Effect of Bismuth Subsalicylate vs Placebo on Use of Antibiotics Among Adult Outpatients With Diarrhea in Pakistan: A Randomized Clinical Trial.

Importance: Many of the 4.5 billion annual episodes of diarrhea are treated unnecessarily with antibiotics; prevalence of antibiotic resistance among diarrheal pathogens is increasing. Knowledge-based antibiotic stewardship interventions typically yield little change in antibiotic use. Objective: To compare antibiotic use among adult outpatients with diarrhea given bismuth subsalicylate (BSS) or placebo. Design, Setting, and Participants: This randomized clinical trial took place from April to October 2014. Participants were patients aged 15 to 65 years with acute, nonbloody diarrhea from 22 outpatient clinics in Karachi, Pakistan. Participants were interviewed about symptoms and health care utilization during the 5 days after enrollment. Group assignment was concealed from participants, field staff, and the statistician. Primary analysis occurred from August to September 2015. Interventions: Participants were randomly assigned (1:1) to receive BSS or placebo for 48 hours or less. Main Outcomes and Measures: Use of systemic antibiotics within 5 days of enrollment. Secondary outcomes included measures of duration and severity of illness.
Results: Among eligible patients, 39 declined to participate, 440 enrolled, and 1 enrolled participant was lost to follow-up, for a total of 439 patients included in the analysis. Median (interquartile range) participant age was 32 (23-45) years and 187 (43%) were male. Two hundred twenty patients were randomized to BSS and 220 were randomized to placebo. Overall, 54 participants (12%) used systemic antibiotics (16% in the placebo group and 9% in the BSS group); all antibiotic use followed consultation with a physician. Use of any antibiotic was significantly lower in the BSS group (20 of 220 vs 34 of 219 patients; odds ratio [OR], 0.54; 95% CI, 0.30-0.98), as was use of fluoroquinolones (8 of 220 vs 20 of 219 patients; OR, 0.38; 95% CI, 0.16-0.88). Rates of care seeking and hospitalization were similar between groups and no difference was detected in timing of diarrhea resolution. However, those in the BSS group less commonly received intravenous rehydration (14 of 220 vs 27 of 219 patients; OR, 0.48; 95% CI, 0.25-0.95) and missed less work (median [interquartile range], 0 [0-1] vs 1 [0-1] day; P = .04) during follow-up. Conclusions and Relevance: This study found less antibiotic use among participants given BSS for acute diarrhea in a setting where antibiotics are commonly used to treat diarrhea. Encouraging health care professionals in such settings to recommend BSS as frontline treatment for adults with diarrhea, and promoting BSS for diarrhea self-management, may reduce antibiotic use and rates of antibiotic resistance globally. Trial Registration: ClinicalTrials.gov identifier: NCT02047162.

RCT Entities:   Population Interventions Outcomes

Introduction

The World Health Organization has declared antimicrobial resistance a global crisis.[1] Inappropriate use of antibiotics, an important driver of antibiotic resistance, is common globally. Although acute respiratory infections are the primary condition for which antibiotics are prescribed inappropriately in the United States,[2,3] globally, many of the estimated 4.5 billion annual episodes of diarrhea are inappropriately treated with antibiotics.[4] In South Asia, most ambulatory patients with acute diarrhea may be treated with antibiotics,[5,6,7] often in conflict with national practice guidelines,[8] and the prevalence of multidrug resistance among enteric pathogens is high.[9,10,11] Antibiotic resistance can lead to treatment dilemmas and increased morbidity and mortality among infected populations and rapidly spread worldwide through globalized markets and frequent international travel.[12,13,14,15,16]

Optimization of antibiotic use is a primary strategy in the global action plan on antibiotic resistance; however, traditional stewardship efforts have had limited success.[1,17] Evidence suggests that patients and health care professionals can be redirected toward nonantibiotic medications to satisfy the desire to actively treat the illness, relieve symptoms, and permit the patient to return to work and other activities sooner. For example, among children given oral rehydration therapy for diarrhea, adding zinc to the treatment regimen led to decreased antibiotic use.[18,19] Because many adult patients with diarrhea in Pakistan are treated with antibiotics,[20] but zinc is not indicated for adult patients with diarrhea, other nonantibiotic medications that relieve symptoms may help reduce inappropriate use of antibiotics. Bismuth subsalicylate (BSS), a generic, nonprescription antidiarrheal with mild, local antimicrobial activity, can reduce the duration of gastrointestinal symptoms among children and adults with diarrhea.[21,22,23,24,25] We studied the effect of BSS on use of systemic antibiotics for acute diarrhea among adult outpatients in Pakistan.

Methods

Setting and Participants

Twenty-two general-care outpatient clinics staffed by allopathic physicians in Karachi, Pakistan, participated in this study. We targeted low- to middle-income, multiethnic neighborhoods because we believed diarrhea incidence would be greater than in higher-income neighborhoods, and recruitment would therefore proceed more quickly. Patients were enrolled from April to June and August to October 2014, which encompassed Karachi’s rainy season. We did not enroll participants during July because most staff and laypersons were observing Ramadan. The trial protocol is available in Supplement 1.

Within each clinic, we screened all clinically stable adults who presented with a chief concern of nonbloody diarrhea for less than 48 hours and who were older than 14 years and younger than 66 years (Figure 1). We excluded patients who were pregnant or lactating, reported taking an antidiarrheal or antimicrobial medication during the previous 48 hours, were allergic to aspirin, or had previously enrolled in the study. To facilitate patient follow-up, we enrolled only the first eligible, consenting patient in each clinic each day.

Figure 1.

Flow of Participants Through the Study

Patients could report multiple exclusion criteria.

Staff from Health-Oriented Preventive Education, a Karachi-based nongovernmental organization that provides health care, education, and advocacy for low-resource individuals in Pakistan, performed all nonlaboratory field work.[26]

Ethics

Although study physicians were instructed not to provide participants with antibiotics at enrollment, participants received oral rehydration solution and were encouraged to seek care if their condition worsened. Study staff were also instructed to refer participants for additional care if they reported feeling worse during follow-up interviews. Participants provided written informed consent before enrollment. The study protocol and data collection instruments were approved by the institutional review boards of the Centers for Disease Control and Prevention and Health-Oriented Preventive Education. This article follows the Consolidated Standards of Reporting Trials (CONSORT) reporting guideline for randomized trials.[27]

Intervention, Group Assignment, and Concealment

The intervention consisted of chewable tablets each containing 262 mg of BSS (Pepto-Bismol InstaCool Peppermint; Procter & Gamble). Placebo, produced by the manufacturer of the intervention drug, was identical in appearance, flavor, and packaging but did not contain BSS. Each patient received 32 tablets of the study drug. We instructed participants to take 2 tablets every 30 to 60 minutes as needed for diarrhea, not to exceed 8 doses (16 tablets) per day during the first 48 hours after enrollment. Patients were asked to discontinue the medication when stools became formed. Study staff administered the initial dose of the study medication as directly observed therapy at enrollment; participants were given verbal, written, and pictorial instructions and self-administered subsequent doses as outpatients. All participants also received 2 sachets of oral rehydration solution and instructions for how to prepare and use it.

Participants were randomized 1:1 to BSS or placebo. One of us (A.B.) assigned the allocation sequences, stratified by clinic, using an Excel spreadsheet (Microsoft Corp) to generate random numbers. This investigator labeled masked vials of BSS and placebo by clinic according to the allocation sequence and did not participate in patient selection, data collection, or data entry. We enrolled 20 patients per clinic. Group assignment was triple masked, or concealed from participants, physicians and data collectors, and the principal statistician (B.P.), until primary analysis was complete. While study physicians were asked not to provide antibiotics to participants during the enrollment visit, neither they nor the participants were told the precise purpose of the study.

Outcomes

The a priori primary outcome measure was antibiotic use by study group within 5 days of enrollment. A priori secondary outcome measures included participants’ illness duration and severity, perceived need for antibiotic medications, health care seeking, and satisfaction with the study medication during the 5 days after enrollment. Post hoc analyses investigated associations between diarrhea etiology and antibiotic use; lapse between enrollment and first antibiotic use; ability to participate in typical activities, including work; and number and type of antibiotic classes recommended per clinic visit and patient. To minimize influence on behavior, participants were not informed of the specific objectives of the study.

Data Collection

At enrollment, participants completed a questionnaire about demographic characteristics, including ethnicity (self-identified using investigator-defined options), health status, and current illness, and study physicians recorded clinical data. We asked participants to record on a daily log sheet the dates and times they consumed study medications and passed stools; the log sheet included a modified, pictorial Bristol stool chart with which to classify each stool as diarrheal (types 5-7) or formed (types 1-4).[28] Study staff interviewed participants in their homes approximately 24, 48, and 120 hours after the initial dose of study drug was administered; during each interview, topics included symptoms, care seeking, and self-treatment. Among participants who sought medical care during the follow-up period, we obtained receipts from clinical consultations and prescriptions to verify antibiotic exposures. Because antibiotics can be obtained without prescription in Pakistan, we also assessed antibiotic use by showing participants visual aids that depicted the names and images of 27 locally available antibiotics and asking whether they had taken any of the pictured medications or other antimicrobial medications. At the end of the final interview, we asked participants their perceptions of the study medication and observed the number of tablets remaining.

Laboratory Methods

Participants were asked to provide a fecal specimen before the initial dose of study medication. Specimens were placed in Cary-Blair media and transported within 24 hours of collection to Aga Khan University Laboratory for bacterial culture, identification, and phenotypic antimicrobial susceptibility testing.

Statistical Analysis

We estimated sample size based on the primary end point of antibiotic use during 5 days of observation. While accounting for stratified design and assuming 95% confidence, 1:1 enrollment in BSS and placebo groups, negligible dropout, and antibiotic use in 10% of the BSS group, we required 20 clinics each to enroll 20 participants to achieve 80% power to detect a 10% difference in antibiotic use between groups (ie, 20% of placebo participants use antibiotics). Because resources sufficed when the study began, we added 2 clinics (for a total of 440 participants) to increase statistical power.

All analyses were by assigned group. We used logistic regression to explore multivariable models, which included clinic, age quartiles, sex, wealth, or diarrhea severity at presentation as covariates. Because none of these potential confounders were significant in models assessing each covariate individually, we reported univariate results derived using χ2, Fisher exact, and Wilcoxon rank-sum tests, as appropriate. We constructed Kaplan-Meier curves to analyze resolution of symptoms, and differences between groups were assessed using the log-rank test. Two-sided P < .05 was considered statistically significant. Analyses were performed using SAS statistical software version 9.4 (SAS Institute Inc). Primary analysis occurred August to September 2015.

Results

Among 1344 patients with diarrhea screened for eligibility, we excluded 865 (64%), primarily because diarrhea had been occurring for 48 hours or more at the time of presentation; 63 (4.7%) reported taking an antimicrobial medication during the previous 48 hours (Figure 1). Of 479 eligible patients, 440 enrolled in the study, with 39 declining to participate. One participant was lost to follow-up. Median (interquartile range) participant age was 32 (23-45) years and 187 (43%) were male. In both groups, 37% of participants reported having taken antibiotics for a previous diarrheal illness (Table 1). The sociodemographic and clinical characteristics of participants were similar between study groups; most patients reported at least 5 stools per day, and one-third were moderately or severely dehydrated at the time of enrollment (Table 1). Similar proportions of patients in each group submitted enrollment stool specimens that yielded pathogens: 37 individuals (17%) in the BSS group and 43 (20%) in the placebo group; distribution of pathogens was similar between groups (Table 2). Among the 86 bacterial isolates, 44 (51%) were nonsusceptible to ampicillin and 50 (58%) were nonsusceptible to quinolones.

Table 1.

Characteristics of Participants at Enrollment, by Study Group

Characteristic	No. (%)
	Bismuth Subsalicylate (n = 220)	Placebo (n = 219)
Age, median (IQR), y	30 (23-45)	35 (23-45)
Female	121 (55)	131 (60)
People in household, median (IQR), No.	6 (5-8)	6 (5-8)
Self-rated health as poor or fair	77 (35)	65 (30)
Reported antibiotic use for previous diarrheal illness	82 (37)	81 (37)
Educational attainment
No formal education	50 (23)	55 (25)
Some primary	14 (6)	16 (7)
Completed primary with or without some secondary	50 (23)	54 (25)
Completed secondary	106 (48)	94 (43)
Ownership of
Refrigerator	184 (84)	180 (82)
Television	198 (90)	190 (87)
Mobile phone	218 (99)	211 (96)
Piped water available in home	174 (79)	161 (74)
Ethnicity
Mohajir	131 (60)	124 (57)
Punjabi	32 (15)	27 (12)
Sindhi	15 (7)	17 (8)
Other	42 (42)	51 (23)
Illness characteristics
≥5 Stools/24 h	137 (62)	130 (59)
Vomiting	47 (21)	49 (22)
Self-reported fever	64 (29)	57 (26)
Moderate or severe dehydrationa	73 (33)	73 (33)
Rehydration prescribed at enrollmentb	47 (21)	47 (21)
Previously sought care for current illness	5 (2)	8 (4)

Abbreviation: IQR, interquartile range.

Reported by study physician.

One patient in either group received intravenous rehydration.

Table 2.

Pathogens Cultured From Stool Specimens Collected at Enrollment by Study Group

Culture Result	No. (%)
	Bismuth Subsalicylate (n = 218)	Placebo (n = 214)
Any pathogen detecteda	37 (17)	43 (20)
Aeromonas hydrophila	6 (3)	4 (2)
Campylobacter jejuni	21 (10)	24 (11)
Salmonella species	2 (1)	2 (1)
Shigella speciesb	5 (2)	7 (3)
Vibrio speciesc	6 (3)	9 (4)

More than 1 pathogen was detected in 3 patients in each group.

Seven isolates were reported as Shigella flexneri, 4 as Shigella boydii, and 1 as Shigella dysenteriae.

Twelve isolates were reported as Vibrio cholerae serogroup O1, 2 as nontypeable Vibrio cholerae, and 1 as Vibrio species.

During 5 days of follow-up, 54 of the 439 participants (12%) used antibiotics (Table 3). Odds of any systemic antibiotic use were significantly lower in the BSS group (20 of 220 vs 34 of 219 patients; odds ratio [OR], 0.54; 95% CI, 0.30-0.98), as were odds of fluoroquinolone use (8 of 220 vs 20 of 219 patients; OR, 0.38; 95% CI, 0.16-0.88). All antibiotic use was reported to have followed consultation with a physician; none of 8 participants who consulted only a pharmacist received an antibiotic prescription. Of 76 participants who consulted a physician, 54 (71%) received antibiotics. Treated patients in each group received a median (range) of 2 (1-3) different antibiotic drugs and 24 (44%) received at least 1 antibiotic intravenously. Participants in both groups rarely perceived a need for antibiotics during follow-up: among the 54 participants who received antibiotics, 4 (7%) felt they needed them. Odds of antibiotic use were lower in the BSS group than in the placebo group when any pathogen was isolated from the enrollment stool culture (5 of 37 vs 16 of 43 patients; OR, 0.26; 95% CI, 0.09-0.81), and tended to be lower when toxigenic Vibrio cholerae was isolated, although the difference was not statistically significant (0 of 2 vs 6 of 7 patients; P = .08). Among the 21 culture-positive participants who were treated, 7 (33%) were given an antibiotic to which their isolate displayed resistance. Among the 54 patients who took antibiotics, 47 enrollment stool cultures (87%) yielded either no pathogens or pathogens resistant to the consumed antibiotic.

Table 3.

Illness Severity, Antibiotic Use, and Economic Effects During Follow-up by Study Group

Characteristic	No. (%)		Odds Ratio (95% CI)	P Value
	Bismuth Subsalicylate (n = 220)	Placebo (n = 219)
Illness Severity
Sought medical carea	38 (17.3)	46 (21.0)	0.79 (0.49-1.3)	.33
From physician	34 (15.5)	42 (19.2)	0.77 (0.47-1.3)	.32
From pharmacist	4 (1.8)	6 (2.7)	0.66 (0.18-2.4)	.54
Was hospitalized	6 (2.7)	10 (4.6)	0.59 (0.21-1.5)	.32
During first 24 h, median (IQR)
No. of stools	3 (2-4)	3 (2-4)		.37
No. of vomiting episodesb	0 (0-0)	0 (0-0)		.31
Nonantibiotic treatments during follow-up
Took oral rehydration solutions	25 (11.4)	31 (14.2)	0.78 (0.44-1.4)	.39
Received intravenous rehydration	14 (6.4)	27 (12.3)	0.48 (0.25-0.95)	.03
Took nonantibiotic medications	32 (14.5)	44 (20.1)	0.68 (0.41-1.1)	.13
Antidiarrheal	25 (11.4)	33 (15.1)	0.72 (0.41-1.3)	.26
Antiemetic	11 (5.0)	16 (7.3)	0.67 (0.30-1.5)	.33
Analgesic or antipyretic	7 (3.2)	16 (7.3)	0.42 (0.17-1.0)	.06
Vitamin	4 (1.8)	8 (3.7)	0.49 (0.14-1.7)	.26
Could participate in “all normal activities”b
First 24 h	122 (55.5)	104 (47.5)	1.4 (0.95-2.0)	.10
Second 24 h	192 (87.3)	179 (81.7)	1.5 (0.91-2.6)	.12
Antibiotics
Received any antibiotic	20 (9.1)	34 (15.5)	0.55 (0.30-0.98)	.04
Received intravenous antibioticsb	14 (6.4)	10 (4.6)	1.4 (0.62-3.3)	.53
No. of antibiotic classes prescribed if treated, median (IQR)b,c	2 (1-3)	2 (1-3)		.80
Type of antibiotic receivedb,d
Fluoroquinolone	8 (3.6)	20 (9.1)	0.38 (0.16-0.88)	.02
Nitroimidazole	17 (7.7)	29 (13.2)	0.55 (0.29-1.0)	.06
Tetracycline	7 (3.2)	6 (2.7)	1.2 (0.39-3.5)	>.99
Perceived need for antibiotics	3 (1.4)	5 (2.3)	0.59 (0.14-2.5)	.50
Economic Effects
Missed workb	106 (48.2)	124 (56.6)	0.71 (0.49-1.0)	.09
Days of work missed, median (IQR)b
Among those missing work	1 (1-2)	1 (1-2)		.30
Among all participants	0 (0-1)	1 (0-1)		.04
Incurred additional medical expenditurese	41 (18.6)	52 (23.7)	0.74 (0.46-1.2)	.20
Medical expenditures, median (IQR), Pakistani rupeesf
Among those with expenditures	100 (50-200)	185 (60-405)		.18
Among all participants	0 (0-0)	0 (0-0)		.14

Abbreviation: IQR, interquartile range.

None reported seeking care from a nurse or traditional healer.

Analysis was not prespecified.

Among the 54 participants who took antibiotics. Assessed classes included penicillins, cephalosporins, sulfonamides, quinolones, macrolides, tetracyclines, aminoglycosides, and nitroimidazoles.

Penicillins, macrolides, and nalidixic acid were not reported among any participants.

Includes fees from clinician consultations and associated treatments, and expenditures on additional remedies. Sample size was 41 in the bismuth subsalicylate group and 52 in the placebo group.

One Pakistani rupee was equivalent to approximately 0.01 US dollar at the time of the study.

Similar proportions of participants in each group sought care; physicians predominated as the type of practitioner (Table 3). Sixteen participants were hospitalized (6 of 220 [3%] in the BSS group and 10 of 219 [5%] in the placebo group); none died. Intravenous rehydration occurred about half as often during follow-up among those in the BSS group (14 of 220 vs 27 of 219 patients; OR, 0.48; 95% CI, 0.25-0.95). We did not detect a difference between groups in time to first formed stool or to last diarrheal stool (Figure 2), nor did we measure differences in the proportion of patients with any formed stool, diarrheal stool, nausea, vomiting, or abdominal pain at 24 or 48 hours after enrollment. Similarly, the median cumulative reported number of diarrheal stools did not differ between groups at 24, 48, or 120 hours after enrollment. Participants in the BSS group tended to use additional medications less frequently than did controls, although the difference was not statistically significant (32 of 220 vs 44 of 219 patients; OR, 0.68; 95% CI, 0.41-1.12); antidiarrheals were the most commonly used medication in both groups (Table 3).

Figure 2.

Kaplan-Meier Curves for Time to Last Diarrheal Stool and Time to First Formed Stool by Study Group

Approximately half the participants in each group missed work during follow-up; the median duration of missed work was significantly greater in the placebo group (median [interquartile range] days missed, 1 [0-1] vs 0 [0-1] for the BSS group; P = .04) (Table 3). Although the difference was not statistically significant, those in the BSS group tended to report ability to participate in “all normal activities” more frequently than those in the placebo group during the first and second 24 hours after enrollment. Medical expenditures tended to be higher and occur among more controls compared with those in the BSS group, although the result was not statistically significant (Table 3).

Participants in either group consumed a median of 16 tablets (8 doses) of study medication. Most participants in each group reported liking the study medication (191 of 220 [86%] in the BSS group and 179 of 219 [82%] in the placebo group), primarily because they perceived that it helped relieve their diarrhea (174 of 220 [79%] in the BSS group and 163 of 219 [74%] in the placebo group). Many participants in each group felt the study medication worked as well as antibiotics for diarrhea (92 of 220 [42%] in the BSS group and 81 of 219 [37%] in the placebo group).

Discussion

Globally, progress on improving the appropriateness of prescribing practices and patient demand for antimicrobials has been slow.[1,3,17] Progress may be particularly challenging in settings where such drugs can be obtained without a prescription and where financial incentives exist for clinicians to prescribe them, such as in many parts of Asia.[16,29] Rates of antibiotic misuse are particularly high in Pakistan, where a recent study found antibiotic use among young children on 18% of more than 150 000 observed days.[30] Although respiratory illnesses are the predominant reason for antibiotic misuse in the United States, diarrhea has been documented as the most frequent reason for antibiotic use in less developed settings, and gastrointestinal symptoms were the most common reason for self-medication with antibiotics among adults in Pakistan.[2,30,31] In this study, patients who received BSS during consultation for diarrhea were 46% less likely than those who received placebo to take antibiotics during 5 days of follow-up. In comparison, a 2017 Cochrane review[32] concluded clinician-directed interventions likely reduce inappropriate antibiotic use by 25% or less.

Interventions to improve antibiotic use often attempt to influence rational decision-making among clinicians; however, the effect of such methods tends to be modest and short-lived.[33] Audit and feedback to prescribers, the most common method in US hospitals,[34] has had little effect in outpatient[35] or inpatient[36] settings; supplying practitioners with printed educational materials has also appeared ineffective.[35,37] Efforts to reduce demand among patients may lead to patients reporting more severe symptoms,[38] but appear to reduce antibiotic prescribing rates when implemented as part of interventions to increase shared medical decision-making.[39] Similarly, clinician behavior change following policy-level interventions tends to be limited. Financial incentives may lead to shifts in diagnostic coding,[40] and formulary restrictions to changes in antibiotic choice.[41,42] Provision of guidelines alone has had little effect on prescribing rates for several outpatient conditions.[43,44]

Alternative strategies rooted more strongly in behavioral science could supplement or supplant more traditional strategies to reduce inappropriate use of antibiotics. Nudge methods do not limit choice, do not alter the economic incentives related to the behavior, and do not mandate—all of which can be difficult to implement and may lead to resistance from practitioners and patients[34,45]—but instead rely on supplying a default that many would not bother to resist.[46] Interventions addressing psychosocial drivers of prescribing behavior, such as social norms, physician desire to provide symptomatic relief, and physician perception that patients expect medical management,[47] may work at least as well as interventions that assume clinicians will make rational decisions based on knowledge of guidelines or possible public health effects.[48,49] Posting a commitment letter on the clinic wall was associated with approximately 20% fewer inappropriate antibiotic prescriptions in one US study.[33] Awareness of prescription practices among clinician leaders appears to influence those among others in the community and may also be important to target during nudge campaigns.[34,49] However, the predominance of small, independent clinics may make this strategy difficult to implement in South Asia.

Limitations

This study has limitations. By offering an alternative medication during the clinical encounter, we redirected patients in both study groups away from antibiotics. This nudge likely reduced antibiotic use in both study groups: effort required to obtain antibiotics increased because they were not provided during the initial clinic visit; meanwhile, because diarrhea symptoms often wane naturally within 1 to 2 days of onset, participant motivation to obtain antibiotics likely decreased while they tried the study medication. Indeed, a Cochrane review[50] concluded that, when safe to do so, withholding antibiotics yields greater reduction in antibiotic use while maintaining patient satisfaction and rates of favorable clinical outcomes compared with providing delayed prescriptions. In our study, the proportion of participants using antibiotics was lower than we anticipated: 34 (16%) of patients in the placebo group used antibiotics, compared with 96% of Pakistani children who presented for care for diarrhea at a private clinic in one study[7] and 46% of Pakistani children with diarrhea, whether or not they presented for care, in another.[30] Thus, we likely would have observed a larger relative reduction in antibiotic use had we not controlled antibiotic use at enrollment in the placebo group. Despite this, compared with controls, substantially fewer participants in the BSS group used antibiotics.

The drivers of this relative reduction are unclear. Although previous studies have found reduced duration and severity of symptoms among children and young adults given BSS for diarrhea,[21,23,24,25] we did not detect reductions in diarrhea output or time to resolution of abdominal pain, nausea, or vomiting in this study. This may be because we relied on participants to record numbers and timing of diarrheal episodes and other symptoms rather than hospitalizing participants to obtain more objective and systematic measurements, or because our study population experienced a wider range of diarrheal etiologies. Nonetheless, additional outcome measures suggest that participants in the control group felt less well than those in the BSS group during follow-up. In addition to taking antibiotics more often, participants in the control group were significantly more likely to receive intravenous rehydration, and they missed more work. Although results did not attain statistical significance, such participants were also less likely to report participation in all their typical activities during the first 2 days of follow-up and more likely to take nonantibiotic medications or supplements for their illness, with antidiarrheal medications being the most commonly reported. The additional treatments used in the placebo group might have obscured differences in illness course between the study groups.

We might have further underestimated BSS’s effect on antibiotic use because we included only participants who had not used antibiotics at the time of presentation and who consented to participate in the study. Such persons might have been less likely to use antibiotics than the general population; indeed, only 37% of participants reported having taken antibiotics for any previous diarrheal episode. Despite this, we measured substantially less antibiotic use in the BSS group than among controls. Next, antibiotic use at enrollment or during the study could have been misclassified; however, we used visual aids depicting photos and common names of antibiotics available in Karachi during patient interviews, and because participants reported obtaining all antibiotics used during follow-up from a physician, we were able to verify drug class using clinical records and pharmacy receipts. We did not collect participant employment status, so analyses including missed work and lost income could be driven by imbalances in employment across study groups, despite lack of imbalance among the other characteristics we analyzed. All participants were recruited in Karachi and followed up for only 5 days; generalizability and sustainability of these findings should be verified through additional study.

Conclusions

Effects of interventions to reduce inappropriate antibiotic use are likely to vary by illness and setting[35]; however, we found lower odds of antibiotic use following BSS for acute diarrhea in a setting where antibiotic use for diarrhea is typically high. This effect size approaches the goal of 50% reduction in inappropriate antibiotic use in outpatient settings set in the United States Action Plan for Combatting Antimicrobial Resistance.[51] Because of the challenges inherent in developing new classes of antibiotic medications, and because increased population density and mixing and decreased travel times facilitate the spread of antibiotic resistance,[16] preserving the function of the limited number of existing drugs is critical.[52] In Pakistan, more than half of adults with diarrhea may seek care from a physician.[20] Encouraging health care professionals and pharmacists in settings with high diarrhea incidence to recommend BSS as frontline treatment for adults with diarrhea, and promoting BSS for diarrhea self-management, may reduce antibiotic use and rates of antibiotic resistance globally.