Lowering Expectations: Glucocorticoid Tapering Among Veterans With Rheumatoid Arthritis Achieving Low Disease Activity on Stable Biologic Therapy

Abstract

Objective

In the Steroid EliMination In Rheumatoid Arthritis (SEMIRA) trial, 65% of patients with rheumatoid arthritis (RA) in low disease activity (LDA) on stable biologic therapy successfully tapered glucocorticoids. We aimed to evaluate real‐world rates of glucocorticoid tapering among similar patients in the Veterans Affairs Rheumatoid Arthritis registry.

Methods

Within a multicenter, prospective RA cohort, we used registry data and linked pharmacy claims from 2003 to 2021 to identify chronic prednisone users achieving LDA after initiating a new biologic or targeted synthetic disease‐modifying antirheumatic drug (b/tsDMARD). We defined the index date as first LDA occurring 60 to 180 days after b/tsDMARD initiation. The primary outcome of successful tapering, assessed at day 180 after LDA, required a 30‐day averaged prednisone dose both less than or equal to 5mg/day and at least 50% lower than at the index date. The secondary outcome was discontinuation, defined as a prednisone dose of 0 mg/day at days 180 through 210. We used univariate statistics to compare patient characteristics by fulfillment of the primary outcome.

Results

We evaluated 100 b/tsDMARD courses among 95 patients. Fifty‐four courses resulted in successful tapering; 33 resulted in discontinuation. Positive rheumatoid factor, higher erythrocyte sedimentation rate, more background DMARDs, shorter time from b/tsDMARD initiation to LDA, and higher glucocorticoid dose 30 days before LDA were associated with greater likelihood of successful tapering.

Conclusion

In a real‐world RA cohort of chronic glucocorticoid users in LDA, half successfully tapered and a third discontinued prednisone within 6 months of initiating a new b/tsDMARD. Claims‐based algorithms of glucocorticoid tapering and discontinuation may be useful to evaluate predictors of tapering in administrative data sets.

SIGNIFICANCE & INNOVATIONS.

• In a real‐world rheumatoid arthritis population of chronic glucocorticoid users who reach low disease activity after starting a new biologic or targeted synthetic disease‐modifying antirheumatic drug (b/tsDMARD), 54% successfully tapered glucocorticoids and 33% discontinued them within 6 months.

• Although rates of discontinuation were predictably lower than seen in clinical trials, more than half of new b/tsDMARD initiations in this real‐world population resulted in at least a 50% reduction in glucocorticoid use to less than 5 mg/day.

• Characteristics associated with a greater likelihood of successful tapering included positive rheumatoid factor, higher enrollment erythrocyte sedimentation rate, greater number of conventional synthetic DMARDs at b/tsDMARD initiation, shorter time from b/tsDMARD initiation to index date, and higher average glucocorticoid dose over the 30 days prior to index date.

• Prespecified claims‐based algorithms for glucocorticoid tapering and discontinuation may be useful to evaluate predictors of successful tapering, yielding results that are more clinically applicable than those from a highly restricted clinical trial population.

INTRODUCTION

Up to 80% of patients with rheumatoid arthritis (RA) use glucocorticoids, with 30% to 60% of patients using them long term (1). RA treatment guidelines previously supported adding glucocorticoids when initiating disease‐modifying antirheumatic drug (DMARD) therapy, or in the case of DMARD failure (2). However, mounting evidence suggests even low‐dose glucocorticoids have unacceptably toxicity when used long term, and that difficulty tapering glucocorticoids is common after their initiation, resulting in potentially avoidable chronic use (3, 4, 5). In light of this, the American College of Rheumatology (ACR) now strongly recommends against use of long‐term glucocorticoids when initiating DMARDs for RA and conditionally recommends modifying DMARDs rather than continuing glucocorticoids to maintain low disease activity (LDA) (6).

In 2020, Burmester et al published the Steroid EliMination In RA (SEMIRA) trial, a large double‐blind multicenter randomized controlled trial (RCT) of glucocorticoid discontinuation in patients with RA (7). Unlike earlier clinical trials evaluating time‐limited, protocolized glucocorticoid use as a means to induce remission in early RA, SEMIRA was the first large RCT of glucocorticoid tapering to focus on patients prescribed stable doses for long‐term management of established RA (8). The SEMIRA protocol randomized 259 adult participants in stable LDA on tocilizumab and prednisone 5 mg/day to either continue prednisone for 24 weeks or taper to 0 mg/day by week 16. In the taper group, 65% of patients remained off glucocorticoids and in flare‐free LDA at week 24. However, it remains unclear how these observations compare with rates of discontinuation in real‐world patients in which greater comorbidity is observed and there is no prespecified strategy to guide glucocorticoid tapering. SEMIRA also did not examine whether the 35% of patients who could not discontinue glucocorticoids could instead tolerate a reduced dose.

Prior observational work on this topic is limited to relatively small international cohort studies and has produced variable results. In a single‐center Japanese cohort of 36 patients on stable RA treatment regimen that specifically included prednisolone, 58% were able to discontinue prednisolone within 1 year, but the mean dose among persistent users did not significantly change (9). In a secondary analysis from the Japanese Institute of Rheumatology, Rheumatoid Arthritis (IORRA) database, although use and dose of glucocorticoids decreased significantly in patients with established RA who initiated biologic DMARDs (bDMARDs), 51% remained on glucocorticoids 2 years later (10). In comparison, a subcohort of 97 patients with experience with DMARDs with early RA in the Chinese Treat to TARget in RA (TARRA) cohort had only a 23% cumulative probability of glucocorticoid discontinuation after 1 year (5). None of these study populations replicated SEMIRA inclusion criteria (ie, established RA with stable LDA on a bDMARD plus glucocorticoids) and none examined glucocorticoid dose reduction using predefined criteria.

We aimed to evaluate real‐world rates of glucocorticoid tapering among patients enrolled in the Veterans Affairs Rheumatoid Arthritis (VARA) registry. Compared with the results of the SEMIRA trial, we hypothesized that less than 65% patients would be able to successfully discontinue glucocorticoids given higher rates of comorbidity and less rigorous disease control in routine practice relative to a clinical trial population, but that a roughly equivalent number would be able to reduce their glucocorticoid dose without complete discontinuation.

PATIENTS AND METHODS

Population and data sources

This retrospective cohort study included participants in VARA, a longitudinal multicenter observational cohort study of US Veterans with RA fulfilling 1987 ACR classification criteria. VARA registry data were linked to national VA pharmacy claims data accessed through the VA Corporate Data Warehouse, as previously described (11). All participants provided written informed consent prior to registry enrollment, and each site maintains institutional review board approval.

Registry participants were selected for analysis if they 1) initiated a new single biologic or targeted synthetic DMARD (b/tsDMARD) (6) between May 2003 and July 2021; 2) had documented LDA, defined as a disease activity score in 28 joints (DAS‐28) of less than or equal to 3.2, at least 60 but no more than 180 days after b/tsDMARD initiation; 3) remained on continuous b/tsDMARD therapy for at least 210 days after documented LDA; and 4) met predetermined criteria for chronic glucocorticoid use at b/tsDMARD initiation (Figure 1). If a participant had multiple qualifying b/tsDMARD course, each course was included separately in the analysis.

Figure 1.

Schematic of study design. Biologic and targeted synthetic DMARDs included abatacept, adalimumab, anakinra, certolizumab, etanercept, golimumab, infliximab, rituximab, tocilizumab, baricitinib, upadacitinib, tofacitinib. b/tsDMARD, biologic or targeted synthetic DMARD; DMARD, disease‐modifying antirheumatic drug; GC, glucocorticoid; LDA, low disease activity.

We defined index date as the first recorded LDA between 60 and 180 days after b/tsDMARD initiation. We defined continuous b/tsDMARD therapy as sequential dispensing episodes of a single drug with no gaps of therapy longer than 90 days between the end of one episode and the beginning of the next (11). We operationalized chronic glucocorticoid use as 1) use of prednisone at least 5 mg/day on the day of b/tsDMARD initiation or averaged over the 30 previous days; 2) use of at least 5 mg/day prednisone at index date or averaged over the 30 days prior to index date; and 3) at least two prednisone dispensing episodes, each with days’ supply of at least 28 days. To focus on continuous rather than intermittent glucocorticoid use, we also required at least one prednisone dispensing episode to occur in the 90 days prior to b/tsDMARD initiation, and a second episode to occur 25 to 90 days after the first, within the period 10 days preinitiation to 90 days postinitiation of the b/tsDMARD. We excluded participants who received oral or IV glucocorticoids other than prednisone between −30 and 210 days from index date, with the exception of premedication given as part of a b/tsDMARD administration protocol.

For each prednisone dispensing episode, the prescribed dose for each day was imputed from the following pharmacy claims variables: dispensing episode start date, unit drug dose, days’ supply, and quantity of tablets dispensed. In cases in which multiple dispensing episodes covered a given day, the total daily dose was reported as the sum of the doses from all dispensing episodes involving that day. We calculated an average 30‐day prednisone dose as the sum of the daily prednisone doses over a 30‐day time period divided by 30.

Clinical variable assessment

As part of the VARA registry, treating rheumatologists collected ACR core measures as part of routine clinical care. These included erythrocyte sedimentation rate (ESR, mm/h), C‐reactive protein (CRP, mg/dl), 28‐joint swollen joint count, 28‐joint tender joint count, patient and provider global assessments (0‐100 mm visual analogue scale), and multidimensional health‐assessment questionnaire (12). The aforementioned variables were used to calculate the DAS‐28 (13). Additional variables collected at enrollment were self‐reported race, sex, and smoking status (current, former, never). Anticyclic citrullinated peptide (anti‐CCP, U/ml) antibody and rheumatoid factor (RF, IU/ml) were measured using banked serum collected at enrollment, as previously described (14). Linked pharmacy claims data were used to determine conventional synthetic DMARD (csDMARD) and b/tsDMARD use.

Outcomes and statistical analysis

Our primary outcome of successful glucocorticoid tapering, assessed at day 180 after index date, required both a 50% reduction in 30‐day averaged prednisone dose compared with index date, and a 30‐day averaged glucocorticoid dose of less than or equal to 5 mg/day. Our secondary outcome was discontinuation, defined as a daily prednisone dose of 0 mg/day between day 180 and day 210. We used univariate statistics to compare patient characteristics by fulfillment of the primary outcome.

RESULTS

Of 3216 VARA participants, 100 b/tsDMARD courses among 95 patients (3%) fulfilled study eligibility criteria (Figure 1). Characteristics of analyzed participants are summarized in Table 1, 2. Of these, 93% were male, 81% were white, 83% had a history of smoking, 81% had positive RF, and 82% had positive anti‐CCP antibodies. At b/tsDMARD start, mean (SD) age and RA duration were 64 (8.8) years and 14.4 (12.6) years, respectively, and mean prednisone dose was 11.0 (10.2) mg/day. The b/tsDMARD initiated was a tumor necrosis factor inhibitor in 62 courses (62%), tocilizumab in 15 courses (15%), abatacept in nine courses (9%), and rituximab in nine courses (9%). Common background csDMARDs at the time of b/tsDMARD initiation included methotrexate (40%), hydroxychloroquine (39%), sulfasalazine (14%) and leflunomide (14%). Forty‐six percent of participants were taking at least two csDMARDs at b/tsDMARD initiation (Table 2).

Table 1.

Data attrition table

	Patients	Courses
All VARA participants	3216	—
≥1 b/tsDMARD course between May 1, 2003, and July 31, 2021	1875	7190
b/tsDMARD course ≥180 days duration	1644	3640
Disease activity documented 60‐180 days after b/tsDMARD start	1152	2083
LDA documented 60‐180 days after b/tsDMARD start	571	922
b/tsDMARD course continues through day 210	522	650
≥5 mg/day prednisone equivalent either at b/tsDMARD start or averaged over the 30 previous days	202	218
≥1 glucocorticoid fill ≤90 days before b/tsDMARD start and ≥1 glucocorticoid fill from −10 days to +90 days after b/tsDMARD start, occurring 25‐90 days after a previous fill	164	154
≥5 mg/day prednisone either on date of LDA above, or averaged over the 30 previous days	95	100

Abbreviations: b/tsDMARD, biologic or targeted synthetic disease‐modifying antirheumatic drug; LDA, low disease activity; VARA, VA Rheumatoid Arthritis Registry.

Table 2.

Cohort characteristics stratified by achievement of primary outcome

	Overall cohort (N = 100)	Primary outcome met (n = 54)	Primary outcome not met (n = 46)	Met vs. not met P value
At enrollment
Male sex, n (%)	93 (93)	48 (88.8)	45 (97.8)	0.08
Years of education	13.1 (3.3)	12.9 (3.8)	13.4 (3.2)	0.46
RF positive, n (%)	81 (81)	48 (88.8)	33 (71.7)	0.02
CCP positive, n (%)	82 (82)	44 (81.4)	38 (82.6)	0.88
ESR	22.8 (19)	27.2 (20.2)	17.5 (16.1)	<0.01
CRP	2.3 (5.4)	2.7 (3.8)	1.9 (3.1)	0.44
DAS‐28	3.8 (1.6)	4.0 (1.7)	3.6 (1.5)	0.26
Race, n (%)
White non‐Hispanic	81 (81.0)	42 (77.8)	39 (84.8)	0.58
Black non‐Hispanic	12 (12.0)	7 (13.0)	5 (10.9)
Hispanic	6 (6.0)	4 (74.0)	2 (4.3)
American Indian/Pacific Islander	1 (1.0)	1 (1.9)	0 (0.0)
Smoking, n (%)
Current	26 (26.0)	11 (20.4)	15 (32.6)	0.91
Former	57 (57.0)	27 (50.0)	30 (65.2)
Never	17 (17.0)	8 (14.8)	9 (19.6)
At b/tsDMARD initiation
Age, y	64.0 (8.8)	63.0 (8.9)	65.1 (7.7)	0.23
RA duration, y	14.4 (12.6)	12.4 (11.3)	16.8 (13.8)	0.09
Mean prednisone dose past 30 days, mg/day	8.5 (6.6)	8.8 (6.8)	8.1 (6.4)	0.80
Type of b/tsDMARD initiated, n (%)
Abatacept	9 (9.0)	6 (11.1)	3 (6.5)	0.13
Anti‐TNF	62 (62.0)	36 (66.7)	26 (56.6)
Rituximab	9 (9.0)	2 (3.7)	7 (15.2)
Tocilizumab	15 (15.0)	9 (16.6)	6 (13.0)
Tofacitinib	5 (5.0)	1 (1.8)	4 (8.7)
Number of csDMARDs at b/tsDMARD start, n (%)
0	25 (25)	12 (22.2)	13 (28.2)	0.04
1	39 (39)	16 (29.6)	23 (50.0)
2	29 (29)	20 (37.0)	9 (19.5)
3	7 (7)	6 (11.1)	1 (2.1)
At index date (LDA)
ESR	13.3 (13.6)	14.6 (14.0)	11.7 (12.0)	0.31
CRP	0.8 (1.5)	0.9 (2.4)	0.8 (0.9)	0.76
MDHAQ	0.8 (0.6)	0.9 (0.5)	0.7 (0.5)	0.25
DAS‐28	2.3 (0.7)	2.3 (0.7)	2.3 (0.7)	0.91
Days from b/tsDMARD start to first LDA	130.9 (51.3)	122.7 (48.8)	140.5 (53.0)	0.08
Mean prednisone past 30 days, mg/day	8.0 (5.1)	9.1 (5.9)	6.7 (3.6)	0.02
At end of study period
Prednisone dose, mg/day	3.9 (4.8)	1.1 (2.9)	7.1 (4.9)	<0.01
Prednisone dose reduction, mg/day	−4.6 (10.4)	−8.8 (11.4)	0.2 (6.2)	<0.01

Note: Numbers presented are mean (SD) unless otherwise specified. Nonsignificant variables not presented included the following: year of b/tsDMARD start and type of csDMARD at b/tsDMARD start.

Abbreviations: b/tsDMARD, biologic or targeted synthetic disease‐modifying antirheumatic drug; CCP, cyclic citrullinated peptide; CRP, C‐reactive protein; csDMARD, conventional synthetic DMARD; DAS‐28, disease activity score in 28 joints; ESR, erythrocyte sedimentation rate; LDA, low disease activity; MDHAQ, multidimensional health‐assessment questionnaire; RA, rheumatoid arthritis; RF, rheumatoid factor; TNF, tumor necrosis factor.

The mean (SD) time from b/tsDMARD initiation to index date was 130.9 (51.3) days and median (IQR) was 123 (87) days. The mean (SD) prednisone dose at b/tsDMARD initiation, index date, and primary outcome assessment were 11.0 (10.2) mg/day, 8.5 (8.8) mg/day, and 3.9 (4.8) mg/day, respectively. Mean (SD) prednisone dose reduction between index date and day 180 was −4.6 (10.4) mg/day.

Of the courses fulfilling eligibility criteria, 54 (54%) met the primary outcome of successful glucocorticoid tapering. Of these 54, 33 (33% of total) also resulted in glucocorticoid discontinuation. In univariate analyses, characteristics associated with a greater likelihood of successful tapering included positive RF, higher enrollment ESR, greater number of csDMARDs at b/tsDMARD initiation, shorter time from b/tsDMARD initiation to index date, and higher average glucocorticoid dose over the 30 days prior to index date (all P ≤ 0.05). Characteristics associated with a greater likelihood of discontinuation were similar but also included younger age (P = 0.04).

DISCUSSION

In a real‐world population of chronic glucocorticoid users who achieve LDA after initiating a new b/tsDMARD for RA, 54% of courses achieved a prespecified, claims‐based definition of successful glucocorticoid tapering and 33% discontinued glucocorticoids, within 6 months. Although rates of glucocorticoid discontinuation in this real‐world population were predictably lower than seen in a recent clinical trial (7), the majority of patients studied were able to reduce their glucocorticoid use by 50% and to less than or equal to 5 mg/day.

Our study used prespecified, claims‐based definitions of both successful glucocorticoid tapering and glucocorticoid discontinuation. Rates of glucocorticoid discontinuation in our cohort were comparable with those seen in prior studies, allowing for differences in follow‐up time (5, 9, 10). We also found that, among the 67% of patients who did not discontinue glucocorticoids within 6 months, an additional one in three reduced their average glucocorticoid dose during that time period. The outcome of glucocorticoid tapering without discontinuation, although not well‐studied previously, is a clinically relevant metric of improved disease activity and is often used as a secondary endpoint in DMARD efficacy trials (15). Although the algorithms require further validation, our results preliminarily suggest that claims‐based measures may be useful to classify patients by their ability to taper glucocorticoids, allowing further study of factors associated with taper success (16).

In our cohort, use of more background csDMARDs at b/tsDMARD start was associated with successful glucocorticoid tapering, whereas glucocorticoid dose immediately prior to b/tsDMARD start and DAS‐28 at index date were not. Similar findings were observed in subgroup analyses from the SEMIRA trial and in DMARD‐experienced patients in small international cohorts (5, 7, 10). These results support the use of glucocorticoid‐sparing DMARD treatment to achieve and maintain LDA and facilitate successful glucocorticoid tapering (6). The lack of association between baseline glucocorticoid dose or DAS‐28 and successful tapering is supported by previous work suggesting that provider preferences or noninflammatory pain conditions, rather than disease activity, often drive glucocorticoid use in RA (3, 17). The observed association between higher enrollment ESR and successful glucocorticoid tapering may reflect confounding by age or comorbidities, especially as enrollment CRP and DAS‐28 do not show a similar relationship.

Our study reinforces the limited generalizability of traditional clinical trials to a real‐world population. Only 3% of VARA participants satisfied criteria for inclusion in the current study, a number comparable to the 3% to 7% of patients with RA in clinical practice who meet criteria for trials of treatment escalation (18, 19). In both trial types, exclusions were primarily due to missing disease activity documentation and disease activity measurements that exceeded trial inclusion cutoffs. These results emphasize the need for prospective studies of chronic glucocorticoid users with RA, both to identify potential candidates for glucocorticoid tapering and to develop generalizable, clinically feasible tapering guidelines.

There are additional limitations to this study. As discussed above, we included only participants who met rigorous inclusion criteria, similar to those used in clinical trials of steroid tapering. This reduces the generalizability of results to the larger population of steroid users. Although the demographics of VARA participants reflect those of the broader VA population, demographic differences between Veteran and civilian populations may also limit generalizability. Because of the small size of the data set after applying eligibility criteria, we did not have sufficient power to perform multivariable modeling to adjust for potential confounders. Further work is needed to evaluate whether the factors we identify here remain associated with successful glucocorticoid tapering in larger, multivariable analyses. Because VARA collects disease activity measures during clinical encounters occurring as part of routine clinical care, these measures were not routinely available at the time of b/tsDMARD initiation, nor were they performed at standardized intervals during or after index date. This limited our ability to evaluate the persistence of glucocorticoid tapering and discontinuation and how improved disease control after DMARD escalation may affect the eventual success of glucocorticoid tapering. Although our claims‐based algorithms produced results similar to prior studies, they have not been formally validated using clinical data, eg medical record review or patient‐reported surveys. Like the SEMIRA trial, our study does not examine glucocorticoid tapering among the 20% to 40% of patients who are able to reach LDA without a b/tsDMARD; such patients may have even greater success in tapering glucocorticoids than has been previously demonstrated. This analysis used both claims‐based data (drug dispensing) and data derived from the VARA registry (demographics, seropositivity, disease activity scores, including laboratory values, smoking status). Although prescription dispensing data within the VA is universally captured for this population, misclassification bias related to prescriptions filled in the civilian sector is possible, though unlikely (20).

We demonstrate that, in a population of chronic glucocorticoid users in LDA after escalating b/tsDMARD therapy, more than half tapered glucocorticoids by more than 50%, to a dose of less than or equal to 5 mg/day, within 6 months. One third of these patients were able to discontinue glucocorticoids within 6 months, a rate consistent with that seen previously in international RA cohorts (5, 9, 10) but lower than that seen in a large randomized clinical trial (7). We used prespecified claims‐based algorithms to define glucocorticoid tapering and discontinuation. Once further validated, such algorithms may be used to evaluate predictors of successful tapering in additional registry‐based and administrative data sets, yielding results that are more clinically applicable than those from a highly restricted clinical trial population. Future work will validate the definitions we used to assess glucocorticoid use and tapering, evaluate claims and registry‐based predictors of successful and unsuccessful glucocorticoid tapering, and examine tapering success among patients with RA who reach LDA without requiring a b/tsDMARD.

AUTHOR CONTRIBUTIONS

All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Cannon 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 conception and design

Wallace, England, Baker, Sauer, Cannon.

Acquisition of data

Wallace, England, Baker, Kunkel, Braaten, Petro, Mikuls, Cannon.

Analysis and interpretation of data

Wallace, England, Baker, Rojas, Sauer, Roul, Cannon.

Supporting information

Disclosure Form: