Evaluation of Malaria Rapid Diagnostic Test Performance and pfhrp2 Deletion in Tanzania School Surveys, 2017

ABSTRACT.

As part of malaria nationwide monitoring and evaluation initiatives, there is an increasing trend of incorporating malaria rapid diagnostic tests (mRDTs) in surveys conducted within primary schools to detect malaria parasites. However, mRDTs based on the detection of histidine-rich protein 2 (HRP2) are known to yield false-positive results due to persistent antigenemia, and false-negative results may result from low parasitemia or Plasmodium falciparum hrp2/3 gene deletion. We evaluated diagnostic performance of an HRP2 and pan-parasite lactate dehydrogenase (HRP2/pLDH) mRDT against polymerase chain reaction (PCR) for detection of P. falciparum among 17,051 primary school–age children from eight regions of Tanzania in 2017. According to PCR, the prevalence of P. falciparum was 19.2% (95% CI: 18.6–19.8). Using PCR as reference, the sensitivity and specificity of mRDT was 76.2% (95% CI: 74.7–77.7) and 93.9% (95% CI: 93.5–94.3), respectively. Test agreement was lowest in low transmission areas, where true-positive mRDTs were outnumbered by false-negatives due to low parasitemia. Discordant samples (mRDT-negative but PCR-positive) were screened for pfhrp2/3 deletion by real-time PCR. Among those with a parasite density sufficient for analysis, pfhrp2 deletion was confirmed in 60 samples, whereas pfhrp3 deletion was confirmed in two samples; one sample had both pfhrp2 and pfhrp3 deletions. The majority of samples with gene deletions were detected in the high-transmission Kagera region. Compared with mRDTs, PCR and other molecular methods offer increased sensitivity and are not affected by pfhrp2/3 deletions, making them a useful supplement to mRDTs in schools and other epidemiological surveys.

INTRODUCTION

The WHO recommends the use of parasite-based diagnosis for malaria case management.¹ Malaria antigen–detecting rapid diagnostic tests (mRDTs) serve as an alternative to microscopy in malaria diagnosis, especially in regions where maintaining high-quality microscopy proves challenging. Currently available mRDTs incorporate immobilized antibodies specific to certain antigens, including histidine-rich protein-2 (HRP-2) specific to Plasmodium falciparum, pan-specific or species-specific Plasmodium lactate dehydrogenase (pLDH), or aldolase (specific to all major Plasmodium species: P. falciparum, P. vivax, P. malariae, and P. ovale).^1,2

Because of their ease of use and lack of equipment requirements, mRDTs are increasingly used in malaria school surveys.^3–9 In Tanzania, mRDTs have also been used in school malaria surveys as part of national monitoring and evaluation efforts.¹⁰ However, mRDTs based on the detection of HRP2 can produce false-positive results due to persistent antigenemia and false negative results due to low parasitemia or P. falciparum HRP-2 and -3 (pfhrp2/3) gene deletion.^8,9,11 These inaccuracies can affect the epidemiological stratification of malaria risk areas and targeting of interventions.

Although earlier research has demonstrated high sensitivity and specificity of mRDTs compared with expert microscopy as the gold standard, recent studies have revealed only moderate sensitivity compared with polymerase chain reaction (PCR).¹² The performance of mRDTs can be influenced by various factors, including test quality and technician competence, brand device, user experience, test interpretation, and patient characteristics such as age and fever status.¹¹ Factors such as low parasite densities or presence of parasites with pfhrp2/3 gene deletion may affect the sensitivity of the test, causing false-negative results.^7,9 Also mRDTs using different antigens that are not affected by these deletions, such as pLDH or aldolase, are known to be less sensitive compared with HRP2 mRDTs.^2,9 Conversely, false-positive results can occur when using mRDTs among patients with a recent history of antimalarial treatment.^13,14 This phenomenon can result in an overestimation of the true prevalence of malaria during surveys or diagnostic screenings. To ensure accurate interpretation of mRDT results, it is crucial to consider the potential influence of recent treatment history.

The aim of this study was to evaluate the accuracy of HRP2 and pan-parasite lactate dehydrogenase (HRP2/pLDH) based mRDTs used in a school malaria survey conducted in 2017 that was organized by the Tanzania National Malaria Control Program. The performance of the mRDTs was compared with PCR, a diagnostic tool with high sensitivity in detecting malaria, as the gold standard. Discordances between mRDT and PCR, defined as those samples that were PCR positive but mRDT negative, were assessed for pfhrp2/3 deletions.

MATERIALS AND METHODS

The malaria survey was undertaken in 198 schools from eight of the 26 regions of Mainland Tanzania in 2017, as described elsewhere.^15,16 The regions were selected to represent the heterogeneous transmission seen in Tanzania. Sampling occurred during the rainy season, with 100 students from each school randomly selected for screening. Ethical approval for the school surveys was obtained from the National Medical Research Institute (NIMR; reference no. NIMR/HQR.8a/Vol.IX/2527), Scientific and Ethics Review Committee, and from the University of North Carolina, Chapel Hill (UNC Institutional Review Board nos. 19-1495 and 20-3175). Written informed consent was obtained from parents/guardians of the schoolchildren, and assent was obtained from all children before their enrollment.

Study procedures.

A structured questionnaire was used to collect information on demographic characteristics, use of bed nets, and malaria symptoms in the previous 2 weeks. Finger prick blood samples were used to assess Plasmodium spp. infection in peripheral blood using an mRDT, CareStart^TM Malaria Pf/PAN (HRP2/pLDH) Ag Combo RDT (product code: RMRM 02571; Access Bio, Inc., Somerset, NJ) to detect P. falciparum–specific HRP2 and pan-pLDH antigens; RDTs were considered positive if either band was positive. Children with positive RDT results were immediately treated with artemether-lumefantrine (20 mg of artemether/120 mg of lumefantrine) according to national guidelines. Dried blood spots (DBS) were collected on Whatman filter paper (Whatman plc, Cytiva, Buckinghamshire, United Kingdom) and shipped to the University of North Carolina, Chapel Hill for molecular testing. DNA was extracted from DBS using a chelex extraction method and tested for malaria using real-time PCR for P. falciparum lactate dehydrogenase (pfldh) with an estimated limit of detection of five to 10 parasites per microliter.¹⁶ The detection of P. falciparum has previously been reported and this data was used in this analysis.¹⁶ Discordant samples (pfldh PCR positive but mRDT negative) with an estimated parasitemia of at least 50 parasites per microliter (n = 165) based on a standard curve generated using mock DBS prepared with cultured P. falciparum strain 3D7 parasites were moved forward to testing for pfhrp2/3 deletions using a previously described real-time PCR assay.¹⁷ The real-time PCR assay is capable of detecting mixed infections between intact and deleted parasites, based on a three-cycle difference in amplification (Supplemental Table 1).

STATISTICAL ANALYSES

Demographic and clinical data were summarized as frequencies, percentages, medians, and interquartile ranges (IQR) as appropriate. The diagnostic performance of the CareStart^TM Malaria Pf/PAN mRDT in detecting P. falciparum infection was compared with PCR as the reference standard and kappa statistics was used to assess the agreement of the tests. At the individual level, sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), false-positive rate (FPR), and their 95% CIs were calculated by using the diagt command in Stata version 15.0 (StataCorp LP, College Station, TX). Logarithmic transformation was applied and used to compare median parasite densities of HRP2 positive and negative samples.

RESULTS

As previously described, a total of 17,131 schoolchildren, with a median age of 11 years (IQR: 9–13), in 182 schools from eight regions and 59 districts had PCR or RDT data available.¹⁶ Females represented the slight majority of participants (50.6%). Roughly half of schoolchildren were in regions classified as high transmission strata (51.4%) (Table 1).

Table 1.

Baseline characteristics of schoolchildren in selected regions

Variable	Value (N = 17,131)
Sex (female), n (%)	8,673 (50.6)
Age (years), median (interquartile range)	11 (9–13)
History of fever in the past 2 weeks, n (%)	5,340 (31.2)
Use of antimalarial drugs in the 2 weeks before the survey, n (%)	3,845/5,193 (22.4)
Slept under bed net the previous night, n (%)	3,221 (18.8)
Epidemiologic risk strata,* n (%)
High	8,806 (51.4)
Moderate	2,180 (12.7)
Low	2,952 (17.2)
Very low	3,193 (18.6)

^*Epidemiological risk strata are defined as follows, based on Plasmodium falciparum prevalence in children from the 2014–2015 Tanzanian school malaria parasite survey: “very low” if prevalence <5%, “low” 5% to <10%, “moderate” 10% to <50%, and “high” ≥50%.

Prevalence of P. falciparum and CareStart^TM Malaria Pf/PAN (HRP2/pLDH) diagnostic performance.

The overall prevalence of P. falciparum infection among schoolchildren was 19.2% (95% CI: 18.6–19.8) on the basis of PCR results versus 19.5% (95% CI: 18.9–20.1) as assessed using mRDTs. Of note, given the use of a combination HRP2/LDH RDT, positivity may represent other species as well, but full non-falciparum PCR data are not available for analysis. The majority of these P. falciparum infections exhibited low parasite densities, with 2,676 (81.4%) having less than 50 p/µL and 261 (7.9%) falling within the range of 50 to 100 p/µL—both below the typical mRDT limits of detection—as well as 351 (10.7%) having parasite density >100 p/µL.

Table 2 displays the sensitivity, specificity, PPV, and NPV of mRDTs across different transmission strata using PCR as the gold standard. Sensitivity and PPV were moderate across schools in the high and moderate transmission strata but less than 50% in the low transmission strata where the number of true positives (both mRDT and PCR positive) was small; specificity and NPV remained high across all strata. In areas with high and moderate transmission, mRDT showed substantial concordance with PCR outcomes, indicated by a kappa value of approximately 0.70. However, in low transmission areas, where false positives outnumbered true positives, the agreement was found to be only fair. The ratio of false positive to false negative mRDT results was similar across strata (1.1, 1.0, and 1.2 in the high, moderate, and low transmission strata, respectively).

Table 2.

Diagnostic performance of CareStart™ using PCR as the gold standard in school malaria surveys in Tanzania, 2017

Test	PCR
	Overall		High Transmission Strata		Moderate Transmission Strata		Low Transmission Strata
	+ (n)	– (n)	+ (n)	– (n)	+ (n)	– (n)	+ (n)	– (n)
CareStart™ +	2,489	839	2,219	689	244	105	26	44
CareStart™ –	777	12,946	627	5,192	107	1724	38	2,843
Sensitivity (95% CI)	76.2 (74.7–77.7)		78.0 (76.4–77.8)		69.5 (64.4–74.3)		40.6 (28.5–53.6)
Specificity (95% CI)	93.9 (93.5–94.3)		88.3 (87.4–89.1)		94.3 (93.1–95.3)		98.5 (98.0–98.9)
PPV (95%CI)	74.8 (73.3–76.3)		76.3 (74.7–76.3)		69.9 (64.8–74.7)		37.1(25.9-49.5)
NPV 95%CI) (%)	94.3 (93.9–94.7)		89.2 (88.4–90.0)		94.2 (93.0–95.2)		98.7 (98.2-99.1
Kappa, P-value	0.70, <0.0001		0.66, <0.0001		0.64, <0.0001		0.37, <0.0001

+ = positive; – = negative; NPV = negative predictive value; PCR = polymerase chain reaction; PPV = positive predictive value.

Eighty schoolchildren had missing rapid diagnostic test data and are not included in the table.

Across the study, 48.0% (405/839) of false-positive tests occurred in those who reported fever in the previous 2 weeks. Among those with data on previous antimalarial use, 87% (263/302) of false-positive tests versus 79% (534/673) of true positives occurred in those who reported taking antimalarials in the previous 2 weeks (P = 0.003). In the remaining 13% (39/302) who did not take an antimalarial but had a false-positive test, 23.1% (8/39) reported fever in the previous 2 weeks.

Prevalence of pfhrp2/3 gene deletions in the study areas.

From the full dataset of more than 17,000 samples with PCR and RDT data, we pulled all samples that were RDT negative but had a calculated parasitemia of 50 p/µL or above by qualitative PCR. We obtained a sample set of 165 specimens, all of which were subjected to pfhrp2/3 multiplex real-time PCR testing. Among these, 15 samples failed to amplify in the pfldh channel and were excluded from the dataset resulting in 150 samples for subsequent analysis. Any sample that was identified with a deletion for either pfhrp2 or pfhrp3 in the first run was then repeated with the assay (n = 67 samples). We considered a sample deleted if the repeated run was concordant. If it was discordant, we classified it as indeterminant (NA). The results of the multiplex real time PCR calls are shown in Supplemental Table 1. The mean pfldh CT value for samples where we were able to make a deletion call was 30.4 (range: 25.4–32.7). A total of 60 samples containing either pfhrp2 and/or pfhrp3 gene deletions (one double deleted; one pfhrp3 deletion; and 58 pfhrp2 deletions) were found. The majority of pfhrp2 and pfhrp3 deletions (n = 50, >80%) were detected from samples collected in the high transmission area of Kagera, with a scattering of pfhrp2 deletions in other regions (Figure 1; Table 3). Estimated prevalence of pfhrp2 deletions, based on the number of PCR pfldh-positive samples in each region, ranged from 0 in Mtwara to 6.1% in the Kagera region. There was no difference in parasite densities between samples with and without pfhrp2 deletions (geometric mean parasitemia of 166 [95% CI: 130–212] versus 191 [95% CI: 154–238] parasites/µL for pfhrp2 positive and pfhrp2 negative samples, respectively, P = 0.40).

Figure 1.

Regional pfhrp2 deletion prevalence among schoolchildren in Tanzania. The prevalence of pfhrp2 deletion was estimated based on assessment of rapid diagnostic test–negative isolates with parasite densities >50 p/µL, with all pfldh polymerase chain reaction–positive isolates used as the denominator. The Kagera region in northwest Tanzania exhibited the highest prevalence (6.1%) of pfhrp2-deleted parasites.

Table 3.

The proportion of Plasmodium falciparum samples with pfhrp2/3 deletions

Region	No. PCR pfldh-Positive Samples	pfhrp2 Deletion	pfhrp3 Deletion
Kagera	816	50 (6.1%)	2 (0.2%)
Mara	699	2 (0.3%)	0
Mtwara	602	0	0
Rukwa	253	2 (0.8%)	0
Tabora	505	3 (0.6%)	0
Tanga	399	3 (0.8%)	0
Total	3,274	60 (1.8%)	2 (0.06%)

PCR = polymerase chain reaction.

The number of PCR positive samples based on pfldh was used as the denominator to estimate prevalence of deletions per region. However, deletion assessment was limited to those samples that were malaria rapid diagnostic test–negative but had >50 parasites/µL based on PCR. A total of 165 samples were assessed for deletions, with 150 exhibiting pfldh amplification during gene deletion assessment, indicating adequate parasite DNA present for assessment.

DISCUSSION

The use of rapid malaria tests in school malaria surveys can be immensely beneficial due to their ability to deliver fast results, their requirement for minimal training, and ability to be used in regions with limited access to laboratory facilities. Many malaria-endemic regions have increasingly adopted the use of mRDTs, initially designed for point-of-care use, where symptomatic patients typically exhibit higher parasitemia levels compared with asymptomatic schoolchildren.^3–6 A better understanding of how these tests function as a diagnostic for screening asymptomatic schoolchildren is needed. The use of schoolchildren surveys provides critical information for malaria control programs because this has become the age range with the highest burden of malaria in much of Africa,¹⁸ and they are implemented in many sub-Saharan African countries. The goal of this study was to better characterize the performance of mRDTs relative to a gold standard of molecular detection, allowing for the assessment of the impact of false positive and false negative detection of parasitemia by the mRDT. In addition, we leveraged discordant samples to identify a new focus of pfhrp2 gene deletions in an area of high transmission in Tanzania.

In this study, we saw that the diagnostic performance of mRDT for detection of asymptomatic malaria infections among schoolchildren showed moderate sensitivity (76%) and good agreement (kappa = 0.70) compared with PCR. These findings are comparable with previous studies, which showed moderate sensitivity of mRDTs when PCR was used as a standard reference.¹⁹ However, in low transmission areas, the performance of mRDTs was worse, similar to previous studies.¹⁴ Here, mRDTs has a sensitivity of only 40% for detecting asymptomatic parasitemia among schoolchildren, with poor concordance with PCR (kappa = 0.37). This highlights the need to explore and use more sensitive tools capable of detecting infections with low parasite densities.²⁰

Although the malaria prevalence detected by PCR (19.2%) and mRDT (19.5%) did not differ significantly, there was a notable lack of agreement between the two methods due to both false-negative and false-positive mRDT results. False-negative RDTs could be attributed to the identification of very low-density malaria infections by PCR, with only 2,489 of 3,266 (76%) of PCR positives samples being positive by mRDT overall. Indeed, false-negative tests from low-density infections outnumbered true positives in low transmission areas (38 false negative versus 26 true positive). This was offset by false-positive tests, the majority of which occurred in the setting of previous antimalarial use or recent fever, likely stemming from persistent antigenemia following prior treatment or recently cleared infection. Overall, 839/3,328 (25%) of the mRDT-positive infections were negative by PCR. These observations align with similar studies that indicate factors such as low-density infections and the age groups being tested, such as schoolchildren and adults, can impact the diagnostic performance of mRDTs.^7–9

The occurrence of pfhrp2 and/or pfhrp3 gene deletions can also be a contributing factor to false-negative mRDT results.^2,21–23 In this study, only 60 samples were found to have a gene deletion of pfhrp2 and/or pfhrp3. Thus, the majority of false-negative testing by mRDT is likely due to low-density infections. This finding is consistent with recent studies conducted in Tanzania and Kenya.^24,25 However, the presence of this level of pfhrp2/3 deletion was surprising, especially given the location in Tanzania where it was found. The majority of these samples (>80%) were collected from the Kagera region, but others were also from regions with high transmission strata. It has been postulated that pfhrp2/3 deletions are likely to become more of a problem in areas of low transmission.²⁶ Although these deletions were found in asymptomatic infections, these findings highlight the need for a more comprehensive survey to determine the prevalence of pfhrp2/3 gene deletions among individuals seeking healthcare for fever/malaria symptoms at healthcare facilities in Kagera region. In addition, genomic analysis of isolates identified in this region could help us determine whether this is a new origin of pfhrp2/3 deletions or if these represent spread from other areas of Africa or elsewhere globally.²⁷

The study has limitations, notably that the samples analyzed were collected more than 6 years before the study took place. This temporal gap introduces a potential disparity between the characteristics of the samples at the time of collection and the current circumstances. Additionally, the extended storage of samples before conducting molecular assays poses a risk of underestimating the molecular identification of parasites. In addition, limiting deletion calls to samples with qPCR parasite densities >50 p/µL reduces the risk of deletion misclassification but prevents us from determining the true prevalence of deleted strains. Our estimates should be viewed as a lower bound. Finally, studies of asymptomatic participants are insufficient to inform policy change; the WHO surveillance protocol recommends a 5% prevalence threshold among symptomatic malaria cases to trigger use of non-HRP2 based RDTs.²⁸

CONCLUSION

In conclusion, the mRDTs demonstrated moderate sensitivity but high specificity when compared with PCR for screening asymptomatic schoolchildren for malaria in higher transmission zones. This performance characteristics of the mRDTs worsened with analysis of low transmission settings, likely because of common low parasitemia infections. Additionally, the prevalence of pfhrp2 and or pfhrp3 gene deletions was found to be low overall, but infection by deleted strains unexpectedly clustered in the Kagera region, an area of high transmission. Although mRDTs have been extremely useful, the addition of collecting a DBS to allow molecular methods for understanding the parasite population is highly beneficial. However, the marginal gains in sensitivity from molecular tests may be outweighed by the practical challenges of implementing these sophisticated and costly technologies on a large scale in resource-poor settings.

Note: Supplemental materials appear at www.ajtmh.org.