Strengthening long-lasting insecticidal nets effectiveness monitoring using retrospective analysis of cross-sectional, population-based surveys across sub-Saharan Africa

Bed nets averted 68% of malaria cases in Africa between 2000 and 2015. However, concerns over insecticide resistance, bed net durability and the effectiveness of long-lasting insecticidal nets (LLIN) are growing. To assess the effectiveness of LLINs of different ages and insecticides against malaria, we conducted a population-based, cross-sectional study using data from 162,963 children younger than 5 years of age participating in 33 Demographic and Health and Malaria Indicator Surveys conducted in 21 countries between 2009 and 2016. We used Bayesian logistic regression to estimate associations between LLIN age, insecticide type, and malaria. Children sleeping under LLINs the previous night experienced 21% lower odds of malaria infection than children who did not (odds ratio [OR] 0.79; 95% Uncertainty Interval [UI] 0.76–0.82). Nets less than one year of age exhibited the strongest protective effect (OR 0.75; 95% UI 0.72–0.79), and protection weakened as net age increased. LLINs containing different insecticides exhibited similar protection (ORdeltamethrin 0.78 [0.75–0.82]; ORpermethrin 0.79 [0.75–0.83]; ORalphacypermethrin 0.85 [0.76–0.94]). Freely-available, population-based surveys can enhance and guide current entomological monitoring amid concerns of insecticide resistance and bed net durability, and be used with locally-collected data to support decisions on LLIN redistribution campaign timing which insecticide to use.


Introduction
Background/rationale 2 Explain the scientific background and rationale for the investigation being reported Scientific Background: We provide a comprehensive summary of previous work on the relationship between insecticide resistance and malaria, including a number of trials. We note that these studies tend to be conducted in a small number of settings, limiting generalizability. Given that bed net decisions are often made at the country level, we argue that more work at the population level is needed.
Rationale: We write in the introduction that insecticide resistance is expanding across Africa, and that there is growing concern regarding the durability of bed nets in field conditions.

Objectives 3 State specific objectives, including any prespecified hypotheses
We state that our objective is to "examine the relationship between agriculture, the mosquito population, and malaria risk using data from a population-based cross-sectional survey of children under 5 years of age living in the Democratic Republic of Congo…and contemporaneous entomological monitoring data collected over time across DRC's ecological zones." We state that our hypothesis is that increasing exposure to agriculture is associated with increased malaria risk, and seek to understand how changes in vector behaviour may be a mechanism underlying this hypothesized increase.

Study design 4
Present key elements of study design early in the paper We state both in the title and in the introduction that this is a cross-sectional study. We further describe the study populations (children under 5 years of age) in detail, including sample sizes and selection criteria in the methods section.
Setting 5 Describe the setting, locations, and relevant dates, including periods of recruitment, exposure, follow-up, and data collection We describe in detail the country settings, the year each survey was conducted, and include a table (Table 1) with summary measures for the outcome and exposure measures of interest for each survey.
Participants 6 (a) Give the eligibility criteria, and the sources and methods of selection of participants We describe the eligibility criteria in detail in the methods section. Briefly, they are children under 5 years of age living who were tested for malaria by RDT, either slept under an LLIN or did not sleep under any net, and had no missing covariate information. Only 126 individuals had missing data.

Variables 7
Clearly define all outcomes, exposures, predictors, potential confounders, and effect modifiers. Give diagnostic criteria, if applicable We define outcomes and exposures in the description of the study population, and dedicate a separate section to confounding variables and how they were measured.
Data sources/ measurement 8* For each variable of interest, give sources of data and details of methods of assessment (measurement). Describe comparability of assessment methods if there is more than one group We describe the source of each variable (outcome, exposure, or confounder).

Bias 9
Describe any efforts to address potential sources of bias We address sources of bias in previous work, how our work helps to address such biases, and further discuss possible bias in our work. For example, we note that our measures of the exposures are subject to misclassification bias and/or reporting bias.

Study size 10
Explain how the study size was arrived at We provide a description of the selection criteria and provide a study flow diagram as Supplementary Figure 1 in our study.

Quantitative variables 11
Explain how quantitative variables were handled in the analyses. If applicable, describe which groupings were chosen and why We provide a detailed description of how quantitative variables were handled (e.g. centering and scaling). We also describe that groupings were chosen based on how the data were collected (e.g. with regard to the housing characteristics variables).
Statistical methods 12 (a) Describe all statistical methods, including those used to control for confounding We describe our statistical methods used in the main text of the paper, together with a discussion of weakly informative prior distributions and how we used them to have the model yield parameter estimates within an epidemiologically relevant range.

(b) Describe any methods used to examine subgroups and interactions
We stratified by survey and considered the effects of nets by age and by insecticide separately, as well as their interaction (c) Explain how missing data were addressed We state that this is a complete case analysis, as there were only 126 study subjects out of 169 013 with any missing data.
(d) If applicable, describe analytical methods taking account of sampling strategy In our description of the methodological approach, we note that we specify a multilevel model to account for the sampling strategy of the survey. Our model has the following general form:

= +
where is the observed malaria outcome for child in survey cluster , is a × row vector of covariates for child in cluster , is a × vector of regression coefficients linking the covariates to the response (through a logit link), while represents a unique cluster-level random effect.

(e) Describe any sensitivity analyses
We implement 4 different multilevel Bayesian models to investigate whether or not different bed net exposures yield better fits to the data. We further describe how we assess model fit and provide fit statistics in Supplementary Table 2.

Results
Participants 13* (a) Report numbers of individuals at each stage of study-eg numbers potentially eligible, examined for eligibility, confirmed eligible, included in the study, completing follow-up, and analysed This information is included in the methods section, as well as in supplementary figure 1.

(b) Give reasons for non-participation at each stage
The mothers for all eligible participants assented to their children being included in the study.

(c) Consider use of a flow diagram
We include a flow diagram in supplementary figure 1.
Descriptive data 14* (a) Give characteristics of study participants (eg demographic, clinical, social) and information on exposures and potential confounders We include this information in Tables 1 and 2 (b) Indicate number of participants with missing data for each variable of interest A total of 126 individuals had missing data.

Outcome data 15* Report numbers of outcome events or summary measures
We begin the results section by summarizing the outcome and exposure measures for the entire sample, and further provide these summary measures by survey, since we also stratify our analysis by survey.
Main results 16 (a) Give unadjusted estimates and, if applicable, confounder-adjusted estimates and their precision (eg, 95% confidence interval). Make clear which confounders were adjusted for and why they were included We do not include a discussion of unadjusted estimates owing to space limitations. Additionally, our literature review indicated that confounding is an important limitation of studies on the insecticide-malaria relationship, and we therefore focus on addressing this confounding by including confounders that are otherwise unavailable in other studies.

(b) Report category boundaries when continuous variables were categorized
We did not categorize continuous variables.
(c) If relevant, consider translating estimates of relative risk into absolute risk for a meaningful time period Not applicable. We report Odds Ratios.
Other analyses 17 Report other analyses done-eg analyses of subgroups and interactions, and sensitivity analyses We report results for the 3 models that yielded similar fit. We present the results from the 4 th model in the supplementary appendix, and report the main results (i.e. those from the best-fitting model) in the main text.

Discussion
Key results 18 Summarise key results with reference to study objectives We provide a broad summary that our findings suggest that bed nets treated with different insecticides and of different ages are effective across Africa, but that there is variability across countries.

Limitations 19
Discuss limitations of the study, taking into account sources of potential bias or imprecision. Discuss both direction and magnitude of any potential bias We discuss potential bias in our discussion. Specifically, we discuss that misclassification and reporting bias may undermine our results. Further, we note that we cannot draw inferences on the adult population, since adults are not tested for malaria.

Interpretation 20
Give a cautious overall interpretation of results considering objectives, limitations, multiplicity of analyses, results from similar studies, and other relevant evidence We work to ensure a cautious interpretation by using cautious language (i.e. the words "suggest" and "may" and "appear"), e.g.: -"…our analysis based on 2011 survey data suggested that nets treated with deltamethrin exhibited a weak protective effect…" -" The effect of using nets of different ages appears to vary across surveys." -"permethrin-treated nets appear to provide little or no protective benefit (OR 0.89, 95% UI 0.72 -1.10)" "-Thus, the observed lack of effectiveness in a net treated with a given insecticide may not be due to resistance, but to IRS or other pesticide spraying that kills mosquitoes before they have the opportunity to make contact with a net."

Generalisability 21
Discuss the generalisability (external validity) of the study results We note that one of the strengths of this study is that it relies on populationbased surveys of children under 5 years of age, suggesting that the results are generalizable to the population of children under 5 years of age.

Other information
Funding 22 Give the source of funding and the role of the funders for the present study and, if applicable, for the original study on which the present article is based We include the following statements in the manuscript: The sponsors of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication. Parental consent for children's participation in the DHS and MIS surveys was obtained by the DHS Program.

Supplementary Figure 1. Study Flow Diagram
Supplementary