DOI 10.1002/art.34536 2012, American College of Rheumatology Sensitivity and Specificity of the Classification of Psoriatic Arthritis Criteria in Early Psoriatic Arthritis Laura C. Coates, 1 Philip G. Conaghan, 1 Paul Emery, 1 Michael J. Green, 2 Gamal Ibrahim, 3 Helen MacIver, 3 and Philip S. Helliwell 1 Objective. To assess the sensitivity and specificity of the Classification of Psoriatic Arthritis (CASPAR) Study Group criteria in early psoriatic arthritis (PsA) and to compare them with the sensitivity and specificity of the Moll and Wright criteria. Methods. The CASPAR Study Group criteria were applied to patients with early PsA (<24 months symp- tom duration) and to control patients with other new- onset inflammatory arthritides. Both groups were naive to all disease-modifying antirheumatic drugs. The gold standard diagnosis was confirmed by the consulting rheumatologist using radiography and magnetic reso- nance imaging where required. Proportions of patients and control patients meeting the criteria were compared using McNemars tests. Results. We recruited a total of 111 patients with early PsA and 111 control patients with other forms of inflammatory arthritis (82 with rheumatoid arthritis, 13 with undifferentiated arthritis, 9 with spondylarthritis, 4 with inflammatory osteoarthritis, and 3 with crystal arthritis) to the study. The sensitivity of the CASPAR Study Group criteria in classifying early PsA was 87.4% compared to 80.2% for the Moll and Wright criteria. The specificity for both criteria was 99.1%. When con- sidering different cut points for the CASPAR Study Group criteria, the best cut point for classification remained a score of >3 as in the original CASPAR Study Group analysis. Considering a score of >2 gave a higher sensitivity of 99.1% but resulted in a drop in specificity to 94.6%. Regression analysis determined that psoriasis and rheumatoid factor negativity were the most important features that differentiated PsA, followed by nail psoriasis and current or previous dactylitis. Conclusion. The CASPAR Study Group criteria are more sensitive than the Moll and Wright criteria in classifying early PsA. Although their sensitivity for early PsA is lower than that for established disease, the CASPAR Study Group criteria are valid for use as inclusion criteria for trials in early PsA. The Classification of Psoriatic Arthritis (CASPAR) Study Group criteria were developed as new classifica- tion criteria for psoriatic arthritis (PsA) using a large cohort of patients with PsA and controls with other inflammatory arthritides. They include characteristic features of PsA and have been shown to have high sensitivity and specificity in a population with estab- lished disease (1). However, one of the limitations of the CASPAR Study Group criteria is that they have not been validated in early disease. A retrospective survey of the Toronto database found that the CASPAR Study Group criteria had high sensitivity for identifying early disease within their PsA clinic population (2). However, that study included only patients referred to a specialist tertiary referral clinic; therefore, there is a significant risk of overestimating the sensitivity of the criteria. Studies in Italy and Sweden have also investigated the Dr. Coates was recipient of a Clinical Research Fellowship from Arthritis Research UK (grant number 18364). 1 Laura C. Coates, MBChB, PhD, MRCP, Philip G. Conaghan, MBBS, PhD, FRCP, FRACP, Paul Emery, MA, MD, FRCP, Philip S. Helliwell, MA, DM, PhD: University of Leeds and NIHR Leeds Musculoskeletal Biomedical Research Unit, Leeds, UK; 2 Michael J. Green, MBChB, MRCP: York Teaching Hospital NHS Foundation Trust, York, and Harrogate and District NHS Foundation Trust, Harrowgate, UK; 3 Gamal Ibrahim, ABMS, MSc, FRCP, Helen MacIver, MBChB, MRCP: St. Lukes Hospital and Bradford Teaching Hospital NHS Foundation Trust, Bradford, UK. Dr. Emery has received consulting fees, speaking fees, and/or honoraria from Pfizer, MSD, Abbott, Roche, UCB, Celgene, and Bristol-Myers Squibb (less than $10,000 each). Address correspondence to Philip S. Helliwell, DM, PhD, Division of Musculoskeletal and Rheumatic Diseases, Chapel Allerton Hospital, Chapeltown Road, Leeds LS7 4SA, UK. E-mail: p.helliwell@leeds.ac.uk. Submitted for publication July 15, 2011; accepted in revised form May 3, 2012. 3150 sensitivity of the criteria in early PsA (3,4), but as yet no studies have been completed with a control population to address the sensitivity and specificity of the CASPAR Study Group criteria in early disease. The criteria cannot currently be recommended for use in studies of early PsA. The other key issue raised with the CASPAR Study Group criteria is the entry criterion that states that the patients must have inflammatory articular disease (joint, spine, or entheseal). Particularly in early PsA, the pattern of disease may be evolving and may be different from that of established disease. This study aimed to emulate the methodology of the original CASPAR study and to recruit both patients with early PsA and control patients with other inflam- matory arthritides to allow calculation of the sensitivity and specificity of the criteria in this population of patients with early PsA. The Moll and Wright criteria, which state that patients have PsA if they have inflam- matory arthritis, psoriasis, and (usually) are negative for rheumatoid factor (RF) (5), were used for compar- ison. In addition, an assessment of arthritis, enthesitis, and axial involvement was included to investigate the pattern of disease in treatment-naive patients with early arthritis. PATIENTS AND METHODS Patient and control selection. Patients and controls were recruited from Early Arthritis Clinics and from new referrals to rheumatology clinics in 4 hospitals based in the Yorkshire region. Consecutive clinic attendees newly diag- nosed as having PsA were enrolled in the study. All PsA patients were required to have active inflammatory musculo- skeletal disease comprising arthritis and/or enthesitis and/or dactylitis. For each patient enrolled, a consecutive control patient with new-onset inflammatory musculoskeletal disease was also enrolled. The gold standard diagnosis was confirmed by the consulting rheumatologist using radiography and mag- netic resonance imaging (MRI) where required. All patients and controls were required to have disease duration of 2 years (from onset of symptoms) and to be disease-modifying antirheumatic drug naive. Data collection. Data were recorded on standardized forms that included demographics, date of onset of arthritis, duration of early morning stiffness, activity of arthritis/ enthesitis and axial disease, and application of the CASPAR Study Group criteria. Activity of arthritis and enthesitis was assessed using clinical examination. All patients underwent a swollen joint count (66 joints), a tender joint count (68 joints), and clinical assessment of entheses, including the entheseal sites required for the Leeds Enthesitis Index (6), the Maas- tricht Ankylosing Spondylitis Enthesitis Score (MASES) (7), and sites used in the Infliximab Multinational Psoriatic Arthri- tis Controlled Trial (IMPACT) and IMPACT 2 clinical trials (bilateral Achilles tendon and plantar fascia) (8,9). Entheses were considered to have active enthesitis if they were tender and/or swollen. The presence of axial involvement was assessed clinically by questioning the patient about inflammatory back pain using the Calin criteria (10). All patients underwent plain radiography of the hands and feet and additional imaging as deemed clinically necessary by the treating physician. Patients with symptoms suggestive of inflammatory back pain underwent radiography and MRI as clinically indicated. Results of all relevant investigations were recorded, including levels of inflammatory markers, IgM-RF and anticyclic citrullinated peptide antibodies, presence or absence of typical PsA-like periosteal new bone formation on radiographs, and evidence of axial involvement identified with radiographs or MRI by consensus of radiologists and the referring rheumatologist in each center (11). When considering patterns of arthritis, oligoarthritis was defined as 5 joints with active disease and polyarthritis as 5 joints with active disease. Small joints were defined as temporomandibular joints, sternoclavicular joints, acromiocla- vicular joints, carpometacarpal joints, metacarpal joints, prox- imal and distal interphalangeal (DIP) joints, metatarsophalan- geal joints, and interphalangeal joints of the toes (total of 58 joints). Large joints were defined as the glenohumeral joints, elbows, hips, knees, and ankle joints (total of 10 joints). The symmetry number was calculated as the number of joints as symmetric pairs/total number of joints involved. Symmetry was defined as a symmetry number of 0.5 as previously described (12). Statistical analysis. Values for sensitivity and specific- ity for the CASPAR Study Group criteria in established disease are 0.91 and 0.99, respectively. Prior to the develop- ment of the CASPAR Study Group criteria, the Moll and Wright criteria were traditionally used, and the values for sensitivity and specificity for the Moll and Wright criteria in the CASPAR study were 0.99 and 0.92, respectively. Assuming that in early case definition the emphasis is on specificity, the null hypothesis is that the specificity for the CASPAR Study Group criteria will be equivalent to that for the Moll and Wright criteria (H A H O , respectively). The alternative hypothesis is that the specificity for the CASPAR Study Group criteria differs from that for the Moll and Wright criteria (H A H O ). According to Obuchowski (13), the number of patients required to test the alternative (2-tailed) hypothesis with an alpha value of 0.05 and a beta value of 0.20 is 111. Therefore, 111 patients and 111 consecutive control patients were required. Sensitivity and specificity of the CASPAR Study Group criteria and Moll and Wright criteria were compared using McNemars tests for paired variables. Receiver operating characteristic (ROC) curves were used to estimate the area under the curve (AUC) for both criteria and the optimal cut point for the CASPAR Study Group criteria for the diagnosis of early PsA. Univariate and multivariate forward stepwise binary logistic regression was used to identify which features included in the CASPAR Study Group criteria were indepen- dently associated with PsA. Significance testing was performed USE OF CASPAR CRITERIA IN EARLY PsA 3151 using McNemars tests for categorical variables and Mann- Whitney U tests for continuous variables. RESULTS We recruited a total of 111 patients with early PsA and 111 control patients with other forms of inflammatory arthritis (82 with rheumatoid arthritis [RA], 13 with undifferentiated arthritis, 9 with spondy- larthritis, 4 with inflammatory osteoarthritis, and 3 with crystal arthritis) to the study. The baseline features of these patients are shown in Table 1. The sensitivity of the CASPAR Study Group criteria (score 3) in classi- fying early PsA was 87.4% compared to 80.2% for the Moll and Wright criteria (P 0.008). The specificity for both criteria was 99.1%, with only 1 control patient fulfilling both criteria for PsA. Using ROC analysis, the AUC for the CASPAR Study Group criteria was 0.991 compared to 0.896 for the Moll and Wright criteria (Figure 1). When consid- ering different cut points for the CASPAR Study Group criteria, the best cut point for classification remained a score of 3 as in the original CASPAR Study Group analysis (Table 2). Considering a score of 2 gave a higher sensitivity but resulted in a drop in specificity that would not be ideal for classification criteria. When considering the individual components of the CASPAR Study Group criteria, 96.4% of PsA patients had current or previous psoriasis or a family history of psoriasis with 84% having current psoriasis. Dactylitis and nail psoriasis were highly discriminatory, as only 1 control patient each had these features. Univariate binary logistic regression analysis determined that all of the features of the CASPAR Study Group criteria, with the exception of radiographic evidence of new bone formation, were significant in distinguishing Table 1. Baseline features of the PsA patients and control patients recruited to the CASPAR study* PsA patients (n 111) Control patients (n 111) Age, median (IQR) years 44 (3455) 52 (3866) Disease duration, median (IQR) months 8 (514) 5 (39.25) Consultant diagnosis, no. of patients PsA 111 0 RA 0 82 UA 0 13 SpA 0 9 ReA 0 6 IBD related 0 2 Axial SpA 0 1 Inflammatory OA 0 4 Crystal arthritis 0 3 Arthritis, % 100 99 Enthesitis, % 67 52 Axial symptoms, % 5 1 Early morning stiffness, median (IQR) minutes 60 (20120) 60 (30120) Fulfilling Moll and Wright criteria for PsA, % 80 1 Fulfilling CASPAR Study Group criteria for PsA, % 87 1 * PsA psoriatic arthritis; CASPAR Classification of Psoriatic Ar- thritis; IQR interquartile range; RA rheumatoid arthritis; UA undifferentiated arthritis; SpA spondylarthritis; ReA reac- tive arthritis; IBD inflammatory bowel disease; OA osteoarthri- tis. P 0.05 versus control patients. Figure 1. Receiver operating characteristic curve for the Classifica- tion of Psoriatic Arthritis (CASPAR) Study Group criteria and the Moll and Wright criteria predicting psoriatic arthritis diagnosis. Table 2. Receiver operating characteristic curve analysis of the Moll and Wright criteria and the CASPAR Study Group criteria* Criteria Score Sensitivity Specificity Moll and Wright Positive 80.2 99.1 CASPAR Study Group criteria score 1 100 42.3 2 99.1 94.6 3 87.4 99.1 4 42.3 100 5 15.3 100 * CASPAR Classification of Psoriatic Arthritis. 3152 COATES ET AL PsA patients from control patients. Multivariate forward stepwise logistic regression determined that current or previous psoriasis, RF negativity, nail psoriasis, family history of psoriasis, and current or previous dactylitis were all associated with PsA. Table 3 shows in more detail the proportion of patients fulfilling each individ- ual aspect of the CASPAR Study Group criteria. When considering the diagnosis of inflammatory articular disease (joint, spine, or entheseal) required for the first aspect of the CASPAR Study Group criteria, there were some minor differences between the PsA patients and control patients. Nearly all patients had arthritis. Enthesitis was significantly more frequent in PsA patients than in control patients (67% versus 52%; P 0.029). There was a trend toward higher enthesitis counts in PsA patients compared to control patients, but this was not significant. When considering the 3 clinical indices for enthesitis (MASES, Leeds Enthesitis Index, and IMPACT) individually, there was again a trend toward higher enthesitis scores in PsA patients com- pared to control patients, but this was only significant when using the IMPACT tool (P 0.022). There were no individual entheseal sites or other combinations of entheseal sites that could distinguish PsA patients from control patients. Spondylitis appeared more frequently in PsA patients, but the difference was not significant due to small numbers (5% versus 1%; P 0.055). Further analysis examined the pattern of arthritis seen in PsA patients and patients with other inflamma- tory arthritides (Table 4). The average tender and swollen joint counts and the proportions of people with oligoarticular and polyarticular disease did not differ significantly between PsA patients and control patients. Involvement of large and small joints was also similar when comparing PsA patients to all control patients. The 82 control patients with RA were found to have significantly higher joint counts (P 0.008), principally due to an increased number of small joints involved. More than 60% of RA patients had predominantly small joint disease. However, DIP joint involvement was sig- nificantly more common in PsA patients than in all control patients (32% versus 16%; P 0.007) and more Table 3. Proportion of PsA patients and control patients fulfilling each aspect of the CASPAR Study Group criteria* PsA patients (n 111) Control patients (n 111) Any psoriasis 96 12 Current psoriasis 84 4 Previous psoriasis 6 1 Family history of psoriasis 18 8 Nail psoriasis 38 1 Rheumatoid factor negative 96 47 Current or previous dactylitis 28 1 Radiographic evidence of new bone formation 2 0 * Values are the percent of patients. See Table 1 for definitions. P 0.05 versus control patients. Table 4. Disease patterns in PsA patients, all patient controls, and RA patient controls* PsA patients (n 111) All patient controls (n 111) RA patient controls (n 82) Proportion with arthritis, % 100 99 100 Oligoarthritis, % 30 24 15 Polyarthritis, % 70 76 85 DIP joint involvement, no. (%) 35 (32) 18 (16) 15 (18) Tender joint count, median (IQR) 7 (316) 8 (419) 12 (622) Swollen joint count, median (IQR) 5 (29) 5 (312) 7 (314) Joints with active disease, median (IQR) 9 (417) 10 (520) 12.5 (723) Predominantly small joint arthritis, no. (%) 59 (54) 59 (54) 50 (62) Predominantly large joint arthritis, no. (%) 51 (46) 50 (45) 31 (38) Small joints involved, median (% of total) 7 (12) 9 (16) 11.5 (20) Large joints involved, median (% of total) 1 (10) 1 (10) 1.5 (15) Symmetry number, median (IQR) 0.5 (00.78) 0.67 (0.290.86) 0.73 (0.490.89) Symmetric disease, no. (%) 58 (52) 72 (65) 62 (76) Proportion with enthesitis, % 67 52 52 Enthesitis count, median (IQR) 2 (04) 1 (03) 1 (04) Proportion with spondylitis, % 5 1 0 * DIP distal interphalangeal (see Table 1 for other definitions). P 0.05 versus all control patients. P 0.05 versus RA control patients. USE OF CASPAR CRITERIA IN EARLY PsA 3153 common in PsA patients than in RA patients alone (18%). It has been suggested that a symmetry number of 0.5 defines symmetric disease (12). The symmetry number was significantly higher in all control patients than in PsA patients (P 0.003), and this difference was even more pronounced when comparing RA control patients with PsA patients (P 0.0001). The remaining control patients (without RA) had a lower median symmetry number, but this was not significant when compared to PsA patients (0.27 versus 0.5; P 0.58). A higher proportion of control patients, particularly those with RA, had symmetric disease compared to PsA patients. When we compared PsA patients to all control patients, there was no significant difference in the proportion with symmetric disease (P 0.11); however, there was a significant difference when we compared PsA patients to RA control patients (P 0.001). DISCUSSION These data confirm that the CASPAR Study Group criteria have greater sensitivity than the Moll and Wright criteria in the identification of early PsA. The most important difference between the 2 criteria is that the CASPAR Study Group criteria allow a diagnosis of PsA to be made in patients without skin psoriasis. The CASPAR Study Group criteria use additional features such as dactylitis, nail psoriasis, and positive family history to increase their sensitivity. This is particularly relevant in early disease, when not all features of PsA may have become manifest. In particular, 20% of people presenting with PsA do not have skin psoriasis, thereby excluding them from a diagnosis using the Moll and Wright criteria. All RF-negative patients who fulfill the Moll and Wright criteria will also fulfill the CASPAR Study Group criteria, so there is significant overlap between the 2 sets of criteria. In comparison with their sensitivity for detecting established disease, the sensitivity of the CASPAR criteria remains slightly lower for detecting early PsA. A short duration of disease means that many patients will not yet have experienced all of the typical features of PsA that they may develop during the course of their disease. In particular, there is a much lower incidence of typical radiographic changes that can be identified in patients with early PsA. The vast majority of patients with new onset of inflammatory arthritis have normal- appearing radiographs at presentation. This problem has been identified in previous cohorts of patients with early PsA, in whom little new bone formation could be identified (3). When considering the pattern of inflammatory articular disease, a higher proportion of PsA patients had enthesitis compared to control patients. Most of the clinical indices and clinical examinations of individual enthesis sites were not able to discriminate between PsA patients and control patients. It is known that clinical assessment for tenderness near an enthesis correlates poorly with ultrasound or MRI in identifying enthesitis. Therefore, this may reflect a failure of the clinical entheseal indices rather than a similar picture of enthesi- tis in the PsA patients and control patients. More PsA patients had axial disease, although this difference was not significant due to small numbers affected in both groups. Patterns of arthritis showed that PsA patients had a lower total number of involved joints compared to RA control patients, and this difference was principally explained by a higher number of involved small joints in RA patients, confirming the identification of RA as a predominantly small joint polyarthropathy. Symmetric joint disease was less common in PsA patients than in RA control patients, and there was a significant differ- ence in symmetry number between PsA patients and all control patients. The most significant limitation of this study re- lates to the selection of the PsA patients. PsA patients were recruited only if they were diagnosed by an expe- rienced rheumatologist as having PsA, and few patients presenting with undifferentiated arthritis were included in this study. This may be less important when consid- ering the use of the CASPAR Study Group criteria as classification criteria for enrollment in clinical trials, as originally intended. The results of this study support their use as inclusion criteria for future research studies of recent-onset PsA. However, these data will result in an overestimate of the sensitivity of the criteria if their use is considered as a diagnostic or screening tool in a general Early Arthritis Clinic population. Nearly all of the patients and controls recruited into this study had active arthritis, although this was not a mandatory inclusion criterion. Patients presenting with isolated enthesitis or dactylitis are less common and may have a delay in diagnosis that would have excluded them from participation in this study. The other limitation is the identification of ideal controls. Again, controls are more likely to have a certain diagnosis, such as RA, and patients with undifferentiated arthritis were not seen in significant numbers. It is also difficult to identify a sig- 3154 COATES ET AL nificant number of controls presenting with axial spon- dylarthritis who have 2 years of symptom duration. This work supports the use of the CASPAR Study Group criteria as classification criteria for future research in early PsA. Although the sensitivity of these criteria is slightly lower than that in established disease, the specificity is reassuring. Their use as diagnostic criteria in cases where diagnosis is unclear cannot be recommended based on this study, but future work evaluating their use in an unselected Early Arthritis Clinic population would address this issue. A different clinical phenotype cannot be relied upon in isolation to identify early PsA, but different patterns are seen in PsA patients compared to patients with other inflammatory arthritides. Enthesitis and DIP joint involvement are more frequent in early PsA compared to other forms of inflammatory arthritis; however, a significant proportion of controls (including a significant proportion of patients with RA) also have clinical enthesitis. Spondylitis is likely to be more frequent in PsA than in RA; however, it cannot be relied upon to differentiate PsA from other forms of inflammatory arthritis. Patterns of arthritis confirm the typical view that RA is a symmetric predom- inantly small joint polyarthropathy, whereas patients with PsA generally present with fewer involved joints, particularly fewer involved small joints, and with less symmetric disease. AUTHOR CONTRIBUTIONS All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Helliwell had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study conception and design. Coates, Helliwell. Acquisition of data. Coates, Conaghan, Green, Ibrahim, MacIver, Helliwell. Analysis and interpretation of data. Coates, Conaghan, Emery, Helliwell. REFERENCES 1. Taylor W, Gladman D, Helliwell P, Marchesoni A, Mease P, Mielants H, and the CASPAR Study Group. Classification criteria for psoriatic arthritis: development of new criteria from a large international study. Arthritis Rheum 2006;54:266573. 2. Chandran V, Schentag CT, Gladman DD. Sensitivity of the classification of psoriatic arthritis criteria in early psoriatic arthri- tis. Arthritis Rheum 2007;57:15603. 3. DAngelo S, Mennillo GA, Cutro MS, Leccese P, Nigro A, Padula A, et al. Sensitivity of the classification of psoriatic arthritis criteria in early psoriatic arthritis. J Rheumatol 2009;36:36870. 4. Lindqvist UR, Alenius GM, Husmark T, Theander E, Holmstrom G, Larsson PT. The Swedish early psoriatic arthritis register 2-year followup: a comparison with early rheumatoid arthritis. J Rheumatol 2008;35:66873. 5. Moll JM, Wright V. Psoriatic arthritis. Semin Arthritis Rheum 1973;3:5578. 6. Healy PJ, Helliwell PS. Measuring clinical enthesitis in psoriatic arthritis: assessment of existing measures and development of an instrument specific to psoriatic arthritis. Arthritis Rheum 2008;59: 68691. 7. Heuft-Dorenbosch L, Spoorenberg A, van Tubergen A, Landewe R, van ver Tempel H, Mielants H, et al. Assessment of enthesitis in ankylosing spondylitis. Ann Rheum Dis 2003;62:12732. 8. Antoni C, Krueger GG, de Vlam K, Birbara C, Beutler A, Guzzo C, et al. Infliximab improves signs and symptoms of psoriatic arthritis: results of the IMPACT 2 trial. Ann Rheum Dis 2005;64: 11507. 9. Antoni CE, Kavanaugh A, Kirkham B, Tutuncu Z, Burmester GR, Schneider U, et al. Sustained benefits of infliximab therapy for dermatologic and articular manifestations of psoriatic arthritis: results from the Infliximab Multinational Psoriatic Arthritis Con- trolled Trial (IMPACT) [published erratum appears in Arthritis Rheum 2005;52:2951]. Arthritis Rheum 2005;52:122736. 10. Calin A, Porta J, Fries JF, Schurman DJ. Clinical history as a screening test for ankylosing spondylitis. JAMA 1977;237:26134. 11. Rudwaleit M, Jurik AG, Hermann KG, Landewe R, van der Heijde D, Baraliakos X, et al. Defining active sacroiliitis on magnetic resonance imaging (MRI) for classification of axial spondyloarthritis: a consensual approach by the ASAS/OMER- ACT MRI group. Ann Rheum Dis 2009;68:15207. 12. Helliwell PS, Hetthen J, Sokoll K, Green M, Marchesoni A, Lubrano E, et al. Joint symmetry in early and late rheumatoid and psoriatic arthritis: comparison with a mathematical model. Arthri- tis Rheum 2000;43:86571. 13. Obuchowski NA. Sample size calculations in studies of test accuracy. Stat Methods Med Res 1998;7:37192. USE OF CASPAR CRITERIA IN EARLY PsA 3155