Enteric Infection at Flare of Inflammatory Bowel Disease Impacts Outcomes at 2 Years

Abstract

Background

Outcomes of inflammatory bowel disease (IBD) following flare complicated by enteric infection (EI) are limited by follow-up duration and insufficient assessment of the role of non-Clostridioides difficile pathogens. We compared 2-year IBD outcomes following flare with and without EI.

Methods

We performed a retrospective cohort study of adults evaluated with stool PCR testing for IBD flare. Subjects were stratified by presence of EI at flare and were matched for age, sex, and date to those without EI. The primary outcome was a composite of steroid-dependent IBD, colectomy, and/or IBD therapy class change/dose escalation at 2 years. Additional analyses were performed by dividing the EI group into C. difficile infection (CDI) and non-CDI EI, and further subdividing non-CDI EI into E. coli subtypes and other non-CDI EI.

Results

We identified 137 matched subjects, of whom 62 (45%) had EI (40 [29%] CDI; 17 [12%] E. coli). Enteric infection at flare was independently associated with the primary outcome (adjusted odds ratio, 4.14; 95% confidence interval [CI], 1.62-11.5). After dividing EI into CDI and non-CDI EI, only CDI at flare was independently associated with the primary outcome (adjusted odds ratio, 4.04; 95% CI, 1.46-12.6). After separating E. coli subtypes from non-CDI EI, E. coli infection and CDI at flare were both independently associated with the primary outcome; other EI was not.

Conclusions

Enteric infection at flare—specifically with CDI—is associated with worse IBD outcomes at 2 years. The relationship between E. coli subtypes at flare and subsequent IBD outcomes requires further investigation.

Key Messages.

What is already known?

Enteric infection at IBD flare is associated with worsening of IBD up to 1 year after flare.

What is new here?

Enteric infection at flare is associated with worse IBD outcomes 2 years postflare. While C. difficile infection is implicated in this relationship, E. coli may be an additional pathogen of concern.

How can this study help patient care?

Our results may help further prognosticate IBD course following flare complicated by infection and suggest that further investigation of the impact of E. coli enteric infection at IBD flare on IBD clinical outcomes may be of value.

Introduction

The identification of an enteric infection (EI) at the time of an inflammatory bowel disease (IBD) flare is common, affecting over a quarter of patients with IBD relapse.¹  Clostridioides difficile infection (CDI) in particular is implicated in up to 16% of IBD flare episodes and is associated with higher morbidity, mortality, related healthcare costs, and longer length of stay than infection-free flare.^1–5 Although it is standard practice to evaluate for CDI during IBD relapse,⁶ multiplex gastrointestinal stool polymerase chain reaction (PCR) panels (GI panels) are increasingly utilized to evaluate for non-CDI EI. Further, a positive GI panel is associated with IBD disease activity⁷ and a lower likelihood of IBD therapy escalation at flare.^1,7–10 However, the contribution of non-CDI EI at flare to subsequent IBD outcomes—especially compared with CDI—is less clear. Current studies on this topic are few, limited by short-term follow-up and focus on inpatient outcomes such as hospitalization, emergency department (ED) visits, and emergent surgical intervention.^7,8,10 Whether non-CDI EI during flare is also associated with aspects of long-term IBD outcomes commonly seen in the outpatient setting, such as steroid-dependent disease or IBD therapy class or medication dose change, is not well-defined. Clarifying any such association would further inform the utility of testing for non-CDI EI at flare.

In this study, we aimed to compare 2-year IBD outcomes following IBD flare with and without enteric infection as detected by PCR-based stool assays.

Methods

Patient Selection

We performed a retrospective cohort study of the electronic medical records at NYU Langone Health, an urban quaternary care institution serving New York City as well as the tri-state area of New York, New Jersey, and Connecticut. Adults with a history of IBD who underwent inpatient or outpatient stool PCR testing with both a GI PCR panel and a C. difficile PCR test between 2013 and 2019 for suspected IBD flare were included in the study. Inflammatory bowel disease flare was defined as acute diarrhea prompting stool testing, endoscopy, and/or intensification of IBD therapy. Acute diarrhea—defined as 3 or more unformed stools in 24 hours for less than 14 days—was ascertained by provider chart documentation. The 2 stool PCR tests included in the study were the FilmArray GI pathogen panel (BioFire Diagnostics, Salt Lake City, UT) and C. difficile infection stool PCR assay (Xpert C. difficile, Cepheid, Sunnyvale, CA). The IBD diagnosis was evaluated using International Classification of Disease 10 coding¹¹ and confirmed by manual chart review. Subjects without acute diarrheal symptomatology were excluded from the analyses.

Variables and Definitions

The following demographic data were collected: age, sex, race, ethnicity, IBD subtype, Charlson’s comorbidity index,¹² tobacco use, body mass index, and prior CDI. Data on testing context including setting of test (outpatient, ED, inpatient) and markers of disease severity (albumin at testing, C-reactive protein [CRP] at testing, colitis on endoscopy, and colitis on cross-sectional imaging) were also collected. Inflammatory bowel disease therapy at time of flare was recorded and divided into the following classes: aminosalicylates, glucocorticoids, immunomodulators (6-mercaptopurine, azathioprine, methotrexate), antitumor necrosis factor (TNF), anti-interleukin (IL)-12/IL-23, anti-integrin, cyclosporine, and Janus kinase (JAK) inhibitors. Inflammatory bowel disease therapy escalation at the time of flare was recorded, including type(s) of new therapy class(es) initiated. Treatment with antibiotic therapy—defined as whether the subject was prescribed a full course of antibiotics in response to the stool test result—was also recorded. Enteric infection was defined as a positive result on stool PCR assay in a subject with acute diarrhea. Only the first stool assay test was included for each subject. To maintain uniformity among subjects, only the index case was used if a patient had multiple documented enteric infections.

Outcome and Statistical Analyses

The primary outcome was the occurrence of a composite outcome by 2 years postflare, comprising steroid-dependent disease, colectomy, and/or IBD therapy class change or dose escalation. Steroid-dependent disease was defined as one or more attempt(s) to wean the subject from oral steroid therapy resulting in flare symptoms requiring reinitiation of oral steroids¹³ or documented as such by the treating gastroenterologist. Inflammatory bowel disease therapy class change or dose escalation was noted in comparison to the IBD medication regimen immediately after resolution of the index flare episode. Secondary outcomes examined included the composite outcome by 1-year postflare, as well as each individual component of the composite outcome by 1- and 2-years postflare. To account for variation in timing of medical care, outcomes occurring up to 3 months beyond the 1- and 2-year mark postflare were included.

Patients were stratified by the presence of EI at flare. Subjects were matched by age, sex, and testing date within 1 year of those with negative EI testing at flare. Test date matching was performed to approximate similar IBD therapy practices between the comparison groups.

A univariate correlation analysis was performed with all variables and the presence of EI at flare. A logistic multivariate regression analysis with the primary outcome was then performed. Certain variables were included a priori: age, sex, race, Charlson’s comorbidity index, IBD subtype, location of testing, antibiotic therapy, and EI at flare. To account for confounding between a positive PCR stool test and antibiotic use, an interaction variable was created with both infection status and antibiotic use. The reference variable for this interaction variable was negative stool PCR testing and no antibiotic treatment. Those variables demonstrating statistical significance in the univariate analysis were added to the a priori variables in the multivariate analysis.

An additional multivariate analysis was performed by dividing the positive EI test group into CDI and non-CDI EI. The purpose of this analysis was to characterize the influences of CDI vs non-CDI EI on any association between overall positive EI testing at flare and the primary outcome. A second subanalysis was performed to similarly evaluate the most common non-CDI EI, E. coli subtypes, by further dividing the non-CDI EI group in a similar manner.

Categorical variables were analyzed using the χ² test, and continuous variables were analyzed using either the Mann-Whitney test or the ANOVA test and reported with median and interquartile range values. A P value <0.05 was considered statistically significant. All statistical analyses were conducted using R version 3.3.3.¹⁴

Results

Characteristics and Outcomes

Of 233 patients with IBD evaluated from 2013 and 2019 for flare, 63 (27%) had positive EI testing at flare. C. difficile was the most common pathogen, with 40 (63.5%) infected subjects (overall prevalence of 17%), followed by E. coli subtypes in 17 (27%) subjects. Forty-eight (76.2%) of 63 patients with positive EI testing received a course of antibiotic treatment. Of 15 subjects with positive EI testing not receiving antibiotic treatment, the most common reasons were patient declination, viral infection, and self-resolution of symptoms (Supplementary Table 1). One patient with positive EI testing had chronic diarrhea and was excluded from the study. After age, sex, and date matching, 137 patients were included in the study (62 with positive EI testing, and 74 with negative stool testing). Seventy-seven patients (56.2%) were female, and 63 (46%) had Crohn’s disease (CD), while 69 (50.4%) had ulcerative colitis (UC) (Table 1).

Table 1.

Baseline demographics.

	Flare Without EI (n = 74)	Flare With EI (n = 63)	P
Median age at flare, years (IQR)	41.0 (27.5-59.3)	39.2 (29.8-59.0)	.929
Female, n (%)	44 (59.5)	33 (52.4)	.405
Race, n (%)			.356
White	56 (75.7)	51 (81.0)
Black/African-American	8 (10.8)	2 (3.2)
Asian	2 (2.7)	3 (4.8)
Other	8 (10.8)	7 (11.1)
IBD subtype, n (%)			.579
Crohn’s disease	32 (43.2)	31 (49.2)
Ulcerative colitis	40 (54.1)	29 (46.0)
Indeterminate colitis	2 (2.7)	3 (4.8)
Tobacco use, n (%)	5 (6.8)	4 (6.3)	.647
BMI, median (IQR)	25.6 (21.6-29.2)	23.7 (21.8-28.1)	.934
Inpatient testing, n (%)	32 (43.2)	11 (17.5)	.001
Charlson’s Comorbidity Index, median (IQR)	0 (0-1)	0 (0-2)	.755
Low albumin (<3.5g/dl) at flare, n (%)	17 (23.0)	4 (6.3)	.016
Elevated CRP (>=10 mg/L) at Flare, n (%)	37 (50.0)	18 (28.6)	.022
Low albumin or elevated CRP at flare, n (%)	43 (58.1)	20 (31.7)	.006
Prior C. difficile infection, n (%)	0 (0)	0 (0)
CT imaging with colitis, n (%)	14 (18.9)	9 (14.3)	.649
Colitis on endoscopy, n (%)	36 (48.6)	27 (42.9)	.354
C. difficile positive, n (%)	0 (0)	40 (63.5)
Non-CDI EI positive, n (%)	0 (0)	27 (42.9)
IBD medications prior to flare
Oral 5-aminosalicylates, % (n)	36 (48.6)	23 (36.5)	.153
Oral steroids, % (n)	20 (27.0)	14 (22.2)	.516
Immunomodulators (6-MP, AZA, MTX), % (n)	14 (18.9)	7 (11.1)	.206
TNF inhibitor, % (n)	19 (25.7)	16 (25.4)	.970
Anti-IL-12/IL-23, % (n)	2 (2.7)	2 (3.2)	.870
Anti-integrin, % (n)	5 (6.8)	4 (6.3)	.924
Cyclosporine, % (n)	0 (0)	0 (0)
JAK inhibitor, % (n)	2 (2.7)	0 (0)	.189
IBD therapy escalation at flare, n (%)	46 (62.2)	21 (33.3)	<.001
Oral 5-aminosalicylates, % (n)	10 (13.5)	8 (12.7)	.888
Steroids, % (n)	33 (44.6)	14 (22.2)	.006
Immunomodulators (6-MP, AZA, MTX), % (n)	0 (0)	1 (1.6)	.277
TNF inhibitor, % (n)	8 (10.8)	1 (1.6)	.030
Anti-IL-12/IL-23, % (n)	1 (1.4)	0 (0)	.354
Anti-integrin, % (n)	3 (4.1)	1 (1.6)	.393
Cyclosporine, % (n)	1 (1.4)	0 (0)	.354
JAK inhibitor, % (n)	1 (1.4)	0 (0)	.354
Antibiotic Course at flare, n (%)	6 (8.1)	48 (76.2)	<.001
IBD Medications at 1 Year
Oral 5-aminosalicylates, % (n)	25 (33.8)	25 (39.7)	.475
Steroids, % (n)	10 (13.5)	15 (23.8)	.120
Immunomodulators (6-MP, AZA, MTX), % (n)	6 (8.1)	7 (11.1)	.550
TNF inhibitor, % (n)	19 (25.7)	21 (33.3)	.326
Anti-IL-12/IL-23, % (n)	12 (16.2)	10 (15.9)	.957
Anti-integrin, % (n)	13 (17.6)	5 (7.9)	.096
Cyclosporine, % (n)	0 (0)	0 (0)
JAK inhibitor, % (n)	5 (6.8)	1 (1.6)	.141
Colectomy Within 1 Year, n (%)	6 (8.1)	5 (7.9)	.971
IBD therapy dose escalation or class change at 1 year, n (%)	33 (44.6)	38 (60.3)	.067
IBD medications at 2 Years
Oral 5-aminosalicylates, % (n)	24 (32.4)	23 (36.5)	.617
Steroids, % (n)	12 (16.2)	13 (20.6)	.515
Immunomodulators (6-MP, AZA, MTX), % (n)	3 (4.1)	5 (7.9)	.334
TNF inhibitor, % (n)	20 (27.0)	19 (30.2)	.686
Anti-IL-12/IL-23, % (n)	15 (20.3)	12 (19.0)	.858
Anti-integrin, % (n)	8 (10.8)	6 (9.5)	.804
Cyclosporine, % (n)	0 (0)	0 (0)
JAK inhibitor, % (n)	5 (6.8)	2 (3.2)	.343
IBD therapy dose escalation or class change at 2 years, n (%)	38 (51.4)	41 (65.1)	.105
Colectomy within 2 years, n (%)	7 (9.5)	7 (11.1)	.750
Composite outcome at 1 year, n (%)	39 (52.7)	45 (71.4)	.025
Composite outcome at 2 years, n (%)	44 (59.5)	48 (76.2)	.038

In the univariate correlation analyses, proportions of IBD subtypes were similar between groups (Table 1). Subjects with negative stool testing were more likely to have a low albumin level (<3.5 mg/dL) or elevated CRP (≥10 mg/L) at flare (Table 1). There were no differences in colitis on endoscopy or cross-sectional imaging between groups. Subjects with negative EI testing were more likely to be hospitalized for the flare episode than subjects with positive EI testing (43.2% vs 17.5%; P = .001; Table 1). Additionally, those with positive EI testing at flare were less likely to have IBD therapy escalation than those testing negative (33.3% vs 62.2%; P < .001; Table 1; Figure 1); this was especially true for corticosteroid medications. The remainder of variables in the univariate analysis were comparable between groups (Table 1).

Figure 1.

Enteric infection at flare is associated with less initial IBD therapy escalation and more IBD therapy nonresponse.

In the univariate correlation analysis evaluating the primary outcome, patients with positive EI testing at flare were more likely to meet the composite outcome at 2 years than patients with negative stool testing at flare (76.2% vs 59.5%; P = .038; Table 1; Figure 1). Additionally, subjects with positive EI testing at flare were also more likely to meet the composite outcome at 1-year postflare than those with negative stool testing at flare (71.4% vs 52.7%; P = .025; Table 1; Figure 1). The remaining secondary outcomes were comparable between groups.

Multivariate Analyses

Two variables were added to the a priori variables in the multivariate analysis: new IBD therapy at time of flare and low albumin or elevated CRP at flare. After controlling for age, sex, race, Charlson’s comorbidity index, location of testing, IBD subtype, new IBD therapy at flare, and low albumin or elevated CRP at flare, the presence of any EI by stool PCR testing complicating flare was independently associated with a higher risk of meeting the composite outcome of steroid-dependent IBD, IBD therapy class or dose change, or colectomy at 2 years (adjusted odds ratio [aOR] 4.14; 95% confidence interval [CI], 1.62-11.5; Table 2). New IBD therapy at flare was also independently associated with the primary outcome (Table 2).

Table 2.

Multivariate regressions predicting composite outcome of steroid-dependent IBD, colectomy, or medication dose escalation/class change at 2 years when comparing infection-complicated flare to infection-free flare. Abbreviations: CDI (Clostridioides difficile infection); CI (confidence interval); CRP (C-reactive protein); IBD, inflammatory bowel disease; OR, odds ratio.

Variable	OR (95% CI)
Age	0.986 (0.241-16.3)
Male sex	0.985 (0.432-2.26)
Race
White	1
Black/African-American	0.545 (0.113-2.71)
Asian	0.333 (0.036-2.48)
Other	1.09 (0.238-6.20)
Charlson’s Comorbidity Index	0.961 (0.650-1.40)
Outpatient location	0.733 (0.205-2.57)
IBD subtype
Crohn’s disease	1
Ulcerative colitis	0.747 (0.304-1.80)
Indeterminate	2.01 (0.233-45.8)
Elevated CRP or low albumin at flare	0.830 (0.306-2.24)
New IBD therapy at flare	2.97 (1.07-8.98)a
Infection status and antibiotics
Negative stool PCR testing/ no antibiotics prescribed	1
Negative stool PCR testing/antibiotic course prescribed	1.29E7 (1.62E-32-NA)
Any positive PCR stool test/no antibiotic prescribed	2.80 (0.747-11.9)
Any Positive PCR Stool Test/Antibiotics Prescribed	4.14 (1.62-11.5)a

^aStatistically significant.

A second multivariate analysis was performed by subdividing patients with positive EI testing at flare into CDI and non-CDI EI. All other variables were kept the same. After controlling for age, sex, race, Charlson’s comorbidity index, location of testing, IBD subtype, new IBD therapy at flare, and elevated CRP or low albumin at flare, only subjects with treated CDI at the time of flare were more likely to meet the primary outcome (aOR, 4.04; 95% CI, 1.46-12.6; Table 3) compared with those with negative stool testing at flare who did not receive antibiotics. Non-CDI EI at flare—regardless of antibiotic treatment status—was not independently associated with the primary outcome compared with those with negative stool testing at flare (Table 3).

Table 3.

Multivariate regressions predicting composite outcome of steroid-dependent IBD, colectomy, or medication dose escalation/class change at 2 years when comparing CDI vs non-CDI vs infection-free flare. Abbreviations: CDI, Clostridioides difficile infection; CI, confidence interval; CRP, C-reactive protein; IBD, inflammatory bowel disease; OR, odds ratio.

Variable	OR (95% CI)
Age	0.985 (0.277-22.1)
Male Sex	0.878 (0.372-2.07)
Race
White	1
Black/African-American	0.601 (0.124-3.00)
Asian	0.455 (0.049-3.57)
Other	1.10 (0.239-6.23)
Charlson’s Comorbidity Index	0.998 (0.671-1.47)
Outpatient location	0.720 (0.201-2.53)
IBD Subtype
Crohn’s disease	1
Ulcerative colitis	0.731 (0.295-1.77)
Indeterminate	2.61 (0.280-64.5)
Elevated CRP or low albumin at flare	0.880 (0.326-2.37)
New IBD therapy at flare	2.67 (0.907-8.41)
Infection Status and Antibiotics
Negative stool PCR testing/no antibiotics prescribed	1
Negative stool PCR testing/antibiotic coursei prescribed	3.54E7 (5.30E-50-NA)
CDI PCR positive/no antibiotic prescribed	4.18E7 (1.36E-69-NA)
CDI PCR positive/antibiotic course prescribed	4.04 (1.46-12.6)
GI panel positive/no antibiotic prescribed	1.47 (0.303-7.26)
GI panel positive/antibiotic course prescribed	4.25 (0.914-25.1)

A subanalysis was performed further subdividing the positive EI test group into those with CDI, infection with E. coli subtypes, and any other EI. In the univariate analysis comparing by infection type, the composite outcomes at 1 and 2 years approached but did not cross the statistical significance threshold (Supplementary Figure 1). In a multivariate analysis controlling for the same demographic variables, treated E. coli infection and treated CDI were both independently associated with the primary composite outcome compared with negative EI testing at flare not receiving antibiotic therapy.

Discussion

In this study of IBD patients with flare, EI as detected by stool PCR testing—specifically CDI—was associated with subsequent steroid-dependent disease, colectomy, and IBD therapy class change or dose escalation at 2-years postflare. Non-CDI EI was not associated with adverse IBD outcomes at 2-years postflare, although E. coli subtypes at flare may be associated with worse IBD outcomes at 2 years.

Prior comparisons of IBD outcomes following flare with and without any EI by stool PCR assay have not demonstrated differences between groups in 90-day surgery, hospitalization rates, and mortality.^1,7 Our study showed sustained differences between positive and negative EI testing at flare in outpatient-focused IBD outcomes at 1 and 2 years. The prevalence of CDI in our study (17%) was otherwise comparable to the 13% to 14% seen in these studies.^1,7 Thus, we attribute this difference from prior data to differences in studied outcomes and a longer follow-up duration. The difference in serologic markers of disease severity at time of flare between groups raises the possibility that those with positive EI testing could have been undertreated for IBD at the time of flare. Although the association between the primary outcome and positive EI testing at flare remained after controlling for serologic markers of IBD severity and new IBD therapy at flare in the multivariate analysis, we were unable to determine the contribution of baseline IBD disease control, if any, to this relationship. Patients in our study seemed to have similar baseline disease control overall compared with current data; compared with a prior study on non-CDI EI in IBD flare, our study had similar rates of biologic exposure at testing (26% in this study; 18% in Axelrad et al) and steroid exposure at testing (25% in this study; 23% in Axelrad et al).⁸

Since antibiotic use is a close confounding variable to the presence of infection, a composite variable combining the 2 was utilized. Any EI at flare that was treated with antibiotics was independently associated with the primary outcome (Table 2), and CDI treated with antibiotics was independently associated with the primary outcome (Table 3). It was rare for patients with a negative PCR stool test to receive antibiotics and for patients with CDI to not receive antibiotic treatment, which led to a wide confidence interval for these groups (Tables 2-3).

The association between positive EI testing at flare and the primary outcome of this study appeared to be driven by CDI, rather than non-CDI EI. There are limited data comparing IBD disease course following flare complicated by CDI and non-CDI EI—and on IBD outcomes following non-CDI EI in general. In a retrospective study of 95 patients, flare with non-CDI EI was associated with longer time to subsequent flare over a 12-month period compared with CDI-complicated flare and flare testing negative for infection.¹⁰ While our study did not identify a protective relationship between non-CDI EI at flare and subsequent IBD disease activity, our results suggest that disease course following non-CDI EI at flare is likely at least comparable to that of negative EI testing at flare. This finding is similar to data from a 2018 study evaluating 163 non-CDI enteric infections in IBD flare by Axelrad and coauthors; at a median follow-up of 10.5 months, non-CDI EI at flare did not correlate with numerous IBD outcomes compared with negative EI testing at flare.⁸ Our findings extend the duration of their conclusions and suggest that negative EI testing at flare and flare with non-CDI EI may have comparable outcomes at 2 years.

While non-CDI EI at flare may not herald a worse IBD disease course, our findings do suggest an association between E. coli subtype infection at flare and subsequent IBD activity. The E. coli subtypes were the most common non-CDI infection in our study, which is comparable to prior data.^1,7,15 In the multivariate analysis separating the positive EI test group into CDI, infection with E. coli subtypes, and other infection, E. coli infection was independently associated with the primary outcome compared with those with negative EI testing—even after controlling for demographics, testing location, serologic markers of inflammation, and whether new IBD therapy was initiated at flare. E. coli colonization is already implicated in many aspects of IBD.¹⁶ Although limited by sample size, our findings suggest that E. coli may be a distinct infectious pathogen to evaluate for during IBD relapse. This is further supported by data from a recent study that found a higher proportion of enteropathogenic E. coli by stool PCR assay in patients with active CD compared with quiescent CD (7.1% vs 1.9%).⁷ However, not all studies are consistent with this association: presence of E. coli on stool PCR assay at IBD flare was associated with a less severe disease course in a separate study.¹⁷ The association between E. coli subtype infection at flare and subsequent IBD disease course—and whether E. coli infection at flare influences disease course or is simply a marker of disease severity—needs to be examined further in a prospective manner and with a larger study population.

The relationship between CDI during IBD flare and subsequent worse IBD outcomes is well-established.^1,3,4,18 However, current studies are limited by a 12-month duration of follow-up. In this study, we demonstrate a higher likelihood of adverse IBD outcomes up to 2 years after a flare complicated by CDI. Our findings further underscore the importance of testing for CDI in IBD patients with flare symptoms to assist with long-term risk stratification. Simultaneously, the role of testing for non-CDI EI remains unclear. Even when controlling for IBD therapy escalation at flare, non-CDI EI—except for E. coli infection—was not associated with worse IBD outcomes at 2 years in our study. Our findings suggest that while infection with an E. coli subtype may be an exception, the majority of non-CDI EI may not portend poor long-term IBD disease control and may not require diagnosis in all IBD relapse. Prospective data are needed to further characterize the role of non-CDI EI in subsequent IBD outcomes and, consequently, the utility of GI panel testing in IBD flare.

This retrospective study was limited by sample size; it may have consequently been underpowered for the individual components of the primary outcome, as only the composite outcome was statistically significant. Also, because of our patient selection method, those patients with flare who did not undergo stool-based testing were not included in the study. While this population is likely very small as it would be rare for patients with suspected IBD flare to not undergo stool-based testing, we cannot exclude the possibility that such subjects might have different characteristics and outcomes from those evaluated in our study. We additionally did not record alternative diagnoses for those with negative stool testing, nor evaluate any potential impact of disease activity prior to flare, type of new IBD treatment at index flare, IBD disease location, or recurring enteric infection on our results.

This study was further limited by its definition of IBD flare; we did not require objective (biomarker, radiographic, endoscopic, or histologic) evidence of inflammation as part of our definition of flare; thus, our study may bias towards including patients without luminal inflammation. While we adjusted for serologic markers of IBD severity and new IBD therapy at flare, we were unable to fully account for the possibility that subjects with negative EI testing were undertreated for IBD at the time of flare presentation compared with those with positive EI testing. Evaluation of disease control prior to IBD flare, as well as objective evidence of inflammation, should be performed in future prospective studies.

Finally, the GI panel may not reliably distinguish between infected and colonized subjects. However, GI panels are more frequently positive in active IBD; in a study of 333 IBD patients and 52 healthy subjects, a positive GI panel was present in nearly one-third of those with active disease, while only present in 8% of those with quiescent disease (P < .001) and in 14% of healthy controls (P = .01).⁷ These data suggest that positive stool PCR testing likely represents true EI.

In conclusion, positive EI testing—specifically for CDI—at the time of flare is associated with IBD therapy class change, dose escalation, colectomy, and steroid-dependent disease 2 years after IBD flare. While this association is primarily related to CDI, infection with E. coli subtypes may also play a role. Non-CDI/non-E. coli EI at flare likely has a comparable subsequent IBD disease course to infection-free flare. Our findings further support the distinction between CDI and non-CDI EI in this patient population, extend the duration of the relationship between CDI and poor IBD outcomes, and suggest E. coli as a potential pathogen of concern during flare. Prospective long-term studies comparing CDI, infection with E. coli subtypes, other enteric infections, and negative EI testing at flare are needed to further elucidate the impact of enteric infection at flare on subsequent IBD disease course.

Supplementary Data

Supplementary data is available at Inflammatory Bowel Diseases online.

Author Contribution

A.D.V.—conceptualization, lead; data curation, lead; formal analysis, lead; methodology, lead; writing—original draft, lead.

S.H.—data curation, supporting; writing—review and editing, supporting.

J.E.A.—conceptualization, supporting; methodology, supporting; supervision, lead; writing—review and editing, lead.

Funding

J.E.A. receives research support from the Crohn’s and Colitis Foundation, the Judith & Stewart Colton Center for Autoimmunity, and NIH NIDDK K23DK124570.

Conflicts of Interest

J.E.A. has received research grants from BioFire Diagnostics; consultancy, advisory board fees, or honorarium from BioFire Diagnostics, Adiso, Abbvie, Pfizer, BMS, and Janssen; and holds US patent 2012/0052124A1. A.D.V.: No disclosures. S.H.: No disclosures.