Longitudinal Associations between Cardiorespiratory Fitness and Cognition in Older Adults: A 10-Year Follow-Up from the Generation 100 Study

Abstract

Introduction:

Cardiorespiratory fitness (CRF) is associated with better cognition in older adults. However, the long-term associations between CRF levels and cognitive outcomes remain unclear. This study aimed to investigate the association between baseline CRF, changes in CRF (from baseline to year 1 and year 3) and cognition in older adults.

Methods:

We included 770 cognitively healthy older adults (367 females; mean age: 72.0 ± 2 years at baseline) from the Generation 100 Study. All participants had complete data on CRF and important confounding factors (age, sex, educational attainment, and self-rated memory at baseline). CRF was measured directly as peak oxygen uptake using ergospirometry. Cognition was assessed using the Montreal Cognitive Assessment (MoCA) 3-, 5-, and 10 years after baseline. Linear mixed models were used to assess the relationship between CRF at baseline, changes in CRF, and MoCA score over time.

Results:

The mean MoCA score at 3-, 5-, 10-year follow up was 25.3, 25.0, and 24.8, respectively. Higher baseline CRF predicted 0.06 higher MoCA score at year 3 (95% CI: 0.02–0.09) and 0.07 at year 5 (95% CI: 0.03–0.11) but was not associated with MoCA score at year 10 (β: 0.02, 95% CI: −0.02 to 0.06). Neither 1-year nor 3-year changes in CRF were associated with MoCA scores at any later time points.

Conclusions:

Changes in CRF over 1 and 3 years were not associated with cognition in healthy older adults. However, higher baseline CRF was linked to better cognition up to 5 years later, suggesting that achieving and maintaining an age-relative high CRF before entering the 8th decade of life might benefit cognitive aging.

The global population is aging, and with increasing age comes a gradual decline in cognitive abilities such as memory, attention, and executive functions (1,2). While these changes, commonly termed cognitive aging, are a part of normal aging, they are distinct from neurodegenerative conditions such as dementia and Alzheimer’s disease, which involve more severe and progressive cognitive degeneration (3). Understanding the factors that influence cognitive aging is key to promoting brain health and potentially delaying the onset of neurodegenerative conditions. To this end, cardiorespiratory fitness (CRF), which reflects the integrated functions of the cardiovascular and pulmonary systems in delivering oxygen to working muscles (4), has received increasing attention. Emerging evidence links higher CRF, preferably measured directly as maximal oxygen consumption (VO_2max; L/min or mL/kg/min) to better cognitive outcomes in older adults (5–11). The CRF hypothesis suggests that greater fitness enhances oxygen delivery and utilization in brain regions responsible for executive functioning, memory, and attention, thereby supporting cognitive health (12). A recent review showed that older adults with high CRF (26–30 mL/kg/min) performed better on executive tasks than those with lower CRF (17–21 mL/kg/min), and that only those with high CRF exhibited increased brain activity responses during these tasks (13). Altogether, these findings offer compelling support for the potential cognitive benefits of having a high CRF level in older age.

CRF declines with advancing age, with studies reporting a reduction of approximately 2 mL/kg/min per year in older adults (14,15). However, exercise training has been shown to improve CRF even among individuals aged 60 years and older (12,16,17). This raises the important question of whether CRF improvements through exercise training during older age can help slow cognitive aging. The current evidence remains mixed (11,12,18–20). Interventions vary widely in design, duration, and type of exercise and cognitive assessment methods, factors that likely contribute to the inconsistency in findings regarding the relationship between CRF, exercise-induced improvement in CRF, and cognition in older adults. Furthermore, most randomized controlled trials within this context have been relatively short, typically lasting 8 weeks to 24 months (19,20). Such durations may be insufficient, as the neurological adaptations from exercise training could take years to manifest (21,22). Therefore, long-term studies assessing the effect of CRF on cognitive trajectories in older adults are needed. Concurrently, a better understanding of the association between changes in CRF and cognition in older adults could provide valuable insights into the temporal dynamics of CRF and cognitive health, informing more effective monitoring and intervention strategies.

Therefore, in this prospective observational study, we investigated whether baseline CRF and changes in CRF over 1 and 3 years were associated with cognitive function assessed up to 10 years later. We hypothesized that a higher baseline CRF would be associated with better cognitive function, and changes in CRF over 1 and 3 years could influence cognitive outcomes in later years.

METHODS

The Generation 100 Study

The Generation 100 Study was a randomized controlled trial (ClinicalTrials.gov: NCT01666340) (23), in which all residents of Trondheim Municipality born between 1936 and 1942 were invited to participate (n = 6966). In total, 1567 participants (aged 70–77, 790 women) were included in the study. The main inclusion criterion, besides age and residency, was being able to undertake physical exercise training (23). Hence, the exclusion criteria included medical conditions or disabilities preventing exercise training according to the intervention or study completion (23). All participants provided signed informed consent at baseline.

In the Generation 100 Study, participants were stratified by sex and cohabitation status and randomized following a 2:1:1 ratio into either a control group (n = 780) or an exercise training intervention group, which was further randomized to follow an exercise training program with high-intensity interval training (HIIT; n = 400) or moderate-intensity continuous training (MICT: n = 387) for 5 years (23). The control group was asked to follow the national recommendation of 30 min of daily physical activity (24). The exercise training group engaged in either HIIT or MICT exercises on a semiweekly basis. The HIIT sessions included a 10-min warm-up followed by 4 × 4 intervals at 85%–95% of maximal heart rate (HR_max), while the MICT sessions involved 50 continuous min at 70% of HR_max.

Study Design and Population

This is a prospective cohort study using data from the Generation 100 Study. In contrast to the main study, which was conducted as a randomized controlled trial, in this substudy, all participants were treated as a single cohort, regardless of their original group allocations. This substudy was covered by the broad consent provided in the Generation 100 Study and obtained its own ethical approval from the Regional Committee for Medical Research Ethics (REK 183708) in 2024.

To be included in this substudy, the participants must have had complete data on 1) CRF at baseline, and after 1 and 3 years; 2) highest educational level at baseline; 3) self-rated memory at baseline. Participants with 4) symptoms of neurodegenerative disorders at baseline; or 5) confirmed diagnosis of dementia during follow-up were excluded. Conditions 4 and 5 were assessed by a neurologist using the participants’ hospital records from baseline testing until December 2021. Overall, a total of 770 presumably cognitively healthy participants were included (Fig. 1).

FIGURE 1.

Flowchart for study population and timeline. AD, Alzheimer’s disease; CRF, cardiorespiratory fitness.

Clinical Measurements and Questionnaires

Baseline assessments were conducted from 2012 to 2013, with the full physical test battery detailed in the Generation 100 Study protocol (23). Similar assessments using the same protocols were repeated after 1, 3, 5, and 10 years of follow-up, with the 10-year assessment concluding in October 2023. Here, we summarize the testing protocols relevant to this substudy. All assessments were administered by qualified personnel, and all testing equipment underwent frequent quality control procedures at St. Olav’s Hospital, Trondheim, Norway.

Participants’ age (years) and sex were obtained from the Norwegian National Population Registry. In this registry, “sex” refers to legal classification at birth (man/woman) (25). For clarity, we refer to this variable as “gender” (man/woman) in our analyses.

Body mass (kg) was measured using a bioelectrical impedance analyzer scale (InBody 720, BIOSPACE, Seoul, Korea).

Questionnaires were used to obtain information on the participant’s highest education level and self-rated memory at baseline. The educational levels were categorized as primary (less than 10 years), secondary (between 11 and 13 years), and tertiary (14 or more years) education. Self-rated memory was dichotomized as “good” or “poor.”

Cardiopulmonary exercise testing (CPET) to assess VO_2max was conducted at the Core Facility NeXT Move at the Norwegian University of Science and Technology, Trondheim, Norway. Of 770 participants included in this substudy, the majority completed testing on a Woodway treadmill (Woodway PPS MED, Waukesha, WI), while a small number of participants (< 3%) with reduced functional capacity or extreme lower limb pain used a cycling ergometer (Lode B.V Medical Technology, Groningen, Netherlands). The software used was mainly Cortex Metamax II (Leipzig, Germany), but some tests (<4%) were performed using the Oxycon Pro system (Erich Jaeger, Hoechberg, Germany) for logistical reasons.

Before and during the test, participants wore a face mask (Hans Rudolph, United States) connected to a gas analyzer and a heart rate monitoring device (RS100, Polar Electro Oy, Kempele, Finland). All participants performed a 10-min warm-up at a moderate intensity, defined as 70% of their HR_max or a rating between 11 and 13 on the Borg’s scale of perceived exertion (26). Following the warm-up, each participant followed an individualized protocol, in which an increase of 2% in inclination or 1 km/h in speed (or 20 watts on the cycling ergometer) was applied approximately every minute until exhaustion, or criteria for reaching VO_2max were met (23). A true VO_2max was achieved if the VO₂ values increased less than 2 mL/kg/min despite increased workload, and the respiratory exchange rate ratio reached ≥1.05. If a participant failed to reach VO_2max due to exhaustion, a VO_2peak value was reported instead. The reported VO_2max/peak values were calculated as the average of the three highest consecutive VO₂ values registered during the final stage of the CPET. In our substudy, 63% of tests met the criteria for VO_2max; thus, we use the term VO_2peak instead. Assessment of VO_2peak was performed using the same protocol, software and devices at baseline, and after 1 and 3 years of follow-up. The 1-year and 3-year changes in VO_2peak (mL/kg/min) were calculated as the difference between VO_2peak measured at year 1 and baseline, and VO_2peak measured at year 3 and baseline, respectively.

In this study, cognitive screening was done using the Norwegian Version 7.1 of the Montreal Cognitive Assessment (MoCA; Supplemental Fig. S1A, Supplemental Digital Content, https://links.lww.com/MSS/D332). MoCA was originally designed to detect MCI in older adults (27); however, its utility as a monitoring tool for general cognition has been validated in similar populations (28). The MoCA consists of different tasks, designed to evaluate various aspects of cognition, including short-term memory, working memory, visuospatial abilities, executive functions, attention, concentration, language fluency, and orientation to time and place (27). Altogether, the tasks add up to a maximum score of 30. Higher scores indicate better cognitive functioning (27). The MoCA was administered at the 3-, 5-, and 10-year follow-ups for all interested participants.

Statistical Analysis

Descriptive baseline statistics are presented as mean (standard deviation [SD]) for continuous variables and frequency (percentages) for categorical variables. Student’s t tests or one-way analysis of variance tests were used to assess differences in VO_2peak between different strata at baseline. Changes in VO_2peak were assessed using paired sample t tests. Before testing, the normality of data was assessed visually using Q–Q plots and with the addition of a Shapiro–Wilk’s test when visual inspection was insufficient. The Mann–Whitney U tests were used for data not following the normal distribution. Similarly, Levene’s test was used to assess homoscedasticity and Welch’s t tests were used for comparisons with unequal variances. The results from these analyses are reported as mean difference and 95% confidence interval (CI).

Descriptive statistics for the dependent variable, MoCA score, were presented as mean (SD). To see how MoCA scores changed from years 3 to 5, years 5 to 10, and years 3 to 10, paired sample t tests were used using complete case analyses. Results were reported with difference in means (SE_diff), degrees of freedom (df), SD, and 95% CI. All assumptions were met.

Linear mixed models (LLMs) were used to estimate the association between baseline VO_2peak, baseline-to-year-1 change, and baseline-to-year-3 change in VO_2peak, and MoCA score measured at 3 timepoints (3-, 5-, and 10-year follow-ups). LLM is a statistical method to analyze longitudinal data that includes both fixed effects (which refer to population-level effect considered to be constant across all individuals, such as time) and random effects (which account for variability between individuals) (29). LLM is equipped to handle missing longitudinal data for the normally distributed continuous outcome variables (29). Thus, all participants (n = 770) could be included regardless of whether they had a MoCA score at all timepoints.

Specifically, Model 1 examined baseline VO_2peak’s association with MoCA score. Participants were treated as random effects, while timepoint (categorical: years 3, 5, and 10) and baseline VO_2peak as fixed effects. Interaction terms between timepoint and baseline VO_2peak were included as a main covariate to assess whether the association between baseline VO_2peak and MoCA score varied over time. All analyses were also adjusted for baseline age, gender, education level, and self-rated memory. To facilitate interpretation, the model was repeated three times, each time changing the reference timepoint.

Model 2 assessed the impact of changes in VO_2peak on MoCA score, using a similar structure to Model 1. Here, 1-year change and 3-year change in VO_2peak replaced baseline VO_2peak as the main predictors. For each participant, changes in VO_2peak were determined by subtracting the baseline value from the VO_2peak measured at each follow-up. The fixed effects included timepoint, VO_2peak change, and an interaction term between timepoint and VO_2peak change. Model 2 was repeated six times; once for each combination of timepoint (years 3, 5, and 10) and duration of VO_2peak change (1 and 3 years). Similarly, all analyses were adjusted for baseline age, gender, education level, and self-rated memory.

Results were reported as estimated coefficients (β) and 95% CI. Normality of residuals was checked by visual inspection of Q–Q plots. All analyses were performed using IBM SPSS Statistics, Version 29.0.1.0 (SPSS Inc., Chicago, IL).

RESULTS

This study included 770 participants (48% female) with a mean (SD) age of 72.0 (2.0) years at baseline. The descriptive statistics for participants’ VO_2peak at baseline, stratified by relevant characteristics, are presented in Table 1. At baseline, mean (SD) VO_2peak for all participants was 30.3 (6.5) mL/kg/min. On average, men had 18% higher VO_2peak at baseline than women (Table 1). Additionally, having a tertiary education and poor self-rated memory at baseline was associated with a higher VO_2peak at baseline (Table 1). Similar descriptions for VO_2peak at year 1 and year 3 are provided in Appendix B (Supplemental Table S1B, Supplemental Digital Content, https://links.lww.com/MSS/D332, and Supplemental Table S2B, Supplemental Digital Content, https://links.lww.com/MSS/D333).

TABLE 1.

Descriptive statistics for participants’ characteristics and VO_2peak at baseline.

	n (%)a	Baseline VO2peak (mL/kg/min)b
All participants	770 (100)	30.3 (6.5)
Gender
Man	403 (52)	32.9 (6.5)*
Woman	367 (48)	27.3 (5.0)	Ref.
Age (years)	72.0 (2.0)
Education level (years)
Primary (≤10)	54 (7)	29.5 (6.2)	Ref
Secondary (11–13)	281 (37)	29.47 (5.6)*
Tertiary (≥14)	435 (56)	30.9 (6.9)*
Memory self-assessment
Poor	193 (25)	31.0 (6.4)*
Good	577 (75)	30.0 (6.5)	Ref.

^aData presented as count (percentage).

^bData presented as mean (standard deviation).

^*P < 0.05, compared with the reference group (ref.) within each category.

One-Year Change and 3-Year Change in VO_2peak

Our study found that 1 year after baseline, the mean VO_2peak increased by 1.8 mL/kg/min (95% CI: 1.5–2.0) (Fig. 2). However, after three years, the mean VO_2peak returned to the baseline level (mean difference: 0.2 mL/kg/min; 95% CI: −0.5 to 0.1) (Fig. 2).

FIGURE 2.

Trend in the mean VO_2peak over 3 years. Data are means with standard deviation. 0 = baseline. *P < 0.001, compared with the baseline.

MoCA Scores over 7 Years

Among 770 participants, 750 had MoCA score at year 3, 665 at year 5, and 454 at year 10. The mean (SD) MoCA score 3-, 5-, and 10 years after baseline was 25.3 (3.0), 25.0 (3.0), and 24.8 (3.2), respectively.

On average, MoCA scores decreased by 0.4 points (SD: 2.5, df = 649, 95% CI: −0.6 to −0.2) from year 3 to year 5; 0.7 points (SD: 2.8, df = 438, 95% CI: −0.9 to −0.4) from year 5 to year 10; and 0.9 points (SD: 2.9, P < 0.001, 95% CI: −1.2 to −0.6) from year 3 to year 10.

The Association between Baseline VO_2peak and MoCA Scores in Later Years

LLM from Model 1 revealed that each unit (mL/kg/min) higher in VO_2peak at baseline predicted an additional 0.06 points in MoCA score at year 3 (95% CI: 0.02–0.09) and 0.07 points in MoCA score at year 5 (95% CI: 0.03–0.11). However, VO_2peak at baseline was not associated with MoCA score at year 10 (β: 0.02, 95% CI: −0.02 to 0.06) (Table 2).

TABLE 2.

Association between baseline VO_2peak and MoCA measured 3, 5, and 10 years after baseline.

	MoCA Score
	β	95% CI
Baseline VO2peak
Year 3	0.06	0.02–0.09
Year 5	0.07	0.03–0.11
Year 10	0.02	−0.02 to 0.06

The Association between Changes in VO_2peak and MoCA Scores in Later Years

LLM from Model 2 revealed that the 1-year change in VO_2peak was not associated with MoCA scores at any later time points (year 3: β: 0.01, 95% CI: −0.05 to 0.07; year 5: β: 0.01; 95% CI: −0.05 to 0.08; year 10: β: 0.02, 95% CI: −0.02 to 0.06) (Table 3).

TABLE 3.

Association between 1 year in VO_2peak and MoCA measured at 3, 5, and 10 years after baseline.

	MoCA Score
	β	95% CI
1-year change in VO2peak
Year 3	0.01	−0.05 to 0.07
Year 5	0.01	−0.05 to 0.08
Year 10	0.02	−0.02 to 0.06

Similarly, the 3-year change in VO_2peak was also not associated with MoCA scores at year 3 (β: 0.03, 95% CI: −0.03 to 0.08), year 5 (β: 0.00, 95% CI: −0.06 to 0.05) or year 10 (β: 0.02, 95% CI: −0.04 to 0.08) (Table 4).

TABLE 4.

Association between 1 year in VO_2peak and MoCA measured at 3, 5, and 10 years after baseline.

	MoCA Score
	β	95% CI
3-year change in VO2peak
Year 3	0.03	−0.03 to 0.08
Year 5	0.00	−0.06 to 0.05
Year 10	0.02	−0.04 to 0.08

DISCUSSIONS

In the present study, a higher baseline CRF was associated with better cognition up to 3 and 5 years later but not at the 10-year follow-up. There was no association between changes in CRF over 1 and 3 years and cognitive function in later years.

Our finding of an association between CRF at baseline and cognition up to 3 and 5 years later supports the notion that CRF is a reliable predictor for cognitive health in older adults (30,31). A prior substudy by Sokołowski et al. (10) using data from the Generation 100 Study found a similar association between baseline CRF and performance on selective cognitive tasks. In that study, various cognitive domains (i.e., memory, attention, and executive functions) were assessed by a self-administered web-based neuropsychological test platform called Memoro (10). Our finding also suggests that controlling for age, gender, educational levels, and memory status, participants who achieved a higher baseline CRF might be able to maintain certain neurocognitive advantages compared to their less fit counterparts. Although statistically significant, the predicted changes in MoCA scores were small (approximately 0.06–0.07 points per year) and therefore unlikely to represent clinically meaningful cognitive change. This interpretation aligns with prior literature indicating that such small changes in MoCA scores reflect cognitive stability rather than improvement (32). Clinically, it is noteworthy that participants with higher CRF at baseline maintained relatively stable MoCA scores over time, contrary to the expected age-related decline in this age group of around one point over a 10-year period, as typically observed among cognitively healthy older adults aged 70–80 years (32).

Similar to our study, Hayes et al. found that older adults with directly measured VO_2peak in the higher quartiles (mean age: 63.3 years, SD: 5.9; mean VO_2peak: 36.0 mL/kg/min, SD: 5.4) had better executive functioning (P < 0.05), episodic memory (P = 0.052), and visual memory (P < 0.001) compared with those in the lower quartiles (mean age: 64.9 years, SD: 8.4; mean VO_2peak: 23.9 mL/kg/min, SD: 6.4) (30). Other observational studies find a similar association between baseline CRF and cognition (6–8,33).

Interestingly, our study found that this association, while it persisted for up to 5 years, did not extend to 10 years after the baseline. It is plausible that a higher baseline CRF might confer neuroprotective effects for these older adults; however, it alone might not be sufficient to mitigate cognitive decline typically observed in their eighth or ninth decade of life. Certain declines in both general cognition and selective cognitive processes are to be expected with increasing age (2). A few other observational studies that include participants in similar or older age groups and with similar or longer follow-up durations have reported an association between baseline CRF and cognitive outcomes later in life (34–36). However, these studies differ from our study in important aspects, making direct comparison challenging. Specifically, they may include a broader age range (35), rely on estimated CRF (37), or report significant association only with selective cognitive outcomes, but not with general cognitive functions assessed by a tool such as the Mini-Mental State Examination (35), which is a cognitive assessment tool comparable to MoCA (37). Altogether, our study extends the current literature by suggesting a potential time-limited protective role of CRF on cognition in older adults in their 70s, highlighting the need for sustained or later-life interventions to preserve cognitive health into the 8th or 9th decades of life.

However, whether CRF improvement in this period of life can counteract the effect of aging on cognition warrants further investigation. Our study did not find any association between changes in CRF over 1 and 3 years and cognition at any subsequent time points. This finding aligns with other studies that also found a lack of association between CRF improvements on cognitive functioning in older adults (18,19,38). It could be that improving CRF levels earlier in life and carrying a high CRF into old age might be more beneficial for cognitive aging than the improvement in CRF itself in this period. A longitudinal prospective cohort study (39) following 2747 participants aged 18–30 at baseline found that a higher baseline CRF predicted better performance in cognitive tests assessing memory (standard error: 0.03, P < 0.0001) and psychomotor functions (standard error: 0.13, P < 0.0001) measured 25 years later. On the other hand, our study’s inclusion of cognitively intact older adults with a mean baseline VO_2peak value higher than those previously reported for similar age groups in the United States (40) (men: mean VO_2peak: 21.2 mL/kg/min, SD: 6.5; women: mean VO_2peak 17.5 mL/kg/min, SD: 4.2) and Korea (41) (men: mean VO_2peak: 27.2 mL/kg/min, SD: 5.6; women: mean VO_2peak: 23.9 mL/kg/min, SD: 4.4) may have masked the effect of CRF changes on cognition in this population. Additionally, the increase in mean VO_2peak of our participants after 1 year (5.9%) was smaller compared with previous interventions that found an association between CRF improvement and enhanced cognitive functions among sedentary older adults (12,42). For example, a CRF increase of 27% following 4 months of aerobic exercise training in older adults aged 55–70 (12) or 24.3% following 6 months of HIIT in older adults aged between 65 and 75 (42) have been previously reported. It is possible that for already fit older individuals, a much higher increase in CRF might be necessary to achieve cognition-modifying effects. Current evidence in the literature regarding the effect of fitness improvement and exercise training on cognitive health in older adults remains inconsistent (19,43).

We concur that it is important to consider not only the magnitude of change in CRF but also the duration over which that change occurs, and how they relate to cognitive outcomes. Another prior Generation 100 substudy by Zotcheva’s (n = 945) (9) revealed that an increase of approximately 3.5 mL/kg/min in VO_2peak from baseline to year 5 was associated with an additional 0.46 points on MoCA at year 5 (beta value: 0.46, 95% CI: 0.25–0.67). The substudy by Sokolowski (n = 106) found that a higher CRF during intervention period was associated with better processing speed (β: 0.02, 95% CI: 0.01–0.04) and a positive change in CRF from baseline to year 5 was associated with better working memory (β: 0.03, 95% CI: 0.00–0.05) but not the other cognitive abilities (i.e., spatial memory, verbal memory, pattern separation, processing speed, and planning ability) (10). Together with these findings, we consider it a possibility that in healthy older adults aged 70 or above, examining changes in CRF over a period longer than 3 years may be necessary to better understand the impact of CRF improvement on cognitive trajectories. However, interpretations across substudies within the same study population should be made with caution due to variations in study designs (observational vs RCTs (9,10)), inclusion criteria, and statistical approaches. Currently, an optimal timeframe for CRF monitoring in relation to cognitive health outcome, as well as the threshold of change required to impact cognition, has not yet been established, considering the inconsistencies in current literature.

Strengths and Limitations

The results from our study should also be interpreted with consideration for several limitations. First, participants who joined the Generation 100 Study were considerably healthier, fitter, more active, and more educated compared to those who did not (17), which may limit the study’s generalizability. We further excluded participants with missing CRF data and participants with diagnosed dementia, potentially undermining the neuroprotective effects of CRF in people with neurodegenerative disorders (5). Second, the statistical power in the main study was calculated to detect a difference in mortality rate; as a result, the findings from our present study might be underpowered for detecting associations between CRF and cognition, and causality cannot be inferred due to the observational design. Finally, the exercise training intervention was not considered in our analyses despite the randomized nature of the Generation 100 Study. This might have affected the observed relationship between baseline VO_2peak or VO_2peak change and cognition since exercise training, especially HIIT, is known to increase CRF (17). This is, however, unlikely since the other two substudies using data from the Generation 100 Study did not find an effect of the exercise training intervention (HIIT or MICT or both combined) on cognition (9,10) despite the significantly higher mean CRF in the HIIT group compared to the control group (mean difference: 0.7 mL/kg/min, P = 0.02) and MICT group (mean difference: 0.7 mL/kg/min, P = 0.04) at the end of the exercise training intervention (17).

Our study also had several notable strengths. First, longitudinal VO_2peak data were measured directly, proving an accurate presentation of our participants’ CRF level (44). Second, our study assessed MoCA at 3 time points with different intervals between each test, which has been suggested to eliminate the practice effect commonly found in tests and retests administered less than 12 months apart (45). Altogether, these strengths helped improve the robustness and credibility of our findings.

CONCLUSIONS

Our study found that higher baseline CRF was associated with better cognition 3 and 5 years later but not at the 10-year follow-up. In contrast, changes in CRF over 1 and 3 years had no association with cognition in subsequent years. This suggests that maintaining high CRF levels earlier in life and sustaining them into older age could provide some neuroprotective benefits. However, these benefits may not be sufficient to mitigate the impact of aging on cognition. Additionally, we cautiously recommend that future studies investigating the relationship between CRF changes and cognition in healthy older adults should consider monitoring periods longer than 3 years to capture potential long-term associations.

This work was supported by the Research Council of Norway, the Norwegian University of Science and Technology (NTNU), the Liaison Committee between the Central Norway Regional Health Authority and NTNU, the Central Norway Regional Health Authority, St Olavs Hospital, Trondheim, Norway, and the DAM Foundation via the Norwegian Brain Council.

The authors thank all participants for their invaluable contribution to this study. The authors also appreciate the excellent assistance from the Clinical Research Facility at St. Olavs Hospital and the NeXt Move Core Facility at NTNU. Finally, the authors acknowledge the dedicated efforts of employees and students from their research group during data collection in Generation 100 Study. No conflicts of interest were disclosed. A. N. L. carried out the literature search, performed data analysis and interpretation, and primarily drafted and revised the manuscript; S. L.: Conceptualization, supervision, literature search, data analysis and interpretation; D. E. B. collected data, supervised the study, conducted the literature search, participated in data analysis and interpretation, and revised and refined parts of the manuscript; T. L. W. supervised the work and critically reviewed the manuscript, M. S. supervised the work and critically reviewed the manuscript, A. H. K. conceptualized the study, supervised the work and critically reviewed the manuscript; H. H. B. reviewed hospital records, conducted the literature search, participated in data analysis and interpretation, and critically reviewed the manuscript; D. S. conceptualized the study, supervised the work, conducted the literature search, participated in data analysis and interpretation, critically reviewed the manuscript, and obtained funding; A. R. T. conceptualized the study, supervised the work, conducted the literature search, participated in data analysis and interpretation, critically reviewed the manuscript, and obtained funding. All authors have read and approved the final version of the manuscript and agree with the order of presentation of the authors. The corresponding author attests that all listed authors meet authorship criteria and that no others meeting the criteria have been omitted. The results of the study are presented clearly, honestly, and without fabrication, falsification, or inappropriate data manipulation. The results of the present study do not constitute endorsement by the American College of Sports Medicine.