Potential impact and cost-effectiveness of future ETEC and Shigella vaccines in 79 low- and lower middle-income countries

While diarrhea mortality in children has declined over the last two decades, there has been a slower decline in diarrheal episodes. Repeated diarrheal episodes are associated with childhood stunting, which leads to increased mortality risk from infectious diseases. Vaccine candidates are under development for enterotoxigenic Escherichia coli [ETEC] and Shigella, important enteric pathogens in children in low income countries. These future vaccines could significantly reduce diarrheal burden, prevent ETEC- and Shigella-induced stunting, and stunting-associated mortality. We developed a cost-effectiveness model for two putative standalone ETEC and Shigella vaccine candidates to evaluate vaccine impact on mortality, morbidity, stunting, and stunting-associated deaths from other infectious diseases. We modeled impact over the first ten years after vaccine introduction in children under five years old living in 79 low and low-middle income countries. ETEC and Shigella diarrhea would cause an estimated 239,300 [95% UL: 179,700–309,800] and 340,300 [256,500–440,800] child deaths, respectively, from years 2025 to 2034. Most of these deaths would occur in AFRO countries. ETEC and Shigella moderate-to-severe diarrheal episodes would result in over 13.7 [8.4–19.0] and 21.4 [13.1–29.8] million stunted children, respectively. Introducing ETEC or Shigella vaccine each with 60% efficacy could prevent 92,000 [61,000–129,000] ETEC and 126,600 [84,000–179,000] Shigella direct deaths and 21,400 [11,300–34,800] ETEC- and 34,200 [18,000–56,000] Shigella-induced stunting deaths. ETEC ICERs ranged from $2172/DALY [1457–4369] in AFRO to $19,172/DALY [12,665–39,503] in EURO. Shigella ICERs ranged from $952/DALY [632–2001] in EMRO to $640,316/DALY [434,311–1,297,192] in EURO. Limitations of this analysis include uncertainty of vaccine efficacy, duration of protection, and vaccine price. Inclusion of other infectious disease mortality due to stunting provides a more accurate assessment of total ETEC and Shigella disease burden and increased the projected impact and cost-effectiveness of vaccination. Introducing vaccines only in high burden countries and regions could substantially reduce cost without substantially reducing impact.


Introduction
Globally, diarrhea remains the second leading cause of mortality, accounting for approximately 500,000 deaths annually in children under five years old [1]. However, diarrheal mortality in children has declined by 34.3%, with similar declines in Shigellosis (33.8%) and enterotoxigenic E. coli (ETEC) infection (38.1%) from 2005 to 2015 [2]. While diarrheal mortality has declined, morbidity and mortality continue to plague many low-and lower middleincome countries (LMICs).
In addition to rotavirus, other pathogens have a substantial role in diarrheal burden. The Global Enteric Multicenter Study (GEMS) found that of 22 diarrheal pathogens, four-rotavirus, Shigella, ETEC, and Cryptosporidium-were associated with moderate-tosevere diarrhea (MSD), accounting for 70% of cases in 0-4 year olds [3]. This study also documented an increased mortality risk for MSD ETEC cases and increased stunting risk in cases associated with both ETEC and Shigella [4]. ETEC and Shigella have been among the top four causes of diarrhea associated with years lost to disability (YLDs) worldwide [5]. A recent study evaluating global ETEC and Shigella burden found that these pathogens have a sizeable global burden, especially in the World Health Organization (WHO) designated African and the Eastern Mediterranean regions [6].
As diarrheal mortality has declined, there is increased focus on diarrhea morbidity in pediatric populations. Repeated, non-fatal episodes of diarrhea from infection by certain pathogens are thought responsible for reduced linear growth and childhood stunting [7][8][9], increasing mortality risk from other infectious diseases [10]. Results from GEMS showed that MSD episodes shifted the height-for-age z-scores downward [3,11], increasing a child's risk of stunting. In countries where stunting is highly prevalent, ETEC-and Shigella-induced stunting impacts even more children [12].
The burden from these pathogens necessitates new prevention strategies. A potential and highly beneficial prevention strategy would be the use of vaccines to prevent ETEC and Shigella infection, which are currently under development with Phase 1 and Phase 2 data available shortly in non-infant populations. In addition to reducing burden worldwide, the successful rollout of rotavirus vaccines has shown that enteric vaccines are deployable to endemic countries. Studies describing rotavirus burden, vaccine impact, and cost-effectiveness have contributed to global and country decisions to accelerate rotavirus vaccine introduction [13,14]. Therefore, investigating how ETEC and Shigella vaccination could impact high burden countries or populations is not only important in guiding vaccination programs, but may in turn spur action by policymakers.
In this study, we conducted a vaccine impact and costeffectiveness analysis for ETEC and Shigella vaccines to identify high-need areas and to capture the full potential value of these vaccines. We evaluated the impact of vaccination on the mortality, morbidity, number of stunted children, and stunting-associated deaths from other infectious diseases, to understand the expanded impact of these vaccines.

Population and time frame
We included seventy-nine countries from a previous analysis of ETEC and Shigella burden with a full description of country inclusion and exclusion criteria [6] (Table 1). We aggregated national estimates by WHO regions to identify trends ('AFRO': African region, 'AMRO': Region of the Americas, 'EMRO': Eastern Mediterranean Region, 'SEARO': Southeast Asian Region, 'WPRO': Western Pacific Region). We assumed all countries would introduce the vaccines in 2025. We examined 10 annual birth cohorts of children over the first five years of life. Population estimates of under five children were based on UN Population Division country estimates and projections from 2025 to 2034 [15].

Etiological fraction
ETEC and Shigella burden is dependent upon the fraction of episodes and deaths attributable to each [3,16]. As previously described [6], we used culture-based etiological estimates [3] and adjusted them for under-detection using estimates derived from molecular methods based on Liu et al. [17,18]. We applied an adjustment of 1.5 times for ETEC and 2.0 times for Shigella to the culture results [20].

Diarrhea mortality and morbidity estimates
Similar to our burden study, we used the mid-point of 2015 diarrheal mortality estimates from two sources: the Global Burden of Disease study (GBD) at the Institute for Health Metrics and Evaluation (IHME) [19] and the WHO Maternal Child Epidemiology Estimation (MCEE) project [20]. We then used the etiological fraction for ETEC and Shigella to calculate pathogen-attributable deaths and adjusted diarrheal mortality to account for countries that introduced rotavirus vaccine before 2014 [6]. We projected mortality and morbidity estimates from 2015 to 2034. Because diarrheal mortality rates have declined over time, we estimated annual rates of decline for non-rotavirus under-five diarrheal mortality in each country using data from Child Health Epidemiology Reference Group from 2000 to 2013 [21]. We calculated diarrhea morbidity estimates using WHO region-specific estimates of diarrhea episodes and the etiological fractions for ETEC and Shigella [22]. We assumed morbidity declined at a rate of 0.45% per year. This decline percentage was calculated from YLDs from diarrheal disease from 1990 to 2010 from the GBD [23].

Effects of ETEC-and Shigella-induced stunting
We applied the methods from Anderson et al. [6] to determine the effects of ETEC and Shigell-induced stunting. First, we calculated the shift in child height-for-age z-scores from ETEC and Shigella episodes using GEMS results (Table 1). In the absence of reliable community-level estimates for diarrhea treatment in countries included in this study, we assumed 22% of diarrheal episodes where care was sought at a health facility were considered MSD. The proportion of child diarrheal episodes where caretakers sought care were taken from the most recent Demographic and Health Surveys (DHS), available for 70 countries [24]. In countries without a country-level estimate, we substituted the corresponding WHO regional average. Based on the mean of estimates for the countries included in this analysis, we calculated that 47% of caretakers would visit a health facility after the child experienced an episode of diarrhea (Table 1). Thus, we assume that 10% (22% Â 47%) of child diarrheal episodes were moderate-to-severe in each region. We used the same the approach as in Anderson et al. [6] to estimate the number of child deaths from infections for which stunting is a risk factor (pneumonia, malaria, measles, and diarrhea [25]). We did not project rate of change of pneumonia, malaria and measles burden over time.

Outcomes measures
Outcome measures included: diarrheal episodes; direct deaths; children stunted; stunted children dying from other infectious diseases; and Disability-adjusted Life Years (DALYs). Stunting is a risk factor for other infectious disease deaths and not directly included in DALY calculations. All mortality outcomes, including other infectious disease deaths due to induced stunting, were translated to DALYs using standard techniques [26,27]. We calculated DALYs using non-uniform age-weighting and a 3% annual discount rate. We calculated all outcomes annually and cumulatively from time of introduction.

Vaccines
We evaluated the impact of potential ETEC and a Shigella standalone vaccine candidates when introduced nationally in 79 countries. We assume each vaccine is 60% efficacious in preventing deaths and MSD episode. Our other assumptions were that protection was conferred after the third dose and that there was no protection for partially vaccinated children. We assumed that vaccine effectiveness does not wane, but we also assumed that there was no effectiveness after five years of age or herd protection. For coverage, we used country specific 2015 DPT3 coverage estimates [28] and only included estimates for children estimated to receive a full course (all three doses).
As vaccine price is uncertain, we used the Gavi Rotarix price of approximately $2.00 per dose [29] as our basis. For our study, we assumed study vaccines would cost $3.30 per dose, and we varied this price in the sensitivity analysis to assess price impact on costeffectiveness.

Costs
As there are no published medical costs for ETEC and Shigella diarrhea for the 79 study countries, we used country specific estimates of direct medical costs of illness associated with inpatient and outpatient care for MSD episodes. We assumed 1 of 8 outpatients with MSD were referred for inpatient care [13,30].
Direct medical costs were based on WHO-CHOICE Service Delivery Unit Cost estimates and commodity costs. For outpatient medical costs, we used country specific cost per outpatient visit at a primary hospital. We used country specific daily costs at hospitals and a four-day stay to calculate inpatient cost. We assumed that outpatients receive six oral rehydration solution packets per day for two days, and inpatients receive six packets per day and two IVs during a four day hospital stay [31]. On average, the illness cost per episode was $10.05 for outpatients and $82.25 for inpatients. These estimates were triangulated against country specific estimates, when available in the literature [32][33][34][35][36][37][38][39][40][41][42][43][44][45][46][47]. In most cases, modelled estimates were aligned with empirical estimates.
Using data available from Portnoy et al. [48], we generated vaccine administration costs. All costs were in 2016 US dollars and discounted (3%). Our cost-effectiveness estimates were from the health system perspective.

Cost-effectiveness
We calculated vaccination cost (V) cumulated over the first 10 years (t) after introduction starting in 2025 for each country (c) based on vaccine administration cost, vaccine price, and quantity (birth cohort times coverage rate with 10% vaccine wastage). We calculated averted costs (A) based on population, vaccine coverage, efficacy, and access to care and medical costs in each country (c). We calculated net costs (N) for each region (r) as: We calculated vaccine benefits (B) for each region (r) based on the sum of population, coverage (C), efficacy (E), and DALY burden (D) in each country (c) cumulated over the first 10 years (t) after introduction. We calculated the number of children fully vaccinated each year by multiplying the annual birth cohort by the assumed vaccine coverage. We projected benefits for the first five years of life for children vaccinated in each annual birth cohort.
Our primary cost-effectiveness measure was the regional Incremental Cost-Effectiveness Ratio (ICER r ), which is the aggregated country-level incremental costs associated with introducing each vaccine divided by aggregated country-level health benefit.
Our comparator scenario was no vaccination. We calculated ICERS with and without other infectious disease burden due to stunting for each country and region annually and cumulatively.
We presented results using two thresholds for costeffectiveness as compared to GDP [49]. We presented results for countries where ICERS are less than 3 times their GDP, which are considered cost effective and results for countries where ICERS are less than GDP, which has historically been used to determine if an intervention is 'highly' cost-effective.

Sensitivity analysis
In order to assess the impact of uncertainty and changes in key input variables, we conducted two types of sensitivity analysis using SimVoi [50]. First, we used one-way sensitivity analysis to demonstrate the impact of individual input parameters on vaccination cost-effectiveness. These results are shown in tornado diagrams, with each horizontal band showing the effect of varying each parameter between high and low values. Second, we conducted a probabilistic sensitivity analysis (PSA) to show the overall impact of input parameter uncertainty (Table 2) on our estimates of key outcomes. Monte Carlo analysis using 10,000 iterations was conducted and we included estimated upper and lower 95% uncertainty limits (2.5% and 97.5%) for key outputs in brackets after our estimates. We reported results of the sensitivity analyses as a range of the difference between high and low ICER estimates. We also analyzed results for Gavi-eligible countries only. These 48 countries account for over 60% of the total ETEC and Shigella burden (Tables 3 and 4 [14; 37] countries, respectively (Table 5), and 'very cost-effective' threshold (ICER < GDP) in 6 [1; 8] and 11 [3; 14] countries, respectively ( Table 6). The majority of these countries are in AFRO (Fig. 1         (continued on next page)   Fig. 2A and B). Variation in projected mortality change (range; $1326/DALY [ETEC] and $923/DALY [Shigella]) was also influential on ICERs.

Discussion
Our analysis is the first evaluating the impact and costeffectiveness of potential ETEC and Shigella vaccine candidates in children in low and lower-middle income countries. The impact of vaccination on stunting and deaths from other infectious diseases makes ETEC and Shigella vaccination more compelling and cost-effective. These vaccines could avert over 274,000 ETEC and Shigella attributable deaths in the first decade, a 47% reduction in mortality. Including benefits of stunting averted results in much lower ICERs and increases the number of countries where vaccination is cost-effective.
Our results suggest that there is heterogeneity in vaccine impact and cost-effectiveness across regions and by Gavi eligibility. While the majority of the averted burden for both ETEC and Shigella vaccines is in AFRO, substantial burden is also averted in EMRO. Introducing vaccines only in high burden countries and/regions could reduce cost without substantial reductions in health impact. This is clearest when calculating impact for those countries where vaccines are cost-effective. The number of deaths averted per 100,000 fully vaccinated children for an ETEC vaccine increases, while still averting the majority of preventable total ETEC deaths. Gavi-eligible countries may also benefit greatly as these countries experience a large share of projected disease burden. Vaccine introduction in these countries alone could have a substantial impact on burden. It will be important for country policy makers to understand disease heterogeneity when evaluating whether or not to introduce these vaccines.
There is evidence that childhood stunting may be associated with chronic diseases such as heightened prevalence of high blood pressure, impaired fasting glucose, and increased body mass index [9]. If ETEC and Shigella infections increase chronic disease through induced stunting, vaccination could be even more impactful and cost-effective than initially realized. In addition to chronic disease risk, early childhood stunting has been shown to be associated with decreased earnings and fewer completed years of school  [51]. If ETEC and Shigella vaccination could reduce stunting and these long-term health and development consequences, then vaccination is more economically viable than indicated in the results from this study. The model incorporates projections based on population forecasts and past diarrheal mortality trends that could contribute to underestimation of vaccine impacts. First, while the past two decades have seen consistent declines in diarrheal mortality [52], additional factors may alter this trend. Global urbanization rates are increasing in low-and lower-middle income countries, placing larger populations at risk as growth overwhelms infrastructure, forcing many into living in underserved informal settlements. Risk for infectious diseases increase in these environments due to overcrowding with lack of safe sanitation and clean water [53]. While moving to urban areas could increase access to health care and therefore reduce mortality, this is highly dependent on individual household economic status and their community's urban infrastructure which is highly variable within and between cities. There is some evidence that a move to urban areas reduces access to health care [54]. Second, increasing antibiotic resistance and climate change may reverse gains made over the past few years by increasing risk of exposure, severe disease, and death. Third, vaccination impact could be underestimated if these vaccines induce herd protection. Fourth, dmLT adjuvant is in current formulation which has potential to improve protection. Evidence of mucosal immune responses induced by ETVAX vaccine may provide some protection against ETEC colonization factor antigens not included in the vaccine [55].
There are several limitations to this analysis. First, there are no vaccine trial-derived measures of vaccine efficacy and duration. As these vaccine candidates are under development, we do not know their price, thus our cost-effectiveness estimates may be higher or lower than projected. Furthermore, our cost-effectiveness estimates are dependent on assuming that these two vaccines would have the same price, efficacy and coverage.  Table 2) are displayed at the end of the corresponding bar. Price per dose is varied by $US ranging from a low estimate of $2/dose to a high estimate of $7/dose. 'Etiological fraction' is variation in the fraction of overall diarrheal mortality attributed to ETEC and Shigella diarrhea. 'Mortality change' is variation in the rates diarrheal mortality projected in years 2025-2034. 'Induced stunting' refers to the number of other infectious disease deaths caused by ETEC or Shigella induced stunting. 'Hospitalized fraction' is variation in the fraction of children hospitalized (1 in 8 referred to inpatient facility).
No treatment-seeking or hospitalization rates exist for children experiencing ETEC and Shigella episodes in most of our study countries. We used DHS data on treatment seeking for diarrhea and assumed a proportion of those children are hospitalized. We also adjusted our fraction of MSD episodes based on GEMS methodology. Thus, we have assumed that all MSD cases access treatment which may underestimate the true number of child MSD. We modelled cost of illness based on available estimates-this approach may over-or under-estimate the medical cost averted due to vaccination in many countries. For most input parameters, we have limited information on the true degree of uncertainty and thus we rely on simple distribution types and measures of dispersion. We did not estimate waning in our model or the impact of partial protection from receiving less than three doses.
A final limitation is the rate of decline for diarrheal mortality. We assumed mortality decline mirrors the decline seen during 2000-2013. However, studies are actively evaluating whether the rate of decline is decreasing. Incorporating these findings may alter our estimates of vaccine impact by increasing the number of deaths averted.
This analysis considered the potential impact and costeffectiveness of vaccines that are in development. It shows the importance of including the expanded impact of vaccination to prevent stunting and resulting deaths due to other infectious disease. To capture the true value of these vaccines, it is important to quantify the expanded effects of enteric pathogens. Burden was concentrated in a few regions, indicating that introduction in these countries could potentially avert the majority of disease burden globally. As these vaccines undergo further clinical development, it will be important to reassess the potential vaccine impact as data on efficacy, duration of protection, and diarrheal impact on stunting becomes available.