Impact of Persistent Hip or Knee Pain on OverallHealth Status in Elderly People: A LongitudinalPopulation StudyJILL DAWSON,1 LOUISE LINSELL,2 KRINA ZONDERVAN,2 PETER ROSE,2 ANDREW CARR,3

TONY RANDALL,2 AND RAY FITZPATRICK2

Objective. To investigate hip or knee symptoms in older persons from a longitudinal, population perspective, and todetermine the impact of persistent hip or knee pain on general health status over time.Methods. A postal questionnaire was sent to a random sample of 5,500 individuals ages >65 years containing the ShortForm 36 (SF-36) general health survey, Lequesne hip and knee indices, and a hip/knee pain severity item. Respondentsreporting hip or knee symptoms at baseline received an identical questionnaire 12 months later. Respondents wereclassified into a persistent pain group with either hip or knee pain at both baseline and followup, and a nonpersistent paingroup who reported hip or knee pain at baseline but no pain at followup.Results. At baseline, 1,305 (40.7%) of 3,210 eligible respondents reported hip or knee pain. At 1 year, 1,072 (82.1%) of1,305 individuals responded, of whom 820 (76.5%) remained symptomatic (the persistent group). In multivariate analysis,baseline factors identified as strongly related to having persistent pain were maximum Lequesne score (odds ratio [OR]1.09, P < 0.001), maximum hip/knee pain score (OR 1.61, P < 0.001), and number of painful hip and knee joints atbaseline (OR 1.48, P � 0.004). Following adjustment for age, sex, and baseline score, differences in mean SF-36 changescores of the 2 groups were significant for all dimensions except for mental health.Conclusion. In older persons, a symptomatic hip or knee frequently progresses in terms of worsening symptoms andaccrual of other symptomatic hip or knee joints. The impact of persistent symptoms on general health is substantial.

KEY WORDS. Hip; Knee; Longitudinal study.

INTRODUCTION

Demographically, individuals over 65 years of age are thefastest growing age group. This population growth hasconsiderable implications for health service and socioeco-nomic planning during the next 2 decades (1,2). There hasbeen much debate concerning the extent to which in-creases in longevity will be offset by increased morbidityat older ages (3). Diseases that are associated with im-

paired mobility are of particular concern due to their po-tential to increase social isolation and ultimately tothreaten a person’s ability to live independently. Irrespec-tive of their precise underlying etiology, symptomatic hipsor knees are associated with considerable mobility impair-ment in elderly persons (4). For these reasons, populationstudies of the natural history of such problems representvital pieces of information to the planners of future healthand social services.

A number of population-based studies have reported theprevalence of individuals affected by hip or knee pain, orby hip or knee osteoarthritis (OA) specifically (5–11).These studies focus on the hip or the knee, rather thanboth. Although the age profile of study participants variesbetween these different studies, all nevertheless agree thatthe prevalence of hip or knee disease is highest amongthose who are older than 65 years of age.

Although much information is now available concern-ing the prevalence of hip and knee symptoms in elderlypersons, less is known about the natural history of suchproblems. There are a number of reasons for this. One isthat studies of hip or knee problems vary regarding the

Supported by a grant from the NHS Executive (South-EastRegion).

1Jill Dawson, DPhil: Oxford Brookes University, Oxford,United Kingdom; 2Louise Linsell, MSc, Krina Zondervan,DPhil, MRC, Peter Rose, MB, Bchir, FRCGP, Tony Randall,MA, MRCGP, Ray Fitzpatrick, PhD: University of Oxford,Oxford, United Kingdom; 3Andrew Carr, FRCS: NuffieldOrthopaedic Centre, Oxford, United Kingdom.

Address correspondence to Jill Dawson, DPhil, Reader inHealth Services Research, School of Health and Social Care,Marston Road Campus, Oxford Brookes University, OxfordOX3 0BP.

Submitted for publication August 18, 2004; accepted inrevised form January 26, 2005.

Arthritis & Rheumatism (Arthritis Care & Research)Vol. 53, No. 3, June 15, 2005, pp 368–374DOI 10.1002/art.21180© 2005, American College of Rheumatology

ORIGINAL ARTICLE

368

definition of the problem, or take as their starting pointdifferent stages in the care pathway. Different studiestherefore also tend to consider different subgroups of thepopulation, so that, for instance, clinically-based studieswill naturally exclude individuals with hip or knee symp-toms who have not sought help. Understanding the naturalhistory of painful hip or knee conditions is also limited bythe fact that such symptoms may be caused by a number ofdifferent diseases or acute injuries, with different likelyoutcomes. In contrast, studies that are based upon specificconfirmed diagnoses, e.g., OA, are limited because theygenerally exclude individuals with hip or knee symptomsthat do not appear to have the particular underlying dis-ease of interest.

For all of these reasons, there is a dearth of longitudinal,population-based information of older persons with hip orknee symptoms. Such studies do well to focus on bothjoints because elderly persons with hip or knee symptomsfrequently have more than one symptomatic joint (12).

This report describes findings from a longitudinal sur-vey of hip and knee symptoms in elderly persons. Thecurrent study has 2 main goals, to examine the 12-monthcourse of hip and knee symptoms among a cohort of indi-viduals who reported hip or knee pain at baseline, and toassess the impact of persistent hip or knee pain on generalhealth status over time.

PATIENTS AND METHODS

Local research ethics committee approval was obtained forthe study (Applied and Qualitative Research Ethics Com-mittee reference A01.060).

Study population. A random sample of 5,500 Oxford-shire residents, ages 65 years and older, was obtained fromthe Oxfordshire Health Authority register representingJanuary 2002. Full details concerning the sample size cal-culations are provided elsewhere (12). A postal question-naire and cover letter were sent to all participants within a2-week period in April 2002 and followed up with 2 postalreminders (including a second copy of the questionnaire).Respondents who reported hip or knee symptoms at base-line were sent an identical followup questionnaire 12months later.

Questionnaire. The questionnaire was divided into ageneral section, a hip section, and a knee section. Thegeneral section contained a small number of demographicitems and the anglicized version of the Short Form 36(SF-36) general health questionnaire (13). The SF-36 con-tains 36 items and is widely used as a generic health statusinstrument. It provides scores on 8 dimensions: physicalfunctioning, role limitations due to physical problems,bodily pain, social functioning, general mental health, rolelimitations due to emotional problems, energy/vitality,and general health perceptions representing the last 4weeks. There is also an item that addresses health changeduring the last 12 months: “Compared to one year ago,how would you rate your general health now?” Scores foreach dimension range from 0 (poor health) to 100 (goodhealth).

The hip section began with a screening question using amodified version of the question used in the NationalHealth and Nutrition Examination survey (14), which hasalso been used in other studies: “During the past 12months, have you had pain in or around either of your hipson most days for 1 month or longer?” (6,9–11). Respon-dents who reported symptoms were asked additional ques-tions about their hip problem, including: “Do you knowthe cause of your hip pain?”

Severity during the last 4 weeks was assessed separatelyfor the right and left hip using the hip and knee disabilityindices defined by Lequesne et al (15,16). This produces acomposite measurement score ranging from 1 to 24 pointsbased on 11 items concerned with the presence of pain,discomfort, maximum walking distance, and ability tofunction. Patients were also asked to rate pain severity ineach hip during the last 4 weeks on a scale ranging from“none” to “very severe.” Information regarding the pres-ence of serious comorbidity was assessed using the ques-tion: “Do you currently have another health problem thatis at least as bad as the problem with your hip?”

The corresponding knee section of the questionnairewas identical to the hip section except that the word“knee” substituted for the word “hip,” and 5 items differedin the Lequesne index. The order of the hip and kneesections was reversed in half of the questionnaires to en-sure that the completion rate for the screening question ineach section was not biased by its position in the ques-tionnaire. The followup questionnaire was identical to thebaseline questionnaire.

Statistical analysis. Pearson’s chi-square test was usedto test for differences in baseline characteristics betweenrespondents to the followup questionnaire and those lostto followup. Respondents with followup data were classi-fied into 2 groups for analysis: a persistent pain group whoreported either hip or knee pain at both baseline andfollowup survey, and a nonpersistent pain group whoreported hip or knee pain at baseline but no pain at follow-up. Three severity groups, conforming to the thresholdvalues suggested by Lequesne et al (15), were definedseparately for the hips and knees: a mild-moderate group(Lequesne score 1–7), a severe group (Lequesne score8–13), and an extremely severe group (Lequesne score�14). Similarly, 3 pain groups (based on the pain severityitem) were defined as mild, moderate, and severe. Whenrespondents reported bilateral symptoms, the most severeand most painful joint was used for these classifications.

The association of various sociodemographic, generalhealth, and joint-specific characteristics with having per-sistent hip or knee pain at followup was examined usingPearson’s chi-square test. Factors with a P value less than0.2 in the univariate analysis were evaluated in a multi-variate logistic regression model to investigate which fac-tors were most strongly predictive of persistent pain. Gen-eral mobility items (taken from the Lequesne index) werenot included in the model in order to avoid collinearity.Model fit was assessed using the Hosmer and Lemeshowgoodness-of-fit test, and the classification statistic (propor-

Impact of Hip or Knee Pain on Overall Health 369

tion of observations correctly classified) was computed(17).

The 8 dimensions of the SF-36 were calculated for thepersistent and nonpersistent groups at baseline and atfollowup. The mean change scores for each dimensionwere calculated and found to be normally distributed forboth groups. Analysis of covariance was used to estimatethe difference in the group mean change scores. This ana-lysis included an adjustment for the patients’ baselinescores to correct for the phenomenon of regression to themean (18,19). All analyses were conducted using STATAversion 8.0 (Stata, College Station, TX).

RESULTS

Self-reported symptomatic hips and knees at baselineand followup. Of the 5,500 individuals originally se-lected, 119 patients (2.2%) were deceased and 342 patients(6.2%) were no longer living at the given address. Of theremaining 5,039 eligible participants, 1,348 (26.8%) didnot respond, 201 (4.0%) were unable to participate, and149 (3.0%) declined participation. Therefore, 3,341(66.3%) individuals completed and returned the initialquestionnaire. Age- and sex-specific response rates variedas follows: of respondents ages 65–74 years, 73.1% werewomen and 70.6% were men; of respondents ages 75–84,

62.3% were women and 68.3% were men; and of respon-dents ages 85 years and older, 43.9% were women and54.5% were men.

Figure 1 shows the study flow of participants based ontheir response to the screening question at baseline and at1-year followup. At baseline, 263 (8.3%) of 3,175 respon-dents reported hip pain only, 695 (21.8%) of 3,194 re-ported knee pain only, and 347 (11.3%) of 3,076 reportedhip and knee pain. Overall, 1,072 (82.1%) of 1,305 symp-tomatic individuals had responded at 1-year followup,with no difference in the response rates between the 3baseline groups (hip pain only, knee pain only, hip andknee pain; P � 0.27). Table 1 shows the baseline sociode-mographic and other characteristics of respondents at1-year followup versus those patients lost to followup.

Hip and knee symptoms reported at followup are shownin the final stage of Figure 1. Of the 1,072 symptomaticindividuals who responded to the second questionnaire, atotal of 820 (76.5%) remained symptomatic (the persistentpain group), and 252 (23.5%) reported no hip or knee pain(the nonpersistent pain group) at 1-year followup. Amongindividuals who had reported hip pain only at baseline, 83(37.4%) of 222 had no reported hip pain at followup and38 (17.1%) of 222 reported both hip and knee pain. Amongindividuals who had reported knee pain only at baseline,164 (28.6%) of 574 had no reported knee pain at followup

Figure 1. Study flow diagram based on response to screening question at baseline and 1-year followup.

370 Dawson et al

and 38 (12.0%) of 574 reported both hip and knee pain.The majority of respondents reporting both hip and kneepain at baseline remained symptomatic in at least 1 jointtype at followup (238 [86.2%] of 276 respondents).

The proportion of respondents reporting hip or kneepain was the same at both baseline and followup, regard-less of whether the hip or knee section was positioned firstin the questionnaire (P � 0.2 in all cases).

Factors associated with persistent hip or knee pain.The characteristics of individuals in the persistent paingroup compared with those in the nonpersistent paingroup are presented in Table 2. Sex, smoking status, andadditional health problems were not significantly associ-ated with having persistent hip and knee pain in univari-ate analyses (P � 0.2); therefore, these factors were notincluded in the multivariate model. All other variables inTable 2 with P � 0.2 were included except for the generalmobility items (which form part of the Lequesne indexalready included in the model). Factors identified in themultivariate analysis that were most strongly related tohaving persistent pain were maximum baseline Lequesnescore (odds ratio [OR] 1.09, 95% confidence interval [95%CI] 1.04–1.14, P � 0.001), maximum baseline pain score

(OR 1.61, 95% CI 1.31–1.99, P � 0.001), and number ofpainful hip and knee joints at baseline (OR 1.48, 95% CI1.13–1.93, P � 0.004). The Hosmer-Lemeshow goodness-of-fit test indicated a satisfactory model fit (P � 0.45), andthe percentage of observations correctly classified by themodel was 79%.

Change in overall health status over a period of 1 year.The mean change scores for the SF-36 dimensions arereported in Table 3. The individuals in the persistent paingroup had baseline scores that were consistently lower(poorer) than those in the nonpersistent pain group acrossall dimensions. The mean change scores for the persistentpain group were all negative values, indicating that theoverall health status in this group had deteriorated overthe course of the year, whereas most of the mean changescores for the nonpersistent pain group were positive,suggesting an improvement in most aspects of generalhealth. The estimated differences in the mean changescores of the 2 groups, adjusting for age, sex, and baselinescore, were significant for all dimensions except for mentalhealth. A negative difference in mean change score indi-cates that the persistent pain group experienced more de-

Table 1. Baseline characteristics of respondents at 1-year followup compared with respondents lost to followup*

Lost to followupNo./total (%)

1-year followupNo./total (%) P†

Age group, years65–74 98/233 (42.1) 620/1,072 (57.8)75–84 92/233 (39.5) 383/1,072 (35.7) � 0.001�85 43/233 (18.5) 69/1,072 (6.4)

SexMale 89/233 (38.2) 452/1,072 (42.2)Female 144/233 (61.8) 620/1,072 (57.8) 0.27

Lives alone 0.002Yes 91/221 (41.2) 318/1,047 (30.4)

Homeowner � 0.001Yes 137/217 (63.1) 810/1,036 (78.2)

EducationHave school qualifications 62/217 (28.6) 370/994 (37.2) 0.02Have a degree 18/213 (8.5) 120/969 (12.4) 0.11Have professional qualifications 37/207 (17.9) 252/967 (26.1) 0.01

Current regular smokerYes 24/219 (11.0) 67/1,020 (6.6) 0.02

Body mass index �30 (obese)Male 15/82 (18.3) 59/424 (13.9) 0.30Female 30/128 (23.4) 126/580 (21.7) 0.67

Severity group at baseline (Lequesne)Mild-moderate 32/156 (20.5) 276/829 (33.3)Severe 58/156 (37.2) 332/829 (40.1) � 0.001Extremely severe 66/156 (42.3) 221/829 (26.7)

Pain score group at baselineMild 56/187 (30.0) 329/947 (34.7)Moderate 82/187 (43.9) 403/947 (42.6) 0.38Severe 49/187 (26.2) 215/947 (22.7)

Other health problems at least as bad as hip/knee problemYes 99/158 (62.7) 494/830 (59.5) 0.46

* Totals � 233 in lost to followup group, or � 1,072 in 1-year followup group indicate missing values for a characteristic.† Pearson’s chi-square test.

Impact of Hip or Knee Pain on Overall Health 371

terioration and less improvement in health compared withthe nonpersistent pain group over the 1-year period.

These findings were also largely reflected in the SF-36 itemthat asks about health change in the previous 12 months.Among the respondents with persistent hip or knee pain, 68(8.5%) of 802 regarded their health as better, 511 (63.7%) of802 thought their health was the same, and 223 (27.8%) of802 thought their health was worse than the year before. Thecorresponding proportions for those who were not symptom-atic at followup were 39 (16.1%) of 243 respondents, 164(67.5%) of 243 respondents, and 40 (16.5%) of 243 repon-dents, respectively. The difference between the 2 groups wassignificant (P � 0.001, Pearson’s chi-square test).

DISCUSSION

In a previous analysis, our findings showed that �40% ofpersons age �65 years have hip or knee pain lasting 1month or more in a given year, and that symptoms fre-quently affect more than 1 joint, usually involving bothlegs (12). In addition, in cross-sectional analyses, a signif-icant dose-response relationship was demonstrated in thatstudy for worsening general health status, according to thenumber of weight-bearing joints that were reported to besymptomatic.

Our current analysis reports longitudinal findings fromthe same study, which evaluated individuals at 12-

Table 2. Baseline characteristics of people with persistent hip or knee pain compared with those who did not have symptomsat 1-year followup*

Characteristics

Nonpersistent hipor knee painNo./total (%)

Persistent hipor knee painNo./total (%) P†

Age group, years 0.00765–74 167/252 (66.3) 453/820 (55.2)75–84 70/252 (27.8) 313/820 (38.2)�85 15/252 (6.0) 54/820 (6.6)

Sex 0.74Male 104/252 (41.3) 348/820 (42.4)Female 148/252 (58.7) 472/820 (57.6)

Lives alone 65/247 (26.3) 253/800 (31.6) 0.11Is a homeowner 208/245 (84.9) 602/791 (76.1) 0.004Education

Have school qualifications 99/231 (42.9) 271/763 (35.5) 0.04Have a degree 39/231 (16.9) 81/738 (11.0) 0.02Have professional qualifications 72/230 (31.3) 180/737 (24.4) 0.04

Current regular smoker 14/241 (5.8) 53/779 (6.8) 0.59Body mass index �30 (obese)

Male 11/97 (11.3) 48/327 (14.7) 0.40Female 20/141 (14.2) 106/439 (24.2) 0.01

Other health problem(s) at least as bad as hip/knee problem 101/176 (57.4) 393/654 (60.1) 0.52Joints affected at baseline � 0.001

Hip(s) only 69/252 (27.4) 153/820 (18.7)Knee(s) only 145/252 (57.5) 429/820 (52.3)Both hip(s) and knee(s) 38/252 (15.1) 238/820 (29.2)

Number of joints (hips and knees) affected at baseline � 0.0011 163/252 (64.7) 360/815 (44.2)2 74/252 (29.4) 345/815 (42.3)3 or 4 15/252 (6.0) 110/815 (13.5)

Severity group at baseline (Lequesne) � 0.001Mild-moderate 94/169 (55.6) 182/660 (27.6)Severe 56/169 (33.1) 276/660 (41.8)Extremely severe 19/169 (11.2) 202/660 (30.6)

Pain score group at baseline � 0.001Mild 117/211 (55.5) 212/736 (28.8)Moderate 78/211 (37.0) 325/736 (44.2)Severe 16/211 (7.7) 199/736 (27.0)

Used painkillers for hip/knee pain in last 4 weeks 53/174 (30.5) 358/663 (54.0) � 0.001Months since onset of hip/knee problem, median/total (IQR) 24/125 (12–84) 60/480 (24–120) � 0.001‡Maximum distance can walk is �100 yards 10/178 (5.6) 70/689 (10.2) 0.06Use a walking aid 45/207 (21.7) 268/739 (36.3) � 0.001Cannot go up and down a flight of stairs 6/210 (2.9) 31/740 (4.2) 0.38

* Totals � 252 in nonpersistent group, or � 820 in persistent group indicate missing values for a characteristic. IQR � interquartile range.† Unless otherwise indicated, P values determined by Pearson’s chi-square test.‡ By Wilcoxon rank-sum test.

372 Dawson et al

months’ followup who had symptomatic hips or knees atbaseline.

Our analysis revealed that of those respondents whoreported hip or knee pain at baseline, 76.5% remainedsymptomatic 12 months later. Approximatley 14% of therespondents who had hip or knee pain only at baselinereported having pain in both types of joints 12 monthslater.

In a multivariate analysis, factors found to be strongpredictors of hip or knee pain persisting for at least 1 yearincluded the severity of symptoms and disability reportedat baseline (Lequesne and pain score of the most severejoint) and the total number of hip and knee joints that weresymptomatic at baseline.

A striking finding was that hip or knee pain that hadpersisted for at least one year led to a significant differencein an individual’s general health within that time period(as measured by the SF-36 general health survey instru-ment), compared with the general health status of individ-uals whose hip and knee symptoms had resolved withinthat same year. This was the case for all health statusdimensions (with the exception of mental health), al-though the most striking difference was the effect on phys-ical role limitation and pain dimensions. Nevertheless,although the evidence presented here is consistent withcausal direction (from persistent joint pain to worseningbroader aspects of health), we cannot rule out the possi-

bility that in some individuals, hip or knee symptomsreported at baseline represented only 1 element of a morewidespread condition that was associated with generaldecline over the course of a year.

There are a number of limitations to this study. The firstconcerns our definition of persistent pain, because re-ported hip or knee pain could theoretically have lasted foronly 1 month in the year preceding the baseline question-naire and for 1 month during the followup year. We aretherefore unable to specify in detail whether the pain ineach symptomatic individual was recurrent, episodic, orchronic/incessant.

A second limitation concerns the respondent’s responserates. Study respondents at baseline had a relatively youngage distribution for individuals ages 65 and older. In par-ticular, there was a low initial response rate in elderlywomen, although we did adjust for age and sex in theanalysis of the SF-36 scores. Further bias in this regard wasintroduced at followup, as nonrespondents to the fol-lowup questionnaire were more likely to be in the oldestage group, to live alone, and not to be a home owner. Theyalso appeared to be less educated, more likely to be regularsmokers, and more likely to have more severe hip and kneepain at baseline. It is likely that many of those lost tofollowup had died before the followup questionnaire wassent out, but we were unable to confirm this possibility.

Large-scale population studies of health status in elderly

Table 3. Change in SF-36 scores of participants with persistent hip or knee pain compared with those who did not havesymptoms at 1-year followup*

SF-36 dimension

Nonpersistent hip orknee pain (n � 252)

Mean � SD

Persistent hip orknee pain (n � 820)†

Mean � SD

Estimated differencein mean change scores

(95% CI)‡ P

Physical function Baseline 64.5 � 26.4 49.7 � 28.7Followup 65.5 � 26.7 46.0 � 28.5Change 0.97 � 20.1 �3.7 � 15.7 �7.3 (�10.2, �4.3) � 0.001

Role limitation (physical) Baseline 62.0 � 41.1 47.1 � 43.2Followup 67.3 � 38.9 40.4 � 41.9Change 5.3 � 39.1 �6.8 � 41.1 �18.8 (�24.3, �13.2) � 0.001

Pain Baseline 64.0 � 22.8 51.4 � 23.3Followup 71.8 � 21.8 49.9 � 22.6Change 7.8 � 22.4 �1.6 � 20.1 �14.7 (�17.4, �11.9) � 0.001

Social function Baseline 83.8 � 22.1 73.2 � 28.9Followup 83.9 � 24.6 71.0 � 29.0Change 0.15 � 21.3 �2.3 � 23.8 �5.7 (�9.0, �2.5) 0.001

Mental health Baseline 78.1 � 16.6 75.5 � 16.7Followup 78.0 � 15.4 74.9 � 16.7Change �0.14 � 14.0 �0.59 � 14.0 �1.4 (�3.3, 0.62) 0.18

Role limitation (emotional) Baseline 84.0 � 29.7 74.2 � 39.0Followup 80.3 � 33.8 69.8 � 41.1Change �3.7 � 36.2 �4.5 � 39.9 �5.3 (�10.8, 0.10) 0.05

Energy and vitality Baseline 56.8 � 18.7 50.2 � 20.3Followup 57.4 � 19.1 48.8 � 20.2Change 0.65 � 13.7 �1.4 � 14.7 �3.6 (�5.9, �1.3) 0.002

General health perception Baseline 62.3 � 18.5 57.6 � 20.1Followup 63.5 � 19.3 57.1 � 19.4Change 1.2 � 14.0 �0.54 � 13.5 �2.7 (�4.8, �0.70) 0.009

* SF-36 � Short Form 36; 95% CI � 95% confidence interval.† People with missing values for an SF-36 dimension at baseline or followup were excluded from the analysis.‡ Analysis of covariance adjusted for age, sex, and baseline score. Negative differences in mean change scores imply that the persistent pain groupexperienced more deterioration/less improvement in general health than the nonpersistent pain group over 1-year period.

Impact of Hip or Knee Pain on Overall Health 373

persons are uncommon in the UK, particularly longitudi-nal studies. Based on this methodology, this study contrib-utes additional information to the understanding of jointproblems and mobility in elderly persons. Our sample isbroadly representative of elderly persons in southeast En-gland. This area is also somewhat more affluent than theUK as a whole (12).

Two particular findings emerged that are worth empha-sizing because they have implications for further researchor for health and social care provision. The first is thatindividuals ages �65 years who have 1 symptomatic hipor knee joint frequently progress, not only in terms ofworsening symptoms affecting the index joint, but also interms of the accrual of other symptomatic hip and kneejoints. Reasons why additional joints become symptomaticare likely to relate to underlying etiology and specificassociated risk factors (although the relationship betweenadvanced age and prevalence of symptomatic arthritis, forinstance, is not well understood at present) (4,20–22), butother possible reasons may include the presence of behav-ioral factors (such as reduced levels of exercise) (23) andpsychological factors (24). Such considerations were be-yond the scope of this study. However, in terms of main-taining mobility and independence, the reasons why thenumber of symptomatic weight-bearing joints tends to in-crease in some elderly persons and not in others warrantfurther research, together with research into potential in-terventions to prevent or slow down this process.

The second main finding was that even during 1 year,the impact of having persistent hip or knee pain on aperson’s overall general health is substantial in the elderlycompared with individuals whose hip or knee symptomsresolve, and that improvements in overall general healthcan be demonstrated in those whose hip and knee symp-toms disappear. This finding is supported by evidencefrom other studies that show that decrements in lowerextremity impairment are a significant predictor of disabil-ity in activities of daily living (4,25).

Our findings indicate that improved pain care for symp-tomatic hips and knees, or more rapid access to jointreplacement surgery, could substantially improve thequality of life of many elderly persons, and possibly slowdown the accrual of further lower joint problems. With arapidly increasing elderly population, we need to invest inresources that help to prevent, treat, or at least minimizehip and knee symptoms because they are widespread andpersistent in this age group and will increasingly representa huge burden to the National Health Service and to soci-ety as a whole.

REFERENCES

1. Evandrou M, Falkingham J. Looking back to look forward:lessons from four birth cohorts for ageing in the 21st century.Popul Trends 2000;99:27–36.

2. Khaw KT. How many, how old, how soon? BMJ 1999;319:1350–2.

3. Manton KG. Past and future life expectancy increases at laterages: their implications for the linkage of chronic morbidity,disability, and mortality. J Gerontol 1986;41:672–81.

4. Hughes SL, Dunlop D, Edelman P, Chang RW, Singer RH.Impact of joint impairment on longitudinal disability in el-derly persons. J Gerontol 1994;49:S291–300.

5. Fear J, Hillman M, Chamberlain MA, Tennant A. Prevalenceof hip problems in the population aged 55 years and over:access to specialist care and future demand for hip arthro-plasty. Br J Rheumatol 1997;36:74–6.

6. Frankel S, Eachus J, Pearson N, Greenwood R, Chan P, PetersTJ, et al. Population requirement for primary hip-replacementsurgery:a cross-sectional study. Lancet 1999;353:1304–9.

7. Peat G, McCarney R, Croft P. Knee pain and osteoarthritis inolder adults: a review of community burden and current useof primary health care. Ann Rheum Dis 2001;60:91–7.

8. Tennant A, Fear J, Pickering A, Hillman M, Cutts A, Cham-berlain MA. Prevalence of knee problems in the populationaged 55 years and over: identifying the need for knee arthro-plasty. BMJ 1995;310:1291–3.

9. O’Reilly SC, Muir KR, Doherty M. Knee pain and disability inthe Nottingham community: association with poor health sta-tus and psychological distress. Br J Rheumatol 1998;37:870–3.

10. McAlindon TE, Cooper C, Kirwan JR, Dieppe PA. Knee painand disability in the community. Br J Rheumatol 1992;31:189–92.

11. Juni P, Dieppe P, Donovan J, Peters T, Eachus J, Pearson N, etal. Population requirement for primary knee replacementsurgery: a cross-sectional study. Rheumatology (Oxford) 2003;42:516–21.

12. Dawson J, Linsell L, Zondervan K, Rose P, Randall T, Carr A,et al. Epidemiology of hip and knee pain and its impact onoverall health status in older adults. Rheumatology (Oxford)2004;43:497–504.

13. Brazier JE, Harper R, Jones NM, O’Cathain A, Thomas KJ,Usherwood T, et al. Validating the SF-36 health surveyquestionnaire: new outcome measure for primary care. BMJ1992;305:160–4.

14. Anderson JJ, Felson DT. Factors associated with osteoarthritisof the knee in the first national Health and Nutrition Exami-nation Survey (HANES I): evidence for an association withoverweight, race, and physical demands of work. Am J Epi-demiol 1988;128:179–89.

15. Lequesne MG, Mery C, Samson M, Gerard P. Indexes of se-verity for osteoarthritis of the hip and knee: validation–valuein comparison with other assessment tests. Scand J Rheuma-tol Suppl 1987;65:85–9.

16. Lequesne MG. The algofunctional indices for hip and kneeosteoarthritis. J Rheumatol 1997;24:779–81.

17. Hosmer DW Jr, Lemeshow S. Applied logistic regression. NewYork: John Wiley & Sons; 2000.

18. Twisk JW. Applied longitudinal data analysis for epi-demiology: a practical guide. Cambridge: Cambridge Univer-sity Press; 2003.

19. Vickers AJ, Altman DG. Statistics notes: analysing controlledtrials with baseline and follow up measurements. BMJ 2001;323:1123–4.

20. Bagge E, Bjelle A, Eden S, Svanborg A. A longitudinal study ofthe occurrence of joint complaints in elderly people. AgeAgeing 1992;21:160–7.

21. Dougados M, Gueguen A, Nguyen M, Berdah L, Lequesne M,Mazieres B, et al. Radiological progression of hip osteo-arthritis: definition, risk factors and correlations with clinicalstatus. Ann Rheum Dis 1996;55:356–62.

22. Felson DT, Zhang Y, Hannan MT, Naimark A, Weissman BN,Aliabadi P, et al. The incidence and natural history of kneeosteoarthritis in the elderly: the Framingham OsteoarthritisStudy. Arthritis Rheum 1995;38:1500–5.

23. Van Baar ME, Assendelft WJ, Dekker J, Oostendorp RA,Bijlsma JW. Effectiveness of exercise therapy in patients withosteoarthritis of the hip or knee: a systematic review of ran-domized clinical trials. Arthritis Rheum 1999;42:1361–9.

24. Van Baar ME, Dekker J, Lemmens JA, Oostendorp RA, BijlsmaJW. Pain and disability in patients with osteoarthritis of hip orknee: the relationship with articular, kinesiological, and psy-chological characteristics. J Rheumatol 1998;25:125–33.

25. Jette AM, Branch LG, Berlin J. Musculoskeletal impairmentsand physical disablement among the aged. J Gerontol 1990;45:M203–8.

374 Dawson et al