It is widely accepted that participation in physical activity is associated with physical, psychological and social benefits 1. Physical activity not only reduces the risk for cardiovascular disease 2 (a major cause of mortality in developed countries) but is also recommended for the management of obesity and the treatment of mental illness 3. Globally, a vast number of large-scale public health campaigns is aimed at promoting physical activity. Exercise is also widely recommended by health care professionals in the prevention and management of chronic health conditions such as osteoarthritis (OA). However, our knowledge of the effect of physical activity on risk for developing OA is limited.

OA is the most common joint disorder affecting the elderly. In particular, radiographic knee OA affects at least 30% of people aged over 60 years 4 and is a major cause of functional disability 5. With our population ageing, the prevalence of OA in the developed world is expected to increase and it is anticipated that OA will become the fourth leading cause of disability in the coming decades 6. A population-wide initiative, such as promotion of physical activity, has the potential to contribute inadvertently to the growing burden of this disease.

Epidemiological studies that have investigated the effect of physical activity on the knee joint have reported conflicting findings. Although some studies have reported that physical activity is associated with risk for knee OA 789, other studies have shown that physical activity may have no effect 1011 or may even protect the knee joint from degenerative changes 1213. These conflicting findings may be a result of individual variation in response to exercise and/or different methodology employed by studies to measure knee structure and physical activity.

We propose that there may be subgroups of individuals who differ in their response to physical activity within our community. Although some individuals may have specific characteristics that enable them to exercise without increasing their risk for knee OA, others may require these individual factors and/or their exercise programme to be modified before they can commence or continue to exercise safely. In this review, we examine the roles played by a number of parameters that may mediate the relationship between physical activity and knee OA. We focus on the role of physical activity in the risk for developing OA, and do not consider its effect on those individuals with established OA. We also suggest that the methodology used to assess knee joint structure and physical activity might have contributed to the conflicting results between studies of physical activity and knee joint health.

We conducted electronic searches of the Medline and EMBASE databases between 1980 and September 2007 to identify relevant studies for this review. The search involved the use of MeSH (medical subject headings) and 'free text' words, including physical activity, exercise and knee osteoarthritis, and was limited to studies relating to humans and published in English. The reference lists of relevant articles were also screened to identify additional studies. We identified epidemiological studies that investigated the effects of age, sex, body mass index (BMI), knee injury and/or knee alignment on the relationship between physical activity and risk for developing radiological, symptomatic and physician diagnosed OA.

Our Medline and EMBASE searches identified 193 and 601 potentially relevant papers, respectively. We identified 12 studies that met our inclusion criteria; nine of these examined the effect of age and/or sex (Table 1), four investigated BMI (Table 2) and four considered knee injury or alignment (Table 3). A variety of methods were used to examine knee OA. Seven studies 11121415161718 used radiographic methods; four 8101319 based their assessment on self-reported, physician diagnosed OA; three 111920 examined self-reported symptoms; and a further two 1516 involved a clinical assessment. All studies used a self-report measure to examine physical activity (Tables 1, 2 and 3).

The use of different methods to measure the development of OA may also have contributed to inconsistency in results between studies of physical activity and knee joint health. Although some studies used radiographic measures of joint structure, including joint space width and/or the presence of osteophytes 71114, others have based their assessment on individuals' self-reported symptoms 20 or self-reported presence of physician-diagnosed OA 813 (Table 1). In addition, two studies 1619 performed clinical examinations of the knee joint.

There are methodological issues associated with each of these techniques, with self-reported data potentially influenced by poor recall, and the accuracy of physician diagnosed OA questionable. Moreover, indirect measurement of the joint space width as a surrogate for articular cartilage has proven to have inherent problems with respect to validity, reliability and sensitivity to change 434445. Thus, studies that have relied on radiographic measures of joint space may have missed an effect of physical activity on structural abnormalities, including cartilage defects and bony enlargement, which cannot be detected radiographically.

An over-reliance on the presence of osteophytes to diagnose knee OA may also have influenced the results of radiographic studies examining knee joint structure. The presence of osteophytes is required for diagnosing knee OA using the Kellgren and Lawrence grading system 46, and measurement of osteophytes is associated with greater reproducibility than that of joint space narrowing 47. Even if osteophytes are more prevalent than joint space narrowing in physically active individuals, it is possible that this finding represents the effect of musculoskeletal traction forces produced by exercise, not direct cartilage damage to the knee joint 48.

Recently, a small number of studies have used magnetic resonance imaging (MRI) to assess the relationship between physical activity and knee joint structure. Although radiographs use ionizing radiation to provide a two-dimensional approximation of articular cartilage, MRI visualizes joint structures from a three-dimensional perspective without exposing the patient to radiation. MRI is recognized to be a valid, accurate and reproducible tool for measuring articular cartilage volume and cartilage defects 49.

A cross-sectional study of 176 community-based women 50 revealed that participation in vigorous activity (activity leading to sweating, shortness of breath, or an increased pulse rate) was associated with greater medial tibial cartilage volume. A longitudinal study of healthy, recreational long-distance runners 51 demonstrated no significant knee structural change at 6 to 8 weeks after a marathon. Moreover, there is accumulating evidence demonstrating that physical inactivity adversely affects cartilage development in children 52 and predisposes to rapid cartilage loss in adults 53. Thus, MRI studies to date have consistently shown a beneficial effect of physical activity on knee joint cartilage 5051525354.

In summary, although several techniques are available to measure knee joint structure in studies of physical activity, use of these techniques might have resulted in conflicting results in terms of the effect of physical activity on risk for OA. Radiographic assessment of both joint space narrowing and osteophytes is associated with issues of measurement and interpretation. We propose that a technique that is reliable and valid, examines different features of OA, and is sensitive to change in both healthy and OA populations must be consistently used in longitudinal investigations if the influence of physical activity, and of other potentially mediating factors, on knee joint health is to be clearly defined.

It has been suggested that use of self-report surveys to measure physical activity has contributed to inconsistency in findings between studies examining the relationship between physical activity and knee joint health. Although surveys are easy to implement in large cohorts and provide important physical activity data, they are associated with over-reporting of activity 55 and reduced accuracy for moderate intensity physical activity 56, poor recall over longer periods 55 and only moderate reproducibility 57. An alternative, accelerometry, provides an objective and reliable measure of the frequency, duration and intensity of physical activity 58. Although expensive and impractical for use in large cohorts, we propose that future longitudinal studies consider use of accelerometry in random samples of participants. This will allow investigators to confirm the accuracy of self-reported survey data, to ensure that those in different physical activity groups actually differ in their types and/or levels of activity, and to confirm that participants' physical activity patterns remain consistent over the course of the study.

Assessment of physical activity is further complicated by the different features of physical activity, including the type, intensity, frequency and duration, that can be examined. For instance, although Cheng and coworkers 8 assessed participants based on the distance they walked or jogged per week (for example, high activity was jogging or walking ≥20 miles/week), Rogers and colleagues 13 used a variety of self-reported activities to categorize participants into low/moderate or high joint stress groups. A small number of studies have performed subanalyses to investigate the effect of different aspects of physical activity on risk for OA. For instance, Hootman and coworkers 10 examined the association between hip/knee OA and training frequency, pace and total weekly mileage in a subgroup of participants, but they identified no association. It is clear that direct head-to-head comparisons of different dosages of physical activity, in particular those recommended by national health bodies, using reliable and valid measures are required.

In this review we examine possible reasons for the conflicting results arising from studies of physical activity and knee joint health, and we propose possible approaches that may be used in future investigations (Table 4). Novel methods that can be used to examine knee structure directly from health through to disease may overcome some of the problems associated with the use of radiography to examine knee joint structure. A more comprehensive examination of various knee structures and the implementation of objective and accurate assessments of physical activity may also enhance our understanding of the mechanism by which physical activity affects the knee joint.

Addressing these methodological issues will allow studies to explore the role of biophysical factors within an individual that may influence the effect of physical activity on risk for OA. There is evidence that factors such as age, sex, body mass and previous knee injury may influence knee joint health in physically active individuals. Rather than a uniform approach to the implementation of physical activity, we may find that certain types of exercise have different effects on different people and that individually tailored exercise programmes are needed to allow exercise to commence safely. In a society that is ageing and being encouraged to be physically active, such programmes may have the potential to reduce the growing burden of OA.

BMI = body mass index; MRI = magnetic resonance imaging; OA = osteoarthritis.

The authors declare that they have no competing interests.