F. Birrell 1,3, M. Lunt 1, G. Macfarlane 2 and A. Silman 1

Rheumatology 2005;44:337 341 Advance Access publication 9 November 2004 Association between pain in the hip region and radiographic changes of osteoarthritis: results from a population-based study F. Birrell 1,3, M. Lunt 1, G. Macfarlane 2 and A. Silman 1 doi:10.1093/rheumatology/keh458 Objectives. The relationship between hip pain and radiographic change in the population is unclear due to lack of agreed definition for hip pain and difficulties in obtaining radiographs from asymptomatic random samples. Our objective was to assess the relationship between hip pain and radiographic change in osteoarthritis (OA) in a population sample aged over 45. Methods. One thousand and seventy-one responders to a postal questionnaire using a recently validated approach to defining hip pain were stratified into hip pain-positive and -negative groups and samples of each were X-rayed and scored for OA using both minimum joint space and the Croft score. The association between pain and X-ray score was estimated, weighting back to the age and gender distribution of the original population. Results. Hip pain prevalence was 7% in males and 10% in females. Severe OA was present in 16% of those with and 3% of those without pain. Adjusting for age and gender, there was a very strong association of pain with severe OA [odds ratio (OR) 17.4, 95% confidence interval (CI) 3.0 102], but no association with mild/moderate OA (OR 1.4, 95% CI 0.4 4.7). By contrast, only 22% of men aged 45 54 with severe OA had current pain, though in older age groups the proportions with pain were higher (54 70%). Conclusions. Hip pain is relatively infrequent in the general population compared with the published reports of other regional pain syndromes. Mild/moderate radiographic change is very frequent and not related to pain, whereas severe change is rare but strongly related. In younger males, severe radiographic change is much less likely to be associated with pain. Regional pain syndromes such as anterior knee pain are very commonly reported in general population surveys [1, 2]. Although in the majority of those subjects reporting pain there is considered to be no underlying pathology, in clinical practice the reporting of such pain is frequently a signal for radiographic examination to exclude the presence of underlying degenerative disease as the source of the pain. In population-based epidemiological studies, any associations between pain reporting (for example, of the knee) and the presence of osteoarthritic changes have been weak [3]. Pain in the hip region appears to be less common than other regional pain syndromes [4]. Hip pain is, however, important for the affected individual as it impacts on key functions, such as walking. The importance of hip pain for health-care providers results from its association with osteoarthritis (OA) and the potential need for joint replacement. Hip OA is the most frequent indication for hip arthroplasty [5] and unremitting pain the most important clinical determinant of need for surgery [6]. The hip, however, is a difficult region to define topographically compared with, for example, the shoulder or knee. Referred pain from the back and disorders of structures in and around the inguinal and pelvic areas all might present with pain in similar areas to that resulting from true hip disease. Similarly, compared with more superficial joints, subjects have difficulty in localizing the hip as the source of their pain. Indeed, there are no well-accepted definitions of hip pain for epidemiological studies. We have recently shown that the optimal definition of hip pain is one that requires the subject both to use the word hip in referring to the site of their pain in addition to indicating the presence of pain in the bathing trunk area on a validated preshaded manikin [7]. Such a stringent definition demonstrated much stronger associations with indicators of hip disease, such as restriction in movement and health-care utilization, than less rigid definitions requiring either just the verbal or just the manikin definition [8]. However, the strength of the relationship between radiographic change and hip pain in the general population is not known. Such studies require the radiographic investigation of asymptomatic subjects and, for this reason, there are few data. We have therefore investigated the relationship between radiographic damage and hip pain in a large population sample. We have specifically addressed two related questions: (i) given the presence of radiographic change, how frequent is hip pain? and (ii), given the presence of hip pain, how frequent is radiographic change? Thus, we wanted to determine from the first question what proportion of severe OA, for example, would be missed by consideration of those only with current hip pain; and from the second question the value of undertaking a radiograph in those with pain. Methods Design This was a two-stage population screening survey: subjects completed a questionnaire asking about hip pain, and samples of those with and without pain were invited for radiographic 1 arc Epidemiology Unit and 2 Unit of Chronic Disease Epidemiology, School of Epidemiology and Health Sciences, Manchester University Medical School, Manchester M13 9PT, UK. 3 Present Address: University of Newcastle, Cookson Building, Framlington Place, Newcastle NE2 4NH, UK. Submitted 5 July 2004; revised version accepted 30 September 2004. Correspondence to: A. Silman: a.silman@manchester.ac.uk 337 Rheumatology Vol. 44 No. 3 ß British Society for Rheumatology 2004; all rights reserved

338 F. Birrell et al. assessment. Associations between pain and OA were then calculated based on weighting the X-ray results of those participating in the second phase to the population that were initially sent a questionnaire. Subjects The sampling frame was the population listing from a general practice population in the south of Manchester and has been described in detail elsewhere [8]. This investigation was restricted to subjects aged over 45, and a total of 1496 subjects were selected for mailing a questionnaire. Ascertainment of hip pain The presence of current hip pain was determined using a postal questionnaire. Two questions were used to determine the occurrence of hip pain. The subjects were asked (i) if they had had pain in the hip for a period lasting more than 24 h in the previous 1 month, and (ii) to indicate on a manikin with a preshaded bathingtrunk [7] area if they had had pain in that area during that period. A positive respondent was one who reported hip pain using both criteria. Subjects not responding to the questionnaire were sent a postcard reminder within 2 weeks and a further questionnaire, if necessary, after a further 2 weeks. The questionnaire also collected data on a number of other aspects of hip pain and general health. Radiographic follow-up Random samples of those reporting and not reporting hip pain were invited to attend a local hospital for a pelvic radiograph. An anteroposterior pelvic film was taken in a standardized fashion at a single centre. The X-rays were read blind to pain group by two readers and graded using both the Croft modification of the Kellgren and Lawrence (K&L) grading system [9]. The minimum joint space (MJS) was also measured using a ruler, as previously described by Croft et al. [9]. For the purposes of analysis, the scores were stratified into three groups: (i) subjects with no OA (defined as a K&L grade of <2 and MJS >2.5 mm); (ii) subjects with mild to moderate OA (defined as a K&L score of at least 2 or MJS <2.5 mm but not satisfying criteria for severe OA); and (iii) subjects with severe OA (defined as a K&L score of at least 4 or MJS <1.5 mm). Analysis The subjects invited for radiographic investigation were not a random sample of the original population but were weighted towards studying a greater proportion of hip pain-positive compared with hip pain-negative subjects. Further, both age and gender were strongly related to the presence of any radiographic change. Thus, for the purposes of analysis, the subjects were stratified by age (in 10-yr age groups) and gender. We therefore assumed that in each of the two pain groups the distribution of radiographic changes in those studied was the same as that in those subjects with the same age and gender profile, who answered the questionnaire but were not selected or attended for X-ray. We used the sampling weights for these two variables to estimate the population occurrence of radiographic change. The data are thus presented first as the estimated population prevalence of radiographic change in subjects with and without pain, and secondly as the estimated population prevalence of hip pain in subjects with and without radiographic change. The confidence intervals (CIs) for the population prevalences were calculated using the svytab procedure in Stata 8.0 (Stata Corporation, College Station, TX, USA), which uses a standard formula for calculating the standard error of a prevalence in a two-stage sampling design such as this [10]. To examine the influence of pain on X-ray grade, logistic regression analyses were undertaken, adjusting for age and stratified by gender, where appropriate. The study was approved by the Local Research Ethics Committee and written subject consent was obtained. Results A total of 1496 subjects were mailed a questionnaire and, after the two reminders, completed questionnaires were received from 1071. A further 123 were notified as being sent to the wrong address. Thus, the final adjusted response rate of those receiving the questionnaire was 78%. The responders comprised 556 males and 515 females. The age- and gender-specific hip pain prevalences are shown in Table 1. Overall, the prevalence was 6.8% in the males and 10.3% in females. As shown, the prevalence was higher in females at all ages except for those over the age of 75. A total of 82 subjects with hip pain and 308 without hip pain were invited to attend for radiographic screening (women aged <50 were not invited), of whom 56 (68%) and 147 (48%), respectively, attended. An analysis within each pain group, comparing those who attended for X-ray with those who did not, showed no evidence of any participation bias in relation to the severity indicators derived from the questionnaire. Thus in the hip pain group, compared with those who did not, those who attended for X-ray had similar visual analogue pain scores as well as similar proportions of those using analgesics and requiring a walking stick for their pain. The distributions of radiographic score between those with and without pain are shown in Table 2. There was a marked difference between the two groups, as expected ( 2 ¼ 12.9, P ¼ 0.002). The frequency of severe change was substantially larger in the group with pain (16 vs 3%), and the absence of change was more common in those without pain. However, mild/moderate OA was very common in this population, with an equal high frequency of such changes in both those with and without pain. Age and gender were expected to have powerful effects on the likelihood of radiographic change. Indeed, there were substantial differences in the occurrence of radiographic change (independent of pain frequency) between the genders and with age. Allowing for the proportions with and without hip pain, the estimated population prevalence of radiographic change by age and gender, together with their 95% CIs, are shown in Table 3. These data show that severe change was exceptionally rare in women at TABLE 1. Prevalence of hip pain by age and gender Age group (yr) n Hip pain % n Hip pain % 45 54 220 10 4.6 227 22 9.7 55 64 169 12 7.1 140 14 10.0 65 74 117 7 6.0 95 12 12.6 75 84 50 9 18.0 53 5 9.4 TABLE 2. Frequency of radiographic change by presence/absence of pain No OA Radiographic severity Mild/moderate OA Severe OA No pain (n ¼ 147) 60 (41) 83 (56) 4 (3) Pain (n ¼ 56) 16 (29) 31 (55) 9 (16)

Hip pain and radiographic change 339 TABLE 3. Estimated (weighted) population prevalence of radiographic change Radiographic severity Age group (yr) None Mild Severe None Mild Severe 45 54 33 (9, 72) 60 (25, 88) 7 (1, 30) 45 (23, 70) 55 (30, 77) 0 55 64 52 (27, 75) 46 (23, 71) 2 (1, 8) 33 (19, 50) 67 (50, 81) 0 65 74 53 (28, 77) 45 (21, 70) 3 (1, 9) 55 (31, 78) 43 (21, 68) 1 (0, 9) 75 84 35 (13, 67) 52 (35, 69) 5 (2, 12) 28 (7, 67) 72 (33, 93) 0 Data are % (95% CI). TABLE 4. Estimated prevalence of radiographic damage by hip pain status: men No pain all ages. Further, and unexpectedly, there was very little influence of age on the occurrence of OA in men. It should be noted, however, that these prevalences, once stratified by age and gender, are based on very small numbers and as a result the CIs are wide. The estimated prevalences of radiographic change in the population were then stratified by pain status (Table 4). were excluded from this analysis, given the virtual absence of severe OA from this group. The data show that in each age group those men with pain were substantially more likely to have severe OA, with no influence on the likelihood of mild to moderate OA. A weighted multinomial logistic regression analysis was undertaken, adjusting for age, and this showed that the odds ratio (OR) for mild/moderate OA in those with pain was 1.4 (95% CI 0.4 4.7), with a much greater association with severe OA (OR 17.4, 95% CI 3.0 102). The final analysis was to evaluate the estimated population prevalence of hip pain in those with and without evidence of OA (Table 5). As discussed above, despite the higher prevalence of pain in the women, there were virtually no women in this sample with severe OA. In men, hip pain was more likely to be reported in those with severe OA, but not by those with mild OA. Hip pain Severity Age group (yr) None Mild Severe None Mild Severe 45 54 34 (8, 74) 61 (24, 89) 5 (1, 33) 23 (3, 76) 45 (10, 85) 33 (6, 78) 55 64 53 (27, 77) 47 (22, 73) 1 (0, 6) 38 (12, 74) 38 (12, 74) 24 (5, 64) 65 74 56 (29, 80) 42 (19, 70) 1 (0, 7) n/a 71 (29, 94) 29 (6, 71) 75 84 44 (27, 63) 53 (34, 71) 3 (1, 13) 44 (14, 80) 30 (7, 71) 26 (6, 67) Data are % (95% CI). n/a, numbers too small for calculation. TABLE 5. Estimated (weighted) prevalence of hip pain in subjects with and without radiographic damage Prevalence of hip pain: % (95% CI) Age group (yr) None Mild Severe 45 54 3 (0, 28) 3 (1, 17) 22 (2, 84) 55 64 5 (1, 19) 6 (1, 21) 70 (12, 98) 65 74 n/a 10 (3, 28) 67 (11, 97) 75 84 22 (5, 64) 6 (3, 37) 54 (6, 95) 45 54 n/a 15 (5, 38) n/a 55 64 18 (7, 38) 5 (2, 15) n/a 65 74 4 (1, 19) 20 (7, 44) n/a 75 84 n/a 14 (4, 44) n/a n/a, numbers too small for calculation. Pain was much less frequent in young men with severe radiographic change (estimated prevalence 22%) and much higher in the other age groups. The age-adjusted ORs for reporting pain in those with mild to moderate OA and for those men with severe OA are the same as above. The data for women, by contrast, do show an association between mild to moderate OA and pain: OR 2.8, 95% CI 0.94 8.6. Discussion In summary, these data have highlighted that the frequency of hipregion pain as defined in the community is low compared with other regional pain syndromes, such as those affecting the knee [1, 2]. The criteria for hip pain used were stringent, though, as others have shown, the prevalence of hip pain, using a broaderbased question, was much higher. Thus, a study from Oxford [4] in subjects over 65 revealed a prevalence of 19%, approximately twice that observed in the same age group in the present study. Similarly, the frequency of severe radiographic change in the population is low, especially in women, whereas the frequency of mild to moderate OA changes are more common in this age group than a normal X-ray appearance. The major findings relate to the limited association between pain and X-ray. As an example, despite an obvious relationship between severe change and pain in males, we estimate that 1 in 20 men aged 45 54 without current hip pain have severe radiographic OA. Examining the results the other way round showed that important proportions of men, especially in the youngest age group investigated (45 54), with severe radiographic OA do not have current pain. By contrast, there is no important relationship between current pain and milder forms of OA. There are a number of limitations in interpreting these data. First, and most importantly, the numbers in each age and gender group with X-ray data are small. Age and gender did have substantial influences on both the occurrence of radiographic change and probably the relationship between pain and OA. The number of individuals in this community survey who were X-rayed was too small to provide robust estimates in the age gender groups, and such strata-specific proportions are subject

340 F. Birrell et al. to wide CIs. It is, however, not easy to obtain pelvic X-rays from true population samples, which is why alternative strategies have been used, such as reading films from intravenous urograms [9]. Secondly, and perhaps surprisingly, we found only one woman from the 109 with X-rays who had severe OA, and as a consequence we are unable to usefully comment on the relationship between pain and severe OA in women. Clinical experience demonstrates that severe hip OA clearly exists in women and the virtual absence of this state in the current cohort might reflect the population scarcity of this grade of OA, or possibly a selective non-response or non-participation of those women with the worst disease. The latter seems intuitively unlikely as typically it is those with the relevant health problem who are more likely to respond to survey participation. Thirdly, the nature of the study design meant that hip pain subjects were more likely to be sampled for radiological examination. As one aim was to ascertain the relationships between pain and X-ray change in the whole population, we assumed that that the former would be the same between those who were and were not X-rayed. It is difficult to be certain about the correctness of this assumption, though analysis of certain characteristics related to hip pain severity in those with hip pain who were and were not X-rayed does not suggest any important selection bias. The response rates to the questionnaire were high and should not have influenced the external validity of these results. The participation rates for those selected for X-ray were lower and selectively different when those with and without pain were compared. It is obviously difficult to encourage those who were pain-free to attend for radiographic examination. It is possible that those who did may have selected themselves on characteristics such as previous injury [11], family history [12] or obesity [13], all of which might have resulted in a bias towards those with OA changes. The study population was also drawn from a single general practice population, and again the associations between hip pain and X-ray change may not be applicable to other groups, which might have a different underlying prevalence of disorders leading to either hip pain or OA. Thus, farming communities have been shown to have a higher prevalence of OA hip [14]. There are a number of strengths of the present study. First, we used a validated definition for hip pain which has been demonstrated to have greater construct validity with measures of hip disease than definitions relying either on the use of the word hip or on a pain manikin alone [9]. Secondly, the X-rays were taken and read in a standardized and blinded fashion; we have previously reported on our reliability of reading hip X-rays [7]. Thirdly, the population answering the first questionnaire was large, with a high response. It was also possible within the present study to obtain a number, albeit limited, of X-rays on those with pain-free hips, a task of increasing difficulty as both ethical and population concerns about X-rays increase in society. The prevalence of mild to moderate OA changes was, as noted above, very high and indeed such change was more common than a normal radiograph. Further, such changes were not associated with pain. The major conclusion from such observations is that these degrees of radiographic change, although they might represent an underlying anatomical change, are perhaps inappropriately considered as representing the disease state of clinical OA. Therefore, the term mild moderate OA to describe such changes is probably misleading. It was interesting to note that the likelihood of reporting pain, given the presence of severe OA, was substantially less likely in younger men. This observation would add weight to the concept that other factor(s) other than the degree of radiographic damage contribute to pain. One explanation, impossible to test in a crosssectional study, is that age is a surrogate for disease duration and that the younger men have had changes for a shorter period, and it is the longevity of the change that contributes to pain. This, however, would not explain why the rates of pain are not highest in the oldest ages. Ageing itself may be accompanied by an increase in pain perception, although in population surveys the increased reporting of musculoskeletal pain with ageing is only modest [15]. As noted in the Introduction, there are few studies with which to compare these results. A study of intravenous urograms from 1315 men aged 60 75 was used to provide a population surrogate for identifying the frequency of different aspects of radiographic hip OA. In total, 29% of a subset of 759 of these men, who provided information, reported ever having had hip pain, though again it was only in those with radiographic appearances at the extreme (2%) end of the radiographic severity spectrum that there was an association between pain and radiographic feature [9]. There are, however, some interesting messages that emerge. First, it would be inappropriate to conclude that an elderly patient with hip region pain is inevitably likely to have OA. Secondly, even severe radiographic change is relatively infrequently accompanied by continuing hip pain, and it is important to be cautious before linking the pain with the X-ray. Thirdly, the majority of younger male subjects in the community with severe OA will be pain-free, and in subjects with other indicators of hip disease, such as known childhood hip disorder, prior injury or limitation in movement, severe OA could still be present. Rheumatology Key messages There is a limited relationship between the presence of hip pain and the degree of radiographic change in the community. Severe hip OA in younger men is frequently pain-free. Acknowledgements This study was funded by the UK Arthritis Research Campaign (arc) and F.B. was an arc research training fellow in epidemiology. We are grateful to Dr Coope (The Waterhouse, Bollington Medical Practice) and Dr Sanders (Cheadle Medical Practice) and their staff for access to their patient population. The survey was coordinated by Liz Nahit, who also assisted in the reading of the radiographs. The authors have declared no conflicts of interest. References 1. Mcalindon TE, Cooper C, Kirwan JR et al. Knee pain and disability in the community. Br J Rheumatol 1992;31:189 92. 2. O Reilly SC, Muir KR, Doherty M. Knee pain and disability in the Nottingham community: association with poor health status and psychological distress. Br J Rheumatol 1998;37:870 3. 3. Hannan MT, Felson DT, Pincus T. Analysis of the discordance between radiographic changes and knee pain in osteoarthritis of the knee. J Rheumatol 2000;27:1513 7. 4. Dawson J, Linsell L, Zondervan K et al. Epidemiology of hip and knee pain and its impact on overall health status in older adults. Rheumatology 2004;43:497 504. 5. NIH Consensus Development Panel on Total Hip Replacement. Total hip replacement. JAMA 1995;273:1950 6.

Hip pain and radiographic change 341 6. Hadorn DC, Holmes AC. The New Zealand priority criteria project. Part 1: overview. Br Med J 1997;314:131 4. 7. Birrell F, Croft P, Cooper C, Hosie G, Macfarlane GJ, Silman A. Radiographic change is common in new presenters in primary care with hip pain. PCR Hip Study Group. Rheumatology 2000;39: 772 5. 8. Birrell F, Lunt M, Macfarlane G et al. Defining hip pain for population studies. Ann Rheum Dis 2004. In press. 9. Croft P, Cooper C, Wickham C et al. Defining osteoarthritis of the hip for epidemiologic studies. Am J Epidemiol 1990;132:514 22. 10. Cochrane WG. Sampling techniques. 3rd edition. New York, NY: John Wiley & Sons, 1977. 11. Felson DT. An update on the pathogenesis and epidemiology of osteoarthritis. Radiol Clin North Am 2004;42:1 9. 12. MacGregor AJ, Antoniades L, Matson M, Andrew T, Spector TD. The genetic contribution to radiographic hip osteoarthritis in women: results of a classic twin study. Arthritis Rheum 2000;43:2410 6. 13. Gelber AC. Obesity and hip osteoarthritis: the weight of the evidence is increasing. Am J Med 2003;114:158 9. 14. Croft P, Coggon D, Cruddas M et al. Osteoarthritis of the hip an occupational disease in farmers. Br Med J 1992;304:1269 72. 15. Symmons DPM. Population studies of musculoskeletal morbidity. In: Silman AJ, Hochberg MC, eds. Epidemiology of the rheumatic diseases, 2nd edition. Oxford, UK: Oxford University Press, 2001.