Abstract

Introduction

Osteoarthritis (OA) is a major cause of disability worldwide ¹. Over the past years, interest has grown among the scientific community, pharmaceutical companies, and regulatory agencies in the development of drugs that might influence the natural history of structural changes in OA by preventing, retarding, or reversing cartilage breakdown. Interest exists, therefore, in identifying a valid, dichotomous outcome variable that reflects the natural history of structural changes in OA. In particular, interest has grown in using the requirement of total joint replacement (TJR) as a “hard” endpoint ^2,3. Limitations exist, however, in the use of such an outcome. Performance of TJR is a measure of utilization and not of a health state. Numerous non-health related factors have been shown to influence utilization including patient race, ethnicity, income, activity level and preferences among others, and other non-musculoskeletal health factors influence the decision to undergo TJR including comorbidity ^2-10. Thus, a better alternative might be to change “time to TJR” to “time to fulfill the criteria for TJR” ¹¹. In this context and as described elsewhere ^12,13, an international working group was created under the auspices of Osteoarthritis Research Society International (OARSI) and Outcome Measures in Rheumatology Clinical Trials (OMERACT). The group's charge was to elaborate a set of criteria defining a state corresponding to recommendation for TJR in patients with symptomatic knee and hip OA, for use in clinical trials evaluating potential disease-modifying drugs and other interventions in OA. It was decided that the domains of pain, physical function and joint structure on radiographs ^14-16 would be combined as a surrogate measure of outcome. The consensus was to consider the level of symptoms (i.e., pain and function) at one point, and a definition of radiological progression between 2 time-points ¹⁶.The final binary outcome could then be used as a definition for “responders/non-responders” in OA clinical trials. For each of these domains, a categorical outcome needs to be used to render combination of the domains feasible. To this end, it is necessary to categorize or dichotomize the continuous variables pain and functional disability.

Thus, the objective of the present study was to define cut points for both pain and functional disability, leading to a joint replacement indication. To this end, a data-driven approach, based on real patient data, was chosen.

This article presents the results of a large cross-sectional study performed to define cut-point levels for pain and functional disability among patients with hip or knee OA being evaluated by orthopaedic surgeons for possible need of TJR. The goal was to use these cut offs to develop a theoretical indication for TJR, in hip and knee OA.

Patients and Methods

Results

Patient characteristics (Table 1)

Table 1

Patients' characteristics

	All patients N=1909	Knee patients N=1130	Hip patients N=779
Age, years	66.4±10.9	67.5±10.4	64.9±11.4
Sex, N (%) women	1086 (58.1)	657 (58.9)	429 (56.9)
OA symptom duration, years	5.4±6.9	6.3±7.7	4.1±5.5
Body mass index, kg/m2	29.9±6.3	31.0±6.8	28.3±5.2
Pain, ICOAP score	51.6±22.3	50.3±22.0	53.3±22.6
Functional disability, HOOS-PS/KOOS-PS scores	56.5±20.0	55.5±18.8	57.8±21.5
Pain, WOMAC subscale	54.0±21.0	52.5±20.8	56.3±21.1
Function, WOMAC subscale	57.0±20.5	55.2±20.2	59.5±20.8
Radiographic joint space narrowing, N (%)*
< 25%	95 (10.7)	67 (13.0)	28 (7.8)
25-50%	131 (14.7)	95 (18.5)	36 (9.6)
50-75%	274 (30.8)	159 (31.0)	115 (30.6)
> 75%	389 (43.8)	192 (37.4)	197 (52.4)

Open in a separate window

Results are presented as mean±standard deviation unless otherwise mentioned. Pain and functional disability were linearly transformed to 0-100 scores where 100=worst state.

HOOS-PS/KOOS-PS: Hip/knee knee disability and Osteoarthritis Outcome Score physical function short-form. WOMAC: Western Ontario and McMaster Universities Osteoarthritis Index.

^*X-ray scoring was only available for 889 patients and % are % of available data.

In all, 1974 patients were included; 1909 had an answer for the gold standard question and were analysed: 1130 knee OA and 779 hip OA patients (Table 1). The patients were included in Europe (N=1050), Australia (N=394), the United States of America (N=261), and Canada (N=204). Supplementary file 1 shows the characteristics of the patients from the different centers.

Mean age of the patients was 66.4 (standard deviation, SD: 10.9) years, 58.1% were women, mean OA duration was reported as 5.4 (SD 6.9) years. Of the 1909 patients, 628 (55.6%) knee patients and 574 (73.7%) hip patients were recommended for TJR. The recommendation was mainly related to the surgeon stating TJR was indicated (91.7% of indications) and much less often to the answers ‘although the symptoms are severe enough, the patient declined surgery’ (4.0%) or ‘there were comorbidities’ (4.3%). The frequency of indication for TJR varied across countries, from 33.8% with an indication for surgery among the patients from the Italian center, to 87.9% among the patients from the Czech Republic center.

Pain assessed by ICOAP and functional disability by HOOS-PS/KOOS-PS (Table 2)

Table 2

Symptom levels and radiographic severity according to recommendation for TJR

	Knee OA: TJR+ N=628	Knee OA: TJR − N=502	Hip OA: TJR+ N=574	Hip OA: TJR − N=205
Pain, ICOAP score	53.7 (52.0, 55.5)	45.9 (44.0, 47.9)	57.3 (55.6, 59.1)	42.4 (39.1, 45.7)
Functional disability, HOOS-PS/KOOS-PS scores	58.1 (56.6, 59.7)	52.3 (50.5, 54.0)	61.4 (59.8, 63.1)	47.4 (44.1, 50.7)
Pain, WOMAC subscale	56.4 (54.8, 57.9)	47.3 (45.3, 49.4)	59.8 (58.3, 61.4)	45.9 (42.4, 49.3)
Function, WOMAC subscale	59.0 (57.4, 60.5)	50.3 (48.3, 52.4)	63.3 (61.7, 64.9)	48.7 (45.3, 52.2)
Duration of symptoms at the current level, months	11.0 (6.2,15.2)	5.9 (2.3, 10.2)	6.9 (3.5, 11.5)	5.9 (2.1, 10.6)
Radiographic joint space narrowing, N (%)*
< 25%	3 (1.2)	64 (23.6)	3 (1.1)	25 (22.3)
25-50%	24 (10.0)	70 (25.8)	14 (5.3)	22 (19.6)
50-75%	85 (35.4)	74 (27.3)	75 (28.5)	39 (34.8)
> 75%	128 (53.3)	63 (32.3)	171 (65.0)	26 (23.2)

Open in a separate window

Results are presented as mean (95% confidence interval) except for radiographic results. Pain and functional disability were linearly transformed to 0-100 scores where 100=worst state. TJR+: indication for TJR. TJR-: no indication for TJR. For other abbreviations please see Table 1.

^*% of available data

Scores for pain and functional disability were not normally distributed (Figure 1 for online version only), but showed a wide spread in severity of symptoms. Pain had the following distribution in knee OA: mean±SD 50.3±22.0, median 50.0 (first quartile=31.8, third quartile =68.2, range 0-100) and in hip OA mean±SD 53.3±22.6, median 54.5 (first quartile=4.1, third quartile=70.5, range 0-100). Functional impairment had the following distribution in knee OA: mean±SD 55.5±18.8, median 51.2 (first quartile=42.0, third quartile=66.6, range 0-100) and in hip OA mean±SD 57.8±21.5, median 55.9 (first quartile=41.7, third quartile=74.8, range 0-100).

Pain and functional disability levels and their duration, for those who did versus did not receive a TJR recommendation, are shown in Table 2. Patients meeting the gold standard had higher symptom levels. For knee/hip patients pooled, mean pain was 55.5 [95% confidence interval 54.2, 56.8] for those with TJR recommendation versus 44.9 [43.2, 46.6] for those without TJR recommendation (p<0.0001). Mean functional impairment was 59.8 [58.7, 60.9] for those with TJR recommendation versus 50.9 [49.3, 52.4], for those without TJR recommendation (p<0.0001). However, there was a wide overlap in symptom levels between groups: almost 50% of patients in the lowest decile of symptom scores were considered candidates for TJR, whereas only 75% of patients in the highest decile were considered candidates (Figure 2 for online version only).

Univariate ROC curves

Taking pain and function separately, in the pooled hip/knee population, it was not possible to determine relevant cut points defining recommendation for TJR (Figure 1). The AUCs for the ROC curves for pain and function versus the gold standard were 0.64 [95% confidence interval, 0.61, 0.67] and 0.63 [0.60, 0.66}, respectively. Thus, we had to accept the null hypothesis (i.e., that pain and functional disability levels do not distinguish patients with versus without a recommendation for TJR). The cut points had low diagnostic properties: e.g., for a specificity of 0.90, the sensitivity was only 0.23 for pain and 0.24 for function; i.e., the positive and negative likelihood ratios were only (1.17; 0.43) for pain and (1.18; 0.42) for physical disability. After stratifying on radiographic severity, the AUCs were not much improved (AUCs ranging from 0.65 to 0.68 for pain, and 0.60 to 0.63 for function, respectively) and cut points assessed had low diagnostic properties (data not shown).

Open in a separate window

Figure 1

Ability of pain and functional impairment severity to predict indication for TJR in 1909 hip or knee OA patients

1a: pain, 1b: functional disability

AUC for curve 1a: 0.64 (95% confidence interval, 0.61-0.67), AUC for curve 1b: 0.63 (95% confidence interval, 0.60-0.66).

Sensitivity analyses

Several sensitivity analyses including use of alternate measures, specifically the WOMAC pain and function subscales (supplementary file 2); and changing the gold standard to true indication for TJR (i.e., not considering patients with severe status but comorbidites or patient refusal as recommendations for TJR), did not modify the results (data not shown).

The 75^th percentile technique gave 89 as the value of the sum (pain+function) defining 75% of the population which had an indication for TJR (respectively, 87 and 92, for knee and hip). When applying the cut point of 89 to the whole population, 59% of the patients were above that level; specificity was 0.51, sensitivity was 0.66, the positive and negative likelihood ratios were (1.34; 0.86).

Excluding patients from the United Kingdom did not modify the conclusions (data not shown).

However, analysing the participants with hip or knee OA separately, the association between symptoms and surgery was stronger in the hip than in the knee (Figure 2). The AUCs of the ROC curves of the sum pain+function were higher in hip OA (AUC, 0.70, 95% CI, 0.66-0.75) than in knee OA (AUC, 0.60, 95% CI, 0.56-0.64). However, even so, for hip patients the cut-points assessed had low diagnostic properties; e.g. the cutpoint leading to a specificity of 90% was 66 (sum pain+function); for that cutpoint, for a specificity of 0.92, the sensitivity was only 0.31; i.e., the positive and negative likelihood ratios were only (1.35; 0.23).

Open in a separate window

Figure 2

Ability of the sum (pain+function) to predict indication for TJR, in knee and hip OA separately

1a: knee, 1b; hip.

AUCs for the ROC curves: 1a 0.60 (95% confidence interval, 0.56-0.64), 1b 0.70 (95% confidence interval, 0.66-0.75).

We also showed when analysing the centers separately (regrouped by country), that in certain centers pain and function were more strongly related to receipt of a TJR recommendation than in other centers (e.g., AUC of the ROC curves for the sum pain+function in hip OA, 0.75 to 0.86 in centers in Canada, France, Germany and Australia as compared with 0.54 to 0.67 in centers in The United Kingdom, the United States, the Netherlands and the Czech Republic,(supplementary file 3).

Discussion

This large-scale international study was launched to determine whether self-reported measures of pain and function could be used to accurately identify patients with OA whose surgeons recommended them for total hip or knee arthroplasty. The first conclusion of this work is that, indeed, among patients with hip and knee OA referred to an orthopedic surgeon, the level of symptoms was higher among patients for whom TJR was indicated by the orthopedic surgeon. Both the level of pain and self-reported functional impairment were independently, though weakly, predictive of the surgeon's recommendation for TJR. The second conclusion is that we could not find a cut point for pain and or physical disability that accurately discriminated across different countries, patients who did versus did not receive a TJR recommendation, as the AUCs for ROC curves were low (< 0.65). Radiographic severity, when available, was a strong predictor of recommendation for TJR but stratifying by radiographic joint status did not modify our conclusions.

Factors consistently predicting TJR are symptom levels and radiographic severity. Less consistent predictors have included gender, age and current treatment ^{2, 4-9}. In the present study, the mean values of pain and function in the group of patients considered candidates for surgery by the surgeons were consistent with previously reported data in this area ^27-32. We confirmed here that both pain and functional disability are independent predictors of recommendation by a surgeon for TJR; however, previous studies did not include a control group to attempt to determine cut points for patient-reported outcomes. In this study, using the ICOAP and HOOS-PS/KOOS-PS, though pain and function were correlated (as could be expected), the correlation was only moderate, which indicates that pain and function using these scores are not redundant when analysing OA patients. Furthermore, we also found that the duration of the symptoms at their current level was an important factor explaining indication for TJR. Other predictors included radiographic severity, stiffness (assessed by WOMAC) and OA disease duration (data not shown).

Despite the fact there was a difference in the level of symptoms between the two groups (candidate for surgery yes/no) the overlap between the two groups prevented us from proposing a specific cut-off. Indeed, among these OA patients referred to an orthopaedic surgeon, most patients were symptomatic. However, the surgeons often decided that surgery was warranted even among the less symptomatic patients (around 50% of the patients in the lower decile of symptoms were recommended for TJR, Figure 2 online), or that surgery was not warranted even if the symptoms were severe (only around 75% of these patients were considered surgery candidates). This indicates that the level of symptoms in this population was not the only driver for such a TJR indication ^30-32. Possibly, the surgeons paid greater attention to the radiographic severity than to the symptom levels ^{10, 27, 33} and several studies have indicated a discordance between radiographs and symptoms in lower-limb OA ^{27, 34-37}. In the present study however, stratifying the analyses on radiographic severity did not modify our conclusions. Finally, the present results indicated a stronger relationship between symptoms and surgical indication in hip OA than in knee OA. It is possible that the questionnaires used, the ICOAP and KOOS-PS/HOOS-PS ^17-21 (which were not seen by the orthopaedic surgeon) may assess different aspects of symptoms, than what the orthopaedic surgeon usually assesses in the clinic; however, it is reassuring to note that these new tools gave results very similar to the WOMAC sub-scales. Indeed, the sensitivity analyses performed using WOMAC data confirmed our main results. Perhaps also, other data related to patient-reported outcomes could be relevant in the indication for TJR, such as worsening of symptoms (e.g., minimal clinically important deterioration); however, we did not collect change in status in this study, but only status at one time point, and persistence of that status, since we felt that a decision for TJR would be more strongly based on status than on change. Clearly, in addition to symptomatic severity, many other factors are as strong or stronger determinants of surgery ^9,38-41. Furthermore, perhaps other aspects of symptomatic severity are taken into account in the surgeons' decision, e.g. the duration of symptoms (whereas questionnaires have a short time-frame), or the ongoing symptomatic treatment of the patient that may influence his/her current level of symptoms⁴. Finally, the surgeons may have based their surgical decision on joint mobility or peri-articular amyotrophy ³⁸, which were not assessed here. There were clear differences across centers and countries; these might be explained by several elements, including differences in the health care systems, or characteristics/training of the surgeons. In all, the current study confirms the wide variability in the indication for TJR suggested by studies of who actually receives a TJR; these results provide evidence that variability in surgeons' recommendations and practices are an important contributor to the clinical variability among TJR recipients.

This study has strengths and weaknesses. It is a large, international study which enhances the external validity of our results. On the other hand and as could be expected, there were differences across centers and countries in terms of symptomatic severity and in terms of the frequency of indication for TJR as assessed by the orthopaedic surgeon ^9,32. We do not believe this is an important limitation to the present results. Indeed, the objective here was to develop international criteria reflecting a level of OA symptoms and disability at which point TJR should be considered, for use as outcome measure in clinical trials. In this context, it was necessary to include patients from different backgrounds. In this study, one possible bias is that only symptomatic OA patients were included since the patients had to have definite OA to be included, and were in fact seeing an orthopaedic surgeon, generally to discuss a surgical indication for their target joint (we do not have information regarding if the patients were coming for the first time, or for return visits). Therefore the present study did not include many asymptomatic patients which may explain the low predictive power of symptomatic severity here. Indeed, symptom thresholds associated with TJR in a more heterogeneous sample (including asymptomatic patients) might be relevant for defining endpoints for observational studies. Nevertheless, the patients in the study presented with a wide range of symptomatic severity, and only about half of them were considered candidates for TJR. Several statistical techniques and sensitivity analyses were performed, to further confirm the internal validity of our results; and the study was not underpowered.

In this study, the gold standard was the surgeon's opinion regarding need for TJR ^{13, 28}. We considered that if surgery was recommended or if the surgeon considered symptoms were severe enough for surgery (although because of comorbidity ¹⁰ or patient refusal ⁹, the patient was not referred for surgery), a state of indication for TJR was attained. However, we did not collect data regarding actual carrying-out of surgery in these patients, which may differ widely^2-10.

In conclusion, this large study indicates that among patients referred to an orthopaedic surgeon to discuss TJR, the level of symptoms was higher among patients for whom TJR was indicated by the surgeon, but there was no cut point for pain and functional disability allowing to discriminate between patients with or without an indication for TJR.

Supplementary Material

Acknowledgments

Footnotes

Contributor Information

Laure Gossec, Paris Descartes University, Medicine Faculty; APHP, Rheumatology B Department, Cochin Hospital, Paris France.

Simon Paternotte, Paris Descartes University, Medicine Faculty; APHP, Rheumatology B Department, Cochin Hospital, Paris France.

Jean Francis Maillefert, Department of Rheumatology, Dijon University Hospital, Dijon, F-21078, France, University of Burgundy, Dijon, F-21079, France, INSERM U887, Dijon, F-21079, France.

Christophe Combescure, University of Geneva, Switzerland.

Philip G Conaghan, Section of Musculoskeletal Disease, University of Leeds & NIHR Leeds Musculoskeletal Biomedical Research Unit, Leeds, United Kingdom.

Aileen M. Davis, Division of Health Care and Outcomes Research and Arthritis and Community Research and Evaluation Unit, Toronto Western Research Institute, Toronto, Ontario and Departments of Rehabilitation Science and Health Policy, Management and Evaluation, University of Toronto.

Klaus-Peter Gunther, Department of Orthopaedic Surgery, University of Dresden, Germany.

Gillian Hawker, Division of Rheumatology, Department of Medicine, Women's College Hospital; Clinical Epidemiology and Health Care Research Program, Department of Health Policy, Management and Evaluation, Faculty of Medicine, University of Toronto, Canada.

Marc Hochberg, Division of Rheumatology & Clinical Immunology, Department of Medicine and Division of Gerontology, Department of Epidemiology and Preventive Medicine, University of Maryland School of Medicine, Baltimore, Maryland, USA.

Jeffrey N Katz, Department of Orthopedic Surgery and Division of Rheumatology, Immunology and Allergy, Brigham and Women's Hospital, Harvard Medical School; and Department of Epidemiology, Harvard School of Public Health, Boston, MA USA.

Margreet Kloppenburg, Department of Rheumatology, Department of Clinical Epidemiology, Leiden University Medical Center, Leiden, Netherlands.

Keith Lim, Department of Rheumatology, St Vincent's Hospital & Western Hospital, Melbourne, Australia.

L. Stefan Lohmander, Department of Orthopaedics, Clinical Sciences Lund, Lund University, Lund, Sweden.

Nizar N. Mahomed, Division of Orthopaedic Surgery, University of Toronto, Toronto Western Hospital, Toronto, Canada.

Lyn March, Institute of Bone and Joint Research, University of Sydney, Royal North Shore Hospital, St Leonards, Australia.

Karel Pavelka, Institute of Rheumatology, Charles University Prague, Czech Republic.

Leonardo Punzi, Rheumatology Unit, Department of Clinical and Experimental Medicine, University of Padova, Italy.

Ewa M. Roos, Institute of Sports and Clinical Biomechanics, University of Southern Denmark, Denmark.

Lydia Sanchez-Riera, Institute of Bone and Joint Research, University of Sydney, Australia.

Jasvinder A. Singh, Birmingham VA Medical Center and University of Alabama, Birmingham, AL, USA.

Maria E. Suarez-Almazor, Section of Rheumatology, University of Texas, M.D. Anderson Cancer Center, Houston, Texas USA.

Maxime Dougados, Paris Descartes University, Medicine Faculty; APHP, Rheumatology B Department, Cochin Hospital, Paris France.

References