Multiplex stool polymerase chain reaction tests (SPTs) simultaneously test for many enteric pathogens. However, the clinical significance of a positive result, particularly in the context of chronic gastrointestinal disease, remains controversial.
To determine whether SPT results correlate with findings on colon biopsies obtained within a week of SPT or with clinical features.
We reviewed 261 colon biopsies during a 15-month period that were obtained within a week of SPT, along with available clinical information, from patients with and without chronic idiopathic inflammatory bowel disease (CIIBD). Statistical analysis was used to test associations between SPT result, histologic features, and clinical variables.
The most commonly detected pathogens were Clostridium difficile, enteropathogenic Escherichia coli, and norovirus. The presence of underlying CIIBD did not correlate with a positive SPT result or with a specific pathogen. Positive SPT result was significantly associated with neutrophilic activity, pseudomembranes, and increased intraepithelial lymphocytes. In addition, the presence of C difficile on SPT was significantly associated with pseudomembranes and neutrophilic activity. There were no other statistically significant relationships between SPT result and any other histologic abnormality. Only about half of SPT positive results were acted on clinically, and most patients with CIIBD were managed as having a presumed IBD flare.
SPTs have many advantages; however, interpretation of results, particularly in the background of chronic gastrointestinal disease, remains a challenge. Therapeutic decisions influenced by a positive SPT result should integrate biopsy findings, clinical data, and other laboratory testing to avoid inappropriate treatment.
Infectious gastroenteritis is a significant public health problem in both industrialized and developing countries.1,2 It is estimated that approximately 178.8 million cases of infectious gastroenteritis occur in the United States each year, and the overall mortality rate from infectious gastroenteritis tripled between 1985 and 2005 in the United States alone.1,3 Elderly patients and those with increased body mass index are at particular risk for hospitalization due to infectious gastroenteritis, which is of considerable importance given our aging and increasingly obese population.4 Furthermore, enteric infections are important causes of morbidity in immunocompromised patients and those with chronic idiopathic inflammatory bowel disease (CIIBD).5
It can be challenging to detect and identify enteric pathogens when patients present with diarrhea, especially hospitalized patients. Stool cultures are limited by low sensitivity, and several days are often required to get results. Some recent studies have underscored this point by comparing traditional stool culture methodology to various molecular platforms. For example, the sensitivity for detection of Shiga toxin–producing Escherichia coli by molecular methods was 93% to 95%, as compared to 33% for stool culture and enzyme immunoassay–based detection.6 Similar results have been shown for Salmonella species, Shigella species, and Campylobacter species.7–9 Stool culture results have also historically been affected by the fact that many patients empirically receive antibiotics before obtaining stool specimens for culture.10,11 For all of these reasons, the relatively recent application of multiplex polymerase chain reaction (PCR) assays to stool specimens has engendered a great deal of excitement among laboratory personnel and clinicians.
The FilmArray Gastrointestinal PCR Panel (BioFire Diagnostics, Salt Lake City, Utah) is a widely used rapid multiplex PCR platform that received federal US Food and Drug Administration approval in 2014.12 It tests liquid stool for 22 types of viral, bacterial, and parasitic pathogens, including Campylobacter species, Clostridium difficile (toxin A/B), Salmonella species, Shigella species, Yersinia enterocolitica, Vibrio species, pathogenic E coli subtypes (enteropathogenic [EPEC], Shiga-like toxin–producing, enteroaggregative, enterotoxigenic, and O157), Plesiomonas shigelloides, Cryptosporidium parvum, Cyclospora cayetanensis, Entamoeba histolytica, Giardia lamblia, adenovirus F 40/41, astrovirus, norovirus GI/GII, rotavirus A, and sapovirus (FilmArray Gastrointestinal Panel Instruction Booklet). Per the FilmArray package inserts, the overall sensitivity is 98.5% and the specificity is 99.2%, figures that have been closely approximated in large studies in the literature.8,13–15 In addition, the sensitivity and specificity have been reported as greater than 99% for many of the individual pathogens, including Salmonella spp, Y enterocolitica, astrovirus, rotavirus A, and sapovirus.8,13 This technology promises to improve pathogen detection and identification, and also offers markedly decreased turnaround time (about an hour from the time the specimen is loaded onto the machine) when compared with older techniques such as microbial culture or stool ova and parasite examination.
Although the use of stool multiplex PCR technology is increasing in frequency, the clinical significance of a positive stool PCR test result is unfortunately not always clear. It may be difficult to determine whether detection of a pathogen by stool multiplex PCR represents an infection requiring treatment, colonization by a nonpathogenic organism, or a nonviable pathogen.16 Interpretation of results can be even more challenging when patients have multiple medical conditions requiring polypharmacy, or underlying CIIBD, which may alter the microbiome. The goal of this study was to determine how the results of a stool PCR test, specifically the FilmArray Gastrointestinal Panel, correlate with clinical findings and histologic abnormalities in biopsy material from the lower gastrointestinal tract in patients with and without inflammatory bowel disease (IBD).
With institutional review board approval by the University of Michigan, we retrospectively identified all adult (>18 years of age) individuals in a 15-month period (April 2016–June 2017) who underwent colonoscopy with mucosal biopsy within 1 week or less of having a FilmArray stool PCR test. Cases limited to sampling of a neoplasm or polyp were excluded. Initially, 301 biopsies were identified from 288 unique patients. To minimize confounding factors affecting patient symptoms, clinical impression, and histologic features, the patients who had undergone bone marrow transplant, had presumed gastroenteritis secondary to mycophenolate and checkpoint inhibitor therapy, or had a prior diagnosis of microscopic colitis were excluded from further analysis. Biopsies from the remaining patients, consisting of 261 biopsies from 251 unique patients, were then subdivided into those with and without IBD, and this comprised the study group.
All biopsy samples (routinely processed, formalin-fixed, paraffin-embedded specimens stained with hematoxylin-eosin) were reviewed by 3 pathologists for the presence of chronic and/or active colitis, increased apoptotic crypt epithelial cells, pseudomembranes, granulomas, ulcers, increased intraepithelial lymphocytes, and the presence of organisms or viral inclusions. Electronic medical records were reviewed for clinical findings including presenting symptoms, comorbid conditions, treatment, and outcome.
χ2 analysis or Fisher exact test (when expected values were less than 5) were used to test associations between the stool PCR test results and histologic features. Significance was determined if P < .05. All analyses were conducted with SAS (version 9.4, SAS Institute).
Clinical and Demographic Features
The 251 patients in the study group consisted of adult patients with a mean age of 46 years (range, 18–89 years). One hundred thirty-eight patients (138 of 251, 55%) had underlying CIIBD (both ulcerative colitis and Crohn disease), whereas the remaining 113 patients had no known history of established CIIBD but did have a variety of underlying chronic medical conditions including a subset with immunosuppression due to systemic disease or solid organ transplant. Presenting symptoms were some combination of diarrhea, hematochezia, and/or abdominal pain in virtually all patients.
Stool PCR Results
Altogether, stool PCR testing was positive in 53 of 261 samples (20%), all from unique patients; of these, 28 of 53 samples (53%) were from patients with chronic idiopathic IBD. The pathogens detected are listed in Table 1. Clostridium difficile (N = 30 of 53, 57%), EPEC (N = 9 of 53, 17%), and norovirus (N = 7 of 53, 13%) were the most commonly identified pathogens, both in patients with and without underlying CIIBD, while the remaining organisms detected were identified in 2 or fewer samples each. A positive PCR result for C difficile also correlated with a positive enzyme immunoassay (EIA) toxin/antigen test in 15 of 30 patients (8 CIIBD patients and 7 non-CIIBD patients), whereas EIA was negative in 14 of 30 patients (1 patient did not have a concurrent EIA) (Table 2). Overall, there was no significant difference in pathogen detection rates between those with and without CIIBD (28 of 147 [19%] versus 25 of 114 [22%], respectively), nor was there a significant association between the specific pathogen detected and the presence of underlying IBD.
Stool cultures were performed for only 3 patients in the cases that had both a biopsy and a positive stool PCR test result, either for purposes of culture and sensitivity testing or because state public health requirements mandated it. Two stool PCR tests that were positive for Salmonella spp and one that was positive for Shigella sonnei, were also positive on stool culture.
Histologic Findings and Correlation With Stool PCR
Biopsies obtained within 1 week of a positive stool PCR result were reviewed for specific histologic features and compared with those of patients with and without CIIBD who had a negative PCR result (Table 1). Abnormal histologic findings were identified in 43 of 53 biopsies (81%) taken from patients with a positive PCR test result, compared with 142 of 208 biopsies (68%) from patients with a negative PCR test result. In patients without IBD, 8 of 25 PCR-positive patients (32%) were found to have a histologically normal biopsy, whereas normal findings were only identified in 2 of 28 patients with IBD (7%) who had a positive PCR test result.
The most common histologic abnormality was neutrophilic activity (Figure 1), which was present in 38 of 53 biopsies (72%) associated with a positive PCR test result and was more frequently identified in biopsies from patients with positive stool PCR test results than in biopsies from patients with negative PCR test results (P = .03). In addition, pseudomembranes (Figure 2, A and B) and increased intraepithelial lymphocytes (Figure 3) were identified in 6 of 53 (11%) and 7 of 53 (13%) biopsies, respectively, in patients with a positive PCR result compared with only 4 of 208 (2%) and 3 of 208 (1%) biopsies in patients with a negative PCR result (P < .001 and P = .006, respectively). Pathogens detected in cases with increased intraepithelial lymphocytes included C difficile (2 cases), norovirus, astrovirus, Cryptosporidium species, and EPEC. No other histologic features were statistically associated with a positive PCR result compared to control cases with a negative PCR test result. However, when comparing patients with and without CIIBD separately, only neutrophilic activity and lymphocytosis remained significant in non-CIIBD patients with a positive PCR test result (P = .045 and P < .001), whereas pseudomembranes alone remained significant in CIIBD patients (P < .001) with a positive PCR test result. Altogether, 6 of 10 patients (60%) with pseudomembranes on biopsy had a positive PCR test result for C difficile, though this comprised just 20% (6 of 30) of C difficile–positive patients overall. Neutrophilic activity, however, was seen in most (24 of 30, 80%) C difficile–positive cases. Interestingly, no significant differences in histologic findings were present between C difficile PCR-positive cases with and without an associated toxin positivity by EIA (Table 2).
When evaluating histologic findings in PCR-positive CIIBD patients compared to PCR-negative CIIBD patient controls, a few subtle differences were notable. A normal biopsy finding was slightly less common (2 of 28, [7%] versus 19 of 119 [16%]) and neutrophilic activity was slightly more common (25 of 28 [89%] versus 90 of 119 [76%]) in patients with a positive PCR test result, though neither of these were statically significant and the severity of activity overall did not substantially differ between cases and controls. Pseudomembranes, however, were exclusively identified in CIIBD patients with a positive PCR finding for C difficile and not seen in PCR-negative CIIBD patients. No other histologic features, including the severity of underlying chronic disease, significantly differed between the 2 groups.
All patients with a positive PCR result were inpatients at the time of evaluation, and the hospital course in each case was reviewed for overall clinical impression, treatment, symptomatic improvement/resolution at the time of discharge, and follow-up data, if available (Table 3). At the time of PCR testing and mucosal biopsy, 17 of 53 patients were known to have previously received antibiotics before hospital admission, and several additional patients received antibiotics at the time of admission immediately before testing. This information was difficult to reliably ascertain in every case, however, as multiple patients were transferred to our institution from an outside hospital. Thirty-two patients (32 of 53) were immunosuppressed at the time of admission (25 patients with CIIBD on immunosuppressive therapy, 5 patients with solid organ transplants, and 2 patients with human immunodeficiency virus). From review of the medical records, the PCR results were interpreted to represent a presumed clinically significant infection in only 28 of 53 patients (53%), 18 with C difficile infection (15 with positive EIA, 2 with negative EIA, and 1 without EIA performed). In the remaining cases, clinically significant infection was thought to be less likely or unlikely, often because patients responded to treatment for CIIBD or the concomitant C difficile EIA was negative. Thirty-seven patients (37 of 53, 70%) ultimately received treatment for infection; in 9 cases, however, the antibiotics were initiated for indications other than the positive stool PCR result. Twenty-four patients with CIIBD (24 of 28, 86%) were also treated for presumed CIIBD flare during admission. Symptom resolution was documented in 39 of 53 patients (74%). Of those with positive C difficile testing, 3 of 30 patients developed recurrent C difficile infection requiring additional courses of antibiotics and 3 of 30 underwent fecal microbiota transplant with ultimate resolution of symptoms.
The purpose of this study was to determine whether the results of multiplex stool PCR studies correlated with any specific clinical features or histologic abnormalities in colon biopsy samples obtained within a week or less of the stool PCR test. To our knowledge, this is the first study of its type. The top 3 pathogens that were detected were C difficile, EPEC, and norovirus, similar to previous studies.8,16 We found that the presence of neutrophilic activity, pseudomembranes, and intraepithelial lymphocytes on biopsy were each significantly associated with a positive stool PCR test result, which is not surprising given that these are typical histologic features of infection on colon biopsies.17 The association between a positive stool PCR test result and intraepithelial lymphocytes or neutrophilic activity remained significant for patients without IBD when the subgroup analysis was performed, whereas only the association between a positive PCR result and pseudomembranes remained significant for patients with CIIBD. When the data were analyzed by individual pathogen, the presence of C difficile by stool PCR was significantly associated with pseudomembranes and neutrophilic activity, but not with other features such as mucosal necrosis. There were no significant associations between results of stool PCR test and the other histologic abnormalities that we evaluated.
Previous studies have also evaluated the use of stool PCR tests to determine the prevalence of gastrointestinal infections in patients with pre-existing CIIBD.18–20 Patients with active disease are more likely to have positive stool PCR test results than those with inactive disease, and patients taking multiple immunosuppressive medications are also more likely to have a positive stool PCR test result.18 Similar to our data, the most commonly detected organisms in this patient population in previous literature were C difficile and E coli species.19 The use of steroids and immunomodulators are independently associated with the development of C difficile infection in these patients as well, and recurrent infection is fairly common. Among CIIBD patients who are experiencing flares, the prevalence of C difficile infection is much higher than for those with inactive CIIBD or patients without CIIBD,21–24 though the rate of detection did not differ in hospitalized patients with and without CIIBD who underwent testing in the current study. Interestingly, it has been previously noted that biopsy samples from CIIBD patients with C difficile infection may not show the classic findings of pseudomembranous colitis typically seen in non-CIIBD patients,24 and though pseudomembranes were a significant finding in this subpopulation in our study, it was absent in several of the biopsy samples taken from CIIBD patients who did have a positive PCR result for C difficile.
Norovirus and pathogenic E coli strains have also been previously associated with CIIBD flares, and these associations are topics of ongoing research.19,20,25 A positive stool PCR result in these patients often affects disease management, in that patients are less likely to have immunosuppressive therapy increased, or to have additional immunosuppressive drugs added to their regimens.19,20,26 The effect of targeted antimicrobial therapy in this situation remains controversial, however, and some studies have found that outcomes were equivocal between those patients who received targeted antibiotic therapy and those who did not.20 Within our own study population of patients with CIIBD and a positive stool PCR result, only about half (54%) were presumed to be clinically significant infections, though 75% of these patients did receive antibiotic therapy during their admission and most were treated for possible CIIBD flare.
There remains ongoing concern about whether or not positive stool PCR test results indicate true infection, colonization, or even the detection of a nonviable organism, particularly in the case of C difficile, for which the sensitivity is 98.8% and the specificity is 97% with the BioFire assay.13,27 Many patients with stool PCR-positive results but C difficile toxin A/B immunoassay negative findings are ultimately clinically determined to be colonized, rather than infected, with C difficile.28,29 One study showed that among adult inpatients with suspected C difficile infection, virtually all infection-associated morbidity and mortality occurred in patients with positive toxin immunoassay results, and patients with a positive stool PCR test result but a negative toxin immunoassay had outcomes similar to those of patients who tested negative for C difficile by any method.29 Other organisms detected by stool PCR assays in our study and previous studies, including EPEC and norovirus, have been shown to colonize renal transplant patients without causing histologic changes, and studies in pediatric patients have shown a high rate of positive stool PCR results in asymptomatic controls.27,30 These prior studies, as well as our data, emphasize that exclusive reliance on stool molecular tests, without correlation with other pertinent tests and clinical information, could result in overdiagnosis or underdiagnosis of gastrointestinal infections and subsequent inappropriate therapy.
Our study has several limitations. Because patients with significant underlying gastrointestinal disease, such as CIIBD or transplant-associated conditions, are more likely to be extensively evaluated for their diarrhea, our cohort is heavily influenced by patients with these underlying diseases. This makes it very challenging to discern how much of the histologic abnormality is due to possible infection, and how much is due to underlying gastrointestinal disease, and thus it is not surprising that most of the histologic features of the biopsies did not show statistically significant correlation with stool PCR results. In addition, many patients with CIIBD and positive stool PCR test results were having a clinically apparent flare, and were thus treated for their CIIBD, confounding our ability to ascertain whether or not they had a clinically significant infection or response to antimicrobial therapy. Although few of our patients were specifically treated with antibiotics before colonoscopy as a result of the stool PCR test, many were given antibiotics for other reasons (eg, transplant prophylaxis, therapy for acute urinary tract infection), and thus the effects of antibiotics on the study results are difficult to ascertain. Anecdotally, antibiotics are not believed to affect stool PCR results to the same extent that they affect traditional stool culture, but to our knowledge there are no large studies as of yet that examine this in detail.31 By limiting the time between stool PCR test and colonoscopy to 7 days, however, we believe that we minimized the chances that histologic abnormalities had entirely resolved between a positive stool PCR test result and the colonoscopy.
In summary, our study demonstrated that the presence of activity, pseudomembranes, or intraepithelial lymphocytes on biopsy was significantly associated with a positive stool PCR test result overall. In addition, the presence of C difficile on stool PCR test significantly correlated with pseudomembranes and neutrophilic activity. Similar to previous literature, the top 3 pathogens that were detected were C difficile, EPEC, and norovirus. There was no significant correlation between a positive stool PCR assay and any other individual histologic features. There was also no statistically significant association between stool PCR test result and the presence of underlying IBD. The role of many of these pathogens in the outcomes of patients with CIIBD and organ or bone marrow transplant is worthy of further study, as is the role of stool PCR test in patients who do not have any underlying comorbid conditions.
Ultimately, stool PCR tests have many positive attributes, including rapid turnaround time, high sensitivity, and ease of performing the test. There are other benefits of this testing methodology as well, including decreased time in isolation, more appropriate and more timely prescribing of antibiotics, and decreased imaging studies, in addition to using this technology to study the changing epidemiology of pathogens in patients undergoing allogeneic stem cell transplants.31–33
However, interpretation of results, particularly in the background of chronic gastrointestinal disease, remains a challenge. Our data indicate that therapeutic decisions should not be made on the basis of a positive stool PCR test result alone, but should be influenced by biopsy findings, clinical data, and other laboratory testing in order to avoid overtreatment of infection or undertreatment of underlying CIIBD.19,20,26,34
The authors gratefully acknowledge Rhonda Yantiss, MD, Liron Pantanowitz, MBBCh, Drew Pratt, MD, and Rondell Graham, MBBS, for their editorial assistance.
The authors have no relevant financial interest in the products or companies described in this article.
These data were presented in part at the Annual Meeting of the United States and Canadian Academy of Pathology; March 19–24, 2018; Vancouver, British Columbia, Canada.