High-intensity Exercise and Cognitive Function in Cognitively Normal Older Adults: A Randomised Clinical Trial

Belinda Brown (  b.brown@murdoch.edu.au ) Murdoch University https://orcid.org/0000-0001-7927-2540 Natalie Frost University of Western Australia Stephanie R. Rainey-Smith Edith Cowan University James Doecke CSIRO: Commonwealth Scienti c and Industrial Research Organisation Shaun Markovic Murdoch University Nicole Gordon Murdoch University Michael Weinborn University of Western Australia Hamid R. Sohrabi Murdoch University Simon M. Laws Edith Cowan University Ralph N Martins Edith Cowan University Kirk I Erickson University of Pittsburgh Jeremiah J Peiffer Murdoch University


Introduction
Physical inactivity is considered the greatest modi able risk factor for dementia (1); however, attempts to provide de nitive evidence from randomised-controlled trials (RCTs) of a link between exercise and enhanced cognition have been inconsistent. Indeed, a 2015 Cochrane review (2) of RCTs concluded there is insu cient evidence, in cognitively normal older adults, to suggest an effect of exercise on cognition.
While a more recent meta-analysis (3), assessing a similar cohort, identi ed positive effects of exercise when session durations were in excess of 45 min and at least of moderate intensity. These con icting results indicate the need for greater clarity for the use of exercise as a method for preventing cognitive decline, speci cally the precise parameters needed for improving brain health.
When compared with the total volume of physical activity, observational work has reported a stronger association between objectively measured intensity of physical activity and cognitive function (4,5).
Furthermore, acute bouts of high-intensity exercise improve memory and executive function to a greater extent than moderate-intensity continuous exercise bouts (6,7). Although previous work in the area is promising, the importance of exericise intensity in enhancing cognitive health requires rigorous examination in RCTs. A recent 12-week intervention in older adults demonstrated greater improvements in memory after undertaking high-intensity compared with moderate-intensity exercise, or a stretching control (8). The use of high-intensity exercise is safe in older populations (9) and provides a timeeffective method to increase physical health, yet, until more consistent and rigorous evidence is available, the widespread use of high-intensity exercise to enhance cognitive health will continue to be questioned.
Variability across studies might also be explained by factors moderating exercise-induced changes in cognition. Genetic factors, such as the apolipoprotein E (APOE) ε4 allele and the brain-derived neurotrophic factor (BDNF) Val66Met single nucleotide polymorphism, may modulate the relationship between exercise and brain health (10)(11)(12). The literature in these elds is predominantly sourced from observational studies, which have contributed to largely inconsistent ndings. In addition, variability in cognitive response may be due to variability in cardiorespiratory tness change following exercise (13,14). Evidence from RCTs is needed to gain a greater understanding of these potential mediating and moderating effects on cognition.
The current proof-of-concept RCT was designed to provide a head-to-head comparison of work-matched moderate-intensity and high-intensity exercise on cognition in cognitively normal older adults. We hypothesised that both intervention groups would receive bene ts to cognition, but the high-intensity group would receive additional bene t beyond the moderate-intensity group in a dose-dependent fashion. Based on the hypothesis that increases in cardiorespiratory tness are an important factor in the relationship between exercise and cognition, we investigated whether change in tness is associated with improved cognition. Finally, we investigated whether targeted genetic factors (APOE ε4 carriage and BDNF Val66Met) moderate the effect of exercise on cognition, and the relationship between altered tness and cognitive changes. Based on previous literature, we hypothesised that APOE ε4 carriers and BDNF Val66Met carriers would receive the greatest bene t from exercise, in terms of cognitive improvements (12,15).

Trial design
The Intense Physical Activity and Cognition (IPAC) study was a single-site parallel randomised controlled trial conducted between October 2016 and September 2019 at Murdoch University and the Australian Alzheimer's Research Foundation, Western Australia. An open access protocol paper for the IPAC study has been published previously (16). Participants were randomised to either six-months of supervised high-intensity exercise, supervised moderate-intensity exercise, or an inactive control group.
The IPAC study is registered with the Australian New Zealand Clinical Trials Registry (ACTRN12617000643370). The human research ethics committees at Murdoch University and Edith Cowan University approved the conduct of this study, and all participants provided written informed consent.

Participants and Randomisation
Participants were recruited between October 2016 and November 2017 from a number of sources, including media advertisement, yers, and word-of-mouth. A full list of inclusion and exclusion criteria and calculations of sample size have been described previously (17). A block randomisation protocol (conducted by a researcher who was not collection outcome data) was used to randomly assign participants to one of the following three groups: high-intensity exercise, moderate-intensity exercise, or a control group.

Interventions
Participants that were allocated to an exercise group (i.e., high-intensity or moderate-intensity) engaged in exercise two times per week (under the supervision of an Accredited Exercise Physiologist) for 6 months. Each exercise session lasted 50 minutes and was conducted on a cycle ergometer (Wattbike Pro; Wattbike, Australia). Target intensity was set using the 6 to 20 Borg Scale of Perceived Exertion (18).
Further details of the exercise interventions can be found within the supplementary material.
Adherence to the intervention was measured via session attendance. In addition, exercise intensity was calculated for each participant in the moderate-and high-intensity groups: the percentage of peak aerobic power (measured via a graded exercise test) was calculated for each session (not including the warm-up, cool-down, and for the high-intensity group, the recovery between intervals). The percentage of peak aerobic power in the initial three months were calculated using baseline peak aerobic power output, while months four to six were calculated using peak aerobic power output from the mid-intervention tness test.
Participants randomised to the control group were provided with an information session regarding the bene ts of exercise for overall physical health and known bene ts to the brain. Participants in the control group did not receive any exercise instructions.

Procedures and outcome measures
Full methodological detail regarding outcome measures can be found in the supplementary material.

Cognitive assessment
A comprehensive battery of neuropsychological tests was administered to all participants at baseline, 6months, and 18-months. Composite scores for global cognitive function, attention, episodic memory and executive function were calculated.

Physical assessment
At baseline, 3-, 6-, and 18-months, all participants underwent a cycling-based graded exercise test to quantify peak aerobic capacity (VO 2 peak) and peak power. All participants also underwent a dual-energy X-ray absorptiometry (DXA) scan, using a Hologic Discovery Bone Densitometer (Hologic, USA), in order to quantify volume of fat, muscle and bone tissue in the body.

Statistical methods
Analyses were conducted in R statistical computing packages version 3.6.2 (20) and Statistical Package for the Social Sciences Version 24 (IBM). Data were inspected to determine parametric testing was appropriate for all physiological and cognitive variables.

Descriptive statistics
Descriptive statistics were calculated to compare baseline information across study groups. Analyses of variance (for continuous variables) and chi-square analyses (for categorical variables) were conducted to identify differences.

Intervention group analysis
All participants that completed a baseline assessment were included in the intention-to-treat (ITT) analyses, regardless of adherence to session attendance, or study withdrawal. To examine the effect of study group on cognition over time, a series of linear mixed models (LMMs) were conducted. Repeated cognitive composite scores were entered as dependent variables, and age, gender, education, time (years), group, and time*group were entered as xed factors, and participant identi cation number as a random factor, into the model. Post-hoc group comparisons were conducted for any signi cant time*group interactions (high-intensity as reference group). LMMs were conducted for baseline and 6-month data only, and then again for baseline, 6-and 18-month data. We report raw mean change scores from baseline and 95% con dence intervals, and unstandardized beta coe cients (B) and their standard error.
A positive B represents a positive slope for the moderate/high-intensity groups, compared with the control group.
The per-protocol analyses were run in the same manner as the LMMs described above, yet this was only conducted for participants that attended at least 75% of exercise sessions.
We ran LMMs (for ITT data only) to investigate the moderating effects of BDNF Val66Met and APOE ε4 carriage on cognitive performance over the intervention. Each LMM had an additional interaction of either BDNF*time*group or APOE *time*group entered into separate models.

Individual variability analysis
The relationships between change in cognition and change in cardiorespiratory tness (VO 2 peak) from pre-to post-intervention within the high-intensity group only, and for all study participants, were examined. Residualised change scores were generated by entering post-intervention score as the dependent variable and pre-intervention score as the independent variable. Linear models were run with the residualised cognitive change as the dependent variable and residualised tness change as the independent variable (age, gender, and education as covariates) (21). For all study participants, the linear models were re-run with the inclusion of either BDNF* tness change or APOE* tness change. The cohort was then strati ed by BDNF Val66Met carriage or APOE ε4 carriage and the linear models re-run.

Results
One hundred and eight participants were enrolled, with ninety-nine completing all baseline assessments and subsequent randomisation to a study group ( Fig. 1; descriptive data, Table 1). Seven participants withdrew during the six-month intervention. Individuals excluded from the per-protocol analyses (n = 13; based on low adherence or withdrawal from study) reported higher education (15.3 ± 2.1 y), compared with those included (13.9 ± 2.3 y; t = 2.17, p < 0.05; eTable 1).

Adherence to prescribed intervention
There was no difference in exercise session attendance between the high-intensity (85.5 ± 12.4%) and moderate-intensity (86.3 ± 9.8%) groups.
The high-intensity group maintained 120.6 ± 25.1% of peak aerobic power during the high-intensity intervals, while the moderate-intensity group cycled continuously at 70.1 ± 16.3% of peak aerobic power.
There were no serious adverse events recorded.

Page 13/21
The current study compared the impact of six months of supervised high-and moderate-intensity exercise on cognition in a group of cognitively normal older adults. Our intervention successfully delivered high-intensity exercise and resulted in greater increases in cardiorespiratory tness, compared with the moderate-intensity group. We did not observe any bene cial effects to cognition when comparing group performance from pre-to post-intervention. However, changes in cardiorespiratory tness from pre-to post-intervention were associated with changes in global cognition and executive function in the high-intensity group, and the entire cohort. Furthermore, we observed moderating effects of the BDNF Val66Met polymorphism, whereby the relationship between change in cognition and change in tness was only evident in Val/Val homozygotes (i.e., non-Met carriers).
When examining the group-level data, we found no effect of either the high-or moderate-intensity exercise interventions on cognitive performance. These ndings are inconsistent with a recent RCT which demonstrated improvements on a single high-interference memory task, following a 12-week highintensity exercise intervention, compared to moderate-intensity exercise (8). As the cohort within the current study and that investigated by Kovacevic et al. were similar for age, cognitive status, and health, a methodological difference between studies may instead account for the disparate ndings. It is possible that more frequent exercise, at least thrice-weekly (delivered by Kovacevic and colleagues), is required to induce cognitive bene t, even with a shorter intervention period (12 weeks). Indeed, the induction of neurotrophic factors (e.g., BDNF) may be required on a more 'frequent' basis to contribute to detectable neural bene ts (22). It would not be surprising that duration and frequency, in addition to intensity, play an important role in exercise-induced cognitive response. Thus, future studies are required to further elucidate the optimal exercise parameters for bene ting cognition.
Within the entire cohort, and high-intensity group alone, increases in cardiorespiratory tness were associated with improvements in global cognitive performance and executive function, with moderate to high effect sizes observed. These data support prior work that has reported relationships between exercise-induced improvements in cardiorespiratory tness and cognitive changes (13,21,23,24). The high-intensity exercise intervention in the current study increased cardiorespiratory tness levels greater than prior similar RCTs in older adults (typically 10-15% increases) (8,13,25). It therefore remains puzzling as to why the observed associations between changes in cardiorespiratory tness and cognition did not yield group-level differences in cognitive outcomes. While it is possible that individuals with the poorest baseline cognition and tness levels were more likely to experience tness-associated cognitive improvement irrespective of exercise intervention, our statistical analysis at least partially corrected for this potential bias. It is also important to consider whether the cognitive assessments used within this generally high-functioning sample were sensitive enough to detect differences between groups.
The associations between changes in cardiorespiratory tness and global cognition, executive function, and episodic memory were strongest in BDNF Val/Val homozygotes. Increases in BDNF levels is one of the most well-supported mechanistic theories underlying the relationship between exercise and brain health (26). BDNF is synthesized in cells as a precursor molecule (pro-BDNF), which undergoes proteolytic cleavage to yield the mature form. Carriage of the BDNF Met allele can negatively alter the processing of pro-BDNF to mature BDNF in neurons: based on our data, it is possible that exercise may not be potent enough to counteract this detrimental phenotype. Consistent with our ndings, a recent systematic review on this topic revealed Val/Val homozygotes are more likely to gain bene t from exercise in terms of better memory performance, compared with Met carriers (27). Moreover, we also observed a relationship between tness and global cognition in APOE ε4 carriers, but not non-carriers; however, the interaction term APOE* tness change was not signi cant (indicating that a moderating effect does not exist). It is likely that combinations of genetic factors in uence the relationship between cardiorespiratory tness and cognition. Previous studies have detected cumulative effects of APOE ε4 carriage and BDNF Val66Met carriage on cognitive decline (28); however, our study was not su ciently powered to examine the APOE*BDNF interaction. Consequently, appropriately powered exercise interventions coupled with hypothesis-driven genetic investigation may reveal more on the ability of exercise to either provide added bene t to those with optimal genetic factors, or alternatively counteract detrimental genotypes.

Strengths and limitations
The lack of an active control group within the current study contributed to differences in exposure (e.g., social interactions) between our exercise groups and inactive control groups. Nevertheless, as we did not detect group differences across cognitive outcomes, this is unlikely to have affected the results reported here. Our cohort was a generally homogenous sample of highly educated, Caucasian older adults living in the community, and our results may not be applicable to the wider population. Strengths of the study include the three-group design that aimed to detect intensity differences, and our ability to objectively examine intensity levels throughout the intervention. Indeed, our use of rate of perceived exertion to monitor within-subject intensity proved to be an effective method in our older adult cohort, allowing individuals to maintain appropriate intensity targets without the need for frequent testing or the use of monitoring equipment (e.g., heart rate monitors).

Conclusions
We found that in a cohort of cognitively normal older adults, six-months of supervised high-intensity and moderate-intensity exercise did not directly contribute to improvements in cognition, compared to an inactive control group. Our data did, however, suggest improvements in cardiorespiratory tness are important for inducing cognitive change, and that genetics may play a role in this response. Overall, our data does not provide evidence that high-intensity exercise can contribute to cognitive change in all individuals. Future work in this eld should be appropriately designed and powered to examine numerous factors that could contribute to individual variability in response to intervention; ultimately leading to individualised prescription of exercise to induce cognitive change and ultimately reduce dementia risk.
Declarations oversight of exercise delivery, interpretation of study ndings and provided a critical review of the manuscript. Linear relationships between change in cardiorespiratory tness (residuals) and change in (A) Global cognition (residuals); (B) Episodic Memory (residuals); (C) Executive Function (residuals); from pre-to immediately post-intervention (6 months) in BDNF Val66Met carriers and non-Met carriers Abbreviations: BDNF Val66Met, brain-derived neurotrophic factor Valine66Methionine single nucleotide polymorphism; VO2peak, peak aerobic capacity ( tness measurement).

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.