Comparable Rates of Glucocorticoid-Associated Adverse Events in Patients With Polymyalgia Rheumatica and Comorbidities in the General Population

Abstract

Objective.

To investigate the use of glucocorticoids (GCs) and related adverse events (AEs) in a long-term, geographically defined cohort of patients with polymyalgia rheumatica (PMR).

Methods.

Using a population-based inception cohort, details of GC therapy were abstracted from medical records of all patients diagnosed with PMR in 2000–2014. Age- and sex-matched comparators without PMR were identified from the same underlying population. Cumulative and daily dosage of GC, rate of disease relapse, occurrence of GC-related AEs, and rate of GC discontinuation were analyzed.

Results.

The study included 359 patients with PMR and 359 comparators. The median time to taper below 5 mg/day for 6 months was 1.44 years (95% confidence interval [95% CI] 1.36–1.62), while the median time to permanent discontinuation was 5.95 years (95% CI 3.37–8.88). The mean ± SD cumulative dose of GC at 2 and 5 years was 4.0 ± 3.5 grams and 6.3 ± 9.8 grams, respectively. The mean ± SD daily dose of GC at 2 and 5 years was 6.1 ± 7.6 mg/day and 7.2 ± 9.5 mg/day, respectively. There were no differences in rates of AEs between patients with PMR and comparators for diabetes mellitus, hypertension, hyperlipidemia, or hip, vertebral, or Colles fractures (P > 0.2 for all). Cataracts were more common in patients with PMR than comparators (hazard ratio 1.72 [95% CI 1.23–2.41]).

Conclusion.

Relapse rates in PMR are highest in the early stages of therapy. Despite often protracted therapy, with the exception of cataracts, the rates of studied morbidities linked to GC are not more common in PMR than comparators.

Introduction

Systemic glucocorticoids (GCs), generally in the form of oral prednisone, are the most widely used and effective treatment for polymyalgia rheumatica (PMR) (1). Although usually effective, GCs may result in a variety of adverse events (AEs). These may include the development of cataracts, hypertension, and osteoporotic fractures (2). As a result, physicians endeavor to minimize the dosage of GCs prescribed for patients. Avoidance of GCs is almost unfeasible due to the lack of other effective drugs for PMR, and leaving the disease untreated can result in anemia, immobilizing pain, fatigue, and depression (1,3).

Clinical features forming the basis for PMR diagnosis and classification include the presence of bilateral shoulder and hip girdle pain, morning stiffness, and elevated markers of inflammation (4). As the disease progresses, flares of PMR are characterized by the exacerbation of musculoskeletal symptoms and the elevation of inflammatory markers like erythrocyte sedimentation rate (ESR) and/or C-reactive protein (CRP) level (5). While it is widely reported that the average length of treatment for PMR patients can range from 1–3 years, treatment is protracted in many patients (3).

Although there are other drugs used in the treatment of PMR, to date none have shown reduction in PMR disease activity comparable to GC (6,7). Methotrexate can be used as a GC-sparing agent in patients who are resistant to GC with best, but only modest, efficacy at the beginning of treatment (6,7). Anti–interleukin 6 therapies as possible GC-sparing agents are currently being explored (6). There is a clear need for improved therapies to control the disease symptoms and reduce the morbidity associated with PMR and current GC treatment. In pursuit of this goal, better information about GC burden and flare rates are needed. This study aims to provide essential data on GC dosage, length of treatment, and frequency of PMR relapses.

Materials and methods

Study population.

The population of Olmsted County, Minnesota, is well suited for investigation of the epidemiology of PMR because comprehensive unit medical records for all residents seeking medical care are available. The potential of this data system for population-based studies has been described (8). This system assures virtually complete ascertainment of all clinically recognized cases of PMR among Olmsted County residents.

The PMR cohort included individuals who first fulfilled 2012 European League Against Rheumatism/American College of Rheumatology classification criteria for PMR between January 1, 2000 and December 31, 2014 (4). Case ascertainment for this cohort was previously described (9). All cases were physician diagnosed. The presence of other diseases that could explain the symptoms, such as active rheumatoid arthritis, systemic lupus erythematosus, polymyositis, chronic infection, or multiple myeloma, was considered an exclusion criterion. A comparison cohort of subjects without PMR with similar age and sex, and indexed to calendar year of PMR diagnosis of the patients with PMR, was drawn randomly from the same population.

Data collection.

Patients in the PMR cohort were tracked retrospectively through their medical records until December 31, 2015, migration from Olmsted County, death, or a change in rheumatologic diagnosis or other diagnosis (e.g., treatment for cancer or other autoimmune disease) requiring protracted GC therapy. The tracking of all dosage data and complications was censored when PMR was reclassified into another rheumatic autoimmune disease, because GC dosage and complications were no longer reliably attributable to PMR. Additionally, 11 patients were removed because of prior history of giant cell arteritis (GCA), and an additional 5 who developed GCA after PMR were censored at GCA diagnosis.

Each medical record was reviewed to identify the dosage of GC used, as well as potential GC-related AEs. Data were obtained from inpatient/outpatient visits, phone conversations, and in some instances correspondences with local medical groups/hospitals. The GC dosage prescribed (in prednisone equivalent) was tracked from the start of GC therapy until the patient entered remission. Remission was defined as permanent discontinuation of GC without relapse. A relapse was defined as an increase in GC dosage due to symptoms, ESR/CRP level, or both. Short bursts of GC used for chronic obstructive pulmonary disease, surgery preparation, and/or any other nonautoimmune inflammatory conditions were not included.

GC-related AEs included diabetes mellitus, hypertension (requiring medical therapy), hyperlipidemia, cataracts, symptomatic lumbar and thoracic vertebral fractures, Colles fracture, intertrochanteric/femoral neck hip fractures, and avascular necrosis, which were manually abstracted from medical records for patients with PMR. The presence of these conditions was documented as either preexisting or newly developed. All complications were physician diagnosed. Preexisting conditions were only tracked for hypertension, hyperlipidemia, cataracts, and diabetes mellitus, as patients with these conditions were not at risk of developing them due to GC therapy. Hyperlipidemia was defined as either a total cholesterol level ≥200 mg/dl or a low-density lipoprotein cholesterol level ≥160 mg/dl. Hypertension was defined as systolic blood pressure level of ≥140 mm Hg and/or diastolic blood pressure of ≥90 mm Hg that required the use of antihypertensive medication. GC-related complications were also electronically retrieved from billing data for subjects in both the PMR and non-PMR cohorts.

Statistical analysis.

Descriptive statistics (means, percentages, etc.) were used to summarize the data. Patients could have multiple relapses during followup. Kaplan-Meier methods were used to estimate rates of GC discontinuation and achievement of <10 or <5 mg/day for 6 months. Cox models adjusted for age, sex, and calendar year were used to examine the association between initial dose of GC and GC discontinuation. Cumulative incidence methods adjusted for competing risk of death were used to estimate the rates of complications over time. Patients with preexisting diagnoses were excluded from the calculation of each complication rate. Comparisons of complication rates between cohorts were performed using Cox models adjusted for age, sex, and calendar year of PMR incidence/index date. Rates of relapse were analyzed using person-year methods. Differences in relapse rates according to age, sex, or disease duration were examined using quasi-Poisson regression models to account for overdispersion. Smoothing splines were used to depict the trends in relapse rates according to time since the diagnosis of PMR, to allow for nonlinear effects. Analyses were performed using SAS software, version 9.4, and R software, version 3.1.1.

Results

The cohort included 359 patients (63.8% female, median age 74.9 years). The median followup time was 5.8 years (total 2,236.4 person-years). The mean ± SD initial dose of prednisone was 16.9 ± 7.1 mg/day (Table 1).<<T1>> Seven patients refused GC therapy. At 1 year the mean ± SD dose was 5.9 ± 5.6 mg/day. The median dose increased was consistently approximately 5 mg/day over time among patients who continued to require GC therapy (Table 1). The median 1-year cumulative dosage of prednisone was 2.7 grams and at 10 years was 4.2 grams.

Table 1.

Distribution of current and cumulative prednisone-equivalent glucocorticoid (GC) dosage in patients with polymyalgia rheumatica

	Starting dose	1 year	2 years	5 years	10 years
Current dose, mg/day*
No.	352	229	130	57	16
Mean ± SD	16.9 ± 7.1	5.9 ± 5.6	6.1 ± 7.6	7.2 ± 9.5	9.1 ± 14.1
Median	15.0	5.0	4.4	5.0	5.0
Quartile 1, quartile 3	15.0, 20.0	2.8, 7.6	2.5, 7.0	3.0, 6.8	3.0, 9.1
Range	5.0–60.0	1.0–60.0	1.0–60.0	1.0–60.0	1.0–60.0
Cumulative dose, grams†
No.	359	334	302	201	78
Mean ± SD	0.0 ± 0.0	2.9 ± 1.8	4.0 ± 3.5	6.3 ± 9.8	9.4 ± 25.1
Median	0.0	2.7	3.4	3.9	4.2
Quartile 1, quartile 3	0.0, 0.0	1.9, 3.6	2.1, 5.4	2.1, 7.9	1.9, 8.4
Range	0.0–0.1	0.0–21.8	0.0–43.7	0.0–109.4	0.0–218.9

^*Among patients on GC at each time point.

^†Among all patients being followed, still under observation at each time point, including patients who are no longer on glucocorticoid therapy.

The median time required for the patients to permanently discontinue steroids was 5.95 years (95% confidence interval [95% CI] 3.37–8.88). Only 19% of patients discontinued GC by 1 year after PMR onset (95% CI 15–23), 37% (95% CI 31–42) discontinued by 2 years, and 50% (95% CI 44–55) discontinued by 5 years. Only 58% discontinued GC use by 10 years after the onset of disease (95% CI 51–64). The median time required to maintain a dose of <10 mg/day for 6 months was 0.81 years (95% CI 0.77–0.86) and to maintain <5 mg/day for 6 months was 1.44 years (95% CI 1.36–1.62). At 10 years of followup, at least 97% of patients reached a GC dose of <5 mg/day for 6 months (95% CI 86–99) (Figure 1, top panel).<<F1>> Initial dose of GC was not associated with time to permanent discontinuation (hazard ratio [HR] 1.06 per 5 mg/day increase [95% CI 0.96–1.18], adjusted for age, sex, and calendar year of PMR incidence). However, the initial dose of GC was significantly associated with a longer time to maintain a GC dose <10 mg/day for 6 months (HR 0.86 per 5 mg/day increase [95% CI 0.78–0.96]) or <5 mg/day for 6 months (HR 0.90 per 5 mg/day increase [95% CI 0.81–0.99]). Results for cumulative dose during the first 30 days were similar to those for the initial dose (data not shown).

Figure 1.

Patients’ times to dose reduction and relapse. yr = year; PMR = polymyalgia rheumatica.

The relapse rate was 31.5 per 100 person-years. The highest relapse rate occurred within the first 2 months of diagnosis (133.4 relapses per 100 person-years [95% CI 70.5–252.3]). From 2 months to 1 year after initiation of therapy, the relapse rate declined to 76.4 per 100 person-years (95% CI 52.2–111.9), in years 1–2 was 53.0 per 100 person-years (95% CI 34.4–81.9), and declined to 29.9 per 100 person-years (95% CI 16.2–55.3) in years 2–3. Sex had no influence on the relapse rate, but it varied by age. The highest relapse rates were seen in patients ages 60–69 and 70–79 years (40.0 and 40.2 per 100 person-years, respectively). Patients ages >80 years experienced a relapse rate of 25.8 per 100 person-years, and patients ages 50–59 years had a relapse rate of 34.2 per 100 person-years. Elevated ESR/CRP level alone does not account for many relapses (85 of 704 relapses); the majority of relapses were accompanied by typical musculoskeletal symptoms of PMR.

The non-PMR comparison cohort included 359 patients with similar age (median 75.0 years) and sex (63.8% female) to the PMR cohort. Manually abstracted and electronically retrieved complication data showed good agreement among the patients with PMR. Agreement ranged from 78% for hyperlipidemia to 99% for hip fractures (Table 2).<<T2>> As expected, disagreement was primarily due to additional complications identified electronically that were not verified by manual abstraction. These ranged from 1% for hip fractures to 21% for hyperlipidemia. For comparison purposes, the electronically retrieved diagnoses were used for both groups.

Table 2.

Glucocorticoid-related complication rates in patients with polymyalgia rheumatica (PMR) and comparators without PMR^*

	Electronic coding, %		No. of events§		Cumulative incidence (95% CI)
Complication	Agreement†	Excess events	Before	After	At 5 years, PMR	At 5 years, non-PMR	HR (95% CI)
Diabetes mellitus	81	18	93/104	38/39	9.5 (5.8–13.1)	11.5 (7.3–15.5)	0.85 (0.54–1.33)
Hypertension	85	13	255/253	42/47	30.4 (20.2–39.2)	33.4 (23.3–42.2)	0.79 (0.52–1.21)
Hyperlipidemia	78	21	276/268	28/30	24.0 (13.6–33.1)	26.6 (16.6–35.4)	1.03 (0.61–1.74)
Cataracts	85	10	203/197	77/66	41.0 (32.1–48.7)	27.3 (19.8–34.2)	1.72 (1.23–2.41)
Hip or femoral neck fracture	99	1	NA	21/25	3.9 (1.8–6.0)	5.1 (2.7–7.4)	0.80 (0.45–1.43)
Symptomatic vertebral fracture	90	3	NA	22/17	3.9 (1.8–6.0)	2.4 (0.7–4.0)	1.27 (0.67–2.39)
Colles fracture	99	3	NA	15/13	2.4 (0.7–4.0)	2.5 (0.8–4.2)	1.11 (0.53–2.33)
Any fracture	90	4	NA	49/49	9.5 (5.8–13.1)	11.5 (7.3–15.5)	0.85 (0.54–1.33)

^*95% CI = 95% confidence interval; HR = hazard ratio; NA = not applicable.

^†Agreement between electronic coding and manual abstraction.

^‡Excess events by coding.

^§PMR incidence/non-PMR index date.

The cumulative incidence of each serious complication at 5 years after PMR incidence/non-PMR index date was similar in both groups for all complications except for cataracts. The cumulative incidence of cataracts at 5 years after PMR incidence was 41.0% (95% CI 32.1–48.7) compared to 27.3% (95% CI 19.8–34.2) among non-PMR comparators. Thus, patients with PMR had higher rates of cataracts than comparators (hazard ratio 1.72 [95% CI 1.23–2.41], adjusted for age, sex, and calendar year). No patients developed avascular necrosis.

Discussion

GCs continue to be the mainstay of treatment for PMR. While effective, disease flares are frequent, GC therapy is protracted, and serious AEs of GC therapy are common. In this study, the median time to GC discontinuation was 5.95 years, while the time to reach <5 mg/day for 6 months was 1.44 years.

Although with shorter duration of followup, other PMR studies have reported relapse rates similar to what was observed here. Relapse rates ranging from 21–61% have been reported in studies using GC for the treatment of PMR (10). A British study also demonstrated that most relapses for PMR occur within the first year of treatment (11). The current study showed the rate of relapse dropping precipitously between the first and second year of treatment. The British study observed that approximately 54% of relapses occur during tapering of the GC dose. That study also reported that 61% of the PMR patients observed relapses after the first 2 months of treatment (11).

The current study considered any increase in GC dosage as a result of symptoms, elevated inflammatory markers, or both as a relapse, which may result in what seem to be higher relapse rates when compared to studies with more stringent definitions of relapse. The aforementioned study defined a relapse as a patient experiencing minimum pain and stiffness in the shoulder or hip girdle requiring increase in GC (11).

Despite the known risk of complications that can arise from the use of GC in these patients, little is known about the occurrence of serious AEs in patients treated for protracted periods of time. Other authors have reported similar complication rates for hyperlipidemia, diabetes mellitus, and hypertension (16.0%, 13.5%, and 51.1%, respectively) (1,6,12). The median followup time was 7.8 years, with an interquartile range of 3.3–12.4 (12). This result is similar to the complication rates determined by this study. A study conducted in Italy reported a fragility fracture rate of 26.6% in patients who were treated with GC for >2 years, similar to that of the current study (24.9% [95% CI 18.1–31.0]), despite methodologic differences (13).

Importantly, the risk of many GC-related AEs such as cardiovascular disease, osteoporotic fractures, and others is related to dose and duration and cannot be assumed to fall to baseline once GC use is discontinued. Comparator subjects had less risk for developing cataracts than patients with PMR, but rates of other comorbidities were similar. This similarity may result from the generally low GC dosages used. While a uniform, prospective treatment protocol is not used across all Olmsted County providers, initial GC doses were within the range recently endorsed by an international work group (14). The initial GC dose was not associated with time to permanent discontinuation, but higher initial doses were associated with a longer time to taper prednisone below <10 mg/day for 6 months. The current study has the strength of examining occurrence after initiation of GC therapy, using cumulative incidence methods to properly account for the variable amount of followup for each patient and the competing risk of death, rather than reporting raw percentages.

Strengths of the study include the complete data capture and ability to interrogate the individual medical record for the outcomes of interest over a long period of observation. Dosing information could be obtained using medical records from clinical visits that are more likely to reflect the actual use of GC than prescription data. The population of Olmsted County is largely of northern European ancestry, which has the highest incidences of PMR and so is representative of this target population (8).

Like all retrospective, observational studies, this study has inherent limitations. Collection of GC dosing data was based on clinical documentation; individual pill counts and daily reporting of GC use by the patients was not possible. In addition, electronic data were used to define serious complications, due to resource limitations. Thus, complication event rates may be slightly overestimated. However, this fact should have little impact on comparisons between patients with PMR and comparators. The high event rate for cataracts may be due to concerns about coexistent GCA that prompted eye exams soon after PMR diagnosis. Thus, the cataracts detected after PMR diagnosis quite possibly existed prior to GC initiation.

Our findings reveal much about the management of PMR with GC. Although the GC dose may be readily tapered to <5 mg/day prednisone equivalent, many patients nevertheless remain on GC therapy for long periods. While patients with GCA and PMR both reach <5 mg/day doses at roughly the same time, patients with PMR tend to have more difficulties discontinuing GC (14). Patients relapse frequently during the early stages of treatment. Serious complications from GC are common, but with the exception of cataracts, in long-term followup, rates of studied morbidities that may be linked to GC use did not appear to be higher than in the general population. Without testing for adrenal insufficiency, we cannot determine whether the prolonged GC use addresses the disease or GC withdrawal symptoms. Possibly the current PMR treatment paradigm is incorrect, and patients should be treated with higher GC doses early in the treatment regimen (6, 14). Whether subclinical large vessel vasculitis contributes to this protracted use is unclear. Improved therapies for PMR are needed to minimize dosage and related AEs.

SIGNIFICANCE & INNOVATIONS.

• A significant proportion of patients with polymyalgia rheumatic (PMR) require low-dose glucocorticoid (GC) therapy for many years to control symptoms of the disease. Disease flares are common, especially during the early stages of therapy.

• Patients on lengthy GC treatment experience a high rate of clinically significant GC-related complications, including hypertension, fracture, diabetes mellitus, hyperlipidemia, and cataracts. However, with the exception of cataracts, the rates of studied morbidities linked to GC are not more common in PMR than comparators.

• This study provides a benchmark for the management experience of PMR in routine clinical practice and for the development of future therapies, which are needed to improve treatment and reduce the burden of GC-related complications.