Statistics from Altmetric.com
- OA, osteoarthritis
- OR, odds ratios
- RR, relative risk
- HR, hazard ratio
- 95%CI, 95% confidence intervals
- ACL, anterior cruciate ligament
- Exp, exercise group
- Con, control group
Osteoarthritis (OA) often limits activities of daily living—for example, climbing stairs, dressing1—and can prevent participation in the labour force for younger patients.1 Patients seek advice from family doctors, internists, rheumatologists, and orthopaedic surgeons.
Recently sponsored symposia in both the United States2 and Canada (co-sponsored by the Canadian Institutes of Health Research and the Canadian Arthritis Network, Toronto, Ont, April 2002) suggest that OA is a complex syndrome—that is, constellation of symptoms and signs with multiple causes—that involves the balance between cartilage synthesis and degradation, and affects all tissues surrounding the joint. That being said, the question remains as to which factors are directly related to the cause of OA and are modifiable so that doctors may counsel patients appropriately.
In the case of primary OA—that is, excluding genetic diseases, severe biomechanical abnormalities, post-septic arthritis, etc—many healthcare professionals believe the major cause of OA is “wear and tear”—that is, gradual thinning of the articular cartilage due to repeated weight bearing activity of the joints—and that therefore OA is caused and worsened by exercise. However, in 1999, Hurley3 reviewed the basic science evidence and proposed that properly contracting muscles are the main force absorber for the joint, and that muscle dysfunction is the most important modifiable mediating factor for primary OA. Because regular exercise improves muscle function, this hypothesis predicts that exercise would not increase the incidence of or worsen OA. Hurley also suggested that whereas the wear and tear hypothesis predicts that cartilage thinning will be the first sign of OA, the muscle dysfunction hypothesis predicts that sclerosis would be the first sign. Finally, in the case of injury, the muscle dysfunction hypothesis predicts that injuries to muscles in a leg may increase the risk of OA in joints not immediately adjacent to the injured muscle because impact forces are not being properly absorbed. The wear and tear hypothesis suggests that injuries would only increase the risk of OA if articular cartilage injury occurs at the time of injury, or is more likely to occur after injury—for instance, anterior cruciate ligament (ACL) instability. The specific objective of this systematic review is to determine the clinical evidence in support of and against the hypotheses that exercise related OA is caused by (a) wear and tear or (b) muscle dysfunction.
The reader should not forget that OA is multifactorial and that there are other causes of OA. As such, there are two important limitations to the scope of this article. Firstly, it focuses on both hip and tibiofemoral OA and does not discuss patellofemoral OA, or OA in other areas of the body. Secondly, regardless of the initiating event of OA in a particular patient, the articular cartilage is eventually destroyed. The mechanism of articular cartilage destruction is also beyond the scope of this article.
MATERIALS AND METHODS
A systematic review of the literature was carried out. Medline and SportDiscus databases were searched using the strategy (osteoarthritis or osteoarthrosis) AND (activity or exercise or injury). Based on titles and abstracts, all potentially pertinent articles were retrieved and reviewed. The bibliographies of all articles retrieved were reviewed for additional references, and a search of Citation Search Index was conducted to find any article that may have cited one of the key articles previously retrieved. Data were abstracted by one person using a standardised form, and verified with a second reading by the same person at least four weeks later. This review is limited to exercise related primary OA, and studies investigating OA secondary to injury or previous surgery were not included in the results.
Results are presented as odds ratios (OR) or relative risks (RR) or hazard ratios (HR) with 95% confidence intervals (95%CI) in parentheses unless otherwise specified. Because many studies lacked the necessary power to determine if the differences were statistically significant, relying on p values or confidence intervals might result in a β error (incorrectly indicating that the differences between groups were not important). Therefore, the emphasis in this review is on the direction and magnitude of the changes—that is, are the changes clinically relevant?— rather than whether a study had significant results.
Because the clinical studies reported different outcomes, used widely differing methodology, etc, a qualitative synthesis was more appropriate than an attempt to provide an overall summary statistic for the estimate of the effect.
Twenty three clinical articles (representing 18 studies) related to exercise and OA were retrieved. Table 1 presents studies on running, table 2 presents studies on football, and table 3 presents studies on other sports. Where studies reported on more than one type of exposure, the relevant details are repeated under each section and the duplication noted.
Overall, the three cross sectional running studies concluded that exercise is not associated with OA,4–6 and the three case-control running studies found mixed results but overall suggested that some higher intensity activities may be associated with the development of OA.7–9
With respect to historical cohort studies on running, there was no increased risk of OA in runners in four of seven historical cohort studies. This was true for (a) 27 elite Danish male orienteering runners compared with hospital controls,10 (b) 60 Finnish male elite runners compared with hospital controls,11 (c) 504 US college varsity cross country runners compared with varsity swimmers,12 and (d) 1282 Finnish ex-elite male endurance athletes after controlling for previous injury (three papers published on the same cohort13–15).
In one study showing a possible increased risk of OA in runners,16 running pace was a better predictor than running mileage even though the wear and tear hypothesis would predict that OA should increase with each vertical impact—that is, step—more so than horizontal speed. Horizontal speed would be important if the running technique was suboptimal, and the runner placed the foot in front of the body at heel strike, thereby creating a large breaking force. However, this breaking force slows the runner down and therefore would not be expected to correlate with running speed.
Another historical cohort study suggested an increased risk in runners younger than 50 who run >20 miles a week.17 An effect of mileage was not seen in subjects older than 50, which again is contrary to what would be predicted by the wear and tear hypothesis. In the remaining study showing a possible increase in OA,18 osteophytes were associated with elite exercise, but the OR for joint space narrowing was close to 1 for the knee (1.2, 95%CI 0.7 to 1.9) and for the hip (1.6, 95%CI 0.7 to 3.5). Within the control population, moderate exercise was not associated with joint space narrowing of the hip or knee, although there was a trend toward decreased joint space of the hip in the higher participation category (1.8, 95%CI 0.73 to 3.48).
In the only prospective study, Lane and colleagues19–21 found no difference in the development or progression of OA between 41 runners and matched controls after two, five, or nine years. In another study that simply categorised exposure as “sport participation”, there was again a lack of progression of OA.22
Besides pure running, team sports such as soccer have also been implicated as a cause of OA. Although Klunder et al23 found a higher proportion of radiographic hip OA in soccer players, 13/30 patients with OA had previous injuries compared with only 3/19 controls. Lindberg et al24 found hip OA was higher only in the elite soccer players (14.1% for elite, 4.2% for non-elite, and 4.2% for control).
In summary, these findings suggest that moderate intensity impact sports do not cause or worsen OA. OA in high intensity or elite sports could be due to a threshold effect—that is, wear and tear only occurs after a threshold—or some other factor, and a closer examination is warranted.
Kulula’s group13–15 found that the risk of hip or knee disability was only increased in elite team sports (previous injury not controlled for13). When the same cohort of athletes was compared with 1403 controls without controlling for previous injury,14 OA was increased in all types of athletes (OR range 1.73–2.17), but the greatest increase occurred in wrestling (OR 2.73, 95%CI 1.63 to 4.64), weight lifting (OR 2.74, 95%CI 1.27 to 5.9), soccer (OR 2.1, 95%CI 1.2 to 3.8), and ice hockey (OR 4.2, 95%CI 2.2 to8.0). Three of four of these exposures do not involve impact, suggesting that wear and tear is not a likely cause. In a subsequent study of a subgroup of the same population but now controlling for previous injury,15 the risk was now considerably less (OR 1.2, 95%CI 1.0 to 2.3) and much less than the risk associated with previous injury (OR 6.0, 95%CI 1.3 to 27.8). The presence of previous injury may also partially explain the higher rate of OA in the previously mentioned Lindberg study.24 Using the same population, the subsequent publication25 noted that 33% of elite soccer players with previous meniscectomy or ACL tear developed knee OA compared with 11% in those without these injuries. The same may also be true for hip OA, but this type of analysis has yet to be published.
The results of this literature review strongly suggest that regular mild-moderate impact exercise does not increase the risk of OA, and that there is some evidence that it does not increase symptoms in patients with mild-moderate OA. This evidence supports the muscle dysfunction hypothesis as a cause of OA over the wear and tear hypothesis.
The wear and tear hypothesis predicts that any type of impact such as running would increase OA, or worsen it once developed. However, the clinical evidence suggests that recreational running and soccer do not increase the risk of OA. In the basic science literature, canine cartilage adapts favourably to moderate running,26 and running did not worsen immobilisation induced OA in rabbit knees,27 which is consistent with the prospective study reported by Lane et al.21 In addition, the finding that degeneration occurs with forced exhaustive running in dogs28,29 is also consistent with the muscle dysfunction hypothesis because exhaustion will prevent the muscles from absorbing force. Although some might believe that marathon running could be analogous to forced exhaustive exercise in dogs, most marathon training is done at much lower mileage. Although subjects may be tired, they are not exhausted. The actual marathon is run only a few times a year, whereas the dogs were run to exhaustion regularly.
Most of the subjects in the clinical studies in this review had intact menisci, and presumably no major malalignment. In subjects with previous meniscectomy, Roos et al30 reported no effect of exercise on the incidence of OA. This contradicts the basic science finding that running increased the risk of OA in meniscectomised sheep.31 Although there were no studies on the effect of exercise in subjects with malalignment, Sharma et al32 reported that disease progression occurs more rapidly in this population. How does the muscle dysfunction hypothesis relate to these populations? The wear and tear hypothesis predicts that cartilage damage precedes bone sclerosis. However, the reverse occurred in adult rabbit knees subjected to one hour impulse loading a day.33 The sclerosis was associated with numerous healing trabecular fractures, suggesting that the principle force absorber in anaesthetised animals is not cartilage but bone. This is supported by in vitro findings suggesting that articular cartilage does not absorb force,34 but does redistribute force.35–37 If enough microtrabecular damage occurs over a short period of time, sclerosis would occur as an adaptation—that is, damage would be less likely in sclerotic bone.38 Within this paradigm, malalignment and meniscectomy could increase the risk of OA30,32 because they prevent the normal redistribution of force—that is, even in normal knees, the muscles do not absorb 100% of the force—which makes micro-damage more likely to occur. Finally, the sclerotic changes in underlying bone stiffness may increase the stress on articular cartilage,39 which would lead to increased degenerative changes in both meniscal and articular cartilage.
Although the findings suggest that recreational sports are innocuous with respect to developing OA, they do suggest that participation in elite sports increases the risk of OA. This occurred in impact sports, such as soccer, and also in non-impact sports, such as weightlifting and hockey. Unlike the wear and tear hypothesis, the muscle dysfunction hypothesis predicts these results through the increased risk of injury that occurs with elite sports and the subsequent muscle dysfunction that occurs with injury. In support of these findings, others have found that young adults with previous knee injury are more likely to develop OA,22,40 and that previous hip injury increases the risk of hip OA.40
There are three possible mechanisms by which previous injury could increase the risk of OA. Firstly, the damage may occur at the time of the injury and OA develops over the subsequent years. Secondly, the associated ligamentous instability with major injury leads to recurrent articular cartilage damage. Finally, the associated muscle dysfunction with injury leads to recurrent articular cartilage damage because the impact forces are no longer being absorbed appropriately.
If damage occurred at the time of injury, and the wear and tear hypothesis is correct, then articular cartilage damage should overlie areas of bone damage. However, there was no correlation between the location of a femoral bone bruise and articular or meniscal cartilage damage observed at surgery for ACL reconstruction.41 The possibility of “sub-clinical articular damage” remains theoretical at the present time. Finally, Felson et al42 recently found a strong correlation between location of bone marrow oedema on magnetic resonance images and progression of OA. If bone marrow oedema is indeed a strong predictor of progression, it suggests that bone injury is an early sign of damage. Future research should examine the subgroup of patients who had osteophytes without joint space narrowing at baseline to see if bone marrow oedema preceded the articular cartilage damage.
If ligamentous instability of the joint after an ACL tear causes OA, then ACL reconstruction should minimise the risk of OA. However, clinical studies (albeit with their limitations) suggest that it may not.43,44 Other authors have suggested that it is the underlying bone injury that occurs at the time of ACL rupture that is the cause of OA. Yet, OA is produced in dogs and cats by isolated transection of the ACL without associated bone damage at the time of injury.36,45,46 In the muscle dysfunction hypothesis, the loss of proprioception information from the ACL would result in increased force transmission to the bone, and increased OA. Further, evidence from biomechanical studies reveal an increase in loading of the non-transected knee, which does not develop OA,47 which again suggests that normal muscles can absorb the regular amounts of stress and strain across a joint and that “wear and tear” is not the cause of OA in uninjured limbs.
The muscle dysfunction hypothesis is based on the finding that muscle fatigue increases the impact forces crossing a joint,48,49 which suggests that properly contracting muscles are the main absorber of force. Whether the muscle cannot contract properly because of age or fatigue or disuse atrophy, or because of injury induced weakness (strains) or loss of proprioception (ACL tears), the effect is the same; more force is transmitted to the bone, which leads to increased microtrabecular damage, which leads to sclerosis, which could lead to changes in the stresses and strains across the articular cartilage, and then joint space narrowing. The added stress would then lead to the characteristic changes observed in periarticular tissue. Note that this hypothesis would predict an increased risk of OA with less severe injuries than are usually accounted for in studies—for example, quadriceps contusion could lead to increased risk of OA even though there was no ligamentous damage—and also the greater risk of hip OA compared with knee OA in soccer players23—that is, groin strains occur often in soccer but rarely with running. In addition, it would predict a higher rate of hip OA in subjects with knee injuries, and vice versa because the muscles of the thigh would be expected to absorb force across both joints. However, this analysis has not yet been published.
Other activity and obesity
The objective of this article was to assess the risk of OA with exercise. Although a detailed discussion of the risk of OA with exposure to various occupations is beyond its scope, the muscle dysfunction hypothesis can explain findings in this area as well. Briefly, if a person is forced to work when fatigued or injured—for example, a farmer—the muscles no longer absorb the forces crossing a joint and there would be an expected increase in microtrabecular damage, then sclerosis, and then OA. For example, in the study by Lau et al,8 for those subjects with occupational exposures that required climbing 15 flights of stairs or more, the OR for developing OA was 5.1 (95%CI 2.5 to 10.2) for women and 2.5 (95%CI 1.0 to 6.4) for men in the entire study, but 34.0 (95%CI 4.7 to 248.4) overall for those with previous injury. Similarly, the OR for developing OA in those subjects with occupational exposures that required lifting ⩾10 kg more than 10 times a week was 2.0 (95%CI 1.2 to 3.1) for the entire group and 25.9 (95%CI 8.1 to 82.4) for those with previous injury.
Finally, obesity is a well recognised risk factor for OA.2,50 The muscle dysfunction hypothesis explains this relation as well. The added weight means that muscles must absorb even more force and therefore must be stronger and have greater endurance or there will be a “relative dysfunction”. However, obesity is associated with physical inactivity and therefore relative muscle dysfunction. With respect to mortality, most of the evidence suggests that obesity is not related to mortality if there is adjustment for physical fitness.51 Future studies should explore whether the relation between obesity and OA is similar to that between obesity and mortality.
The muscle dysfunction hypothesis that was originally proposed based on basic science evidence is supported by the clinical literature as well. This includes:
Major injuries are associated with a high rate of OA.
Because muscles provide the “dynamic” joint stability during movement, some signs of OA—that is, osteophytes and capsular thickening—may be an attempt by the body to increase joint stability in the presence of muscle dysfunction induced dynamic instability.
A wide variety of elite sports, but not recreational exercise, are associated with OA. This effect is greatly reduced when major injuries are controlled for. Because elite athletes often play while injured—that is, on weak muscles—the muscle dysfunction hypothesis predicts that there would still be an increase in risk if minor injuries are not controlled for—for example, groin strain in soccer and hip OA.
The most important implication of the muscle dysfunction hypothesis is that proper rehabilitation after an injury may be important in the prevention of OA. A study designed to definitively test the role of muscle dysfunction would require detailed prospectively collected data, controlling for proper rehabilitation after major and minor injuries using appropriate strength testing and close supervision. That being said, the hypothesis that best explains the evidential relation between exercise and OA currently available today is the muscle dysfunction hypothesis.
Funding by the Chercheur-Boursier Clinicien program of the Fonds de Recherche en Santé du Québec is gratefully acknowledged.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.