Time Trends in Glucocorticoid Use in Rheumatoid Arthritis During the Biologics Era: 1999–2018

Abstract

Objectives:

To examine time trends in glucocorticoid (GC) use among patients diagnosed with rheumatoid arthritis (RA) during the biologic era.

Methods:

A population-based inception cohort of RA patients diagnosed during 1999 – 2018 was followed longitudinally through their medical records until death, migration or 12/31/2020. All patients fulfilled 1987 American College of Rheumatology classification criteria for RA. GC start and stop dates were collected along with dosages in prednisone equivalents. The cumulative incidence of GC initiation and discontinuation adjusted for the competing risk of death was estimated. Cox models adjusted for age and sex were used to compare trends between time periods.

Results:

The study population included 399 patients (71% female) diagnosed in 1999 – 2008 and 430 patients (67% female) diagnosed in 2009 – 2018. GC use was initiated within 6 months of meeting RA criteria in 67% of patients in 1999–2008 and 71% of patients in 2009–2018, corresponding to a 29% increase in hazard for initiation of GC in 2009–2018 (adjusted hazard ratio [HR]: 1.29; 95% confidence interval [CI]: 1.09–1.53). Among GC users, similar rates of GC discontinuation within 6 months after GC initiation were observed in patients with RA incidence in 1999 – 2008 and 2009 – 2018 (39.1% versus 42.9%, respectively), with no significant association in adjusted Cox models (HR: 1.11; 95% CI: 0.93–1.31).

Conclusion:

More patients are initiating GCs early in their disease course now compared to previously. The rates of GC discontinuation were similar, despite the availability of biologics.

INTRODUCTION

In the 1950s glucocorticoid (GCs) became the “wonder drug” for the management of patients with rheumatoid arthritis (RA).(1) They have been widely used ever since then due to being cost effective and having potent anti-inflammatory properties that aid in providing rapid symptomatic relief.(2) However, prolonged use of GCs has been linked to various adverse effects, including osteoporosis, fractures, gastrointestinal bleeding, cataracts, glaucoma, infections and cardiovascular disease.(3, 4) These adverse effects have led to more cautiousness in the long-term use of GCs.

Since the beginning of the 21st Century, several biologic agents (e.g., tumor necrosis factor alpha antagonists, interleukin 6 inhibitors, B-cell inhibitors, and T-cells inhibitors) have been shown to reduce symptoms, delay progression, and improve physical function and quality of life in patients with RA. These agents, along with an increased understanding of RA and the progression of the disease, have had a profound impact on the treatment of RA. There has been a large emphasis on the early recognition of disease, early initiation of treatment using a treat-to-target approach and aggressive control of disease activity through intensive use of conventional synthetic disease modifying anti-rheumatic drugs (DMARDs), biologic DMARDs and targeted synthetic DMARDs. GCs were first addressed in the 2015 American College of Rheumatology (ACR) treatment guidelines, which conditionally recommended low dose (≤10 mg/day) and short duration (<3 months) with an acknowledgment that there was a lack of evidence for these recommendations.(5) The 2021 guidelines strengthened these recommendations: 1) “Initiation of a conventional synthetic DMARD without short- term (<3 months) GCs is conditionally recommended over initiation of a conventional synthetic DMARD with short- term GCs for DMARD- naive patients with moderate- to- high disease activity”, and 2) “Initiation of a conventional synthetic DMARD without longer term (≥3 months) GCs is strongly recommended over initiation of a conventional synthetic DMARD with longer-term GCs for DMARD-naive patients with moderate-to-high disease activity”.(6) The 2022 EULAR recommendations modified the 2019 recommendations to clarify that short-term GC use is preferred and subsequent tapering of GCs should reduce the dosage of GCs to zero.(7)

Despite the advent of improved alternative therapies and guidelines recommending limiting use of glucocorticoids, it is unclear whether the use of GCs have changed in real-world clinical practice. In this study, we aimed to examine trends in GC use in a population-based inception cohort of patients with RA and to determine whether these have changed over the past 2 decades. We hypothesized that during the “biologic era” the use of GCs has declined due to the availability of new therapeutic options and increased awareness of the potential negative impact of GCs.

METHODS

Study Population.

This retrospective cohort study was conducted with the population of Olmsted County, Minnesota, using the resources of the Rochester Epidemiology Project.(8) This unique system ensures virtually complete ascertainment of all clinically recognized cases of RA among the residents of Olmsted County and includes details of all inpatient and outpatient encounters, including medication use and dosages, providing an opportunity to study therapeutic trends in a well-defined RA population. Using this resource, a population-based incidence cohort of patients diagnosed with RA between January 1, 1999, and December 31, 2018, among residents of Olmsted County, Minnesota and ≥18 years of age was assembled as previously described.(9) All cases fulfilled the 1987 American College of Rheumatology criteria (ACR) for RA and were followed from the date of criteria fulfilment longitudinally through their complete medical records until death, migration or 12/31/2020.(10) This study was approved by the Mayo Clinic (#17-002593) and Olmsted Medical Center Institutional Review Boards (#017-omc-17).

Data Collection.

RA disease characteristics assessed included rheumatoid factor (RF) positivity, anti- cyclic citrullinated peptide (anti-CCP) antibody, erythrocyte sedimentation rate (ESR) (at RA incidence and highest during first year after RA incidence) and radiographic erosions during the first year after RA incidence. DMARD use was defined as uninterrupted treatment lasting ≥30 days with a particular csDMARD (methotrexate, sulfasalazine, hydroxychloroquine, leflunomide) or biologics. All diagnostic codes from all healthcare providers were retrieved for up to 5 years prior to the RA incidence date. Using 2 codes ≥30 days apart and our previously published ICD-9/10 code lists, 55 morbidities of interest in patients with RA were defined.(11) Multimorbidity was defined as the presence of two or more morbidities and substantial multimorbidity was defined as the presence of five or more morbidities.

Definition of Glucocorticoid exposure.

All generic and trade names of GC medications were used to identify exposures documented in the medical records. Exposures from inpatient, outpatient, emergency department and nursing home encounters were ascertained. Use of oral, intravenous and intramuscular GCs, their doses in prednisone equivalents (categorized as 0–5 mg/day, 6–10 mg/day, 11–15 mg/day, 16–20 mg/day, or 21+ mg/day), dates of initiation and discontinuation, and duration of therapy were recorded. Intra-articular and inhalational GC use was not abstracted. Furthermore, some patients (<3%) that were documented with an as needed basis (PRN) for GC medication were also classified as continual use. For patients using GC for asthma or chronic obstructive pulmonary disease, data were only abstracted when the daily dose was 5 mg or more.

Data on RA disease characteristics, use of DMARDs and GCs were collected through medical record review by trained nurse abstractors according to a prespecified and pretested protocol.(2) A given dosage of GC was assumed to have continued until it was discontinued or changed. If the recorded dosage was higher or lower at a subsequent visit, then the date of this change was assumed to have been the date of that subsequent visit. GC discontinuation was defined as the first discontinuation of GC for ≥90 days. A second outcome of GC reduction to low dose defined as the earliest date of discontinuation or reduction to low dose (≤5mg/day) GCs was also examined among those who did not start on a low dose at GC initiation. For long intervals between physician visits, the subjects were assumed to have continued taking a given dosage of a GC if it was indicated as a current medication at both the current and former visits, with no documented plan of tapering or other alternation. Changes in the oral GC dosage that remained within a dosage category were not documented.

Statistical Analysis.

Descriptive statistics were used to summarize the demographic characteristics of the cohort by time period (1999 – 2008 vs 2009 –2018) with comparisons performed using chi square and rank-sum tests. Patient comorbidities at baseline were summarized according to time period and absolute standardized differences were presented. Standardized differences greater than 0.2 were considered meaningful, but no formal statistical comparisons were made. Cumulative incidence of GC initiation, discontinuation, and reduction was computed according to sex, time period, and age at RA incidence with adjustment for the competing risk of death. For graphical representation, inverse probability weighting was used to adjust for age, sex and time period to ensure comparability across groups. All patients were considered for the GC initiation outcome. Those who initiated GC prior to meeting RA criteria were considered as initiators at RA incidence (i.e., the day they met RA criteria). Each of the GC discontinuation outcomes was assessed among those who initiated GCs. The combined outcome of discontinuation or reduction was evaluated among the patients who were not on low dose GC at initiation. The current prevalence of GC use was estimated as the proportion of subjects on GCs from among those under observation during each month of RA duration, adjusted for mortality to avoid survival bias. Trends in prevalence were depicted using smoothing splines to allow for non-linear effects.

Cox models adjusted for age, sex and time of RA incidence period were used to examine the associations between patient characteristics and the three outcomes of GC initiation, GC discontinuation and GC reduction. When assessing the associations between comorbidities and discontinuation outcomes, comorbidities and multimorbidity status indicators were included as non-reversible time-dependent variables in the analysis such that the number of comorbidities was allowed to remain constant or increase during follow-up. To limit the false discovery rate, adjusted p-values were calculated using Benjamini-Hochberg’s linear step-up method when assessing individual comorbidities only.(12) Proportional hazards assumptions were checked using Schoenfeld residuals. To address a violation of the proportional hazards assumption in the analysis of time to GC discontinuation, logistic regression was used to assess the association between sex and odds of event occurring within 6 months of initiation, with those who died or were lost to follow-up before 6 months excluded from the analysis. Cox models were then used to estimate the multiplicative increase in hazard for event associated with sex in those who were under follow-up and had not experienced the event at 6 months. In secondary analyses, sex was included in the model as a stratification variable. Results from the Cox models and logistic models were summarized as hazard ratios and odds ratios respectively, with 95% confidence limits and p-values. Analyses were performed using SAS version 9.4 (SAS Institute, Cary, NC, USA) and R 4.0.3 (R Foundation for Statistical Computing).

RESULTS

The study population comprised 829 (69% female) patients; 399 (71% female) with incident RA in 1999 – 2008 and 430 (67% female) in 2009 – 2018, and median followup of 12.0 and 3.8 years, respectively (Table 1). Median age (55 and 57 years, respectively), and the proportion of patients who were rheumatoid factor and/or anti-CCP positive (66% and 62%, respectively) were similar between groups. There was a higher prevalence of DMARD initiation during the first year of RA incidence in the 2009 – 2018 compared to the 1999 – 2008 cohort (94% vs 86%; p<0.001). Patients in the 2009 – 2018 cohort tended to have more comorbidities compared to the 1999 – 2008 cohort (median [Q1, Q3] 4 [2, 7] versus 3 [1, 5]; p=0.003). Intramuscular and intravenous GC use was included in our analyses with oral GC use. Only 96 (12%) patients had at least one intramuscular or intravenous GC use. GC use for chronic conditions other than RA occurred in 56 (7%) patients, and a subset of these with GC use of ≥5mg/day were included in our analyses.

Table 1 –

Characteristics of 829 Patients with Incident Rheumatoid Arthritis in Olmsted County, Minnesota, in 1999–2018^*

	1999–2008 (N=399)	2009–2018 (N=430)	P Value
Age, years	55 (43, 66)	57 (45, 67)	0.30
Female sex	284 (71%)	289 (67%)	0.22
Rheumatoid factor and/or anti-CCP positive (n=828)	261 (66%)	268 (62%)	0.33
BMI, kg/m2	28.3 (24.1, 32.9)	28.5 (24.9, 34.0)	0.13
Current smoker	60 (15%)	74 (17%)	0.40
CRP, mg/L (n=734)	8.0 (2.9, 22.0)	6.8 (2.9, 18.7)	0.84
ESR, mm/hour (n=811)	18 (9, 31)	16 (7, 33)	0.42
Highest ESR, mm/hour (n=803)†	24.5 (12, 40.5)	22 (10, 40)	0.21
Erosions†	112 (28%)	101 (24%)	0.13
DMARDs†	343 (86%)	403 (94%)	<.001
Total number of comorbidities	A
Median (Q1, Q3)	3 (1, 5)	4 (2, 7)	0.003
0–1	107 (2,%)	104 (24%)
2–4	163 (41%)	140 (33%)
5+	129 (32%)	186 (43%)
Duration of followup, years	12.0 (8.9, 15.2)	3.8 (2.9, 5.4)	-

^*Continuous variables are summarized as median (Q1, Q3) and compared using Kruskal-Wallis rank-sum tests. Categorical variables are summarized as number (percentage) and compared using Pearson’s chi-square tests.

^†During the first year after RA incidence.

RA: rheumatoid arthritis; BMI = body mass index; anti-CCP= anti-cyclic citrullinated peptide; CRP= C-reactive protein; ESR = erythrocyte sedimentation rate; DMARDs = disease-modifying antirheumatic drugs.

The cumulative incidence of GC initiation was higher in the 2009 – 2018 cohort compared to the 1999 – 2008 cohort (Figure 1). Cox model results confirmed these findings (adjusted HR 1.29, 95% CI 1.09, 1.53) (Table 2). Among the 702 patients who initiated GC, similarly low rates of GC discontinuation within 6 months after GC initiation were observed in patients with RA incidence in 1999 – 2008 and 2009 – 2018 (39.1% versus 42.9%, respectively) (Figure 1). The rate of GC discontinuation for at least 90 days did not differ in the 2009 – 2018 cohort compared to the 1999 – 2008 cohort (adjusted HR 1.11, 95% CI 0.93, 1.31). The cumulative incidence of reduced dosage to ≤5 mg/day of prednisone equivalence (or complete discontinuation) for at least 90 days among the 623 patients who did not initiated GC at a low dose was also similar in the two time periods (Figure 1). Despite the higher rates of GC initiation in the 2009–2018 cohort, the prevalence of GC use was similar in the two cohorts with a slightly lower prevalence of GC use in the 2009–2018 cohort after the first year (Figure 2A). Even in the 2009–2018 cohort, a large proportion of patients with RA (52%) remained on GCs beyond 3 months, and 30% were still on GCs after 2 years.

Figure 1 –

Adjusted cumulative incidence estimates for GC initiation, discontinuation, and discontinuation or reduction to 0–5 mg/day according to date of RA incidence 2009–2018 (black line) versus 1999–2008 (grey line), sex: male (blue) versus female (red), and age at RA incidence: < 50 (green), 51–64 (brown), and 65+ (teal) years. Estimates are adjusted for age and sex using the method of inverse propensity weighting.

Table 2 –

Associations between patient characteristics and glucocorticoid (GC) initiation and discontinuation^*

Characteristics	GC initiation, N=829	GC discontinuation (90 days), N=702	Reduced dose (0–5 mg/day) for 90 days, N=623
	HR (95% CI)	HR (95% CI)	HR (95% CI)
RA incidence 2009–2018 (vs 1999–2008)	1.29 (1.09, 1.53)	1.11 (0.93, 1.31)	0.7 (0.83, 1.14)
Sex, female (vs male)	0.92 (0.77, 1.11)	0.83 (0.64, 1.07)§	1.24 (1.04, 1.47)
Age, per 10-year increase	1.06 (1.00, 1.13)	0.91 (0.86, 0.96)	1.04 (0.99, 1.10)
Rheumatoid factor and/or anti-CCP positive, n=828/701/622	1.06 (0.91, 1.24)	0.85 (0.72, 1.02)	0.98 (0.82, 1.16)
BMI per 5-kg/m2 increase	1.00 (0.95, 1.06)	0.99 (0.94, 1.05)	0.97 (0.91,1.02)
Current smoker	1.10 (0.90, 1.35)	0.65 (0.51, 0.82)	0.85 (0.68, 1.05)
CRP per 5 mg/L increase; n=734/626/559	1.02 (1.00, 1.03)	0.97 (0.95, 0.99)	0.98 (0.96, 0.99)
ESR per 10 mm/hour increase, n=811/691/613	1.07 (1.03, 1.11)	0.91 (0.87, 0.95)	0.95 (0.91,0.99)
Highest ESR per 10 mm/hour increase, n=215/288/91**	0.97 (0.86, 1.09)	0.94 (0.87, 1.00)	0.93 (0.84, 1.04)
Erosions, n=226/295/91**	1.11 (0.71, 1.71)	0.92 (0.64, 1.32)	0.77 (0.46, 1.29)
DMARDs, n=226/295/91**	1.23 (0.73, 2.09)	1.05 (0.57, 1.95)	1.23 (0.48, 3.15)
Multimorbidity (time-dependent)
0–1 comorbidities		Referent	Referent
2–4 comorbidities		1.09 (0.85, 1.40)	0.98 (0.77, 1.23)
5+ comorbidities		1.28 (0.99, 1.66)	1.12 (0.88, 1.42)

^*Results are from Cox models adjusted for time period, age, and sex unless noted and are presented as hazard ratios with 95% confidence intervals. Analyses for GC discontinuation were limited to patients who initiated GC during follow-up. Analyses for GC reduction were limited to subjects who initiated GC at a dose higher than 5 mg/day. Subjects with missing data were excluded from analysis of the given characteristic. When not all data were available, the number of subjects included in each analysis is provided.

^§To satisfy distributional assumptions, the association between sex and GC discontinuation was assessed beginning at 6 months after RA incidence date among those under follow-up who did not discontinue GC during the first 6 months (n=403).

^**During the first year of RA incidence. Analyses were limited to subjects under observation and event free at 1 year.

RA = rheumatoid arthritis; BMI = body mass index; anti-CCP= anti-cyclic citrullinated peptide; CRP= C-reactive protein; ESR = erythrocyte sedimentation rate; DMARDs = disease-modifying antirheumatic drugs; HR = Hazard ratio; CI = Confidence interval

Figure 2 –

Estimated prevalence of GC use following RA incidence according to (A) the date of RA incidence 1999–2008 (grey line) versus 2009–2018 (black line), (B) sex: male (blue) versus female (red), and (C) age at RA incidence: < 50 (green), 51–64 (brown), and 65+ (teal) years. Plotted lines are smoothers.

The Cox model estimated a 6% increase in GC initiation (HR 1.06, 95% CI 1.00, 1.13) and a 9% reduction in the GC discontinuation for every 10-year increase in patient age at GC initiation (HR 0.91, 95% CI 0.86, 0.96) (Table 2 and Figure 1). This resulted in a higher prevalence of GC use among older patients (age 65+ years) compared to younger patients (Figure 2C).

Initiation of GC did not differ between the sexes (Table 2 and Figure 1). However, female sex was associated with a higher rate of discontinuation within the first 6 months (adjusted OR 1.64, 95% CI 1.17, 2.31) but not after 12 months (HR 0.83, 95% CI 0.65, 1.07) (Figure 1). This results in a higher prevalence of GC use in males than in females in the first 12 months after RA incidence (Figure 2B).

In secondary analyses, after adjusting for age, sex, and calendar year of RA incidence, increased baseline ESR was associated with increased rate of GC initiation (adjusted HR 1.07 per 10 mm/hr increase, 95% CI 1.03, 1.11) and decreased rate of discontinuation (adjusted HR 0.91, 95% CI 0.87, 0.95) (Table 2). In those who initiated GC, patients who indicated current smoking at RA incidence were less likely to discontinue GC compared to non-smokers (adjusted HR 0.65, 95% CI 0.51, 0.82). There were no significant differences in GC initiation or discontinuation between those with RF and/or APCA positivity vs negativity. Patients with 5+ morbidities were somewhat more likely to discontinue GCs, but this association did not reach statistical significance (Table 2). When assessing the association between individual comorbid conditions and GC discontinuation, there were only 2 comorbidities with associations that were statistically significant after controlling the false discovery rate; obesity (adjusted HR 1.42, 95% CI 1.13, 1.78; 0.05<p<0.1), and headaches (adjusted HR 1.48, 95% CI 1.16, 1.88; 0.05<p<0.1) (Supplemental Table 1).

DISCUSSION

Our findings demonstrate that a higher proportion of RA patients are initiating GCs early in their disease course now compared to previously. Furthermore, current RA patients had no significant improvement in GC reduction or discontinuation compared to previously. In fact, a large portion of patients required long-term use of GCs and continued their GC for over 2 years. Older patients with RA were more likely to initiate GCs and less likely to discontinue GCs. There was no difference in the initiation of GCs by sex. However, females had a higher discontinuation rate within the first 12 months, afterwards discontinuation rates of GC were similar in females and males.

This study extends our previous report by Makol et al. on GC use in patients with RA in 1980–2007, which found more patients were starting GCs in 1995–2007 compared with 1980–1994 with no difference in GC discontinuation between the two time periods.(2) These trends are consistent with our extended findings for the 2009–2018 time period. Crossfield et al. found GC prescribing remained stable over 1998–2018 in a population of RA patients in the UK despite increases in DMARD usage.(13) Our study also noted increases in DMARD initiation rates over time. Laugesen et al. reported a decline in the prevalence of GC use in Denmark in 1999–2014, but this study was not limited to patients with RA and also included GC injections. (14) Changing goals of treatment including more early, aggressive treatment, the treat-to-target strategy and heightened expectations regarding pain control may be the key drivers of the higher rates of GC initiation in the recent years.

While we found that a substantial proportion of patients were using GCs longer than the 3-month short-term use mentioned in the ACR and EULAR guidelines, the optimal dose and duration of GCs in patients with RA remains controversial.(6, 7) Of course, the GC use recommendations were conditional in the 2015 ACR guidelines and the 2019 EULAR recommendations suggested tapering but did not clearly state discontinuation of GCs after short-term use was the goal. Furthermore, the time period of this study was prior to either of these recommendations. As the guidelines noted, 3 months may not be enough time for a patient to be on a GC for optimal bridging therapy. A minimum of 6 months has been suggested based on recent evidence from clinical trials.(1, 15) However, a recent systematic review noted only 10% of patients in clinical trials remained on GCs for more than 12 months.(16) Furthermore, a recent meta-analyses of clinical trial data found GC discontinuation after bridging was more successful than previously thought with lower dosing and shorter schedules associated with more successful discontinuation.(17) These findings contrast with our observational study where 30% of patients remained on GCs for at least 2 years.

Our study found that patients age 65+ years were often continued on a reduced dosage of GCs for longer periods of time. This finding was consistent with a recent report from a Canadian population.(18) Furthermore, a recent report from the Glucocorticoid Low-dose Outcome in RheumatoId Arthritis (GLORIA) randomized placebo-controlled trial provided strong evidence of the benefit of GC on disease activity and slowing of joint damage progression over 2 years in patients aged 65+ years, along with quantification of the trade-off regarding harms including infections and other adverse events.(19) This new evidence aligned with previous suggestions that elderly patients should be on lower GC doses due to the potential loss of muscle mass and plasma volume that occurs with aging.(20) However, there remain ongoing concerns regarding polypharmacy and possible interactions between GCs and other medications in elderly patients with multiple comorbidities.(21)

While GC initiation did not differ by sex, females were more likely than males to discontinue GCs during the first year after RA incidence. Reasons for this difference are unknown and may be due to differences in disease activity/severity or to greater concerns among women regarding side effects (e.g., weight gain), which were not accounted for in this analysis. Alternatively, this could be related to our previous findings of delays in meeting RA criteria for women with seronegative RA or unknown delays in seeking healthcare among men.(22)

The strengths of our study include its population-based design with standardized case ascertainment and use of a comprehensive medical record linkage system (in-patient and out-patient care from all local providers) providing an opportunity to study a large cohort of RA patients with long and complete follow up of all subjects. Study limitations include the retrospective, observational nature of this study with its inherent biases that relied on clinical information recorded in medical records. Patients prescribed GCs “as needed” were included as GC users, but this would have little impact on the results as it occurred in <3% of patients. Also, physician recommendations do not always directly translate into patient compliance, and data on patient adherence were not available. In addition, information on intra-articular use of GC was not collected in this study, but frequent and substantial intra-articular GC use would be rare. Finally, patterns of GC use may differ in other rheumatology practices, institutions, or geographic areas.

In conclusion, our findings demonstrate that more patients started GCs early in their disease course in 2009–2018 compared to the previous decade with no change in reduction or discontinuation rates over the “biologic era”. These trends likely reflect more early, aggressive treatment, which will hopefully improve long-term outcomes. However, the large proportion of patients who remained on GCs for more than 2 years (well beyond the 3 months recommended in the current ACR and EULAR guidelines) are worrisome due to the known negative effects of long-term GC use. Real world use of GCs in patients with RA is not currently aligned with management guidelines. Future studies should evaluate whether prolonged GC prescriptions have continued despite the 2019 guidelines and investigate the reasons for such practices.