Association Between Steroid-Sparing Therapy and the Risk of Perianal Fistulizing Complications Among Young Patients With Crohn Disease

This comparative effectiveness analysis of commercial administrative claims uses propensity score matching to investigate the effectiveness of medical therapy for reducing risk of perianal fistulizing complications among young people with Crohn disease.


Introduction
Crohn disease (CD) is a chronic condition, with estimated prevalence in the Western world ranging from 250 to 1300 cases per 100 000 individuals. 1,2 It can cause severe destructive transmural intestinal inflammation, 3,4 which may create fistulas that penetrate through the bowel wall. Fistulas develop more commonly among those with childhood-onset CD (30%) vs adult-onset CD (15%-20%). [5][6][7][8] Fistulas provide a path for feculent material to invade adjacent tissues, causing septic complications or soiling of the skin via cutaneous openings. 5,9,10 The most common site for fistula development is in the perianal region, including perianal fistulas, perirectal abscesses, rectovaginal fistulas, and fistulas to the scrotum or labia, collectively referred to as perianal fistulizing complications (PFCs). 5,9,11 These complications can be severe, causing major negative consequences, including reduced quality of life (QOL), long-term detrimental health consequences, and increased health care expenditures. [11][12][13][14][15] Despite improvements in medical therapies, PFCs remain difficult to treat and commonly reoccur. 5,[16][17][18] About 70% of patients with PFCs eventually undergo surgical interventions, which often provide only temporary relief. 6,17,19 Despite all available medical and surgical interventions, 8% to 19% of patients with PFCs eventually undergo permanent diverting ostomy. 17,19,20 Prevention of PFC development through effective medical treatments could lead to better outcomes. However, the epidemiology and progression of PFCs are not well understood, and evidence for effective preventive strategies is lacking.
Steroid-sparing therapy (SST), including immunomodulators and anti-tumor necrosis factor α (anti-TNFα) medications, improves many CD outcomes, but it remains to be determined whether PFCs are preventable and, if so, which medical therapies are most efficacious. [21][22][23] We hypothesized that SST use would be associated with prevention of PFCs among patients who were free from such complications at CD diagnosis. Prevention of PFCs could lead to improved QOL and reduced health care use for approximately one-fourth of patients with CD. 2,[24][25][26] With that in mind, we designed a study to assess the effectiveness of SST for preventing PFCs among children and young adults with newly diagnosed CD. We chose this population because of their high incidence of PFCs. [5][6][7][8]

Study Population
This study identified patients aged 5 to 24 years when diagnosed as having CD between July 1, 2001, and June 30, 2014. Excluded were children younger than 5 years at diagnosis (n = 93) because CD with very early onset often manifests differently and is more likely to represent an underlying immune defect with CD-like characteristics, [28][29][30][31] and patients older than 24 years old at diagnosis (n = 8794) because pediatricians may provide care up to age 24 years. We considered the index date of CD diagnosis to be the first occurrence of any office-based evaluation and management, emergency consultation, or inpatient evaluation and management, consultation, or observation claim with a diagnosis of CD (eAppendix 1 in the Supplement). The dates of analysis were October 2017 to February 2020.
Confirmatory CD diagnoses were required with at least 3 CD claims within 2 years of the index date. 32,33 Patients were required to have at least 6 months of continuous insurance enrollment within any of the Optum included health plans before the index date (from January 1, 2001) and 2 years after the index date (through June 30, 2016) to capture treatments and outcomes. Patients who had ulcerative colitis diagnosis before CD diagnosis were excluded (n = 304) to avoid ambiguity of CD diagnosis. Because there was no pharmaceutical coverage indicator available in the data set, patients who did not have any prescription claims during their continuous enrollment period were considered as not having pharmaceutical coverage and were excluded from the study (n = 29).
Because anti-TNFα medications are contraindicated in patients with congestive heart failure, 34 patients with congestive heart failure before or at the time of CD diagnosis were excluded (n = 1).
Patients were also excluded if they had a claim with perianal or genital fistula or abscess diagnosis, underwent perianal fistula-related surgical procedures before or at the index date (n = 203), or had immunomodulator or anti-TNFα use (n = 155) 90 days before the index date (Figure 1).

Exposure
The exposure of interest was whether patients with CD initiated SST before development of PFCs.
Steroid-sparing therapy was defined as immunomodulators (thiopurines and methotrexate) and/or anti-TNFα medications (infliximab, adalimumab, and certolizumab pegol) and was identified using the National Drug Code and the Healthcare Common Procedure Coding System codes reported in pharmacy or medical claims (eAppendix 1 in the Supplement). Because immunomodulators require about 90 days for full effectiveness, patients were considered as having initiated immunomodulators if they began the immunomodulators during the period between 90 days before the index date and 90 days before either developing PFCs or the end of study period. In contrast, anti-TNFα can take effect almost immediately, so patients were considered as having initiated anti-TNFα therapy if they began receiving anti-TNFα during the period between 90 days before the index date and before either developing PFCs or the end of the study period. SST was categorized as immunomodulators alone, anti-TNFα alone, or anti-TNFα plus immunomodulators.

Outcomes
The primary outcome was PFC development within 2 years after the index date. Perianal fistulizing complication development was identified by previously validated claims-based case definition (perianal/genital fistula/abscess or seton/fistulotomy) using claims. 33 The earliest date of PFC diagnosis or perianal complication-related surgical procedure was considered as the PFC date. Time from CD diagnosis to PFC development was calculated from the index date to the PFC date or to the end of the study period if no PFC development. There is no validated means of assessing complexity or severity of PFCs in administrative claims data. Consequently, we used ostomies (colostomy, ileostomy, or enterostomy) as markers of PFC severity because patients with complex or recalcitrant PFCs commonly require diverting ostomy to facilitate healing. 19 Because ostomy can be performed for reasons other than PFCs, we limited the assessment to ostomy that first occurred in claims at or after PFC development but before the end of the 2-year study period.

Covariates
Patient-level covariates included sex, race/ethnicity (categorized as white, black, Hispanic, or Asian individuals or unknown), age at CD diagnosis, education level of the primary insurance policyholder (categorized as high school or less, college or higher, or unknown), household income (categorized as <$40 000, $40 000-$49 999, $50 000-$59 999, $60 000-$74 999, $75 000-$99 999, Ն$100 000, or unknown), geographic region, and year of CD diagnosis. Comorbid conditions (eg, arthritis and gastrointestinal bleeding) were included to adjust for all potential confounders; these were identified through International Classification of Diseases, Ninth Revision, Clinical Modification diagnosis codes using medical or inpatient claims before or at the index date. Other medications were identified through the National Drug Code or Healthcare Common Procedure Coding System codes using pharmacy, medical, or inpatient claims data (including antibiotics and systemic corticosteroids) taken before the initiation of SST or PFC development or the end of the study period, whichever occurred first.

Statistical Analysis
Descriptive analyses were conducted for all patients included in the study cohort. To assess the consequences of SST initiation in preventing PFC development within 2 years of the index date, propensity score matching was performed between patients who did or did not initiate SST to create a subcohort adjusted for all available potential confounders (eAppendix 2 in the Supplement).
Covariate balance was checked before and after propensity score matching with χ 2 tests. Crude 2-year rates of remaining free from PFCs after CD diagnosis were estimated for patients initiating and not initiating SST using the Kaplan-Meier method. Log-rank P < .05 was considered statistically

Study Cohort
The study identified 2214 patients aged 5 to 24 years who were diagnosed as having CD between   (Figure 2).

Patient Characteristics
After adjusting for sociodemographic and clinical characteristics (

Sensitivity Analyses
To  After propensity score matching, patients who received steroid-sparing therapy (SST) were less likely than those who did not (no SST) to develop PFCs (hazard ratio, 0.41; 95% CI, 0.33-0.52; P < .001). Numbers in parentheses represent PFC events. Note that SST may have been started at any time after Crohn disease (CD) diagnosis but before PFC development or the end of the study. Graphical assessment indicates proportionality assumption was met. 0.39-0.59; P < .001). Because patient insurance coverage after age 18 years may have changed, we performed additional sensitivity analysis limiting the population to patients 18 years or younger at CD diagnosis. After propensity score matching, 484 individuals were left in each treatment group. In this younger cohort, we similarly found that SST use was associated with a 50% reduction in risk of PFCs within 2 years (HR, 0.50; 95% CI, 0.37-0.69; P < .001).
After propensity score matching and adjusting for other covariates, an era association was found, with a 43% increased risk of PFC development in patients diagnosed as having CD in 2009 to 2014 compared with 2001 to 2005 (HR, 1.43; 95% CI, 1.08-1.88; P = .01). To address the possibility Looking back at other changes across these periods, there were more patients excluded because PFCs were found at or before CD diagnosis in the later years of the study. Over time, the proportion of patients with PFCs present at or before CD diagnosis increased from 6.3% (

Discussion
This study used a large, privately insured cohort to investigate the effectiveness of SST for reducing the risk of PFC development among young patients with CD. Consistent with prior pediatric evidence, 7 we found that almost 1 in 5 patients who were initially free from PFCs at CD diagnosis developed PFCs within the following 2 years. We noted that the use of SST was less common than guidelines recommend 36 ; however, using a propensity score-matched cohort, we found that introduction of SST was associated with a 59% lower risk of developing PFCs among young patients with newly diagnosed CD. We also found evidence that SST use may reduce the severity or complexity of PFCs if they do develop. Among patients who developed PFCs, a smaller proportion of those who had previously initiated SST subsequently underwent diverting ostomy compared with those who had not initiated SST before PFC development.
These are important findings because PFCs can be devastating, with long-lasting major negative consequences on QOL. 14,15 They can cause pain, feculent drainage, fecal incontinence, and dyspareunia and are often associated with negative body image and depression. 15,37,38 Perianal fistulizing complications are difficult to treat; they often require invasive surgical procedures but still commonly reoccur. 5,39,40 Despite optimal medical and/or surgical therapy, there remains a substantial risk of requiring a permanent ostomy. Therefore, evidence-based strategies for preventing PFC development among patients with CD are greatly needed.
Although ample evidence exists about the effectiveness of various medical therapies for treatment of PFCs because of CD, 16,40 this investigation is the first study, to our knowledge, that directly evaluates whether PFCs may be preventable. A small study 41 provided preliminary indications that perianal fistulas may be preventable. In that open randomized trial, combined immunosuppression was associated with improved outcomes compared with routine care (n = 133), and fewer patients developed perianal fistulas in the treatment group (not statistically significant).
However, in post hoc analysis of patients who underwent follow-up endoscopy (n = 46), fewer patients who achieved mucosal healing developed perianal fistulas, suggesting that PFCs may be preventable if CD is well treated. 42 Although mucosal healing is not captured in administrative claims data, our study similarly found that patients treated with SST (more effective than non-SST 21,22,43 ) and especially with combination therapy (more effective than either immunomodulators or anti-

Limitations
This study has important limitations to acknowledge. As with all retrospective studies, there is potential for confounding by indication. This risk can be mitigated by propensity score methods, but they cannot completely remove all potential confounders. Other important limitations relate to a lack of clinical information in administrative claims data. As a consequence, we were unable to assess CD location or endoscopic severity or whether patients or their family members were smokers.
Propensity score-matched analyses may minimize the risk of more smokers in one group than the other but cannot fully obviate this problem. Another potential limitation of this study is that we excluded patients without any recorded prescriptions because they may have had no pharmacy benefits. It is also possible that some patients who did not take any prescription medications but instead used diet or herbal therapies may have been mistakenly excluded for appearing to have no pharmacy benefits. We could not detect the consequences of these treatments with our study design. These may be important contributors or confounders that will require further examination in prospective studies. Despite these limitations, this study was rigorously conducted and presents the first compelling evidence to date that PFCs may be preventable with the use of SST.

Conclusions
Among young patients with CD, those who initiated SST were 59% less likely to develop PFCs than those who were not treated with SST in the 2 years after CD diagnosis. Furthermore, among those who developed PFCs, fewer SST users underwent ostomy than SST nonusers. These results indicate that PFCs may be preventable or less severe with effective medical therapy. We also found that the use of SST was lower than expected, being used by only slightly more than half of young patients with CD who did not have PFCs at diagnosis. The reason for this infrequent use of medications that are considered standard of care is unknown and requires further investigation. The study findings support existing guidelines, which recommend SST use for all patients with CD. 36,[47][48][49] Most important, we believe that these results provide an evidence base on which to develop strategies for prevention of PFCs in young patients with CD.