Cost-effectiveness of Vitamin A supplementation among children in three sub-Saharan African countries: An individual-based simulation model using estimates from Global Burden of Disease 2019.

BACKGROUND: Vitamin A Supplementation (VAS) is a cost-effective intervention to decrease mortality associated with measles and diarrheal diseases among children aged 6-59 months in low-income countries. Recently, experts have suggested that other interventions like large-scale food fortification and increasing the coverage of measles vaccination might provide greater impact than VAS. In this study, we conducted a cost-effectiveness analysis of a VAS scale-up in three sub-Saharan African countries.
METHODS: We developed an individual-based microsimulation using the Vivarium simulation framework to estimate the cost and effect of scaling up VAS from 2019 to 2023 in Nigeria, Kenya, and Burkina Faso, three countries with different levels of baseline coverage. We calibrated the model with disease and risk factor estimates from the Global Burden of Disease 2019 (GBD 2019). We obtained baseline coverage, intervention effects, and costs from a systematic review. After the model was validated against GBD inputs, we modeled an alternative scenario where we scaled-up VAS coverage from 2019 to a level that halved the exposure to lack of VAS in 2023. Based on the simulation outputs for DALYs averted and intervention cost, we determined estimates for the incremental cost-effectiveness ratio (ICER) in USD/DALY.
FINDINGS: Our estimates for ICER are as follows: $860/DALY [95% UI; 320, 3530] in Nigeria, $550/DALY [240, 2230] in Kenya, and $220/DALY [80, 2470] in Burkina Faso. Examining the data for DALYs averted for the three countries over the time span, we found that the scale-up led to 21 [5, 56] DALYs averted per 100,000 person-years in Nigeria, 21 [5, 47] DALYs averted per 100,000 person-years in Kenya, and 14 [0, 37] DALYs averted per 100,000 person-years in Burkina Faso.
CONCLUSIONS: VAS may no longer be as cost-effective in low-income regions as it has been previously. Updated estimates in GBD 2019 for the effect of Vitamin A Deficiency on causes of death are an additional driver of this lower estimate of cost-effectiveness.

Introduction

Vitamin A deficiency (VAD) is a risk factor for several major causes of death in children, including diarrheal diseases and measles [1]. Vitamin A supplementation (VAS) has historically been an intervention of interest for the prevention of xerophthalmia and night blindness as well as reduction of child mortality caused by measles and diarrhea in many low-income countries [2, 3]. Globally among children under 5, VAD is responsible for 9.9% [95% UI; 8.3, 11.6] of deaths due to diarrheal diseases and 12.0% [0.5, 23.1] of deaths due to measles [1]. Over the past few decades, efforts to increase coverage have decreased the burden of disease in children [1, 3]. From 1995 to 2005, the number of preschool-age children with VAD dropped from 251 million to 190 million globally [3]. However, in 2019, over 3 million disability-adjusted life years (DALYs) were still lost due to VAD globally [1]. Reducing the prevalence of VAD in children under 5 years of age could be an important part of the strategy to meet Sustainable Development Goal 3.2 of reducing child mortality to at most 25 per 1,000 live births by 2030 [4].

VAS has been an intervention that was popular for its low cost, varying between 0.50–1.50 USD per capsule based on spatial and temporal differences [5-7] while maintaining high efficacy in lowering VAD. It is commonly distributed as part of Child Health Days (CHD), which is a mass campaign designed to deliver immunizations and nutritional supplements to children. CHDs often last for a week, take place semiannually, and are a national effort to locate children though community-based or outreach campaigns. Two supplements every year for children up till the age of five is enough to lower the risk of death associated with deficiency [8]. Systematic reviews have shown that the relative risk of VAS on all-cause mortality and on night blindness are 0.88 [0.83, 0.93] and 0.32 [0.21, 0.50], respectively [9].

In recent years, some experts have suggested providing other interventions like the measles vaccine and large-scale food fortification in conjunction with VAS in order to reduce the incidence associated with measles and diarrhea [9]. Experts still vigorously debate the best course of action to tackle VAD moving forward [10]. There is very little recent literature on the cost-effectiveness ratio of VAS, and many of these papers do not consider VAS individually but rather as one among a group of interventions [11]. Using simulation results, we can determine the cost-effectiveness of VAS as a standalone intervention. Leveraging data from the Global Burden of Disease (GBD) 2019 study, we set out to estimate the cost-effectiveness of VAS in an individual-based model using the Vivarium microsimulation framework.

With a Vivarium model calibrated to estimates from the GBD 2019 study, we projected the impact of increasing coverage of VAS in three sub-Saharan African countries with different levels of existing coverages: Burkina Faso, Kenya, and Nigeria. We aimed to estimate the impact of scaling up this intervention by measuring the change in exposure to VAD and the corresponding change in DALYs from year to year. We compared the outcomes of the interventions in the three countries to investigate how different levels of baseline coverage affect the cost-effectiveness of scale-ups. Furthermore, we calculated the DALYs averted and costs of the scale-up to add up-to-date, quantitative evidence to support decisions on which interventions would be worth pursuing.

Materials and methods

Vivarium model

We used Vivarium, an individual-based, discrete-time, Monte Carlo simulation framework developed by the Simulation Science team at the Institute of Health Metrics and Evaluation [12]. In the Vivarium model, we track a population of simulants over the period of the simulation. For each individual simulant, we track whether they are supplemented, have been exposed to VAD, have a disease, and are alive on every day of the simulation. With this framework, we can calculate the cost-effectiveness of increasing the coverage of VAS by running two distinct simulations. First, we run a baseline simulation that reflects the scenario where the coverage of VAS remains constant over the duration of the simulation. Next, we simulate the alternative scenario where the coverage of VAS increases steadily over the course of the simulation. We track the DALYs and cost of each scenario, and we can then obtain the DALYs averted and additional cost.

Vivarium includes modular components that incorporate distributions and data from the Global Burden of Disease (GBD) [1]. These components include an intervention model, a risk exposure model, a risk-effect model, a cause model, a mortality/morbidity model, and a cost model (). Each individual component requires inputs to be modelled. For example, the risk exposure model would need risk exposure prevalence; the risk-effect model would need relative risks for risk-cause pairs; the cause model would need disease prevalence and incidence; the mortality model would need cause-specific mortality rates and all-cause mortality rates. We used incidence, prevalence, mortality, and relative risk data from GBD 2019 to calibrate the risk-effect model, the cause model, and the mortality/morbidity model. We conducted a meta-analysis to determine the inputs in the other models.

VAS in children is meant to reduce VAD, which in turn, affects measles and diarrheal diseases (). We used Vivarium to model the two causes of death (measles and diarrheal diseases) as well as the risk factor of Vitamin A deficiency by calibrating the simulation to estimates from GBD 2019 that produced relative risks for risk-cause and cause-outcome pairs. We use compartmental models to describe the dynamics of causes of death. The measles component utilized a Susceptible-Infected-Recovered (SIR) model with a 10-day duration for the infected state. While all individuals are initially susceptible to measles, some fraction may be exposed to it over time. Those who recover from measles will never be susceptible to it again, so we use an SIR model to describe it. The diarrheal disease component used a Susceptible-Infected-Susceptible (SIS) model because individuals who get diarrhea will still be susceptible to it after they recover. At each time step, based on the appropriate health models and country-level data from the GBD database, the simulation determined whether the state of each individual changed with regards to either of the two diseases. We chose a step size of 1 day to be able to capture the possibility of being inflicted with a disease on successive days in the case of diarrhea.

At model initialization, Vivarium assigns the age, sex, and VAD status of simulants. VAD is determined as a serum retinol concentration of less than 0.70 μmol/L. We model VAD as a binary categorical variable. At each time step, the simulation samples probability distributions using a Monte Carlo method to determine changes in a simulant’s status. These changes include whether an individual receives VAS, experiences incidence of diarrheal disease, measles, or changes to their VAD status.

First, we determined the initial characteristics of the population that we would simulate. We used an open cohort of 1,000,000 individuals between the ages of 0 and 59 months. While we tracked individuals in the simulation starting from birth, we only simulated the effect of supplementing individuals between 6–59 months, the ages eligible for VAS. We stopped tracking simulants whose age exceed 59 months. For both the baseline and alternative scenarios, we ran one simulation each from 2017 until 2023. For the alternative scenario, we modeled a scale-up of VAS coverage starting in 2019. We allowed the alternative simulation to start two years prior to scale up to verify that the model outputs were calibrated to the GBD 2019 estimates relevant to Vitamin A. When determining the impact and cost-effectiveness of the intervention, we only consider the time period from 2019 to 2023 in the simulations.

Although the GBD 2019 Study includes estimates of the exposure and relative risk of VAD, as well as the coverage of VAS, it does not include information on the effectiveness of VAS to reduce VAD. We modelled VAS as a dichotomous attribute, with one possibility being the simulant received two doses of Vitamin A in one year and the other being that they did not receive the supplementation at all. We modeled the scale-up of coverage of VAS as stepping up from the baseline to the target value at equal intervals each year from 2019 to 2023 (held constant throughout each year). We determined the baseline coverage and intervention effectiveness using a systematic review. We employed a fixed-effect meta-analysis to synthesize the values in the literature review (see Vitamin A Supplementation Coverage and Vitamin A Supplementation Effect sections below). A fixed-effect meta-analysis assumes that our studies are measuring a shared effect value that deviates from study to study due to only sampling error. To generate values for coverage and relative risk, we weighted the values from different studies based on their sample sizes, where studies with more individuals have larger weights.

Finally, we calculated the cost for the intervention from the perspective of the group distributing the doses, including the costs of the capsule, training, and transportation. We determined the number of doses supplied by the simulation and multiplied that value by a unit cost. The unit cost was inferred from a literature search on VAS [5-7].

Simulation inputs

Vitamin A supplementation coverage

For our estimate of VAS baseline coverage, we considered using the UNICEF database of VAS coverage by country for our estimates. However, due to the variation of estimates from year to year for a few of our countries of interest (specifically Nigeria and Kenya), we decided to use other sources.

GBD has a covariate for VAS coverage measuring the proportion of children who received one dose of Vitamin A over the last six months. In other work, this measure has been used as a proxy for the coverage of two doses over the past twelve months [8]. We used the GBD covariate and a separate literature search to determine the baseline coverage for each location. If the results of the literature search were similar to the GBD covariate, we used the value of the GBD covariate for the baseline coverage.

For the literature search, we included DHS surveys and articles from Google Scholar and PubMed. We included only sources that defined supplementation to be two doses within twelve months or one dose in six months. We used a fixed-effect meta-analysis based on each source’s coverage and sample size to determine the value for the baseline coverage in each region [13]. For sources that did not provide a sample size, we assigned to them the smallest sample size of all the other papers.

The target coverage in all three countries was chosen so that the exposure to the lack of VAS was halved to represent an intervention scale-up of similar relative intensity for each country. We ran simulations for two scenarios: one representing the “business as usual” scenario with VAS coverage held constant at the 2017 level until 2023, and the other representing the VAS scale-up from the existing coverage in 2019 to the target coverage by 2023. We compared the deaths and DALYs in the scenarios in each country to find the impact of VAS scale-up.

Vitamin A supplementation effect

To determine the efficacy of VAS in reducing VAD as a risk factor, we performed a literature search of online articles. We used a combination of the search terms “Vitamin A supplementation,” “Vitamin A deficiency,” and “relative risk” in Google Scholar and PubMed to locate relevant articles. Of the articles found in these databases, we filtered out the ones that did not use the standard definitions for VAS (frequency of capsule distribution is twice per year) and VAD (serum retinol concentration less than 0.70 μmol/L) [8]. Additionally, we included articles that provided a value for the relative risk of VAS on VAD and excluded those that provided only relative risk of other outcomes like mortality, night blindness, and causes of death (more details in ).

We quantified the effect of a lack of VAS on VAD in terms of the relative risk, meaning the ratio of VAD prevalence among those without VAS to those with VAS. By conducting a fixed-effect meta-analysis on these online sources, we determined distributions for our simulation to decide which individuals would be afflicted with VAD based on their supplementation status.

Intervention unit cost

We calculated the overall cost of both the baseline and alternative scenarios by finding the number of doses distributed in the simulation and multiplying that by a unit cost. We created an observer in the simulation that reported the total number of supplemented years. One supplemented year corresponds to an individual who was alive, between 6–59 months old and was provided the two doses of Vitamin A over a one-year time period. We doubled the number of supplemented years to determine the number of doses distributed in a particular year of the simulation.

We determined the unit cost from a recent study on costs of Vitamin A interventions for the semiannual CHD program [7]. It considered costs for the capsule, training, wages, and transportation to provide the unit cost per dose. After adjusting this value by inflation, we arrived at a cost per dose of $0.60 USD. We took the product of the simulation’s doses and the unit cost to determine the overall cost per scenario. The additional cost of the intervention was calculated by comparing the costs of the baseline and intervention scenarios.

Analysis of simulation results

We tallied the population stratified on sex, year, and age with regards to our outcomes of interest, including years of life lost (YLL) and years lived with disability (YLD). From our simulation we added the number of YLLs and YLDs to find the total quantity of disability-adjusted life years (DALYs) throughout the simulation. We arrived at the number of DALYs averted in each location by finding the difference between the baseline and alternative scenarios. Similarly, we calculated the additional cost of the intervention in each region. Using these two values, we estimated the Incremental Cost-effectiveness Ratio (ICER), which is ratio of the change in cost to the DALYs averted. The ICER is useful for putting the health impact in perspective of the additional cost of the intervention.

The main two sources of uncertainty that we model in our simulation are parameter uncertainty and stochastic uncertainty.

We modelled our input parameters from GBD and literature searches as probability distributions. We utilized a Monte Carlo technique to draw samples from the distributions to generate random values. For each input, we propagated parameter uncertainty through the model and captured the outcomes (including DALYs averted, additional cost, and ICER) for each draw. We used a total of 100 draws for each country and took the median to represent the results for each scenario. We chose to summarize our results using the median instead of the mean as the median is resistant to outliers. Using the outcomes of the 100 draws, we also generate an uncertainty interval for the DALYs averted, additional cost, and ICER values.

Stochastic uncertainty is meant to model the variability across scenarios among identical individuals within the population that have the same age and sex. We wanted to run the baseline and alternative scenarios with the same stochastic variation, to allow us to examine the differences in the outcomes due to parameter uncertainty, while minimizing random noise. Vivarium uses common random numbers to reduce this variance between individuals [14]. Common random numbers ensure that the values drawn from an input parameter distribution are the same for both the baseline and alternative scenario in each draw.

Verification and validation

We validated our simulation by comparing its estimates for health outcomes in the years 2017–2018 with the data collected in the GBD 2019 study. For each of the three countries, we compared the rate of YLLs and YLDs with regards to the two diseases that are affected by VAS. For each metric and disease, we stratified our results by sex and by age group (post neonatal, and 1–4 years) as we believed that these variables would substantially affect the rate of YLLs and YLDs.

In addition, we performed some more validations over the intervention years. We confirmed the coverage of VAS scales up in the alternative scenario and remains constant in the baseline scenario. We compared the relative risk of VAD with the value provided by GBD and verified that the effectiveness of VAS on reducing VAD matches the relative risk given by the meta-analysis.

Results

We found that the existing literature backed the GBD covariate for coverage in Burkina Faso [1, 15–18], so we used the GBD value for our simulation input. However, for Kenya and Nigeria, we judged the GBD covariate for supplementation coverage to be outdated and used the results of the literature search.

Overall, we found that the baseline coverage for Nigeria was 32.1% [31.8, 32.3] [19-26], for Kenya it was 55.2% [54.7, 55.7] [25-31], and for Burkina Faso it was 88.4% [85.3, 91.9] [1]. The values we used in the simulation for baseline coverage were 32%, 55%, and 88% for Nigeria, Kenya, and Burkina Faso, respectively. As a result, for Nigeria the target coverage was 66%, for Kenya it was 77%, and for Burkina Faso it was 94%.

Our literature search for the relative risk of lack of VAS on the risk factor VAD provided 1,020 hits in Google Scholar and 33 hits in PubMed. After checking the articles based on our eligibility criteria, four sources eventually went into the fixed-effect meta-analysis. We found that the relative risk was 1.48 [1.05, 2.08] [32-35].

We used GBD 2019’s estimates to model the relationship between risk factors and causes as well as the relationship between causes and outcomes ( [1].

Simulation outputs

The simulation results show benefits of scaled-up interventions in each of the countries involved. Over the seven-year period, 21 [5, 56] DALYs per 100,000 person-years were averted in Nigeria, 21 [5, 47] DALYs per 100,000 person-years were averted in Kenya, and 14 [0, 37] DALYs per 100,000 person-years were averted in Burkina Faso ().

As coverage increased, the DALYs averted per year rose as well for all countries (see for yearly simulation outputs). In 2017 and 2018, the number of DALYs averted were zero, when supplementation was yet to begin. Starting from 2019, the DALYs averted increased at a constant rate in all three regions depending on the size of the yearly coverage increase.

For each nation, the simulation records the number of supplemented years (the total number of person-years the simulation decided to supplement individuals with the intervention) which takes into account the current coverage. As the scale-up was smallest in Burkina Faso, the additional cost to implement the intervention was the least while Nigeria had the greatest cost. The additional cost per 100,000 person-years amounted to $17,880 [17,800, 17,960] in Nigeria, $11,390 [11,350, 11,430] in Kenya, and $3,100 [3,060, 3,120] in Burkina Faso (). The additional cost per year increased at a rate proportional to the increase in coverage for each specific intervention.

Despite averting the fewest DALYs over the course of the simulation, the intervention in Burkina Faso has the lowest median ICER among the countries. Overall, the intervention over the seven-year period would have an ICER of $860/DALY [320, 3530] in Nigeria, $550/DALY [240, 2230] in Kenya, and $220/DALY [80, 2470] in Burkina Faso ().

Validation

We validated the simulation output for each nation in 2017 and 2018 by comparing it to GBD estimates from that year. We stratified the simulation results by age, sex, and cause for the comparison. In particular, the two age groups that we validated were post neonatal and 1–4 years of age since the intervention is only provided to children between 6–59 months of age. The mean rate of YLLs and YLDs from that year over 100 draws are within 10% of the GBD value for both age groups and sexes for measles. Those values for diarrheal diseases are within 20% of the average in Nigeria and Burkina Faso but are well within the range of the parameter uncertainty estimated by GBD.

Discussion

By comparing the effect of a scale-up of VAS coverage in three countries with varying baseline coverages, we measured the changes in the cost-effectiveness of different levels of scale-ups. In Nigeria, where the baseline coverage was the smallest, we simulated an intervention coverage scale-up from 32% to 66%, while in Burkina Faso, where the baseline coverage was the largest, we simulated an intervention coverage scale-up from 88% to 94%. The model found that scale-ups in all three nations produce similar rates of DALYs averted, all with overlapping uncertainty intervals. A surprising result is that although Nigeria had a larger scale-up than Kenya, the two nations had the same median DALYs averted. This can be attributed to the increased prevalence of VAD in Kenya compared to Nigeria (see ).

In addition, the model predicted that Nigeria has the most additional cost whereas Burkina Faso has the fewest. Similarly, the intervention in Burkina Faso is the most cost-effective and the scale-up in Nigeria is the least. However, we note that the model’s estimates for ICER have wide and overlapping uncertainty intervals. Because ICER is the ratio between additional cost and DALYs averted, one factor alone may not sufficiently reflect the trend in the cost-effectiveness of an intervention. Although Burkina Faso had the fewest DALYs averted, the intervention had the lowest ICER because it has the smallest cost. There is substantial uncertainty in the DALYs averted and ICER values as well, but in all settings the median ICER is between 100–1000 USD/DALY.

Overall, our simulation study finds that VAS is not as cost-effective as previously reported in the literature. The cumulative ICER estimates for the nations in our simulation are approximately ten times larger than previous cost-effectiveness analyses of this intervention [11, 36]. Edejer’s study from 2005 considers the cost-effectiveness of different combinations of nutrition-related interventions in sub-Saharan Africa. The authors found that VAS, Zinc Supplementation, Measles vaccination, and pneumonia case management together have an ICER of $85/DALY [11]. However, this study measures the effectiveness of VAS combined with other interventions, and therefore may not be directly comparable with our results. Another analysis by Chow and colleagues in 2010 found that VAS by itself had an ICER ranging between $23-50/DALY within India [36]. The differences in these values compared to our simulation’s outputs may be due in part to the combination of interventions or differing locations of the studies.

However, we believe that the increase in ICER can be mainly attributed to the decrease in VAD over time and new estimates for the risk effect of VAD on causes of death. For example, among children under 5 years of age in Nigeria, VAD was responsible for 5,850 DALYs per 100,000 person-years in 2005 and 1,860 DALYs per 100,000 person-years in 2010, but it only caused 690 DALYs per 100,000 person-years in 2019 [1]. A lower prevalence of VAD would allow fewer DALYs to be averted, which would result in a larger ICER.

Changes in the modelling of VAS in GBD 2019 might have further reduced the impact of VAS. GBD 2019 used a new meta-analysis method for micronutrient modelling that was not applied in GBD 2017 [37]. The results of the new modelling approach reduced the effect of VAD on measles and diarrheal diseases by an order of magnitude [1, 38]. GBD 2019 also found that lower respiratory infections (LRI), a cause of death associated with VAD in GBD 2017, did not have sufficient evidence to estimate a causal relationship with VAD [1]. As a result, we chose not to include LRI in our model. We ran a separate simulation that was calibrated to GBD 2017 estimates and included LRI. This model found that ICERs were more similar to previous literature: Nigeria had an ICER of 41 [26, 61], Kenya had an ICER of 62 [46, 97], and Burkina Faso had an ICER of 35 [22, 59]. This illustrates the importance of the inputs to our model and how it can affect the results of our simulation.

Our cost-effectiveness analysis has two major strengths: the flexibility of the Vivarium framework, and the ability to incorporate uncertainty into the model. The Vivarium framework allows us to create models that are very flexible while calibrating population-level parameters to GBD estimates. The model used in our simulation only considered VAS, which is sometimes not possible in the field because individuals may be subject to multiple interventions simultaneously. In fact, much of the literature on cost-effectiveness of nutritional supplements combines different interventions when estimating the ICER. Additionally, the microsimulation model allows us to vary the number of draws and individuals in each simulation. This gives us the ability to incorporate uncertainty in our simulations.

Our approach has two limitations due to the scope of our model. In addition to lowering VAD by increasing VAS coverage, campaigns often boost the intake of VA as well. While GBD estimates for VAD include data on VA intake, we did not model VA intake explicitly. Second, our model does not incorporate the YLDs stemming from xerophthalmia and night blindness. However, the YLDs are small compared to the YLLs of the causes that we did include, so we believe that this omission does not greatly affect our results.

Another area for future improvement is our cost model. We determined the overall cost of the intervention using a singular value for the average unit cost of VAS (see for more details on the literature search). We used sources from journal publications only, but there are other sources that claim the unit cost might be larger. The WHO One Health Tool estimates that the unit cost can range between $2.45–3.17 USD. Increases in unit cost would create a proportional increase in ICER. For example, if the unit cost of VAS were five times larger, the additional cost and ICER would increase by a factor of five as well. Another drawback of our cost model is that it assumes that the intervention cost increases linearly as coverage expands. However, costs are likely to vary in different nations and even subnational regions based on accessibility, budget, and the size of the program. As a result, cost has a superlinear relationship with intervention coverage in the real world. Although our simulation can take uncertainty into account, it might not have enough complexity to consider all the factors of a realistic cost model. Up-to-date cost data as well as studies analyzing the nonlinear costs of supplementing individuals in isolated regions would improve the quality of this analysis.

Conclusion

In this study, we calculated the ICER as a measure of the cost-effectiveness of scale-ups of VAS coverage in three countries using an individual-based microsimulation calibrated to match GBD estimates at the population level. We found that the intervention was not as cost-effective as it has been reported previously. This is due to falling levels of VAD among children in our countries of interest and lower risk effects for causes of death due to VAD in GBD 2019.

Simulation inputs.

(DOCX)

Click here for additional data file.

Yearly results.

(DOCX)

Click here for additional data file.

Transfer Alert

This paper was transferred from another journal. As a result, its full editorial history (including decision letters, peer reviews and author responses) may not be present.

22 Dec 2021

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: https://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols. Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols.

We look forward to receiving your revised manuscript.

Kind regards,

Susan Horton

Academic Editor

PLOS ONE

Journal Requirements:

1. When submitting your revision, we need you to address these additional requirements.

Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

https://journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and

https://journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

2.Thank you for stating the following in the Funding information of your manuscript:

(The Bill and Melinda Gates Foundation (OPP-1170133) funded this project. The funders had no role in the decision to publish.)

We note that you have provided funding information that is not currently declared in your Funding Statement. However, funding information should not appear in the Acknowledgments section or other areas of your manuscript. We will only publish funding information present in the Funding Statement section of the online submission form.

Please remove any funding-related text from the manuscript and let us know how you would like to update your Funding Statement. Currently, your Funding Statement reads as follows:

(ADF received funding from the Bill and Melinda Gates Foundation (https://www.gatesfoundation.org) under grant OPP-1170133. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.)

Please include your amended statements within your cover letter; we will change the online submission form on your behalf.

3. Please include captions for your Supporting Information files at the end of your manuscript, and update any in-text citations to match accordingly. Please see our Supporting Information guidelines for more information: http://journals.plos.org/plosone/s/supporting-information.

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Partly

Reviewer #2: No

**********

2. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: I Don't Know

Reviewer #2: No

**********

3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

Reviewer #2: Yes

**********

4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #2: No

**********

5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: Please see the attached - as this version does not include the hyper links the reviewer provided:

The paper is very well thought out, very relevant, and timely in its topic. VAS is an essential child survival strategy in populations where VAD is a public health problem among children 6-59 mos of age. Despite it being very effective to reduce morbidity and mortality associated with VAD, it is an intervention that has been in place in some countries for over 20 years, and commitment to maintain such programs is waning in some countries. Cost-effectiveness data is critical to engaging policy makers, and advocating for continuation of this life-saving public health intervention especially as health systems are being stretched to the limits of their resources. For this reason we really welcome this analysis, and the rigour to which the authors brought to it.

Understanding that the authors simulated the scale up of VAS coverage from a hypothetical baseline of coverage to demonstrate that across three different settings, the effort to scale up the intervention is very cost effective.

I recommend however that this paper be revisited to consider the following:

1) The GBD 2019 significantly revised the methodology used and cited the following major changes that resulted in a very large reduction in # of DALYs due to VAD (reduced to 3.3M globally, and specifically 2.63M DALY’s in children U5):

• Vitamin A deficiency and vitamin A supplementation were modelled in ST-GPR to achieve improved time trends.

• Vitamin A supplementation estimates are now age-sex specific since supplementation campaigns target children.

• The age-specific stunting SEV was added as a covariate for vitamin A deficiency, alongside the three used last year: SDI, the availability of retinol activity equivalent (RAE) units in foods, and newly updated vitamin A supplementation.

• The evidence on vitamin A deficiency as a risk factor for diarrhoea, measles, and lower respiratory infections (LRIs) was re analysed and evaluated using MR-BRT. LRIs were removed as an outcome due to insufficient evidence, and the relative risks for diarrhoea and measles were updated. Notably, we no longer adjust relative risks for background vitamin A deficiency prevalence.

Attached to this review is the Lancet ref, as well please see this recently published discussion paper: Basis for changes in the global burden of disease estimates related to vitamin A and zinc deficiencies in the GBD 2017 and 2019 studies | Public Health Nutrition | Cambridge Core

Could the authors consider re-running the analysis with the GBD 2019 data? If not, I strongly recommend this be acknowledged in the introduction as well as in the Discussion as a limitation, and consider what the ICER would be if the GBD 2019 was the source instead.

2) Line 67 references a paper that compared measles vaccination and therapeutic VAS in reducing mortality due to measles. While I am unable to access the full paper referenced (#10), it appears that that specific paper was considering the number of capsules given to treat a case of measles (in accordance with WHO Guidance on the therapeutic uses of VAS) – whereas the authors are comparing this to the two doses provided in PREVENTIVE programs. While preventive VAS can reduce incidence of measles by 50% (Imad et al 2017) – it is not found to reduce mortality due to measles.

3) Line 67, in addition to the above also has the following error: Large scale Food fortification does not reduce deaths associated with measles, LRI and diarrhoea – instead it is a strategy to increase the consumption of Vitamin A through the daily diet, and as such, can be considered as a means to reduce VAD in the population over time. Vitamin A supplementation on the other hand, only temporarily reduces VAD in the population, and so if a child consumes two high dose supplements, the supplement is protecting them from the effects of VAD (caused by the low intake of VA from the diet). As a result, VAS and fortification (and other dietary strategies) are complementary to each other – not alternatives. The use and interpretation of serum retinol distributions in evaluating the public health impact of vitamin A programmes | Public Health Nutrition | Cambridge Core

4) Line 72 – refers to VAS as a treatment – however VAS programs are a preventive public health intervention. Recommend replacing “treatment” with the term “intervention”

5) Figure 1: this would change if GBD 2019 were considered instead

6) Line 128 – the cohort is referred to as children 0-59 months of age – however the eligible ages for VAS is 6 months to 59 months of age.

7) Line 139: VAS does not treat VAD. Recommend the change to “effectiveness of VAS to reduce VAD”

8) Line 161 that describes the source of the coverage data used in the simulation. Is there a reason why the authors did not consider using the globally available UNICEF database of national VAS coverage data by country? This data is administrative data, carefully reviewed and curated by UNICEF Data and Analytics, and publishes coverage by year, by country, by semester, and two-dose coverage –and can be found here: Vitamin A Deficiency in Children - UNICEF DATA The user can also download the full excel dataset using the link on that page. This would be the better source for actual VAS coverage in a given country in a given year, as this is also the globally accepted source for VAS coverage data included in the Global Nutrition Report (GNR). If the authors do not decide to use the UNICEF data, I recommend the reason be included in the paper.

9) As a reviewer, my expertise is in the area of VAS programs, and Vitamin A nutrition and public health and I do not consider myself qualified to comment on statistical analysis, which is why I answered Q#2 in that manner.

10) Line 326: recommend the authors provide more context to make the connection to the Edejer paper and the relevance. In addition, suggest the authors also clarify if the $85/DALY result was all of the interventions combined, or each intervention at a time etc.

11) Line 360: could the authors build this out a little bit more to describe some additional cost drivers that are relevant to VAS programs. Costs are more likely to vary due to a) the delivery mechanism/or platform used in each country to deliver the VAS to children twice/year (routine, or campaign, or what kind of campaign), and b) the strength of the health system. Some countries need to spend a lot of money to deliver VAS to every child because the health system is not already reaching them. Did the authors feel there was sufficient data on costs in the literature? If not, could the authors make a recommendation of some kind – and comment on how this would improve the quality of future analysis such as this.

Reviewer #2: Large Issues:

1. The core results emerge logically from the assumptions and modeling approach – if more children receive VAS, there will be fewer VA-preventable deaths, and if program costs are linear then the cost-effectiveness parameters look promising. However, there are reasons to question the mortality-reducing impacts of VAS in today’s world (estimates are old and new ones cannot be generated for ethical reasons), and especially to question the linearity of costs of expanding coverage. There are reasons why coverage is so low in Nigeria, for example, and these suggest increasing costs as coverage increases, and perhaps some non-cost-related thresholds that need to be considered.

2. There is a big difference between inadequate intake of VA and VAD. A thorough analysis of the impact of a VAS campaign should consider both.

3. The authors use the suggestions that (e.g.) LSFF might be a better way of delivering VA to children to justify their analyses. However, no modeling is done to capture the cost-effectiveness of alternative VA deliver platforms, so the justification lacks punch. Vosti et al. (2020) and others have done such work.

4. It is not exactly clear what the authors mean when they say VAS; what are the VAS distribution details and associated cost functions? There are many ways to manage VAS programs, and the impacts and costs vary dramatically depending on the approach chosen.

5. Most countries in sub-Saharan Africa have found campaign-based VAS to be very expensive – indeed, if the international community were not covering the costs of these campaigns (including all inputs and many of the management costs), they would not be undertaken. So, what the authors suggests as unit costs seem not to jibe with reality; indeed, the country-level estimates of program expansion costs (even after adjusting for target population size) are an order of magnitude or more smaller than estimates offered by others. See Schott et al. 2021 for an example of the structure of costs for MNP distribution – not the same intervention, but the cost structure will be similar for VAS campaigns.

6. The authors note the importance of addressing uncertainty in their modeling efforts, but uncertainty regarding the impact of VAS on VAD-attributable mortality and (especially) regarding scale-up costs seem not to have received such attention.

7. The authors contend that VAS is a highly cost-effective intervention; new literature and ongoing experience in the field suggest that this may not be the case. This is an especially contentious claim based on linear costs of scaling up VAS programs.

Smaller Issues:

1. It is not clear why these three countries were chosen.

2. The authors admit that there are no YLD parameters in this simulation exercise, so best to exclude them. Indeed, I would suggest presenting results in terms of deaths averted (rather than DALYs), since no comparisons are being made with other interventions with different mortality/disability structure.

3. It is not clear what ‘individual-based’ means, nor it is clear what calibrations were undertaken to ‘match’ (how well?) GBD estimates. What are SIR and SIS models? What does fixed-effect metanalysis mean, in the several places in which the term is used? Generally, the paper is not accessible to those outside the IHME and their sphere of influence; this can be fixed, but some additional effort will be required.

**********

6. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

Reviewer #2: No

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

Submitted filename: PLOS review_Dec20.docx

Click here for additional data file.

Submitted filename: GBD 2019_Annex vitamin-A-deficiency.pdf

Click here for additional data file.

Submitted filename: PLOS ONE Review 12 21 21.docx

Click here for additional data file.

29 Jan 2022

Response to reviewers for Vitamin A Supplementation

Reviewer 1

The paper is very well thought out, very relevant, and timely in its topic.

Thank you very much for this positive assessment of our work.

VAS is an essential child survival strategy in populations where VAD is a public health problem among children 6-59 mos of age. Despite it being very effective to reduce morbidity and mortality associated with VAD, it is an intervention that has been in place in some countries for over 20 years, and commitment to maintain such programs is waning in some countries. Cost-effectiveness data is critical to engaging policy makers, and advocating for continuation of this life-saving public health intervention especially as health systems are being stretched to the limits of their resources. For this reason we really welcome this analysis, and the rigour to which the authors brought to it.

We thank the review again!

Understanding that the authors simulated the scale up of VAS coverage from a hypothetical baseline of coverage to demonstrate that across three different settings, the effort to scale up the intervention is very cost effective.

I recommend however that this paper be revisited to consider the following:

1) The GBD 2019 significantly revised the methodology used and cited the following major changes that resulted in a very large reduction in # of DALYs due to VAD (reduced to 3.3M globally, and specifically 2.63M DALY’s in children U5):

• Vitamin A deficiency and vitamin A supplementation were modelled in ST-GPR to achieve improved time trends.

• Vitamin A supplementation estimates are now age-sex specific since supplementation campaigns target children.

• The age-specific stunting SEV was added as a covariate for vitamin A deficiency, alongside the three used last year: SDI, the availability of retinol activity equivalent (RAE) units in foods, and newly updated vitamin A supplementation.

• The evidence on vitamin A deficiency as a risk factor for diarrhoea, measles, and lower respiratory infections (LRIs) was reanalysed and evaluated using MR-BRT. LRIs were removed as an outcome due to insufficient evidence, and the relative risks for diarrhoea and measles were updated. Notably, we no longer adjust relative risks for background vitamin A deficiency prevalence.

Attached to this review is the Lancet ref, as well please see this recently published discussion paper: Basis for changes in the global burden of disease estimates related to vitamin A and zinc deficiencies in the GBD 2017 and 2019 studies | Public Health Nutrition | Cambridge Core

Could the authors consider re-running the analysis with the GBD 2019 data? If not, I strongly recommend this be acknowledged in the introduction as well as in the Discussion as a limitation, and consider what the ICER would be if the GBD 2019 was the source instead.

We thank the reviewer for identifying this improvement to our results. We were indeed able to rerun the analysis with the GBD 2019 data and found that it substantially changed the outcomes of the simulation!

Using GBD 2019’s data, we found that the risk effects for causes of death due to VAD were significantly reduced. We also excluded LRI from our model and this contributed to the higher ICERs that we now report in our manuscript. Specifically, our model reported an ICER of $860/DALY [320, 3530] in Nigeria, $550/DALY [240, 2230] in Kenya, and $220/DALY [80, 2470] in Burkina Faso.

Because this is much larger than what others have reported previously, we revised the tables, discussion, and conclusion of our paper to reflect our new interpretation based on the model. We pointed out the changes of the latest estimates from GBD 2019 and how these inputs affected our model. We compare our results briefly with the outputs we reported earlier when we submitted a model that was calibrated with estimates from GBD 2017. For reference, below is the original set of results from GBD 2017 and GBD 2019:

Country Overall DALYs Averted Per 100,000 Person-Years Overall Additional Cost Per 100,000 Person-Years (USD / 100,000 Person-Years) Cumulative ICER (USD/DALY)

Nigeria 440 (290, 680) 17,670 (17,560, 17,730) 41 (26, 61)

Kenya 180 (120, 250) 11,340 (11,300, 11,380) 62 (46, 97)

Burkina Faso 90 (50, 140) 3,120 (3,090, 3,140) 35 (22, 59)

Table 1: Results from GBD 2017 inputs

Country Overall DALYs Averted Per 100,000 Person-Years Overall Additional Cost Per 100,000 Person-Years (USD / 100,000 Person-Years) Cumulative ICER (USD/DALY)

Nigeria 21 (5, 56) 17,880 (17,800, 17,960) 860 (320, 3530)

Kenya 21 (5, 47) 11,390 (11,350, 11,430) 550 (240, 2230)

Burkina Faso 14 (0, 37) 3,100 (3,060, 3,120) 220 (80, 2470)

Table 2: Results from GBD 2019 inputs

2) Line 67 references a paper that compared measles vaccination and therapeutic VAS in reducing mortality due to measles. While I am unable to access the full paper referenced (#10), it appears that that specific paper was considering the number of capsules given to treat a case of measles (in accordance with WHO Guidance on the therapeutic uses of VAS) – whereas the authors are comparing this to the two doses provided in PREVENTIVE programs. While preventive VAS can reduce incidence of measles by 50% (Imad et al 2017) – it is not found to reduce mortality due to measles.

We thank the reviewer for pointing out this area for improvement in our exposition. We have updated the text and reference to make the mechanism by which measles vaccination could impact the cost-effectiveness of VAS more precise.

3) Line 67, in addition to the above also has the following error: Large scale Food fortification does not reduce deaths associated with measles, LRI and diarrhoea – instead it is a strategy to increase the consumption of Vitamin A through the daily diet, and as such, can be considered as a means to reduce VAD in the population over time. Vitamin A supplementation on the other hand, only temporarily reduces VAD in the population, and so if a child consumes two high dose supplements, the supplement is protecting them from the effects of VAD (caused by the low intake of VA from the diet). As a result, VAS and fortification (and other dietary strategies) are complementary to each other – not alternatives. The use and interpretation of serum retinol distributions in evaluating the public health impact of vitamin A programmes | Public Health Nutrition | Cambridge Core

We thank the reviewer for identifying this point, which we have now clarified in the text. We agree that VAS and LSFF should be considered complementary approaches.

4) Line 72 – refers to VAS as a treatment – however VAS programs are a preventive public health intervention. Recommend replacing “treatment” with the term “intervention”

We thank the reviewer for pointing out this issue. We have corrected this description of VAS in the manuscript.

5) Figure 1: this would change if GBD 2019 were considered instead

As we switched to a model with GBD 2019 estimates, we have now also updated Figure 1 to reflect the model diagram without LRI as a cause of death associated with VAD.

6) Line 128 – the cohort is referred to as children 0-59 months of age – however the eligible ages for VAS is 6 months to 59 months of age.

We thank the reviewer identifying this point of confusion. Our model does track individuals between the ages of 0-59 months as more individuals with age 0 are added to the simulation based on the population birth rate. However, we only simulate the effect of VAS on children between ages 6-59 months. Our explanation of the cohort on line 128 was confusing and it has been modified to describe our model more explicitly.

7) Line 139: VAS does not treat VAD. Recommend the change to “effectiveness of VAS to reduce VAD”

We thank the reviewer for pointing out this issue. We have changed the wording to accurately reflect the relationship between VAS and VAD.

8) Line 161 that describes the source of the coverage data used in the simulation. Is there a reason why the authors did not consider using the globally available UNICEF database of national VAS coverage data by country? This data is administrative data, carefully reviewed and curated by UNICEF Data and Analytics, and publishes coverage by year, by country, by semester, and two-dose coverage –and can be found here: Vitamin A Deficiency in Children - UNICEF DATA The user can also download the full excel dataset using the link on that page. This would be the better source for actual VAS coverage in a given country in a given year, as this is also the globally accepted source for VAS coverage data included in the Global Nutrition Report (GNR). If the authors do not decide to use the UNICEF data, I recommend the reason be included in the paper.

We thank the reviewer for pointing out this improvement for our method description. We found that the coverage in the UNICEF database for two of our countries of interest (Nigeria and Kenya) varied tremendously from year to year. If we used the coverage value from a specific year for the simulation, it was not clear to us how accurate it would be given that the measured coverage might shift largely the next year. In addition, for the third nation (Burkina Faso), the UNICEF database estimated that the coverage was 97-99 percent, which is very large and perhaps unlikely.

For these reasons, we chose to conduct our own literature search for coverage in our countries of interest. We have included this explanation in the text.

9) As a reviewer, my expertise is in the area of VAS programs, and Vitamin A nutrition and public health and I do not consider myself qualified to comment on statistical analysis, which is why I answered Q#2 in that manner.

10) Line 326: recommend the authors provide more context to make the connection to the Edejer paper and the relevance. In addition, suggest the authors also clarify if the $85/DALY result was all of the interventions combined, or each intervention at a time etc.

We appreciate the reviewer’s advice to add more detail to this part of the discussion to justify the comparison to the Edejer paper. We have added this explanation and added some more context to the study. We also clarified that the $85/DALY result is for all the interventions combined.

11) Line 360: could the authors build this out a little bit more to describe some additional cost drivers that are relevant to VAS programs. Costs are more likely to vary due to a) the delivery mechanism/or platform used in each country to deliver the VAS to children twice/year (routine, or campaign, or what kind of campaign), and b) the strength of the health system. Some countries need to spend a lot of money to deliver VAS to every child because the health system is not already reaching them. Did the authors feel there was sufficient data on costs in the literature? If not, could the authors make a recommendation of some kind – and comment on how this would improve the quality of future analysis such as this.

There are two changes that we make based on this recommendation. First, we clarify that the mechanism that we model in our simulation is a national campaign-based program, meant to model Child Health Days where VAS is distributed as one among many interventions. We have added this in the introduction and methods of the paper. In the discussion, we add more detail regarding the need for up-to-date and accurate cost data in the literature, and how that would improve our study.

Reviewer 2

Review of ‘Cost-effectiveness of Vitamin A Supplementation among Children in Three sub-Saharan African Countries: An Individual-based Simulation Model Using Estimates from the Global Burden of Disease 2017’

Summary Assessment: This is an interesting modeling exercise that has the potential to make useful contributions to the literature. However, substantial improvements are required to the current manuscript.

Thank you for reviewing our paper.

Large Issues:

1. The core results emerge logically from the assumptions and modeling approach – if more children receive VAS, there will be fewer VA-preventable deaths, and if program costs are linear then the cost-effectiveness parameters look promising. However, there are reasons to question the mortality-reducing impacts of VAS in today’s world (estimates are old and new ones cannot be generated for ethical reasons), and especially to question the linearity of costs of expanding coverage. There are reasons why coverage is so low in Nigeria, for example, and these suggest increasing costs as coverage increases, and perhaps some non-cost-related thresholds that need to be considered.

There are two parts to this, which we consider separately. Regarding the impact of VAS on mortality, we agree that ethics prohibit updating old estimates directly, and therefore we believe that by combining measured effects of VAS on VAD and the GBD-estimated effects of VAD on measles and diarrheal diseases (and maybe LRI, see response to previous reviewer about change between GBD 2017 and 2019) we have obtained the most reliable estimate of the impact of VAS on child health.

Regarding the cost, we acknowledge that our assumption of linearity is a simplification, and agree with the reviewer that the marginal costs of increasing VAS coverage will be nonlinear. Although we already acknowledged this in the Discussion section at a limitation of our model, we have revised the organization and wording to further emphasize this. We have additionally explained how a different cost estimate would affect our analysis.

2. There is a big difference between inadequate intake of VA and VAD. A thorough analysis of the impact of a VAS campaign should consider both.

We thank the reviewer for pointing out the distinction between inadequate intake of VA and VAD. GBD focuses on VAD, although it includes data on VA intake in that estimate. Usually, strategies that involve fortification are designed to directly increase VA intake, so we did not model it in this manuscript. We have clarified this as a limitation to our model in the discussion.

3. The authors use the suggestions that (e.g.) LSFF might be a better way of delivering VA to children to justify their analyses. However, no modeling is done to capture the cost-effectiveness of alternative VA deliver platforms, so the justification lacks punch. Vosti et al. (2020) and others have done such work.

We thank the reviewer for identifying this important issue. Although we feel that modelling other Vitamin A delivery mechanisms is important, we believe that it may be beyond the scope of this paper. We hope to address this in future work—our simulation is designed to be modular and allow an apples-to-apples comparison of LSFF and VAS approach (or even combinations!).

4. It is not exactly clear what the authors mean when they say VAS; what are the VAS distribution details and associated cost functions? There are many ways to manage VAS programs, and the impacts and costs vary dramatically depending on the approach chosen.

We appreciate the reviewer’s help to clarify the description of our simulated distribution program and cost methods. We consider this in two parts. First, we have added in the introduction and methods of our paper that our simulation is meant to model the Child Health Days distribution program, where VAS is distributed as one among many interventions.

Additionally, our paper describes the way we use results from our simulation to calculate the overall cost of the intervention under the “Intervention Unit Cost” subsection of the methods. It illustrates what the simulation outputs (Supplemented Days), what the outputs mean, and how they are processed to determine intervention cost.

5. Most countries in sub-Saharan Africa have found campaign-based VAS to be very expensive – indeed, if the international community were not covering the costs of these campaigns (including all inputs and many of the management costs), they would not be undertaken. So, what the authors suggests as unit costs seem not to jibe with reality; indeed, the country-level estimates of program expansion costs (even after adjusting for target population size) are an order of magnitude or more smaller than estimates offered by others. See Schott et al. 2021 for an example of the structure of costs for MNP distribution – not the same intervention, but the cost structure will be similar for VAS campaigns.

We thank the reviewer for their suggestion to check the unit cost for our model to be consistent with other estimates. Because the unit cost directly affects our ICER calculations, it is important to use a reliable value. Based on the literature for VAS costs, we believe that our unit cost is appropriate. Below we include a table describing estimates for VAS unit costs for campaign-based programs in sub-Saharan African countries.

Paper Cost Notes

Neidecker-Gonzales et al. 2007 0.51 USD (Ghana)

0.61 USD (Zambia) Other countries are also considered in this literature review. Unit costs vary tremendously from country to country, so we decided only to use values from sub-Saharan African countries.

Kagin et al. 2015 0.51 USD (Cameroon) The costs analysis in this study found that the wages, training, and communication took up most of the cost. The cost per capsule was 0.03 USD.

Horton et al. 2018 0.62 USD (Senegal) This paper reported the cost as 728.5 FCFA per child. Here, we used the conversion provided in the paper of 1 USD = 584 FCFA and divide by 2 for a cost per dose.

This paper also finds the largest share of the cost comes from health workers’ wages.

Although we do not have data specifically for our regions of interest, we believe these estimates are appropriate. There are examples for cost varying from country to country based on budget and wages for health workers. Neidecker-Gonzales et al. 2007 hypothesize an interesting relationship between unit cost and GDP in their paper:

Because the countries in our analysis are towards the left end of the graph, we believe that this reinforces our cost estimate.

We also found that VAS tends to be much cheaper than MNP. MNP consists of 15 micronutrients including iron, zinc, folic acid, Vitamins A, C, D, B1, B6, etc., so the cost of the supplements would be larger. According to Schott et al. 2021, the cost of a sachet is 0.25-0.59 USD, and the cost of a packet is 7.40-17.83 USD per child. In addition, Kagin et al. 2015 compares the costs of VAS and MNP; they find that while it costs 0.06 USD for two VA capsules themselves, the cost of one year of MNP packets is 3.60 USD.

There are other reasons why estimates for MNP may be higher than VAS. We considered VAS in the context of Child Health Days (CHD), which involve the distribution of many different interventions in addition to VAS—this perhaps reduces the average wage and management cost reported for VAS. Another possibility may be that Schott et al. 2021 considers opportunity costs (only one of the three papers for VAS do), which is estimated to be 17-20% of the cost they report.

In going back and looking for more estimates on unit cost data, we found that the WHO One Health Tool (OHT) provided an estimate of 2.45-3.17 USD. As we believed that the assumptions of this estimate differ substantially from our scenario for CHD, we did not change our unit cost estimate, but we added some reflection on it in the limitations section.

Getting the unit cost right is very important for this analysis as it directly affects the ICER results of the simulation. Although we believe that our current cost estimate agrees with the values in the evidence base at this time, we appreciate the reviewer’s inquiry about our unit cost, and we have added more detail in our discussion to explain the limitations of our cost model. We have highlighted the variability in the unit cost from the WHO OHT as well as different subnational regions in the country. We also added an explanation of how changing the unit cost would proportionally change the ICER estimate.

6. The authors note the importance of addressing uncertainty in their modeling efforts, but uncertainty regarding the impact of VAS on VAD-attributable mortality and (especially) regarding scale-up costs seem not to have received such attention.

We address this issue in two parts. For the impact of VAS on the pipeline, we model the effect of VAS on VAD, the effect of VAD on causes of disease, and the effect of causes on DALYs. The effect of VAS on VAD is determined by a meta-analysis that includes uncertainty. The final two parts of the pipeline are taken from GBD estimates (which include uncertainty estimates).

For the scale-up costs, we did not include uncertainty as part of the scale-up schedule because we modelled the interventions as step functions. To keep it simple, we assumed certainty in the scale-up schedule. For the uncertainty of the unit cost estimate, the paper that provided this cost estimate did not include uncertainty, so we did not add an uncertainty interval either.

For the outcomes of the simulation, we have added an explanation regarding the uncertainty calculation for the outcomes (including DALYs averted and scale-up cost) in our model under “Analysis of Simulation Results” in the methods section.

7. The authors contend that VAS is a highly cost-effective intervention; new literature and ongoing experience in the field suggest that this may not be the case. This is an especially contentious claim based on linear costs of scaling up VAS programs.

We thank the reviewer for pointing out this issue. After changing our modelling to be in line with GBD 2019 estimates (please see Reviewer 1’s first suggestion), our simulation outcomes suggest that the intervention is actually less cost effective than what was reported in previous studies. We have modified our discussion and conclusion to analyze these new results.

Smaller Issues:

1. It is not clear why these three countries were chosen.

We thank the reviewer for pointing out this issue. We chose these three countries because they had different coverage levels for VAS, so the outcomes of their scale-ups would serve as an interesting point of comparison across countries. We have now stated this purpose in the introduction to make the rationale behind the choice of countries clearer.

2. The authors admit that there are no YLD parameters in this simulation exercise, so best to exclude them. Indeed, I would suggest presenting results in terms of deaths averted (rather than DALYs), since no comparisons are being made with other interventions with different mortality/disability structure.

We appreciate the reviewer’s suggestion to present the ICER value to be in cost per death averted. Although we do not incorporate YLDs for xerophthalmia, we include YLDs for diarrheal diseases (as well as LRI in the GBD 2017 model). These YLDs are included in our DALYs estimate. Although we agree that DALYs are often unclear, we prefer to report ICER in dollars per DALY in our analysis as it is the conventional unit used in the literature.

3. It is not clear what ‘individual-based’ means, nor it is clear what calibrations were undertaken to ‘match’ (how well?) GBD estimates. What are SIR and SIS models? What does fixed-effect metanalysis mean, in the several places in which the term is used? Generally, the paper is not accessible to those outside the IHME and their sphere of influence; this can be fixed, but some additional effort will be required.

We thank the reviewer for identifying that clarifying terminology would make this paper accessible to a wider audience. We have added explanations for these terms in our revisions.

Submitted filename: Response to Reviewers.docx

Click here for additional data file.

11 Mar 2022

20 Mar 2022

Response to reviewers for Vitamin A Supplementation

Reviewer 1

One outstanding question for me: The VAD prevalence data indicated in Table 1 seems off by quite a bit - for instance WHO estimates for VAD among PRESAC children in Nigeria is in the range of 15-25% -- whereas Table 1 states it as 1.2%. What is the data source for the VAD prevalence used in the analysis? GBD 2019 also indicates the analysis no longer corrects for background VAD - whereas the explanation of the difference in results is partially due to the differences in VAD prevalence. Please review before proceeding.

We thank the reviewer for pointing out this issue in Table 1. We reviewed this issue and found that indeed the prevalence for VAD in Table 1 deviated from previous estimates.

These values were generated from the GHDx GBD Results tool. We found that the values for VAD prevalence were not updated properly in GHDx for GBD 2019 due to a last-minute resubmission. We updated Table 1 to accurately reflect the inputs for VAD prevalence from GBD 2019. We also added prevalence values for Measles and Diarrheal Diseases that were inputs in the simulation.

Fortunately, the values for VAD risk exposure that we used for the simulation were updated already to include the correct GBD estimates, so this change does not affect the outputs of the simulation.

Reviewer 2

The updated manuscript is much improved; indeed, it is much clearer and the conclusions have been completely reversed! There are still some additional work that could be done (e.g., additional sensitivity analysis regarding some key model parameters and assumptions), but to my mind, the authors have met the criteria for publication. Congrats to all!

Thank you very much for this positive assessment of our work.

Submitted filename: Response to Reviewers.docx

Click here for additional data file.

22 Mar 2022

Cost-effectiveness of Vitamin A Supplementation among children in three sub-Saharan African countries: an individual-based simulation model using estimates from Global Burden of Disease 2019

PONE-D-21-34978R2

Dear Dr. Kannan,

We’re pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it meets all outstanding technical requirements.

Within one week, you’ll receive an e-mail detailing the required amendments. When these have been addressed, you’ll receive a formal acceptance letter and your manuscript will be scheduled for publication.

An invoice for payment will follow shortly after the formal acceptance. To ensure an efficient process, please log into Editorial Manager at http://www.editorialmanager.com/pone/, click the 'Update My Information' link at the top of the page, and double check that your user information is up-to-date. If you have any billing related questions, please contact our Author Billing department directly at authorbilling@plos.org.

If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they’ll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

Kind regards,

Susan Horton

Academic Editor

PLOS ONE

Additional Editor Comments (optional):

Reviewers' comments:

29 Mar 2022

PONE-D-21-34978R2

Cost-effectiveness of Vitamin A Supplementation among children in three sub-Saharan African countries: an individual-based simulation model using estimates from Global Burden of Disease 2019

Dear Dr. Kannan:

I'm pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department.

If your institution or institutions have a press office, please let them know about your upcoming paper now to help maximize its impact. If they'll be preparing press materials, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information please contact onepress@plos.org.

If we can help with anything else, please email us at plosone@plos.org.

Thank you for submitting your work to PLOS ONE and supporting open access.

Kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Dr. Susan Horton

Academic Editor

PLOS ONE

Table 1

GBD 2019 estimates for risk factors and causes associated with VAD by country for ages 0–5.

	Prevalence (%)			PAF of VAD with respect to Cause		DALYs by Cause (per 100K Person-Years
Nation	VAD	Measles	Diarrheal Diseases	Measles	Diarrheal Diseases	VAD	Measles	Diarrheal Diseases
Nigeria	9 (5, 13)	0.13 (0.05, 0.31)	3.3 (2.7, 3.9)	4.8 (0.2, 10.8)	1.6 (0.2, 3.3)	69 (43, 107)	75 (1, 150)	547 (83, 1,163)
Kenya	48 (34, 61)	0.05 (0.02, 0.12)	3.0 (2.5, 3.6)	17.2 (0.9, 32.6)	5.2 (1.0, 9.7)	170 (104, 257)	86 (1, 253)	757 (104, 1190)
Burkina Faso	35 (25, 47)	0.04 (0.01, 0.09)	4.1 (3.3, 4.9)	11.9 (0.5, 24.4)	4.3 (0.7, 8.5)	269 (160, 402)	525 (9, 1581)	877 (137, 1988)

Abbreviations: GBD = Global Burden of Disease study; PAF = Population Attributable Fraction; VAD = Vitamin A Deficiency; DALY = Disability-Adjusted Life Year

Table 2

Overall outcomes for simulated intervention by country.

Country	Overall DALYs Averted Per 100,000 Person-Years	Overall Additional Cost Per 100,000 Person-Years (USD / 100,000 Person-Years)	Cumulative ICER (USD/DALY)
Nigeria	21 (5, 56)	17,880 (17,800, 17,960)	860 (320, 3530)
Kenya	21 (5, 47)	11,390 (11,350, 11,430)	550 (240, 2230)
Burkina Faso	14 (0, 37)	3,100 (3,060, 3,120)	220 (80, 2470)

Abbreviations: DALY = Disability-Adjusted Life Year; USD = US Dollar; ICER = Incremental Cost-effectiveness Ratio