The results from our large, prospective population-based cohort study showed a small, non-significant, trend of higher risk of knee replacement due to OA with higher leisure time physical activity, assessed in men and women at a mean age of 58 years. For hip OA, an opposite trend was found, and for women, higher leisure time physical activity was associated with a reduced risk of hip replacement. Walking was associated with a reduced risk for hip replacement, specifically for women. Our findings indicate a possible difference in effects of physical activity on the risk of knee and hip replacement due to OA. Walking, which involves low joint loading, may have a protective role for hip replacement. The possible gender difference in effects of physical activity, denote that men and women should be analyzed separately.
There was a trend of higher risk for knee replacement due to OA with higher leisure time physical activity over 11 years. However, the results were non-significant, and the risk for knee replacement increased with higher physical activity by a factor less than two (adjusted RRs between 1.08 and 1.44), indicating that leisure time physical activity was not a major risk factor for knee replacement/osteotomy. Wang et al.  recently reported that increasing levels of leisure time physical activity were associated with an increased risk of knee replacement due to OA over approximately 10 years. The effect seemed to be related to vigorous activity but not to less vigorous activity or walking . Although their values reached statistical significance, the effect was small (HRs between 1.08 and 1.47)  and similar to that observed in our study. Previous case–control studies reported either an increased risk of knee replacement with high exposure to sports , or a reduced risk with increasing cumulative hours of recreational physical activity . Longitudinal prospective studies, with similar follow-up time as in our study, reported that physical activity appeared to have little effect on OA risk in middle-aged and older subjects, measured as radiographic or symptomatic knee OA , or self-reported physician-diagnosed knee or hip OA .
For the hip, there was a small, non-significant, trend of lower risk for OA with higher leisure time physical activity over 11 years, indicating that leisure time physical activity was not a risk factor for hip replacement. This is in line with previous prospective cohort studies, reporting that leisure time physical activity was not a risk factor for hip replacement due to OA [9–11]. The values in the present study reached statistical significance for women, showing a 34% lower risk of hip OA in those with the highest compared with the lowest physical activity (lowest RR 0.66, lowest CI 0.48). This indicates a protective role of leisure time physical activity for the incidence of hip replacement in women. Further studies are needed to confirm this possible gender difference.
Walking was associated with a reduced risk for hip replacement, specifically in women. Physical activity involving higher joint loads (soccer, weight lifting), regular or intense exercise (elite, ex-athlete, physical education teachers), and frequent knee-bending activities may be associated with an increased risk for OA [1, 2, 21–24], although this risk is less than that for previous joint injury and overweight . Physical activity involving lower joint loads, (long-distance running, swimming, walking, golf), and moderate exercise do not appear to increase the risk of knee or hip OA development or progression, or may even have a protective role [2, 4, 5, 8, 22, 24, 25]. Participation in sports activities, such as ball games, was not particularly common in our cohort, which is likely related to the age of the participants (mean 58 years at baseline). Therefore, comparison of activities involving high or low joint loading could not be performed.
The large size of the cohort and the prospective design are major strengths of the present study . The participants were representative of the eligible population, and those who had been surgically treated due to hip or knee OA before the baseline examination were excluded. We used a case definition of knee replacement (or high tibial osteotomy) or hip replacement due to OA; a definition which is highly related to the disease burden of OA. A limitation of this definition is that only a small proportion of the total OA population undergoes knee or hip replacement (“the tip of the iceberg”). However, it has the advantage of an unambiguous relationship with the OA disease burden. So, from our results, we can draw conclusions for this group only, and not for those with possible symptomatic or radiographic OA and less severe disease, or those with severe OA that for a variety of reasons (e.g., not willing to consider TJR) have not undergone joint replacement . The presence and effects of such patient selection bias cannot be ruled out, but it is difficult to value whether this would cause an over- or underestimation of our results.
The Swedish hospital discharge register was used for case-retrieval. This register was active during the entire follow-up period and covers all Swedish hospitals. A validation study reported that at least 95% of primary knee and hip joint replacements were included in the register, and that the diagnostic misclassification was about 5% for hip replacement . Primary OA was the diagnosis for more than 85% and 75% of all primary knee and hip joint replacements, respectively, in Sweden during the follow-up period . We therefore believe that the bias in our study due to misclassification of OA is small, likely not influencing the results.
The progression from symptomatic OA to a joint replacement occurs over approximately a decade , suggesting sufficient follow-up time in our study. However, some persons entering the present study may have had early OA symptoms, possibly limiting their physical activity. This may have caused an underestimation of our results.
Obesity is a risk factor for OA of the knee or hip. In the same cohort as that used in the present study, the risk for knee or hip OA was 4–7 and 2–3 times higher, respectively, in people with obesity compared to those with normal weight . BMI was well documented, and adjusted for in the present study.
There are a number of potential limitations to our study. History of joint trauma was not recorded at the time of the baseline examination. Previous knee injury is a well known risk factor for knee OA , and when adjusted for, regular physical activity may protect against knee OA . Also for hip OA, a history of hip joint trauma is considered a risk factor [2, 24]. Because hip injuries are less common than knee injuries, the impact of previous hip injury is likely small in the current study. However, we cannot rule out that previous joint injury, particularly to the knee, may have affected our results. This may constitute one explanation for the opposing trends on the effects of physical activity and the risk of future knee and hip replacement. This is in line with a previous prospective cohort study, reporting that higher levels of physical activity were associated with an increased risk of knee replacement due to OA, whereas no such association was found for the hip . The different anatomical characteristics and function of the hip and knee joints may comprise another reason for the possible different effects of physical activity on knee and hip replacement.
Another weakness of our study, and of previous studies [9–11], is that sports participation earlier in life was not adjusted for. High exposure to sports earlier in life was related to an increased risk for knee or hip OA, defined as joint replacement [19, 21] radiographic OA [31, 32], or medically-diagnosed self-reported hip OA , implying that lifetime physical activity should be adjusted for.
The influence of occupational physical activity on knee or hip OA is more consistent than that for leisure time physical activity. Systematic reviews conclude some evidence of an association between high work load and knee or hip OA [1, 2, 33, 34]. Since we do not have data for the participants’ previous occupational physical activity, we could not control for this in the analysis. Because an urban population was included, there were likely few participants with high risk occupations, such as farming [35, 36]. The interplay between physical activity at work and during leisure time and the association with knee and hip OA may be important to consider in future studies. Men with high exposure to both sports activities and occupational/leisure time physical work load may have an increased risk of knee replacement due to OA . It was also reported that men whose jobs included both knee bending and higher physical demands had an increased risk of radiographic incident OA .
Self-report surveys to measure physical activity are associated with some limitations, such as, over-reporting, reduced accuracy for moderate physical activity, and moderate reproducibility . Accelerometry provides an objective and reliable measure of the frequency, duration and intensity of physical activity, but is impractical to use in large cohorts . Although validity and reliability were reported for the questionnaire used [16, 17], these studies were small and further validation may be required.
Studies of this cohort  and other cohorts [38–40] report that higher physical activity, assessed by the questionnaire that we used, or a 3–point scale, was associated with a lower risk of cardiovascular diseases, implying that these questionnaires are sensitive in detecting important health risks. These questionnaires, and other questionnaires that are used in patients with hip or knee OA , measure the aerobic aspect of physical activity, which is relevant for cardiovascular diseases but may not be appropriate for OA. Because current instruments may not have adequate measurement properties for measuring physical activity as risk factor for OA development , this may be a reason for failing to find a consistent association with knee or hip OA in our study and in previous prospective cohort studies [6, 7, 9–11]. Questionnaires accurately assessing physical activity in terms of type, frequency, intensity, and joint load seem to be relevant and important for studying the association with OA. Such questionnaires may also determine whether the association of physical activity with OA is linear or not, as it may be assumed that both very high and very low joint loads are risk factors for OA.