The potential of using multiplex PCR testing for the diagnosis and management of infectious diarrheal illness has been extensively documented in a variety of settings [13-15, 17, 20-21], but so has its potential for yielding results of unclear clinical significance, including false-positives and multiple positives per sample [18-19, 22-23]. This tendency is inherent in the technology, and is a result of its reliance on the amplification of genetic loci which may persist in the gut during asymptomatic carriage, or be transferred horizontally between enteric bacteria, particularly E. coli pathotypes and other enteric bacteria [16-17, 25-27]. In clinical use, this creates a conflict between the desire to expedite diagnostic results in patients with infectious gastroenteritis, and the potential to over-treat non-pathological results.
Co-detection of multiple pathogens, particularly multiple E. coli pathotypes, was common in our study. This is similar to findings in prior studies of multiplex PCR panels for gastrointestinal sample analysis, in which co-detections were reported in 16-28% of positive samples [22-27]. The FilmArray GI panel is noted to be more prone to detecting mixed infections than the Luminex xTAG system (Luminex Corporation, Austin, TX), the only other FDA-approved multiplex PCR gastrointestinal panel [22]. Whether such detections represent true coinfections of viable organisms or merely colonization cannot be determined based on these tests alone, and decisions regarding treatment continues to be a clinical issue. Standard bacterial cultures continue to have a key role in confirming PCR diagnoses, but published rates of successful culture confirmation of multiplex PCR findings range widely. In one meta-analysis, culture confirmation was 63.9% for Campylobacter (CI: 39.8-84.9%), 48.4% for Salmonella (CI: 27.8-69.3%), 73.4% for Shigella (CI: 38.1-97.1%), and 75.0% for E. coli O157 (CI: 50.9-91.3%); overall, the PCR methods detected approximately 1.5 times more pathogens compared to standard stool cultures [18].
Overestimation of C. difficile infections is well-documented risk of using PCR-based diagnostics. In one study of the FilmArray GI panel, only 43% of samples where C. difficile was detected were subsequently found to produce detectable toxin by enzyme immunoassay (EIA) [28]. In our study we found that physicians did not consistently order a confirmatory EIA in patients who had C. difficile detected by the FilmArray GI panel. This has since been addressed with quality improvement interventions in our institution, and a consistent two-step reflex testing algorithm instituted. Physicians have also been educated about refraining from ordering the GI multiplex panel when a high pre-test suspicion for CDI exists. Additional aggressive institutional and regional efforts in antibiotic stewardship and infection prevention aimed at reducing CDI incidence have been ongoing. It is possible that these efforts contributed to the difference in LOS and antibiotics use we observed, though the relatively short period of time captured (2015-2016) makes it less likely that such changes in practice or other secular trends (such as the tendency towards shorter LOS overall) entirely account for our findings.
Several recent multi-center studies have compared the FilmArray GI panel to other PCR assays or conventional culture methods with mixed results. While initial studies showed high sensitivity and specificity of FilmArray GI panel when compared to reference methods [16-17, 22, 25], other real-world studies have found more variation in the rates of pathogen detection and reproducibility, depending on the practices of specific centers [23-24, 26-27]. The use of the FilmArray GI panel has implications regarding other aspects of patient care, including use of hospital resources and infection control. Beal et al. found a modest reduction in LOS, days of antibiotics, abdominal imaging studies, and costs per patient associated with the use of the FilmArray GI panel; however, in that study the FilmArray GI panel results were not communicated to physicians in a timely manner, likely reducing their impact on management [29]. In a cost-benefit analysis, the use of the Luminex xTAG gastrointestinal panel was associated with increased diagnostic costs but break-even points and cost savings could be realized contingent on using the test results to reduce isolation days [30]. In practice, physicians frequently fail to follow up on these results with appropriate adjustment of isolation status [31]. Some experts favor restricting the use of multiplex PCR panels to initial evaluations or to patients with severe symptoms in order to improve cost-effectiveness [32-33].
In our study, the introduction of FilmArray GI panel testing did not appear to affect overall time to disposition (either hospital admission or discharge from the ED), or time to empiric antibiotics. These findings are not altogether surprising, as initial management is more likely to be driven by clinical factors, including patient characteristics and illness severity, than by the findings of a single diagnostic test. Administrative and logistical factors might also affect the time to admission or discharge. Overall LOS was significantly shorter after the introduction of FilmArray GI panel testing, as was the time to initiate appropriate optimal antibiotics. These findings suggest that the use of the FilmArray GI panel as part of the initial evaluation of patients with acute diarrhea of suspected infectious etiology may result in cost savings.
The role of multiplex PCR in antibiotic stewardship remains mixed. On one hand, a positive result in the absence of proper clinical interpretation may lead to unnecessary treatment. The high proportion of patients in our study who received treatment for organisms found on the FilmArray GI panel alone (not confirmed by culture results) may speak to the propensity of physicians to treat reflexively based on reported test results. On the other hand, in the era of stool culture-based diagnosis alone, only 37% of patients in our sample were started on the optimal medication right away, whereas after the introduction of the FilmArray GI panel this increased to 52%, and optimal treatment occurred sooner in their hospital course. By enabling physicians to avoid starting antibiotics where not indicated (e.g. viral enteritis or certain E. coli types), and targeting therapy to pathogen earlier in the treatment course, this methodology can support the aims of antibiotic stewardship.
Our study has several limitations. The number of patients meeting inclusion criteria was unequally affected by miscoding (i.e. patient charts initially selected based on diagnostic codes, which were then excluded), resulting in an especially low number in the control group, and compromising the overall statistical power. By its very nature, a retrospective review is dependent upon correct coding, and it is likely that some patients were incorrectly diagnosed and their charts could not be accessed. The high prevalence of CDI-related diagnoses in our pre-FilmArray group may reflect a sampling bias in which cases of suspected infectious diarrhea were not coded as such in the absence of a specific pathogen (in turn reflecting the low diagnostic yield of stool cultures) and thus overlooked when screening patient charts for inclusion in the study. We adjusted for these issues to the best of our ability by using more conservative statistical methods.