Comparative Evaluation of Enteric Bacterial Culture and a Molecular Multiplex Syndromic Panel in Children with Acute Gastroenteritis

Although enteric multianalyte syndromic panels are increasingly employed, direct comparisons with traditional methods and the inclusion of host phenotype correlations are limited. Luminex xTAG gastrointestinal pathogen panel (GPP) and culture results are highly concordant.

ABSTRACT

Although enteric multianalyte syndromic panels are increasingly employed, direct comparisons with traditional methods and the inclusion of host phenotype correlations are limited. Luminex xTAG gastrointestinal pathogen panel (GPP) and culture results are highly concordant. However, phenotypic and microbiological confirmatory testing raises concerns regarding the accuracy of the GPP, especially for Salmonella spp. A total of 3,089 children with gastroenteritis submitted stool specimens, rectal swab specimens, and clinical data. The primary outcome was bacterial pathogen detection agreement for shared targets between culture and the Luminex xTAG GPP. Secondary analyses included phenotype assessment, additional testing of GPP-negative/culture-positive isolate suspensions with the GPP, and in-house and commercial confirmatory nucleic acid testing of GPP-positive/culture-negative extracts. The overall percent agreement between technologies was >99% for each pathogen. Salmonella spp. were detected in specimens from 64 participants: 12 (19%) by culture only, 9 (14%) by GPP only, and 43 (67%) by both techniques. Positive percent agreement for Salmonella spp. was 78.2% (95% confidence interval [CI], 64.6%, 87.8%). Isolate suspensions from the 12 participants with specimens GPP negative/culture positive for Salmonella tested positive by GPP. Specimens GPP positive/culture negative for Salmonella originated in younger children with less diarrhea and more vomiting. GPP-positive/culture-negative specimen extracts tested positive using additional assays for 0/2 Campylobacter-positive specimens, 0/4 Escherichia coli O157-positive specimens, 0/9 Salmonella-positive specimens, and 2/3 Shigella-positive specimens. For both rectal swab and stool samples, the median cycle threshold (C_T) values, determined using quantitative PCR, were higher for GPP-negative/culture-positive samples than for GPP-positive/culture-positive samples (for rectal swabs, 36.9 [interquartile range {IQR}, 33.7, 37.1] versus 30.0 [IQR, 26.2, 33.2], respectively [P = 0.002]; for stool samples, 36.9 [IQR, 33.7, 37.1] versus 29.0 [IQR, 24.8, 30.8], respectively [P = 0.001]). GPP and culture have excellent overall agreement; however, for specific pathogens, GPP is less sensitive than culture and, notably, identifies samples false positive for Salmonella spp.

INTRODUCTION

Bacterial enteric pathogens continue to cause substantial morbidity worldwide (1, 2). Although conventional stool culture is classically used to identify bacterial enteropathogens in clinical microbiology laboratories, such methods have drawbacks, including prolonged turnaround times and a reliance on diverse selective and enrichment media, identification tests, and expertise (3). In contrast, nucleic acid amplification techniques (NAAT) can detect multiple pathogens within hours (4). However, questions remain about their postmarket accuracy when evaluating a range of enteropathogens across diverse populations and regions (5). Published data have limited relevance to North American children, as most reports have emerged from low- and middle-income countries (5–7), focused primarily on adults (4, 6, 8–12), did not compare NAAT results to standard culture results (7–10, 13), applied inconsistent testing protocols (4, 8, 11, 14), or analyzed single pathogens (15). No studies have included host phenotypes in interpreting discordant results.

Here, we determine the agreement for the bacterial pathogens of interest between stool bacterial culture and the Luminex xTAG gastrointestinal pathogen panel (GPP; Luminex Molecular Diagnostics, Austin, TX, USA) NAAT platform in a systematically tested cohort of children with acute gastroenteritis in Alberta, Canada. Secondarily, we evaluate discordant samples by analyzing clinical phenotypes and determine GPP and culture accuracy through additional testing.

MATERIALS AND METHODS

Population.

This prospective cohort study was conducted as part of the Alberta Provincial Pediatric EnTeric Infection TEam (APPETITE) project (16). Eligible children were enrolled through the Alberta Children’s Hospital (Calgary, AB, Canada) and Stollery Children’s Hospital (Edmonton, AB, Canada) emergency departments (EDs) and a provincial nursing triage telephone advice line (Health Link) (17).

Eligible participants were <18 years of age and had had ≥3 episodes of vomiting and/or diarrhea in the preceding 24 h and <7 days of symptoms. Children were excluded if they were enrolled in the previous 14 days, unable to participate in follow-up, had significant psychiatric comorbidities or neutropenia, or were critically ill. Children recruited through Health Link did not require medical attention, and supportive care at home was recommended.

Approval was granted by the University of Calgary and University of Alberta research ethics boards. Caregivers provided informed consent; assent was obtained when appropriate.

Outcome measures.

The primary outcome was agreement, measured as overall percent agreement, positive percent agreement (PPA), and Cohen’s ĸ, between stool bacterial culture and the GPP for bacterial pathogens sought by both detection methods: Campylobacter spp., Escherichia coli O157, Salmonella spp., and Shigella spp. Yersinia enterocolitica was not included because of negligible detection rates.

Secondary outcomes focused on Salmonella spp. and consisted of a comparison of the clinical phenotypes of patients with concordant and discordant results and three confirmatory tests. First, GPP-negative/culture-positive isolates were tested using the GPP. Second, an in-house real-time quantitative PCR (RT-qPCR) assay was performed on nucleic acid extracts of specimens GPP positive/culture positive and GPP negative/culture positive for Salmonella spp. to confirm the initial concordant result and to evaluate the relationship between the cycle threshold (C_T) value and pathogen identification. Third, GPP-positive/culture-negative specimens were tested with the in-house RT-qPCR and a different commercial assay (the Prodesse ProGastro SCSS assay; Hologic Inc., San Diego, CA) (18) on a SmartCycler II instrument (Cepheid, Sunnyvale, CA) to confirm or refute the presence of Salmonella spp., Shigella spp., and Shiga toxin-producing E. coli (STEC) (see “Confirmatory testing” below for details).

Specimen acquisition.

Two rectal swab specimens were obtained from each ED participant (FLOQSwab; Copan Italia, Brescia, Italy) (19): one was placed in 2 ml of modified Cary-Blair transport medium and used for bacterial culture, and the other was placed into a sterile tube without medium and tested (off-label) using the GPP. Stool samples were collected in sterile containers (V302-F; Starplex Scientific Inc., ON, Canada). If they were not provided prior to discharge, stool samples were collected by the parents at home and placed in the same sterile containers described above, which were retrieved by a courier service.

Health Link participants received specimen collection kits by courier. Specimens were collected per ED protocols and were returned to the clinical microbiology laboratory by a study-funded courier service within 12 h of collection. All specimens were placed in coolers with ice packs while in transit to the laboratory. Bacterial cultures were inoculated per Public Health Laboratory (ProvLab) protocols as soon as possible following specimen arrival, after which both the rectal swabs and stool specimens were stored at −80°C until molecular testing was performed.

Specimen processing.

(i) Culture. The FLOQSwab rectal swab specimens were vortexed for 30 s, and 100 μl of transport medium was transferred onto MacConkey agar with crystal violet (Dalynn Biologicals, Calgary, AB, Canada), sheep blood agar (Oxoid, Thermo Fisher Scientific, Waltham, MA, USA), Hektoen agar (Dalynn Biologicals, Calgary, AB, Canada), cefsulodin, irgasan, and novobiocin agar (Dalynn Biologicals, Calgary, AB, Canada), CHROMagar O157 with 2.5 mg/liter potassium tellurite (CHROMagar, Paris, France), and Campylobacter blood-free selective agar (Dalynn Biologicals, Calgary, AB, Canada) plates. Overnight enrichment was performed by adding ∼200 μl of bulk stool to mannitol selenite broth followed by culture on Salmonella-Shigella (Dalynn Biologicals, Calgary, AB, Canada) and Wilson-Blair (ProvLab) agar. Campylobacter plates were incubated microaerobically (in 6.0% O₂, 7.1% CO₂, 3.6% H₂, and 83.3% N₂ at 42°C for up to 72 h); all other media were incubated in atmospheric oxygen (35°C, 24 h). A quantity of stool samples of <1 g was considered insufficient and not subjected to culture. Identification of isolates was performed per routine laboratory confirmation protocols using an API 20E system (bioMérieux Inc., USA) in ProvLab, Calgary, AB, Canada, while a Vitek MS system (bioMérieux, St-Laurent, QC, Canada) and supplemental biochemical methods were used in ProvLab, Edmonton, AB, Canada. Salmonella serotyping was performed using the Check & Trace Salmonella assay (Check-Points, Netherlands) (20). All positive cultures were skimmed and frozen at −80°C.

(ii) Molecular diagnostics. The GPP is a qualitative multiplex molecular-based syndromic panel that identifies nine bacterial targets (Campylobacter spp., Clostridioides [formerly Clostridium] difficile toxin A/B, E. coli O157, enterotoxigenic E. coli [ETEC], Shiga toxin-producing E. coli [STEC] stx₁ and stx₂, Salmonella spp., Shigella spp., Vibrio cholerae, Yersinia spp.) (8, 21). GPP testing was performed at a single site (ProvLab, Edmonton, AB, Canada). Immediately before nucleic acid extraction, stools were thawed and aliquots were made, and the aliquots were then refrozen at −80°C. Rectal swab samples remained frozen until extraction. Samples were batch tested two to three times a week, depending on specimen volumes.

The FLOQSwab rectal swab specimens were vortexed in 500 μl of NucliSENS lysis buffer, and 300 μl of this lysate was added to 700 μl of lysis buffer in Bertin SK38 solid grinding lysis bead tubes with 10 μl of bacteriophage MS2 (each from Luminex Molecular Diagnostics, Toronto, ON, Canada). Similarly, 100 to 150 mg or 100 μl of solid or liquid stool, respectively, was suspended in a final volume of 1 ml of lysis buffer. Two hundred microliters of prepared lysates was extracted using a NucliSENS easyMAG system (bioMérieux, Marcy l’Etoile, France). Total nucleic acids were eluted in a volume of 70 μl and tested per the recommendations for the xTAG GPP kit. All extracts were stored at −80°C until confirmatory testing.

Confirmatory testing.

(i) Testing of GPP-negative/culture-positive Salmonella isolates. Salmonella species isolates were cultured overnight (37°C) on sheep blood agar, and a single colony was transferred into Bertin SK38 tubes containing 1 ml of NucliSENS lysis buffer. Lysis and extraction were performed per the recommendations for the testing of stool by the xTAG GPP.

We also performed RT-qPCR using a TaqMan chemistry-based quantitative PCR (qPCR) targeting a conserved region of the Salmonella invA gene. The forward primer (CTGCGGTACTGTTAATTAC), reverse primer (GAACGTGGCGATAATTTC), and dual-quenched probe (6-FAM-CGGCATCGG/ZEN/CTTCAATCAAGA-Iowa Black FQ; where 6-FAM is 6-carboxyfluorescein) (Integrated DNA Technologies [IDT], Coralville, IA, USA) were designed using Beacon Designer (version 8.20) software (Premier Biosoft, Palo Alto, CA, USA). Five-microliter nucleic acid extracts were used in a 20-μl total qPCR volume with 2× PrimeTime gene expression master mix (IDT, Coralville, IA, USA) with a 0.222 μM final probe concentration and a 0.333 μM final primer concentration. A fast cycling protocol with an initial 95°C polymerase activation step for 3 min, followed by 40 cycles of 95°C for 5 s and 60°C for 30 s, was performed on an Applied Biosystems 7500 Fast instrument. Positive controls with DNA extracted from Salmonella enterica serovar Enteritidis and no-template controls were integrated into each qPCR run. qPCR assay optimization included evaluation against a specificity panel of four clinical Salmonella serotypes (Salmonella Enteritidis, Salmonella Heidelberg, Salmonella Typhimurium, Salmonella Newport), Yersinia enterocolitica (ATCC 9610), Shigella sonnei (clinical isolate A79), and Shigella flexneri (ATCC 12022).

(ii) RT-qPCR with nucleic acid extracts of GPP-positive/culture-positive specimens. As described above for GPP-negative/culture-positive Salmonella isolates, we performed RT-qPCR to confirm the result for GPP-positive/culture-positive specimens and to validate our assay.

(iii) Evaluation of GPP-positive/culture-negative specimens. The frozen DNA extracts used on the Luminex xTAG GPP were thawed and tested using the Prodesse ProGastro SCSS assay (Hologic Inc., San Diego, CA) on a SmartCycler II instrument (Cepheid, Sunnyvale, CA) per manufacturer guidelines. The latter tests for Salmonella, Shigella, and Campylobacter nucleic acids and Shiga toxin 1 and 2 genes (18).

As described above, we performed RT-qPCR to evaluate specimens GPP positive/culture negative for Salmonella spp. We performed a similar evaluation for Shigella spp., STEC, and E. coli O157. The primers and probes used are described in Appendix S1 in the supplemental material.

Statistical analysis.

Test results were categorized as positive or negative, with no indeterminant readings. We did not differentiate between specimen types because rectal swabs have diagnostic capabilities similar to those of stool samples (19). Only participants whose specimens underwent culture and GPP testing were analyzed.

Overall percent agreement between the GPP and bacterial culture was determined for the pathogens of interest (22). For assessing positive response rates, PPA was calculated as [(A)/(A + C)] × 100%, where A is the number of specimens with concordant positive results and C is the number of specimens with GPP-negative/culture-positive results (22). PPA and overall percent agreement were chosen over sensitivity and specificity per U.S. Food and Drug Administration guidelines when evaluating diagnostic tests in the absence of a gold standard (22). Cohen’s κ value was calculated to measure the extent of agreement (23). Clinical characteristics were compared using a chi-square test and the Kruskal-Wallis H test, as appropriate. The Mann-Whitney U test was used to compare RT-qPCR C_T values of the rectal swab and stool specimen groups GPP positive/culture positive and GPP negative/culture positive for Salmonella spp.

We did not use imputation because data from only seven participants were incomplete; we did include all available data from these participants in our analysis. SPSS (version 24.0; IBM Corp., Armonk, NY) was used to perform analyses. A single 2-tailed P value was used to assess differences among all three groups, and the significance level was set at 0.05.

RESULTS

Specimens from 3,089 (88.0%) of the 3,511 participants enrolled between 12 December 2014 and 31 March 2018 underwent GPP and bacterial culture (Fig. 1). Ninety-one (3.0%) and 88 (2.9%) of these specimens, respectively, were positive for at least one of Campylobacter spp., E. coli O157, Salmonella spp., or Shigella spp.

FIG 1.

Flow diagram of potential study participants and enteropathogen detection methods. GPP, gastrointestinal pathogen panel.

Primary outcome.

The overall percent agreement was >99% for each individual bacterial target and 98.9% (95% confidence interval [CI], 98.5%, 99.3%) for all targets combined (Table 1). PPA ranged from 78.2% for Salmonella spp. (95% CI, 64.6%, 87.8%) to 100% for E. coli O157 (95% CI, 51.7%, 100%). Overall, PPA was 83.0% (73/88; 95% CI, 73.1%, 89.8%). Cohen’s ĸ was >0.70 for E. coli O157, Shigella spp., and Salmonella spp. and almost perfect for Campylobacter spp. (ĸ = 0.89).

TABLE 1.

PPA and overall percent agreement between the Luminex xTAG GPP and bacterial culture of stool and rectal swab specimens^a

Bacterial targets	Luminex xTAG GPP result	No. of specimens with the following bacterial culture result:		PPA (95% CI)	Overall % agreement (95% CI)	Cohen’s ĸ
		Positive	Negative
Campylobacter spp.	Positive	20	2	87.0 (65.3, 96.6)	99.8 (99.6, 99.9)	0.89
	Negative	3	3,064
Escherichia coli O157	Positive	6	4	100 (51.7, 100)	99.9 (99.6, 99.9)	0.75
	Negative	0	3,079
Salmonella spp.	Positive	43	9	78.2 (64.6, 87.8)	99.3 (98.9, 99.6)	0.80
	Negative	12	3,025
Shigella spp.	Positive	5	3	100 (46.3, 100)	99.9 (99.7, 99.97)	0.77
	Negative	0	3,081
Total	Positive	73	18	83.0 (73.1, 89.8)	98.9 (98.5, 99.25)	0.81
	Negative	15	2,983

^aThe bacterial targets included in this table were identified by both the Luminex xTAG gastrointestinal pathogen panel (GPP) and bacterial culture.

Secondary outcomes for Salmonella.

The most frequently identified pathogen, Salmonella spp., was detected in 64 participants (Fig. 2). Salmonella spp. were detected in 80% (51/64) of submitted rectal swabs (positive results, n = 5 for GPP only, n = 12 for culture only, n = 34 for both) and 84% (43/51) of submitted stool specimens (positive results, n = 7 for GPP only, n = 8 for culture only, n = 28 for both). Salmonella Enteritidis accounted for 48% (31/64) of the Salmonella spp. identified by culture (Table 2). Sixty-seven percent (43/64) of the participants with Salmonella spp. detected in swab or stool produced concordant GPP and culture test results; 19% (12/64) were GPP negative/culture positive, and 14% (9/64) were GPP positive/culture negative. Rectal swabs and stool samples had similar detection rates for Salmonella spp., with the exception of GPP-positive/culture-negative specimens, where only 33% (3/9) of the rectal swabs but 88% (7/8) of the stool samples were positive (Table 3).

FIG 2.

Confirmatory testing flow diagram of Salmonella species-positive specimens. All available specimens were tested. Participants may have submitted both a swab and a stool sample. GPP, gastrointestinal pathogen panel; qPCR, quantitative PCR; †, GPP-negative/culture-positive samples were not retested with the commercial assay.

TABLE 2.

Serotype analysis of Salmonella species-positive specimens on the Luminex xTAG GPP and bacterial culture

Salmonella serovar	No. (%) of specimens with the following result:
	Total	Luminex xTAG GPP positive/culture negative	Luminex xTAG GPP negative/culture positive	Luminex xTAG GPP positive/culture positive
Untyped or nontypeablea	10 (15.6)	9	0	1
Salmonella Agona	1 (1.6)	0	0	1
Salmonella Alachua	1 (1.6)	0	0	1
Salmonella Anatum	1 (1.6)	0	1	0
Salmonella Bareilly	1 (1.6)	0	0	1
Salmonella enterica subsp. enterica	4 (6.3)	0	0	4
Salmonella enterica subsp. salamae (II)55:k:z39	1 (1.6)	0	0	1
Salmonella Enteritidis	31 (48.4)	0	7	24
Salmonella Heidelberg	1 (1.6)	0	0	1
Salmonella Infantis	2 (3.1)	0	1	1
Salmonella Newport	1 (1.6)	0	0	1
Salmonella Reading	2 (3.1)	0	0	2
Salmonella Rubislaw	1 (1.6)	0	0	1
Salmonella Schwarzengrund	1 (1.6)	0	0	1
Salmonella Thompson	1 (1.6)	0	0	1
Salmonella Typhi	1 (1.6)	0	0	1
Salmonella Typhimurium	3 (4.7)	0	2	1
Salmonella Virchow	1 (1.6)	0	1	0
Total	64 (100)	9 (14.1)	12 (18.8)	43 (68.8)

^aOf these, isolates were detected in only 9/10 specimens by the Luminex xTAG GPP. The final specimen was positive for a Salmonella sp. on GPP and culture, but the serovar was not identified.

TABLE 3.

Comparison of rectal swab and stool specimen results for Salmonella spp. grouped by diagnostic testing approach^b

Salmonella species-positive group	Specimen type	No. of specimens with the indicated result/total no. of specimens tested (%)
		Luminex xTAG GPP positive	Culture positive	qPCR positive
GPP positive/culture positive	Stool	34/38 (89.5)	31/38 (81.6)	36/38 (94.7)
	Rectal swab	36/43 (83.7)	38/43 (88.4)	39/43 (90.7)
GPP positive/culture negative	Stool	7/8 (87.5)a	0/8 (0)	0/8 (0)
	Rectal swab	3/9 (33.3)a	0/9 (0)	0/9 (0)
GPP negative/culture positive	Stool	0/5 (0)	5/5 (100)	1/5 (20.0)
	Rectal swab	0/12 (0)	8/12 (66.7)	4/12 (33.3)

^aP = 0.05. No other comparisons achieved statistical significance.

^bGPP, gastrointestinal pathogen panel; qPCR, quantitative PCR.

Participants in the group GPP positive/culture positive for Salmonella spp. were older (median age, 38 months [interquartile range {IQR}, 13, 63 months]) than those in the GPP-positive/culture-negative group (median age, 7 months [IQR, 5, 16 months]) (Table 4). The latter group of children had significantly more vomiting episodes in the 24 h before enrollment (median, 6 [IQR, 2, 13]) than those with concordant positive specimens (median, 1 [IQR, 0, 3]) or GPP-negative/culture-positive specimens (median, 1 [IQR, 0, 4]). All 43 GPP-positive/culture-positive specimens were provided by children who presented with diarrhea; only 44% (4/9) of GPP-positive/culture-negative children had had diarrhea prior to enrollment (P < 0.001). The maximum number of diarrheal episodes per 24-h period before enrollment was greater in children with specimens with concordant results (median, 10 [IQR, 7, 15]) than in children with GPP-positive/culture-negative specimens (median, 0 [IQR, 0, 9]) (P = 0.002).

TABLE 4.

Clinical characteristics and enteropathogen codetection among participants with concordant and discordant Salmonella species test results^g

Clinical feature	Luminex xTAG GPP positive/culture positive		Luminex xTAG GPP positive/culture negative		Luminex xTAG GPP negative/culture positive		P value
	No. of specimens	Value	No. of specimens	Value	No. of specimens	Value
Age (mo)b	43	38 (13, 63)	9	7 (5, 16)	12	31 (11, 66)	0.006
Time since symptom onset (h)b ,c	43	84 (45, 102)	9	49 (13, 66)	12	56 (31, 87)	0.05
No. (%) of patients with vomitinga	43	23 (54)	9	7 (78)	12	6 (50)	0.44
Maximum no. of vomiting episodes in any 24-h periodb	43	1 (0, 3)	9	6 (2, 13)	12	1 (0, 4)	0.02
Duration of vomiting at time of enrollment (h)b	43	2 (0, 75)	9	16 (2, 66)	12	0.2 (0, 37)	0.47
No. (%) of patients with diarrheaa	43	43 (100)	9	4 (44)	12	9 (75)	<0.001
Maximum no. of diarrhea episodes in any 24-h periodb	43	10 (7, 15)	9	0 (0, 9)	12	7 (1, 12)	0.002
Duration of diarrhea at time of enrollment (h)b	43	83 (37, 101)	9	0 (0, 65)	12	34 (1, 87)	0.02
No. (%) of patients with fevera	43	32 (74)	9	6 (67)	12	6 (50)	0.14
Maximum pain level in prior 24 hb ,d	39	7 (5, 8)	7	4 (0, 8)	11	8 (5, 10)	0.15
No. (%) of specimens with non-Salmonella pathogens presenta ,e	43	11 (26)	9	6 (67)	12	3 (25)	0.05
Median (IQR) qPCR CT value
Rectal swab	39	30 (26, 33)	0	NA	4	37 (34, 37)	0.002f
Stool	36	29 (25, 31)	0	NA	1	38	0.05f

^aP value by the chi-square test across all three groups.

^bP value by the Kruskal-Wallis H test across all three groups.

^cDefined by the time of onset of either vomiting or diarrhea, whichever was greater.

^dPain level was reported by the parents of the participants on a scale ranging from 0 (no pain at all) to 10 (worst pain imaginable) in the 24 h prior to enrollment (only for patients enrolled in the emergency departments).

^eThis included detection of any of the following targets by GPP, the gastroenteritis virus panel, and/or stool culture in any of the specimens (i.e., swab, stool): adenovirus, astrovirus, norovirus, rotavirus, sapovirus, Aeromonas, Campylobacter, C. difficile, V. cholerae, E. coli O157, ETEC, Shigella, STEC, Yersinia, Cryptosporidium, Entamoeba, or Giardia.

^fMann-Whitney U test for comparison of two groups.

^gGPP, gastrointestinal pathogen panel; qPCR, quantitative PCR; NA, not available.

Confirmatory testing.

Five of 5 (100%) stool specimens and 8/8 (100%) rectal swab specimens GPP negative/culture positive for Salmonella isolates were positive by confirmatory testing by GPP (Fig. 2). Salmonella qPCR of GPP-negative/culture-positive extracts was positive for 4/12 (25%) rectal swab specimens and 1/5 (20%) stool specimens. Of the concordant GPP-positive/culture-positive specimens, 36/38 (94.7%) stool samples and 39/43 (90.7%) rectal swab specimens were positive on qPCR, with median cycle threshold (C_T) values being 29.0 (IQR, 24.8, 30.8) for stool and 30.0 (IQR, 26.2, 33.2) for rectal swabs (P = 0.02 for the 32 paired specimens). Median C_T values were higher for GPP-negative/culture-positive than for GPP-positive/culture-positive Salmonella rectal swab (36.9 [IQR, 33.7, 37.1] versus 30.0 [IQR, 26.2, 33.2]; P = 0.002) and stool specimen extract (36.9 [IQR, 33.7, 37.1] versus 29.0 [IQR, 24.8, 30.8]; P = 0.001).

Retesting of GPP-positive/culture-negative specimen extracts yielded the same results on both the Prodesse ProGastro SCSS and in-house qPCR assays (except for Campylobacter spp., for which in-house primers were not developed): 0/4 E. coli O157, 0/9 Salmonella species, and 2/3 Shigella species isolates were positive for the bacteria detected initially on the GPP. Both Campylobacter species-positive specimens were negative on commercial PCR testing (Table 5).

TABLE 5.

Analysis of specimens Luminex xTAG GPP positive/bacterial culture negative for Campylobacter spp., Escherichia coli O157, Salmonella spp., and Shigella spp. on commercial and in-house PCR assays

Bacteria with positive test result on GPP	No. of specimens	Commercial assay retest result	In-house PCR result
Campylobacter spp.	2	Negative	NAa
Escherichia coli O157	4	Negative	Negative
Salmonella spp.	9	Negative	Negative
Shigella spp.	2	Positive	Positive
Shigella spp.	1	Negative	Negative

^aNA, not applicable (our in-house assay did not include primers for Campylobacter spp.).

DISCUSSION

We found excellent overall agreement between the GPP and bacterial culture. For Salmonella spp., however, GPP-positive/culture-negative children differed significantly from those for whom the results were concordant. Confirmatory analyses demonstrated that GPP-positive/culture-negative specimens likely are false positive, while GPP-negative/culture-positive specimens reflect the inability of the assay to detect Salmonella spp. when small quantities of nucleic acid are present.

A meta-analysis comparing a variety of multiplex gastrointestinal panels with standard microbiology methods reported that PPA ranged from 68% when NAAT methods provide the benchmark to 93% when conventional methods are considered the gold standard (24). However, the high PPA obtained when conventional methods served as the benchmark was inconsistent across pathogens, a concern that we identified as it relates to Salmonella spp. A Vietnamese report identified a lack of specificity of Salmonella species detection using the GPP (5), potentially due to a high rate of Salmonella carriage (25) or amplification of DNA from non-Salmonella strains. In a U.S. multicenter study, 14% of Salmonella spp. detected by GPP could not be confirmed by gold standard methods (26).

Given that specimens GPP negative/culture positive for Salmonella had higher C_T values than GPP-positive/culture-positive specimens, our data do not support the hypothesis that culture fails to detect pathogens because it is less sensitive. While prolonged storage under suboptimal conditions may permit nucleic acid degradation (27) and, hence, a lower yield on NAAT platforms, in our study, all samples were processed within 7 days of receipt and both aliquots and extracts were stored at −80°C at all times. This approach led to our finding a PPA of 78%, which is aligned with previous reports (27, 28). While the inability to detect some pathogens on the GPP could reflect inefficient nucleic acid extraction or inhibition, many of the same nucleic acid extracts yielded positive results with our in-house assay and the alternate platform. Our RT-qPCR results and our retesting of culture-positive Salmonella isolates on the GPP lead us to believe that the discrepancies represent a sensitivity threshold, with there being less target pathogen nucleic acid present in the GPP-negative/culture-positive group.

The clinical presentations of participants GPP positive/culture negative for Salmonella spp. were at variance with those of individuals with classic Salmonella infection (29). Retesting of these specimens using commercial and in-house assays provided negative results, indicating that the GPP results were likely false positive. All specimens GPP positive/culture negative for E. coli O157 were also negative when tested with an alternate commercial and in-house assay, suggesting an overall suboptimal specificity for these targets. Concerns regarding false positives due to cross-reactivity with commensal bacteria have also been reported with the BioFire FilmArray gastrointestinal panel (4). Other possible causes for false positives on the NAAT include cross contamination, amplification of a target from a non-Salmonella strain, or detection of Salmonella spp. in a child who is a carrier. Contamination can be addressed by engineering or procedure modifications, and non-Salmonella species amplification can be minimized by including corroborative loci. For example, in the case of E. coli O157, the presence of conserved loci for intimin (eae) (30) and a component of the O157 side chain synthesis cluster (rfbE) (31) increases confidence that a Shiga toxin gene signal originates in a bona fide pathogen. Nonetheless, these findings highlight the concern regarding false-positive results when a multiplex NAAT is employed and the importance of considering such results with the clinical presentation and codetected pathogens in mind (3, 12).

Although we might be underpowered to detect problems with the identification of E. coli O157 and Shigella spp. by the GPP, the panel identified all culture-positive participants. The rapid identification of these two pathogens is of considerable value. Detecting E. coli O157 infection would prompt the withholding or discontinuation of antibiotics (32) and the reversal of dehydration using parenteral fluids (33–35); the detection of Shigella would appropriately prompt antibiotic therapy (36, 37). Evidence supporting this assertion was provided by a study of the BioFire FilmArray GI panel, which reduced the time to initiation of antimicrobial therapy by 50 h and the time to the discontinuation of therapy in STEC-infected children by 47 h (12).

Selenite enrichment broth was employed as part of our stool culture protocol to enhance the identification of Salmonella spp. (38). By comparison, false-negative results by NAAT may relate to technical issues, including nucleic acid extraction efficiency. Future iterations of NAAT-based multianalyte syndromic panels should consider protocol modifications that enhance diagnostic accuracy, including alternative approaches to extraction, modified amplification or cycling conditions, and the incorporation of additional pathogen loci to reduce the number of panel-positive/culture-negative results.

Given our data, the existing literature, and the adjustments required for NAAT assays, one needs to consider how postmarket monitoring of the performance of multianalyte assays should be performed. Specifically, if laboratories abandon culture detection of enteric pathogens in favor of multianalyte assays, there is little opportunity to compare recovery rates between the two methodologies. Hence, the failure to detect pathogens by multianalyte assays will go unverified. The U.S. Food and Drug Administration has postmarket surveillance mechanisms for drug safety and efficacy, especially for those drugs approved via accelerated pathways (39, 40), but no analogous mechanism exists for diagnostic devices. Our findings urge continuing assessment of the accuracy of multianalyte assays for bacterial enteric pathogens.

To our knowledge, this is the first study to compare GPP with bacterial culture employing samples from an entirely pediatric population in a high-income country. It is also the only study to incorporate the clinical phenotype into the interpretation of discordant specimens. Our study nonetheless has limitations. Despite our large sample size, we detected only 64 participants with Salmonella species-positive specimens, and Campylobacter species-, STEC-, and Shigella species-positive specimens were even less common. The low prevalence of bacterial enteropathogens has implications for our reported percent agreement and κ calculations. Participants were recruited from only two western Canadian cities, so our findings cannot be automatically generalized to locations with a different epidemiology. We, unfortunately, do not have data to enable calculation of the precise time interval between specimen receipt at the laboratory and culture setup. It should also be noted that sampling error may have occurred due to the small volume of nucleic acid extract used and that all specimens underwent a freeze-thaw cycle; both of these operational elements could explain negative GPP and qPCR results in the setting of low bacterial loads.

In conclusion, overall GPP results had excellent concordance with those of bacterial culture, but the PPA was suboptimal for the shared bacterial targets. In particular, Salmonella species identification with the GPP was prone to false positives and negatives. These results have clinical and public health implications. While the GPP platform and other NAAT assays have the potential to provide valuable and credible results, their current accuracy requires additional validation before universal abandonment of culture diagnostics. Careful consideration of the context of the illness in patients whose stool tests yield positive results is required.