To confirm the reliability and applicability of the Polymyalgia Rheumatica Disease Activity Score (PMR-AS), and to establish a threshold for remission.
To confirm the reliability and applicability of the Polymyalgia Rheumatica Disease Activity Score (PMR-AS), and to establish a threshold for remission.
First, 78 patients with PMR (50 women/28 men, mean age 65.97 years) were enrolled in a cross-sectional evaluation. The PMR-AS, patient's satisfaction with disease status (PATSAT; range 1–5), erythrocyte sedimentation rate (ESR; first hour), and a visual analog scale of patients' general health assessment (VAS patient global; range 0–100) were recorded. Subsequently, another 39 PMR patients (24 women/15 men, mean age 68.12 years) were followed longitudinally. Relationships between the PMR-AS, PATSAT, ESR, and VAS patient global were analyzed by the Kruskal-Wallis test, Spearman's rank correlation, and kappa statistics. PMR-AS values in patients with a PATSAT score of 1 and a VAS patient global <10 formed the basis to establish a remission threshold.
PMR-AS values were significantly related to PATSAT (P < 0.001), VAS patient global (P < 0.001), and ESR (P < 0.01). PATSAT and VAS patient global were reasonably different (κ = 0.226). The median PMR-AS score in patients with PATSAT score 1 and VAS patient global <10 was 0.7 (range 0–3.3), and the respective 75th percentile was 1.3. To enhance applicability, a range from 0 to 1.5 was proposed to define remission in PMR. The median ESR in these patients was 10 mm/hour (range 3–28), indicating external validity.
We demonstrated the reliability, validity, and applicability of the PMR-AS in daily routine. Moreover, we proposed a remission threshold (0–1.5) founded on patient-dependent parameters.
In 2003, the European League Against Rheumatism (EULAR) response criteria for polymyalgia rheumatica (PMR) were published comprising a core set of disease activity markers for monitoring therapeutic responses in patients with PMR, namely, erythrocyte sedimentation rate (ESR) or C-reactive protein level (CRP), representing the acute-phase response, and a visual analog scale (VAS) of the patient's pain, physician's global assessment, morning stiffness, and the ability to elevate the upper limbs, representing the clinical situation (1). Until that time, the gold standard of monitoring PMR included measurement of the acute-phase response and patient's global assessment (2, 3), despite the fact that PMR may also exist without elevated acute-phase reactants (4). In view of the crucial role pain plays during the course of the disease, it was decided that a change in pain intensity must occur, whereas only 3 of the other 4 parameters must change, to indicate improvement or deterioration (1).
As a consequence of the development of the EULAR response criteria for PMR, it was considered useful to have a disease activity index for PMR available. Combining the EULAR response criteria core set parameters, a disease activity score (the PMR Activity Score [PMR-AS]) was generated, which applied an easy-to-calculate formula (Figure 1) (5). This index, expressing disease activity as a number, provides advantages with respect to the comparability of patients, primarily because recent disease activity can be described without a baseline observation (5).
The PMR-AS proved to be an easily applicable and valid tool for disease activity monitoring in patients with PMR, and was shown to mirror patients' satisfaction with disease status (6). However, another important issue was addressed, namely, the definition of a remission threshold according to the PMR-AS.
We, together with many others including the World Health Organization, define remission as the disappearance of the signs and symptoms of the disease despite ongoing therapy. When this occurs, the disease is said to be in remission. A remission can be temporary or permanent. Thus remission can also be regarded as being very similar to a complete clinical response (7, 8). Therefore, we performed a 2-step observational study in daily clinical routine, namely, in a hospital-based outpatient clinic and a private office, with the goal of proposing a threshold of remission as well as confirming the correlation of the PMR-AS with the past gold standards of disease activity monitoring.
As the first step, 78 patients (50 women/28 men, ages 38–84 years, mean age 65.97 years) with PMR as defined by the criteria developed by Bird et al (9, 10), and without evidence of giant cell arteritis (excluded by clinical means ), were enrolled in a cross-sectional evaluation from July to December 2003. Mean disease duration in these patients was 8 months (range 1–34 months), and no patient met the American College of Rheumatology (formerly the American Rheumatism Association [ARA]) criteria for rheumatoid arthritis (RA) (12) at presentation or at the point of assessment. All but 9 of the 78 patients were treated with corticosteroids only (2.5–25 mg prednisolone equivalent). The respective 9 patients were assessed before the initiation of corticosteroids at their first presentation.
At a regular control visit, disease activity was assessed by administering the PMR-AS and patients were classified with respect to the PMR-AS disease activity categories (5). At the same time, patient satisfaction with disease status (PATSAT; ranging from 1 [excellent] to 5 [completely unsatisfactory], according to the Austrian school marking system), ESR (first hour, normal up to 15 mm/hour, not age adjusted), and the patient's assessment of general health measured on a 100-mm VAS (VAS patient global; ranging from 0 [perfect] to 100 [dreadful]) were recorded. In RA, PATSAT proved to be highly correlated with established disease activity indexes, also taking into account the patient's attitude and thoughts (13). The Austrian school marking system resembles a Likert scale and is well established; therefore, misinterpretations by the patients are highly unlikely.
The purpose of this part of the study was to confirm the correlation of the PMR-AS with ESR and the patient's assessment of general health (14). Moreover, to obtain a second patient-dependent parameter for defining remission in the next step, the relationship of PATSAT with the patient's global assessment of general health and with the PMR-AS was examined.
As the second step, 39 patients with PMR (24 women/15 men, mean age 68.12 years, range 41–85) were followed longitudinally from July 2003 to December 2004, resulting in 199 assessments. All of these patients were treated in a private rheumatologic practice. Their mean disease duration was 10 months (range 1–43 months). As far as it was possible to determine retrospectively, all patients had elevated acute-phase reactants at the onset of their disease. Similar to the first cohort, all patients met the Bird et al classification criteria for PMR (9, 10) and did not show clinical signs of giant cell arteritis (11). Moreover, no patient enrolled in the study met the ARA classification criteria for RA at any time point throughout the observation period (12). Patients were assessed at least twice throughout the observation period for PMR-AS, ESR (first hour), VAS patient global, and PATSAT, as well as the respective changes. All patients were treated with corticosteroids (median dose 3.75 mg prednisolone equivalent, range 2.5–25), and 9 assessments (5%) were performed after patients stopped corticosteroid treatment. Two patients were additionally treated with methotrexate (25 mg/week).
We postulated that a PATSAT score of 1 (excellent) and a VAS patient global score <10 mm would characterize a patient in remission accurately enough, and the respective PMR-AS values were considered for the preliminary definition of a remission status. ESR was not part of the PMR-AS originally; however, it can be considered a surrogate if CRP values are not available, calculated as ESR × 0.1 (5). Thus its value could be considered as another control for the accuracy of the PMR-AS value, which should indicate remission.
Because corticosteroids are known to exert a rapid and sustained response in patients with PMR (15), in this second cohort we evaluated the relationship between PMR-AS values and daily corticosteroid dose, as well as an eventual influence exerted by corticosteroid dose on PATSAT or patient's global assessment.
Statistical evaluation was performed by SPSS for Windows, version 11.0 (SPSS, Chicago, IL). Testing for normal distribution was primarily performed by applying the Kolmogorov-Smirnov accommodation. Internal consistency of the PMR-AS was assessed by calculating standardized item alpha (16). Because all parameters were found to be not normally distributed, nonparametric tests were applied, such as Spearman's rank correlation and the Kruskal-Wallis test. The relationship between PMR-AS values and daily corticosteroid dose was investigated by regression analysis, and an eventual influence on PATSAT or patient's assessment was investigated by applying kappa statistics after categorizing by creating quintiles of the daily corticosteroid dose and patient's assessment. To achieve better comparability of PATSAT and the patient's global assessment, the latter was automatically categorized and subsequently both categorical variables were compared by creating crosstabs and calculating kappa values.
The median PMR-AS score in the first group (78 cross-sectionally evaluated patients) was 3.75 (range 0.5–29.2). Median PATSAT was 2 (range 1–5) and median VAS patient global score was 15 (range 0–95), whereas the median ESR (first hour) was 15 mm/hour (range 2–88).
PMR-AS values at the different stages of PATSAT were statistically significantly different (P < 0.001) according to the Kruskal-Wallis test (Figure 2). Applying Spearman's rank correlation, a pronounced statistically significant relationship between PMR-AS values and the VAS patient global (r = 0.76, P < 0.001) and between the PMR-AS and ESR (r = 0.382, P < 0.01) was found, indicating external validity of the PMR-AS when compared with the previous standards of disease activity monitoring (Figure 3A and 3B).
To obtain information as to whether or not PATSAT and VAS patient global are strongly related to one another, the values of the VAS patient global were categorized into quintiles. Subsequently, crosstabs between these quintiles and PATSAT were calculated (Figure 4). The crosstabs demonstrated reasonable differences between the 2 parameters; kappa value was 0.226, indicating only a weak relationship between the parameters. Commonly, kappa values >0.65 are regarded as indicators of a strong relationship. Additionally, we related PATSAT and VAS patient global to the categories of disease activity by creating crosstabs, and both parameters, with PATSAT (y = 0.972) at a reasonably higher degree, were seen to be in congruence with the disease activity categories.
The main objective of the second part of this study was to propose a threshold for remission of PMR because such a definition was not established yet. In 39 patients, 199 assessments could be analyzed. The median PMR-AS value was 4.05 (range 0–35.0), the median PATSAT was 2 (range 1–5), and the median VAS patient global was 24.5 (range 0–95). It is important to remark that all patients achieved a 70% response according to the EULAR response criteria at least once (1).
Internal consistency testing of the PMR-AS in this population revealed a standardized item alpha of 0.812, indicating high reliability of the score (16). Factor analysis by principal component analysis demonstrated that the PMR-AS constitutes a monodimensional instrument. Additionally, the PMR-AS was not age related, but women had significantly higher PMR-AS values (P < 0.005 according to the Mann-Whitney U test).
We had postulated that a PATSAT score of 1 (excellent) and a VAS patient global score <10 mm accurately characterize remission in a patient. Therefore, patients were categorized accordingly. In 24 of the 199 assessments (12 patients, 6 women/6 men), an excellent PATSAT score and a VAS patient global score <10 mm were found simultaneously. The median PMR-AS in this respective group was 0.7 (range 0–3.3). In another 36 assessments, one of these prerequisites was fulfilled, and in 139 assessments none had been met (Figure 5). PMR-AS values in these 3 groups were statistically significantly different according to the Kruskal-Wallis test (P < 0.001). The 75th percentile of the median PMR-AS (0.7) in the 12 patients meeting the postulated definition of remission was 1.3. Therefore, a range of 0–1.3 could have been proposed to describe remission according to the PMR-AS. To enhance the applicability of this definition, however, we decided to extend the range from 0 to 1.5 to define a remission-like status in patients with PMR.
As a primary control, the median ESR (first hour) in the patients with a PMR-AS score of 1.5 was 8 (range 3–28), indicating external validity of the threshold of remission. Given the average age of the patients investigated, even the maximum value of the ESR can be regarded as being within the normal range (17).
Regression analysis revealed a highly significant statistical relationship between the daily corticosteroid dose and PMR-AS values (r = 0.681, P < 0.0001), which indicates that PMR-AS values constituted one of the major issues for adjusting daily corticosteroid doses. To reveal eventual influences of the daily corticosteroid dose on PATSAT or patient's assessment, quintiles of the daily corticosteroid doses were calculated and related to PATSAT and the quintiles of patient's assessment. The kappa value for the relationship between daily corticosteroid dose and PATSAT was 0.061 (not significant) and was 0.043 (not significant) for the relationship with patient's assessment.
The 2 objectives of this investigation were to prove the applicability of the PMR-AS and its external and internal validity, and to establish a threshold for the definition of remission or complete clinical response in patients with PMR. Regarding the first objective, the PMR-AS was cross-sectionally compared with the previous standards of disease activity monitoring, namely, ESR and VAS patient global (14). As in the preliminary study, which was part of the very first published article of the PMR-AS, the score proved to be as reliable as ESR measurements and VAS patient global assessment for expressing the disease activity of patients with PMR (5).
In addition, another goal was to define differences, at least in part, between patient satisfaction with disease status and patients' assessment of global health in order to apply 2 patient-dependent parameters as standards for the definition of remission in the second part of the study. Whether patients are indeed able to distinguish between general health status and their satisfaction with the status of a distinct disorder is a widely discussed question. The results obtained here, as expressed by the low kappa value, indicate clearly that these specific patients were able to differentiate between the 2 questions, which encouraged us to use both parameters as standards for defining remission.
In addition, ESR values, which are not part of the PMR-AS, were chosen as a control of external validity. Because physician's global assessment of disease status is indeed one of the components of the PMR-AS, it was inappropriate to use this parameter as a control, as in the case of CRP level, which also constitutes a component of the score. Validating a composite score by relating the score to one of its components may obviously flaw the results.
Daily corticosteroid doses were found to be highly significantly related to PMR-AS values mirroring daily routine. However, kappa statistics did not reveal any influence of corticosteroid dose on PATSAT or patient's assessment. The routine therapeutic strategy is to bring the patient as soon as possible into the best clinical situation achievable after diagnosing PMR. This method of handling PMR therapy appears to be the main reason for the lack of a relationship between corticosteroid dose and patient-dependent parameters.
It seems convincing that it is more appropriate to define a remission threshold by applying patient-dependent parameters as controls. The PMR-AS already incorporates CRP level, symptoms of the disease (such as pain, morning stiffness, and the ability to raise the upper limbs), and physician's assessment (5). Moreover, the extent to which the patient is satisfied with the result of the physician's efforts is of greater importance. Therefore, we postulated that excellent satisfaction with disease status constituted the first prerequisite for remission, while a reasonably good general health assessment by the patient constituted the second prerequisite (18). General health assessment also includes possible side effects or the status of concomitant diseases, which may be influenced by the treatment of PMR in this case (19). Can we really speak of remission if improvement of the respective disease is achieved at the expense of serious impairments of the patient's general well-being?
The internal consistency and external validity of the PMR-AS were proved during the first part of this study. Subsequently, the proposal that a PMR-AS value of 1.5 or lower indicates remission of PMR, regardless of the amount of medication administered, is justified. The results for the relationship between corticosteroid doses and the patient-dependent parameters strongly support this assumption.
If a complete clinical response can be maintained after stopping treatment, one may speak of a cure, which is indeed achievable in many patients with PMR (20). It could be extrapolated that a patient achieving a PMR-AS score of 1.5 or lower without treatment can be regarded as being in a status of cured PMR. The patients achieving PMR-AS values indicating remission also had a normal ESR, indicating congruence of the remission threshold with an objective parameter of inflammation, which can be seen as proof of external validity.
In summary, it was possible to confirm external and internal validity of the PMR-AS during this study. In addition, the comparability of the PMR-AS with other parameters to monitor PMR disease activity was proven. These parameters, however, present the disadvantage of not representing a reproducible and categorical parameter. The reliability of the originally established disease activity categories was also proved when compared with the patient's attitudes, and in addition it was possible to establish a remission threshold based on patient parameters and confirmed by an objective parameter of inflammation. Thus the proposal of a PMR-AS remission threshold constitutes an enhancement of the tool's applicability for disease activity monitoring. In combination with the PMR response criteria, a better description of response to new therapies should also be possible. We anticipate that better monitoring of disease activity and treatment responses in patients with PMR will result in improved clinical decision making and ultimately in improved patient care.
Dr. Leeb 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. Leeb, Rintelen, Bird.
Acquisition of data. Leeb, Rintelen, Sautner, Fassl.
Analysis and interpretation of data. Leeb, Rintelen, Sautner, Fassl.
Manuscript preparation. Leeb, Rintelen, Sautner, Fassl, Bird.
Statistical analysis. Leeb, Rintelen.