Age-dependent associations between physical activity and cognitive function in adults aged 70–100: A cross-sectional and longitudinal study

Abstract

Background

While the protective effects of physical activity (PA) on cognitive function and risk for cognitive decline are well documented, it remains unclear if the preventive effect continues at advancing age.

Objective

This study aims to examine the association between PA and cognitive impairment between ages 70–100

Methods

The Jerusalem Cross-sectional and Longitudinal Study (1990–2023) prospectively follows a representative community-dwelling cohort born 1920–1921, combining follow-up of the original cohort with repeated recruitment of same–birth cohort participants at each age. During the study, assessments were performed at ages 70, 78, 85, 90, 95, and 100 (n = 604, 1024, 1222, 729, 508, 205, respectively). Comprehensive assessment included Mini-Mental State Examination (MMSE) defining cognitive impairment (≤23/30), and self-reported PA, dichotomized to Sedentary (PA-S) (<4 h weekly) versus Active (PA-A) (at least >4 h a week).

Results

At ages 70, 78, 85, 90, 95, and 100 the frequency of PA-A was 53.5%, 77.0%, and declined to 64.3%, 16.0%, 25.5%, and 27.1%, respectively. In a longitudinal analysis, PA-S at baseline was found to have a statistically significant association with cognitive impairment at a later age between ages 78–85, 85–90: OR 2.61 (95%CI 1.38–4.82); OR 2.90 (95%CI 1.63- 5.11), respectively. At ages 90–95 a similar not statistically significant effect was found.

Conclusions

PA is associated with better cognitive aging and lower risk for cognitive impairment in advanced age. Inclusion of very old adults extends evidence into an age rarely examined and demonstrates that promotion of PA remains relevant even at advanced ages.

Introduction

Cognitive impairment is a major driver of disability, loss of independence and mortality in older adults.^1,2 Prevention and risk reduction through lifestyle factors are central to public health strategies. Physical inactivity has been consistently identified as a modifiable risk factor for multiple adverse outcomes in older adults, including cognitive decline. Systematic evidence demonstrates that physically active older adults have lower risks of cognitive impairment, dementia, functional disability, falls, cardiovascular disease, and mortality compared with inactive counterparts.^3,4 International guidelines emphasize lifestyle-based interventions, particularly regular physical activity, as a cornerstone of strategies to support cognitive health in older age. ⁵

This study focuses on the association between physical activity and cognitive status, specifically the risk for cognitive impairment, across different ages. Importantly, the very old population (aged 85 and above) are often underrepresented in cohort studies, leaving uncertainty whether associations observed in younger elderly populations generalize to the oldest-old. Some prospective studies specifically targeting very old adults have examined physical activity in relation to incident mild cognitive impairment or dementia, demonstrating feasibility but also highlighting that evidence in this age range remains comparatively sparse.^6,7 With the growing oldest-old population at high risk of cognitive impairment, this study examines whether the association between physical activity and cognitive status varies with age.

Methods

Basic methodology and sampling

The Jerusalem cross-sectional and Longitudinal Study on aging was conducted between the years 1990 to 2021 and included participants aged 70 to 100 who were born between June 1920 to May 1921 and reside in Jerusalem. Participants were randomly selected from the electoral registry at the age of 70 and were conducted under the auspices of the Institute for Aging Research at Hadassah University Medical Center, Mount Scopus, Jerusalem. The study is described in detail elsewhere.^8,9

In brief, the primary objective of the study was to investigate aging trajectories and identify risk factors associated with the aging process in this population. To address substantial attrition inherent to very old populations, the study combined a longitudinal cohort design with repeated recruitment from the same birth cohort. The original cohort included individuals born in 1920–1921 who were first assessed at age 70 (1990–1991) and followed longitudinally at ages 78 (1997–1998), 85 (2005–2006), 90 (2010–2011), 95 (2015–2016), and 100 (2020–2021). To maintain an adequate sample size and representativeness at older ages, additional participants from the same underlying source population (Jerusalem residents born in 1920–1921) were recruited at each assessment wave. Thus, each wave comprised both surviving participants from earlier assessments and newly enrolled individuals from the same birth cohort. This design enabled longitudinal follow-up within individuals while also preserving cross-sectional representativeness at each age. To evaluate potential selection bias introduced by ongoing recruitment, key indicators, including survival rates and hospitalization rates, were compared between study participants, listed birth cohort individuals and those who refused to participate. These analyses demonstrated comparable outcome patterns, supporting the representativeness of the sample and the assumption that combining incident recruits with previously followed participants did not materially bias the results.

Comprehensive interviews and home assessments were conducted with participants at ages 70, 78, 85, 90, 95, and 100 (number of participants at each age was 604, 1024, 1222, 729, 508, and 205, respectively). Of the total sample at each follow-up wave, the number of participants continuing from a prior assessment was 374 at age 78, 569 at age 85, 507 at age 90, 125 at age 95, and 56 at age 100. The variation in sample size across age waves reflects both substantial attrition inherent to very old populations, primarily due to mortality, and the study design, which incorporated cross-sectional and longitudinal study to attempt in replenishing the participants in each wave but was partial and did not fully offset losses over time. Moreover, the number of individuals available for recruitment from the underlying birth cohort decreased at older ages. Consequently, sample sizes varied across waves and declined at more advanced ages. The interviews were carried out by trained personnel with experience in geriatrics and included a social questionnaire, functional status, performance in daily activities [activities of daily living (ADL) and instrumental ADL], self-assessment of health status, and a medical questionnaire. The medical questionnaire included a physical examination, past and present diagnoses, physical assessments, medication usage, depression evaluations, and cognitive assessment (Mini-Mental State Examination, MMSE). In cases in which individuals were cognitively unable to respond to interviewees’ questions, responses were obtained with the assistance of a legal guardian.

Measurements and data collection

The primary independent study variable of interest in this study was physical activity. This was a self-reported measurement and assessed according to participants response to the question “how often are you physically active?” and answers being (a) less than 4 h weekly, (b) 4 h weekly, (c) vigorous sports at least twice weekly (e.g., jogging, swimming), and (d) regular physical activity (e.g., walking at least an hour daily). This 4-item question was adapted from the Gothenburg study population of 70-year-olds. ¹⁰ Physical activity was dichotomized to sedentary (answer a) versus physically active (answers b, c, and d). This cutoff is in accordance with the world health organization's guidelines ¹¹ and was justified statistically, accounting for distribution and frequency of responses. ¹²

Additional health variables

As part of the Jerusalem Longitudinal Study, a broad range of variables was collected, as previously described in detail elsewhere. ⁸ Variables relevant to the present analyses included ADL, self-rated health (SRH), and depression.

ADL was assessed using six domains: transferring, mobility, dressing, bathing, continence, toileting, and eating. ¹³ Participants were asked about their ability to perform each activity, and dependence in ADL was defined as dependence in at least one of the six domains. SRH was evaluated using the question, “How do you rate your state of health?”, with responses categorized dichotomously as good or poor. Depression was assessed using the Brief Symptom Inventory. ¹⁴ Participants rated, on a scale from 0 to 4 (0 = none, 4 = greatly), the extent to which during the previous month they experienced: (a) loneliness, (b) lack of interest, (c) thoughts of ending life, (d) depressed mood, (e) hopelessness about the future, and (f) feelings of worthlessness. Depression was defined according to established criteria ¹⁵ as a total score of ≥6.

Outcomes

Cognitive status. Cognitive status was measured using the MMSE ¹⁶ and examined as a continuous variable from 0 to 30. participants were classified as cognitive functioning categories based on their MMSE score. Normal functioning was defined if MMSE was 29 and above, mild cognitive impairment (MCI) for those with MMSE between 24 and 28, and cognitive impairment for those scoring 23 and below.

Mortality. Death was the primary outcome obtained from annual review of death, which is 100% certificated and issued by the Ministry of Interior.

Deterioration in cognitive status. Deterioration over time in cognitive status was measured among each two consecutive study populations, i.e., participants who participated at both age 70 and age 78 (n = 374), participants at age 78 and 85 (n = 569), 85 and 90 (n = 507), 90 and 95 (n = 125) and 95 and 100 (n = 56). Cognitive impairment deterioration was defined as the new onset of functioning cognitive impairment (MMSE≤23) at follow-up among participants with MMSE>23 at baseline.

Statistical analysis

Baseline characteristics are presented among active and sedentary participants, each age. A comparison is performed using chi-square test. This analysis was repeated for the cross-sectional population at each age, and an additional sensitivity analysis was performed for participants appearing in at least two consecutive cohorts, named the longitudinal sample. MMSE scores are presented as mean, SD, Median and interquartile range (IQR) at each age. A comparison between those active and sedentary was performed using rank sum Wilcoxon test. In addition, the association between physical activity and cognitive impairment was tested at each age using two-sample z-test for proportions (normal approximation).

For participants appearing in at least two consecutive cohorts, named the longitudinal sample, the association between physical activity and new onset cognitive impairment (among those without cognitive impairment at a younger age) was tested for each two consecutive cohorts, using two-sample z-test for proportions (normal approximation).

Logistic regression was performed to test the association between physical activity at a younger age and cognitive impairment at an older age, adjusting for cognitive impairment at the younger age. In addition, when possible, an adjusted model was performed, adjusted for gender, education, depression and ADL. As a sensitivity analysis, similar logistic regression models were repeated, excluding participants already suffering from cognitive impairment at the younger age.

Statistical analyses were performed using R 4.4.2. All p-values were two tailed, and p < 0.05 was considered significant.

Results

Baseline characteristics

The proportion of participants who self-reported being physically active varied across ages, increasing from 53.5% at age 70 to 77.0% at age 78, and subsequently declining to 64.3%, 16.0%, 25.5%, and 27.1% at ages 85, 90, 95, and 100, respectively.

As shown in Table 1, which is a cross-sectional analysis, physically active participants consistently had higher education and reported fewer financial difficulties compared with sedentary participants across all age groups. Higher physical activity rates are observed among men at ages 78, 90, and 95. Active participants also reported significantly better SRH, lower rates of depression, and greater independence in ADL. Similar results were found when analyzing baseline characteristics of the longitudinal sample.

Table 1.

Baseline characteristics by physical activity status.

	Age 70		Age 78		Age 85		Age 90		Age 95		Age 100
	sedentary	active	sedentary	active	sedentary	active	sedentary	active	sedentary	active	sedentary	active
All	212	244	203	679	410	738	551	105	369	126	145	54
Gender
Female	48.6%(103)	42.2%(103)	71.9%(146)	44.2%(300)	63.9%(262)	49.6%(366)	58.6%(323)	49.5%(52)	63.4%(234)	46.8%(59)	68.8% (99)	61.1% (33)
Male	51.4%(109)	57.8%(141)	28.1%(57)	55.8%(379)***	36.1%(148)	50.4%(372)***	41.4%(228)	50.5%(53)	36.6%(135)	53.2%(67)**	31.2% (45)	38.9% (21)
Married	67%(140)	71.8%(173)	47.7%(93)	63.5% (422)***	37.3% (152)	46.3% (341)*	29.2% (159)	35.2%(37)	70.8% (243)	59%(69)*	15.2% (22)	14.8% (8)
High Education	50.5%(107)	67.2%(164)**	40.4%(82)	62.3% (423)***	44.9% (184)	65% (480)***	59.2%(326)	84.8%(89)***	48.2%(178)	67.5%(85)**	54.5% (79)	57.4% (31)
Money Problem	24.6%(51)	19.1%(46)	32.6%(56)	13.1% (83)***	28.6% (113)	12.1% (89)***	16.3%(88)	1%(1)**	12%(41)	2.4%(3)*	5% (7)	0% (0)
Poor SRH	45.5% (95)	13.2% (32)***	71.3% (139)	34.6% (229)***	53.8% (200)	25.9% (189)***	49.8% (255)	15.4%(16)***	34.8% (117)	15.1%(18)***	29.3% (29)	9.6% (5)*
Depression	22.7% (42)	10.3% (23)*	37%(34)	18.4% (96)	52.6% (192)	28.3% (206)***	43%(192)	14.6%(15)**	23.9% (68)	22.1%(25)	38% (35)	31.2% (15)
Problem in ADL Katz	9.2%(19)	1.7%(4)**	36%(67)	4.4%(28)***	72.8% (287)	18.7% (136)***	62.5% (208)	13.1%(8)***	56.6% (193)	16.1%(20)***	86.9% (126)	37% (20)***
Hypertension	49.1% (104)	30.3% (74)**	65%(132)	51.1% (347)**	72.9% (298)	71.1% (525)	76.3% (415)	59.2%(61)**	80.2% (275)	73.6%(89)
Ischemic heart disease	28.8% (61)	23.4% (57)	38.4%(78)	33%(224)	37.7% (154)	37.1% (274)	26%(143)	27.6%(29)	39.9% (137)	36.4%(44)	33.8% (49)	35.2% (19)
T2D	13.7% (29)	12.3% (30)	21.2%(43)	16.1% (109)	26.6% (109)	16.7% (123)**	17.5%(59)	11.8%(11)	18.2% (61)	8.9%(11)*	19.3% (28)	18.5% (10)

*p < 0.05;**p < 0.001;***p < 0.0001.

SRH: self-rated health; ADL: activities of daily living; T2D: type 2 diabetes.

MMSE in a cross-sectional analysis

In a cross-sectional analysis, at ages 70 and 78, physical activity was not associated with higher cognitive performance, as measured by MMSE. For both ages the mean MMSE was high (28 and above) for both active and sedentary participants (Table 2). On the contrary, at ages 85 and above, this association was statistically significant with Mean (SD) MMSE scores among active versus sedentary participants of 27.7 (2.9) versus 24.5 (5.1) at age 85, 27.9 (2.3) versus 22.6 (9.3) at age 90, 26.4 (2.9) versus 24.2 (5.5) at age 95, and 23.6 (6.8) versus 17.0 (9.1) at age 100. Similarly, among physically active participants there were significantly lower rates of cognitive impairment, defined as MMSE≤23 (Figure 1), at all ages besides 70 (p < 0.0001).

Table 2.

Mean (SD) of Mini-Mental State Examination scores, by physical activity status.

	Physical activity	N	Mean (SD)	Median (IQR)	p
Age 70	Sedentary	87	28.8 (2.7)	30 (29, 30)	0.1562
	Active	138	29.5 (0.9)	30 (29, 30)
Age 78	Sedentary	48	28 (2.9)	29 (27.5, 30)	0.1334
	Active	323	28.9 (1.4)	29 (28, 30)
Age 85	Sedentary	225	24.5 (5.1)	26 (21.1, 28.9)	<0.0001
	Active	555	27.7 (2.9)	28.9 (27, 30)
Age 90	Sedentary	383	22.6 (9.3)	27 (21, 29)	<0.0001
	Active	87	27.9 (2.3)	28 (27, 29)
Age 95	Sedentary	295	24.2 (5.5)	26 (22, 28)	0.0026
	Active	116	26.4 (2.9)	27 (25, 28)
Age 100	Sedentary	145	17.0 (9.1)	19.5 (11, 25)	<0.0001
	Active	54	23.6 (6.8)	26 (22.7, 28)

Figure 1.

Rate of cognitive impairment, by physical activity status.

Physical activity and cognitive impairment in a longitudinal analysis

The protective effect of physical activity on cognitive impairment was analyzed on a longitudinal sample, among participants who were examined in two consecutive studies. Physical activity at a younger age was significantly associated with less new onset of cognitive impairment at an older age for ages 78 to 85 (37.2% cognitive impairment among sedentary versus 15.6% among active, p = 0.008) and 85 to 90 (53.2% cognitive impairment among sedentary versus 21.8% among active, p = 0.001). Similar, but non-significant, associations were found at age 90 to 95, p = 0.079 (Figure 2).

Figure 2.

New onset of cognitive impairment, by physical activity status *New onset of cognitive impairment among participants without cognitive impairment at a younger age.

The effect of physical activity on cognitive impairment was further analyzed in a multivariate analysis. Being sedentary was found to be associated with higher rates of cognitive impairment in an older age, after adjusting for cognitive status at the younger age. This was statistically significant for ages 78 to 85 [odds ratio of 2.61 95%CI = (1.38–4.82)], at ages 85 to 90 [odds ratio of 2.90 95%CI = (1.63- 5.11)]. After adjusting for additional relevant risk factors, gender, education, depression and ADL, these results remain statistically significant (Figure 3). This association was not statistically significant at ages 90 to 95 (p = 0.121).

Figure 3.

Logistic regression for the effect of physical activity at a younger age on future cognitive impairment. *“Unadjusted” model 1 was adjusted only for cognitive impairment at a younger age; **“Adjusted” model was adjusted for cognitive impairment at a younger age, as well as gender, education, depression and ADL.

As a sensitivity analysis, regression models adjusted for gender, education, depression and ADL were repeated only among those who did not suffer from cognitive impairment at a younger age, to test the protective effects of physical activity on new onsets of cognitive impairment. The results show similar trends but in many cases the sample size of this subdivision as well as the number of events in each model were very small and the results were not statistically significant.

Discussion

This population-based study extends existing evidence by demonstrating that the association between physical activity and cognitive outcomes persists from age 70 to 100 years, an age range largely under-represented in previous research. Our results are consistent with prior large observational studies and meta-analyses showing that higher levels of physical activity are associated with lower risk of cognitive decline and dementia, particularly in younger-old populations.^3,4,17 Proposed biological mechanisms include improved cerebral perfusion, enhanced neuroplasticity and neurogenesis, reduced systemic inflammation, and improved cardiometabolic health, all of which are closely linked to cognitive aging.^18–20 In addition, physical activity may indirectly support cognitive function by preserving functional independence, reducing depressive symptoms, and improving overall health status, pathways that are themselves strongly associated with cognitive outcomes in later life. ²¹ In the present study, physically active participants consistently reported better self-rated health, lower rates of depression, and greater independence in activities of daily living, supporting the plausibility of both direct and indirect protective mechanisms. However, the association between physical activity and lower risk of cognitive impairment remained significant after adjustment for depression and ADL dependency, suggesting that the observed relationship is not solely mediated by psychological well-being or functional status.

In addition to its association with cognitive outcomes, Physical activity has also been consistently associated with reduced mortality in older adults. In the Jerusalem Longitudinal Study, higher physical activity levels were linked to lower all-cause mortality independent of baseline health status and other confounders, with survival benefits observed even among individuals who became active at advanced old age.^22,23

This study demonstrated that the effect of physical activity on cognitive state is likely modified by age. Specifically, in the longitudinal analysis, physical activity was not associated with less cognitive impairment from age 70 to 78, likely reflecting ceiling effects of the MMSE score and relatively low prevalence of cognitive impairment at these ages. Similar null findings in younger elderly populations have been reported in prior studies where baseline cognition is high, and follow-up periods are relatively short. ²⁴ On the contrary, at the following stages of aging an association was found between physical activity and reduced risk for cognitive impairment at subsequent examinations, particularly between ages 78 to 85 and 85 to 90, even after adjustment for baseline cognitive status and other relevant risk factors. This effect was less pronounced among the very old (age 95 to 100). Evidence from studies specifically targeting very old populations suggests that physical activity may help preserve cognitive reserve or delay the clinical expression of neurodegenerative pathology at advanced ages.^6,7 This effect may be particularly important between ages 85 to 95, when cognitive heterogeneity increases and dementia prevalence rises, physical activity emerges as a strong differentiating factor. The inclusion of adults aged 95 and 100 in this study allows insight regarding the very old adults, who are often underrepresented in epidemiological studies, leading to uncertainty regarding whether risk factors identified in younger elderly populations apply to the oldest-old. ²⁵ In this cross-sectional study, the association between physical activity and one's cognitive state was strong both at ages 95 and 100 but there was no evidence of a protective effect of physical activity at age 95 to cognitive impairment at age 100 in the longitudinal analysis.

Several limitations should be considered when interpreting these findings. First, Cognitive status and risk for cognitive impairment in this study were assessed using MMSE alone. Although widely used, MMSE has recognized limitations, including limited sensitivity to subtle cognitive changes, which may result in misclassification or underestimation of early cognitive impairment. Second, physical activity was assessed by self-reported response and is therefore subject to recall and social desirability bias. Notably, individuals with cognitive impairment may have reduced reliability of self-report; however, in such cases, responses were obtained with the assistance of a legal guardian, which may have reduced this source of bias. Similar limitations are relevant for other self-reported measurements, such as ADL.

Finally, as is common in studies of very old adults, survivorship bias cannot be excluded. Participants who survived to the oldest ages may represent a healthier subset of the original population. Nevertheless, the mixed longitudinal design, with repeated recruitment from the same birth cohort at each assessment wave, was specifically implemented to preserve representativeness across ages, and previous analyses demonstrated comparable mortality and hospitalization patterns across recruitment groups. In addition, the use of well-established measures such as the MMSE, self-reported physical activity, and ADL facilitate comparison with prior large-scale epidemiological studies of aging.

To appropriately plan interventions and preventive strategies, it is important to understand when during the adult life course physical activity is most associated with cognitive impairment risk. By demonstrating consistent cross-sectional and longitudinal associations between physical activity and cognitive outcomes at ages 85 to 90 and beyond, this study extends existing evidence into an age range where data remain sparse.

From a public health perspective, these findings align with and strengthen international recommendations that emphasize physical activity as a cornerstone of dementia risk reduction across the life course. ⁵ Importantly, they suggest that promoting physical activity remains relevant even at very advanced ages and should not be limited to younger or healthier older adults.

Conclusions

In summary, this study provides evidence that physical activity is associated with better cognitive status and reduced cognitive impairment risk in advanced old age, with effects becoming increasingly apparent from age 78 to 90. These findings underscore the importance of including very old adults in research on lifestyle-based cognitive impairment prevention and support the continued promotion of physical activity as a key component of healthy cognitive aging across the entire later-life spectrum.

Data availability statement

The data supporting the findings of this study are available from the corresponding author upon reasonable request.