Identification of Widespread Antibiotic Exposure in Patients With Cholera Correlates With Clinically Relevant Microbiota Changes

Abstract

Background: A first step to combating antimicrobial resistance in enteric pathogens is to establish an objective assessment of antibiotic exposure. Our goal was to develop and evaluate a liquid chromatography-ion trap mass spectrometry (LC/MS) method to determine antibiotic exposure in patients with cholera.

Methods: A priority list for targeted LC/MS was generated from medication-vendor surveys in Bangladesh. A study of patients with and those without cholera was conducted to collect and analyze paired urine and stool samples.

Results: Among 845 patients, 11% (90) were Vibrio cholerae positive; among these 90 patients, analysis of stool specimens revealed ≥1 antibiotic in 86% and ≥2 antibiotics in 52%. Among 44 patients with cholera and paired urine and stool specimens, ≥1 antibiotic was detected in 98% and ≥2 antibiotics were detected in 84%, despite 55% self-reporting medication use. Compared with LC/MS, a low-cost antimicrobial detection bioassay lacked a sufficient negative predictive value (10%; 95% confidence interval, 6%-16%). Detection of guideline-recommended antibiotics in stool specimens did (for azithromycin; P = .040) and did not (for ciprofloxacin) correlate with V. cholerae suppression. A nonrecommended antibiotic (metronidazole) was associated with decreases in anaerobes (ie, Prevotella organisms; P < .001).

Conclusion: These findings suggest that there may be no true negative control group when attempting to account for antibiotic exposure in settings like those in this study.

Keywords: Vibrio cholerae; AMR; Bangladesh; Diarrhoea; LC/MS; antimicrobial resistance; cholera; diarrhea; mass spectrometry.

Figure 1.

Responses to the question “Name the top three antibiotics you sell for the treatment of diarrheal disease” among 62 rural medication vendors. “Other” refers to amoxicillin (1 respondents; median cost, $0.06/tablet), ampicillin (1; $0.06/tablet), erythromycin (5; $0.10/tablet or $0.76/liquid dose), and nitazoxanide, an antiparasitic medication (3; $0.13/tablet). Pricing for ciprofloxacin, metronidazole, azithromycin, is tetracycline are provided in the text.

Figure 2.

Frequency of antibiotic detection in paired urine (UR; left) and stool (ST; right) samples from patients with cholera, using liquid chromatography/mass spectrometry, self-reported prior medication use (REP), and bioassay with urine extracts (BIO). Dark blue denotes positive results or self-report, light blue denotes trace detection, and white denotes negative results or no self-report. For urine and stool specimens, the number of antibiotics detected per sample is shown at right, and the number of times (%) an antibiotic was detected across samples is shown at the bottom. AM, amoxicillin; AZ, azithromycin; CF, ceftriaxone; CI, ciprofloxacin; ER, erythromycin; ME, metronidazole, TE, tetracycline.

Figure 3.

Frequency of antibiotic detection in stool specimens from patients with cholera and impact on Vibrio cholerae. A, Frequency of antibiotic detection by liquid chromatography/mass spectrometry and self-report (REP) among all patients with cholera from whom stool specimens were collected. The number of times (%) an antibiotic was detected across samples is shown at the bottom. The red lines at left denote positivity for phage, and the black line at right denotes an unknown history. Dark blue, positive result or self-report; light blue, trace detection; white, negative result or no self-report. B, Paired comparisons of total bacteria (16S ribosomal DNA) and V. cholerae by nanoliter quantitative polymerase chain reaction analysis, and the relative abundance of Vibrio species with (Abx+) and without (Abx-) detection of the specified antibiotic. Bars and whiskers denote medians and interquartile ranges, respectively. Analyses were restricted to samples without detection of trace antibiotic and phage. See Supplementary Table 2 for further details. AM, amoxicillin; AZ, azithromycin; CF, ceftriaxone; CI, ciprofloxacin; CT, cycle threshold; ER, erythromycin; ME, metronidazole, TE, tetracycline. *P < .05, by the Mann-Whitney U test, with adjustment for multiple comparisons. ^aSamples negative for ciprofloxacin were also negative for metronidazole.

Figure 4.

Ciprofloxacin is associated with minimal microbiota changes. A, Comparison of stool samples from patients with cholera in which ciprofloxacin was or was not detected; samples with phage or trace antibiotic detection were removed. P = .158, by permutational multivariate analysis of variance (Bray-Curtis dissimilarities). Brachyspira species are in brown. B, Relative abundance of taxa previously shown to be associated with early, middle, and late phases of recovery from cholera, with (Abx+) and without (Abx-) antibiotic detection. No significant differences were detected by the Mann-Whitney U test (α = 0.05). P values were adjusted for multiple comparisons. Bars and whiskers denote medians and interquartile ranges, respectively. Analyses were restricted to samples without detection of trace antibiotic and phages. See Supplementary Table 2 for further details. Bacteroid, Bacteroides; Enterob, Enterobacteriaceae (family); Enteroc, Enterococci; Prevotel, Prevotella; Roseb, Roseburia; Streptoc, Streptococci. ^aSamples negative for ciprofloxacin were also negative for metronidazole. ^bEnterobacteriaceae is grouped with genera associated with the early phase of recovery because the family contains Escherichia, which is known to be associated with this phase.

Figure 5.

Metronidazole is associated with relative decreases in anaerobes, including Prevotella species. A, Comparison of stool samples from patients with cholera in which metronidazole was or was not detected; samples with phage or trace antibiotic detection were removed. P = .001, by permutational multivariate analysis of variance (Bray-Curtis dissimilarities). Prevotella species are in purple. B, Relative abundance of taxa previously shown to be associated with early, middle, and late phases of recovery from cholera, with (Abx+) and without (Abx-) antibiotic detection. Bars and whiskers denote medians and interquartile ranges, respectively. Analyses were restricted to samples without detection of trace antibiotic and phage. See Supplementary Table 2 for further details. *P < .05 and ***P < .001, by the Mann-Whitney U test, with adjustment for multiple comparisons. Bacteroid, Bacteroides; Enterob, Enterobacteriaceae (family); Enteroc, Enterococci; Prevotel, Prevotella; Roseb, Roseburia; Streptoc, Streptococci.

Figure 6.

Azithromycin is not associated with major taxonomic changes other than suppression of Vibrio cholerae. A, Comparison of stool samples from patients with cholera in which azithromycin was or was not detected; samples with phage or trace antibiotic detection were removed. P = .041, by permutational multivariate analysis of variance (Bray-Curtis dissimilarities). Vibrio species are in blue. B, Relative abundance of taxa previously shown to be associated with early, middle, and late phases of recovery from cholera, with (Abx+) and without (Abx-) antibiotic detection. No significant differences were detected by the Mann-Whitney U test (α = 0.05). P values were adjusted for multiple comparisons. Bars and whiskers denote medians and interquartile ranges, respectively. Analyses were restricted to samples without detection of trace antibiotic and phage. See Supplementary Table 2 for further details. Bacteroid, Bacteroides; Enterob, Enterobacteriaceae (family); Enteroc, Enterococci; Prevotel, Prevotella; Roseb, Roseburia; Streptoc, Streptococci. ^aEnterobacteriaceae is grouped with genera associated with the early phase of recovery because the family contains Escherichia, which is known to be associated with this phase.