Global patterns of submicroscopic Plasmodium falciparum malaria infection: insights from a systematic review and meta-analysis of population surveys

Abstract

Background: Adoption of molecular techniques to detect Plasmodium falciparum infection has revealed many previously undetected (by microscopy) yet transmissible low-density infections. The proportion of these infections is typically highest in low transmission settings, but drivers of submicroscopic infection remain unclear. Here, we updated a previous systematic review of asexual P falciparum prevalence by microscopy PCR in the same population. We aimed to explore potential drivers of submicroscopic infection and to identify the locations where submicroscopic infections are most common.

Methods: In this systematic review and meta-analysis we searched PubMed and Web of Science from Jan 1, 2010, until Oct 11, 2020, for cross-sectional studies reporting data on asexual P falciparum prevalence by both microscopy and PCR. Surveys of pregnant women, surveys in which participants had been chosen based on symptoms or treatment, or surveys that did not involve a population from a defined location were excluded. Both the number of individuals tested and the number of individuals who tested positive by microscopy or PCR, or both, for P falciparum infection were extracted. Bayesian regression modelling was used to explore determinants of the size of the submicroscopic reservoir including geographical location, seasonality, age, methodology, and current or historical patterns of transmission.

Findings: Of 4893 identified studies, we retained 121 after screening and removal of duplicates. 45 studies from a previous systematic review were included giving 166 studies containing 551 cross-sectional survey microscopy and PCR prevalence pairs. Our results show that submicroscopic infections predominate in low-transmission settings across all regions, but also reveal marked geographical variation, with the proportion of infections that are submicroscopic being highest in South American surveys and lowest in west African surveys. Although current transmission levels partly explain these results, we find that historical transmission intensity also represents a crucial determinant of the size of the submicroscopic reservoir, as does the demographic structure of the infected population (with submicroscopic infection more likely to occur in adults than in children) and the PCR or microscopy methodology used. We also observed a small yet significant influence of seasonality, with fewer submicroscopic infections observed in the wet season than the dry season. Integrating these results with estimates of infectivity in relation to parasite density suggests the contribution of submicroscopic infections to transmission across different settings is likely to be highly variable.

Interpretation: Significant variation in the prevalence of submicroscopic infection exists even across settings characterised by similar current levels of transmission. These differences in submicroscopic epidemiology potentially warrant different approaches to targeting this infected subgroup across different settings to eliminate malaria.

Funding: Bill & Melinda Gates Foundation, The Royal Society, and the UK Medical Research Council.

Figure 1

Prevalence of infection by PCR vs microscopy in 267 prevalence survey pairs and model fits Bayesian Markov chain Monte Carlo methods were used to fit a linear relationship between PCR prevalence and microscopy prevalence on the log odds scale. (A) 387 microscopy and PCR prevalence surveys were identified in this study and previous systematic reviews. The fitted model relationship (purple line) and the 95% credible interval of the mean (light purple shaded area). (B) The prevalence ratio (ie, the proportion of PCR positive individuals also detectable by microscopy) according to underlying PCR prevalence for each of the 387 survey microscopy–PCR pairs (points) used to fit the full model. The estimated mean prevalence ratio (purple line) and 95% credible interval of the mean (light purple shaded area) are also shown. (C) Box plot of the prevalence ratio disaggregated by global region. For each region, the size of the point reflects the number of individuals tested by microscopy and PCR. Thick coloured bar on the box plot represents the weighted mean prevalence ratio for each global region. Thin line indicates the median, box indicates IQR, and whisker limits span 1·5× the IQR.

Figure 2

Global variation in the prevalence ratio and the relative size of the submicroscopic reservoir Microscopy and PCR prevalence in included surveys (points), the model-fitted relationship (coloured line) and 95% credible interval (shaded area) for Asia and Oceania (A), west Africa (B), east Africa (C), and South America (D). (E) The model-fitted average microscopy: PCR prevalence ratio by PCR prevalence for each of the four regions (coloured line) and 95% credible interval (shaded area). Coloured countries on each regional map indicates countries for which studies were identified during the systematic review.

Figure 3

The effect of historical and current transmission intensity on the prevalence of submicroscopic malaria infection in Africa (A) Map detailing the African countries and associated administrative unit 1 level regions for which prevalence surveys were identified, as well as their assigned transmission archetypes based on historical and current transmission intensity (high high=historically high and currently high; high low=historically high and currently low; low low=historically low and currently low). (B) The prevalence ratio of surveys in each transmission archetype (points; n=90 for high high, n=99 for high low, and n=40 for low low), and the modelled average prevalence ratio (coloured line) with 95% credible interval (shaded area).

Figure 4

The influence of age on submicroscopic malaria infection (A) Box plot of age disaggregated prevalence survey data for young children (0–5 years, purple points, n=49) older children (6–15 years, pink points, n=62), and adults (>15 years old, blue points, n=53). For each age group, the size of the point reflects the number of individuals tested by microscopy and PCR. Thick coloured bar on the boxplot represents the weighted mean prevalence ratio for each age group. Thin line indicates the median, box indicates IQR, and whisker limits span 1·5× the IQR. (B) The prevalence ratio in surveys where age-disaggregated data (points) were available by age group, showing the fitted model relationship (coloured lines) and the 95% credible interval (shaded areas).

Figure 5

The potential contribution of submicroscopic infections to onwards transmission according to current and historical transmission intensity Potential contribution of the submicroscopic reservoir to onwards transmission for each of the transmission archetypes: historically high and currently high (A), historically high and currently low (B), and historically low and currently low (C) if microscopic infections are either 2×, 5×, or 20× more infectious than submicroscopic infections.