Mapping of etiologies of computed tomography-proven acute colitis: a prospective cohort study

Our objective was to describe the etiologies of acute colitis and to identify patients who require diagnostic endoscopy. Patients with symptoms of gastrointestinal infection and colonic inflammation on CT were prospectively included. Those immunosuppressed, with history of colorectal cancer or inflammatory bowel disease (IBD), were excluded. Microbiological analysis of the feces was performed using PCR assays BD-Max and FilmArray (GI panel,) and fecal cultures. Fecal calprotectin was determined. Patients with negative BD-Max underwent colonoscopy. One hundred and seventy-nine patients were included. BD-Max was positive in 93 patients (52%) and FilmArray in 108 patients (60.3%). Patients with infectious colitis (n = 103, 57.5%) were positive for Campylobacter spp. (n = 57, 55.3%), Escherichia coli spp. (n = 8, 7.8%), Clostridioides difficile (n = 23, 22.3%), Salmonella spp. (n = 9, 8.7%), viruses (n = 7, 6.8%), Shigella spp. (n = 6, 5.8%), Entamoeba histolytica (n = 2, 1.9%) and others (n = 4, 3.9%). Eighty-six patients underwent colonoscopy, which was compatible with ischemic colitis in 18 patients (10.1%) and IBD in 4 patients (2.2%). Fecal calprotectin was elevated in all patients, with a mean concentration of 1922.1 ± 2895.6 μg/g, and was the highest in patients with IBD (8511 ± 9438 μg/g, p < 0.001). After exclusion of patients with infectious etiology, a fecal calprotectin > 625 μg/g allowed identifying patients with IBD with an area under ROC curve of 85.1%. To conclude, computed tomography-proven colitis was of infectious etiology in 57.5% of patients. The main pathogens identified were Campylobacter spp. (55.3%), Clostridioides difficile (22.3%) and Salmonella spp. (8.7%). Ischemic colitis (10.1%) and IBD (2.2%) were seldom represented. No colorectal cancer was found.

Definition of the etiologies of acute colitis. The episode of acute colitis was considered as infectious if either the routine BD-Max or the multiarray PCR assay FilmArray GI panel was positive. Escherichia coli spp. EAEC, EPEC and ETEC identified only by FilmArray GI panel were considered as carriage due to ongoing doubts regarding their pathogenicity. Pathogens were further identified by fecal culture. If BD-Max was negative, diagnosis was obtained using colonoscopy with or without biopsies. In non-infected cases, we defined the etiology of colitis as ischemic, or secondary to colorectal cancer or IBD in the presence of suggestive macroscopic aspect and/or histopathology. Patients later registered in the Swiss IBD cohort were considered to have IBD. To avoid missing a secondary diagnosis in patients with infectious etiology, medical files were reviewed at the date of the inclusion of the last patient to rule out later diagnosis of IBD and/or colorectal cancer. Undetermined colitis, in which the etiology of acute colitis was unknown, was defined as normal colonoscopy and, if available, normal histology. Statistical analysis. All statistical analyses were performed using STATA (version 13, StataCorp LP, College Station, USA) and GraphPad PRISM (version 7, GraphPad Software, La Jolla California USA). The null hypothesis was rejected at p < 0.05. Network representation of pathogens identified using the FilmArray GI panel was performed using Cytoscape (version 3.8.1).
Differences between groups. Differences between groups (etiologies of colitis) were determined using the Student's two-sided t-test or the Pearson's chi-squared test, as appropriate. Continuous variables were transformed into categorical variables if required. Variables were expressed as proportions for categorical variables and means for continuous variables; 95% confidence intervals (95% CI) and standard deviations (SD) were reported, as appropriate.
Receiver operating characteristic (ROC) curve. To define the optimal cut-off point of fecal calprotectin allowing identifying patients suffering from IBD, a ROC curve was drawn. Standard error was calculated according to the DeLong method. The optimal cut-off point was determined using the Liu method.

Results
Inclusion process. From December 2016 to December 2019 (3 years), 300 patients were potentially eligible for inclusion in the study. Sixty-two patients (20.7%) were excluded for fulfilling at least one exclusion criteria (history of IBD: 14 patients, immunosuppression: 14 patients, inability to provide informed consent: 10 patients, no possibility for follow-up (foreign resident): 10 patients, liver cirrhosis with ascites: 8 patients, history of colorectal cancer: 4 patients, refuse to participate: 2 patients), leaving 238 patients for inclusion. Among those, biological samples were unavailable (improperly harvested or stored) in 33 patients, colonoscopy was not performed during hospitalization or was postponed after the acute phase in 18 patients, and 8 patients refused colonoscopy. Therefore, 59 patients (representing 24.8% drop-outs) were further excluded, leaving 179 patients for final analysis (Fig. S2) Fig. S4).
Colonoscopy. Eighty-six patients with negative BD-Max underwent colonoscopy within 6.2 ± 2.9 days after computed tomography. Colonoscopy was macroscopically and histologically compatible with ischemic colitis in 20 patients (23.3%) and with IBD in 4 patients (4.7%) (Fig. S5). Two of the patients with signs of ischemic colitis had positive samples and were therefore considered as having infectious colitis. Eight patients (9.3%) had polyps that were removed (6 patients with low grade dysplasia adenomas, 2 with hyperplastic polyps   Table 3).
Matching with the Swiss IBD cohort. No additional diagnosis of IBD was identified from the Swiss IBD cohort.
Primary outcome: etiologies of computed tomography-proven acute colitis. One hundred and three (57.5%) patients had positive routine PCR and/or positive BioFire FilmArray GI panel and were therefore considered as suffering from infectious colitis. Isolated EAEC, EPEC or ETEC were considered as carriage and were therefore not considered as infectious etiology (8 patients). Eighteen patients (10.1%) with negative microbiological examination of the feces showed macroscopic and histological characteristics of ischemic colitis and were classified as suffering from ischemic colitis. Four patients (2.2%) showed characteristics of IBD on biopsies. Three were diagnosed with ulcerative colitis and one with Crohn's disease. Finally, fifty-four patients (30.2%) had negative microbiological examination of the feces and no sign of chronic inflammation on colonoscopy and/      Fig. S4). When analyzing only patients with negative microbiological examination of the feces, ROC curve analysis of calprotectin indicates an optimal cut-off point for diagnosis of IBD at 624.8 μg/g (sensitivity: 100%, specificity: 62%, area under the ROC curve: 0.81).

Discussion
The medical literature is dramatically scarce regarding the definition of acute colitis and its etiologies. Numerous publications refer to the term infectious colitis to describe viral and/or bacterial episodes of diarrhea without any confirmation of colonic inflammation by imaging modalities, therefore mostly referring to episodes of gastroenteritis 9,10 . Further, most studies reporting the prevalence of bacterial pathogens during episodes of diarrhea identified low proportions of positive samples 11,12 , reflecting the low sensitivity of standard fecal cultures and the non-targeted selection of patients. To our knowledge, only one study investigated symptomatic patients admitted in emergency with CT-proven episode of colitis but suffered from information bias, with no broad screening of the feces and endoscopy reserved to a not clearly defined subgroup of patients 13 . Therefore, we thought that the recent emergence of PCR-based multi-array assays could constitute a significant advance in the field of etiological research for colitis by decreasing the number of false-negative results and guiding antibiotic therapy.
In the present study, we systematically screened the feces of patients with CT-proven acute colitis for infectious etiologies using the FilmArray GI panel, besides the routine PCR workflow (BD-Max). Out of the 179 included patients, 103 (57.5%) had positive routine PCR and/or positive FilmArray GI panel. The FilmArray GI panel allowed identifying a pathogen in 60.3% of patients, a proportion that is higher than the proportions reported in episodes of gastroenteritis detected using the same panel [14][15][16] , which is of 39.7% as reported by a recent systematic review and meta-analysis 17 . This higher proportion of identified pathogens might be notably explained by a more precise selection of patients in our cohort, which included symptomatic patients with radiologic evidence of colonic inflammation. Of note, studies applying better selection of patients with gastrointestinal symptoms obtained a proportion of positive samples of up to 71% using the same assay 18 . Similarly, the spectrum of pathogens encountered in the present study was different than the spectrum of pathogens identified during episodes of gastroenteritis. For instance, the prevalence of Campylobacter spp. in our study population was of 31.8% (representing 55% of the 58% of patients with infectious colitis). We previously published a systematic review and meta-analysis reporting on the prevalence of different pathogens in patients with gastroenteritis identified with the same multiplex PCR assay as in our cohort study, and found a prevalence of Campylobacter spp. of 11.8% 17 . We believe that the difference may result from a different population selection. For instance, in the present cohort study, we included only patients with symptomatic colonic inflammation demonstrated by CT, and not only patients suffering from "gastroenteritis", which may regroup a wide variety of symptoms and clinical presentations. Moreover, we noted that several patients had co-infections with different pathogens (co-detection by molecular panel). This was previously documented in fecal samples collected in other settings, with co-infection rates ranging from 9.8 to 38% 14,15,[19][20][21] , and sometimes detection of up to 6 pathogens 22 . This co-detection may be the result of synchronous co-infection by different pathogens or of false positives. This latest possibility is, however, less likely as previous studies have reported that the multiplex PCR assay used in our study had a specificity ≥ 97.1% when compared to fecal cultures and molecular methods 22 . Also, in our study, the majority of positive identification was confirmed by additional methods performed in parallel, as reported in Table 2.
Fecal calprotectin is a protein produced by neutrophils that accumulates in the feces in case of bowel inflammation or infection. Recent applications for calprotectin include the early diagnosis of IBD (calprotectin showed a 93% sensitivity and a 86% specificity to distinguish IBD from irritable bowel syndrome 23 ) and the detection of recurrence of both Crohn's disease and ulcerative colitis [24][25][26] . Similarly, calprotectin concentration is increased during bacterial diarrhea 27 or colorectal cancer. In our study, we showed that fecal calprotectin was elevated in patients suffering from CT-proven colitis with a mean concentration of 1922.1 ± 2895.6 μg/g, therefore showing the important colonic mucosal inflammation in these patients.
Further, we performed colonoscopy in patients with negative routine PCR panel examination (BD-Max) of the feces (accounting for 48% of the cohort). Colonoscopy allowed diagnosing ischemic colitis in 10.1% of included patients, and IBD in 2.2% (three patients with ulcerative colitis and one with Crohn's disease). No colorectal cancer was found. Moreover, in our cohort, 30.2% of included patients had negative microbiological examination of the feces and no sign of chronic inflammation on colonoscopy and/or biopsies, and were considered as having an episode of colitis of undetermined etiology. Calprotectin concentration was significantly lower in these patients than in the rest of the cohort, but still, 50% of patients with an undetermined episode had a calprotectin concentration equal to or superior to 250 μg/g, which demonstrates an active inflammation of the gastrointestinal tract. Of note, 45.2% of them had eosinophilic infiltration of the mucosae on histology. We think these findings are compatible with a subacute inflammation, explaining the absence of positive microbiological examination of the feces, false-negative result of the microbiological examination of the feces (which is unlikely as several methods of identification were performed in parallel), infection by a pathogen not detected by the methods of identification performed in the present study or, alternatively, iatrogenic origin (for example, NSAID consumption).
We believe that patients suffering from acute colitis should be guided to the appropriate diagnostic and therapeutic cares, with the objective to generate savings by shortening hospital stays and by improving prescription of additional tests, including colonoscopy. We therefore think that an adequate and early patient stratification has important medical and economical values in this setting. Therefore, considering that patients suffering from IBD require specialized treatment and follow-up, we aimed at identifying patients suffering from IBD to better target the subpopulation of patients needing diagnostic endoscopy. First, introduction of a PCR assay with more www.nature.com/scientificreports/ targets allowed to better identifying patients with infectious etiology. For instance, FilmArray GI panel was positive in 20.9% of patients with otherwise negative routine PCR (BD-Max). Pathogens identified were mostly pathogens not looked for by routine PCR and/or culture. Second, we have introduced fecal calprotectin, which is currently used for the early diagnosis of IBD allowing to distinguish them from functional disorders 23 and the detection of recurrence [24][25][26] . We found that calprotectin was elevated in patients with episodes of infectious colitis(as previously reported in patients with bacterial diarrhea 27 ), but also in patients with ischemic colitis and undetermined colitis. The highest values were, however, reported in patients with a first episode of IBD manifesting as an episode of colitis. We then determined the optimal cut-off value allowing identifying patients with IBD and determined that value to be of 625 μg/g. Applying that value as a cut-off in our cohort would have allowed reducing the number of potentially unnecessary colonoscopy from 82 to 29, therefore resulting in a decrease of 64.6% in the number of colonoscopies. The present study has the following strengths: (1) its prospective design reduced selection bias and missing patients otherwise eligible for inclusion, (2) the introduction of thorough screening methods for microbiological examination of the feces applied for the first time in patients with acute colitis and performed in parallel, (3) reporting for the first time clinical and para-clinical presentations, as well as etiologies, of patients with computed tomography-proven episode of colitis. Its limitations are the following: (1) the number of drop-outs (24.8%) due to protocol violation, mostly due to absence of sampling, but which was in the range of the proportion planned in our preliminary sample size calculation, (2) the absence of colonoscopy in patients with positive microbiological examination of the feces, which could have led to miss potential cases of IBD and/or colorectal cancer. On this latest aspect, however, cross-matching of our cohort with the Swiss IBD cohort did not identify additional cases of IBD. And finally (3) potential patients' selection as some patients with CT-proven acute colitis may have been treated as outpatients if not referred to the recruiting hospital.