Seasonality and heterogeneity of malaria transmission determine success of interventions in high-endemic settings: a modeling study

Abstract

Background: Malaria transmission is both seasonal and heterogeneous, and mathematical models that seek to predict the effects of possible intervention strategies should accurately capture realistic seasonality of vector abundance, seasonal dynamics of within-host effects, and heterogeneity of exposure, which may also vary seasonally.

Methods: Prevalence, incidence, asexual parasite and gametocyte densities, and infectiousness measurements from eight study sites in sub-Saharan Africa were used to calibrate an individual-based model with innate and adaptive immunity. Data from the Garki Project was used to fit exposure rates and parasite densities with month-resolution. A model capturing Garki seasonality and seasonal heterogeneity of exposure was used as a framework for characterizing the infectious reservoir of malaria, testing optimal timing of indoor residual spraying, and comparing four possible mass drug campaign implementations for malaria control.

Results: Seasonality as observed in Garki sites is neither sinusoidal nor box-like, and substantial heterogeneity in exposure arises from dry-season biting. Individuals with dry-season exposure likely account for the bulk of the infectious reservoir during the dry season even when they are a minority in the overall population. Spray campaigns offer the most benefit in prevalence reduction when implemented just prior to peak vector abundance, which may occur as late as a couple months into the wet season, and targeting spraying to homes of individuals with dry-season exposure can be particularly effective. Expanding seasonal malaria chemoprevention programs to cover older children is predicted to increase the number of cases averted per treatment and is therefore recommended for settings of seasonal and intense transmission.

Conclusions: Accounting for heterogeneity and seasonality in malaria transmission is critical for understanding transmission dynamics and predicting optimal timing and targeting of control and elimination interventions.

Keywords: Heterogeneity Mathematical modeling; Malaria; Seasonality.

Fig. 1

Inferring EIR from infant and child conversion rates. a Reported EIR from Matsari site (purple), EIR inferred from fitting to infant asexual parasite densities (green), and EIR of a high-biting population inferred from fitting to child asexual parasite densities (yellow). b Infant parasite densities observed (red) and simulated (blue) under reported EIR. c Infant parasite densities observed (red) and simulated (blue) after adjusting dry-season EIR. d Parasite densities in children between 4 and 8 years of age observed (red) and simulated (blue) with inferred EIR from adjusting to infant parasite densities. e Parasite densities in children between 4 and 8 years of age observed (red) and simulated (blue) after including heterogeneous dry-season biting and calibration of immune model

Fig. 2

Asexual parasite density by month and age group in the Matsari site after calibrating immune parameters. Red: reference data from the Garki Project; blue: calibrated simulation

Fig. 3

The infectious reservoir of malaria in a seasonal setting stratified by age, detectability of infection status, and susceptibility to dry-season biting. Results presented are means of 50 stochastic realizations. Annual EIR in panels A and B is 130. a Total human infectiousness varies through the year in a seasonal setting. b Contribution to the human infectious reservoir (normalized total human infectiousness) by age group, detectability, and dry-season biting risk varies through the year. c Annual average total human infectiousness under the same seasonality increases with EIR. d Contribution to the annual average human infectious reservoir by age group, detectability, and dry-season biting risk varies with EIR

Fig. 4

Optimal deployment of IRS. a Effectiveness of IRS depends on both IRS half-life and timing of spray campaigns relative to the peak biting season. Timing has the strongest impact on moderately long-lasting IRS. Top: daily count of adult vectors and EIR. Bottom: fractional reduction in prevalence one year after an IRS campaign with 50% coverage in a population with homogeneous transmission. Mean and standard deviation of 50 stochastic realizations are shown for each half-life and campaign start date. b The matrix of shaded boxes represents prevalence reduction in a population experiencing heterogenous dry-season biting with the coverage in population experiencing dry-season biting along the vertical axis, and the fraction of the same population in the entire population along the horizontal axis. Each box within the matrix is shaded to reflect the mean number of clinical cases averted during the year after the campaign is started for 50 stochastic realizations. The optimal start month for each campaign given the coverage and fraction of the group experiencing dry-season biting is indicated in the box. Regions shaded with gray oblique lines indicate coverage in the total population was either above or below the target 50% overall coverage for the corresponding coverage and fraction of the group with dry season biting

Fig. 5

Impact of SMC, MDA, and MTAT campaigns depends on campaign timing and coverage. a Top: daily count of adult vectors and EIR. Bottom: clinical cases averted during the year after the first round of drugs distributed as compared to a baseline scenario with no drug campaigns. SMC, MDA and MTAT campaigns were simulated with 50% coverage. See “Methods” section for details on how drug campaigns were configured. Mean and standard deviation of 50 stochastic realizations are shown for each campaign start date. Arrows indicate the campaign start date that results in the maximum number of clinical cases averted for each campaign type. b Cases averted, number of tests and treatments distributed, and cases averted per test or treatment when each drug campaign type is distributed at optimum timing, at 50% coverage. c Cases averted by age group (under 5 years, 5 to 10 years, over 10 years) by campaign type and coverage achieved when each drug campaign type is distributed at optimum timing

Fig. 6

Targeting SMC, MTAT or MDA campaigns to the subpopulation experiencing dry-season biting does not increase the number of clinical cases averted in the entire population. Each campaign was started at the optimal time shown in Fig. 5a. Coverage in the group experiencing dry-season biting was varied while adjusting coverage in the remaining population such that overall coverage in each campaign round was 50%. Solid and dotted lines indicate mean and standard deviation of 50 stochastic realizations respectively