This population-based prospective study describes the contribution of multiple osteoarthritic correlates of QoL over five years of observation. Physician diagnosed OA of the back and pain at all sites were independent and stable correlates of QoL, and pain at multiple sites has an additive deleterious effect on QoL. With the exception of the back, pain appeared to mediate the association between diagnosed OA and QoL. Radiographic osteoarthritis was not associated with QoL.
In this study, the strongest musculoskeletal correlate of QoL was pain. Pain is a priority for patients seeking care  and thus it is perhaps not surprising that pain largely mediated the association between doctor diagnosed OA and QoL. Further, pain assessed at one site in cross sectional studies is known to be associated with poorer QoL, [19, 20] but no studies that have looked at pain at many sites. Our data suggests that pain at all sites measured independently contribute to QoL, there is a dose response association between number of pain sites and QoL, and severity of pain is also related to QoL. Our data suggests that pain is very common in older adults in the community. Given that pain at individual joints and overall number of sites of joint pain were associated with poor QoL, this suggests that interventions to reduce the frequency and intensity of pain may be effective in improving QoL at the population level.
While there are some inconsistencies in the three analyses, the most weight should be put on the analysis over five years as it uses all the data and therefore is the most powerful. These results confirm and extend the findings of Woo et al., 2009,  where pain at the back, hip (men only) and knees was associated with QoL over time in ethnic Chinese. Pain in the shoulders and back were the most important factors in our analyses, but knees, hips and even hands and feet were significant. The inconsistency with the hip may, in part, be due to patients have difficulty locating the correct anatomical position of the hips,  or that pain in the knee can actually be referred from the hip . Knee pain was of borderline significance in cross-sectional analyses but became significant over time.
Diagnosed OA of the back was also an independent correlate of poor QoL (both in cross-sectional and longitudinal analyses), but diagnosed OA of the neck, shoulders, hips, hands, knees and feet were not once adjustment was made for the multiple sites of OA and for pain. This suggests that while pain mediates the associations between diagnosed OA and QoL at sites other than the back (neck, shoulders, hands, hips, knees and feet), the association between diagnosed OA of the back and QoL is only partially mediated by pain. It is well known that psychological factors such as depression are associated with chronic back pain but unfortunately we were not able to assess these in the current study.
There was no association between radiographic osteoarthritis and QoL at baseline, after adjusting for age, sex and BMI. This suggests that radiographic findings make no independent contribution to QoL, consistent with other studies which showed that the association between radiographic OA of the hand and function was largely mediated by pain,  and that pain is a better predictor of disability than radiographic change [24–26]. This differs from the findings of other studies [9, 27], who found that radiographic OA was associated cross-sectionally with different disease-specific measure of QoL, after adjustment for pain and other covariates. Both of these measures of QoL had pain as a subscale, so this may explain why they found an association yet we did not. A strength of our study is that, unlike Norimatsu and colleagues, we have collected (self-reported) diagnosis of OA and radiographic findings separately (in addition to pain), and while finding them to be correlated, when both diagnosis and radiography appear together in one model, radiographic findings are no longer associated with QoL. Our data demonstrates that diagnosis of OA reflects more than radiographic evidence of joint damage, but that with the exception of diagnosed OA of the back, is not independent of pain.
Strengths of this study include the random population-based sampling and comprehensive data collection, and five-year period of observation, providing excellent external validity for our findings. Limitations include absence of information on psychological factors, such as diagnosed mental health conditions or psychological distress: this limits our ability to consider such conditions as covariates or effect modifiers, but our model is robust whether or not the mental health component of QoL is included, suggesting this is not a major issue. Additionally, the initial response rate of 57%, while lower than desirable, is similar to other comparable Australian studies,  and a lower response rate does not mean that relationships between outcome and exposure are necessarily biased . Participants who did not continue in the study were older, heavier, with pain at more sites at baseline than the remaining participants. This should reduce the observed effect size of our findings, but since few associations were of borderline significance this should not have altered our conclusions. We did not seek to confirm doctor-diagnosed cases of arthritis, and therefore participants may have under-or over-reported diagnosed arthritis, and the extent to which this may have affected the findings of the study is unclear. However, use of self-reported doctor diagnosed OA appears to be a reasonable proxy for OA, as JSN was more common in participants reporting doctor-diagnosed OA at the hips and knees (hips OR 2.3, p < 0.001; knees OR 1.6, p = 0.023), and osteophytes more common in participants reporting knee (OR 4.10, p < 0.001), but not hip OA (OR 0.94, p = 0.83). We had X-rays only of the hips and knees, and so are not able to assess the association between ROA and QoL at other anatomic sites. However, unless the causal pathways at other sites are substantially different to those at the knees and hips, it is unlikely that radiographic OA at these sites would add any new information to the models.