Background
Osteoarthritis (OA) is ranked the 11
th highest contributor to global disability with the hip and knee representing the highest proportion of the OA burden [
1]. Joint replacement is a proxy outcome measure for severe, symptomatic end-stage OA [
2]. The frequency of these procedures is rising in Western countries [
3,
4]. In total, 12 920 hip and knee replacements were performed in Norway in 2013, which represented a 20 % increase from 2008 [
5]. The lifetime risk of symptomatic OA is also high and has been estimated to 25 % and 45 % for hip and knee, respectively [
6,
7]. The high prevalence of OA, its social costs, and its impact on physical function and quality of life underlines the importance of identifying factors influencing its management and prevention. Known risk factors for hip and knee OA include older age, obesity, heavy, physical workload, joint injury and physical activity [
8‐
10]. However, the relationship between leisure time physical activity (LPA) and the development of hip and knee OA [
10,
11] or future joint replacement [
12‐
15] has not yet been clearly defined. Previous cohort studies have found the risk for total knee replacement (TKR) to increase with high levels of LPA [
12,
15], while others have found no overall relationship to either TKR or total hip replacement (THR) [
13,
16]. Intensive exercise and sport participation have more consistently showed an increased risk of later THR and TKR in former athletes [
14,
17]. The inconsistent results from previous studies may to a large extent be related to differences in study design, source population, definition of LPA and definition of OA. Few prospective studies have in the same population compared the risk of THR or TKR in relation to LPA as the main exposure of interest [
12,
15]. The main objective of the present study was, therefore, to investigate the incidence and association between levels of LPA and the risk of severe OA, defined as THR or TKR, due to primary OA in a large population-based cohort.
Discussion
In this population-based cohort study we found that LPA was associated with increased risk of THR due to primary OA. For TKR, LPA was associated with increased risk for women only. The effect was first and foremost related to high LPA, as no significant association was observed for less vigorous LPA. This is, to our knowledge, the first study to present the association between LPA and the risk of THR and TKR in relation to different age groups.
In contrast to previous population-based cohort studies, we found an association between high LPA and subsequent THR [
12,
13,
15]. This suggests that mechanical load or abnormal stress induced by high LPA may affect the hip and knee in the same way, although the underlying anatomical structures differ. Our findings are however in line with studies that have found an increased risk of THR after sports participation [
14,
17]. The high incidence of THR in our study gave us statistical power to detect any potential associations, while other studies may have been underpowered [
13]. For TKR, we found an increased risk of related to high LPA for women only. Higher incidence [
25] and prevalence [
26,
27] of OA has been found in women as compared to men, especially after menopause. Inherent biological and hormonal differences offer a possible explanation for this variation between men and women. The incidence of both THR and TKR have also been reported to be higher in women than men in Norway [
28,
29], which agrees with the results in our study.
In an Australian cohort study, high LPA was associated with increased risk of TKR (HR 1.46, 95 % CI 1.13–1.87), but not for THR [
15]. However, they had no detailed information on occupational physical demands, which may have left residual confounding related to workload. The follow-up time in this study was in average 4.8 years (SD 0.7), which may have been somewhat short to follow the onset and progression of OA to joint replacement. As they propose, LPA may have different effects across the spectrum of disease from a healthy joint to severe end-stage OA [
15]. By the short follow-up they might have limited the associations to the effects of LPA in those with already OA.
Two previous Norwegian cohort studies found no association between LPA and THR or TKR [
13,
16]. However, in these two studies [
13,
16], LPA were not the main exposure of interest. Their definition of high LPA were highly sports related, requiring participation in hard training or competitions regularly and several times a week. The authors state that the small number of participants and joint replacements in the most active category may partly be a reason for not finding any effect of LPA on THR or TKR [
16].
Similarly, a Swedish cohort study found no consistent relationship between LPA and THR or TKR other than showing a protective role of high LPA for women on the risk of THR (HR 0.66, 95 % CI 0.48–0.89) [
12]. They had no information on workload, which may have affected the direction and magnitude of association. Physical workload, in comparison to LPA, has consistently shown an increased risk of hip and knee OA for both men and women [
30,
31]. We found a protective effect of moderate LPA among women on the risk of TKR in the age-adjusted analyses (HR 0.80, 95 % CI 0.66–0.96). However, this effect was attenuated and non-significant after adjusting for workload and BMI in the full model. It is possible, as proposed earlier [
12], that absence of adequate measurement tools to quantify LPA leads us to misclassify LPA in relation to OA. Self-reported measures may not capture the type and true intensity of LPA, although if validated they are often the most feasible to use in large cohort studies.
The association we found between LPA and joint replacement was first and foremost present for high LPA. We cannot rule out the possibility that participants with high LPA included a subgroup of individuals who engaged in intensive exercise and sports, beyond recreational activities. Previous studies have shown that those who exercise more regularly or intensely have an increased risk of hip and knee OA, although the risk seems to be highly related to previous joint injury as well as being overweight [
9,
32,
33]. Participation in specific sports, such as soccer, ice hockey, cross country skiing and team handball has shown to be associated with OA and later joint replacement [
14,
17,
34]. A knee injury to the anterior cruciate ligament, especially with a combined injury to the menisci, has been shown to result in a significantly higher risk of radiographic knee OA [
35]. Consequently, as we were interested in joint replacements due to primary OA, all joint replacements due to previous injury (e.g. sequela after ligament or menisci injury) and conditions other than primary OA were censored.
Michaelsson et al. [
14] investigated a cohort of long-distance cross-country skiers and found that participation in multiple races and faster finishing time was associated with increased risk of THR or TKR. They proposed that we might overestimate the impact of high LPA if previous injury is not sufficiently accounted for because these two can act together. However, adjustment for previous disease and injury did not substantially change their results. Another study investigated the prevalence of hip and knee OA and subsequent joint replacement in former male athletes. They found an increased odds ratio after adjusting previous injury in addition to age, BMI and occupational load [
17]. Thus, we may both over- and underestimate the effect of LPA by including injury into the analyses.
Age was a strong confounder in our analyses, which changed the direction and magnitude of risk. In the stratified analyses, we found no association of LPA and joint replacement for those at 60 years or older at baseline. With ageing come comorbidities which could act as contraindications for joint replacement, but also reduce the level of LPA. This is in line with previous studies in which LPA was found to have little effect on the risk of developing hip or knee OA in middle-aged and older people [
11,
36]. In the sample from the Framingham Offspring cohort without knee OA at baseline, they found that recreational activities like walking or jogging for exercise neither protected nor decreased the risk of radiographic or symptomatic knee OA, regardless of being overweight or not [
11]. Cheng et al. [
36] found that high LPA was associated with OA of the hip and knee only in young men (age 20–49).
Similar results have been found in studies investigating BMI, where weight gain in early adulthood (between age 20–40 at screening) was associated with increased risk of both THR and TKR [
37,
38]. The cumulative effect of excess bodyweight over several decades compared to those who gain weight later in life was proposed as one possible explanation for the increased risk of OA and later joint replacement [
39]. This may offer a potential explanation for LPA as well, suggesting that exposure at certain age may be more important than at other ages [
40]. Active individuals may assert the necessity for, and undergo, arthroplasty at a younger age to retain an active lifestyle both at work and in leisure time, and might be considered more eligible for surgery than older individuals due to less comorbidity. A delay of surgical intervention has shown to potentially have an impact on the success rate of surgery. Consequently, the number of operations performed in younger people has increased over the recent years [
41].
Strengths and limitations
The main strengths of our study include a large population-based cohort, data available from clinical measurements and questionnaires, long follow-up and a prospective design. However, we measured LPA at baseline only and had no information about change in LPA during follow-up or prior LPA that could have affected the association with joint replacement. As the LPA questions in HUNT did not include information on type of activity performed, we cannot say anything about the risk related to any specific activity or sport. Self-reporting LPA may be prone to overestimation compared to more objective measures of physical activity like accelerometers [
42], at the same time lack of precision in measuring LPA may have biased the effects towards null.
We used joint replacement as a proxy for severe OA of the hip and knee. The limitation of this case definition is that only a small portion of the total OA population undergoes joint replacement, therefore we are not able to explain the full burden of hip and knee OA. The majority of hospitals in Norway are publicly funded and free of charge for the patient. Consequently, we do not think that socio-economic status represents a significant confounder to the provision of, nor the seeking of surgery. We excluded participants with joint replacement before baseline, but we cannot be certain that we detected all THRs and TKRs prior to HUNT2. This might be especially relevant for some of the older participants with a joint replacement before the register was fully operational for hip and knee in 1989 or 1994, respectively. Differential misclassification may have occurred if in older participants with joint replacements prior to baseline were distributed unequally across LPA-groups. However, the number of undetected joint replacements in our study period is expected to be insignificant.
We performed separate sensitivity analyses to investigate reversed causation. After excluding participants with self-reported OA at baseline, the effect of LPA on THR and TKR became somewhat stronger. This may suggest reversed causation, where subjects with OA are more likely to report low levels of LPA at baseline. Consequently, our main results may be considered as a conservative estimate, as we may have underestimated the true effect of LPA. We also have to acknowledge that multiple testing might have given significant findings that actually were due to chance. However, it was only for high LPA that we found an increased risk of joint replacement. This relationship was consistent both in the main analyses and sensitivity analyses.
We did not have information about previous injury, which has been a limitation in previous cohort studies as well [
12,
15]. However, joint replacements secondary to injuries, e.g. ligament or menisci injuries, were censored in the analyses to protect us from misclassifying the diagnosis of the outcome (primary OA). If previous injury is not sufficiently accounted for, as mentioned, we might have over- or underestimated of the true risk of LPA. However, if LPA causes injury which subsequently results in OA and joint replacement, then injury is more likely a mediator and adjustments should not be made for the total effect of LPA.
Participants who were excluded from the analyses because of missing data on LPA were older, more often women, and had somewhat higher BMI and more comorbidities at baseline compared to those included in the analyses. A great proportion of participants who were missing data on LPA were also missing workload data (71 %). These findings are in line with previous non-participation studies from HUNT [
43,
44]. Participants with LPA and workload data were in contrast younger, healthier and more often in work. Thus, our findings are first and foremost generalizable to similar study samples.
Even though our findings indicated higher risk of future joint replacement related to high LPA, there are undeniably beneficial effects of being physically active. Weight-bearing activities have beneficial effects for bone health, general health and mortality, and moderate to high intensity of bone-loading is recommended, including a combination of endurance training and resistance exercises [
45,
46]. Exercise has also been shown to reduce pain and improve physical function in those already diagnosed with hip or knee OA [
47,
48] and potentially reduce or delay the need for joint replacement [
49]. However, we acknowledge the importance of identifying subgroups that are at a higher risk of injury and early development of OA in order to prevent the increasing demand of joint replacements and future burden of OA.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
All authors, MBJ, AIH, VB, OF, AL, GF, LN, JAZ and KS, have contributed to the conception and design of the study, interpretation of data, revision of the manuscript, and approval of the final draft. VB contributed with statistical expertise. VB and MBJ performed the statistical analyses.