Mathematical modelling of the impact of expanding levels of malaria control interventions on Plasmodium vivax

Abstract

Plasmodium vivax poses unique challenges for malaria control and elimination, notably the potential for relapses to maintain transmission in the face of drug-based treatment and vector control strategies. We developed an individual-based mathematical model of P. vivax transmission calibrated to epidemiological data from Papua New Guinea (PNG). In many settings in PNG, increasing bed net coverage is predicted to reduce transmission to less than 0.1% prevalence by light microscopy, however there is substantial risk of rebounds in transmission if interventions are removed prematurely. In several high transmission settings, model simulations predict that combinations of existing interventions are not sufficient to interrupt P. vivax transmission. This analysis highlights the potential options for the future of P. vivax control: maintaining existing public health gains by keeping transmission suppressed through indefinite distribution of interventions; or continued development of strategies based on existing and new interventions to push for further reduction and towards elimination.

Fig. 1

Calibration of model to cross-sectional and longitudinal surveys from PNG and the Solomon Islands. For the cross-sectional surveys, the data are presented as age-stratified estimates of prevalence with 95% confidence intervals. The same definition of clinical malaria was used in all cross-sectional surveys: high density parasitaemia and fever (≥38 °C) in the last 48 h. The number of individuals in each cross-section is denoted N. For the longitudinal surveys in the bottom row, the data are presented as Kaplan–Meier estimates of proportion infected with 95% confidence intervals. The number of individuals included in longitudinal follow-up is denoted N and the total number of samples denoted s. The solid curves show the posterior median model prediction, and the shaded regions denote the 95% credible intervals

Fig. 2

Validation of model to data from systematic reviews. Data are shown as points and model predictions as solid lines. a Relationship between PvPR_LM and clinical incidence based on data reviewed by Battle et al.. Studies included in this review came from throughout the world, had various age ranges, and differences in the frequency of active case detection (ACD) for clinical cases of P. vivax. b Relationship between EIR and PvPR_LM compared to data reviewed by Battle et al. plus data from a study in Papua New Guinea by Burkot et al.. Orange points denote studies where infectious mosquitoes were confirmed as P. vivax positive, and green points denote studies where mosquitoes were identified as infectious but without Plasmodium species identification. In some studies, a range was provided instead of a point estimate. c Relationship between PvPR_LM and PvPR_PCR compared to data reviewed by Moreira et al.. Data are from throughout the world and are based on studies with various age ranges. d Model-predicted relationship between PvPR_LM and hypnozoite prevalence. As it is currently not possible to directly detect hypnozoites, this relationship cannot be formally compared to any data

Fig. 3

Predicted PvPR_LM in Papua New Guinean provinces using individual-based model. Model predictions are based on the median of 100 stochastic simulations. Data are from household prevalence surveys in randomly selected villages, and surveys from a number of sentinel villages either before or after LLIN distribution. The black curves denote the model-predicted scenario if LLINs are not replaced. In the LLIN campaigns, nets are assumed to be distributed every 3 years, with 50% of nets still in use after 19.5 months. Primaquine (PQ) or tafenoquine (TQ) with accompanying G6PD screening are assumed to be include in first-line treatment regimens from 2020, with 50% of individuals experiencing a clinical episode of P. vivax being tested and treated

Fig. 4

Predicted impact of combinations of interventions in Papua New Guinea. Model predictions are based on the median of 100 stochastic simulations. a, b Estimated PvPR_LM in Papua New Guinea provinces based on household surveys from 2010 and 2014. c Model-predicted prevalence in 2025 under a scenario where LLINs are distributed every 3 years at 80% coverage levels. d Combinations of interventions required to obtain pre-elimination (defined as prevalence < 0.1%) by 2025. Grey shading indicates that the interventions considered were predicted not to be sufficient to reduce prevalence to <0.1%

Fig. 5

Individual-based model projections of P. vivax transmission in New Ireland under a range of intervention scenarios. Model predictions are based on the median of 100 stochastic simulations. a LLIN campaign every 3 years at 80% coverage. b LLIN campaigns plus introduction of tafenoquine (TQ) into first-line treatment regimen from 2020 such that 50% of clinical cases are treated. c LLIN campaigns plus mass drug administration (MDA) with tafenoquine at 80% coverage in 2020. d LLIN campaigns with TQ incorporated in first-line treatment regimen and TQ MDA. e Estimated clinical cases per 1000 person years for the period 2020–2022 under the intervention scenarios considered. f Estimated number of TQ doses per person over the period 2020–2022

Fig. 6

Compartmental representation of P. vivax transmission model in humans. Infected individuals can be in one of three compartments depending on whether blood-stage parasitaemia is detectable by PCR (I_PCR), light microscopy (I_LM) or has high density with accompanying fever (I_D). A proportion of individuals that progress to a symptomatic episode of P. vivax will undergo treatment with a blood-stage drug (T) leading to clearance of blood-stage parasitaemia and a period of prophylactic protection (P) before returning to the susceptible state (S). The superscript k denotes the number of batches of relapse causing hypnozoites in the liver. Red arrows denote new blood-stage infections arising from either new mosquito bites or relapses. Each square denotes a compartment and the circles denote the dependence of transition rates between compartments on levels of anti-parasite immunity (A_p) and levels of clinical immunity (A_c)