Background
Malaria is a major public health problem that still causes significant morbidity and mortality in developing countries [
1,
2]. In 2019, an estimated 228 million cases and 405,000 deaths were recorded annually [
3]. Most malaria cases were in the World Health Organization (WHO) African region (213 million or 93%) [
3]. An estimated 90% of all global malaria mortality is also in sub-Saharan Africa, mainly in children aged under 5 years [
4].
In Ethiopia, approximately 75% of the land mass is estimated to be malarious with about 52 million people being at risk of malaria.
Plasmodium falciparum (70%) and
Plasmodium vivax (30%) were the dominant
Plasmodium species in Ethiopia as measured by microscopy [
5]. Malaria transmission in Ethiopia is seasonal and unstable with the peak transmission season from September to December, following the main rainy season from June/July to September [
5,
6].
Rapid accurate diagnosis followed by effective treatment is very important for malaria control. For decades, light microscopy remains the “golden standard” method for detecting and identifying malaria parasites although it requires training and experience [
7]. The use of rapid diagnostic test (RDT) that detects malarial antigen is vital especially in resource-poor settings; it is simple to use and needs a little expertise, it doesn’t require electricity and results can be obtained in few minutes [
7‐
9]. Rapid diagnostic tests are immunochromatographic lateral flow devices detect parasite antigens, such as histidine rich protein 2 (HRP2) that detect only
P. falciparum and
Plasmodium lactate dehydrogenase (pLDH) or aldolase that detect all
Plasmodium species [
7]. Histidine rich protein 2 (HRP-2) and pLDH are the most commonly detected malarial antigens [
7,
9].
Molecular detection of DNA/RNA using polymerase chain reaction (PCR) is another alternative method for malaria diagnosis. The PCR techniques includes conventional PCR, nPCR, qPCR and multiplex PCR [
2]. It is more sensitive than microscopy and RDTs for malaria detection, but requires well-trained staff, sophisticated laboratory equipments and a good quality assurance system [
10,
11]. The loop-mediated isothermal amplification (LAMP) is a recently developed molecular technique to that is simpler and faster with an excellent diagnostic accuracy [
2].
Selection of diagnostic tests should consider affordability, number of tests to be performed, equipment required, trained staff, besides diagnostic method accuracy. However, a test method must have sufficient level of accuracy although the test method is practical and affordable and perhaps the only possibility in a certain situation. Therefore, this study aimed to investigate the published studies of diagnostic accuracy of RDTs, microscopy, LAMP and PCR for the malaria diagnosis in Ethiopia.
Methods
Search strategy and eligible studies
The preferred reporting items for systematic review and meta-analysis’ (PRISMA) guidelines was used to report this systematic review and meta-analysis. Studies that assessed the diagnostic accuracy of RDTs, microscopy, LAMP and PCR for the detection of malaria parasites were eligible for this review. Case–control studies were excluded to overcome an over-estimation of the sensitivity and specificity of tests [
12]. RDTs detecting any type of antigens in any format and from any manufacturer were eligible. Molecular diagnostic tests (PCR) in any format using
Plasmodium DNA and/or RNA amplification were also eligible to be included.
Studies were searched on electronic databases such as PubMed, PubMed central, Science direct databases, Google scholar, Scopus, proceedings of health professional associations [such as Ethiopian medical laboratory association (EMLA), and Ethiopian public health laboratory associations’ (EPHLAs)]. The search was performed from September to October, 2020 using key words including “malaria”, “P. falciparum”, “P. vivax”, “P. ovale”, “P. malariae”,“diagnosis”, “RDT”, “HRP-2”, “pLDH” “PCR”, “microscopy” “diagnostic test accuracy”, “systematic review”, “meta-analysis” and “Ethiopia”. These search terms were also combined with each other using Boolean operators (AND, OR) to retrieve all relevant studies. Overlapped studies found in more than one databases were excluded. The reference lists of included studies were searched to retrieve additional studies.
Selection of studies
Studies were selected based on their title and abstract by two authors (DGF and YA) independently. Duplicated, studies published before the year 2000 and studies without reference test methods were removed. Studies considered relevant by at least one of the two authors were considered in further review. Disagreements about the eligibility of studies were solved by all authors after the detailed discussions on the pre-set inclusion and exclusion criteria. Studies found only in abstract form were included only after obtaining full article by communicating the authors whenever their contact is available.
Data extraction and management
Pre-designed data extraction form was developed on Microsoft excel 2010 by authors based on the objective of this review. Two authors (DGF and YA) collected the required data from the included studies independently. Information about studies (title, authors, journal), study design, descriptions of reference and index tests and data for 2 × 2 tables were collected. Methodological quality of the included studies was assessed using the Quality Assessment of Diagnostic Accuracy Studies QUADAS-2 tool [
13] by two reviewers independently (DGF, NY). This tool assesses bias in patient selection, index test, reference standards, and flow/timing areas. A risk of bias summary and graph was generated in Review Manager 5 (RevMan version 5.4.1). When a study compares more than two tests, multiple 2 × 2 tables were extracted from a single study. In this case for each test comparisons, separate quality items were considered within a single study.
Statistical analysis
The estimates of sensitivity and specificity and their 95% confidence interval were plotted in forest plots using Review Manager 5.4.1 [
14]. Review manager plots do not provide summary points and heterogeneity measures are not sufficiently provided. Therefore, Midas in Stata 14.0 was used to calculate the summary estimates of the sensitivity, specificity, positive likelihood ratio, negative likelihood ratio, and diagnostic odds ratio. Diagnostic accuracy tests are expected to show considerable heterogeneity. As a result, hierarchical summary receiver operating characteristic curve (HSROC) was used [
15]. Midas was also used to assess heterogeneity of the included studies. When the heterogeneity was significant or I
2 > 50%, Meta-disc 1.4.0 software was used to explore whether a threshold effect existed. Meta-regression was performed to investigate the potential sources of heterogeneity. The covariates investigated for possible source of heterogeneity were sample size, sampling method, blinding status, study population and target antigens.
Discussion
There are several strategies used to achieve malaria control and elimination. These are accurate and prompt diagnosis, measuring the impact of intervention and effective treatment [
2]. In malaria endemic areas, there are a significant proportion of asymptomatic malaria carriers due to the decrease in a patient’s parasitaemia. Microscopy is an appropriate method for detecting and identifying malaria parasites and has been the golden standard method for malaria diagnosis for decades. However, it requires training and experience of microscopist.
Rapid diagnostic tests that detect malarial antigens (HRP-2, pLDH, Aldolaes) are an alternative malaria diagnosis method especially in resource-poor settings; it needs a little expertise and doesn’t require electricity. Molecular detection of malaria using PCR is another advanced diagnostic method. Although this method requires well-trained staff and well-structured laboratory infrastructure, it is more sensitive than microscopy and RDTs. The aim of this study was to compare and analyse the performance of RDTs, microscopy, LAMP and PCR. An extensive search was performed for all the available studies regardless of study areas and clinical presentation of tested individuals in Ethiopia. In the present study, 26 studies were analysed to estimate the accuracy of RDTs for diagnosing malaria. With the exception of one study, all the included studies were done between 2009 and 2020.
Statistical heterogeneity was tested using the
I2 statistic, which measures the variation across studies due to inter-study heterogeneity. Heterogeneity was expected to be related to the method of test reading, the patients levels of parasitaemia, and the comparators. There was significant heterogeneity among the studies included in this meta-analysis. Meta-regression showed variables such as blinding status and target antigens were the source of heterogeneity. Subgroup analyses of studies based on target antigen showed that HRP2 based RDTs demonstrated higher sensitivity than HRP2/pLDH antigen based kits. This was in agreement with reports of previous studies that showed HRP2-based RDTs are more sensitive compared with pLDH-based RDTs [
44,
45].
In this study there was a study that target HRP-2 & pan-aldolase. It showed a sensitivity of 95% and specificity of 89%. This finding was lower than a study that reported a sensitivity of 97.4%, and a specificity of 100% that target
P. vivax-specific aldolase [
46].
In the present study, the summary estimate of sensitivity and specificity of RDTs using microscopy as a golden standard method were 95.05% (95% CI 92.95–96.55%) and 96.47% (95% CI 94.69–97.67%), respectively. In this study the sensitivity was lower than a systematic review and meta-analysis study conducted in India that reported a sensitivity of 97.0% (95% CI 95.0–98.0%). On the other hand, the specificity of this study was almost similar with a study conducted in India (specificity = 96.0% (95% CI 93.0–97.0%) [
47]. In contrast to the summary estimate of sensitivity, the summary estimate of specificity of this study (96.47%) was almost similar with a report from a study that compared RDTs with microscopy and PCR among pregnant women (94%) [
15].
The AUC of RDT using microscopy as a reference method was 0.99 (95% CI 0.98–1.00). This indicates that RDTs are diagnostic test methods with an excellent specificity and sensitivity. Therefore, RDTs can be used as an alternative malaria diagnosis method for microscopy especially in resource poor settings.
The PCR techniques used as a reference test for evaluating RDTs includes qPCR, nested PCR, qRT-PCR and Semi-nested Multiplex PCR. Rapid diagnostic tests had an AUC of 0.83 (95% CI 0.79–0.86). However, RDT showed lower sensitivity (37%) in a study conducted among mixed population of symptomatic and asymptomatic individuals. Similarly the sensitivity of RDT was 51% in a study conducted among clinical and sub-clinical patients. This might be due to the reason that among asymptomatic individuals and sub-clinical patients the parasitaemia level expected to be low and usually submicroscopic. However, the specificity of RDT in these study groups (asymptomatic individuals and sub-clinical patients) was 100% and 94%, respectively. This indicates RDT still has a good diagnostic accuracy for malaria parasite detection.
In this study, the performance of microscopy was also evaluated using PCR as a reference test. Microscopy showed a sensitivity that varies between 39% and 97% and a specificity between 80% and 100%. However, two of the studies conducted among asymptomatic individuals and clinical and sub-clinical patients were showed lower sensitivities. Overall, microscopy demonstrated a very good diagnostic accuracy for the malaria parasite detection (AUC = 0.95 (95% CI 0.93–0.97).
Rapid diagnostic test (RDT) has lower sensitivity (37%–88%) but had good specificity (93%–100%) with the exception of one outlier (28%) using PCR as a reference test methods. A decrease in sensitivity might be due to low parasitaemia of individuals that may influences the detection ability of RDTs.
Rapid diagnostic tests showed higher diagnostic odds ratio when compared with microscopy than PCR. It had 525.67 (95% CI 299.89–924.41) times higher odds of obtaining positive result in diseased individuals than in non‐diseased. On the other hand, it showed 40.22 (9.23–175.28) times higher odds of positive test in positive individuals than negative individuals. This can be explained by the fact that performance of RDT is much closer to microscopy. This was supported by the present study finding as it showed an excellent diagnostic accuracy (AUC = 0.99).
The present study showed that LAMP had an excellent sensitivity and quite good specificity when PCR is used as a reference test. This was in line with a systematic review and meta-analysis of diagnostic accuracy of LAMP methods which showed a sensitivity and specificity of > 95% in majority of the studies [
15].
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.