cortex). Small average cognitive benefits are detectable via meta-analyses, but variability across individuals remains high. The Brain-EXTEND trial addresses this gap by testing whether moderate-to-vigorous aerobic exercise improves hippocampal-dependent learning in older adults and clarifying mechanisms. The study had three aims: (1) determine whether chronic exercise-induced changes in hippocampal–default network functional connectivity (FC) relate to improvements in hippocampal-dependent learning (spatial, associative, context) more than in non-hippocampal learning; (2) test whether acute, rapid effects of a single bout of moderate-to-vigorous exercise on hippocampal-cortical FC predict longitudinal changes in FC and learning after six months; and (3) investigate physiological mediators and moderators, particularly cardiorespiratory fitness (CRF) changes, alongside vascular, behavioral (sedentary, sleep), and genomic/molecular determinants of CRF and trainability.

Preclinical evidence: Aerobic exercise enhances hippocampal neurogenesis, synaptic plasticity, and learning in rodents, including aged mice. Acute exercise rapidly modulates hippocampal plasticity markers (e.g., immediate-early genes) and dentate gyrus function. Human evidence: Exercise training can increase hippocampal volume and enhance brain network integrity in older adults. However, large RCTs show mixed cognitive outcomes, with meta-analyses indicating modest average benefits and substantial inter-individual variability. Acute exercise in humans alters large-scale brain network FC measurable by fMRI and may predict training-induced changes in cognition and connectivity. CRF correlates with hippocampal-cortical FC and cognition cross-sectionally, suggesting physiological adaptations may mediate cognitive benefits. The literature motivates testing acute brain responses as predictors of long-term change and examining CRF and vascular/genetic factors as mediators/moderators.

Design: Randomized, single-blind, two-group clinical trial (NCT03114150) with an acute cross-over exercise assessment preceding randomization to a 24-week chronic training intervention. Inactive older adults (55–80 years), including non-demented participants and some with MCI per stratified MoCA cutoffs, were enrolled and completed pre- and post-intervention assessments of CRF, MRI, and cognition. Enrollment: May 2018 to Nov 28, 2022; final post-tests by June 2023. COVID-19 (Mar–Jul 2020) prompted temporary suspension of in-person visits and some remote modifications (e.g., remote orientation and MoCA). Randomization: Participants first completed two acute exercise sessions (within-subjects: light vs moderate-to-vigorous [ModVig], counterbalanced) with MRI pre/post each session. Then, participants were re-randomized to chronic training: ModVig vs light (active control). Randomization was blocked by age (55–67, 68–80) and sex, with independent sequences for acute order and chronic group. Timeline: ~4 weeks pre-testing (including acute sessions), 24 weeks training (3 sessions/week), ~2 weeks post-testing (total ~28–30 weeks). Participants and recruitment: Recruited within ~1-hour drive of University of Iowa via mailings, mass email, word of mouth, community outreach, and advertisements. Inclusion/exclusion criteria detailed in Table 1 (e.g., age 55–80, inactive <60 min/week moderate intensity over past 6 months, MoCA ≥20, MRI eligible, not high cardiovascular risk per ACSM, no major neurological or psychiatric diseases, etc.). Screening involved online/phone pre-screen and in-person/remote orientation plus maximal exercise test. Sample characteristics: Baseline n=122 (completed at least one MRI, maximal exercise test, neuropsychological assessment, T1); randomized n=116; completed training n=107 (ModVig n=55, Light n=52). MCI prevalence at baseline: 9.8% (n=12) balanced across groups. Cardiorespiratory fitness and vascular health: Symptom-limited maximal exercise test to exhaustion on upright cycle ergometer with 12-lead ECG and breath-by-breath gas exchange to determine VO2max (True One, ParvoMedics). Criteria for VO2max: meet ≥2 of 3 (RER ≥1.10; Borg RPE ≥17; HR within 10 bpm of age-predicted max or ≥80% predicted). 86% of completed tests met VO2max criteria (88% among randomized). Resting supine BP pre-test, BP measured every 2-min stage and at peak. Additional vascular assessments included central arterial stiffness and central hemodynamics; blood draws for metabolic and other biomarkers (details in supplemental materials). Physical activity and sleep: Wrist-worn ActiGraph GT3X Link for 7–10 days at baseline, with monthly repeats during training to post-testing. Participants maintained logs for wear times and sleep. Brain imaging: MRI on 3T GE Discovery 750W (32-channel coil) and Signa Premier 750W (48-channel coil). Participants: Discovery n=52 (42.6%), Premier n=70 (57.4%); no group differences in demographics or CRF; scanner model balanced across groups; pre/post scanner consistency (DD, PP) or switch (DP) balanced (χ²(2)=1.39, p=.50). Sequences: structural (T1, T2, diffusion), resting-state BOLD (aging-sensitive networks), T1ρ (neuronal metabolism), breath-hold BOLD for cerebrovascular reactivity. fMRI acquisition occurred 1 hour before and 30 minutes after each acute session, and 1 hour at post-intervention. Cognition: Two ~2-hour sessions pre- and post-training. Primary hippocampal-dependent learning tasks: (1) Spatial navigation (allocentric/hippocampal-dependent cognitive map vs egocentric/route learning) in a virtual maze (WorldViz Vizard, Autodesk 3ds Max) adapted from Head & Isom and Allison et al.; (2) Associative learning with integrative paired associates vs configural response learning (face-pair tasks; Center for Vital Longevity Face Database); (3) Context acquisition via cued task-switching with Separate vs Overlap response-set conditions, varying hippocampal demands. Standardized measures included Rey Auditory Verbal Learning Test (immediate/delayed), vocabulary (WRAT reading), and processing speed (letter comparison, pattern comparison, DSST). Alternate forms used where applicable. Acute exercise protocol: Two supervised 30-min cycling sessions in counterbalanced order: light intensity (<40% HRR) and ModVig (peak 70–80% HRR). HRR = [(Peak HR – Rest HR) × %] + Rest HR. Resting-state fMRI acquired pre and post each session. HR continuously monitored; self-reported affect, arousal, and RPE recorded pre, every 5 minutes during, and post (self-report results in supplemental). Sessions separated by ≥48 hours. Chronic training intervention: 24-week supervised program, 3 sessions/week. ModVig: progressive protocol with warm-up/cool-down and staged increases: weeks 1–4 build to 40 min/session; weeks 6–12 at 60–70% HRR; weeks 13–24 at 70–80% HRR (with progressive build-up). Light: functional flexibility/mobility with stretching, 20 min light aerobic (<40% HRR), aiming to maintain HR ≤40% HRR. Both groups wore Polar M200 HR monitors; sessions pre-programmed with target HR; data synced to Polar Flow; adherence feedback provided. RPE used alongside HR to calibrate targets if discrepancies at baseline test. Supervision: Initially first two weeks in-lab; thereafter up to two home sessions/week. During COVID-19 restrictions, all sessions conducted at home; upon resumption, option to start in lab with precautions, then transition to more home sessions. Outcomes and fidelity checks: Acute manipulation checks for HR and RPE across timepoints; chronic fidelity via HR across training phases and CRF change. Adherence measured as number of completed sessions (out of 72). Statistical analyses: Repeated-measures ANOVAs for acute RPE and %HRR with factors Condition (ModVig vs Light) × Time Point (Warm-up, TP1–TP4, Cool-down), plus order (day1/day2) as between-subjects; Huynh-Feldt correction for sphericity violations. For training HR, rmANOVA across phases with Group × Training Phase. CRF intervention effect: Linear regression predicting post-intervention VO2max including age, sex, baseline VO2max, and Group (1=ModVig, 0=Light). Missing post-intervention VO2max values (n=24 total; 12 per group) addressed via multiple imputation (MAR; details in supplemental). Significance threshold p<0.05. CONSORT: Screened n=856; consented n=217; randomized to acute order n=129; completed acute day 1 n=122, day 2 n=119; randomized to training groups n=116; discontinued n=9 (5 ModVig, 4 Light); completed training n=107 (55 ModVig, 52 Light).

• Sample and baseline fitness: Randomized participants (n=116) were older adults (mean age ~63.4 years; 65.5% female). Baseline VO2max averaged 20.6 ml/kg/min (7th percentile for age/sex), with no baseline VO2max difference between training groups (t(114) = -1.06, p = .29). 86% of completed maximal tests met VO2max criteria (88% among randomized).
• Acute exercise manipulation check: Robust separation of intensity conditions.
  • RPE: Condition effect F(1,111)=393.168, p<.001, η²=.310; Time Point (TP) effect F(3.380,375.131)=256.295, p<.001, η²=.205; Condition×TP interaction F(3.884,431.174)=163.113, p<.001, η²=.088. RPE increased during ModVig, peaking at TP4 at 14.5 (“hard”), and remained 8–9 (“light”) during light intensity.
  • %HRR: Condition effect F(1,107)=375.108, p<.001, η²=.387; TP effect F(3.086,330.180)=226.165, p<.001, η²=.148; Condition×TP interaction F(3.220,344.557)=135.115, p<.001, η²=.069. ModVig showed progressive %HRR increases from TP1 to TP4; light remained low across TPs.
• Chronic training fidelity (HR across phases): rmANOVA showed Training Phase effect F(1.717,181.954)=125.898, p<.001, η²=.543, and Group×Training Phase interaction F(1.717,181.954)=61.818, p<.001, η²=.368. ModVig maintained higher %HRR than Light across phases, consistent with targets (50–60%, 60–70%, 70–80% HRR).
• Adherence: Average adherence 94% in both groups. Median adherence: ModVig 99% (IQR 7%); Light 97% (IQR 7%). During COVID-related adjustments, total sessions completed averaged 69 (ModVig; 85% at home) and 71 (Light; 83% at home).
• CRF (VO2max) intervention effect: Linear regression (controlling for age, sex, baseline VO2max) showed a significant Group effect favoring ModVig over Light at post-intervention: b=1.42 ml/kg/min, t(63.16)=2.46, 95% CI [0.27, 2.59]; results consistent in non-imputed data.
• Imaging platform balance: No differences in demographics, education, sex, or CRF between scanner models (Discovery vs Premier); pre/post scanner model balance across groups (χ²(2)=1.39, p=.50).

The trial achieved its primary implementation goals, demonstrating clear differentiation between light and moderate-to-vigorous intensities at both acute and chronic timescales. Acute sessions elicited significantly higher perceived exertion and heart rate responses under ModVig, validating the acute paradigm used to probe rapid hippocampal-cortical FC changes. Chronic training fidelity was high, with excellent adherence and HR profiles matching targeted intensity progressions. Importantly, the ModVig group exhibited significantly greater gains in CRF compared with the light-intensity control, supporting the premise that exercise intensity drives physiological adaptation. These results establish the necessary conditions to test the central hypotheses: that acute, rapid changes in hippocampal-cortical FC may predict longer-term FC and memory improvements, and that CRF and related physiological processes mediate cognitive benefits. While cognitive and connectivity outcomes are not reported here, the validated manipulation and strong adherence enhance confidence that subsequent analyses can sensitively evaluate the relationships among acute brain responses, training-induced changes in FC, and hippocampal-dependent learning. Given the centrality of hippocampal network integrity to cognitive aging and Alzheimer’s disease risk, the demonstrated capacity to alter CRF and deliver targeted exercise doses supports the trial’s translational objectives.

The Brain-EXTEND trial was designed to determine the impact of aerobic exercise intensity on hippocampal-cortical functional connectivity and hippocampal-dependent learning, and to link acute brain responses and CRF changes to chronic outcomes. This report confirms successful implementation and fidelity of both acute and chronic exercise paradigms, high adherence, and significantly greater CRF gains with moderate-to-vigorous training versus light-intensity control. These findings provide a strong foundation for forthcoming analyses of cognitive and neuroimaging endpoints and mechanistic pathways, including vascular and molecular markers. Future work will test whether acute hippocampal-cortical connectivity responses predict longitudinal FC and learning changes, and evaluate physiological mediators and moderators (e.g., CRF, arterial stiffness, sleep/sedentary behavior, genomic trainability) to optimize exercise prescriptions for reducing age-related cognitive decline and Alzheimer’s disease risk.

• COVID-19 disruptions necessitated suspension of in-person visits and increased home-based training, potentially introducing variability in supervision and environment, though groups were balanced and adherence remained high.
• Missing post-intervention VO2max data (n=24; 12 per group) required multiple imputation (assumed missing at random), which may introduce uncertainty despite consistent findings in complete-case analyses.
• Resting HR used for HRR calculations was measured seated on the bike prior to the maximal test, potentially elevating target HRR ranges for acute sessions.
• Imaging conducted on two scanner models; although balanced and without baseline differences, multi-scanner acquisition can introduce variance.
• Ethnicity data were incomplete for some participants; the sample was predominantly White, which may limit generalizability.
• Cognitive and connectivity outcome results were not presented in this report, limiting conclusions about memory and brain network changes at this stage.