False-negative malaria rapid diagnostic test results and their impact on community-based malaria surveys in sub-Saharan Africa

Abstract

Surveillance and diagnosis of Plasmodium falciparum malaria relies predominantly on rapid diagnostic tests (RDT). However, false-negative (FN) RDT results are known to occur for a variety of reasons, including operator error, poor storage conditions, pfhrp2/3 gene deletions, poor performance of specific RDT brands and lots, and low-parasite density infections. We used RDT and microscopy results from 85 000 children enrolled in Demographic Health Surveys and Malaria Indicator Surveys from 2009 to 2015 across 19 countries to explore the distribution of and risk factors for FN-RDTs in sub-Saharan Africa, where malaria's impact is greatest. We sought to (1) identify spatial and demographic patterns of FN-RDT results, defined as a negative RDT but positive gold standard microscopy test, and (2) estimate the percentage of infections missed within community-based malaria surveys due to FN-RDT results. Across all studies, 19.9% (95% CI 19.0% to 20.9%) of microscopy-positive subjects were negative by RDT. The distribution of FN-RDT results was spatially heterogeneous. The variance in FN-RDT results was best explained by the prevalence of malaria, with an increase in FN-RDT results observed at lower transmission intensities, among younger subjects, and in urban areas. The observed proportion of FN-RDT results was not predicted by differences in RDT brand or lot performance alone. These findings characterise how the probability of detection by RDTs varies in different transmission settings and emphasise the need for careful interpretation of prevalence estimates based on surveys employing RDTs alone. Further studies are needed to characterise the cost-effectiveness of improved malaria diagnostics (eg, PCR or highly sensitive RDTs) in community-based surveys, especially in regions of low transmission intensity or high urbanicity.

Keywords: Plasmodium falciparum; RDTs; malaria diagnosis; mapping; mathematical modelling; pfhrp2; pfhrp2 deletion; rapid diagnostic tests.

Figure 1

Observed percentage of false negative rapid diagnostic test (FN-RDT) results from DHS and MIS field surveys and expected results based on the WHOFoundation for Innovative New Diagnostics (FIND) lot testing. Red bars depict the weighted percentage of microscopy-positive infections that yielded a negative RDT result in the DHS/MIS surveys, and blue bars depict the expected percentage of microscopy-positive, RDT-negative infections based on a conservative estimate derived from the WHO Product Testing Panel Detection Score (PDS) at parasite densities of 200 parasites/μL. DHS, Demographic Health Surveys; MIS, Malaria Indicator Surveys.

Figure 2

Mean adjusted percentage of false negative rapid diagnostic test (FN-RDT) results at the first administrative region. The percentage of FN-RDT results presented represents the data from the most recent Demographic Health Surveys (DHS) or Malaria Indicator Surveys (MIS) for each country for which both RDT and microscopy results were available, after adjusting for the expected percentage of FN-RDT results based on WHO product testing results.

Figure 3

Distribution and impact of rapid diagnostic test (RDT) brand upon the observed percentage of false-negative (FN) RDT results. In (A) the RDT brand used within the Demographic Health Survey (DHS) years of interest is shown. In (B) the weighted percentage of FN-RDT results at the first administrative region is shown for each RDT brand. The sample size for each region is indicated by the point size, and the mean and 95% CI for each brand is shown in red, with the expected result based on WHO product testing shown in blue.

Figure 4

Hierarchical analysis of covariates associated with an increased risk of a false-negative rapid diagnostic test (FN-RDT) result. The log ORs for each covariate are shown with their 95% CIs as whiskers surrounding each point for (A) cluster-level and individual-level covariates and (B) RDT brand. Log ORs significantly not equal to 1 (probability of the posteriors including zero (pMCMC) <0.05) are shown in blue and were observed for the type of residence (urban vs rural), the age of the individual, the prevalence of malaria within a cluster and the interaction between malaria prevalence and residence type. The reference for the log ORs associated with brand is CareStart Malaria HRP2/pLDH Combo.

Figure 5

Observed relationship between proportion of false-negative rapid diagnostic test (FN-RDT) results and malaria prevalence. The weighted proportion of FN-RDT results at the first administrative region is shown on the y-axis and the weighted malaria prevalence by microscopy on the x-axis. The sample size for each region is indicated by the point size, and the generalised linear model (GLM) relationships with a binomial error structure are displayed as red curves. The relationship is further stratified by residence type, with the relationship within rural and urban areas shown with a solid and dashed line, respectively. The blue curve depicts the expected proportion of FN-RDT results based on WHO product testing results.