JAGS 50:1774–1781, 2002© 2002 by the American Geriatrics Society 0002-8614/02/$15.00

Physical Activity as a Determinant of Change in Mobility Performance: The Longitudinal Aging Study Amsterdam

Marjolein Visser, PhD,* Saskia M. F. Pluijm, PhD,* Vianda S. Stel, MSc,*

Ruud J. Bosscher, PhD,

†

and Dorly J. H. Deeg, PhD*

OBJECTIVES:

This study examined the association of(change in) physical activity and decline in mobility per-formance in older men and women.

DESIGN:

A 3-year prospective study using data of theLongitudinal Aging Study.

SETTING:

Netherlands.

PARTICIPANTS:

Two thousand one hundred nine menand women aged 55 to 85.

MEASUREMENTS:

Total physical activity (expressed as

hours per day and kilocalories per day) and sports par-ticipation were measured using a validated, interviewer-administered questionnaire. Mobility performance was as-sessed using two timed tests: 6-meter walk and repeatedchair stands.

RESULTS:

Mobility performance declined for 45.6% ofthe sample. At baseline, the mean time

�

standard devia-tion spent on total physical activity was 3.0

�

2.1 h/d or719

�

543 kcal/d, and 56.6% of the sample participatedin sports. Sports participation and a higher level of totalphysical activity, walking, or household activity wereassociated with a smaller mobility decline. After 3 years,total physical activity declined, and only 53.4% of thosereporting sports at baseline continued doing so. Continua-tion of physical activity over time was associated with thesmallest decline in mobility. The observed associationswere similar for those with and without chronic disease (

P

�

0.3). The conclusions did not change after adjustment forpotential confounders, including demographic and life-style variables, depression, and cognitive status.

CONCLUSIONS:

Physical activity, and especially a regu-larly active lifestyle, may slow the decline in mobility per-formance. A beneficial effect was observed for sports andnonsports activities, independent of the presence ofchronic disease.

J Am Geriatr Soc 50:1774–1781, 2002.Key words: exercise; functional status; longitudinal study;

mobility; physical activity

P

hysical activity is known to decrease the risk forchronic disease (e.g., diabetes mellitus and heart dis-

ease) and mortality from cardiovascular disease and allcauses in older persons.

1–3

Much less attention has beendevoted to the role of physical activity on functional sta-tus, independent of disease.

Exercise trials in older persons have consistentlyshown the beneficial effect of resistance training on func-tional status.

4–6

Furthermore, positive effects of supervisedlow-intensity exercise programs on functional status havealso been observed,

7–9

but the question remains whetherregular, daily activities in the general older populationmay protect against functional decline.

Epidemiological studies investigating the relationshipbetween physical activity and self-reported disability haveshown a protective effect of higher physical activity on dis-ability onset.

10–12

These studies were limited by their use ofself-reported disability, which is a subjective measure offunctional status influenced by sex, cognitive status, per-ceived mastery, and depressive symptoms;

13–15

conse-quently, substantial misclassification may occur.

16

One study included performance-based measures andshowed a lower risk for performance decline during 2.5years of follow-up in participants who were physically ac-tive at baseline,

17

but this study made no distinction be-tween sports and nonsports activities. Furthermore, nostudies have examined the relationship between change inphysical activity and change in functional performance inolder persons. Physical activity is known to decline withaging in persons aged 65 and older.

18

It has been suggestedthat this change in physical activity pattern influences thechange in functional status.

11

From the *Institute for Research in Extramural Medicine, VU University

Medical Center, Amsterdam, the Netherlands; and

†

Department of Human Movement Sciences, Vrije Universiteit, Amsterdam, the Netherlands.

The research of Dr Visser has been made possible by a fellowship of the Royal Netherlands Academy of Arts and Sciences. The Longitudinal Aging Study Amsterdam was financed primarily by the Netherlands Ministry of Health, Welfare, and Sports.

Address correspondence to Marjolein Visser PHD, Institute for Research in Extramural Medicine, VU University Medical Center, Van der Boechorststraat 7, 1081 BT Amsterdam, the Netherlands. E-mail: m.visser.emgo@med.vu.nl.

JAGS NOVEMBER 2002–VOL. 50, NO. 11

PHYSICAL ACTIVITY AND CHANGE IN MOBILITY

1775

The aim of this prospective study was to examinewhether physical activity is a determinant of future changein mobility performance in a population-based sample ofmen and women aged 55 to 85. Included were total physi-cal activity, sports activity, and daily activities such aswalking and household activities. In addition, whetherchange in physical activity is related to change in mobilityperformance was examined.

METHODS

Study Sample

Data for this study were collected in the Longitudinal Ag-ing Study Amsterdam, a prospective study of persons aged55 to 85. The sampling and data collection proceduresand nonresponse have been described elsewhere in de-tail.

19,20

In summary, a random sample stratified by age,sex, and expected 5-year mortality was drawn from thepopulation registers of 11 municipalities in the Nether-lands. Three thousand one hundred seven subjects wereenrolled in the baseline examination (1992–93) and wererepresentative of the older Dutch population. Record of a3-year follow-up examination was available for 2,545 per-sons (deceased n

�

417, refused n

�

90, ineligible n

�

38,not contacted n

�

17). A total of 323 persons were ex-cluded because of missing or incomplete data on mobilityperformance at baseline or follow-up. Also excluded were113 persons because of missing or incomplete data on phys-ical activity at baseline, leaving 2,109 participants (67.9%of original cohort) available for the statistical analyses. Thelocal medical ethics committee approved the study. All re-spondents gave informed consent at the start of the study.

Compared with the 2,109 persons included in theanalyses, those who were excluded or lost to follow-upwere older at baseline (73.9 vs 69.3), less physically active(total activity 2.6 vs 3.0 h/d, sports participation 44.8% vs56.5%), had more chronic diseases (1.1 vs 0.9), and weremore likely to be female (51.9% vs 46.9%) (all

P

�

.01).

Physical Activity

Information on physical activity was collected at baselineand after 3 years of follow-up. The information was ob-tained using an interviewer-administered questionnairebased on the questionnaires by Voorrips et al.

21

andCaspersen et al.

22

The following activities were addressed:walking outdoors, bicycling, light household activities,heavy household activities, and a maximum of two sportsactivities. A selection of 16 sports categories (with exam-ples ranging from fishing and bowling to running and ski-ing) was presented on a card during the interview. Addi-tional sports activities mentioned by the respondents werealso recorded.

Respondents were asked how often and for how longin the previous 2 weeks they had engaged in each activity.They were also asked whether they considered their physi-cal activity pattern of the previous 2 weeks to be normalcompared with the rest of the year (yes/no). The question-naire has been validated against a 7-day physical activitydiary (n

�

356) and pedometer counts (n

�

296) in per-sons aged 70 to 92, showing correlation coefficients of0.70 and 0.52 (Stel, unpublished data).

Based on self-reported sports activities in the previous2 weeks, a dichotomous variable for sports participationwas created (yes/no). No minimum time was required tobe considered a sports participant. Walking and bicyclingfor transportation purposes were not considered sports ac-tivities because they are common daily activities in theNetherlands. Two measures of total physical activity werecreated. The total time spent on physical activity (in hoursper day) was calculated by multiplying the frequency andduration of each activity in the previous 2 weeks, sum-ming these values across activities, and dividing the sum-score by 14.

23

A second measure of total physical activitywas created that considered different levels of intensity ofactivities. A metabolic equivalent value was assigned toeach activity and was used to calculate the number of kilo-calories per day per kilogram of body weight spent on thatactivity.

22,24

For each participant, the scores of all per-formed activities were summed and multiplied by bodyweight to create an overall physical activity score in kcal/d.For 112 participants, total physical activity in kilocaloriesper day was considered missing because no measured bodyweight was available.

Information on 3-year change in total physical activityexpressed as hours per day was available for 2,049 partic-ipants (96.8% of the 2,109 participants), and 3-yearchange in total physical activity expressed as kilocaloriesper day was available for 1,942 participants (92.1% of the2,109 participants). Change in total physical activity wascalculated as total physical activity level at 3-year follow-up minus the baseline level. Change in sports participationwas categorized into four groups: no sports participationat either examination, sports participation at both exami-nations, discontinued sports activities, and started sportsactivities.

Mobility Performance

Mobility performance was measured at baseline and at3-year follow-up using a timed walking test and chair-stand test. Both tests were performed in the participants’home. To test walking performance, a 3-meter walkingcourse was created with a measuring-line. Participantswere instructed to walk to the end of the course, to turnaround, and to walk back as quickly as possible. Theywere allowed to use assistive devices if needed. To test theability to rise from a chair, persons were asked to foldtheir arms across their chest and to stand up and sit downfive times from a standard kitchen chair as quickly as pos-sible. A trained research nurse recorded the total time tocomplete each of the two tests, and a standard protocolwas used throughout the study.

19

These objective measuresof mobility performance are highly predictive of subse-quent disability, nursing home admission, and mortality.

25

Test-retest correlations of more than 0.89 for walking and0.73 for repeated rising from a chair have been reportedfor these measures.

25

Those completing a test were assigned scores of 1 to 4,corresponding to the quartiles of time needed to completethe test, with the fastest times scored as 4.

16

Those whocould not complete a test (e.g., because they could notwalk, were in a wheelchair, or were physically not capableor because the test was not fully completed or completed

1776

VISSER ET AL.

NOVEMBER 2002–VOL. 50, NO. 11 JAGS

with help) were assigned a score of 0. Mobility perfor-mance was calculated by summing the walk score and thechair stand score. Values could range from 0 (unable toperform both tests) to 8 (time in fastest quartile for bothtests). The change in mobility performance was computedby subtracting the baseline score from the 3-year follow-up score.

Covariates

Baseline sociodemographics included age, sex, and level ofeducation (low, medium, high). The number of self-reporteddiseases was used as a measure of somatic disease. Partic-ipants were asked (yes/no) whether they have or haveever had any of the following diseases or disease events:chronic obstructive pulmonary disease (asthma, chronicbronchitis, pulmonary emphysema), cardiac disease, pe-ripheral atherosclerosis, stroke, diabetes mellitus, arthri-tis (rheumatoid arthritis and osteoarthritis), and cancer.

26

A Dutch version of the Center for Epidemiologic StudiesDepression scale, a 20-item self-report scale, was used tomeasure depressive symptoms.

27,28

A score of 16 points orhigher was used to identify clinically relevant depression.Cognitive function was measured with the Mini-MentalState Examination.

29

A score of 24 or less was used toidentify low cognitive function. Smoking (never, former,current) and alcohol consumption (none,

�

7 drinks perweek,

�

7 drinks per week) were based on self-report.

Statistical Analyses

All analyses were conducted using SAS software version6.12 (SAS Institute Inc., Cary, NC). Change in mobilityperformance and physical activity were tested using Stu-dent paired

t

tests. Reported correlations were Pearsonproduct-moment correlations. Total physical activity, walk-ing, bicycling, and household activity and change in totalphysical activity were each categorized into five groups,each including about 20% of the study sample. The use ofcategories enabled the evaluation of a potential thresholdin the association with change in mobility performance.The most inactive category or the group with the greatestdecline in total physical activity was used as the referencegroup. The association between baseline sports participa-tion and total physical activity with change in mobilityperformance over 3 years was tested using analyses of co-variance. The association between (1) change in sportsparticipation and change in total physical activity and (2)change in mobility performance over 3 years was alsotested using analyses of covariance. To test for trend, thefive groups were entered in the model as an ordinal vari-able. All analyses were adjusted for age, sex, and baselinemobility performance. In a second model, we additionallyadjusted for education, smoking, alcohol consumption,cognitive function, number of chronic diseases, and de-pression. The association between change in total physicalactivity and change in mobility performance was addition-ally adjusted for baseline total physical activity. Possibleinteractions of sports participation or total physical activ-ity with health status, sex, or age were tested by includingproduct terms in the models.

RESULTSChange in Mobility Performance

After 3 years of follow-up (n

�

2,109), the mean

�

stan-dard deviation change in mobility performance of thestudy sample was –0.4

�

1.8 (

P

�

.01). The change in mo-bility performance was normally distributed (Figure 1).The percentage of the study sample that experienced a de-cline in mobility performance was 45.6%.

The baseline characteristics of the study sample andtheir relationship with the change in mobility performanceare shown in Table 1. Younger age, higher education,fewer chronic diseases (

P

�

.09), poorer mobility perfor-mance, and the consumption of seven or more alcoholicdrinks per week at baseline were associated with a smallerdecline in mobility performance during follow-up.

Baseline Physical Activity

At baseline, the mean time spent on total physical activityper day was 3.0

�

2.1 h/d, or 719

�

543 kcal/d. The twomeasures of total physical activity were highly correlated(

r

�

0.93,

P

�

.01). Only 5.4% of the study sample spentless than 30 min/d (the recommended level of physical ac-tivity in the Netherlands) on these activities. At baseline,more than half the study sample (56.5% or n

�

1,192) re-ported participation in one or two sports activities duringthe previous 2 weeks. The sports activities most frequentlyreported were gymnastics (52.3%, defined as the perfor-mance of strength, balance, and stretching exercises athome, sports club, or community center), distance walking(25.8%), and distance bicycling (20.8%). For 76.6% of thestudy sample at baseline, the physical activity pattern of theprevious 2 weeks was representative of the rest of the year.

Baseline Sports Participation and Change inMobility Performance

The relationship between baseline sports participation andchange in mobility performance is shown in Table 2. Partic-ipation in sports activity at baseline was associated with asmaller decline in mobility performance than that of no par-ticipation. After adjustment for physical and psychologicalhealth and lifestyle variables, the difference remained statis-

Figure 1. Distribution of 3-year change in mobility performancescore.

JAGS NOVEMBER 2002–VOL. 50, NO. 11

PHYSICAL ACTIVITY AND CHANGE IN MOBILITY

1777

tically significant (

P

�

.009). Participants who reportedsports activity at baseline were overall more active, evenwhen only nonsports activities (walking, bicycling, andhousehold activities) were considered. Participants who re-ported sports participation spent 2.8

�

1.9 h/d, while par-ticipants who did not report sports activities spent 2.5

�

1.9h/d on these activities (

P

�

.001), but, after additional ad-justment for the time spent on nonsports activity, the associ-ations between sports participation and change in mobilityperformance remained virtually unchanged (not shown).

Baseline Total Physical Activity and Change inMobility Performance

The relationship between baseline total physical activityand change in mobility performance is also shown in Ta-ble 2. After adjustment for age, sex, and baseline mobilityperformance, higher total physical activity expressed as

hours per day was associated with a smaller decline in mo-bility performance (

P

for trend

�

.009). This relationshipdid not change markedly after additional adjustment forother potential confounders. A similar association was ob-served when total physical activity was expressed in kilo-calories per day. Persons in the upper four categories expe-rienced a smaller decline in mobility performance thanthose in the most inactive category. No differences wereobserved between the upper four categories, suggestingthat, above a certain level of total physical activity, no ad-ditional beneficial effect on mobility performance can beexpected. This potential threshold effect suggests that per-sons spending 300 kcal/d or more or 1 hour and 20 min-utes or more per day have a smaller risk for mobility de-cline than the least active persons.

Whether individual nonsports activities were associ-ated with change in mobility performance (Figure 2) were

Table 1. Characteristics of Study Sample (N

�

2,109) and Relationship with 3-Year Change in Mobility Performance Score

Characteristic

Change in MobilityPerformance Score

n Mean SE

GenderWomen 1,121

�

0.45 0.05Men 988

�

0.35 0.06Age

�

65 782

�

0.20 0.0665–74 688

�

0.31 0.07

�

75 639

�

0.74 0.07**Education

Low 844

�

0.56 0.06Medium 1,001

�

0.29 0.06High 262

�

0.32 0.11**Smoking

Never 607

�

0.46 0.07Former 846

�

0.37 0.06Current 470

�

0.31 0.08Missing 186

�

0.58 0.13Alcohol (drinks/week)

0 385

�

0.48 0.091–6 891

�

0.44 0.06

�

7 640

�

0.25 0.07Missing 193

�

0.60 0.13*Center for Epidemiological Studies—Depression Scale

�

16 1,849

�

0.41 0.04

�

16 260

�

0.35 0.11Mini-Mental State Examination

�

24 133

�

0.64 0.16

�

24 1,976 �0.39 0.04Chronic diseases

0 858 �0.30 0.061 795 �0.44 0.06�2 455 �0.52 0.09

Baseline mobility performance score0–2 309 �0.42 0.103–5 854 �0.08 0.066–8 946 �0.96 0.06**

*P � .05, **P � .01, analysis of variance.SE � standard error.

1778 VISSER ET AL. NOVEMBER 2002–VOL. 50, NO. 11 JAGS

also examined. After adjustment for all potential confound-ers, a higher level of walking activity, household activities,and bicycling was associated with a smaller decline in mo-bility performance after 3 years of follow-up. These rela-tionships remained statistically significant after additionaladjustment for baseline sports participation (not shown).

Interaction of Baseline Physical Activity with Health Status, Sex, and AgeA higher physical activity level at baseline may be amarker of good health, potentially explaining the smallerdecline in mobility performance in active persons. Wetherefore examined whether the relationship of baselinesports/total physical activity with change in mobility per-formance was different for healthy participants and partic-ipants with one or more chronic diseases at baseline. Nostatistically significant interaction between baseline sportsparticipation and health status (P � .85) or between base-line total physical activity and health status (P � .59 forphysical activity in h/d and P � .34 for physical activity inkcal/d) was observed. Figure 3 shows the similar associa-tion between total physical activity expressed in kilocalo-ries per day and change in mobility performance for par-ticipants with and without chronic disease at baseline. Nointeraction was observed between baseline sports partici-pation or baseline total physical activity and sex (P � .12)or between baseline sports participation and age (P �.42). Only the interaction between baseline total physicalactivity and age tended to be statistically significant (P �.06), indicating a stronger relationship in participants aged70 and older than in those younger than 70.

Three-Year Change in Physical ActivityDuring follow-up, mean total physical activity declinedfrom 2.9 � 2.1 h/d at baseline to 2.5 � 1.8 h/d at follow-up (P � .0001). For 29.0% of the study sample, total phys-ical activity declined more than 1 h/d, and 14.6% reporteda decline of more than 2 h/d. Mean total physical activity

Table 2. Baseline Physical Activity in Relation to 3-Year Change in Mobility Performance Score

Physical Activity Parameter

Change in Mobility Performance Score

n

Model 1* Model 2†

Mean � Standard Error

Sports participationNo‡ 917 �0.53 � 0.05 �0.49 � 0.06Yes 1,192 �0.31 � 0.05§ �0.30 � 0.05§

Total physical activity (hours/day)�1.3‡ 417 �0.64 � 0.08 �0.61 � 0.081.3–1.9 327 �0.29 � 0.09§ �0.28 � 0.09§

2.0–2.9 489 �0.46 � 0.07 �0.46 � 0.083.0–3.9 482 �0.29 � 0.07§ �0.25 � 0.08§

�4.0 394 �0.31 � 0.08§ �0.28 � 0.09§

P-value for trend .009 .008Total physical activity (kcal/day)

�300‡ 372 �0.77 � 0.09 �0.74 � 0.09300–499 446 �0.31 � 0.08§ �0.33 � 0.08§

500–699 362 �0.19 � 0.08§ �0.19 � 0.08§

700–999 386 �0.29 � 0.08§ �0.28 � 0.08§

�1,000 431 �0.35 � 0.08§ �0.38 � 0.08§

P-value for trend .005 .01

*Adjusted for age, sex, and baseline mobility performance score.†Additionally adjusted for education, smoking, alcohol consumption, depression, cognitive function, and chronic disease.‡Reference.§P � .01 versus reference category.

Figure 2. Three-year change in mobility performance score(with standard error) according to categories of baseline walk-ing, household, and bicycling activity. *P � .05 vs lowest cate-gory. Walking activity: 1 � �5 min/d, 2 � 5–15 min/d, 3 �15–30 min/d, 4 � 30–60 min/d, 5 � �60 min/d. Household ac-tivity: 1 � �30 min/d, 2 � 30–60 min/d, 3 � 60–120 min/d,4 � 120–180 min/d, 5 � �180 min/d. Bicycling activity: 1 � 0min/d, 2 � 1–10 min/d, 3 � 10–20 min/d, 4 � �20 min/d.

JAGS NOVEMBER 2002–VOL. 50, NO. 11 PHYSICAL ACTIVITY AND CHANGE IN MOBILITY 1779

expressed in kilocalories per day (n � 1,942) declined from600 � 493 kcal/d to 547 � 446 kcal/d (P � .0001). Only53.4% of the participants who reported sports activity atbaseline still participated in sports activities at follow-up.The majority (82.9%) of participants not reporting anysports participation at baseline were still not involved insports activities 3 years later. For 72.9% of the study sam-ple at follow-up, the physical activity pattern of the previ-ous 2 weeks was representative of the rest of the year.

Change in Sports Participation and Change inMobility PerformanceIn addition to baseline sports participation, we also exam-ined whether change in sports participation during follow-up was related to change in mobility performance. Table 3shows the relationship between change in sports participa-tion during follow-up and change in mobility perfor-mance. Participants who remained active experienced nodecline in mobility performance (mean change � � 0.01),whereas participants who remained inactive (mean � �0.45,

P � .0001) or stopped sports participation during follow-up (mean change � � 0.49, P � .0001) experienced alarger decline (Model 1). Participants who started sportsparticipation during follow-up tended to have a smallerdecline in mobility performance (mean change � � 0.21)than those who remained inactive (mean change � � 0.45,P � .07). These results did not change after additionaladjustment for other potential confounders (Model 2).

Change in Total Physical Activity and Change inMobility PerformanceThe relationship between change in total physical activityand change in mobility performance is shown in Table 4.After adjustment for potential confounders, the change intotal physical activity expressed in hours per day was asso-ciated with 3-year change in mobility performance (P fortrend � .0004). Persons who had a stable activity patternduring follow-up (change in total physical activity 0 to�1.0 h/d) experienced the smallest decline in mobility per-formance. In this model, baseline total physical activity ex-pressed in hours per day remained independently associ-ated with change in mobility performance (P for trend �.0001). Similar results were observed when change in totalphysical activity was expressed in kilocalories per day. Asmaller decline in total physical activity was associated with asmaller decline in mobility performance (P for trend �.03). Baseline total physical activity expressed in kilocalo-ries per day remained independently associated with mo-bility change (P for trend � .03).

Interaction of Change in Physical Activity with Health Status, Sex, and AgeThis study examined whether the relationship of change insports participation/change in total physical activity withchange in mobility performance was different for healthyparticipants and participants with one or more chronicdiseases at baseline. No statistically significant interactionbetween change in sports participation and health status(P � .86) or between change in total physical activity andhealth status (P � .53 for physical activity in h/d and P �.79 for physical activity in kcal/d) was observed. No statis-tically significant interaction was observed between change

Figure 3. Three-year change in mobility performance score(with standard error) according to categories of baseline totalphysical activity (kcal/d) for participants with and withoutchronic disease at baseline. *P � .05 vs lowest category. 1 ��300 kcal/d, 2 � 300–499 kcal/d, 3 � 500–699 kcal/d, 4 �700–999 kcal/d, 5 � �1000 kcal/d.

Table 3. Three-Year Change in Mobility Performance Score According to Change in Sports Participation During Follow-Up

Change in Sports Participation

Change in Mobility Performance Score

Model 1* Model 2†

n Mean � Standard Error

Remained inactive 773 �0.45 � 0.06 �0.42 � 0.06Remained active 592 �0.01 � 0.06‡ �0.01 � 0.06‡

Stopped 502 �0.49 � 0.07§ �0.46 � 0.07§

Started 155 �0.21 � 0.12 �0.19 � 0.12

Note: Information missing for n � 126.*Adjusted for age, sex, and baseline mobility performance score.†Additionally adjusted for education, smoking, alcohol consumption, depression, cognitive function, and chronic disease.‡P � .001 versus remained inactive.§P � .001 versus remained active.

1780 VISSER ET AL. NOVEMBER 2002–VOL. 50, NO. 11 JAGS

in sports participation or change in total physical activityand sex (P � .33) or age (P � .22).

DISCUSSIONThe results of this prospective study of 2,109 older menand women show a protective effect of physical activity onfuture decline in mobility performance. Timed perfor-mance tests to objectively assess change in mobility perfor-mance were used, which is a major advantage over previ-ous studies investigating the association between physicalactivity and self-reported disability, which is a more sub-jective measure of functional status. Another advantage isthat performance tests can validly define a gradient offunctioning even at the upper end of the functional spec-trum and are therefore able to assess the full range of func-tional status better then self-report measures, whichmainly identify the presence of overt disability.16,25 More-over, our study extends the findings of previous studies byinvestigating the relationship between change in physicalactivity level and change in mobility performance.

Sports participation and an overall higher physical ac-tivity level were associated with a smaller decline in mobil-ity performance. Furthermore, participants who spent moretime on daily activities such as walking and household ac-tivities experienced a smaller decline in mobility perfor-mance. This finding suggests that sports activity per se isnot necessary to influence change in mobility performance.For older persons who are unable or unwilling to partici-pate in vigorous exercise, less-intensive activities may alsohelp to slow down mobility decline.

Physical activity is known to decline with age in olderpersons.18 Therefore, whether change in physical activitywas associated with change in mobility performance mea-sured during the same time period was examined. With re-gard to sports activity, two findings should be noted. First,

participants who remained active in sports activities expe-rienced no decline in mobility performance. Second, par-ticipants who started sports activities between baseline andfollow-up experienced a smaller decline in mobility perfor-mance than those who remained inactive. These resultssuggest that older persons should try to keep up the goodhabit of exercising and that it is never too late to start ex-ercising. The associations were observed over a relativelyshort time frame (3 years). The long-term effect of physicalactivity on mobility performance may be even stronger.

Various mechanisms may explain the observed associ-ation between physical activity and mobility performance.Physical activity increases strength,30 flexibility, endur-ance, and coordination,31,32 which are well-known deter-minants of functional status. In addition, physical activitywill lower the risk for disabling diseases.1–3 Adjustment forbaseline chronic disease did not change our results, andsimilar results were observed in persons with and withoutchronic disease. This suggests that baseline health statuscannot explain our observations. A higher activity levelhas also been associated with a better disease risk profile,(e.g., lower resting heart rate and higher high density lipo-protein cholesterol,33 lower systolic blood pressure,22 anda better glucose homeostasis),34 which may lower the riskfor future diseases and subsequent decline in functionalstatus. Baseline subclinical disease was not considered inthe present study and may be an additional mechanism ex-plaining the observed associations.

Several limitations of the study should be discussed.Considering the design of the study, no causality of the ob-served associations can be inferred. Although physical in-activity has been shown to increase the risk for mobilitydecline, it is also intuitive that poor mobility will lead tophysical inactivity. Physical inactivity and poor functionalstatus may be mutually reinforcing, causing a downward

Table 4. Three-Year Change in Mobility Performance Score According to Change in Total Physical ActivityDuring Follow-up

Physical Activity Parameter

Change in Mobility Performance Score

n

Model 1* Model 2†

Mean � Standard Error

Change in total physical activity (hours/day)��1.5‡ 433 �0.54 � 0.09 �0.45 � 0.09�1.5 to �0.5 404 �0.44 � 0.08 �0.40 � 0.08�0.5 to 0 334 �0.33 � 0.08 �0.39 � 0.090.0 to 1.0 485 �0.05 � 0.07� �0.09 � 0.07§

�1.0 392 �0.24 � 0.08§ �0.22 � 0.08§

P-value for trend .0004 .005Change in total physical activity (kcal/day)

��300‡ 399 �0.49 � 0.09 �0.45 � 0.09�300 to �100 382 �0.36 � 0.08 �0.35 � 0.08�100 to 0 309 �0.31 � 0.09 �0.32 � 0.090 to 200 432 �0.11 � 0.08§ �0.16 � 0.07§

�200 419 �0.22 � 0.08§ �0.22 � 0.08P-value for trend .008 .03

*Adjusted for age, sex, and baseline mobility performance score.†Additionally adjusted for education, smoking, alcohol consumption, depression, cognitive function, and chronic disease.‡Reference.§P � .05; �P � .001 versus reference category.

JAGS NOVEMBER 2002–VOL. 50, NO. 11 PHYSICAL ACTIVITY AND CHANGE IN MOBILITY 1781

spiral ultimately leading to disability and loss of indepen-dence in older persons. Nevertheless, Miller et al.35 haveshown that the relationship between physical activity andchange in self-reported functional limitations is not mod-erated by participants’ preexisting levels of functional lim-itations. Furthermore, physical activity was associated witha lower risk of dependence after 8 years of follow-up inolder persons with self-reported mobility limitations.36

These results suggests that poor mobility does not neces-sarily lead to inactivity and that it is possible to preventfurther disability with physical exercise. A second limita-tion of the study is that about 25% of the respondents re-ported that their physical activity pattern of the previous 2weeks was not representative of the rest of the year. Nev-ertheless, excluding these respondents did not changethe conclusions of the study. Nor were the conclusionschanged when a dummy variable was added to the regres-sion models indicating whether the activity pattern wasrepresentative or not. A third limitation of our study is thatno objective method was used to measure physical activitysuch as the doubly labeled water method or accelerome-ters. However, these methods do not provide informationon the type or the intensity of activities and thereforewould not have allowed us to distinguish sports and non-sports activities.

In conclusion, physical activity, and especially a regu-larly active lifestyle, may slow the decline in mobility per-formance in older men and women. A beneficial effect wasobserved for sports and nonsports activities, independentof the presence of chronic disease.

REFERENCES1. Wannamethee SG, Shaper AG, Walker M. Changes in physical activity, mor-

tality, and incidence of coronary heart disease in older men. Lancet 1998;351:1603–1608.

2. Bijnen FC, Caspersen CJ, Feskens EJ et al. Physical activity and 10-year mor-tality from cardiovascular diseases and all causes: The Zutphen ElderlyStudy. Arch Intern Med 1998;158:1499–1505.

3. Hu FB, Leitzmann MF, Stampfer MJ et al. Physical activity and televisionwatching in relation to risk for type 2 diabetes mellitus in men. Arch InternMed 2001;161:1542–1548.

4. Krebs DE, Jette AM, Assman SF. Moderate exercise improves gait stability indisabled elders. Arch Phys Med Rehabil 1998;79:1489–1495.

5. Jette AM, Lachman M, Giogetti MM et al. Exercise—it’s never too late: Thestrong-for-life program. Am J Public Health 1999;89:66–72.

6. Taaffe DR, Duret C, Wheller S et al. Once-weekly resistance exercise im-proves muscle strength and neuromuscular performance in older adults. J AmGeriatr Soc 1999;47:1208–1214.

7. Lord SR, Ward JA, Williams P et al. The effect of a 12-month exercise trialon balance, strength, and falls in older women: A randomized controlledtrial. J Am Geriatr Soc 1995;43:1198–1206.

8. Brown M, Sinacore DR, Ehsani AA et al. Low-intensity exercise as a modifierof physically frailty in older adults. Arch Phys Med Rehabil 2000;81:960–965.

9. Rubenstein LZ, Josephson KR, Trueblood PR et al. Effects of a group exer-cise program on strength, mobility, and falls among fall-prone elderly men. JGerontol A Biol Sci Med Sci 2000;55A:M317–M321.

10. LaCroix AZ, Guralnik JM, Berkman LF et al. Maintaining mobility in latelife II. Am J Epidemiol 1993;137:858–869.

11. Simonsick EM, Lafferty ME, Phillips CL et al. Risk due to inactivity in phys-ically capable older adults. Am J Public Health 1993;83:1443–1450.

12. Clark DO. The effect of walking on lower body disability among olderblacks and whites. Am J Public Health 1996;86:57–61.

13. Kempen GIJM, Steverink N, Ormel J et al. The assessment of ADL amongfrail elderly in an interview survey: Self-report versus performance-based

tests and determinants of discrepancies. J Gerontol B Psychol Sci Soc Sci1996;51B:254–260.

14. Hoeymans N, Wouters ERCM, Feskens EJM et al. Reproducibility of perfor-mance-based and self-reported measures of functional status. J Gerontol ABiol Sci Med Sci 1997;52A:M363–M368.

15. Merrill SS, Seeman TE, Kasl SV et al. Gender differences in the comparisonof self-reported disability and performance measures. J Gerontol A Biol SciMed Sci 1997;52A:M19–M26.

16. Guralnik JM, Simonsick EM, Ferrucci L et al. A short physical performancebattery assessing lower extremity function. Association with self-reporteddisability and prediction of mortality and nursing home admission. J Geron-tol A Biol Sci Med Sci 1994;49A:M85–M94.

17. Seeman TE, Berkman LF, Charpentier PA et al. Behavioral and psychosocialpredictors of physical performance: MacArthur Studies on Successful Aging.J Gerontol A Biol Sci Med Sci 1995;50A:M177–M183.

18. Bijnen FC, Feskens EJ, Caspersen CJ et al. Age, period, and cohort effects onphysical activity among elderly men during 10 years of follow-up: The Zut-phen Elderly Study. J Gerontol A Biol Sci Med Sci 1998;53A:M235–M241.

19. Deeg DJH, Westendorp-de Serière M, eds. Autonomy and Well-Being in theAging Population I. Report from the Longitudinal Aging Study Amsterdam1992–1993. Amsterdam, the Netherlands: VU University Press, 1994.

20. Deeg DJH, Beekman ATF, Kriegsman DMW, eds. Autonomy and Well-Beingin the Aging Population II. Report from the Longitudinal Aging Study Am-sterdam 1992–1996. Amsterdam, the Netherlands: VU University Press, 1998.

21. Voorrips LE, Ravelli ACJ, Dongelmans PCA et al. A physical activity ques-tionnaire for the elderly. Med Sci Sports Exerc 1991;23:974–979.

22. Caspersen CJ, Bloemberg BPM, Saris WHM et al. The prevalence of selectedphysical activities and their relation with coronary heart disease risk factors inelderly men. The Zutphen Study, 1985. Am J Epidemiol 1991;133:1078–1092.

23. Visser M, Launer LJ, Deurenberg P et al. Total and sports activity in oldermen and women: Relation with body fat distribution. Am J Epidemiol 1997;145:752–761.

24. Ainsworth BE, Haskell WL, Leon AS et al. Compendium of physical activi-ties: Classification of energy costs of human physical activities. Med SciSports Exerc 1993;25:71–80.

25. Guralnik JM, Ferrucci L, Simonsick EM et al. Lower-extremity function inpersons over the age of 70 years as a predictor of subsequent disability. NEngl J Med 1995;332:556–561.

26. Kriegsman DM, Deeg DJ, van Eijk JT et al. Do disease specific characteristicsadd to the explanation of mobility limitations in patients with differentchronic diseases? A study in The Netherlands. J Epidemiol CommunityHealth 1997;51:676–685.

27. Radloff LS. The CES-D scale: A self-report depression scale for research inthe general population. Appl Psych Meas 1977;1:385–401.

28. Beekman ATF, van Limbeek J, Deeg DJH et al. Screening for depression inthe elderly in the community. Using the Center for Epidemiologic Studies De-pression Scale (CES-D) in the Netherlands [in Dutch]. Tijdschrift voor Ger-ontologie en Geriatrie 1994;25:95–103.

29. Folstein MF, Folstein SE, McHugh PRA. “Mini-Mental State.” A practicalmethod for grading the cognitive state of patients for the clinician. J PsychiatrRes 1975;12:189–198.

30. Rantanen T, Era P, Heikkinen E. Physical activity and the changes in maxi-mal isometric strength in men and women from the age of 75–80 years. J AmGeriatr Soc 1997;45:1439–1445.

31. Wood RH, Reyes R, Welsch MA et al. Concurrent cardiovascular and resis-tance training in healthy older adults. Med Sci Sports Exerc 2001;33:1751–1758.

32. Lazowski DA, Ecclestone NA, Myers AM et al. A randomized outcome eval-uation of group exercise programs in long-term care institutions. J GerontolA Biol Sci Med Sci 1999;54A:M621–M628.

33. Young DR, Haskell WL, Jatulis DE et al. Associations between changes inphysical activity and risk factors for coronary heart disease in a community-based sample of men and women. The Stanford Five-City Project. Am J Epi-demiol 1993;138:205–216.

34. Baan CA, Stolk RP, Grobbee DE et al. Physical activity in elderly subjectswith impaired glucose tolerance and newly diagnosed diabetes mellitus. Am JEpidemiol 1999;149:219–227.

35. Miller ME, Rejeski WJ, Reboussin BA et al. Physical activity, functional lim-itations, and disability in older adults. J Am Geriatr Soc 2000;48:1264–1272.

36. Hirvensalo M, Rantanen T, Heikkinen E. Mobility difficulties and physicalactivity as predictors of mortality and loss of independence in the commu-nity-living older population. J Am Geriatr Soc 2000;48:493–498.