Trends in Medical Management of Moderately to Severely Active Ulcerative Colitis: A Nationwide Retrospective Analysis

Abstract

Background

With an increasing number of therapeutic options available for the management of ulcerative colitis (UC), the variability in treatment and prescribing patterns is not well known. While recent guidelines have provided updates on how these therapeutic options should be used, patterns of long-term use of these drugs over the past 2 decades remain unclear.

Methods

We analyzed a retrospective, nationwide cohort of more than 1.7 million prescriptions for trends in prescribing behaviors and to evaluate practices suggested in guidelines relating to ordering biologics, step-up therapy, and combination therapy. The primary outcome was 30-day steroid-free remission and secondary outcomes included hospitalization, cost, and additional steroid usage. A pipeline was created to identify cohorts of patients under active UC medical management grouped by prescribing strategies to evaluate comparative outcomes between strategies. Cox proportional hazards and multivariate regression models were utilized to assess postexposure outcomes and adjust for confounders.

Results

Among 6 major drug categories, we noted major baseline differences in patient characteristics at first exposure corresponding to disease activity. We noted earlier use of biologics in patient trajectories (762 days earlier relative to UC diagnosis, 2018 vs 2008; P < .001) and greater overall use of biologics over time (2.53× more in 2018 vs 2008; P < .00001) . Among biologic-naive patients, adalimumab was associated with slightly lower rates of remission compared with infliximab or vedolizumab (odds ratio, 0.92; P < .005). Comparisons of patients with early biologic initiation to patients who transitioned to biologics from 5-aminosalicylic acid suggest lower steroid consumption for early biologic initiation (-761 mg prednisone; P < .001). Combination thiopurine-biologic therapy was associated with higher odds of remission compared with biologic monotherapy (odds ratio, 1.36; P = .01).

Conclusions

As biologic drugs have become increasingly available for UC management, they have increasingly been used at earlier stages of disease management. Large-scale analyses of prescribing behaviors provide evidence supporting early use of biologics compared with step-up therapy and use of thiopurine and biologic combination therapy.

Key Messages.

What is already known? The landscape of medical management in ulcerative colitis care has changed dramatically as new therapeutic options have become available. Guidelines with limited evidence have been developed with the goal of improving patient outcomes and reducing variation in care.

What is new here? Trends in prescribing behaviors and patterns over time were identified in a large nationwide cohort.

How can this study help patient care? This study quantifies how the contexts where biologics have changed over time and provides evidence supporting several prescribing guidelines relating to biologic ordering, biologic initiation, and combination therapy.

Introduction

Ulcerative colitis (UC) is a widely prevalent inflammatory bowel disease that affects 1.5 million people in North America.¹ Estimates for average annual per-patient costs for UC range from US$6200 to US$11 400 per year, representing an annual burden of US$8 to US$14 billion.² While previously limited in therapeutic management, there have been significant developments in the medical management of UC as our understanding of the disease has progressed.^3-11

As a result, guidelines for initiating, optimizing, and monitoring response to existing therapies have undergone considerable evolution over the last 2 decades.¹⁰ Despite frequent changes in therapeutic options and practice guidelines, the actual prescribing patterns of these drugs in clinical practice are not well characterized. Three major clinical areas in which limited evidence exists to support prescribing practices include (1) standardized ordering of biologics, (2) the use of a step-up approach to therapy, and (3) the use of thiopurine and biologic combination therapy.^3,10

A step-up approach refers to the practice of prescribing less expensive medications before incrementally increasing to medications of greater cost.¹² The 2020 American Gastroenterological Association (AGA) guidelines recommend that, among outpatients with moderate-to-severe UC, the use of biologic agents early is preferable to a 5-aminosalicylic acid (5-ASA)–first step-up approach in most cases. However, given the absence of comparative effectiveness studies examining the step-up approach vs the early-biologic approach, this is a conditional recommendation.

Combination therapy with thiopurines and infliximab or other anti-tumor necrosis factor drugs is recommended by both the AGA 2020 guidelines and American College of Gastroenterology (ACG) 2019 guidelines over monotherapy. However, this recommendation is also based on limited evidence.^3,10

As the guidelines have changed, the impact on actual prescribing patterns in UC is not known. Previous retrospective observational analyses examining prescribing behaviors in the management of UC have been restricted by limited sample size^13,14 or did not associate exposures and outcomes in individual patients.¹⁵ One study utilized administrative records to analyze patterns in prescription drug sequencing in UC management but did not examine if behaviors changed over time and did not associate particular sequences with patient outcomes.^13,14 In this study, we utilize a nationwide dataset of more than 87 000 patients with UC and track drug transitions, steroid consumption, and hospitalizations at an individual level over time. This approach enables us to quantify population-level prescribing behaviors and examine comparative effectiveness of various prescribing patterns. The use of administrative data enables the observation of health system interactions for patients outside a single-provider system.

The purpose of this analysis is multifold: to characterize and describe temporal trends in the prescribing patterns of drugs for UC, understand which patients are receiving these drugs, examine the outcomes of these patients, and evaluate the clinical value of particular prescribing guidelines. We examine how 6 practices related to biologic ordering, step-up therapy, and combination therapy put forth in the AGA guidelines are associated with patient outcomes.

Methods

Study Population

The primary dataset for this study was a de-identified nationwide administrative database with records between 2008 and 2020 at Aetna, a commercial managed health care company. No race, ethnicity, or socioeconomic data were present in the database. Patients were selected based on 2 primary criteria: evidence of active UC management (based on the presence of at least 2 International Classification of Diseases–Ninth Revision or –Tenth Revision codes for UC) and evidence of available prescription records. Patients required 6 months of coverage prior to first observed biologic exposure. Observational analysis of this dataset has been validated for a wide range of phenotypes and outcomes in the past.^16-19

After inclusion, each patient was characterized as a timeline of consecutive, overlapping drug exposures. Annotations of the days’ supply of each prescription were utilized to concatenate consecutive courses of prescriptions and identify simultaneous courses. For each patient, the observation period was defined as the prescriptions ordered after the first UC diagnosis and the ending date, corresponding to the date of first colectomy or the last recorded observation. Within each observation period for each patient, each day was characterized by the set of active prescriptions the patient had, or if the patient was prescribed nothing at all. These timelines were used to assess cohort entry time points, calculate pre-exposure characteristics, and identify postexposure outcomes.

Analysis Pipeline

To compare various cohorts of patients, a generic data processing pipeline was constructed, summarized in Figure 1. This pipeline was used to define all considered cohorts, which were defined by a prescription or prescribing strategy. For a given drug category, the group of patients at first exposure was first identified. Exposures were characterized in the form of filters applied to each patient timeline. For example, in examining the comparative efficacy of adalimumab compared with infliximab or vedolizumab in biologic-naive patients, we identified the population of patients at first exposure to adalimumab who did not have prior exposure to any other biologics and compared their outcomes with the similarly identified infliximab-vedolizumab patients. For every analysis, exposure to high-dose steroids following a UC diagnosis was used as a proxy for the presence of moderately to severely active UC. The date of first exposure was treated as the cohort entry date for each subject and used to determine prior exposures and future outcomes. Prior exposures calculated include amount of prior steroid exposure (stratified by high or low dose), time since first steroid exposure, time since initial diagnosis, drug exposure by category, and combination therapy status, as assessed by intention to treat. Age, sex, and year of therapy were included as additional covariates. Standard definitions of remission were not available in the dataset. Therefore, a surrogate marker of remission was utilized corresponding to the first 30 consecutive steroid-free days observed following an exposure. We chose this method as many studies define maintenance of remission as a steroid free period.^20,21 For intercohort comparisons, remission and hospitalization were assessed using Cox proportional hazards models to identify if either cohort definition had greater odds ratios toward the outcome in question. Ordinary least squares multivariate linear regression models were utilized to assess intercohort differences in steroid-free days, costs, and future steroid consumption. All regression calculations were adjusted for steroid consumption in the 6 months prior to exposure, exposure year, sex, and age. The effect of exposure was assessed in the form of an odds ratio for time-varying outcomes (remission and hospitalization) and in the form of marginal impact for static outcomes (6-month postexposure steroid consumption, cost, and steroid-free days).

Figure 1.

Summary of study parameters and cohort definitions. 5ASA, 5-aminosalicylic acid; UC, ulcerative colitis;.

Ethical Considerations

The Harvard Medical School Institutional Review Board waived the requirement for approval, as it deemed this analysis to not be human subject research.

Results

Baseline Population Characteristics

Among all patients considered, 1 743 559 prescriptions were found included. We examined the pre-exposure characteristics of patients at first exposure to 5-ASA, thiopurines, adalimumab, infliximab, vedolizumab, and ustekinumab. We found segregation between each drug class based on typical indications for use into 3 tiers (Figure 2A) across all evaluated features. 5-ASA represented a first-line therapy, while vedolizumab or ustekinumab was typically preceded by 2 different prescriptions. Within the second-line therapies (thiopurines, infliximab, adalimumab), thiopurines distinguished themselves as being deployed in noticeably earlier points compared with the biologics (224 days earlier; P for thiopurines vs biologics as a group < .001). Furthermore, the majority of patients receiving thiopurine (58%) and biologic (55%, group probability) therapy have had 5-ASA prescriptions in the past, while patients on vedolizumab had similar likelihoods of being previously treated with of thiopurines (31%), adalimumab (29%), or infliximab (31%) (Figure 2B).

Figure 2.

Baseline characteristics of prescribing behaviors. A, Characteristics of patients at first exposure to each drug category. B, Heatmap of drug history at first exposure to each drug category. Values correspond to fraction of patients with history of a given drug at time of exposure. C, Drug share in patient-days over time. Each drug category is characterized by the fraction of prescribed patient-days within that month.

Temporal Trends in Prescribing Strategies

An examination of population-wide prescription frequency between 2008 and 2020 revealed that the overall diversity in prescription share has steadily increased with time (1.25× more, 2018 vs 2008, Shannon entropy ratio) as biologic drugs were introduced and steadily gained in popularity, while the shares of 5-ASA (59% to 51% of patient-days) and thiopurines (14% to 11%) have declined (Figure 2C). Despite the increase in use of biologic drugs, steroid use has remained stable until recently, when a small decrease in prescribing was seen.

While there were insufficient observations to examine year-by-year shifts in ustekinumab patients, they appear to follow the same overall trend. Patients exposed to ustekinumab prior to 2019 had substantially more prior drugs (3.12 ± 0.11 vs 1.81 ± 0.34; P = 3.67 × 10^-7) and steroids (4620 ± 653 mg vs 607 ± 307 mg; P = 6.09 × 10^-3) than patients exposed after 2019. Significant differences in steroid consumption between biologic-naive ustekinumab patients and biologic-naive infliximab or adalimumab patients were not observed (1944 ± 484 mg vs 607 ± 94 mg; P = .313).

Timing of Prescription Choice in Disease Course

Three drugs, adalimumab, vedolizumab, and ustekinumab, were approved for use in UC patients during our observation period (2012, 2014, and 2019, respectively). For adalimumab and vedolizumab, we therefore examined changes in prescribing patterns over time of medications including 5-ASA, thiopurines, and infliximab (Figure 3).

Figure 3.

Temporal changes in prescribing behaviors. A, Prior steroid exposure (milligram equivalents of prednisone) by first year of drug exposure. B, Time since first steroid (days) by first year of drug exposure. C, Time since initial ulcerative colitis (UC) diagnosis (days) by first year of drug exposure. D, Count of distinct prescription therapies prior to exposure to drug of interest.

We found that all drugs, with the exception of 5-ASA, tended to be prescribed at earlier points in patient trajectories at later dates. For both adalimumab and vedolizumab, trends toward earlier exposure began well before either drug was approved.

We further examined the drug profile of patients at first exposure to each biologic (Figure 4). We noted 3 general profiles based on drug class: (1) steroid and 5-ASA use increases in the lead-up to biologic initiation and falls off after, (2) thiopurine use remains constant relative to initiation and increases slightly postadalimumab or infliximab, and (3) other biologics are discontinued postinitiation. As an example, relative to adalimumab initiation, first, 5-ASA use increases from 18% of patients to 31% of patients in the 6 months prior to initiation, before dropping to 23% in the following 6 months, while high-dose steroids go from 5% to 19% to 5% over the same period. Second, thiopurine use rises from 6% to 12% after initiation, while, third, infliximab use drops from 8% to 0% within 60 days of initiation.

Figure 4.

Trends in drug ordering at first biologic exposure for A, adalimumab; B, infliximab; C, vedolizumab; and D, ustekinumab.

Evaluating Guideline Effectiveness

Using the developed cohorts, we next examined 6 individual clinical practice guidelines that were noted as having lower quality of evidence and certainty.¹⁰ For each guideline, we conducted a retrospective cohort study comparing each option.

Order of Biologic Prescriptions

The use of infliximab or vedolizumab was associated with faster remission, substantially lower costs, and more steroid-free days compared with the use of adalimumab (Figure 5A, Table 1). Both groups consumed comparable amounts of steroids during the 6 months following biologic exposure. This effect appeared to be driven by the comparison of adalimumab and infliximab (Supplementary Appendix). Among patients already exposed to infliximab and transitioning to another biologic, we did not observe significant differences in outcomes between those who progressed to ustekinumab and those who progressed to either adalimumab or vedolizumab (Figure 5B, Table 1).

Figure 5.

Partial effect plots for prescription pattern comparisons. Comparison of A, biologic-naive patients at first exposure (FE) with infliximab OR vedolizumab vs adalimumab; B, infliximab-exposed patients at FE with adalimumab OR vedolizumab vs ustekinumab; C, patients at biologic FE stratified by prior 5-aminosalicylic acid (5-ASA) exposure; D, 5-ASA–naive biologic patients with 5-ASA patients who progress to biologics; E, patients at biologic monotherapy FE with patients at biologic and thiopurine combination FE; and F, patients at FE thiopurine monotherapy with patients at thiopurine and biologic combination FE.

Table 1.

Results From Cox Proportional Hazards and Logistic Regression Models for Practice Guidelines

6-Month Outcomes	Biologic Ordering				Step-Up Therapy				Combination Therapy
	First Biologic Exposure (n = 11 096)		Post–Infliximab Exposure (n = 1803)		First Biologic Exposure (n = 6407)		Imputed Step-Up Initiation (n = 2421)		Biologic Exposure (n = 4451)		Thiopurine Exposure (n = 2569)
	Effect of Preferential Adalimumab Exposure	P Value	Effect of Ustekinumab Exposure	P Value	Effect of 5-ASA Exposure	P Value	Effect of 5-ASA Exposure	P Value	Effect of Thiopurine Combination Therapy	P Value	Effect of Biologic Combination Therapy	P Value
Remission, OR (95% CI)	0.92 (0.89 to 0.96)	<.005	0.90 (0.70 to 1.14)	.37	0.72 (0.68 to 0.77)	<.005	1.06 (0.97 to 1.15)	.18	1.47 (1.06 to 2.04)	.02	1.36 (1.09 to 1.70)	.01
Hospitalization, OR (95% CI)	1.01 (0.97 to 1.05)	.69	1.07 (0.84 to 1.35)	.59	0.95 (0.90 to 1.02)	.14	1.06 (0.98 to 1.15)	.16	1.21 (0.88 to 1.67)	.23	1.03 (0.82 to 1.29)	.8
Additional milligrams prednisone consumption (95% CI)	63 (-8 to 135)	.146	-125 (-617 to 366)	.675	232 (112 to 352)	.001	761 (618 to 904)	<.001	193 (-374 to 760)	.576	-601 (-1077 to -125)	.038
Additional costs (95% CI), $	-21 970 (-23 200 to 20 700)	.001	-8189 (-14 400 to 2014)	.029	-3113 (-5279 to -946)	.018	-7546 (-10 200 to -4907)	<.001	16 320 (5849 to 26 800)	.01	14 640 (9122 to 20 200)	<.001
Additional steroid-free days (95% CI)	-4.1	.001	4.3	.504	-2.3	.183	-21.3 (-25.1 to -17.4)	<.001	-7.6	.391	18.8	.003

Comparison of biologic-naive patients at FE with infliximab OR vedolizumab vs adalimumab; infliximab-exposed patients at FE with adalimumab OR vedolizumab vs ustekinumab; patients at biologic FE stratified by prior 5-ASA exposure; 5-ASA–naive biologic patients with 5-ASA patients who progress to biologics; patients at biologic monotherapy FE with patients at biologic and thiopurine combination FE; and patients at FE thiopurine monotherapy with patients at thiopurine and biologic combination FE.

Abbreviations: 5-ASA, 5-aminosalicylic acid; CI, confidence interval; FE, first exposure; OR, odds ratio.

Early Use of Biologics: Step-Up Therapy

In examining the efficacy of step-up therapy, key in our assessment was the assumption that physicians utilizing step-up therapy were doing so deliberately. As shown, 5-ASA is typically deployed at much earlier points in patient trajectories compared with biologics. Consequently, the patients experiencing step-up therapy do not represent typical 5-ASA patients; based on their physiology, they may have been plausible candidates for biologic induction. However, given dataset limitations, it is unknown if a subset of the patients had worsening disease during step-up therapy or not. To adjust for this, we conduct a 2-armed experiment to establish upper and lower bounds for the impact of step-up therapy (Supplement). We observed substantial differences between these populations in terms of remission, hospitalization, and steroid consumption, with step-up patients experiencing worse outcomes.

In examining the lower bound of the effect of step-up therapy (Figure 5C, Table 1), at the time of biologic initiation, we found that patients who had stepped through 5-ASA achieved remission at significantly lower rates and consumed slightly more steroids in the subsequent 6 months than patients with early biologic initiation. In examining the upper bound of the effect of step-up therapy (Figure 5D, Table 1), at the imputed time of step-up initiation, while both 5-ASA and biologic patients had similar rates of remission and hospitalization, the 5-ASA cohort required substantially more steroids and experienced fewer steroid-free days in order to achieve those outcomes. In both cases, the 5-ASA therapies were associated with significantly lower subsequent costs of care.

Combination Biologic and Thiopurine Therapy

The efficacy of combination biologic and thiopurine therapy was compared with both biologic (Figure 5E, Table 1) and thiopurine (Figure 5F, Table 1) monotherapy. The presence or absence of combination therapy was assessed using a previously validated intention-to-treat algorithm.²² In both cases, combination therapy was found to accelerate remission compared with monotherapy. In comparison with thiopurine monotherapy, combination therapy also resulted in subsequent steroid consumption (-601 mg; P = .038) and significantly more steroid-free days (18.8; P = .003), while no such differences were observed relative to biologic monotherapy.

Discussion

In this study, we identified several important trends in the medical management of moderate to severe UC. We also elucidated findings supporting some of the recent guidelines made by the AGA and ACG that were largely based on expert opinion with limited evidence. Overall, we found that there is an increased tendency to prescribe biologics for a more diverse group of patients with UC, as manifested by the use of biologics earlier in patients’ disease course and a wider array of patient profiles receiving these drugs over time. Finally, we highlighted trends in improved patient outcomes that support the use of early biologic therapy (vs 5-ASA–first step-up therapy) and combination therapy.

Our findings confirm prior studies showing that biologic use has increased over time.¹⁵ We also described a previously unreported trend toward earlier biologic use in a patient’s trajectory. Potential explanations for this trend include increased provider comfort with prescribing and monitoring of biologic therapy, increased awareness of more guidelines promoting the use of medications tailored to disease severity,³ and mounting evidence showing the efficacy of biologic therapy in a broader group of patients. Furthermore, we found that more types of patients are currently being considered for biologic therapy compared with years past. Patients with diverse drug history profiles are now getting biologics, and there is no one clear pathway leading to an individual biologic. This may reflect increased understanding of using these drugs in more clinical contexts and increasing evidence to support the efficacy of biologics. Strikingly, we found convergence of thiopurines and biologics in recent (post 2018) years, reinforcing the need for guidance regarding ordering or sequencing of prescriptions.

We uncovered important trends in prescribing patterns for corticosteroid and 5-ASA use as well. In the period just prior to biologic initiation, the use of corticosteroids significantly increases, then falls once again following biologic initiation and exhibits a downward trend (Figure 4). This is consistent with the results of recent studies, which report decreasing population-level trends in mean annual exposure to steroids and a decreasing proportion of patients receiving prescriptions for corticosteroids.^23,24 Our finding of ongoing corticosteroid use in the early phases of biologic initiation may reflect increased appropriate early corticosteroid use for induction of remission as biologic therapies take effect. Similarly, our finding of a downward trend in corticosteroid use after biologic initiation—in the context of population-level decreases in corticosteroid use over time^23,24—suggests that long-term inappropriate steroid use for these patients is indeed decreasing. This finding also suggests that our study uncovered inappropriate steroid prescriptions. A multicenter study examining steroid use in the United Kingdom found that 50% of steroid prescriptions were deemed to be avoidable or inappropriate.²⁵ We also found that, despite the increasing use of biologics to achieve remission, the prevalence of postbiologic 5-ASA prescriptions did not correspondingly decrease. Given the unnecessary nature of continuing 5-ASA therapy in this scenario and the potential side effects, this finding contributes to recent findings in favor of deprescribing.²⁶

Importantly, we found evidence to support current recommendations made by the AGA and ACG regarding step-up therapy and combination therapy. We found that early biologic therapy displays trends toward positive patient outcomes in comparison with step-up therapy. At the time of biologic initiation, patients who did not first step through 5-ASA achieved higher rates of remission and consumed less corticosteroids over the following 6 months. For patients who moved through step-up therapy, those on 5-ASA required more steroids and had fewer steroid-free days prior to achieving remission.

We also presented evidence to support the role of combination therapy with a biologic and thiopurine over monotherapy with either drug. Both the AGA and ACG make low-grade recommendations in favor of combination therapy owing to limited evidence; they also both support combining thiopurines with other biologics based on the purported mechanism for how these 2 drug classes interact with one other.^3,10 We found that, when thiopurines are combined with any biologic, there is a trend toward improved patient outcomes over biologic monotherapy. Of note, we also saw a concomitant rise in thiopurine use after biologic initiation. Finally, among patients who fail infliximab and require a transition to another biologic, we found that outcomes are similarly poor regardless of which subsequent biologic is initiated.

Our study has several limitations. First, although all patients in our analysis had prescription coverage, we were unable to ensure that coverage was comprehensive. Prescriptions or services received from outside our dataset would have been excluded. Second, owing to the administrative nature of our dataset, the resolution of the duration of each prescription was limited. We could not consider physician instructions that may have differed from each prescription claim or address prescription adherence. Dosing information was also unavailable due to incomplete coverage. Assessing the true impact of prescribing patterns like step-up therapy is limited, as we were unable to discern what the physician’s assessment was when deciding to prescribe one medication vs another. While these data allow us to approximate the impact of this behavior, our findings should prompt further investigation into outcomes associated with step-up therapy.

In summary, as the number of therapies approved for UC over the last several years has rapidly grown, several important trends have emerged in the medical management of UC. Biological therapies are being used earlier in patients’ disease courses and being prescribed for more moderate-to-severe disease, with a concomitant downward trend in corticosteroid use as a result. Early biologic initiation and combination therapy with thiopurines and biologics, 2 prescribing practices endorsed by recent guidelines but with limited prior evidence to support them, are supported by our findings and exhibit recent trends toward improved patient outcomes. Further study is needed to assess the direct impact of step-up therapy on patient outcomes and the costs associated with these prescribing practices.

Author Contributions

W.Y. contributed to conceptualization, formal analysis, investigation, methodology, software, validation, visualization, writing—original draft, and writing—review and editing. J.S.M. contributed to conceptualization, methodology, writing—original draft, and writing—review and editing. S.T.R. contributed to conceptualization, methodology, writing—original draft, and writing—review and editing. N.P.P. contributed to data curation and resources. I.S.K. contributed to resources, supervision, and writing—review and editing. D.T.R. contributed to conceptualization and writing—review and editing. G.A.B. contributed to conceptualization, methodology, resources, supervision, writing—original draft, and writing—review and editing. J.D.F. contributed to conceptualization, methodology, resources, supervision, writing—original draft, and writing—review and editing.

Funding

W.Y. is supported by T32HD040128 from the National Institute of Child and Human Development/National Institutes of Health. G.A.B. is supported by a Blavatnik Pilot Grant at Harvard Medical School.

Conflict of Interest

The authors disclose no conflicts.