A joint Bayesian framework for modeling Plasmodium vivax transmission
Plasmodium vivax transmission from humans to mosquitoes depends on the density of gametocytes that in turn depends on asexual parasite replication and gametocyte commitment. These processes are often analyzed separately, despite being biologically linked and measured with substantial uncertainty. We used a joint Bayesian latent-variable model to simultaneously analyze parasite density, gametocyte density, and mosquito infectivity while accounting for measurement error and propagating uncertainty across linked processes. The model was applied to individual-level data from three P. vivax transmission studies conducted in Ethiopia (n = 455). A tenfold increase in gametocyte density was associated with more than a twofold increase in the odds of mosquito infection (odds ratio [OR] = 2.32, 95% credible interval [CrI]: 2.12-2.54). Asexual parasite density was also positively associated with infectivity after accounting for gametocyte density (OR = 1.74, 95% CrI: 1.60-1.90), and inclusion of parasite density improved predictive performance. Pathway decomposition within the joint model indicated that approximately 41% of the parasite-infectivity association operated through gametocyte density. Increasing age was associated with lower asexual parasite density but higher gametocyte density, resulting in minimal overall association with infectivity. Predicted infection probability increased sigmoidally with gametocyte density, remaining low at lower densities before increasing sharply and approaching a plateau at higher densities. Gametocyte density produced the largest predicted changes in the proportion of infected mosquitoes, while asexual parasite density added predictive information not fully captured by measured gametocyte density alone. This approach links molecular parasite measurements with mosquito infection risk while accounting for measurement uncertainty and provides an interpretable framework for studying the P. vivax infectious reservoir. Author SummaryMalaria transmission occurs when mosquitoes ingest sexual-stage parasites, called gametocytes, during a blood meal. In Plasmodium vivax infections, human-to-mosquito transmission depends on linked biological stages, including asexual parasite replication, gametocyte production, and mosquito infection. These processes are closely connected and often measured with uncertainty, making them difficult to study using standard approaches that analyze them separately. In this study, we applied a joint Bayesian model that analyzes parasite density, gametocyte density, and mosquito infectivity together while accounting for uncertainty in laboratory measurements. Using data from three studies in Ethiopia, we quantified how parasite density, gametocyte density, and host characteristics relate to mosquito infection. The analysis showed that measured gametocyte density alone did not fully explain variation in infectivity, and that asexual parasite density provided additional predictive information. We also found that age was associated differently with asexual parasite and gametocyte densities, resulting in little overall association with infectivity. This approach helps link molecular parasite measurements with transmission outcomes and improves understanding of the P. vivax infectious reservoir in endemic settings.