An Assessment of Comparative Medication Durability in Inflammatory Bowel Disease Patients With and Without Co-morbid Psoriasis, Rheumatoid Arthritis, and/or Enteropathic Arthritis

Abstract

Background

Patients with inflammatory bowel disease (IBD) are at increased risk for many co-morbid diseases. However, little is known about durability of IBD medications in patients with co-morbid diseases.

Aims

Determine medication durability in IBD patients with and without psoriasis, rheumatoid arthritis, and/or enteropathic arthropathy.

Methods

All patients with at least three ICD-9 or 10 diagnoses for IBD were included in the cohort. The risk factors of interest were a co-morbid diagnosis of psoriasis (IBD-Ps), rheumatoid arthritis (IBD-RA), and/or enteropathic arthritis (IBD-EA). Medication durability was defined as days of medication use, calculated using order start and stop dates from the electronic medical record. Significant differences were tested using the Wilcoxon rank sum test for continuous variables and Fisher’s exact test or Pearson’s Chi-squared test, as appropriate, for categorical variables. Boxplots were constructed for graphical interpretation of results.

Results

In the psoriasis group, there were 481 patients with 831 medication exposures [131 IBS-Ps (16%), 700 IBD only (84%)]. The median days of medication use were numerically higher in the IBD-Ps group for all therapies [anti-TNF: 1109 vs 861 (p = 0.17); anti-IL-12/23: 984 vs 834 (p = 0.33); JAKi: 682 vs 230 (p = 0.13)], anti-TNF/IM: 370 vs 202 (p = 0.57), except anti-integrin therapy [214 vs 470 (p = 0.08)]. When restricting to UC only, patients with co-morbid again Ps had a significantly shorter duration on anti-integrin therapy (84 vs 456 days, p = 0.02). While not reaching statistical significance, there was a distinctly longer medication duration on JAKi therapy (910 vs 317, p = 0.10). When restricting to patients with CD only, no results reached statistical significance though there was a trend towards longer anti-TNF durability in CD-Ps (1340 vs 1000 days, p = 0.098). There were no differences in medication durability in IBD-RA or IBD-EA patients.

Discussion

Larger studies investigating medication durability of JAKi and anti-integrin therapy in IBD patients with psoriasis would be beneficial given noteworthy trends towards increased and decreased durability, respectively.

Background

Immune-mediated disease occurs when the host immune system mounts an inappropriate immune response against otherwise healthy host tissues. One example of an immune-mediated disease is inflammatory bowel disease (IBD), which can manifest as inflammation in any part of the gastrointestinal tract. IBD patients are significantly more likely to have co-morbid autoimmune diseases, compared to non-IBD patients [1]. The presence of co-morbid autoimmune disease is important to consider when making medical management decisions as therapeutics common to both diseases are preferable. However, in many cases, there may be multiple shared therapeutic classes and it is unclear which class to use as first-line therapy.

Herein, we aimed to determine the durability of IBD medications in patients with or without co-morbid autoimmune diseases [including psoriasis (Ps) and rheumatoid arthritis (RA)] as well as enteropathic arthritis (EA). Multiple studies have confirmed the increased risk of psoriasis in patients with IBD [2–5]. The cause of this increased risk is unclear, but several studies have shown overlapping genetic [6–8], transcriptional [9, 10], and immunological [9, 11] features suggesting shared mechanisms of disease pathogenesis. Correspondingly, IBD and psoriasis share many similar therapeutic targets (i.e., anti-TNF, anti-IL-23). However, there remain distinct therapeutic differences as well, such as anti-IL-17 therapy in psoriasis and anti-integrin therapy in IBD. Ultimately, better data are needed to guide therapeutic decision making for patients with both diseases. The same principles apply to patients with IBD and RA. IBD and RA share genetic [12] and biologic overlap with common therapeutic agents such as anti-TNFs and JAK inhibitors. However, there are also distinct therapeutic mechanisms of action for each disease as is seen with psoriasis. Understanding how to position IBD therapies in patients with overlapping diseases is crucial to providing optimal care.

Herein, we aimed to understand the comparative medication durability of biologic and small molecule therapies in patients with and without co-morbid disease. Medications with longer medication durability are preferential to position as first-line therapy as they would lead to improved patient satisfaction and reduce the morbidity associated with numerous therapeutic switches. Therefore, we sought to report the medication durability of various IBD therapies in patients with and without co-morbid diseases of special interest (Ps, RA, and EA).

Methods

Study Design and Study Population

The study design was a retrospective cohort study including patients with a diagnosis of IBD enrolled in the Michigan Genomics Initiative at the University of Michigan. The risk factors of interest were a co-morbid history of psoriasis, rheumatoid arthritis, and/or enteropathic arthropathy. The study period reflects the entirety of a patient’s medical course, as documented by the electronic medical record (EMR), until the date of data extraction (2021).

The study population included patients with IBD defined as at least three encounters with an ICD-9 (555.x, 556.x) or ICD-10 (K50.00, K50.01x, K50.10, K50.11x, K50.80, K50.81x, K50.90, K50.91x, K51.00, K51.01x, K51.20, K51.21x, K51.30, K51.31x, K51.40, K51.41x, K51.50, K51.51x, K51.80, K51.81x, K51.90, K51.91x) diagnosis code for IBD. A co-morbid diagnosis of psoriasis was defined as at least one ICD-9 (696.1) or ICD-10 (L40.0, L40.1, L40.2, L40.3, L40.4, L40.8, L40.9) code for psoriasis. A co-morbid diagnosis of rheumatoid arthritis was defined as at least one ICD-9 (714.0, 714.2) or ICD-10 (M05.0, M05.1, M05.2, M05.3, M05.4, M05.5, M05.6, M05.7, M05.8, M05.9) code for rheumatoid arthritis. A comorbid diagnosis of enteropathic arthropathy was defined as at least one ICD-9 (713.1) or ICD-10 (M07.6) code for enteropathic arthropathy.

Outcomes of Interest and Covariates

The primary outcome of interest was medication durability, which was defined as days of medication use. Medication durability was calculated using medication order start and stop dates extracted from the electronic medical record. Medication courses of less than 30 days were excluded from analyses due to the inadequate time to measure effect of the therapy. In addition, any biologic or small molecule medication prescription which was ordered more than 400 days after the preceding order was considered a separate course of therapy and was excluded from analyses. Medications of interest included anti-TNF (infliximab, adalimumab, golimumab, certolizumab), anti-integrin (vedolizumab), anti-IL-12/23 (ustekinumab), and JAKi therapy (tofacitinib). As patients could be exposed to one or more medications, each medication exposure was counted as an event. Combination therapy was also evaluated, defined as an anti-TNF with an immune modulator for a duration of more than 90 days. Clinical data were extracted from the EMR in collaboration with the Data Office for Clinical and Translational Research. To limit confounding due to anti-TNF-induced psoriasis (rather than de novo psoriasis), medication courses where a patient’s last anti-TNF prescription was within 180 days of the psoriasis diagnosis were excluded from the psoriasis analyses. Covariates of interest included age, gender, race, and history of IBD-related surgery (as a surrogate for disease severity).

Statistical Analyses

Continuous variables are presented as median values ± IQR. Significant differences were tested using the Wilcoxon rank sum test. Categorical variables were presented as absolute numbers and percentages. Significant differences were tested using Fisher’s exact test or Pearson’s Chi-squared test, as appropriate. All analyses were completed in RStudio [13] with R version 4.1.0 using the tidyverse [14], gtsummary [15], and [16] packages. Boxplots were constructed for graphical interpretation. The study was completed in compliance with the University of Michigan Institutional Review Board, HUM00159951.

Data Availability

All data are incorporated into the article and its online supplementary material.

Results

There were 1135 individuals with IBD in the study cohort. After restriction to the medications of interest and exclusion of suspected anti-TNF-induced psoriasis events, there were 481 patients with 831 medication exposures. Of the 831 medication exposures, 131 occurred in IBD patients with Psoriasis (IBD-Ps) patients and 700 occurred in IBD-only patients (Table 1). Patients with IBD-Ps were younger (39 vs 44 yrs, p = 0.005), more often female (64% vs 48%, p < 0.001), and more racially diverse (p = 0.024) than IBD-only patients. Rates of IBD-related surgery were similar across groups (57% vs 60%, p = 0.6). Rates of medication exposure were different (p = 0.037), with higher rates of anti-TNF/IM (8.4% vs 7%), anti-IL-12/23 (17% vs 12%), and JAK inhibitor (5.3% vs 2.1%) use in IBD-Ps. The medication events included 538 exposures to anti-TNF therapy, 105 exposures to anti-IL-12/23 therapy, 106 exposures to anti-integrin therapy, 22 exposures to JAKi therapy, and 60 exposures to anti-TNF/IM therapy. The median days of medication exposure were higher in the IBD-Ps group for all classes of therapy, except anti-integrin therapy (Fig. 1a). For anti-TNF exposed patients, the median days of medication exposure were 1109 days in IBD-Ps compared to 861 days in IBD only (p = 0.17). For anti-IL-12/23-exposed patients, the median days of medication exposure were 984 in IBD-Ps compared to 834 days in IBD only (p = 0.33). In JAKi-exposed patients, the median days of medication exposure were 682 in IBD-Ps vs 230 days in IBD only (p = 0.13). In anti-TNF/IM-exposed patients, the median days of medication exposure were 370 in IBD-Ps vs 202 in IBD only (p = 0.57). For anti-integrin-exposed patients, the median days of medication exposure were 214 in IBD-Ps compared to 470 in IBD only (p = 0.084).

Table 1.

Demographic and clinical characteristics of IBD patients with or without psoriasis

Variable	IBD psoriasis (n = 131)	IBD only (n = 700)	p value
Age (yrs)	39 (32,51)	44 (34,58)	0.005
Gender			< 0.001
Male	47/131 (36%)	365/700 (52%)
Female	84/131 (64%)	335/700 (48%)
Race			0.024
African American	13/131 (10%)	64/700 (9.3%)
Asian	1/131 (0.8%)	6/700 (0.9%)
Caucasian	108/131 (84%)	615/700 (89%)
Other	6/131 (4.6%)	6/700 (0.9%)
Unknown	3/131 (2.3%)	9/700 (1.3%)
History of IBD-related surgery	75 (57%)	420 (60%)	0.6
Therapy class			0.037
Anti-TNF	81/131 (62%)	457/700 (65%)
Anti-IL-12/23	22/131 (17%)	83/700 (12%)
Anti-Integrin	10/131 (7.6%)	96/700 (14%)
JAK inhibitor	7/131 (5.3%)	15/700 (2.1%)
Anti-TNF + IM	11/131 (8.4%)	49/700 (7%)

Fig. 1.

a Days of medication use, by therapy class, in IBD patients with (IBD-Ps) and without (IBD-only) psoriasis. b Days of medication use, by therapy class, in CD patients with (CD-Ps) and without (CD only) psoriasis. c Days of medication use, by therapy class, in UC patients with (UC-Ps) and without (UC-only) psoriasis

Subset analyses, by IBD subtype (CD, UC), were performed in the psoriasis cohort. When restricting to patients with CD only, there were 355 patients with 632 medication exposures (412 anti-TNF, 98 anti-IL-12/23, 64 anti-integrin, 9 JAKi, 49 anti-TNF/IM). Of the 632 medication exposures, 112 were CD-Ps and 520 were CD only. Again, patients with CD-Ps were younger (39 vs 44 yrs., p < 0.001), more often female (64% vs 50%, p = 0.005), and more racially diverse (p = 0.006). Rates of IBD surgery were not significantly different (62% vs 63%, p = 0.7). Rates of medication exposure were different (p = 0.019), with CD-Ps more often exposed to anti-TNF/IM (9.8 vs 7.3%), anti-IL-12/23 therapy (20 vs 15%), and JAKi therapy (4.5% vs 0.8%). The median days of medication exposure were longer in the CD-Ps group across all classes of therapy [anti-TNF (1340 vs 1000, p = 0.098), anti-IL-12/23 (984 vs 856, p = 0.44), JAKi (450 vs 121, p = 0.41), anti-TNF/IM (370 vs 225, p = 0.76)], except anti-integrin therapy [286 vs 476, p = 0.46]. (Fig. 1b) When restricting to patients with UC only, there were 119 patients with 187 medication exposures (117 anti-TNF, 6 anti-IL-12/23, 40 anti-integrin, 13 JAKi, 11 anti-TNF/IM). Of the 187 medication exposures, 19 were UC-Ps and 168 were UC only. There were no significant differences in age (p = 0.5), race (> 0.9), gender (p = 0.074), or IBD-related surgery (p = 0.11). There were also no signifcant diferences in rates of medication exposure across groups (p = 0.4). UC patients with psoriasis exposed to anti-integrin therapy had signifcantly less durability (84 vs 456 days, p = 0.023), but the sample size for analysis was notably very small (UC-Ps = 2, UC only = 38). (Fig. 1c) There were no signifcant diferences in the median days of anti-TNF (389 vs 524, p = 0.42) or JAKi use (910 vs 317, p = 0.1). However, JAKi use was notably longer in UC-Ps than UC only. Anti-IL-12/23 and anti-TNF/IM use could not be compared given all exposures were UC only.

After restriction to the medications of interest (no restriction for “drug”-induced effect), there were 482 patients with 837 medication exposures for the IBD and rheumatoid arthritis dataset. Of the 837 medication exposures, 112 included a co-morbid diagnosis of rheumatoid arthritis (IBD-RA) and 725 did not (IBD only). IBD-RA patients were older (48 vs 43 yrs, p = 0.004), more likely to be female (74% vs 47%, p < 0.001), more racially diverse (p = 0.031), and were less like to have had an IBD-related surgery (43 vs 62%, p < 0.001). Medication classes were generally similar between the groups (p = 0.053), with the most notable difference occurring with JAK inhibitor therapy which was more likely to be prescribed in IBD-RA (7.1% vs 1.9%). There was no difference in medication durability in any of the therapy classes for IBD-RA vs IBD only [anti-TNF (p = 0.44); anti-IL-12/23 (p = 0.34), anti-integrin (0.29), JAKi (p = 0.56), anti-TNF-IM (p = 0.56)]. (Fig. 2) Comparatively, there were 167 events with a co-morbid diagnosis of enteropathic arthropathy (IBD-EA) and 670 without (IBD only). There was no difference in age (p = 0.6), medication class exposures (p = 0.2), or IBD-related surgery (p = 0.3) across the groups. The IBD-EA group, similar to other analyses, showed female predominance (70% vs 46%, p < 0.001) and increased racial diversity (p < 0.001). Like RA, there was no difference in medication durability in any of the therapy classes for IBD-EA vs IBD only [anti-TNF (p = 0.83), anti-IL-12/23 (p = 0.85), anti-integrin (0.91), JAKi (p = 0.38), anti-TNF-IM (p = 0.85)]. (Fig. 3).

Fig. 2.

Days of medication use, by therapy class, in IBD patients with (IBD-RA) and without (IBD-only) rheumatoid arthritis

Fig. 3.

Days of medication use, by therapy class, in IBD patients with (IBD-EA) and without (IBD-only) enteropathic arthritis

Discussion

With the increasing number of therapeutic options available for the treatment of IBD, it is becoming increasingly important to select the most effective and durable therapy upfront. Little data exist, however, to assist clinicians in making such decisions. In this study, we aimed to determine the comparative durability of several classes of therapy in IBD patients who did and did not have a co-morbid diagnosis of psoriasis, rheumatoid arthritis, and/or enteropathic arthritis.

The results of this study show that systemic immune suppressives, such as anti-TNF, anti-IL-12/23, and JAKi therapies, were all used for longer periods of time in IBD patients who had a concurrent diagnosis of psoriasis compared to IBD patients who did not. This was notably different than anti-integrin therapy, which was used for a shorter period in IBD patients with psoriasis. These results suggest that systemic immune suppressive therapies may be more durable than gut-targeted therapies in IBD patients with psoriasis. This conclusion is somewhat logical as guttargeted anti-integrin therapies are not expected to provide significant immune suppressive effects in other tissues, such as the skin. However, published case reports describing vedolizumab-induced efficacy for palmoplantar pustulosis and vedolizumab-induced development of psoriasis may suggest that this assumption is too simplistic [17, 18]. An alternative reason for the shorter duration could be patient and/or provider preference to switch to therapies that are approved to treat both conditions. Ultimately, larger studies with improved power and longer durations of assessment will be needed to determine if these trends are reproducible. Interestingly, the largest delta between IBD-Ps and IBD-only patients was with JAKi therapy, which was 682 days vs 230 days of medication use. While this change did not reach statistical significance, the sample size was rather small at 22 events. Thus, it is possible that a larger cohort size would reveal a statistically significant difference. JAKi is approved for the treatment of UC [19] and has shown promising results in the treatment of CD [20] and psoriasis. [21–24] Therefore, this may represent an important class of therapy to investigate further for preferential co-management of IBD and psoriasis. When evaluating medication durability by IBD subtype, the reduced medication durability with anti-integrin therapy and increased durability with JAKi therapy seemed to be largely driven by UC with similar, albeit weaker, trends in CD. Because subset analyses led to smaller sample sizes, it is difficult to draw strong conclusions about differences across biologic subtypes, and it would be beneficial to evaluate these trends in larger datasets.

There did not appear to be an incremental advantage to any class of therapy with RA or EA. Somewhat surprisingly, systemic immune suppressives were not consistently associated with longer durability as in psoriasis. For example, anti-TNF therapies had relatively equal durability in both analyses, anti-IL-12/23 therapy had relatively equal durability in EA analyses but less durability in RA analyses, and JAKi had improved durability in RA but reduced durability in EA. Ultimately, no effects reached statistical significance, and therefore, no definitive conclusions can be drawn regarding any preferential therapies.

There are several strengths of this work. First, we applied rigorous exclusion criteria to ensure high-quality medication data. Specifically, we excluded courses of less than 30 days as it would be unclear if medication had ever been started and would be impossible to infer durability. We also excluded repeat courses of the same therapy and events where psoriasis occurred within 180 days of anti-TNF exposure suggesting that the event may have been drug induced. Second, we utilized an outcome of “medication durability” to allow for more complete and uniform data acquisition. The use of a “disease response” variable is often difficult to assign as assessments are not standardized by time or measure (endoscopic, radiologic, or biochemical). This creates large amounts of missing data and the need to use composite outcomes for assessment, which can be limited by heterogeneity. By using length of medication exposure, we allowed for uniformity in the outcome measured, incurred the advantage of increased power with a continuous outcome, and can draw conclusions about the time of effectiveness which is an important and often unaccounted for clinical outcome in IBD research.

There are also limitations of this work. First, the smaller sample sizes for newer biologic and small molecule therapies may have limited the ability to detect small to moderate effect sizes. Furthermore, newer drugs have had longer approval times for co-morbid diseases, which may lead to longer estimates of durability due to differences in the approval timeframes rather than the patient population. Larger and longer duration studies would be beneficial to further interrogate the comparative durability of these drug classes. Second, medication prescriptions were extracted from the electronic medical records which would not have captured prescriptions before being seen at the institution or those written to be given outside the institution such as at a free-standing infusion center. Furthermore, if patients were mandated by their insurance to switch infusions to an outside infusion site, this would have been reflected as a therapy stop and underestimated the true medication durability. Third, medications could have been stopped for alternative indications such as patient choice, insurance issues, or side effects. Fourth, due to small sample sizes, sequential drug therapy and/or drug positioning could not be evaluated. This is an important outcome that would benefit from further investigation in larger studies. Finally, data on first contact with the health care system and entry date into the Michigan Genomics Initiative were not readily available. Therefore, time of entry into the cohort is not well established, which limits the ability to interpret duration of follow-up and timing of the exposure of interest.

In summary, IBD patients with psoriasis spend longer times on systemic immune suppressive therapies and shorter times on gut-directed therapy (anti-integrin). Therefore, preference should likely be given to systemic immune therapies in these patients when possible. While there were a very small number of events on JAKi therapy, there is a trend toward large deltas in medication durability such that this class would benefit from further investigation in larger cohorts.