What is the impact of mass and systematic antibiotic administration on antibiotic resistance in low- and middle-income countries? A systematic review

35 Antibiotic consumption is a key driver of antibiotic resistance (AR), particularly in lowand 36 middle-income countries, where risk factors for AR emergence and spread are rife. However, the 37 potential contribution of mass and systematic antibiotic administration (MDA/SDA) to AR 38 spread is unknown. We conducted a systematic review to provide an overview of MDA/SDA in 39 lowand middle-income countries, including indications, antibiotics used and, if investigated, 40 levels of AR over time. This systematic review is reported in accordance with the PRISMA 41 statement. Of 2438 identified articles, 63 were reviewed: indications for MDA/SDA were 42 various, and targeted populations were particularly vulnerable, including pregnant women, 43 children, HIV-infected populations and communities in outbreak settings. Available data suggest 44 MDA/SDA may lead to significant AR increase, especially after azithromycin administration. 45 However, only 40% of studies evaluated AR. Integrative approaches that evaluate AR in addition 46 to clinical outcomes are needed to understand consequences of MDA/SDA implementation, 47 combined with standardized AR surveillance for timely detection of antibiotic resistance 48 emergence. 49


58
Antibiotic resistance (AR) is one of the greatest threats to global health, particularly in low-and

98
Exclusion criteria were systematic reviews and meta-analyses articles (only used as a source of 99 references in snowball searches), data collection prior to January 1 2000 and studies on MDA for 100 trachoma control, owing to a recently updated systematic review and meta-analysis investigating 101 AR following azithromycin MDA for trachoma control [9]. No language restrictions were 102 applied. 103 Three researchers were involved in the review process (LR, BTH and EDA). One reviewer (LR) 104 assessed article titles for relevance. Two of the three investigators (LR and BTH or EDA) 105 independently reviewed all potentially relevant abstracts. The same process was used for full text 106 screening and quality assessment. Disagreements were resolved by consensus among all parties. 107 For all eligible studies, we extracted details on objectives, methods and MDA/SDA 108 characteristics. If AR was evaluated, epidemiological and microbiological methods were 109 extracted. We stratified studies by target populations and types of antibiotic, and summarized 110 data on AR when evaluated (resistant pathogen prevalence, measures of association). 111 The Critical Appraisal Skills Programme tools based on Cochrane guidelines were used to assess 112 study quality. To assess data extraction quality, two investigators (LR and BTH or EDA) 113 reviewed extracted data for selected articles.

155
Eight studies (22%) investigated co-trimoxazole as SDA in HIV-infected adults [60-68] (or 156 adults and children) and its potential to decrease mortality rates, infections or malaria incidence.

157
The remaining three studies (

208
Of the sixteen studies in which co-trimoxazole was used as SDA, nine evaluated AR.

209
AR was assessed using meta-genomic analysis in two studies. Analysis of rectal swabs from 210 healthy infants from ARMCA showed a significant increase in risk of carrying sulfonamide 211 (RR=8.8, p=0.05) and trimethoprim (RR=3.3, p=0.04) resistance gene determinants relative to 212 the placebo group, while no difference was observed for beta-lactam and macrolide resistance 213 genes [11]. The second study targeted HIV-exposed uninfected infants [39]. In the group treated 214 with co-trimoxazole compared to placebo, the authors showed a decrease of gut microbiome β-215 diversity (diversity in resistance gene composition), increased AR gene α-diversity (resistance  The proportion of Haemophilus influenzae resistant to co-trimoxazole was also higher in the co- Of the five studies using amoxicillin as MDA, AR was evaluated in only one study [11]. While 251 prevalence of beta-lactam, macrolide and trimethoprim resistance genes were not significantly 252 different, prevalence of sulfonamide resistance was higher in the amoxicillin arms compared to

335
Better understanding of mechanisms of AR across species could help to better target particular 336 bacteria while minimizing bystander selection [75]. Microbiological assessment of AR was also 337 highly heterogeneous, and included phenotypic, molecular or metagenomic testing methods.

338
Phenotypic methods can identify resistance of specific organisms to specific antibiotics, and are 339 commonly used to characterize AR among both gram-positive and gram-negative bacteria.