Prevalence of Plasmodium falciparum lacking histidine-rich proteins 2 and 3: a systematic review

Abstract Objective To calculate prevalence estimates and evaluate the quality of studies reporting Plasmodium falciparum lacking histidine-rich proteins 2 and 3, to inform an international response plan. Methods We searched five online databases, without language restriction, for articles reporting original data on Plasmodium falciparum-infected patients with deletions of the pfhrp2 and/or pfhrp3 genes (pfhrp2/3). We calculated prevalence estimates of pfhrp2/3 deletions and mapped the data by country. The denominator was all P. falciparum-positive samples testing positive by microscopy and confirmed positive by species-specific polymerase chain reaction testing (PCR). If microscopy was not performed, we used the number of samples based on a different diagnostic method or PCR alone. We scored studies for risk of bias and the quality of laboratory methods using a standardized scoring system. Findings A total of 38 articles reporting 55 studies from 32 countries and one territory worldwide were included in the review. We found considerable heterogeneity in the populations studied, methods used and estimated prevalence of P. falciparum parasites with pfhrp2/3 deletions. The derived prevalence of pfhrp2 deletions ranged from 0% to 100%, including focal areas in South America and Africa. Only three studies (5%) fulfilled all seven criteria for study quality. Conclusion The lack of representative surveys or consistency in study design impairs evaluations of the risk of false-negative results in malaria diagnosis due to pfhrp2/3 deletions. Accurate mapping and strengthened monitoring of the prevalence of pfhrp2/3 deletions is needed, along with harmonized methods that facilitate comparisons across studies.


Introduction
Despite improvements in malaria control over the past decade, malaria caused an estimated 405 000 deaths worldwide in 2018. 1 In 2010, World Health Organization (WHO) treatment guidelines established that all cases of suspected malaria should be confirmed by microscopy or an antigendetecting rapid diagnostic test before treatment. 2 Malaria rapid diagnostic tests contain one or a combination of antibodies that recognize specific plasmodial antigens. These antigens include histidine-rich protein 2 (HRP2) which is specific to P. falciparum, and genus-and species-specific lactate dehydrogenase or aldolase, which are produced by all four major human-infecting Plasmodium species. 3 The number of rapid diagnostic tests procured has increased significantly, from 10 million in 2002 to 412 million in 2018. 1 The great majority of these tests detect an HRP2 target, alone or with another antigen, with 15 of 16 (94%) WHO-prequalified malaria tests targeting HRP2 for P. falciparum detection. 4 Rapid diagnostic tests targeting HRP2 came to dominate the market because they are generally more sensitive than other assays and tend to be more heat stable. 5,6 The presence of repetitive epitopes in HRP2 provides numerous antibody binding sites and enables the detection of low levels of protein.
The monoclonal antibodies used in HRP2-detecting tests often cross-react with HRP3, encoded by the pfhrp3 gene, 7,8 particularly at parasite counts above 1000 per µL of blood. 9 HRP3 is a structural homologue of HRP2 that shares similar amino-acid repeats. 8,10 Deletions in the pfhrp2 and/or pfhrp3 (pfhrp2/3) genes as a cause of false-negative rapid diagnostic tests was first recognized in 2010 in the Peruvian Amazon basin. 11 Molecular testing by polymerase chain reaction (PCR) confirmed P. falciparum infection, but also that pfhrp2 and pfhrp3 genes were deleted in 41% (61 samples) and 70% (103 samples) of these 148 samples, respectively. 11 Additional analyses have confirmed a significant increase in the frequency of samples showing pfhrp2/3 deletions in the same area. 12,13 More recently, malaria parasites with pfhrp2/3 gene deletions have been documented in other parts of the world including East, 9,14 Central, 15 West 16 and Southern Africa, 17 Asia 18 and the Middle East. 19 Most concerning was a study in Eritrea that reported samples from 62% (31/50) of microscopy-confirmed P. falciparum patients testing negative for pfhrp2. 20 Collectively, these reports suggest a global threat to the continued use of HRP2-based rapid diagnostic tests.
In 2014, recommendations on investigating and accurate reporting of pfhrp2/3 gene deletions were published. 21 Additional criteria have been proposed in more recent studies, including parasite quantification by microscopy or quantitative PCR to rule out false-negative pfhrp2 detection in samples below the limit of detection of the pfhrp2 assay, 9 and analysis of pfhrp3. 22 However, we have found no assessments of the uptake of these recommendations.
There are increasing numbers of reports documenting the threat of mutant parasite genes for malaria case management. However, due to different study designs and laboratory methods it is difficult to compare findings across studies and accurately understand this threat. We aimed to compile all published studies on the prevalence of pfhrp2/3 gene deletions and assess the quality of methods and reporting. We used our findings to paint a global picture of the current status of pfhrp2/3 deletions to guide decisions on the locations and methods of future surveys.

Search strategy and data extraction
We carried out a systematic review according to the Preferred Reporting Items for Systematic Reviews and Meta-analyses statement. 23 We made a search of the online databases of PubMed®, Scopus, LILACS (Literatura Latino-Americana e do Caribe em Ciências da Saúde), WHO Global Index Medicus and the Web of Science for articles published in any language between 1 January 2010 and 20 August 2019. We used the search terms "[histidine* OR hrp* OR pfhrp*] AND [deletion* OR variation OR diversity OR lack] AND [malaria OR falciparum]" to identify articles reporting molecular analysis of P. falciparum parasite samples for pfhrp2/3 deletions. Additional articles were identified through manual searches. Further information about the search criteria are provided in Table 1. Two investigators screened the titles and abstracts of all eligible articles and extracted the following information from the full text: country, study sites, study design, year(s) of data collection, patient symptom status, age range, number of P. falciparum-positive patients, type of blood sample, which samples underwent molecular analysis, number of samples with pfhrp2/3 deletions, laboratory methods (seven items; Box 1) and analysis of flanking genes. Discrepancies in the data were double-checked.

Prevalence estimates
To maximize consistency in calculating prevalence across studies, we used the total number of P. falciparum samples testing positive by microscopy and confirmed P. falciparum-positive by species-specific PCR as the denomina-tor. We did this regardless of whether all or only a subset of patient samples were tested for pfhrp2/3 deletion by molecular analysis or whether it was the denominator reported in the original publication. If microscopy was not performed, we used the number of samples based on a different diagnostic method or PCR alone. We did not make prevalence estimates from case reports. All prevalence estimates in this review were therefore derived using a standardized denominator and not necessarily the same prevalence as reported in the original article.
Where researchers collected samples from multiple countries, or used different sampling methods or time frames, we separated the results by country or sample collection group to present prevalence data as separate studies. We presented compiled results for studies which collected samples at one point in time from multiple sites with the same sampling design. When we combined data from different studies by country,

Assessment of study quality and bias
We assigned a total quality score from 1 to 7 to each study, based on fulfilment of seven criteria for quality of laboratory methods (Box 1). We assessed study bias as a score from 1 (lowest bias) to 4 (highest bias). The values show the potential bias of the derived prevalence estimate from the true prevalence in the population, depending on the sample population (symptomatic, asymptomatic, mixed or unrepresentative) and samples tested for pfhpr2/3 genes (all, discordant only or another subset). Studies which analysed all samples have a lower bias score than those which only analysed discordant or a subset of samples, while studies which included both symptomatic and asymptomatic samples have a lower study bias than those which only analysed samples from symptomatic people or an unrepresentative sample.

Study characteristics
The included studies showed substantial differences in study design, laboratory methods and data reporting.

Sample populations
The number of samples tested for pfhrp2 ranged from 1 to 783, while the denominator of P. falciparum-positive samples ranged from 1 to 3291 ( Table 2). Out of the 55 studies, 36 (65%) analysed blood samples only from people with symptoms of malaria, as part of a prospective or retrospective survey including unbiased cohorts. Samples in these studies were collected from suspected malaria patients presenting to health facilities or through active case detection. Eight studies (15%) included samples from asymptomatic and symptomatic people as part of cross-sectional surveys or malaria screening programmes, while eight other studies (15%) used samples from an unrepresentative sample of participants and three studies (5%) did not specify the symptom status of the participants. One study collected samples from patients with severe malaria only, while one study collected equal numbers of samples from human immunodeficiency virus-positive and -negative children.
In 35 studies (64%) all samples underwent pfhrp2/3 genotyping. Thirteen studies (24%) genotyped discordant samples only. Of these, nine studies analysed only microscopy-positive and HRP2-rapid diagnostic test-negative samples (of which two were case studies including only one sample), while four studies genotyped only samples which were negative by HRP2-rapid diagnostic test and positive by PCR. One article reporting seven studies only genotyped samples showing the lowest HRP2 concentrations by enzyme-linked immunosorbent assay.

Study procedures
Only three studies (5%) fulfilled all seven criteria for quality of procedures (Box 1). While the number of P. falciparum-positive samples was based on microscopy-and PCR-positive results in 45 studies (82%), in nine studies (16%) the denominator was based on PCR results alone, and in one study (2%) it was based on P. falciparum-specific lactate dehydrogenase-based rapid diagnostic tests and confirmed by PCR. The presence of P. falciparum was confirmed most commonly by amplification of the multi-copy 18SrRNA gene. Thirty-four studies (62%) analysed samples from dried blood spots, 13 (24%) used venous blood, seven (13%) used a combination of both and one study (2%) did not provide information on sample type. Fortysix studies (84%) conducted molecular analysis to determine pfhrp3 deletion. One of these studies only genotyped pfhrp3 deletions among samples found to be pfhrp2-negative.
Twenty-one studies (38%) did not amplify any other single-copy genes while 13 studies (24%) amplified one other and 21 studies (38%) amplified at least two other single-copy genes. To rule out negative pfhrp2/3 PCR results being due to parasite density below the limit of detection of the assay, only samples which were positive by other single-copy genes and failed to amplify the pfhrp2/3 gene were considered to be pfhrp2/3-deleted in the 21 studies which conducted this analysis. The most commonly selected genes for confirmation were P. falciparum merozoite surface proteins 1 and 2, and glutamate-rich protein. One study confirmed pfhrp2 Note: One article reported samples collected from three countries but did not present the results separately, so it is presented as one study.
Malaria parasites, prevalence and distribution Rebecca Thomson et al. deletion by testing for pfhrp3. However, while parasite density was measured in 36 studies, only five studies used these results when determining if a sample was pfhrp2/3-negative. In three studies only samples above a chosen parasite density or deoxyribonucleic acid (DNA) concentration were tested for pfhrp2, while in one study all samples below 5 parasites per µL of blood were discounted and in one study samples were only included in the original study if they were above 2000 parasites per µL of blood.

Prevalence estimates
The derived prevalence of pfhrp2 gene deletions in the 55 studies ranged from 0% to 100% (Table 2). Although we present the overall results by study, 14 studies were conducted over many sites and showed geographically heterogenous results. Further details about the results presented by region are provided in the data repository. 51 In Fig. 2 we mapped the geographical distribution of the highest derived prevalence estimate of pfhrp2 gene deletions by study for each country. The highest derived prevalence was above 50% in Colombia, Eritrea and Peru. Fig. 3 plots the weighted average prevalence of pfhrp2 gene deletions for each country and the range by study sites. The weighted average prevalence ranged from 0% to 43%. Average prevalence above 20% was found in Eritrea, Ghana, Nicaragua, Peru and Sudan.
We plotted the prevalence of pf-hrp2/3 gene deletions by sample size (available in the data repository). 52 Five studies had a sample size over 1000, while 36 had sample sizes smaller than 100. All seven studies reporting greater than 50% prevalence of pfhrp2 deletions had a sample size of fewer than 55. Scatter plots of prevalence against time are available in the data repository. 53 Table 2 shows the bias scores of the prevalence estimates from the true prevalence of pfhrp2/3 gene deletions in the population. Six studies (11%) had a bias score of four while two (4%) had a bias score of one.

Discussion
We found that mutant parasites have been reported from all major malariaendemic areas, in asymptomatic and symptomatic P. falciparum-positive patients. However, our results also confirm that the full extent of the threat has not yet been characterized. The limited number of well conducted prevalence surveys in malaria-endemic countries indicate geographical variability in the prevalence of mutations in the pfhrp2 and pfhrp3 genes and do not completely illuminate the factors driving these differences.
The study has limitations. Although we included only published articles, we were aware of other abstracts and doctoral theses for which relevant data on methods were not available. For manuscripts included in the review, we contacted authors to obtain information not included in the manuscripts; this was not always possible, however, and we therefore occasionally made assumptions about the methods. Survey design and sample populations varied greatly across the included studies. Most studies were not purposely designed for investigating the prevalence of gene deletions and relied on convenience sampling or on secondary analyses of existing specimens. These shortcomings limit our ability to draw conclusions that can inform the use of rapid diagnostic testing, but rather provides guidance for future surveys.
Reconciling the different populations and sample sizes across studies is challenging. First, studies of asymptomatic and symptomatic patients require different interpretations and are difficult to integrate. Samples from asymptomatic patients may have lower parasite densities, resulting in less DNA target for amplification and potentially greater risk of falsely reporting pfhrp2 deletions. This risk is especially high when the investigation does not include amplification of other single-copy genes or does not quantify parasite DNA. Furthermore, little is known about the effect of pfhrp2/3 gene deletions on the virulence of malaria infection. If pfhrp2/3 deletions are associated with less virulent infections, there could be a difference in prevalence between symptomatic and asymptomatic infections. We found numerous studies with low sample sizes which may not be representative of the true prevalence of deletions in a population or country.
Second, different selection criteria for pfhrp2/3 genotyping (all malaria suspects or only those with discordant HRP2-based rapid diagnostic test and microscopy and/or PCR results) result in the use of different numerators and denominators for prevalence estimation across studies. Analysis limited to deletions found among discordant samples leads to a higher prevalence of gene deletions being reported. To improve consistency in calculating the prevalence of pfhrp2-negative mutants, we used the total number of samples with confirmed pfhrp2 gene deletions by species-specific PCR as the numerator and the total cohort number of P. falciparum-positive samples by microscopy and/or PCR as the denominator. The WHO-recommended approach of testing only a subset of high-risk (discordant) samples is a more economical way of monitoring the prevalence of gene deletion and targets clinically significant deletions that cause negative test results. WHO recommends using nonexclusive HRP2-based rapid diagnostic tests when the prevalence of pfhrp2/3 gene deletions causing false-negative test results is greater than 5%. 22 Most studies included in this review did not allow us to determine if the WHO threshold had been exceeded. It is well acknowledged that the WHO approach may underestimate the prevalence of pfhrp2 deletions in the parasite population. Samples that are pfhrp2-negative and pfhrp3-positive are not likely to be flagged as high risk or discordant due to the cross-reactivity between HRP2 and HRP3 proteins on many rapid diagnostic test brands. However, this concern does not pose an immediate threat to patients. 54 Ideally, all P. falciparum-positive samples should be used as the denominator, determined either by microscopy or a good quality rapid diagnostic test for detecting P. falciparum-specific lactate dehydrogenase.
The study bias scores show the potential bias of the prevalence estimates from the true prevalence of pfhrp2/3 gene deletions in the population, but not necessarily the bias of deletions causing false-negative results (which is more important for determining the effect on malaria case management). Ultimately, Systematic reviews Malaria  Malaria parasites, prevalence and distribution Rebecca Thomson et al.
high-throughput screening options could become more readily available and more commonly used. If so, the true prevalence of pfhrp2 gene deletions could be determined by molecular testing of all people with suspected malaria regardless of rapid diagnostic test or microscopy results, and those samples confirmed to have pfhrp2 deletions used as the numerator.
Third, recent modelling suggests that the likelihood of finding pfhrp2/3 deletions can vary during the malaria transmission season due to changes in the transmission intensity and multiplicity of infection, whereby a person can be infected with multiple P. falciparum strains. 55 Co-infection with pfhrp2/3negative-and wild-type parasites can prevent detection of gene deletions using current laboratory techniques, leading to an underestimation of the prevalence of pfhrp2/3-negative mutants. Time of year and relation to the transmission season is rarely described in published reports. A publicly available database using prediction models could be useful to help determine the optimal time in the transmission season to conduct a gene deletion survey. 56 Due to the small number of studies, differing populations and often small samples sizes of the reviewed studies, it is difficult to draw robust conclusions on the prevalence of pfhrp2/3 gene deletions in specific areas or to perform meta-analysis from these data. The implementation of more large-scale, robust surveys would enable a better understanding of if, and at what rate, these mutations are increasing in a given area, and would allow for meta-analysis.
Identifying the prevalence of pf-hrp2/3 deletion mutations requires synthesis of several lines of evidence and study procedures that include proper performance of rapid diagnostic tests and careful genotyping methods. While most studies in this review followed some components of published criteria to classify a sample as pfhrp2-deleted, 21 only 3 (5%) of the studies followed the seven recommended criteria proposed in this review. One specific challenge for molecular analyses of pfhrp2/3 is using the absence of amplified products as the indicator of interest, rather than the presence of amplified products. Rigorous methods and appropriate controls must be used to ensure the presence of

Region, country or territory
Notes: the prevalence of phfrp2 gene deletions was calculated using the number of Plasmodium falciparum-positive samples as the denominator. In countries with multiple studies prevalence estimates were weighted based on the number of patients tested. Countries with only one study have no bar shown. Cases studies and a study from Zambia that tested for pfhrp2-leader sequences are not included.

Systematic reviews
Malaria parasites, prevalence and distribution Rebecca Thomson et al.
non-degraded, amplifiable parasite DNA and the lack of amplicon contamination. Improving the accuracy of survey outcomes requires novel molecular-based technology and methods that could detect pfhrp2/3 gene deletions more reliably and efficiently and detect pf-hrp2/3 deletions in samples with mixed infections (such as quantitative-PCR and whole genome sequencing). Not all malaria-endemic countries have the capacity to conduct molecular analysis to a high standard, and establishing such capacity is challenging and costly. In addition, the sensitivity and specificity of PCR assays can be affected by the protocol used, potentially resulting in variations in the results across laboratories following different procedures. For example, lowering the elongation temperature on pfhrp2 assays improved the limit of detection of many previously published assays. 57 WHO has established a network of laboratories capable of conducting this analysis to ensure that samples from prevalence surveys can be performed quickly and procedures harmonized across laboratories. 22 Just over half of the studies amplified both exon 1 and 2 of the pfhrp2 gene, while the rest amplified only exon 2. While the chromosomal break points could theoretically be anywhere within the pfhrp2 gene, it appeared that most samples from Eritrea and Peru have observed deletions in both exon 1 and 2 (Qin Cheng, Australian Defence Force Malaria and Infectious Disease Institute, personal communication, 2019). Therefore, whether analysis of exon 2 alone is sufficient to identify most parasites with pfhrp2 gene deletions requires further analysis of gene deletions from other parasite populations. While not included in the recommendations for pfhrp2/3 molecular analysis, 21 analysis of flanking genes can provide additional information on genetic mutations.
Despite the diversity of study approaches, there appear to be areas of high prevalence of pfhrp2/3 mutant parasites where diagnostic testing based on HRP2 alone would be inadequate. Thus, the need for alternative rapid diagnostic tests is of urgent concern in the Amazon basin and Eritrea, where the prevalence of tests producing falsenegative results among symptomatic patients has forced changes in the diagnostic strategy. 58 Malaria control programmes should remain vigilant for evidence suggesting the presence of pfhrp2/3 gene deletions. Evidence of false-negative rapid diagnostic tests or confirmed pfhrp2/3-negative mutants in neighbouring countries should trigger careful investigation and surveillance.
To improve the quality and relevance of surveys for clinical case management, WHO now provides general guidance on when to prioritize surveys for pfhrp2/3 deletions. 22 WHO has also developed protocols for guiding survey design, data collection and laboratory methods to determine the prevalence of clinically-relevant pfhrp2/3 deletions causing false-negative rapid diagnostic tests. 59 The guidelines aim to ensure that future investigations are implemented to high and comparable standards. Additionally, an up-to-date repository of pfhrp2/3 deletion studies is maintained on the WHO malaria threat map. 60 The specific factors that drive the evolution and spread of pfhrp2/3 mutations are not clear, although mathematical models suggest that selective pressure by HRP2-detecting rapid diagnostic tests over the past decade is likely to have played an important role. 5 Low ma-laria transmission and high frequency of people correctly treated on the basis of diagnosis with HRP2-detecting tests have also been identified as key drivers of the selection of pfhrp2/3-negative mutants. 61 Nevertheless, the existence and rising prevalence of pfhrp2 deletions in Peru, 11,12,33 where HRP2-only methods have never been widely used, along with the local prevalence of pfhrp3 mutations, confirms that selective treatment based on test results is not the only factor driving the evolution of these parasites.
Due to the global reliance on rapid diagnostic tests for malaria diagnosis, a coordinated, multifaceted response to P. falciparum with pfhrp2/3 gene deletions is required. This response should include representative studies of the prevalence and distribution of pfhrp2/3 deletions, more efficient and affordable methods for screening and confirming these deletions, and efforts to standardize and ensure high-quality reporting. Followup surveys in areas with documented pfhrp2/3 deletions will provide insight into the speed at which the mutant parasites are evolving in response to diagnostic pressure and other drivers. Research for the development and commercialization of rapid diagnostic tests based on new or improved non-HRP2 targets is an essential parallel area of work. ■

Country or territory
Year of data collection