Differential diagnostics of polymyalgia rheumatica in a university hospital in Finland

Objectives Diagnosing polymyalgia rheumatica (PMR) can be difficult as many conditions present with similar symptoms and findings. This study aimed to analyse how often the diagnosis of PMR changes during follow-up in a university hospital setting and to determine the most common clinical conditions initially misdiagnosed as PMR. Method All patients with a new primary diagnosis of PMR on at least one visit during the years 2016–2019 were identified from the hospital discharge register of Turku University Hospital, Finland. A diagnosis of PMR was confirmed if the patient met at least one of the five classification criteria, complete clinical follow-up (median 34 months) was compatible with PMR, and no other diagnosis better explained their condition. Results Of the patients initially diagnosed with PMR, 65.5% were considered to have PMR after further evaluation and clinical follow-up. The most common conditions initially diagnosed as PMR were inflammatory arthritides (34.9%), degenerative or stress-related musculoskeletal disorders (13.2%), infection (9.3%), malignancy (9.3%), giant cell vasculitis (6.2%) and other vasculitis (6.2%), and a wide range of other less common diseases. The diagnosis of PMR remained in 81.3% of patients who fulfilled the 2012 American College of Rheumatology/European League Against Rheumatism PMR classification criteria and in 45.5% of patients who did not. Conclusions Diagnosing PMR is challenging, even in a university hospital. One-third of the initial diagnoses of PMR changed during further evaluation and follow-up. There is a substantial risk of misdiagnosis, especially in patients with atypical presentation, and the differential diagnoses of PMR must be considered carefully.

Diagnosing PMR can be difficult as many other conditions present with similar symptoms and findings.Different diagnoses to consider include a wide variety of disorders such as other rheumatic diseases, autoimmune disorders, infections, and malignant diseases.Little research into the difficulties of diagnosing PMR has been conducted in recent years.The aim of this study was to analyse patients diagnosed with PMR in Turku University Hospital and to determine how often a competing diagnosis was found during a clinical follow-up, as well as to find out the most common conditions misdiagnosed as PMR.

Method
All patients with a new primary diagnosis of PMR [International Classification of Diseases, 10th revision (ICD-10) code M35.3] on at least one inpatient or outpatient visit during the years 2016-2019 were identified from the hospital discharge register of Turku University Hospital in Finland.Patient charts were systematically reviewed by JP and SS, and data were collected according to a standardized protocol.The PMR diagnosis was evaluated taking into account the full clinical follow-up (median 34 months).A true positive diagnosis of PMR was confirmed based on whether the patient met at least one of the five classification criteria (4)(5)(6)(7)(8), if the complete clinical followup was compatible with PMR, and if no other diagnosis better explained the patient's condition.If some other diagnosis better explained the patient's condition after follow-up, the diagnosis of PMR was considered not correct, even if the PMR classification criteria were still met.If there was uncertainty or discrepancy between the reviewers about the diagnosis, the final decision was made by JP as a senior rheumatologist.The authors examined how often the primary diagnosis of PMR changed after further diagnostic evaluation or during follow-up, as well as what the most common final diagnoses were.
Collected data included past and current comorbidities of the patients, current symptoms and clinical findings, laboratory and imaging results, maximum dosage of glucocorticoid used, the number of visits to the hospital with the diagnosis of PMR, and whether the diagnosis had been made at a rheumatology clinic or in another healthcare unit in the hospital.A record was also made of whether some other disease would better explain the patient's symptoms and findings during the follow-up.When evaluating the fulfilment of classification criteria, an item was counted as negative if it was not marked as positive in the patient charts, but the absence of rheumatoid factor (RF) and/or anti-cyclic citrullinated peptide (anti-CCP) antibodies had to be confirmed with a negative test result.The fulfilment of ACR/EULAR classification criteria was evaluated without using ultrasound, because ultrasound was not routinely used by all clinicians in our hospital for PMR evaluation during the study years, and negative findings may not have been reported.Morning stiffness was considered relevant if present, given that the duration of morning stiffness was rarely recorded.Other symptoms that were considered typical for PMR were pain and tenderness in the neck, shoulder, upper arm, hip, and thigh areas.
For 73 patients, there were insufficient patient record data available for analysis, which resulted in the reviewer not being able to evaluate the correctness of the diagnosis.For eight patients, the diagnoses in the hospital register were a clear deviation from the physician's record and were considered to have been input incorrectly.These 81 patients were excluded from the final analysis.
Turku University Hospital is a tertiary referral centre of the Hospital District of Southwest Finland, with a population of 480 000 in the catchment area.The patients come into the hospital either through the emergency department or with a referral from other healthcare units.In our study, the duration of symptoms was calculated from the beginning of the symptoms to the first visit to the hospital with a recorded PMR diagnosis, and the duration of follow-up onwards from this visit, even if the diagnosis had already been made earlier in another healthcare unit.
Study data were collected and managed using RED-Cap electronic data capture tools hosted at the University of Turku (9,10).

Statistics
Statistical analyses were performed using R version 3.6.2,with the R base, dplyr, stringr, ggfortify, and survival packages.Continuous variables are expressed as medians with interquartile ranges (IQRs), and categorical variables are described as counts with percentages.When comparing differences between different categories, the p values were analysed using the nonparametric Mann-Whitney U-test for continuous variables and Pearson's chi-squared test for categorical variables.Multivariable logistic regression analysis was performed to study which variables predicted the change in PMR diagnosis.Six variables were selected for this analysis based on their clinical relevance, and these variables are often included in classification criteria.Two-sided p values < 0.05 were considered statistically significant.

Ethical considerations and study permission
This was a non-interventional retrospective study without any direct patient contact, and according to Finnish legislation, no patient consent or ethical committee approval was needed.Permissions for the study were obtained from the Hospital District of Southwest Finland.The legal basis for processing personal data is public interest and scientific research [EU General Data Protection Regulation 2016/679 (GDPR), Article 6(1)(e) and Article 9(2)(j); Data Protection Act, Sections 4 and 6].

Results
The inclusion and exclusion of the study patients are depicted in Figure 1.There were 374 patients in the final analysis.Demographic and clinical characteristics of the study population are presented in Table 1.57.2% of the patients were female, and the median age at diagnosis was 70 years.Most of the patients were diagnosed (79.0%) and treated (81.0%) at a department of rheumatology.The median duration of follow-up was 34.0 (IQR 21.0-50.0)months.
Of the 374 patients, upon follow-up, 245 (65.5%) were considered to have PMR, whereas for 129 patients (34.5%), the follow-up did not support the diagnosis of PMR.
The most common conditions initially diagnosed incorrectly as PMR were inflammatory arthritides Out of the malignancies, 33% (4/12) were carcinomas of the colon, 25% (3/12) were lymphomas, and the rest were individual cases of other malignancies: breast, uterus, kidney, peritoneal carcinosis, and chronic lymphocytic leukaemia.In 75% (9/12) of the patients with malignancies, the symptoms were considered  paraneoplastic and not directly caused by the malignancy.Four of these nine patients received active treatment for their malignancy, and all became asymptomatic after treatment.
The patients had suffered from symptoms for a median of 10.  2).
The diagnosis made in a department of rheumatology was correct in 71.8% (211/294) of the patients, compared to 43.0% (34/79) of the patients for whom the diagnosis had been made in other departments.
Patients whose diagnosis did not change during followup more often had a longer symptom duration before diagnosis (p = 0.016), elevated C-reactive protein (CRP) or erythrocyte sedimentation rate (ESR) values (p < 0.001), and a full response to symptoms (p < 0.0001) and inflammatory values (p < 0.0001) with glucocorticoid treatment compared to those patients whose diagnosis changed during follow-up.These patients with a true PMR diagnosis were also more often diagnosed in the department of rheumatology (p < 0.0001) and had been followed for a longer period (p = 0.011) (Table 1).In a multivariable logistic regression analysis, a change in the PMR diagnosis was statistically significantly predicted if the patient had no morning stiffness [odds ratio (OR) of 1.75, 95% confidence interval (CI) 1.03-2.94],had none of the other typical symptoms of PMR (OR 7.45, 95% CI 2.65-24.53),and had less than full resolution of symptoms with glucocorticoid treatment (OR 3.01, 95% CI 1.83-5.00).If the patient had normal inflammatory values, the OR for the change of diagnosis was 2.27 (95% CI 0.98-5.24).Gender and younger age at diagnosis were not statistically significant predictors of a change in diagnosis (Table 2).

Discussion
PMR is an inflammatory disease that causes muscle pain and morning stiffness, especially in the shoulders and hips, often with elevated inflammatory markers (1)(2)(3).Diagnosing PMR can be difficult, as the diagnosis is mainly clinical without a gold standard to confirm the diagnosis, and many symptoms and findings of PMR may be present in other conditions (1)(2)(3)11).In our study, 65.5% of the patients initially diagnosed with PMR were considered to have PMR after further evaluation and clinical follow-up.
The validity of PMR diagnoses in primary healthcare has previously shown to be 60% in a study in Sweden by Fors et al (12).In a hospital setting, the persistence of PMR diagnoses at the 1 year follow-up has previously varied between 48% and 72% in Italy (13,14), and was 94% in the UK for patients meeting the Jones and Hazleman criteria (15).Compared to these studies, the follow-up period in our study was markedly longer, with a median of 34 months, and 13.7% of the changed diagnoses happened after a 1 year follow-up.Also, unlike these previous studies, the patients in our study were not prescreened since the diagnosis of PMR was the only inclusion criterion.For diagnoses made in a department of rheumatology, the proportion of correct diagnoses in previous studies was 79.1% in Alaska (16) and 69.0% in the UK (17), and these figures correspond to the 71.8% in our study.The proportion of correct diagnoses was higher in the department of rheumatology compared to other departments in the hospital.We have previously shown similar results with rheumatoid arthritis and systemic sclerosis (18,19).
In our study, when the PMR diagnosis changed during follow-up, the diseases that better explained the patient's condition were largely consistent with those described earlier (20,21).Other inflammatory joint diseases were the most common mimics of PMR in 35% of changed diagnoses, with degenerative and stress-related musculoskeletal disorders in second place with 13%.In our study, of the patients for whom the diagnosis changed later, 9.3% had a malignancy, with the median of diagnostic delay being 2 months.Especially with malignancies, minimizing the diagnostic delay is important, since delay may worsen a patient's prognosis.
The results of our study show that patients with atypical presentation of PMR had a greater risk of initial misdiagnosis.In a multivariable logistic regression analysis, patients without typical shoulder and pelvic girdle pain, morning stiffness, or complete glucocorticoid response were especially at a higher risk of misdiagnosis.The presence of atypical features of PMR is a warning sign that should lead to a search for other conditions mimicking PMR, such as infections and malignancies (22).An imaging technique, such as positron emission tomography combined with computed tomography (PET/CT) evaluation, may be considered to identify possible underlying large-vessel vasculitis, especially in patients with marked involvement of the pelvic girdle or with inflammatory low back pain or bilateral diffuse pain in the lower extremities (23).
We also studied the usefulness of different PMR classification criteria in helping to identify patients in whom the risk of misdiagnosis is higher.The diagnosis was considered correct in 81.3% of patients fulfilling the 2012 ACR/ EULAR PMR classification criteria and in 45.5% of patients who did not fulfil these criteria.It should be noted that one of the requirements of correct PMR diagnosis in our study was the fulfilment of at least one set of classification criteria, and that no other diagnosis explained the patient's condition better during the follow-up.The fulfilment of the classification criteria indicates a more typical presentation of PMR, and the diagnosis of PMR changed less often for these patients compared to the patients with atypical presentation.The classification criteria were also often positive for those patients who were later reclassified as having some other disease than PMR.This highlights the fact that classification criteria are not diagnostic criteria (24) and should be applied only to patients in whom an alternative diagnosis responsible for the symptoms has already been excluded with reasonable certainty.
As is often the case with retrospective study designs, we were dependent on the clinician's records on the patient charts, and not all classification criteria items were available for all patients.When evaluating the fulfilment of classification criteria, an item was counted as negative if it was not marked as positive in the patient charts, but the absence of RF and/or anti-CCP antibodies had to be confirmed with a negative test result.We were also unable to group patients according to duration of the morning stiffness, since the exact duration was not routinely recorded in the charts by the clinicians.However, we believe that if the clinician specifically mentioned morning stiffness, it was markedly prolonged in most of these patients.
In Finland, the majority of patients with PMR are treated in primary healthcare, and only those patients with diagnostic uncertainty or poor treatment response are referred to rheumatological consultation.Thus, the patient material in a university hospital and in a department of rheumatology is more complex than for patients treated in primary healthcare, which may account for the relatively high proportion with normal ESR and CRP in this study.The results of our study are limited to a university hospital setting in Finland and are best generalizable to patients in large, centralized hospitals.Long follow-up periods and access to comprehensive medical records were strengths of our study.
Our study demonstrated the challenges of diagnosing PMR, especially when the presentation of PMR was atypical and the patient did not fulfil classification criteria.In this case, the diagnosis often changed during follow-up.Thorough consideration of differential diagnoses is always essential to minimize the risk of misdiagnosis, and this is particularly so with atypical presentation or suboptimal treatment response.

Conclusion
n a university hospital setting, one-third of initial diagnoses of PMR were changed during further evaluation and follow-up.Our findings highlight that thorough consideration of differential diagnosis is essential when diagnosing PMR, especially in patients with atypical presentation where there is a substantial risk of misdiagnosis.Most common differential diagnoses were other inflammatory arthritides, degenerative or stress-related musculoskeletal disorders, infections, and malignancies.

Figure 1 .
Figure 1.Study flowchart, including final diagnosis for patients with a diagnosis of polymyalgia rheumatica (PMR)

Figure 2 .
Figure 2. Proportion of patients with a correct diagnosis of polymyalgia rheumatica (PMR), among all patients and by fulfilment of classification criteria.+, Classification criteria fulfilled; −, classification criteria not fulfilled.

Table 1 .
Demographic and clinical characteristics of the study sample.

Table 2 .
Logistic regression analyses to predict the change in polymyalgia rheumatica (PMR) diagnosis during follow-up.