Effect of recreational physical activities on the development of knee osteoarthritis in older adults of different weights: The Framingham Study




To evaluate the long-term effect of recreational exercise on the development of knee osteoarthritis (OA) in a community-based cohort of older adults, many of whom were overweight or obese.


Subjects were asked about recreational activities including walking or jogging for exercise and working up a sweat, and were asked to compare their activity levels with others. Subjects were then asked about knee pain and weight-bearing anteroposterior and lateral knee radiographs were obtained. Approximately 9 years later, subjects were reexamined for OA. Radiographs were read for OA features in both tibiofemoral and patellofemoral compartments and were scored for tibiofemoral joint space narrowing. To evaluate incident OA, we excluded knees with OA at baseline for all analyses and focused on 3 knee-specific outcomes: incident radiographic OA, symptomatic OA, and tibiofemoral joint space loss. After adjusting for age, sex, body mass index (BMI), knee injury history, and correlation between knees, we evaluated the association of each recreational activity with OA development.


A total of 1,279 subjects underwent both baseline and followup examinations (mean age at baseline 53.2 years). Neither recreational walking, jogging, frequent working up a sweat, nor high activity levels relative to peers were associated with a decrease or increase in risk of OA. Joint space loss was also unaffected by activity. Persons with BMI above the median (27.7 kg/m2 for men and 25.7 kg/m2 for women; mean BMI >30 kg/m2 for both) had no increases in risk of OA by different type of activity.


Among middle-aged and elderly persons without knee OA, many of whom were overweight, recreational exercise neither protects against nor increases risk of knee OA.


Regular exercise is recommended for middle-aged and older persons. The effect of regular exercise on the development of osteoarthritis (OA) in older persons, especially those who are overweight, is unclear.

OA involves all structures within the joint, but cartilage loss is the pathologic hallmark of disease. Because dynamic loading of cartilage has a trophic effect on cartilage (1), frequent dynamic loading, especially in a healthy range, would cause cartilage to become thicker (2), and if cartilage thickness prevents disease, OA might occur less frequently in individuals engaging in regular activity. In most animal studies, weight-bearing exercise has been shown to protect against the development of OA (3, 4), but in one strain of centrally stimulated animals that run almost constantly, there is a high rate of cartilage loss versus control animals (2, 3).

Studies on cartilage thickness in exercising versus nonexercising humans have reported varied results. In a recent short-term trial, Roos and Dahlberg (5) reported that individuals without OA who were randomized to an exercise regimen had a healthier distribution of proteoglycans within cartilage as demonstrated by imaging studies compared with sedentary individuals not participating in exercise. This finding suggests an overall protective effect on the development of OA over a longer period. Triathletes have thicker cartilage in their patellae, but thinner cartilage in their medial femoral condyles, than do age-matched inactive study volunteers (6). In children, vigorous self-reported activity in the previous 2 weeks was associated with an accretion of cartilage compared with children with no reports of vigorous activity (7).

Obesity is a major risk factor for knee OA, and its effect is thought to be due to increased loading (5). Weight-bearing activity and recreational activity may be injurious to the knees of persons who are overweight, but this issue has not been well studied. Although many studies have attempted to determine whether physical activity prevents (or causes) knee OA, the answer is not known, in part because of limitations of studies addressing this issue. Many studies have been cross-sectional with recollected activity (e.g., Can OA now be tied to recalled activity?) (8). Other studies with prospective designs have used self-reported arthritis (9), an entity with questionable validity at best. Individuals who are healthy may exercise more, may visit doctors for health problems less, and may not get the opportunity to have their OA diagnosed.

Few prospective studies exist in which persons have been asked about activity and then followed to see who develops OA. Followup needs to be long enough for OA to develop. Of such studies, there have been 2 different results, each represented by 2 studies. One study tracked young and middle-aged runners for up to 10 years and found no increase in knee OA by serial radiograph (10). Intriguingly, in the most recent followup, runners did not show joint space loss, whereas nonrunners did, a difference that was not significant but suggested that running may preserve cartilage thickness. Another study (11) demonstrated no effect of physical activity on disease occurrence in middle-aged women, although numbers were small, followup was short (4 years), and confidence limits were wide, but results actually suggested that walking protected against joint space loss (odds ratio [OR] 0.60, 95% confidence interval [95% CI] 0.22–1.71). These 2 aforementioned studies not only found no increased risk of OA, but also suggested the possibility of disease protection. In contrast, another study found that elderly adults who self reported high levels of heavy physical activity had an increased risk of radiographic knee OA (12). In this latter study, the only study with a large number of overweight persons, body mass index (BMI) above the median was found to further increase the risk of knee OA among those who exercised. Lastly, a large elderly cohort followed for the development of hip OA (13) demonstrated that elderly women who had been more active in middle age had more hip OA by radiograph.

Discordance of study results has arisen in part because studies were often small and followed insufficient numbers of overweight elderly adults. Studies used nonoptimal questions about specific physical activities so that the relationship of popular physical activities, such as recreational walking, to knee OA remained unknown; and, in one case, focused on hips rather than knees, where effects of exercise may be different. Also, different definitions of OA have been used in different studies.

Ideally, a study examining the effects of physical activity on OA should examine effects on both structural disease and symptomatic disease. Also, studies examining OA of the knee have focused on tibiofemoral disease, whereas much symptomatic disease occurs in the patellofemoral compartment, which should also be examined.

We conducted a longitudinal study of persons from the community who answered a questionnaire about physical activity, underwent knee radiographs, and answered questions about knee symptoms and then had the same knee evaluation ∼9 years later. Our population was older and contained many overweight subjects, permitting us to examine effects of moderate activity on OA incidence and test whether the effect was different in overweight persons.



The Framingham Offspring cohort, assembled in the early 1970s, consists of the sons and daughters of the original Framingham cohort and the spouses of these offspring. Offspring subjects were recruited as part of a study of the inheritance of OA (14). During a callback visit to offspring examination 5 (1993–1994), we obtained a weight-bearing, fully extended radiograph of both knees using a standardized protocol that included outlines of the feet to keep the rotation of the knee constant at followup radiograph. Films were obtained at 0° and at 6° caudad, and the best view of these 2 was selected for the followup examination. Weight-bearing, semiflexed lateral radiographs were also obtained using a protocol recently described (15). Subjects were also asked about knee symptoms, specifically whether they experienced pain, aching, or stiffness in either knee on most days of a recent month. If they answered yes, they were asked which knee was painful.

As part of offspring examination 5, which preceded the present OA examination by 1–2 years, subjects were asked questions about recent physical activity. The focus was on identifying and quantifying activities that subjects engaged in on a regular basis. Questions included, “How many times per week do you engage in intense physical activity (enough to work up a sweat)?” and “How would you compare your activity level to others your age? (Less active? Same as others? More active than others?)” In addition, subjects were asked whether they walked for exercise on a regular basis, and if they said yes they were asked how many miles per session and how many times per week. Similarly, they were asked if they jogged or ran as part of a physical activity program. If they said yes, they were asked the frequency of their running.

In 2002–2005, the same offspring were called back for a followup examination. Subjects underwent the same knee radiographs and the same questions about knee pain were asked, but subjects were not surveyed again regarding activity. They were also asked about history of knee injury and asked to specify which knee had been injured. At both baseline and followup examinations, subjects were weighed using a balance beam scale without heavy clothing or objects in their pockets, and height was assessed on a stadiometer.

After subjects had completed the followup examination, their baseline and followup radiographs were read, paired, and unblinded to sequence by 2 study readers, a bone and joint radiologist (PA) and a rheumatologist (BS), working independently. Readers were blinded to all information about subjects including age, weight, and physical activity level. Readers evaluated the Kellgren and Lawrence (K/L) grade (16) of each knee at both time points and evaluated individual features and whether these features had changed over the followup. Scoring of features was done using the Osteoarthritis Research Society International Atlas (17) for the anteroposterior (AP) view. For the lateral view, we scored features using an atlas created for this project.

After each reader had scored the radiographs, we evaluated if there were important disagreements in the scoring of change. We defined important disagreements as those in scoring joint space loss from baseline to followup in the tibiofemoral or patellofemoral joint on either the AP or the lateral view. For knees where there was important disagreement, we held adjudication sessions to arrive at a final reading. Adjudication sessions were attended by the 2 readers and a third experienced reader (DTF) and each knee was scored by consensus. All evaluations of radiographic progression in the tibiofemoral or the patellofemoral joint space (the latter on the lateral view) were based either on unanimity of opinion or on adjudication of readings.

For K/L grade and its change, we used readings from the bone and joint radiologist, but if adjudication of joint space change altered the appropriateness of the K/L score, then K/L score was also adjudicated. For the bone and joint radiologist, intrareader kappa for K/L grade was 0.82 and interreader kappa was 0.74 (both P < 0.001).

Statistical analysis.

Initially we evaluated which subjects came to the followup, how many were lost to followup, and whether those lost to followup were similar to those who came to the examination. The main focus of this study was the relationship of physical activity assessed at offspring examination 5 with the change in status from the baseline OA examination to the followup period. We looked at 3 different measures of change in OA. First, we examined whether the patient had experienced incident knee OA, which we defined as present when a knee with a K/L grade of 0 or 1 at baseline in the tibiofemoral joint on the AP view progressed to a K/L grade 2 or higher at followup. We also classified a knee as having incident OA if there was new patellofemoral OA using criteria from Felson et al (18). No knees that fit this category had knee replacements at followup. We included patellofemoral OA in our characterization of incident OA but not in terms of progressive narrowing, because change in patellofemoral narrowing is very difficult to read accurately on serial lateral radiographs.

The second radiographic definition focused on tibiofemoral joint space loss. We and others have shown that joint space loss is related to cartilage loss (19). We defined the knee as having had joint space loss when there was a 1-grade change (e.g., 0–1 or 1–2) on either the lateral view or the AP view in the tibiofemoral joint from baseline to followup. Similar to other analyses, we focused on knees without OA at baseline (excluding knees with K/L grade 2 or higher at baseline). We have recently validated joint space progression on the lateral view (15).

Our last arthritis outcome was incident symptomatic OA. As in past work (20), we defined incident symptomatic OA as a knee that did not have the combination of symptoms and radiographic disease at baseline but developed that combination by followup. Therefore, all knees at followup with incident symptomatic OA had both frequent symptoms (knee pain, aching, or stiffness on most days) and a radiograph showing definite OA in either the tibiofemoral or patellofemoral compartment of that knee.

For each of these outcomes, we defined the eligible population at baseline as knees that did not have the specified outcome. For example, we examined symptomatic OA in knees that did not have the combination of symptoms and radiographic OA at baseline. For incident joint space loss, we looked at knees without joint space narrowing at baseline but that also did not have radiographic OA (excluded K/L grade 2 or higher) at baseline. All analyses were knee specific. Knees that already had prevalent disease at baseline were excluded from these analyses, therefore subjects could have contributed anywhere from 0 to 2 knees depending on whether they had baseline disease in a given knee or not.

The main analysis focused on the relationship of exercise to incident OA using the above definitions of OA. For each of these definitions, we started with bivariate analyses in which we divided activity frequency in tertiles based on the overall distribution and then examined the percentage of subjects with incident disease in these tertiles of activity. After bivariate analysis, we performed logistic regression analysis; we adjusted for age, BMI (at baseline), knee injury (identified at the followup examination), sex, and the correlation between the 2 knees using generalized estimating equations. We were concerned that persons with OA in one knee could have altered their physical activity as a consequence and that we would be studying the incidence of disease in their other knee. Therefore, in sensitivity analyses, we excluded knees with unilateral OA and examined only subjects without OA in either knee. Results were the same as those presented here. For this study, written informed consent was obtained from subjects at both baseline and followup, and protocols were approved by the Boston University Medical Campus Institutional Review Board.


Of 1,705 subjects seen at the initial examination, 1,279 (75.0%) underwent radiographs as part of the followup examination, with a mean ± SD interval of 8.75 ± 1.04 years between baseline and followup radiograph. Subjects followed versus those not followed did not differ in age at baseline (mean age 53.2 years [range 26–81] versus 54.2 years [range 29–78]) or in average BMI (mean 27.4 kg/m2 versus 27.8 kg/m2). The prevalence of knee pain on most days at the baseline examination was also comparable (26.0% versus 27.4%), although subjects who attended the followup were less likely to be men than those who did not attend (44.0% versus 53.3%).

Of 2,259 knees eligible for incident radiographic OA, 215 (9.5%) developed incident disease, with most (181 knees) developing disease in the tibiofemoral compartment. Of the knees eligible to develop symptomatic OA, there were 173 (7.2%) incident cases. Lastly, of the knees eligible for incident narrowing, 222 (9.8%) developed narrowing over followup.

Most subjects reported walking regularly for exercise (Table 1), but neither the group walking up to 6 miles nor the group walking ≥6 miles per week had an increase or decrease in their risk of OA. Specifically, neither risk of incident OA, incident symptomatic OA, nor joint space loss was affected by walking for exercise. Generally, the risk of disease was close to null (e.g., adjusted OR 1.10 for incident OA in those who walked the most), with confidence bounds narrow enough to suggest that neither a substantial increase nor a decrease in risk was being missed. In additional analyses, we divided the group that walked the most for exercise (≥6 miles/week) at the median and looked at those with the greatest amount of walking for exercise (≥9 miles/week), finding that they also experienced no increase or decrease in disease risk. When we examined working up a sweat (Table 2) or physical activity of subjects compared with others of the same age (Table 3), our findings were similar: physical activity appeared to have little effect on OA risk.

Table 1. Walking for exercise and incident OA of the knee in Framingham Study subjects*
 All subjectsSubjects at or above median BMISubjects below median BMI
Cumulative incidence of OAAdjusted OR (95% CI)Cumulative incidence of OAAdjusted OR (95% CI)Cumulative incidence of OAAdjusted OR (95% CI)
  • *

    Values are the number/total number (percentage) unless otherwise indicated. OA = osteoarthritis; BMI = body mass index; OR = odds ratio; 95% CI = 95% confidence interval.

  • Adjusted for age, BMI, history of knee injury, and sex.

  • Referent.

Radiographic OA
 No walking for exercise104/1,128 (9.2)178/567 (13.2)129/561 (5.2)1
 Walk <6 miles/week35/385 (9.1)0.99 (0.60–1.62)25/164 (15.2)1.11 (0.61–2.03)10/221 (4.5)0.86 (0.34–2.17)
 Walk ≥6 miles/week68/674 (10.1)1.10 (0.73–1.66)38/298 (12.8)0.95 (0.55–1.62)30/376 (8.0)1.44 (0.72–2.86)
Symptomatic OA
 No walking for exercise91/1,214 (7.5)172/623 (11.6)118/596 (3.1)1
 Walk <6 miles/week30/411 (7.3)0.96 (0.57–1.62)16/176 (9.1)0.69 (0.37–1.30)14/235 (6.0)1.93 (0.78–4.78)
 Walk ≥6 miles/week46/721 (6.4)0.78 (0.49–1.24)32/328 (9.8)0.84 (0.37–1.92)14/393 (3.6)0.78 (0.44–1.38)
Joint space loss
 No walking for exercise111/1,133 (9.8)173/571 (12.8)138/562 (6.8)1
 Walk <6 miles/week35/387 (9.0)0.93 (0.56–1.54)22/165 (13.3)1.01 (0.52–1.95)13/222 (5.9)0.83 (0.37–1.85)
 Walk ≥6 miles/week65/676 (9.6)0.95 (0.62–1.45)35/299 (11.7)0.96 (0.55–1.65)30/377 (8.0)0.99 (0.50–1.96)
Table 2. Working up a sweat and incident OA of the knee in Framingham subjects*
 All subjectsSubjects at or above median BMISubjects below median BMI
Cumulative incidence of OAAdjusted OR (95% CI)Cumulative incidence of OAAdjusted OR (95% CI)Cumulative incidence of OAAdjusted OR (95% CI)
  • *

    Values are the number/total number (percentage) unless otherwise indicated. See Table 1 for definitions.

  • Adjusted for age, BMI, history of knee injury, and sex.

  • Referent.

Radiographic OA
 Do not work up a sweat51/576 (8.9)132/265 (12.1)119/312 (6.1)1
 Sweat <3 times/week69/692 (10.0)1.24 (0.77–2.00)53/353 (15.0)1.48 (0.81–2.71)16/339 (4.7)0.84 (0.7–1.81)
 Sweat ≥3 times/week87/896 (9.7)1.15 (0.72–1.82)54/405 (13.3)1.22 (0.67–2.21)53/491 (6.7)1.08 (0.53–2.22)
Symptomatic OA
 Do not work up a sweat41/613 (6.7)133/294 (11.2)18/319 (2.5)1
 Sweat <3 times/week57/742 (7.7)1.41 (0.82–2.44)43/381 (11.3)1.22 (0.66–2.27)14/361 (3.9)2.1 (0.65–7.00)
 Sweat ≥3 times/week70/968 (7.2)1.23 (0.72–2.10)45/445 (10.3)1.04 (0.55–1.96)23/522 (4.4)2.2 (0.76–6.19)
Joint space loss
 Do not work up a sweat47/577 (8.2)124/264 (9.1)123/313 (7.3)1
 Sweat <3 times/week68/697 (9.8)1.25 (0.78–2.00)50/356 (14.0)1.72 (0.94–3.17)18/341 (5.3)0.75 (0.35–1.61)
 Sweat ≥3 times/week96/899 (10.7)1.29 (0.82–2.02)57/408 (14.0)1.57 (0.87–2.82)39/491 (7.9)1.03 (0.53–2.03)
Table 3. Activity of Framingham subjects compared with others of the same age and incident knee OA*
 All subjectsSubjects at or above median BMISubjects below median BMI
Cumulative incidence of OAAdjusted OR (95% CI)Cumulative incidence of OAAdjusted OR (95% CI)Cumulative incidence of OAAdjusted OR (95% CI)
  • *

    Values are the number/total number (percentage) unless otherwise indicated. See Table 1 for definitions.

  • Adjusted for age, BMI, history of knee injury, and sex.

  • Referent.

  • §

    Confidence intervals not determinable because of 0 cases in 1 group.

Radiographic OA
 Less active than others31/344 (9.0)0.72 (0.42–1.22)29/230 (12.6)0.81 (0.45–1.47)2/114 (1.8)0.32 (0.08–1.37)
 Same as others100/1,014 (9.9)169/485 (14.2)131/529 (5.9)1
 More than others79/832 (9.5)0.94 (0.63–1.40)41/323 (12.7)0.82 (0.48–1.40)38/509 (7.5)1.20 (0.65–2.21)
Symptomatic OA
 Less active than others25/358 (7.0)0.81 (0.46–1.43)25/244 (10.2)0.87 (0.49–1.55)0/114 (0)0§
 Same as others79/1,087 (7.3)147/536 (8.8)114/550 (2.5)1
 More than others63/901 (7.0)0.94 (0.60–1.47)38/353 (10.8)0.63 (0.35–1.16)32/548 (5.8)2.08§
Joint space loss
 Less active than others24/345 (7.0)0.53 (0.30–0.94)21/231 (9.1)0.59 (0.31–1.16)3/114 (2.6)0.39 (0.08–1.83)
 Same as others107/1,018 (10.5)168/489 (13.9)139/529 (7.4)1
 More than others85/836 (10.2)0.89 (0.60–1.31)44/324 (13.6)0.84 (0.49–1.42)41/512 (8.0)0.99 (0.56–1.79)

Only a few of our subjects (n = 68) reported jogging or running regularly for exercise. They too did not appear to be at increased risk of later knee OA. For example, for incident symptomatic OA, the adjusted OR of OA was 1.00, although because the numbers were small, our confidence bounds were wider (95% CI 0.27–3.68), and we could not exclude either a modest protective or increased risk of disease.

We next examined the effect of physical activity on OA in subjects above and below the median of sex-specific BMI. Among men, the group above the median had a mean ± SD BMI of 31.1 ± 3.4 kg/m2 and mean ± SD weight of 208.7 ± 26.0 pounds. For the women above the median, the mean ± SD BMI was 30.7 ± 5.0 kg/m2 and mean ± SD weight was 174.5 ± 31.0 pounds. For men below the median BMI of the cohort, the mean ± SD BMI was 25.2 ± 1.8 kg/m2 and the mean ± SD weight was 172.7 ± 17.7 pounds. For women below the median BMI, mean ± SD BMI was 22.7 ± 1.9 kg/m2 and mean ± SD weight was 130.8 ± 14.1 pounds. As shown in Table 1, we found no relationship of recreational walking to knee OA in overweight subjects. For example, the OR of walking at least 6 miles/week and subsequent incident knee OA was 0.95. Similar findings are shown for working up a sweat (Table 2) and activity level greater than peers (Table 3). Additional analyses in which we examined risk in each sex were no different from results reported here.


In a long-term followup study of a community-based sample without evidence of OA, we found no relationship between recreational walking, jogging, or other self-reported activity and the development of knee OA, regardless of how knee OA was defined. Even though overweight persons in our cohort had an increased risk of developing OA, physical activity did not contribute to this increased risk. However, despite suggestive evidence from small cohort studies that exercise may prevent joint space loss (10, 11), we found no evidence for such an effect. This suggests that in middle-aged and older adults who do not have OA, exercise does not protect against disease development.

Intriguing results about the effects of dynamic weight-bearing loading on knee joint cartilage in healthy persons have been published recently. First, Roos and Dahlberg (5) completed a trial comparing exercise with no exercise in persons at risk of OA and assessed gadolinium enhancement, a technique used to evaluate the concentration of proteoglycans in cartilage. Because early OA consists of a focal loss of proteoglycans, high concentrations would suggest qualitatively normal and healthy cartilage. Roos and Dahlberg (5) found that the exercise group had more normal-appearing cartilage based on proteoglycan distribution than did the nonexercise group, strongly suggesting that exercise would have beneficial qualitative effects on cartilage in persons at risk of disease. In another study, Muhlbauer and colleagues (6) compared triathletes with physically inactive controls in terms of cartilage thickness, reporting that although the patellar cartilage was thicker among the athletes, cartilage in the medial compartment, the site of most weight bearing, was actually thinner, without any evidence of OA having developed in either group.

Magnetic resonance imaging (MRI)–based studies examining effects of physical activity on cartilage thickness make a number of assumptions about cartilage and disease. First, they presume that diffuse cartilage thinning predisposes to OA, a supposition that is unproven. Second, they focus on average thickness as a measure of cartilage loss, when early disease may occur focally and may not be reflected in a measure of diffuse thinness. Lastly, they focus on cartilage to the exclusion of other joint structures, structures that may be affected by exercise. For example, the strength and conditioning of the muscles across the joint may be more important in preventing later OA than cartilage thickness.

Our inquiry, by contrast, attempted to examine all of the different ways in which OA might become manifest. By looking at incident radiographic disease, we chose a conventional definition of the development of structural disease defined using the K/L scale. We also tested symptomatic OA, which would correspond to clinical OA: the development of new, frequent knee symptoms combined with a radiograph showing OA. Lastly, we looked at joint space loss on the radiograph. Radiographic joint space loss has been reported in many studies as a good surrogate for cartilage loss (6, 21–23).

Evaluating joint space loss over time is challenging using radiographs. Although the fully extended AP weight-bearing radiograph is probably no longer acceptable as a single modality for evaluating joint space loss, our use of lateral radiographs to evaluate the tibiofemoral joint space has recently been validated compared with fluoroscopy and also correlates with MRI-based cartilage loss (19). The majority of joint space loss in our cohort was seen on both the lateral and AP views.

There have been 3 other longitudinal prospective cohort studies evaluating physical activity and the development of knee OA (10–12). One of these (12), from a different study sample in the Framingham Study (the original cohort), suggested that self-reported high levels of physical activity were associated with incident radiographic disease, especially in overweight persons. Although this increase in risk was impressive for heavy activity (OR 13.0 for both high BMI and heavy physical activity versus OR 2.1 for lowest tertile of BMI and heavy activity), the difference in risk by weight did not reach statistical significance, given the small numbers of cases. Our current results are based on a larger number of cases and on specific activities. The previous study from Framingham was conducted in a group where physical activity was not assessed in detail, so that it was not possible to evaluate the same questions as were posed here. In addition, the evaluation of knee OA was less detailed. The other prospective study, the Chingford Study (11), constituted a 4-year followup of a smaller group of subjects than reported here. Walking was associated with a diminished risk of joint space loss over 4 years, with an OR of 0.60 (95% CI 0.22–1.71). Confidence bounds of these results were wide, and we believe that our findings are more definitive. The last prospective cohort study focused on runners and age-matched controls (10), and although it found no increase in disease or joint space loss in runners, the numbers of tracked subjects were small.

It is common to use a 1-time survey to measure habitual physical activity in cohort studies. The Framingham survey administered to the Framingham offspring drew questions from validated questionnaires. Scores on this survey have been shown to directly correlate with high-density lipoprotein levels, lower heart rates, and lower BMI (24), suggesting the survey's validity. Generally, there are strengths and weaknesses to a survey approach to assessing physical activity (25, 26). Although surveys can predict important health outcomes, are feasible for use in large cohorts, and can assess activity over a long period, the reproducibility of surveys, depending on the instrument, is only moderate and they sometimes do not capture the activities accounting for a large proportion of the variance in activity levels, a matter not of great relevance to our focused inquiry.

Although there are many strengths of our study, there are a number of limitations. One is that we had no MRI imaging at the baseline evaluation. Second, we did not have enough joggers or runners to evaluate the effect of running on OA. Our results for symptomatic OA do not suggest any effect of running, but confidence bounds are wide.

In conclusion, walking for exercise and other recreational activities in older persons without knee OA do not affect these individuals' risk of developing OA, even if they are overweight. Although dynamic loading may have a trophic effect on cartilage, there is no measurable protective effect of recommended weight-bearing exercise on OA. Physical activity can be done safely without concerns that persons will develop knee OA as a consequence.


Dr. Felson had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study design. Drs. Felson and Zhang.

Acquisition of data. Drs. Felson, Clancy, Sack, and Aliabadi.

Analysis and interpretation of data. Drs. Felson, Niu, and Zhang.

Manuscript preparation. Dr. Felson.

Statistical analysis. Drs. Niu and Zhang.