Objective To compare the performance of published classification/diagnostic criteria for polymyalgia rheumatica (PMR), including the new 2012 European League Against Rheumatism (EULAR)/American College of Rheumatology (ACR) criteria, in a single-centre study.
Methods We studied all consecutive patients with new-onset PMR seen in our centre over 6 years, whose diagnosis was confirmed during a prospective 12-month follow-up period. Subjects were classified by each of the seven different criteria. Sensitivity and specificity were compared. Control population consisted of all consecutive patients aged ≥50 years seen in a 4-year period in our early arthritis clinic who had a 12-month confirmation of a diagnosis of rheumatoid arthritis (RA) or other inflammatory articular diseases.
Results Data were collected from 136 cases and 149 controls, including 94 patients with RA. The most sensitive criteria were the new 2012 EULAR/ACR classification criteria (92.6%). Adding ultrasound (US) specificity increased from 81.5% to 91.3% in total cases and from 79.7% to 89.9% in RA. Bird criteria had a sensitivity of 89.2% but the lowest specificity (40.2% in total cases and 72.5% in RA). Jones and Nobunaga criteria were the most specific criteria (96.7% and 97.8% in total cases and 98.6% and 99.5% in RA) but the less sensitive (63.1% and 58.2%) ones. Overall, discriminatory ability, as reflected by the area under the receiver operating characteristic curve, was better for the 2012 US EULAR/ACR criteria (0.920 in total cases and 0.910 in RA).
Conclusions The new EULAR/ACR criteria in new-onset PMR patients perform best in discriminating PMR from RA and other inflammatory articular diseases. Ultrasound further increases the specificity of the criteria.
- polymyalgia rheumatica
- classification criteria
Statistics from Altmetric.com
The diagnosis of polymyalgia rheumatica (PMR) rests on a combination of clinical symptoms, raised acute-phase reactants, exclusion of other conditions and response to glucocorticoids (GCs).1 At least five sets of diagnostic criteria for PMR have been formulated on the basis of clinical experience and used in clinical practice.2–6 In response to the lack of standardised classification criteria for PMR, in 2012 an international working group developed new European League Against Rheumatism (EULAR)/American College of Rheumatology (ACR) provisional classification criteria for PMR. Musculoskeletal ultrasound (US) examination was part of these new criteria.7
The aim of our study was to compare the performance of the numerous diagnostic/classification criteria sets available for PMR, including the recent 2012 EULAR/ACR classification criteria, in our cohort of patients with early arthritis prospectively followed up.
Patients and methods
All patients with suspected PMR seen at our centre from 2004 through 2009 were prospectively followed up according to a standardised protocol, which included clinical examination, determination of laboratory parameters (including erythrocyte sedimentation rate (ESR), C-reactive protein (CRP), rheumatoid factor (RF), anticitrullinated protein antibodies (ACPA) and antinuclear antibodies (ANA)), disability (assessed by the validated Italian version of the Health Assessment Questionnaire disability index)8 and the core set parameters considered in Leeb's disease activity score.9
US examination of shoulders and hips was performed at the first and subsequent visits according to a standardised protocol10 ,11 by the same rheumatologist (PM) expert in musculoskeletal US using the Esaote MyLab70 (Esaote, Genova, Italy) machine with a 13–5 MHz linear array transducer or a 7–3 MHz convex transducer. The patients were evaluated at disease onset and after 1, 3, 6 and 12 months. Additional visits were scheduled for patients with recurrences/relapses on an ‘as required’ basis.
PMR patients whose diagnosis was confirmed at 12-month follow-up by two of the authors (PM, CS) were included in this study. Patients with associated giant cell arteritis were excluded.
ACPA were determined in 69 (51%) of the 136 identified PMR patients, while RF was evaluated in all patients.
Non-PMR controls comprised patients aged 50 years or older consecutively seen at our early arthritis clinic (EAC) during a 4-year period (2006–2009). All these patients were followed up according to a standardised protocol that included US examination of the shoulder and/or hip in case of shoulder and/or hip pain. At the first visit all patients had a full clinical examination, determination of laboratory parameters (ESR, CRP, RF, ACPA and ANA), the core set of parameters considered in ACR evaluation12 and the duration of morning stiffness. In case of suspected rheumatoid arthritis (RA), radiographs of hands and feet were taken at disease onset and then yearly. The non-PMR control group consisted of all patients who had a 12-month confirmation of a diagnosis of a non-PMR inflammatory articular condition (RA ACR criteria),13 spondyloarthropathy (SpA) (ASAS criteria),14 connective tissue disease (expert opinion and/or ACR criteria when applicable) or disorders characterised by bilateral shoulder pain with or without increased ESR or CRP values.
The total control group was constituted by 94 patients with RA, 29 with SpA (16 had psoriatic arthritis, 8 undifferentiated spondylitis, 2 ankylosing spondylitis and 3 reactive arthritis), 6 with connective tissue diseases (3 Sjøgren syndrome 1 systemic lupus erythematosus and 2 undifferentiated connective tissue disease) and 20 with bilateral shoulder pain with or without elevated inflammatory markers (6 patients had crystal-related arthropathy, 9 mechanical shoulder involvement and 5 fibromyalgia).
All the enrolled patients were steroid-naïve.
The study was reviewed and approved by the local ethics committee (Comitato etico provinciale di Reggio Emilia). Written informed consent was retrospectively obtained from all participating patients.
Continuous variables were compared using t test or Mann–Whitney test when appropriate. Ordinal or dichotomous variables are presented as percentages and were compared by χ2 or Fisher's exact test.
Two-by-two classification tables were generated to estimate sensitivity and specificity.
To evaluate the capacity of the new criteria to discriminate between patients with and without PMR, receiver operating characteristic (ROC) curves with corresponding areas under the curve (AUC) were calculated.
The capacity of the different sets of PMR criteria to discriminate between PMR and non-PMR patients was compared.
Clinical and demographic data of the patients included in the study are reported in table 1.
PMR patients were significantly older, had higher ESR and CRP values at diagnosis and lower frequency of RF and ACPA positivity than control patients. Duration of morning stiffness was longer in PMR patients.
Three PMR patients were diagnosed as RA at the end of the follow-up period and were excluded from the analysis. None of the RA patients had an alternative diagnosis at follow-up.
Considering the individual items of the EULAR/ACR provisional criteria, any of them (including abnormal US findings) was significantly more frequent in PMR patients compared with both the entire controls’ group and the RA patients (table 2).
Table 3 shows sensitivity, specificity and AUC of the new EULAR/ACR provisional criteria and of the previous classification/diagnostic criteria. The most sensitive criteria were the new 2012 EULAR/ACR classification criteria (92.6%). Adding US increased specificity from 81.5% to 91.3% in total cases and from 79.7% to 89.9% in RA. Bird criteria had a sensitivity of 89.2% but the lowest specificity (40.2% in total cases and 72.5% in RA). Jones and Nobunaga criteria were the most specific criteria (96.7% and 97.8% in total cases and 98.6% and 99.5% in RA) but the less sensitive (63.1% and 58.2%). Healey and Chuang criteria had similar intermediate sensitivity (80.3% and 77%) and specificity (81.5% in total cases and 78.3% in RA).
Overall discriminatory ability, as reflected by the area under the ROC curve, was better for EULAR/ACR criteria. The lower limit of 95% CI of US-EULAR/ACR criteria was higher than the upper limits of 95% CI values observed in all other sets of criteria, except for the EULAR/ACR clinical criteria.
When tested against the RA control patients with bilateral shoulder involvement and elevated acute phase reactants (28 patients), EULAR/ACR criteria showed a reduced specificity (0.54 for clinical criteria and 0.68 for US criteria) as compared with the total RA controls (0.86 and 0.91, respectively).
When tested against the SpA control patient subgroup (29 patients), the specificity of clinical EULAR/ACR criteria was 95.7% and of EULAR/ACR US criteria 100%.
We tested the performance of the new 2012 EULAR/ACR provisional classification criteria for PMR comparing them with the former PMR diagnostic/classification criteria using patients from our EAC as controls.2–6 Our results show that the new criteria perform better than the older ones2–6 and in particular discriminate better between PMR and both total and RA patients.
The new EULAR/ACR PMR criteria7 performed better in our study compared with the original cohort used to develop them. Differences in the inclusion criteria of the controls may partially explain this discrepancy. We considered a series of unselected consecutive patients with early inflammatory articular disease, independently from the presence of shoulder pain, which reflects more closely the patients usually seen in clinical practice. In contrast, the presence of new-onset bilateral shoulder pain was required to be included in the control group in the original study.7 When the criteria were tested against RA patients with bilateral shoulder involvement and elevated acute phase reactants, we also found a lower specificity. Early differentiation between PMR and RA with PMR-like onset can be difficult because these conditions may have a similar clinical presentation.1
The first systematic comparison of the criteria available at that time to classify PMR was performed by Bird et al in 2005, who evaluated the sensitivity of four sets of criteria in 213 patients with new-onset PMR attending eight European rheumatological centres.15 The limitations of this study were the short follow-up period (2 weeks) and the absence of a control group of patients with conditions mimicking PMR, which precluded the determination of specificity. The results of this study, in agreement with our own, showed the best sensitivity for the Bird's criteria.15 However, we found that Bird's criteria15 were less specific than the other ones. To perform well, a set of criteria must have the best combination of sensitivity and specificity. In our study, among the old criteria, those that performed better were the Healey5 and Chuang et al4 criteria, which discriminated equally well between PMR and both total controls and RA patients. These two criteria are much the same, the only difference being the inclusion in the Healey criteria5 of the response to GCs, and they have been those most used in clinical trials to identify patients with PMR.
Imaging studies have documented in PMR the predominant involvement of proximal periarticular synovial structures.16–19 The new EULAR/ACR criteria for the first time included shoulder and hip girdle US examination as criteria for classifying PMR patients.7 In our study, US shoulder and hip examination increased the specificity of the new EULAR/ACR criteria not only in distinguishing between PMR and total controls but also in differentiating between RA and PMR patients. Adding US, the increase in the specificity of the new criteria was much higher in our study compared with the original one. A strength of our study is that US was performed by the same person (PL) who is an expert in musculoskeletal US. Musculoskeletal US is quite operator-dependent and requires considerable expertise; therefore a non-homogeneous US examination may have hampered the performance of the US criteria in the original study.
One limit of these new criteria is that patients can be classified as having PMR only if they have both elevated inflammation markers and shoulder girdle complaints. Different epidemiological studies have demonstrated that a percentage of PMR patients ranging from 7% to 20% have normal acute-phase reactants at diagnosis.1 ,20 However, differently from most of the previous criteria that included only ESR as acute-phase reactant, the new criteria also include the CRP, which is more sensitive than the ESR in active PMR in detecting inflammation.20 This permits to increase the number of patients with low-grade inflammation that can be identified by the new criteria.
The limitation of our study is the relatively short duration of follow-up period; however, we are confident that a 12-month longitudinal follow-up allowed a correct diagnosis in most of the patients and controls. A prospective, multicentric study recruiting a larger number of controls with arthritis and other conditions mimicking PMR is required to confirm our findings
We thank Nicolò Pipitone for the critical review of the manuscript.
Handling editor Tore K Kvien
Contributors All authors contributed to conception and design, or analysis and interpretation of data drafting the article or revising it critically for important intellectual content and final approval of the version to be published.
Competing interests None.
Patient consent Obtained.
Ethics approval Comitato etico provincia di reggio emilia.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.