Flow of participants in the Jerusalem Longitudinal Cohort Study. Throughout the 3 study phases, 1861 different participants enrolled in the study, of whom 462 (24.8%) died, 131 (7.0%) withdrew from subsequent phases of re-enrollment, and 76 (4.1%) were unavailable for follow-up. From the overall total of 3884 approaches made during the 3 study phases, 901 participants (23.2%) refused to participate (20.2%, 21.4%, and 25.7% at ages 70, 78, and 85 years, respectively). A further 113 (2.9%) were not enrolled owing to a lack of updated contact information, and 20 (0.5%) failed to enroll owing to an inability to communicate because they spoke a foreign language for which a translator was not available.
Kaplan-Meier survival curves according to physical activity (PA) level. A, Cumulative survival from ages 70 to 78 years according to PA level at age 70 years. B, Cumulative survival from ages 78 to 85 years according to PA level at age 78 years. C, Cumulative survival from ages 85 to 88 years according to PA level at age 85 years.
Kaplan-Meier survival curves according to changes in physical activity (PA) level. A, Cumulative survival from ages 78 to 88 years, stratified according to whether level of PA remained consistent, was low, increased, or decreased from ages 70 to 78 years. B, Cumulative survival from ages 85 to 88 years, stratified according to whether level of PA remained consistent, was low, increased, or decreased from ages 78 to 85 years.
Customize your JAMA Network experience by selecting one or more topics from the list below.
Stessman J, Hammerman-Rozenberg R, Cohen A, Ein-Mor E, Jacobs JM. Physical Activity, Function, and Longevity Among the Very Old. Arch Intern Med. 2009;169(16):1476–1483. doi:10.1001/archinternmed.2009.248
Copyright 2009 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2009
Recommendations encouraging physical activity (PA) set no upper age limit, yet evidence supporting the benefits of PA among the very old is sparse. We examined the effects of continuing, increasing, or decreasing PA levels on survival, function, and health status among the very old.
Mortality data from ages 70 to 88 years and health, comorbidity, and functional status at ages 70, 78, and 85 years were assessed through the Jerusalem Longitudinal Cohort Study (1990-2008). A representative sample of 1861 people born in 1920 and 1921 enrolled in this prospective study, resulting in 17 109 person-years of follow-up for all-cause mortality.
Among physically active vs sedentary participants, respectively, at age 70, the 8-year mortality was 15.2% vs 27.2% (P < .001); at age 78, the 8-year mortality was 26.1% vs 40.8% (P <.001); and at age 85 years, the 3-year mortality was 6.8% vs 24.4% (P < .001). In Cox proportional-hazards models adjusting for mortality risk factors, lower mortality was associated with PA level at ages 70 (hazard ratio, 0.61; 95% confidence interval, 0.38-0.96), 78 (0.69; 0.48-0.98), and 85 (0.42; 0.25-0.68). A significant survival benefit was associated with initiating PA between ages 70 and 78 years (P = .04) and ages 78 and 85 years (P < .001). Participation in higher levels of PA, compared with being sedentary, did not show a dose-dependent association with mortality. The PA level at age 78 was associated with remaining independent while performing activities of daily living at age 85 (odds ratio, 1.92; 95% confidence interval, 1.11-3.33).
Among the very old, not only continuing but also initiating PA was associated with better survival and function. This finding supports the encouragement of PA into advanced old age.
Physical activity (PA) is a modifiable behavior associated with health, functional status, and longevity, and encouraging a physically active lifestyle has become an accepted public health goal.1,2 Most research has focused on middle-aged populations and has described the relationship between PA and survival3-5 and the effect of starting or stopping PA on mortality.6-8 Few longitudinal studies have replicated these findings among older people, and, with few exceptions,9,10 research has not extended beyond age 80 years.11-14 Prior levels of activity15-17 and increasing or initiating PA influences survival among people in their 70s.18,19 However, to our knowledge, there is no evidence that continuing or initiating PA among the very old has long-term health or survival benefits. Despite this lack of substantive evidence, the recently revised recommendations for PA among elderly persons set no upper age limit.2,20
The Jerusalem Longitudinal Cohort Study examined the influence of PA among an aging cohort during 18 years of follow-up and addressed the following questions: Is PA among older adults, including the oldest old (age ≥85 years), associated with better survival? Is PA among the very old associated with functional or health benefits?
The Jerusalem Longitudinal Cohort Study has followed up a birth cohort of Jerusalem residents (born June 1, 1920, through May 31, 1921) from age 70 years at baseline in 1990 to the present. At phases I, II, and III (ages 70, 78, and 85 years, respectively), a total of 605, 1021, and 1222 participants were enrolled (Figure 1). The phase I study sample was augmented at phases II and III with new participants randomly recruited from the same birth cohort. At phase III, 30 participants previously seen at phase I but unavailable for follow-up at phase II reentered the study. Each participant, or his or her legal guardian, provided informed consent, and the Hadassah Hebrew University Medical Center Institutional Review Board approved the study.
Initial aims of the study included the description of aging patterns and risk factors for healthy aging. The study sample, which formed about one third of the total birth cohort, was randomly selected from the electoral registry (a complete register of Jerusalem residents born in 1920 and 1921). Participants underwent comprehensive home assessment at ages 70, 78, and 85 years.21 At each phase, participants were interviewed twice; an occupational therapist interviewed participants about social, demographic, cultural, health behavior, and functional domains, and the study physician gathered data concerning PA level, medical history, physical examination, and cognitive and psychological tests. The number of subjects who underwent both medical and social assessment at each phase, on whom the present study is based, was 457 of 605 (75.5%), 894 of 1021 (87.6%), and 1172 of 1222 (95.9%), respectively.
The representative nature of the study sample was confirmed by finding similar rates of hospital in-patient morbidity, health service utilization, and mortality among the study sample, those who refused to participate, and those not approached to enroll.22
The National Health Survey, conducted by the Central Bureau of Statistics from 2003 through 2004, includes data concerning the frequency of PA among Israelis.23 Among those older than 65 years, 41.3% were physically active (men, 49.1%; women, 35.3%), compared with 28.7% of the overall adult population. Physical activity was defined as participation in a sporting activity (walking, swimming, or other exercising) for at least 20 consecutive minutes in the 3 days before being interviewed. Considering the differences in how PA was defined, the rates of PA among this national sample are close in magnitude to those of our study.
Participants were asked: “How often are you physically active?” The answers were as follows: (1) less than 4 hours weekly, (2) about 4 hours weekly, (3) vigorous sports at least twice weekly (eg, jogging or swimming), and (4) regular PA (eg, walking at least an hour daily). This 4-item questionnaire, introduced in 1990 at the baseline evaluation, was adapted from the Gothenburg population study of 70-year-olds.24 We kept the same questions at follow-up to ensure internal consistency of longitudinal data.
Physical activity was dichotomized as sedentary (answer to 1) vs physically active (answers to 2, 3, and 4). This cutoff was justified statistically, accounting for distribution and frequency of responses.25 Separate analyses examined the 4 levels of PA as discrete variables in their effect on survival and function. Furthermore, at age 78 years, participants were classified according to PA levels at ages 70 and 78 years; at age 85 years, participants were classified according to PA levels at ages 78 and 85 years: consistent (physically active→physically active), decreasers (physically active→sedentary), increasers (sedentary→physically active), and low (sedentary→sedentary).
We examined the following characteristics: sex; origin (Ashkenazi vs Sephardi, as defined by the participant); educational level (in years); self-reported financial status (difficult vs not difficult); self-reported loneliness (often or occasionally vs never); depression using the Brief Symptom Inventory26; self-rated health status (good vs poor); Mini–Mental State Examination27 score; 2 measures of functional status, the first defined as dependence/independence on another person in 1 or more of 6 activities of daily living (ADLs)28 and the second, more sensitive, measure defined as ease/difficulty performing 1 or more of 6 ADLs29; body mass index (calculated as weight in kilograms divided by height in meters squared); smoking pack-years; chronic joint or musculoskeletal pain; medications used; falls in the last year; fractures in the last 7 years; and major diseases (hypertension, ischemic heart disease, diabetes mellitus, renal disease, or history of neoplasm) defined by the International Classification of Diseases, Ninth Edition.30 The study physician made diagnoses following medical assessment, system review, and examination. Similarly, data about basic ADLs were based on self-report and direct observation.
The primary outcome was death, and mortality data were obtained from an annual review of death certificates issued by the Ministry of the Interior from 1990 through 2008. This provided 100% capture of mortality data for those participants who remained in Israel, and since only a negligible number from this age group (<0.1%) leave the country, the accuracy of mortality data was considered to be complete. Secondary outcomes included deterioration over time in functional status, health measures, and new disease onset.
We compared physically active vs sedentary participants for baseline characteristics and follow-up data using χ2 tests for categorical variables and the Wilcoxon rank sum test for continuous variables (Table 1). We used Kaplan-Meier survival curves and log-rank testing to examine the influence of PA as a dichotomous variable (physically active vs sedentary) (Figure 2) and the effect of PA level (with sedentary participants serving as the reference group) at ages 70, 78, and 85 years on survival from ages 70 to 78 years, 78 to 85 years, and 85 to 88 years, respectively. We examined the influence of changes in PA level (ie, consistent, low, increasers, or decreasers) between ages 70 and 78 years on mortality from ages 78 to 88 years, and changes in PA level between ages 78 and 85 years on mortality from ages 85 to 88 years (Figure 3). We used Cox proportional-hazards models to adjust for established mortality risk factors and calculate hazard ratios with 95% confidence intervals (CIs) for mortality. Models examined PA as a dichotomous variable (sedentary vs physically active) (Table 2) and according to the 4 PA levels, comparing levels 2 through 4 with level 1 (sedentary participants) (Table 3). A time-dependent Cox proportional-hazards model31 analyzed the influence of PA together with risk factors as time-dependent variables throughout the study period on mortality from ages 70 to 88 years. This analysis was selected for its ability to account for changes in the levels of PA as well as confounding comorbid factors at the 3 potential time points during follow-up. The proportional hazards assumption was met in all models for PA exposure in each age group. All Cox proportional-hazards models adjusted for the same variables: sex, financial status, origin, body mass index, smoking pack-years, ease of performance in ADLs, hypertension, ischemic heart disease, diabetes mellitus, history of neoplasm, and chronic renal disease.
The effect of PA on the secondary outcomes of functional status and various health measures was examined using bivariate analysis; logistic regression models were performed to determine odds ratios (ORs).
Data storage and analysis were performed using SAS statistical software, 9.1e package (SAS Institute Inc, Cary, North Carolina). All P values were 2-tailed, and P < .05 was considered statistically significant.
A total of 1861 people enrolled in the overall study sample, yielding 17 109 person-years of follow-up for all-cause mortality data. At ages 70, 78, and 85 years, 46.6%, 23.0%, and 36.0%, respectively, were physically active for less than 4 hours weekly; 17.9%, 28.9%, and 36.4%, respectively, for about 4 hours weekly; 17.7%, 37.4%, and 16.7%, respectively, participated in regular PA (eg, walking at least an hour daily); and 18.1%, 10.6%, and 10.9%, respectively, participated in vigorous sports at least twice weekly (eg, jogging or swimming). Baseline characteristics are shown in Table 1. At ages 70, 77, and 85 years, 53.4%, 76.9%, and 64.0%, respectively, were physically active. Men were more active than women: 140 of 250 men (56.0%) vs 104 of 207 women (50.2%) at age 70 years, 384 of 441 (87.1%) vs 304 of 453 (67.1%) at age 78 years, and 376 of 528 (71.2%) vs 374 of 644 (58.1%) at age 85 years. Physical activity was more common among Ashkenazi participants and was associated with increased years of education, lack of financial difficulties, less depression, good self-rated health status, and independence in ADLs. Physically active participants took fewer medications and reported fewer falls or fractures and less chronic joint or musculoskeletal pain.
As a dichotomous variable (sedentary vs physically active), PA was significantly associated with better survival throughout follow-up. The number of deaths among physically active vs sedentary participants was 37 of 244 (15.2%) vs 58 of 213 (27.2%) (P < .001) from ages 70 to 78 years, 179 of 687 (26.1%) vs 84 of 206 (40.8%) (P < .001) from ages 78 to 85 years, and 51 of 750 (6.8%) vs 103 of 422 (24.4%) (P < .001) from ages 85 to 88 years (Figure 2). Thus, the magnitude of the difference was 12%, 15%, and 17% at ages 70, 78, and 85 years, respectively. After adjusting for confounding mortality risk factors, PA at ages 70, 78, and 85 years was significantly associated with lower mortality. Similarly, when treated as a time-dependent variable throughout the entire follow-up period, PA was also associated with better survival rates from ages 70 to 88 years (Table 2). We performed sensitivity analyses, eliminating early deaths (within 2 years from baseline) from the survival analyses at ages 70 to 78 years and 78 to 85 years and the time-dependent analysis from ages 70 to 88 years. No differences in magnitude or direction of PA effect on survival were observed for either unadjusted or adjusted hazard ratios (data not shown).
An analysis of the mortality rate according to the level of PA failed to show a consistent dose-dependent response (Table 3), and no evidence of a linear trend was found for any age cohort.
Kaplan-Meier survival curves, stratified according to change over time in PA levels, showed that not only continuing PA (consistent) but also starting PA (increasers) between ages 70 to 78 years and 78 to 85 years was associated with better survival. The increased mortality rate among decreasers was similar to that of sedentary subjects (low group). The mortality in the consistent, increasers, decreasers, and low groups were 31.6%, 40.6%, 56.3%, and 51.2% from ages 78 to 88 years (P = .04) and were 8.0%, 12.8%, 25.3%, and 25.0% from ages 85 to 88 years (P < .001) (Figure 3). Cox proportional-hazards models were not performed because of insufficient numbers in each category.
Physical activity level was associated with maintaining independent functional status over time, irrespective of increasing age. Among participants who were independent in ease of performing ADLs, physically active participants deteriorated significantly less than sedentary participants between ages 70 and 78 years (33.3% vs 52.3%; P = .01) and deteriorated marginally less from ages 78 to 85 years (81.3% vs 83.2%; P = .70). The opposite was observed for independence in performing ADLs: marginally less deterioration among physically active participants occurred from ages 70 to 78 years (6.8% vs 12.3%; P = .12) and significantly less deterioration from ages 78 to 85 years (26.7% vs 41.8%; P = .01). Logistic regression analyses including all participants in the sample were performed. After adjusting for sex, prior baseline ADL level, body mass index, diabetes mellitus, ischemic heart disease, and hypertension, PA level at age 70 years was associated with increased likelihood of maintaining ease in performing ADLs at age 78 years (unadjusted OR, 2.0; 95% CI, 1.16-3.45; adjusted OR, 1.72; 95% CI, 0.97-3.03) (P = .06), and PA level at age 78 years was significantly associated with increased likelihood of remaining independent while performing ADLs at age 85 years (unadjusted OR, 2.12; 95% CI, 1.26-3.57; adjusted OR, 1.92; 95% CI, 1.11-3.33). Sensitivity analyses were performed, after eliminating early deaths within the first 2 years of follow-up. No change in direction or magnitude of the associations was observed. Regression analyses of functional status as an outcome of PA level were not performed because there were insufficient numbers in each category.
During follow-up from ages 70 to 78 years and 78 to 85 years, physically active participants reported less onset of loneliness (12.2% vs 22.6% [P = .06]; 26.5% vs 44.1% [P = .03]) and poor self-rated health (77.3% vs 63.3% [P = .04]; 63.8% vs 82.6% [P = .07]) than sedentary participants. Neither association remained significant after adjustment for confounders. Physical activity level was also associated with less new onset of hypertension from ages 70 to 78 years (27.7% vs 43.5%; P = .03). However, this association became insignificant after adjusting for baseline hypertension, sex, self-rated health status, ADLs, diabetes mellitus, and ischemic heart disease (OR, 0.68; 95% CI, 0.38-1.22) (P = .19). No other longitudinal associations emerged for baseline health measures.
The results of this 18-year longitudinal cohort study support the hypothesis that not only continuing but also initiating PA among older people delays functional loss and improves survival. Not only was the effect of this benefit similar regardless of increasing age but the magnitude of the difference between physically active and sedentary participants actually increased with advancing age. Maximum survival benefit was observed among the oldest age group, a finding that, to our knowledge, has not been reported previously. Analysis of survival according to PA level found the primary distinction to be between sedentary vs active participants, with no clear dose-dependent effect.
Although the mechanism of the survival benefit is most likely multifactorial, one important finding was the sustained protective effect of PA against functional decline. Maintaining function is a central goal in aging, and awareness is increasing of the intimate relationship between the phenotype of frailty, loss of physiological reserves, and performance-based measures of functional decline as harbingers of preterminal trajectories of illness and mortality. Among older people, PA may be instrumental in delaying the onset of this spiral of decline through its influence on a spectrum of pathways, which may include improved cardiovascular fitness, decelerated sarcopenia, reduced adiposity, and improved immunity together with the suppression of chronic inflammation.31-34 Evidence in support of this has emerged from exercise training interventions, which show positive short-term effects on physiological and performance-based parameters among older adults.35,36
Studies among middle-aged participants have shown that individuals who increased their levels of PA7,8 displayed better survival rates. However, to date, the findings among the “young” elderly have been conflicting (based on an English-language MEDLINE search through March 2009). Gregg et al18 showed that among elderly white women the most recent PA levels were more important predictors of longevity than past levels; however, the benefit of increasing PA did not extend beyond age 75. Similarly, a positive effect on survival was associated with increasing recent PA levels among participants in their early 70s in both the British Heart Study6 and the Zutphen Elderly Study.17 Other studies that failed to show benefits from increasing recent PA levels37 nevertheless described increased mortality associated with decreasing PA levels. In contrast, the Copenhagen Male Study38 found that increased recent PA levels were associated with higher mortality from ischemic heart disease among a sample aged 65 to 74, a finding yet to be reproduced elsewhere.
In our study, being physically active up to age 85 years remained a strong predictor of survival, a robust finding that emerged from Cox proportional-hazards models at ages 70, 78, and 85 years and the time-dependent model spanning 18 years of follow-up from ages 70 to 88 years. Furthermore, changing levels of PA over time were associated with increased survival; the Kaplan-Meier survival trajectories of “increasers” were similar to those of consistently active participants (consistent group), and those of decreasers resembled the trajectories of the consistently sedentary participants (low group). Our findings suggest that not only is PA protective among people who remained active throughout their lives into old age but becoming active during advanced old age is also beneficial, even among previously sedentary people. Although it may be argued that reverse causality is at work, and that PA served as a proxy for good health, nevertheless, the association with survival remained significant after adjustment for comorbidities, functional status, and self-rated health status and after eliminating data from early deaths.
Few studies have examined the relationship of PA and changes in ADLs,29,39,40 and the few longitudinal studies among them29,40 have not extended to include the oldest old. The relationship of PA at age 70 years and the ease of performing ADLs 7 years later was described within our cohort.29 Elsewhere, baseline and increasing PA levels during 16 years of follow-up among long-distance runners improved functional status into their early 70s.40 Unlike Berk et al,40 who examined endurance athletes, our cutoff point included as little as 4 hours per week of PA, which might have attenuated the positive impact of more vigorous activity. However, the analysis of mortality data stratified according to PA level showed that the magnitude of benefit did not significantly differ between the different subgroups, with no evidence of a linear trend between the various PA groups. Of interest, lack of consistency between exercise levels and survival advantage corresponds to findings from Manini et al,41 who confirmed the hypothesis that increasing energy expenditure of regular daily activities was associated with a lower mortality risk among older adults. However, they failed to find a clear dose-dependent association between PA and mortality, suggesting, rather, that self-reported PA questionnaires may underestimate the positive expenditure of regular daily activities.
Physical activity was prevalent among our study sample, and 76.9% were active at age 78 years. These rates are greater than those quoted for North American populations42,43 and support the positive prevailing attitude toward an active lifestyle among the study population and among all elderly Israelis. Generalizability of our results to other cultures requires caution. However, the cohort was extremely varied in place of origin with more than 40 countries of birth reported. Furthermore, the beneficial effect of PA remained after adjusting for origin, socioeconomic status, and educational level, suggesting that the benefits of PA are independent of the cultural milieu.
Although we did not directly investigate the possibility that PA among elderly participants might result in adverse effects, several surrogate measurements suggested that this was not the case. The frequency of falls, fractures, and chronic joint or musculoskeletal pain was consistently lower among physically active participants. These findings support epidemiological data, which suggest that exercise-related injuries actually decrease in frequency among elderly exercisers.44 Similarly, these observations provide circumstantial support for the growing consensus that continued PA does not increase the likelihood of osteoarthritis of the knee and may actually reduce musculoskeletal pain, as shown among elderly long-distance runners who were followed up for more than 2 decades.45,46 The finding that falls and recent fractures were lower among active participants is not surprising, and evidence exists in support of implementing PA as a treatment modality in fall prevention among older adults.47 We found no published evidence suggesting an upper age limit to PA or a maximum recommended level for exercise intensity. In a study of 75-year-olds who achieved their maximum heart rate on a cycle ergometer, although 10% stopped because of rhythm disturbances, no lasting rhythm or ischemic changes were observed.48 In contrast, a wealth of literature exists to support increased oxygen consumption per unit time, cardiac output, peripheral circulation, bone density, muscle mass, improved stability, and flexibility.20
Our findings warrant further investigation: quantification of energy expenditure, together with physiological and performance-based measures, would help to elucidate the mechanism of benefit. There is a need to quantify optimal types and threshold levels for PA recommendations, according to the biological, functional, and comorbid status of the very old.
The study limitations include the diminishing size of the sample during follow-up, a problem inherent in longitudinal aging studies of this duration. This may have introduced an inherent healthy survivor bias. However, sample augmentation at ages 78 and 85 years with randomly chosen recruits from the same birth cohort served to redress the sample's representative nature. An additional limitation concerns the self-reported nature of the PA data, together with a lack of documented validity or reliability of the PA question, or its sensitivity at detecting changes in PA levels. The intensity of PA most likely varied between participants, which would be expected to reduce the power of the study to show the effect of PA. The fact that PA still remains an independent predictor suggests that activity of at least 4 hours per week, daily walking, or participating in sports twice weekly is beneficial for older adults.
In conclusion, our study provides unique evidence that not only maintaining PA but also starting to be physically active across a spectrum of advancing ages, up to and including the oldest old, is associated with improved function and survival. Despite the increasing likelihood of comorbidity, frailty, dependence, and ever-shortening life expectancy, remaining and even starting to be physically active increases the likelihood of living longer and staying functionally independent. The clinical ramifications are far reaching. As this rapidly growing sector of the population assumes a prominent position in preventive and public health measures, our findings clearly support the continued encouragement of PA, even among the oldest old. Indeed, it seems that it is never too late to start.
Correspondence: Jeremy M. Jacobs, MBBS, Department of Geriatrics and Rehabilitation, Hadassah Hebrew University Medical Center, Mt Scopus, PO Box 24035, Jerusalem 91240, Israel (email@example.com).
Accepted for Publication: May 18, 2009.
Author Contributions: All authors had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Stessman and Jacobs. Acquisition of data: Stessman, Hammerman-Rozenberg, Cohen, and Jacobs. Analysis and interpretation of data: Stessman, Hammerman-Rozenberg, Cohen, Ein-Mor, and Jacobs. Drafting of the manuscript: Stessman and Jacobs. Critical revision of the manuscript for important intellectual content: Stessman, Hammerman-Rozenberg, Cohen, Ein-Mor, and Jacobs. Statistical analysis: Stessman, Ein-Mor, and Jacobs. Administrative, technical, or material support: Stessman. Study supervision: Stessman.
Financial Disclosure: None reported.
Funding/Support: The Jerusalem Longitudinal Study has received funds from the Ministry of Labor and Social Affairs of the State of Israel; ESHEL, the Association for the Planning and Development of Services for the Aged in Israel; the National Insurance Institute; and various private, charitable donors. No support was offered by any commercial venture. These funds were used exclusively to support the research effort, primarily as salaries to ancillary staff.
Role of the Sponsors: The sponsors had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; or preparation, review, or approval of the manuscript.
Additional Contributions: Etty Arbel and Liat Cohen provided organizational, secretarial, and logistic assistance, and Aliza Hammerman-Rozenberg, MA, contributed to the statistical analysis.