The LEAD trial took place in Toronto, Canada, from July 2018 to July 2020. Four groups of participants were cluster randomized into the intervention or control group: 1) a combined aerobic and resistance exercise intervention (EX) with additional group-based dietary intervention called the Baycrest Brain-healthy Eating Approach (BBEA) (EX+DIET), or 2) a combined aerobic and resistance exercise intervention (EX) with additional group-based brain health education (BHED) on brain aging and tips to support brain health (EX+ED). The BBEA encompasses didactic nutrition education regarding a specific brain-healthy dietary pattern combined with goal setting and strategy training to promote sustainable dietary change (Bar et al., 2021). BHED acted as a time-matched placebo and was designed to be of equal frequency, duration, and social interaction/engagement as the BBEA. Outcomes were measured at baseline, 6 months (post-intervention) and 12 months (follow-up).
The LEAD study was a sub-study of the Canadian Consortium on Neurodegeneration in Aging (CCNA; ccna-ccnv.ca) and participants were enrolled in the CCNA Comprehensive Assessment of Neurodegeneration and Dementia (COMPASS-ND) study (NCT03402919). COMPASS-ND is a longitudinal study involving comprehensive clinical and neuropsychological testing, as well as neuroimaging to determine cohort membership (e.g. healthy control, SCD, mild cognitive impairment, and various forms of dementia) at baseline and at 2-year follow-up (Chertkow et al., 2019). Refer to Additional file 2: Appendix 1 for a more detailed description of the COMPASS-ND study. Participation in LEAD was contingent on first completing the COMPASS-ND baseline assessments which took place over 4 visits at either the Centre for Memory and Aging or Baycrest memory clinic in Toronto. The LEAD study was a collaborative effort between Baycrest Hospital, Sunnybrook Hospital, and University Health Network’s Toronto Rehabilitation Institute (TRI), all situated in a large urban centre in Toronto, Ontario, Canada. Baycrest was the lead site, responsible for recruitment. The intervention sessions took place at TRI’s comprehensive Cardiovascular Prevention and Rehabilitation Program (CRP).
Participants were recruited through a research volunteer database at Baycrest, investigator talks, community advertisements, and clinically through the memory clinics and TRI’s CRP. Following telephone contact and screening to determine inclusion criteria, participants attended the initial COMPASS-ND assessment to confirm eligibility and provide written consent to participate in COMPASS-ND and LEAD. Once participants completed baseline COMPASS-ND assessments they were enrolled into the LEAD study. Patients referred to TRI’s CRP were informed about the LEAD trial and screened by LEAD researchers during their intake assessment. If they were deemed to be eligible and were interested in participating, they were invited to attend the initial COMPASS-ND assessment.
Participants were aged 60–85 years old with SCD or EMCI and ≥2 vascular risk factors. The initial COMPASS-ND visit was used to establish a research diagnosis. SCD was determined by answering “Yes” to the following Jessen questions: “Do you feel like your memory or thinking is becoming worse?” and “Does this worry you?” (Jessen et al. 2014; Jessen et al. 2014). They also had minimal or no cognitive deficit as indicated by having: (1) a delayed recall score on Story A of the Logical Memory subtest of the Wechsler Memory Scale—Revised (Wechsler, 1997) above the ADNI education-adjusted cutoffs (= 9 for 16+ years of education; = 5 for 8–15 years of education; = 3 for 0–7 years of education); (2) a Montreal Cognitive Assessment (MoCA) total score above 24 (Nasredine et al., 2005) ; (3) a delayed recall score on the Consortium to Establish a Registry for Alzheimer’s Disease (CERAD) word list above 5; and (4) a global Clinical Dementia Rating (CDR) score lower than 1.0. Although our original plan was to recruit only people with SCD, we expanded our inclusion criteria to include EMCI. The ADNI criteria (Aisen et al., 2010) was used to determine EMCI, and included individuals who scored between 20 and 24 on the MOCA and 5 on the CERAD. We report on the group assignment of SCD and EMCI in the manuscript. Participants were required to possess ≥2 of the following vascular risk factors: overweight (BMI>25), or physician diagnosed type 2 diabetes mellitus (T2DM) or pre-diabetes (HbA1c ≥ 6.0%) (Diabetes Canada, 2018); high cholesterol; or hypertension. They were also required to be reasonably sedentary at baseline (less than 75min per week of moderate or vigorous intensity physical activity assessed using the Godin Leisure Time Exercise Questionnaire (Godin & Shephard, 1985), be consuming a reasonably poor quality diet (Additional File 2: Appendix 2: Diet screening questionnaire), and be available for the whole intervention. Participants also had to have a study partner who knew them well enough to complete informant questionnaires (COMPASS-ND requirement). Exclusion criteria were: significant known chronic brain disease; major surgery within the past 2 months; major depression or clinical anxiety disorders; schizophrenia or other major psychiatric disorders; ongoing alcohol or drug abuse; inability to undergo an MRI scan; or contraindications to an exercise program (ACSM, 2000)
Exercise took place at TRI’s CRP in Toronto, Ontario. This clinical program services over 1800 adults/year and has demonstrated success in multiple vascular cohorts, notably coronary artery disease (Marzolini et al., 2008; Marzolini et al., 2012). The program is led by an interprofessional team of physicians, physiotherapists, nurses, kinesiologists, psychologists, and dietitians. All LEAD participants engaged in the standard CRP program which included aerobic and resistance training, education/counseling on exercise and cardiac health, and goal setting. Participants were required to complete 5 aerobic and 2-3 resistance training sessions per week. Participants attended 60-minute group supervised exercise sessions once per week, with the balance of the exercise being completed in the home/community. Thirty minutes of exercise education was provided before every supervised exercise session. Embedded into the exercise education sessions were 2.5 hours total of formal nutrition education which addressed fats, fibre, sodium, and food labels. Study participants also had access to clinical program’s psychosocial counseling upon request.
The initial walking prescription was set at a distance of approximately 1.6 km per day and an intensity equivalent to the ventilatory anaerobic threshold and/or 60-80% of V̇O2peak (mL⋅kg−1⋅min−1). Prescriptions were progressed to a maximum of 6.4 km and then increasing intensity to a maximum of 80% of V̇O2peak. The intensity was re-evaluated based on results of a 3-month CPA, which is a standard CRP assessment. Resistance training was started in week 8 of the program, and was performed 2 to 3 times per week (1-2 times at home in an unsupervised setting). Resistance was provided by hand-held dumbbells, the patient’s body weight, and elastic bands of different thickness. Exercises included 3 lower body, 5 upper body, and 2 trunk-stabilizing exercises. In the first weekly class, a moderate load was selected for each of the ten exercises (the selected load was chosen to yield a value of <15 out of 20, i.e., <hard/heavy, on the Borg 6-20 rating of perceived exertion scale (RPE) on the last repetition of 10 repetitions). At home, the participants were required to perform 1 set of 10 repetitions for each exercise using the same load, twice over the following week and to gradually increase the number of repetitions from 10 to 15 over following weeks. Once they could perform 2 sets of 15 reps comfortably, they were advised to increase the weight/resistance load, but reduce the repetitions to 10, and then build again to 15.
Participants were required to log each exercise session, noting the precise distance walked/jogged, duration, resting and peak heart rate, Borg RPE, and any symptoms experienced during exercise. Participants were trained to measure resting and exercise heart rates during orientation to the program and accuracy was checked each week. Resistance training logs included the Borg RPE of the last repetition of the last set performed, any symptoms encountered, the amount of weight lifted, and the number of repetitions and sets performed for each workout. This record was submitted and cross validated by an exercise specialist during the participant’s weekly visit.
Baycrest Brain-healthy Eating Approach
The BBEA encompassed didactic nutrition education, goal setting and strategy training to promote sustainable dietary change (Table 1). We first defined a dietary pattern which drew on contemporaneous scientific evidence linking diet to improved cognitive functioning in older adults with VRFs. In particular, the recommended dietary pattern encompassed key features of two dietary interventions associated with cognitive benefits as implemented in the PREDIMED (Valls-Pedret et al., 2015) and ENCORE (Smith et al., 2010) trials. Since neither trial published complete nutritional intake data for their cognitive subsamples, targeted and achieved levels of nutritional intake data from the larger trial populations were compared and formed the basis of the recommendations (Epstein et al., 2012; Zazpe et al., 2008). In some cases, evidence from prospective cohort studies, published at the time, was used to further refine the recommendations. The principal foods or food groups that were targeted were total vegetables; raw or leafy green vegetables; cruciferous vegetables; total fruit; berries; unsalted nuts or all natural butters with an emphasis on walnuts; fish or seafood; fatty fish; canned beans or cooked dried beans; meat and poultry; red or processed meat; butter, cream, or high fat dairy spreads; white bread; and processed foods. A translational product called the Brain Health Food Guide (BHFG) (Parrott, 2016; Additional File 2: Appendix 3) was created to educate individuals on the diet recommendations.
The BBEA builds upon a previously established program designed to manage age-related cognitive changes in older adults (Dawson et al., 2014). Once per week, participants attended a 1-hour group session with the study dietitian (Additional File 2: Appendix 4) who underwent formal training by individuals with expertise in this approach. The dietitian used guided discovery to help participants develop and adapt their own goals and identify ways to overcome barriers to implementation. Participants took part in an iterative process of working to attain their personalized diet goals using a meta-cognitive strategy (Goal-Plan-Do-Check). Participants self-selected goals on a weekly basis, which were rationalized within the context of the BHFG and individuals were expected to incrementally improve an aspect of their diet on a weekly basis. Guided discovery, goal-setting and the identified meta-cognitive strategy are key elements of the CO-OP Approach from which they were derived (Dawson, et al., 2017). The didactic education curriculum within the BBEA included some set topics around the BHFG, but also many open sessions where participants were given the chance to pick topics that would help them move towards their goals. There were also brainstorming sessions where the group came up with ideas, and the dietitian only acted as a facilitator, not education provider. Results from the baseline Canadian Diet History Questionnaire II (C-DHQ II) (Csizmadi et al., 2007), describing intake patterns according to BHFG food groupings were distributed to participants during the first or second session to help participants identify areas where their diet could be improved. Twice throughout the second half of the program, participants were allowed to schedule individual meetings with the study dietitian. If participants felt they needed to schedule additional individual sessions, they were accommodated on a case by case basis. Participants submitted weekly logs outlining diet goals, plans, successes and obstacles to help directly identify sequential goals. Every month, participants were asked to complete an eating pattern self-assessment to help them assess the degree of dietary change achieved (Additional File 2: Appendix 5).
Once per week, participants in the BHED group participated in a 1-hour passive group discussion and education session surrounding lifestyle practices to support brain health which acted as the placebo to the BBEA (Additional File 2: Appendix 4). The BHED program was adapted from the DISCOVERY program used by our partner CCNA sub-study ENGAGE (NCT#03271190) as the placebo program to their intervention (Belleville et al., 2019). These classes were designed to be of equal frequency, duration, and social interaction/engagement compared to the BBEA sessions and were manualized to ensure consistency between trainers. During the classes, participants received information on the brain and cognitive processes, the effect of age on cognition, and tips to promote successful aging. Twice throughout the second half of program, participants were allowed to schedule individual meetings with a study dietitian. Individuals in these groups did not participate in strategy training to assist them in undertaking dietary changes nor receive the supplemental BHFG information.
Feasibility measures included recruitment, retention, adherence, and safety. We report data on recruitment sources and the number of participants screened and enrolled. Retention rate post-intervention and after 6 and 12-month follow-up assessments was calculated as the percentage of enrolled participants. Since the assessment protocol was different for those who completed the intervention pre-COVID-19 versus during the COVID-19 pandemic (described below), retention rates through follow-up assessments are described separately by time point. Adherence was measured as the percentage of intervention sessions attended. We also report the percentage of exercise logs and diet self-assessments completed and the number of one-on-one sessions with the dietitian that were scheduled. Following the intervention, participants were asked to complete an anonymous feedback questionnaire (Additional File 2: Appendix 6,7,8) about knowledge/skills gained, the length of sessions, what they found most or least useful, and whether they would recommend the study to other individuals concerned about their memory.
Original LEAD outcome measures included structural and functional MRI, fasting blood biomarkers, cognitive function, anthropometric measurements, dietary adherence, physical fitness, gait and balance assessment, and a variety of questionnaire-assessed lifestyle and psychosocial factors (Additional File 2: Appendix 9). Assessments were spread over 3 visits of approximately 2-3 hours each. These visits took place at Sunnybrook hospital except for fitness testing and blood draws which were done at TRI and the memory clinics, respectively. COVID-19 impacted the 6 month assessments for 3 participants in the EX+DIET intervention and the 12-month assessment for 3 participants in the BHED group. These individuals were invited to take part in a shortened assessment which included a diet assessment, graded exercise test, bloodwork, MRI, and a remote cognitive assessment. This resulted in incomplete data collection and a revised focus on preliminary efficacy, measuring change from baseline to 6 months using outcomes from the shortened assessment.
Participants underwent symptom-limited, graded exercise tests on a cycle ergometer (Ergoline 800 P) or a treadmill (Quinton®) at the discretion of the cardiology technologist and physician depending on balance, mobility, and participant preference. Participants were tested using the same modality at baseline and follow-up. The Bruce protocol was used for patients being tested on the treadmill (Bruce et al., 1973) and for the cycle-ergometer protocol, the workload was increased by 50 kpm per minute. A 12-lead electrocardiogram (Quinton®, Q-Stress system) was monitored continuously and breath-by-breath gas samples were collected and averaged over 20-second periods via calibrated metabolic cart (VMAX Encore and Spectra – CareFusion, Yorba Linda, California, USA). Peak oxygen consumption (V̇O2peak) was determined.
Adherence to the BHFG was measured using the C-DHQ II (Csizmadi et al., 2007). The C-DHQ II is a web-based, 153-item questionnaire that takes approximately 1 hour to complete. Participants completed the questionnaire during the first and last group session so that the study coordinator and dietitians were available to assist. Items from the C-DHQ II were mapped on to items from the BHFG and target ranges for 14 key foods were used to assess dietary adherence. For “foods to include”, we rated participants’ adherence to each of these items as either 1 = greater than or equal to the target, 0.5 = intake greater than equal to 50% of the target, 0 = below 50% of the target. For “foods to limit” we rated participant’s adherence as either 1 = intake less than or equal to the target, 0.5 = intake exceeding the target by an amount less than or equal to 50% of the target, or 0 = intake exceeding the target by an amount greater than 50% of the target. These ratings were summed to calculate a composite Brain-healthy Eating Index (BEI) representing a participant’s adherence to the BHFG (maximum scores = 14).
MRI data were collected according to the Canadian Dementia Imaging Protocol (Duchesne et al., 2019). A 60-minute brain MRI protocol consisted of: 1) high spatial resolution anatomical imaging sequences, 2) an attention-related task-based blood oxygenation level dependent functional MRI (fMRI) that involved button responses to visual stimuli during a Flanker test, 3) two functional sequences during a resting state, one of which used BOLD contrast while the other used arterial spin labeling (ASL) for cerebral blood flow, and 4) additional structural sequences to assess small vessel disease and white matter integrity, i.e. fluid attenuated inversion recovery and diffusion tensor imaging, respectively. The sequences are as follows: 3D T1-weighted MRI, PD/T2-weighted MRI, FLAIR, Gradient Echo, Resting State fMRI (BOLD), DTI, PCASL resting, Attention-based task fMRI (BOLD).
Images were acquired on a 3 Tesla Siemens Prisma scanner with a 12-channel head coil. Each participant’s head was restrained using cushions that fit inside the head coil. High-resolution structural images (T1-weighted three-dimensional magnetization-prepared rapid gradient-echo sequence; 3D-MPRAGE) were acquired with the following parameters: TR/TE = 2300/2.98, FOV = 256mm, slice thickness = 1mm, number of slices = 192. Whole hippocampal segmentation was performed using an established deep learning HippMapp3r algorithm (hippmapp3r.readthedocs.io) that was based on a convolutional neural network (Goubran et al., 2019). It uses a T1-weighted image as the only input and the outputs are segmentation masks for the left and right hippocampi.
Our cognitive testing session lasted 2-3 hours and was based on a harmonized CCNA neuropsychological battery and other cognitive tests to harmonize with a concurrent CCNA sub-study (ENGAGE: NCT#03271190). Of the latter tests, the following were administered at baseline, 6, and 12 months: Direct Assessment of Functional Status – Revised (DAFS-R) (McDougall et al., 2009), Number-letter computer task, Memory toolbox task (Troyer, 2001), Geriatric Anxiety Inventory (GAI) (Pachana et al., 2007), Apathy Inventory (participant version) (Robert et al., 2002). The following were administered only at baseline: Beck Anxiety Inventory (BAI) (Beck et al., 1979), Beck Depression Inventory (BDI)-II (Beck et al., 1979).
Follow-up testing also included a selection of tests that were administered to participants during the CCNA - COMPASS-ND baseline assessment and these data were used to determine pre-post cognitive changes. These tests include the Jessen questions, Montreal Cognitive Assessment (MOCA), Rey Auditory Verbal Learning Test (RAVLT) – immediate and delayed recall (Schmidt, 1996), Trail Making Test (TMT) (Reitan & Wolfson, 1992), Digit Symbol Substitution test (DSST) (Weschler, 1997)), Face-Name Association test - immediate and delayed recall (adapted from a task being used in the CIMA-Q study, www.cima-q.ca/en/home), DKEFS Color Word Interference (Lippa et al., 2010), Geriatric Depression Scale (GDS) (Yesavage et al., 1983), Activities Specific Balance Confidence Scale (ABC) (Powell et al., 1995), Pittsburg Sleep Quality Index (PSQ) (Carpenter, 1998), MAYO clinic fluctuations scale, and Quality of Life – AD scale.
During the pandemic we administered remote cognitive assessments to 3 participants via the Zoom video conferencing application. Tests were administered by a trained assessor and participants were asked to set up their computer/camera in a quiet room without any distractions. A package with all testing forms and a return envelope was mailed to participants who were instructed not to open the package until the beginning of the scheduled session. At the beginning of the assessment, the assessor helped to work out any technical issues and made sure they had the proper camera angles in order to see the participant and the test forms depending on the test. All the follow up tests listed above were administered except for DAFS, Number-letter computer task, and Face-Name Association Task. At the end of the assessment, the participant was instructed to seal the testing forms in the return envelope while in camera view and mail it back. One participant was unable to set up video conferencing so they were sent a package of questionnaires to be filled out on their own and the RAVLT was administered over the telephone. The RAVLT was the only cognitive test for which we had complete data at baseline and 6-months.
Fasting blood samples were collected and analyzed for inflammatory cytokines, oxidative burden and BDNF and APOE status using CCNA platforms. The blood draw was done at the same memory clinic at baseline and follow-up and was drawn according to CCNA protocols. A tube for Vitamin K analysis was collected at all time points. Vitamin K was assessed by high-performance liquid chromatography (HPLC) (Presse et al., 2013). HbA1c and Vitamin K were processed, stored, and analyzed locally. Broader measures of biological markers (e.g., LDL, HDL, triglycerides, etc.) were drawn and stored by the CCNA but were not yet analyzed by the time this manuscript was prepared.
Sample size was originally estimated to observe an overall effect of participation in an exercise intervention on brain structural, functional, and cognitive outcome measures. The a priori plan was to first compare the effect size associated with pre-post changes in individuals in this study, relative to comparable patient populations participating in either the exercise intervention or a stretching placebo control at the TRI and undergoing the same outcome measures. Based on an expected grey matter volume effect size of 0.27 (partial eta-squared = 0.07), 3 time points, 2 groups, power = 0.9, alpha = 0.05 and correlation between repeated measures = 0.5, we required a minimum sample size of 60. The total sample size of 66 was used to accommodate a retention rate of approximately 90% and allow for the following covariates: inflammatory cytokines, oxidative burden, BDNF/APOE status and sex. Our effect size was derived from neuroimaging work that reported a grey matter volume increase in a time by group analysis among older sedentary adults (Erickson et al., 2011) and is comparable to other older adult exercise studies with functional, (Rosano et al., 2010) hemodynamic (Ivey et al., 2011) and cognitive (Baker et al., 2010) outcome measures.
Randomization and blinding procedure
We used cluster randomization whereby 4 groups of participants were randomly allocated into one of the two study arms. We chose to randomize in clusters in order to enrol groups of 6 participants at the same time and complete all their baseline measurements within a 2 to 4-week window before starting the intervention. We did not anticipate it would be feasible to simultaneously enrol and assess 12 participants (required for individual randomization) in COMPASS-ND and LEAD within this timeframe. Randomization was done using the Random.org online software. Assessors were blinded to group assignment and participants were asked not to mention their group assignment to assessors. To keep participants blinded from the hypotheses and group effect expectations, the content of the intervention and the wording of recruitment documents and consent forms did not convey the differences between EX+DIET and EX+ED conditions. Participants were specifically told that the groups would differ in the type of nutrition and brain health education that was provided.
We performed statistical analyses using R (v.3.3.1) (R Core team, 2019). Demographic data are presented as means and standard deviations or frequencies/percentages where appropriate. Frequencies/percentages were used to summarize recruitment, retention, and adherence. Responses to quantitative questions from the surveys were summed and reported. For open ended questions, responses from 2 or more participants with similar themes were grouped together and reported.
Descriptive statistics and effect sizes (Cohen’s d) were used to characterize changes from baseline to 6 months within each group and differences in changes between groups for outcomes included in our shortened follow-up assessment. These included diet (BEI scores), cardiorespiratory fitness (V̇O2peak), glycated hemoglobin (HbA1c), vitamin K, hippocampal volume, and the Rey Auditory Verbal Learning Test (RAVLT) - delayed recall.