Burden of diarrhea and antibiotic use among children in low-resource settings preventable by Shigella vaccination: A simulation study

Background Shigella is a leading cause of diarrhea and dysentery in children in low-resource settings, which is frequently treated with antibiotics. The primary goal of a Shigella vaccine would be to reduce mortality and morbidity associated with Shigella diarrhea. However, ancillary benefits could include reducing antibiotic use and antibiotic exposures for bystander pathogens carried at the time of treatment, specifically for fluoroquinolones and macrolides (F/M), which are the recommended drug classes to treat dysentery. The aim of the study was to quantify the reduction in Shigella attributable diarrhea, all diarrhea, and antibiotic use in the first 2 years of life that could be prevented by a Shigella vaccine. Methods and findings We used data from the Etiology, Risk Factors, and Interactions of Enteric Infections and Malnutrition and the Consequences for Child Health and Development (MAL-ED) study, a birth cohort study that followed 1,715 children with twice weekly surveillance for enteric infections, illnesses, and antibiotic use for the first 2 years of life from November 2009 to February 2014 at 8 sites. We estimated the impact of 2 one-dose (6 or 9 months) and 3 two-dose (6 and 9 months, 9 and 12 months, and 12 and 15 months) Shigella vaccines on diarrheal episodes, overall antibiotic use, and F/M use. Further, we considered additional protection through indirect and boosting effects. We used Monte Carlo simulations to estimate the absolute and relative reductions in the incidence of diarrhea and antibiotic use comparing each vaccination scenario to no vaccination. We analyzed 9,392 diarrhea episodes and 15,697 antibiotic courses among 1,715 children in the MAL-ED birth cohort study. There were 273.8 diarrhea episodes, 30.6 shigellosis episodes, and 457.6 antibiotic courses per 100 child-years. A Shigella vaccine with a mean vaccine efficacy of 60% against severe disease given at 9 and 12 months prevented 10.6 (95% CI [9.5, 11.5]) Shigella diarrhea episodes of any severity per 100 child-years (relative 34.5% reduction), 3.0 (95% CI [2.5, 3.5]) F/M courses for Shigella treatment per 100 child-years (relative 35.8% reduction), and 5.6 (95% CI [5.0, 6.3]) antibiotic courses of any drug class for Shigella treatment per 100 child-years (relative 34.5% reduction). This translated to a relative 3.8% reduction in all diarrhea, a relative 2.8% reduction in all F/M courses, a relative 3.1% reduction in F/M exposures to bystander pathogens, and a relative 0.9% reduction in all antibiotic courses. These results reflect Shigella incidence and antibiotic use patterns at the 8 MAL-ED sites and may not be generalizable to all low-resource settings. Conclusions Our simulation results suggest that a Shigella vaccine meeting WHO targets for efficacy could prevent about a third of Shigella diarrhea episodes, antibiotic use to treat shigellosis, and bystander exposures due to shigellosis treatment. However, the reductions in overall diarrhea episodes and antibiotic use are expected to be modest (<5%).

Based on the different scenarios outlined above, will extend the results from Aim 1a (attributing diarrheal episodes that cause treatment) and from Aim 1b (attributing bystander antibiotic exposures for subclinical pathogens to diarrheal pathogens).Analyses will be performed at the level of diarrheal episode (N=6,677) from Aim 1a.We will then merge in the number of subclinical pathogen exposures to antibiotics per diarrheal episode from Aim 1b.We will use the attributable fractions per episode (AFe) for diarrheal pathogens that were calculated in Aim 1a where a pathogen with an AFe >0.5 is considered the cause of the diarrheal illness.We will then simulate how much antibiotic exposure calculated in Aims 1a (https://osf.io/e4bqr/)and 1b (https://osf.io/3asxh/)can be prevented with a vaccine.To simulate, we will perform Monte Carlo simulations with random sampling of the dataset with replacement to a sample size of 50,000.95% confidence intervals for all parameters will be estimated by bootstrap with 1,000 resamples.

Analyses:
Probabilistically sample observed episodes as "prevented" by vaccine and re-calculate incidence We will start with a vaccine for Shigella with dosing at 6 and 9 months of age.We will assume 60% efficacy 14-days after administration of the second dose.For simplicity, we will assume 0% efficacy prior to this time point.The Monte Carlo simulation will then be performed over the dataset with 50,000 samples under two scenarios: vaccine and no vaccine.For the vaccine scenario, we will probabilistically assign whether diarrheal episodes attributable to Shigella (defined by AFe > 0.5) would have been prevented assuming 60% efficacy.It is important to note that Shigella infections increase with age therefore, 60% efficacy will not directly translate to a 60% reduction in diarrheal illnesses.Additionally, there is low natural immunity (~20%) against subsequent Shigella diarrhea when considering all serotypes and species, so prior infection will not be factored in (Rogawski McQuade 2020 J Infect Dis), at least initially.In a later iteration of the analyses, we may incorporate potential clustering of prevented episodes within children (i.e.all episodes among 60% of children are prevented, rather than 60% of episodes prevented overall, regardless of the child).We may also incorporate the potential for herd immunity among children too young to be vaccinated, partial efficacy after a single vaccine dose, different vaccine schedules (6&9, 9&12, 12&15 months), effects on severity in addition to or instead of incidence, and varying levels of vaccine efficacy.For the no vaccine scenario, no Shigella-attributable episodes would be prevented.We will conduct a sensitivity analysis where we define shigella diarrhea by any detection of shigella rather than AFe > 0.5.
We will estimate the expected reductions due to a Shigella vaccine in the following metrics: 1. Antibiotic treated severe Shigella diarrhea episodes To estimate the expected reductions in #1-7 listed above, we will compare the relative and absolute differences between the vaccine and no vaccine simulations.
Relative difference, is expressed as a ratio of the metric between the vaccine simulations and the no vaccine simulation, which will be calculated as:

𝑣𝑎𝑐𝑐𝑖𝑛𝑒 𝑠𝑖𝑚𝑢𝑙𝑎𝑡𝑖𝑜𝑛 − 𝑛𝑜 𝑣𝑎𝑐𝑐𝑖𝑛𝑒 𝑠𝑖𝑚𝑢𝑙𝑎𝑡𝑖𝑜𝑛 𝑛𝑜 𝑣𝑎𝑐𝑐𝑖𝑛𝑒 𝑠𝑖𝑚𝑢𝑙𝑎𝑡𝑖𝑜𝑛
Absolute difference, which captures the true distance between two values, will be calculated as:

| 𝑣𝑎𝑐𝑐𝑖𝑛𝑒 𝑠𝑖𝑚𝑢𝑙𝑎𝑡𝑖𝑜𝑛 − 𝑛𝑜 𝑣𝑎𝑐𝑐𝑖𝑛𝑒 𝑠𝑖𝑚𝑢𝑙𝑎𝑡𝑖𝑜𝑛 |
As a secondary outcome, we will estimate the average age of Shigella diarrhea under the vaccine and no vaccine scenarios to determine whether the vaccine would be expected to increase the average age of disease.
Future directions: