Performance of the polymyalgia rheumatica activity score for diagnosing disease flares

Authors


Abstract

Objective

To evaluate the effectiveness of the polymyalgia rheumatica activity score (PMR-AS) in diagnosing disease flares.

Methods

Rheumatologists prospectively included 89 patients with PMR (mean ± SD age 74.6 ± 6.2 years, mean ± SD disease duration 1.6 ± 2.2 years). At each visit, the rheumatologist assessed disease activity using a visual analog scale (VAS) and recorded whether a disease flare was diagnosed and/or the glucocorticoid dose changed. Overall, 137 visits including 49 pairs (allowing intraindividual comparisons) were available; a disease flare was diagnosed at 32 visits. We evaluated statistical associations linking flare diagnosis to the PMR-AS, each of its components (VAS, VAS for pain, C-reactive protein, morning stiffness, and elevation of upper limbs), and changes in these parameters between 2 visits.

Results

Associations with disease flare diagnosis were strongest for PMR-AS scores ≥9.35 (agreement 92%, 95% confidence interval [95% CI] 85.8–95.7%, κ = 0.78; sensitivity 96.6%, 95% CI 80.4–99.8; specificity 90.7%, 95% CI 83.2–95.2) and for ΔPMR-AS scores ≥6.6 (agreement 98%, 95% CI 88.0–99.9%, κ = 0.95; sensitivity 100%, 95% CI 74.7–100; specificity 97.1%, 95% CI 82.9–99.8). Other parameters showed weaker diagnostic performance.

Conclusion

This study supplies new evidence that the PMR-AS is useful for monitoring PMR activity in everyday practice and for managing glucocorticoid tapering. PMR activity changes seem even more relevant than absolute values.

INTRODUCTION

Polymyalgia rheumatica (PMR) is a clinical syndrome of the elderly characterized by pain and morning stiffness in the neck and limb girdles, usually with elevations in erythrocyte sedimentation rate (ESR) and serum C-reactive protein (CRP) level. Systemic glucocorticoid therapy is the treatment of choice. Low dosages (10–20 mg/day of prednisone) are sufficient to correct the clinical symptoms and laboratory evidence of inflammation within a few days. The glucocorticoid dose is then tapered according to disease activity. Monitoring of disease activity and detection of flares usually relies on a combination of clinical symptoms with ESR and/or CRP (1). However, the definition of PMR flare is not universally agreed on, and no criteria for PMR relapse have been published. The risk of disease flare is difficult to assess, which may lead physicians to delay glucocorticoid tapering. Delayed tapering unnecessarily prolongs exposure to the adverse effects of glucocorticoids, an obvious concern in elderly individuals. Better tailoring of the glucocorticoid dose to disease activity would substantially improve the treatment of patients with PMR. To achieve this goal, an effective tool for evaluating disease activity is needed.

The European Collaborating Polymyalgia Rheumatica Group developed European League Against Rheumatism (EULAR) response criteria for PMR in 2003 (2). Leeb and Bird (3) used these criteria to create a PMR activity score (PMR-AS) based on 5 variables: morning stiffness (in minutes), ability to elevate the upper limbs (on a scale from 0 to 3), physician's global assessment on a 10-point visual analog scale (VASph), pain severity on a 10-point VAS (VASp), and CRP level (in mg/dl). PMR-AS values <7 indicated low disease activity, values between 7 and 17 indicated moderate disease activity, and values >17 indicated high disease activity. However, a PMR-AS threshold highly predictive of a need for an increased glucocorticoid dose remained to be determined.

In a previous study (4), we used clinical vignettes to evaluate the effectiveness of the PMR-AS in diagnosing flares that required glucocorticoid dose escalation. PMR-AS values ≥7 defined a flare with 98.1% sensitivity and 94.3% specificity. However, a PMR-AS increase >4.2 between 2 visits was better than a PMR-AS score ≥7 in diagnosing flares (4).

The next step was to evaluate the ability of the PMR-AS to diagnose disease flares in a sample of actual patients. Accordingly, the main objective of the current study was to determine the PMR-AS cutoff associated with a high probability of disease flare in a prospective cohort of patients with PMR. The secondary objectives were to evaluate the performance of each of the 5 components of the PMR-AS (morning stiffness, ability to elevate the upper limbs, VASph, VASp, and CRP level) in diagnosing flares and to determine whether diagnosing a flare was more strongly associated with the magnitude of changes in the PMR-AS or its components between 2 visits than with absolute values above the cutoff.

PATIENTS AND METHODS

Study rheumatologists.

The patients were included in the study by 26 rheumatologists who were members of, or participants in, the Inflammatory Joint Disease Working Group of the French Society for Rheumatology (Club Rhumatismes et Inflammation). They recruited consecutive patients meeting study criteria.

Study criteria.

Patients were eligible if they had PMR treated by glucocorticoids. Because no specific diagnostic criteria are available for PMR, we defined PMR for this study as a diagnosis of the disease by a rheumatologist in the absence of other diseases that might mimic PMR.

Study questionnaire.

At each visit, the rheumatologist recorded the following, in addition to age and date of PMR diagnosis: VASp, VASph, morning stiffness, ability to elevate the upper limbs, ESR, and CRP level (the PMR-AS components); whether the patient had synovitis, limb girdle pain, fever, nocturnal awakenings due to PMR, and/or weight loss; whether a flare was diagnosed; and whether the glucocorticoid dose was changed. By definition, a flare is a change from inactive to active disease. For the purposes of our study, a flare was diagnosed if the patient experienced either the first bout of active disease (transition from health to symptomatic disease) or a transition from controlled to uncontrolled disease while receiving glucocorticoid therapy, in the opinion of the rheumatologist. Because no universally accepted criteria are available for diagnosing PMR flares, we asked the study rheumatologists to evaluate their level of confidence, or level of certitude, in their flare diagnosis on a 10-point VAS (VAS confidence).

Statistical analysis.

Results for quantitative variables are reported as the mean ± SD and those for qualitative variables as numbers of positive responses per category (percentages). We computed PMR-AS as indicated by Leeb and Bird (3): PMR − AS = CRP (mg/dl) + VASp (scale 0–10) + VASph (scale 0–10) + (morning stiffness [minutes] × 0.1) + elevation of upper limbs (scale 0–3). The 4 levels of the semiquantitative elevation-of-upper-limbs scale are as follows: 3 = no upper limb elevation, 2 = elevation below the shoulder girdle (90°), 1 = elevation up to the shoulder girdle (90°), and 0 = elevation above the shoulder girdle (>90°). Patient groups were compared using the chi-square test for qualitative variables and the Mann-Whitney U test for quantitative variables. P values less than 0.05 were considered statistically significant.

First, we studied associations linking flare diagnosis to the total PMR-AS score and to each of its 5 components (morning stiffness, ability to elevate the upper limbs, VASph, VASp, and CRP level). Then, we evaluated associations linking glucocorticoid dose escalation to each of these 6 variables. Last, we evaluated associations between flare diagnosis and the magnitude of the change (Δ) in each of the 6 variables between the last 2 physician visits (ΔPMR-AS, Δ[morning stiffness], Δ[ability to elevate the upper limbs], ΔVASph, ΔVASp, and ΔCRP) in the subgroup of patients with at least 2 visits. Sensitivities and specificities were calculated for different cutoff values. Receiver operating characteristic (ROC) curves were plotted. An optimal cutoff for a variable was defined as the value nearest the upper-left point of the ROC curve. When 2 similar cutoff values were found, the one with the best sensitivity was chosen. We also evaluated associations between flare diagnosis and glucocorticoid dose escalation, and we computed Cohen's kappa to assess agreement between flare diagnosis by the rheumatologist and PMR-AS values above the cutoff. Statistical tests were performed using SPSS software, version 13.0 (SPSS, Chicago, IL).

Sample size calculation showed that 62 visits were needed to obtain at least 10% precision in determining agreement between flare diagnosis and either PMR-AS or ΔPMR-AS, with the alpha risk set at 5%.

RESULTS

Study population.

The study rheumatologists prospectively included 89 patients with a mean ± SD age of 74.6 ± 6.2 years and mean disease duration of 1.6 ± 2.2 years (range 0–13 years). These 89 patients had 149 visits, with 32 patients each having at least 2 visits and 57 having a single visit. ESR was used instead of CRP level at 12 visits, which were excluded from the evaluation of the link between PMR-AS or ΔPMR-AS and flare diagnosis or glucocorticoid dose escalation. All PMR-AS parameters were available for 82 patients at their first visit, with 23 flare diagnoses, which were first flares (at the diagnosis) in 8 patients and repeat flares (relapse) in 15; there were 137 visits overall, with 32 flare diagnoses including 8 first flares and 24 repeat flares, and 49 visit pairs (allowing comparisons of 2 visits in the same patients), including 21 with flares. We conducted statistical analyses in each of these 3 populations, because we used statistical tests designed for independent observations, although multiple observations in the same patients were included in the analysis of the overall 137 visits. In the 32 patients with at least 2 visits, the mean ± SD number of visits per patient was 2.87 ± 1.50 (range 2–9) and the mean time between 2 visits was 2.49 ± 2.18 months (range 0.20–11.64); this long period between visits supports independence (or at least limited dependence) of multiple observations in the same patients. The mean ± SD score for physician confidence in flare diagnosis (VAS confidence) was 7.1 ± 2.2.

Comparison between groups with and without flare diagnosis.

The mean values of VASp, VASph, morning stiffness, CRP level, and PMR-AS at the 89 first visits (23 flare diagnoses) and the 149 overall visits (32 flare diagnoses) are presented in Table 1. As expected, all parameters reflecting PMR activity were significantly higher (P < 0.0001) in the group with flare diagnoses compared with the group without flares. At the first visit, the mean ± SD PMR-AS was 25.3 ± 8.6 (range 12.2–52.4) in the flare group and 5.0 ± 4.5 (range 0–23.9) in the group without flares. Synovitis was found at 5 (22%) of the 23 first visits during which flares were diagnosed compared with 2 (3%) of the other 66 visits (P = 0.04), limb girdle pain at 21 (95%) of 22 visits compared with 18 (30%) of 59 (P < 0.0001), fever at 4 (18%) of 22 visits compared with 0 (0%) of 66 (P < 0.0001), nocturnal awakenings at 18 (78%) of 23 visits compared with 7 (11%) of 64 (P < 0.0001), and weight loss at 4 (18%) of 22 visits compared with 7 (11%) of 66 (P > 0.05).

Table 1. Mean values of the total polymyalgia rheumatica activity score (PMR-AS) and of each of its components, and number of patients with each score for elevation of the upper limbs (EUL), in the groups with and without a flare diagnosis at the 89 first visits or the 149 visits overall*
 No. of visitsScore
No flareFlareNo flareFlare
  • *

    Values are the mean ± SD (range) unless otherwise indicated. VASp = pain severity assessed by the patient on a 10-point visual analog scale (VAS); VASph = physician's global assessment of disease activity on a 10-point VAS; MST = morning stiffness; CRP = serum C-reactive protein level.

  • P < 0.0001 for flare versus no flare.

89 first visits    
 VASp (0–10)66232.34 ± 2.00 (0–7)6.16 ± 1.96 (1.5–8.7)
 VASph (0–10)65221.33 ± 1.85 (0–9.5)6.17 ± 2.06 (0–10)
 MST (minutes)63237.10 ± 19.48 (0–120)83.91 ± 55.31 (5.0–240)
 CRP (mg/liter)64237.28 ± 7.82 (0–45)34.80 ± 28.91 (0.5–124)
 PMR-AS60225.04 ± 4.53 (0–23.9)25.26 ± 8.58 (12.2–52.4)
 EUL (0–3)6623  
  06412  
  101  
  2210  
  300  
149 visits overall    
 VASp (0–10)117322.06 ± 2.11 (0–7.5)5.93 ± 2.16 (1–8.7)
 VASph (0–10)116301.04 ± 1.51 (0–9.5)5.81 ± 2.18 (0–10)
 MST (minutes)114327.73 ± 22.72 (0–180)77.75 ± 63.29 (0–240)
 CRP (mg/liter)112315.88 ± 6.62 (0–45)33.08 ± 28.06 (0.4–124)
 PMR-AS108294.42 ± 4.88 (0–28.7)23.22 ± 9.13 (2.04–52.4)
 EUL (0–3)11732  
  011319  
  101  
  2412  
  300  

Flare diagnosis and glucocorticoid dose escalation.

As expected, flare diagnosis was strongly associated with glucocorticoid dose escalation (P < 0.0001). Of the 89 first visits, 83 were concordant for flare diagnosis and glucocorticoid dose (agreement 93.2%; 95% confidence interval [95% CI] 85.4–97.2]); 20 visits had both a flare diagnosis and glucocorticoid dose escalation, and 63 visits had neither a flare diagnosis nor glucocorticoid dose escalation. However, the glucocorticoid dose was increased at 3 visits without flare diagnoses and was not increased at 3 visits with flare diagnoses. Of the 149 visits overall, 138 were concordant for flare diagnosis and glucocorticoid dose (agreement 92.6%; 95% CI 86.9–96.1); 26 visits had both a flare diagnosis and glucocorticoid dose escalation, and 112 visits had neither a flare diagnosis nor glucocorticoid dose escalation. However, the glucocorticoid dose was increased at 5 visits without flare diagnoses and was not increased at 6 visits with flare diagnoses.

Diagnostic impact of clinical and laboratory parameters.

Analysis of the 82 first visits for which all PMR-AS parameters were available.

The ROC curves for associations linking flare diagnosis (n = 23) to PMR-AS, morning stiffness, ability to elevate the upper limbs, VASph, VASp, and CRP level at the 82 first visits are shown in Figure 1A. The best cutoff values for the best parameters, together with their sensitivities and specificities for flare diagnosis, are reported in Table 2. The 3 parameters most strongly associated with flare diagnosis were PMR-AS ≥10.96 (agreement 95.1%, 95% CI 87.3–98.4, κ = 0.88; sensitivity 100.0%, 95% CI 81.5–100.0; and specificity 93.3%, 95% CI 83.0–97.8), morning stiffness ≥17.5 minutes (sensitivity 95.5% and specificity 96.7%), and VASph ≥3.25 (sensitivity 95.5% and specificity 93.3%). Weaker associations were found for the other parameters. ROC curve analysis was also performed on these 82 first visits (Figure 1B) to identify factors associated with glucocorticoid dose escalation (n = 23). The parameter most strongly associated with glucocorticoid dose escalation was VASph ≥3.25 (sensitivity 95.5% and specificity 93.3%). A strong association with glucocorticoid dose escalation was found for PMR-AS ≥8.83 (sensitivity 95.5% and specificity 83.3%). Weaker associations were found for the other parameters (VASp ≥4.75: sensitivity 72.7% and specificity 80.0%; morning stiffness ≥12.5 minutes: sensitivity 85.2% and specificity 88.2%; CRP level ≥11 mg/liter: sensitivity 81.8% and specificity 91.7%; and elevation of upper limbs ≥1: sensitivity 50.0% and specificity 98.3%). PMR-AS values correlated closely with flare diagnosis in the 82 first visits (κ = 0.88) (Table 3).

Figure 1.

Receiver operating characteristic curves of associations linking A, flare diagnosis or B, glucocorticoid dose escalation with polymyalgia rheumatica activity score (PMR-AS), morning stiffness (MST), ability to elevate the upper limbs (EUL), physician's global assessment of disease activity on a 10-point visual analog scale (VASph), pain severity assessed by the patient on a 10-point VAS (VASp), and C-reactive protein level (CRP). Analysis on the 82 first visits. Ref = reference line.

Table 2. Optimal cutoffs with sensitivity and specificity of various parameters for diagnosing flares of polymyalgia rheumatica*
Population/parametersSensitivity, %Specificity, %
  • *

    See Table 1 for definitions.

  • Variation between the 2 last visits.

82 first visits  
 PMR-AS ≥10.9610093.3
 VASph ≥3.2595.593.3
 MST ≥17.5 minutes95.596.7
 CRP ≥8.65 mg/liter81.876.7
 VASp ≥3.886.475.0
 EUL ≥150.098.3
137 visits overall  
 PMR-AS ≥9.3596.690.7
 VASph ≥3.2593.193.5
 MST ≥17.5 minutes86.295.4
 CRP ≥11 mg/liter79.387.0
 VASp ≥3.882.877.8
 EUL ≥141.497.2
49 visit pairs  
 ΔPMR-AS ≥6.610097.1
 ΔVASph ≥2.0592.997.1
 ΔMST ≥12 minutes92.997.1
 ΔVASp ≥1.2510088.2
 ΔCRP ≥16.55 mg/liter85.7100
 ΔEUL ≥171.497.1
Table 3. Cohen's kappa for agreement between flare diagnosis by the rheumatologist and polymyalgia rheumatica activity score (PMR-AS) or ΔPMR-AS values*
Population/PMR-AS valuesNo flare diagnosisFlare diagnosisκ
  • *

    ΔPMR-AS = variation between the 2 last visits in PMR-AS.

82 first visits   
 PMR-AS <10.965600.88
 PMR-AS ≥10.96422 
137 visits overall   
 PMR-AS <9.359810.78
 PMR-AS ≥9.351028 
49 visit pairs   
 ΔPMR-AS <6.63300.95
 ΔPMR-AS ≥6.6115 

Analysis on the 137 visits overall and on the 49 visit pairs.

The ROC curves for associations linking flare diagnosis to PMR-AS, morning stiffness, ability to elevate the upper limbs, VASph, VASp, and CRP level at the 137 visits overall are shown in Figure 2A. The ROC curves for associations linking flare diagnosis to changes in the same variables between 2 visits in the group with 49 visit pairs are displayed in Figure 2B. Strong associations with glucocorticoid dose escalation were found for PMR-AS ≥9.35 (agreement 92%, 95% CI 85.8–95.7, κ = 0.78; sensitivity 96.6%, 95% CI 80.4–99.8; and specificity 90.7%, 95% CI 83.2–95.2) and for ΔPMR-AS ≥6.6 (agreement 98%, 95% CI 88–99.9, κ = 0.95; sensitivity 100%, 95% CI 74.7–100; and specificity 97.1%, 95% CI 82.9–99.8) (Table 2). Flare diagnosis was also strongly associated with VASph ≥3.25 (sensitivity 93.1% and specificity 93.5%), ΔVASph ≥2.05 (sensitivity 92.9% and specificity 97.1%), Δ(morning stiffness) ≥12 minutes (sensitivity 92.9% and specificity 97.1%), and ΔVASp ≥1.25 (sensitivity 100% and specificity 88.2%). Associations were weaker for the other parameters.

Figure 2.

Receiver operating characteristic curves of associations linking flare diagnosis with A, polymyalgia rheumatica activity score (PMR-AS), morning stiffness (MST), ability to elevate upper limbs (EUL), physician's global assessment of disease activity on a 10-point visual analog scale (VASph), pain severity assessed by the patient on a 10-point VAS (VASp), and C-reactive protein level (CRP), and B, variations between the last 2 visits in PMR-AS or its components (dPMR-AS, dMST, dEUL, dVASph, dVASp, and dCRP). Analysis on A, 137 visits overall and B, 49 visit pairs.

ROC curve analysis was performed to identify factors associated with a diagnosis of repeat flare (n = 24) as opposed to first flare (at the diagnosis). The results were similar to those of the ROC curve analysis on all 32 flares (curves not shown). The 2 parameters most strongly associated with repeat flare diagnosis were PMR-AS ≥9.35 (sensitivity 95.2% and specificity 90.7%) and ΔPMR-AS ≥6.6 (sensitivity 100% and specificity 97.1%). Strong associations with repeat flare diagnosis were also found for VASph ≥3.45 (sensitivity 90.5% and specificity 93.5%), Δ(morning stiffness) >12 minutes (sensitivity 88.9% and specificity 97.1%), ΔVASph >2.05 (sensitivity 88.9% and specificity 97.1%), ΔVASp ≥1.95 (sensitivity 100% and specificity 88.2%), and ΔCRP level ≥16.55 mg/liter (sensitivity 77.8% and specificity 100%). Weaker associations were found for morning stiffness ≥19.5 (sensitivity 85.7% and specificity 95.4%), CRP level ≥11 mg/liter (sensitivity 71.4% and specificity 87.0%), and VASp ≥4.85 (sensitivity 81.0% and specificity 84.3%). The ability to elevate the upper limbs was weakly associated with repeat flare diagnosis (elevation of upper limbs ≥1: sensitivity 28.6% and specificity 97.2%, and Δ[elevation of upper limbs] ≥1: sensitivity 55.6% and specificity 97.1%).

ROC curve analysis was also performed to study the associations linking glucocorticoid dose escalation (31 of 137 visits) to PMR-AS or its components (curves not shown). The parameter most strongly associated with glucocorticoid dose escalation was VASph ≥3.25 (sensitivity 92.6% and specificity 90.9%). Strong associations with glucocorticoid dose escalation were found for PMR-AS ≥9.03 (sensitivity 92.6% and specificity 83.6%). Weaker associations were found for the other parameters (VASp ≥3.8: sensitivity 77.8% and specificity 75.5%; morning stiffness ≥12.5 minutes: sensitivity 85.2% and specificity 88.2%; and elevation of upper limbs ≥1: sensitivity 44.4% and specificity 97.3%).

PMR-AS values correlated closely with flare diagnosis (κ = 0.78) (Table 3). Of the 99 visits with PMR-AS values <9.35, all but 1 had no flare diagnosis. However, of 38 visits with PMR-AS values ≥9.35, 10 had no flare diagnosis. The mean ± SD PMR-AS value was 17.7 ± 5.9 (range 11.3–28.6) in these 10 cases. Table 3 shows ΔPMR-AS values (κ = 0.95).

DISCUSSION

This study supports the usefulness of the PMR-AS for diagnosing flares. Thus, the PMR-AS may help rheumatologists and general practitioners to monitor PMR activity and to manage glucocorticoid tapering, particularly as it seems well suited to everyday practice. The PMR-AS is a composite index, which minimizes the impact of the variance associated with each measure (5).

PMR is an inflammatory condition of unknown etiology that is common among older individuals. It is generally benign, since glucocorticoid therapy is nearly always effective. To taper the glucocorticoid dose and to detect flares, physicians usually rely on assessments of symptoms and serial ESR or CRP values (1). However, side effects of glucocorticoids are common in individuals older than 50 years, as shown by the 2.5-fold higher risk of diabetes mellitus, vertebral fractures, femoral neck fractures, and hip fractures in patients with PMR than in the general population of the same age (6). In addition, several studies have found higher prevalences of cardiovascular disease in patients with PMR (7, 8). However, the increased cardiovascular risk may be related to long-term inflammation rather than glucocorticoid therapy (9). Therefore, carefully tailoring the glucocorticoid dose to each patient's needs is crucial in order to strike the best compromise between the risk of side effects and the risks associated with inflammation. Leeb and Bird (3) developed a disease activity score, the PMR-AS, as a tool for improving the assessment of PMR. In an earlier study (4) using clinical vignettes evaluated by 35 rheumatologists, we found that a PMR-AS score ≥7 was 98.1% sensitive and 94.3% specific for flare diagnosis. Leeb and Bird (3) reported that 7 was the cutoff between low and moderate disease activity. Here, we sought to identify the best PMR-AS cutoff for flare diagnosis in prospectively included patients, using the same statistical methods as in our previous study.

The PMR-AS showed excellent performance characteristics for diagnosing PMR flares in prospectively included patients. PMR-AS values ≥9.35 were 96.6% sensitive and 90.7% specific for flare diagnosis. The 9.35 value was a better cutoff than the 10.96 value for several reasons: the PMR-AS value was overestimated in the 82 first visits because patients with first flares (at diagnosis), which were included in these visits, had higher disease activity than did patients with repeat flares; in the ROC curve analysis excluding the visit at diagnosis (first flare), 9.45 was the best PMR-AS cutoff (sensitivity 95.2% and specificity 90.7%); and observations were probably independent or nearly independent, so that analyzing the overall visits was legitimate. An increase ≥6.6 in PMR-AS between 2 visits showed even better diagnostic performance, in keeping with the results of our clinical vignette study (4). Using a single absolute PMR-AS value to diagnose flares is possible only in patients with previously controlled disease; an absolute value >9.35 with previously uncontrolled disease indicates high disease activity but does not necessarily reflect a flare. The magnitude of the PMR-AS change (ΔPMR-AS ≥6.6) seems more relevant because it applies to all stages of the disease; however, the cutoff for this parameter needs to be confirmed by further studies. Most of the individual components performed less well than the total score. Although VASph and ΔVASph performed well, the variability in VASph (4) limited the value of these parameters. Until now, PMR flare was defined as an increase in ESR or CRP level and/or recurrence of clinical symptoms, with no specific criteria regarding the magnitude or nature of the changes. In our previous study, we found that more than 80% of rheumatologists agreed regarding the presence or absence of flares for 78% of the clinical vignettes. The results from the current study suggest that the PMR-AS may be a useful reference standard for diagnosing flares and comparing treatments in future studies.

Of the 99 visits with PMR-AS values <9.35, only 1 had a flare diagnosis. However, 10 visits with PMR-AS values >9.35 had no flare diagnosis. Of these 10 visits, 8 patients had no flare diagnosis by the rheumatologist and no CRP level elevation but had PMR-AS values >9.35 because of high values of VASp and morning stiffness; the other 2 patients had PMR-AS values >9.35 reflecting high disease activity of a recent flare that was not yet fully controlled by the treatment. Patients with PMR are treated chiefly by general practitioners, and the shortage of rheumatologists requires careful patient selection for referral (10). One strategy might involve patient referral by general practitioners to rheumatologists when the PMR-AS is >9. As a first step toward assessing this strategy, we have planned a clinical vignette study to determine whether a cutoff of ∼9 as assessed by general practitioners also correlates closely with PMR flares.

Although flare diagnosis was closely associated with glucocorticoid dose escalation (P < 0.0001), in 6 cases the rheumatologist decided not to increase the glucocorticoid dose despite a flare diagnosis. Other medications were used in these cases, mainly nonsteroidal antiinflammatory drugs or disease-modifying antirheumatic drugs (e.g., methotrexate), probably in an effort to limit side effects from glucocorticoid therapy (11).

Our study has 2 main limitations. First, diagnosis by a rheumatologist was the reference standard for identifying PMR, as opposed to published criteria sets. However, none of the study patients was later diagnosed with another condition. Even the best criteria for PMR (12), including the Bird et al criteria (13) or Chuang et al criteria (14), lack sensitivity and specificity. There is no consensus about diagnoses that must be ruled out or investigations that must be performed before diagnosing PMR. The constitution of an American College of Rheumatology/EULAR study group to work on diagnostic criteria for PMR would be very useful. The second limitation is that external validation of flare diagnoses was not available, as the same rheumatologist determined whether a flare was present and assessed the VASph during the same visit. This fact may have artificially strengthened the correlation between flare diagnosis and PMR-AS values.

Nevertheless, the results of this study support the relevance of the PMR-AS as a tool for assessing PMR activity in everyday practice. The PMS-AS is simple and fast to determine. The cutoffs associated with flares were very similar to those found previously using clinical vignettes (4), and the difference between 2 consecutive visits performed even better than the absolute value. Thus, the PMR-AS is a simple tool that may help both general practitioners and rheumatologists to evaluate the glucocorticoid needs of the individual patient. It may be particularly useful in patients whose ESR and CRP values remain low from disease onset (15). However, a prospective randomized controlled study in an inception cohort of patients with PMR is needed to determine whether closely spaced PMR-AS determinations by rheumatologists and general practitioners improve flare diagnosis and reduce glucocorticoid exposure. The PMR-AS should also prove useful in clinical trials to ensure uniformity across study centers.

AUTHOR CONTRIBUTIONS

Dr. Saraux 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 design. De Bandt, Saraux.

Acquisition of data. Binard, De Bandt, Berthelot, Saraux.

Analysis and interpretation of data. Binard, De Bandt, Berthelot, Saraux.

Manuscript preparation. Binard, Saraux.

Statistical analysis. Binard, Saraux.

Acknowledgements

We are grateful to E. Nowak for assistance with the statistical analysis. We are indebted to the following physicians for their participation in the study: J. P. Benais, L. Beraneck, A. Besson, B. Boyer, D. Caplanne, J. P. Chabellard, P. Chicault, J. F. Debarge, A. E. Dumel, E. Gibert, P. Leca, V. Lemaire, P. Legros, B. Loze, F. Magnan, J. Marcou, J. C. Marquestaut, S. Melac-Ducamp, Q. Vuong Dang, E. Sornay-Rendu, J. Taillandier, and C. Visticot. We thank the following organizations for participating in the study: Association des Rhumatologues Parisiens, Association des Rhumatologues de l'Essonne, Association des Rhumatologues du Val de Marne, Association des Rhumatologues du Finistère, Association des Rhumatologues du Val d'Oise, Club Rhumatismes et Inflammation, Fédération Française de Rhumatologie, and Société de Rhumatologie de l'Ouest.

Ancillary