Background
Malaria is most commonly diagnosed in humans by either microscopy or rapid diagnostic tests (RDTs), which detect antigens produced by malaria parasites [
1,
2]. The antigens histidine rich protein 2 (HRP2) and lactate dehydrogenase (LDH) are those most commonly targeted by RDTs. HRP2 antigen is expressed only by
Plasmodium falciparum malaria, and RDTs targeting
P. falciparum infections typically detect HRP2, while LDH is a constitutive enzyme expressed by all
Plasmodium species. RDTs can specifically detect
P. falciparum infections through antibodies targeting subspecies epitopes on the LDH antigen, as well as all
Plasmodium species through conserved epitopes. Of note, HRP2-based RDTs can exhibit cross-reactivity with HRP3 due to antigenic similarity.
Until recently, malaria RDT evaluation programmes have focused on benchmarking malaria diagnostic tests against parasite density measured either by microscopy or nucleic acid tests, although antigen concentrations of standard samples have recently begun to be included [
3‐
5]. RDTs have been considered to perform equivalently to microscopy in terms of sensitivity and specificity [
6], with a recognition that HRP2-based RDTs may provide false positives due to the long half-life of circulating HRP2 [
1,
2]. The advent of new RDTs, such as the ultra-sensitive HRP2-based Alere™ Malaria Ag P.f RDT (Abbott, South Korea) with a tenfold lower limit of detection for HRP2 than previous RDTs [
7,
8] has incentivized the research community to better understand antigen dynamics in infected populations [
8‐
12]. Concurrently, new assays for antigen quantification have been developed both as research tools on the Luminex platform [
13,
14] and as the commercially available Q-Plex™ Human Malaria Array (Quansys Biosciences, USA) [
9].
This report describes the performance of the ultra-sensitive Alere Malaria Ag P.f RDT (uRDT) on samples collected in a phase 1 clinical trial malaria vaccine study in Bancoumana, Mali. This trial assessed Pfs25M-EPA/Alhydrogel® and Pfs230-EPA/Alhydrogel® individually and in combination for safety and functional immunogenicity in malaria-exposed Malian adults (ClinicalTrials.gov: NCT02334462). The vaccine was given on a 0, 1, 6 month schedule in the 1st year with a booster dose 12 months after dose 3. Given the longitudinal nature of the study cohort, the population level antigen dynamics for both HRP2 and LDH were investigated, as well as relative antigen abundance post-treatment, as determined with the Q-Plex enzyme-linked immunosorbent assay (ELISA).
Methods
Study design and data collection
Individuals in this analysis were part of a double-blind, randomized, phase 1 clinical trial conducted by the Laboratory of Malaria Immunology and Vaccinology (LMIV)/National Institute of Allergy and Infectious Diseases (NIAID)/National Institutes of Health (NIH). Bancoumana is a rural village 60 km from Bamako, Mali, with high prevalence of Plasmodium falciparum. This trial investigated the safety and immunogenicity of Pfs230D1M-EPA/Alhydrogel® and Pfs25M-EPA/Alhydrogel, both transmission-blocking vaccines against P. falciparum. Transmission-blocking vaccines were administered on study days 0, 28, 169, and 540. Blood smears were prepared before each vaccination, at least monthly post vaccination, or when clinically indicated. Starting 1 week after the third and fourth vaccinations, blood smears were prepared twice a week for 6 weeks at the same time that subjects underwent direct skin feeding assays with colony-raised Anopheles coluzzii to assess malaria parasite transmission. Whole blood samples to be analysed in this study were collected at per-protocol scheduled blood draws 1-6 weeks after the fourth vaccination (study days 547, 554, 568, and 582) during peak malaria transmission season (July–December). Between 1 and 21 mL of whole blood was collected at each blood draw for each study participant, all who agreed to have blood samples stored for future research prior to enrollment. Individuals who at any point presented with symptomatic malaria, defined as any parasitaemia by blood smear or RDT positive result with symptoms consistent with malaria, were treated with anti-malarial drugs artemether–lumefantrine (Coartem or Laritem) for uncomplicated malaria and artemether for severe malaria. Whole blood samples for individuals treated with anti-malarials were collected at the same per-protocol frequency as the remainder of the study cohort. Microscopy-positive asymptomatic individuals were not treated, per Malian National Policy on Malaria Control Guidelines.
Sample evaluation
Frozen whole blood samples (n = 622) were sent to PATH’s laboratory (Seattle, WA, USA) for further evaluation. Two ultra-sensitive HRP2-based Alere Malaria Ag P.f RDTs (uRDT), product number 05FK140, lots 05LDB005A and 05LDB004A, were used to test in duplicate each specimen, all of which had been stored at − 80 °C. The test required 5 µL of whole blood and was run following the standard workflow outlined in Das et al. [
15]. A final uRDT result was generated from duplicate uRDT results in agreement only; when results were either discordant or invalid, results were considered not confirmed and excluded from final analyses.
HRP2 and
Plasmodium LDH (pLDH) concentrations were quantified using the Q-Plex Human Malaria Array (4-Plex), which quantifies pLDH by detecting pan epitope [
9]. Standards of recombinant protein with known antigen concentration are run on each plate allowing quantification through standard curves. Ranges of quantification for HRP2 and pLDH were 1.07–16,500 pg/mL and 14.41–525,700 pg/mL, respectively. For numeric analyses, samples with antigen concentrations beyond the limit of quantification (LOQ) for Q-Plex were treated as (upper LOQ) * 2 and (lower LOQ)/2. Thresholds above which samples were defined as antigen positive, determined through receiver operating characteristics analysis to identify the optimal sensitivity and specificity tradeoff, were 2.30 pg/mL for HRP2 and 47.8 pg/mL for pLDH [
9]. Parasite count by microscopy included both gametocytes and asexual parasites and was estimated as parasites per 1000 white blood cells (WBCs), but is reported in parasites/µL, using the conversion of 8000 WBCs/µL [
16]. Gametocyte counts were combined with asexual parasite counts as both have been shown to express HRP2 and LDH [
17].
Statistical analysis
Data compilation and statistical analysis was performed using R 3.6.0 software [
18]. A Bayesian logistic regression model with study level random effects was used to model the relationship between HRP2 concentration and probability of detection by uRDT. A log10 transformation was applied to the HRP2 concentration data and a Gaussian distribution with a mean of zero and standard deviation of three was used for the prior. Four chains of 1000 iterations were ran after a burn-in of 500 iterations, from which median predictions and 95% Bayesian credible intervals (CrI) were taken. A Kaplan–Meier survival curve was generated to estimate probability of a uRDT-positive result for individuals post successful anti-malarial treatment. HRP2 and pLDH dynamics post-treatment were modelled by fitting monophasic and biphasic exponential decay models. A monophasic decay assumes a constant decay rate over time, whereas the biphasic decay model allows for two different decay rates, typically a rapid initial decay followed by a period of slower decay. The functional forms for these two models are:
$${\text{Monophasic: }}\log_{10} \left( {concentration} \right) = k_{1} t + C_{0}$$
$${\text{Biphasic: }}\log_{10} \left( {concentration} \right) = \left\{ {\begin{array}{*{20}l} {k_{1} t} & {\textrm{if}\; t < t_{switch} } \\ {k_{1} t + k_{2} \left( {t - t_{switch} } \right)} & {\textrm{if} \;t \ge t_{switch} } \\ \end{array} } \right\} + C_{0}$$
where
\(t\) = time (in days),
\(k_{1} , k_{2}\) are decay parameters,
\(t_{switch}\) is the switch point between “fast” and “slow” decay, and C
0 is the log10 initial concentration. An individual level random effect was incorporated into each model, accounting for individual variation in antigen concentration at time of treatment (t = 0) and therefore fitting unique values of C
0 to each individual. Models were compared using ANOVA and those that minimized both Akaike information criterion (AIC) and Bayesian information criterion (BIC) were ultimately selected. Predictive intervals were obtained by using the predictInterval function in the R package merTools, which estimates the distribution of all model parameters while incorporating uncertainty in both fixed and random effects. This function was run over 1000 simulations to obtain 95% predictive intervals. Finally, receiver operating characteristic (ROC) curves were calculated to determine optimal thresholds for predicting active versus recently cleared
P. falciparum infection, with thresholds maximizing Youden’s index (the sum of sensitivity and specificity) defined as optimal.
Discussion
This analysis uses data available from both laboratory and field testing in Bancoumana, Mali, to inform three overarching objectives, to: (1) evaluate performance of the uRDT compared to other diagnostic methods, (2) investigate how relative antigen concentrations can classify infections, and (3) better understand the post-treatment dynamics of pLDH and HRP2.
Comparing the HRP2 threshold at which there was a 50% probability of detection by uRDT with the same values from two other studies in Uganda and Myanmar [
7,
8] resulted in unexpected differences. HRP2 concentrations in Uganda were quantified using a Bi-Plex Human Malaria Array, an earlier version of the Q-Plex ELISA with a lower LOD (0.1 pg/mL) [
7], potentially contributing to observed differences in detection thresholds. Other sources of variation could be (but were not confirmed) lot-to-lot variation in performance of the uRDT, variability in class of HRP2 present at the different locations [
20], storage conditions of tests, and/or interpretation of test results. Overall, results indicate a need for further evaluation of the uRDT LOD in the field based on antigenaemia, similar to the large-scale systematic review of co-RDT detection by parasitaemia [
21].
In this Malian population, pLDH had a stronger correlation with parasitaemia than HRP2, with the constraint that parasitaemia was quantified by microscopy and not quantitative polymerase chain reaction (qPCR). This is consistent with findings that residual HRP2 lingers after parasite clearance, whereas pLDH has a shorter half-life and is more indicative of active infection [
22,
23]. This analysis is important in the context of future development or adoption of pLDH-based assays to address emerging
pfhrp2/3 deletions [
25,
26]. In this dataset, there were 12 microscopy-positive, HRP2-negative cases, 4 of which were confirmed to be non-falciparum infections. Of the remaining eight cases, all had low parasite densities (≤ 20 parasites/mL) and only three had a significant pLDH signal (> 300 pg/mL). Further molecular analysis is required to confirm if these are
pfhrp2/3 deletions [
24,
25].
Several first-order kinetics models have previously been used to fit HRP2 dynamics [
22,
26]. Here biphasic exponential decay models were found to best capture pLDH and HRP2 clearance post-treatment [
19]. The nature of biphasic exponential decay (fast, then slow decay) means that previous models may overestimate antigen concentrations initially in the days following treatment [
27].
One of the concerns accompanying introduction of the uRDT is that due to HRP2 persistence, ultra-sensitive HRP2-based diagnostics may lead to overtreatment due to individuals with recently cleared infections testing positive and being treated with anti-malarial drugs when there may be another infection or illness causing fever [
28]. The ability to use a patient’s antigen concentrations to predict if they are in a stage of typical antigen decline post-treatment would be beneficial both to avoid unnecessary retreatment with anti-malarial drugs and to better understand levels of active infection in the population. This need to distinguish between previous versus active infection led us to develop a novel algorithm for distinguishing recently cleared infections from active ones based on both HRP2 concentration and HRP2:pLDH ratios. Although pLDH alone can be a reliable indicator of active infection, it can be difficult to classify pLDH-positive infections as active versus recently cleared without detailed drug treatment histories. Therefore, although perhaps not viable as a standard case management tool, our approach could be used for routine monitoring of drug efficacy at sentinel surveillance sites and to improve estimates of prevalence in cross-sectional surveys. Overall sensitivity of the classification algorithm was promising (77.5%), with predictive power highest for samples with > 100 pg/mL HRP2. In our analysis, HRP2 and pLDH levels below the LOQ were treated as LOQ/2, although most samples likely cleared pLDH within the 21-day post-treatment window due its more rapid clearance dynamics. In order for HRP2:pLDH ratios to become a reproducible metric for distinguishing recently cleared
P. falciparum infections in the future, a standardized protocol for dealing with pLDH values of 0 pg/mL (or < LOQ) in HRP2:pLDH calculations will need to be defined.
This analysis was limited to individuals in a high-transmission
P. falciparum setting and to adults over 18, whereas children under 5 years carry the majority of the global malaria burden [
29]. qPCR data was also not collected in this study. Without being able to account for submicroscopic infections, a significant proportion of the infectious reservoir may be being ignored [
21]. Further analyses should incorporate data from low-transmission settings and on low-density, submicroscopic infections.
Open AccessThis article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit
http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (
http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.