The empirical support for the radical cure strategy for eliminating Plasmodium vivax in China

Abstract

Background: With the recent certification by World Health Organization that the People's Republic of China is malaria-free, it is timely to consider how elimination of malaria was completed in People's Republic of China over the last 7 decades. Of the four widespread species of human malaria, Plasmodium vivax was the last to be eliminated by the national program of China. Understanding the incubation periods and relapses patterns of P. vivax through historical data from China is relevant for planning disease elimination in other malaria-endemic countries, with residual P. vivax malaria.

Methods: We collated data from both published and unpublished malaria parasite inoculation experiments conducted between 1979 and 1988 with parasites from different regions of the People's Republic of China. The studies had at least two years of follow-up. We categorized P. vivax incubation patterns via cluster analysis and investigated relapse studies by adapting a published within-host relapse model for P. vivax temperate phenotypes. Each model was fitted using the expectation-maximization (EM) algorithm initialized by hierarchical model-based agglomerative clustering.

Results: P. vivax parasites from the seven studies of five southern and central provinces in the People's Republic of China covering geographies ranging from the south temperate to north tropical zones. The parasites belonged to two distinct phenotypes: short- (10-19 days) or long-incubation (228-371 days). The larger the sporozoite inoculation, the more likely short incubation periods were observed, and with more subsequent relapses (Spearman's rank correlation between the number of inoculated sporozoites and the number of relapses of 0.51, p-value = 0.0043). The median of the posterior distribution for the duration of the first relapse interval after primary infection was 168.5 days (2.5% quantile: 89.7; 97.5% quantile: 227.69 days). The predicted survival proportions from the within-host model fit well to the original relapse data. The within-host model also captures the hypnozoite activation rates and relapse frequencies, which consequently influences the transmission possibility of P. vivax.

Conclusions: Through a within-host model, we demonstrate the importance of clearance of hypnozoites. A strategy of two rounds of radical hypnozoite clearance via mass drug administration (MDA) deployed during transmission (summer and autumn) and non-transmission (late spring) seasons had a pronounced effect on outbreaks during the malaria epidemics in China. This understanding can inform malaria control strategies in other endemic countries with similar settings.

Keywords: Elimination; Model; People’s Republic of China; Plasmodium vivax; Strategy.

Fig. 1

Locations of seven inoculation studies in five provinces in China

Fig. 2

Minimum and maximum short/long incubation periods in seven inoculation studies. Guangxi (GX), Guizhou (GZ), Henan Kaifeng (HN_KF), Henan Zhengzhou (HN_ZZ), Hunan, Yunnan Mengla (YN_ML), and Yunnna_Weixi (YN_WX). There were no long incubation cases for study in Henan Zhengzhou (HN_ZZ)

Fig. 3

Cluster analysis of incubation period of inoculation studies. Blue dots indicate the first cluster with a mean incubation period of 12.4 days, while red squares denote the second cluster with a mean incubation period approaching 294.2 days

Fig. 4

Spearman’s rank correlation between the number of inoculated sporozoites and relapses. The x-axis denotes number of sporozoites inoculated. The y-axis denotes the number of relapses. Shaded areas represent 95% confidence interval

Fig. 5

Posterior distribution of parameters of within-host model. The x-axis denotes each parameter’s value. The y-axis denotes the density of posterior distribution. The orange lines denote the prior distributions, based on the posterior distributions reported by White and colleagues [32] for the duration of long-latency (d_LL), a uniform prior distribution is assumed. The black lines denote the posterior distributions

Fig. 6

Within-host model prediction of the probability of survival for each relapse. The plot shows a comparison between model predicted times to relapse and the datasets described in Table 2. The curves represent the best-fit prediction for the within-host model. The five numbered curves represent five relapses. The dots indicate the survival probability of each relapse (based on raw data) with a 95% confidence interval (vertical bars). The survival analysis outcomes including 95% confidence intervals for the survival function at the time of each relapse are shown in the Additional file 1: Fig. S2