Background
Plasmodium knowlesi was first recognized as a cause of simian malaria in long-tailed or pig-tailed macaques [
1]. It was also recognized as a cause of human malaria in 1965 [
2]. The large focus on
P. knowlesi as a cause of human malaria was reported in Sarawak, Malaysia, in 2004 [
3]. Since then,
P. knowlesi malaria in humans has been reported throughout Malaysia [
4‐
22] and other Southeast Asian countries including Thailand [
23‐
26], Indonesia [
27‐
30], Singapore [
31,
32], Brunei [
33], Cambodia [
34,
35], Laos [
36,
37], Myanmar [
38], the Philippines [
39], and Vietnam [
40]. Moreover,
P. knowlesi malaria has also been reported in travellers returning from endemic countries [
41‐
54].
Although most
P. knowlesi infections are asymptomatic, approximately 19% of infected patients develop severe infections, including acute kidney injury (AKI) (45.6%), jaundice (42%), and hyperparasitaemia (32.5%), as the common clinical manifestations [
55]. Deaths from
P. knowlesi infection have been linked to delayed parenteral treatment [
56]. In the endemic country of Malaysia, early intravenous artesunate treatment is now recommended for all severe malaria cases to prevent mortality, resulting in a lower death rate during 2010–2014 [
8]. The risk factors associated with
P. knowlesi infection include older age, male sex, plantation work, sleeping outside, and travelling in areas where monkeys live [
21,
55,
57]. A recent study also suggested that the transmission of
P. knowlesi malaria between humans might occur with mosquitoes as vectors, given the presence of family clustering [
14].
The identification or detection of malaria parasites relies on the results of analysis based on microscopy, the standard for malaria diagnosis. However, the use of microscopy to diagnose
P. knowlesi malaria is inaccurate since the morphological features of the early trophozoites of
P. knowlesi resemble those of
Plasmodium falciparum, and the growing trophozoites are similar to the band-form trophozoites of
Plasmodium malariae [
1,
58]. In this study, the rate at which
P. knowlesi is misidentified as
P. malariae by microscopy was estimated and quantified to clarify the inherent disadvantage of solely utilizing microscopy to identify
P. knowlesi infection in endemic and non-endemic areas.
Methods
This study followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) (Additional file
1: Checklist S1) [
59]. The protocol was registered in the PROSPERO International Prospective Register of Systematic Reviews (ID = CRD42020204770).
Search strategy
Searches of potentially relevant articles published from January 1, 1993, to August 17, 2020 were performed in MEDLINE, Scopus and Web of Science. The search terms used were (Plasmodium OR malaria) AND knowlesi AND (microscopy OR microscopic OR blood film OR “blood film” OR “thick film” OR “thin film”) AND (PCR OR “polymerase chain reaction”). The searches aimed to find original articles in any language and ended on August 17, 2020.
Eligibility criteria
Original research articles were eligible to be included in the present study if they were on retrospective or prospective cross-sectional studies and reported: (1) the misidentification of P. knowlesi as P. malariae as identified by microscopy and (2) the confirmation of P. knowlesi cases by molecular methods. Studies/papers were excluded for the following reasons: absence of P. malariae or P. knowlesi as determined by microscopy, absence of P. knowlesi as determined by PCR, microscopic findings of P. malariae/P. knowlesi, P. knowlesi in macaques, submicroscopic P. knowlesi, unextractable data, case–control studies, case reports or case series, clinical trials, conference abstracts, experimental research, guidelines, letters to the editor, test performances, review articles, systematic reviews and use of the same participants or data set as in another study.
Study selection and data extraction
The selection of the included studies according to the eligibility criteria was performed by two of the authors (MK and AM). Any discrepancies between these two authors were resolved by discussion in order to reach a consensus. For each study that was included in the analysis, the following information was extracted: name of the first author, year of publication, study area (years of the survey), study design, age range (years) of the participants, sex (male, %) of the participants, PCR detection for Plasmodium spp., target genes, number of P. malariae and P. knowlesi identified by microscopy (including mixed infections), number of P. malariae and P. knowlesi identified by PCR (including mixed infections), and number of species discordances. Raw data from each study were stored in a standardized datasheet before data synthesis. Data selection and extraction were managed using Endnote Software X7 (Clarivate Analytics, Philadelphia, USA).
Quality of the included studies
The risk of bias for each study was assessed using the Quality Assessment of Diagnostic Accuracy Studies (QUADAS) [
60]. This tool comprises four domains: patient selection, index test, reference standard, and flow and timing [
60].
Statistical analysis
Data from the included studies were analysed using the STATA Statistical Software Version 15.0 (StataCorp LLC, Texas, USA). The number of cases of Plasmodium species discordance (P. knowlesi as P. malariae) as identified by microscopy and the number of P. knowlesi cases identified by PCR were used to analyse the pooled prevalence of the misidentification of P. knowlesi as P. malariae. The pooled prevalence of discordance of the misidentification of P. knowlesi as P. malariae was estimated by a random effects model using the numerator in the prevalence calculation as the number of discordances, and the denominator as the number of PCR-positive malaria cases. The pooled prevalence and 95% confidence interval (CI) of the misidentification were estimated using a random effects model. Subgroup analysis of the study sites was performed to demonstrate any differences in the pooled prevalence in both endemic and non-endemic countries. The existence and level of heterogeneity across the included studies were assessed using Cochrane Q and I2 statistics, respectively. Publication bias was assessed using funnel plot asymmetry and Egger’s test for asymmetry.
Discussion
Although
P. knowlesi is well documented in Malaysia, the pooled quantification of the misidentification of this species as
P. malariae has not been previously described. In this study, such misidentification was quantified using 11 studies [
3,
13‐
19,
23,
24,
27], and it was found that the rate of this misidentification from 2000 to 2015 was 57%, with high heterogeneity among the included studies. Subgroup analysis of the study sites demonstrated a large difference in the misidentification rates. The highest prevalence of misidentification was demonstrated in two states of Malaysian Borneo, Sabah and Sawarak, where
P. knowlesi was endemic in the last decade. In these areas, the number of
P. knowlesi cases was not available until PCR testing was used to identify it in 2004 [
3]. Molecular detection assay, nested PCR assay and real-time PCR test have been described for
P. knowlesi targeting 18S rRNA gene targets [
3,
61], with sensitivity of between 1 and 6 parasites/µl of blood [
62]. From the time that nested PCR was implemented for diagnosis, the number of cases of
P. knowlesi significantly increased, while a small number of
P. malariae cases were still observed by PCR in Sabah during 2008–2011 [
14,
15,
18] and in Sawarak during 2001–2006 [
19]. This indicated that the highest number of
P. malariae cases identified by microscopy in the last decade was caused by the emergence of
P. knowlesi malaria, as these two species are morphologically similar and difficult to distinguish from each other using microscopy. While the highest prevalence of the misidentification of
P. knowlesi as
P. malariae occurred in Sabah and Sawarak, microscopically misdiagnosed cases of
P. malariae were not found in other parts of Malaysia, such as Pahang and Kelantan [
16]. This explained why the prevalence of the misidentification of
P. knowlesi as
P. malariae in the study by Yusof et al
. [
16] in nine states of Malaysia was lower than in studies conducted in Sabah and Sawarak [
3,
13‐
15,
17‐
19].
In areas in which
P. knowlesi was endemic, it was also frequently misidentified as
P. falciparum or
P. vivax malaria by microscopy [
10]. The similarity of
P. knowlesi and
P. falciparum occurs at the stage of young rings of both species, which contain double chromatin dots, multiple-infected erythrocytes, and applique forms [
63]; while the similarity of
P. knowlesi and
P. malariae occurs in the trophozoite, schizont, and gametocyte stages [
63]. The recent decrease in diagnostic discrepancies by microscopy was due to the increased awareness and recognition among microscopists of
P. knowlesi infections in endemic areas. Moreover,
P. malariae is less endemic in Southeast Asia, where the presence malarial parasites with morphology similar to that of
P. malariae coupled with high parasitaemia has been reported as
P. knowlesi infection by default [
16]. Moreover, the World Health Organization (WHO) recommends that all
P. malariae-positive diagnoses by microscopy in
P. knowlesi endemic areas be reported as
P. malariae/
P. knowlesi [
64].
In areas where
P. knowlesi is not endemic, such as Thailand and Indonesia, a low prevalence of the misidentification of
P. knowlesi as
P. malariae by microscopy was observed. Only one case of
P. knowlesi misidentified as
P. malariae from 33 confirmed cases of
P. knowlesi was recorded by Jongwutiwes et al
. during 2008–2009 [
24], and only one such case among ten confirmed cases of
P. knowlesi was recorded by Putaporntip et al
. during 2006–2007 [
23]. There was also a low prevalence of such misidentification by microscopy during 2014–2015 in Aceh, Indonesia [
27], as only three cases of
P. knowlesi were misdiagnosed from 19 confirmed cases of
P. knowlesi as recorded by Coutrier et al
. [
27]. In addition,
P. knowlesi was also misidentified as
P. falciparum and
P. vivax, as reported by studies in both Thailand and Indonesia [
24,
27]. This indicated that microscopists were unable to recognize
P. knowlesi because its ring forms were similar to those of
P. falciparum, or sometimes its growing trophozoites were similar to those of
P. vivax. Misidentification, such as the misdiagnosis of
P. falciparum as
P. knowlesi, might cause the administration of chloroquine, and the resistance of
P. falciparum to chloroquine can increase the likelihood of patient mortality. Further, the misidentification of severe
P. knowlesi as
P. vivax malaria may lead to treatment failure, such as missed parenteral treatments as per national guidelines, which have been reported to be associated with fatal outcomes [
56]. In addition to the misidentification of
P. knowlesi mono-infection, mixed-infections of
P. knowlesi combined with other
Plasmodium species were also recorded in Thailand and Indonesia, such as mixed-infections with
P. falciparum or mixed-infections with
P. vivax malaria, which microscopists reported as
P. falciparum or
P. vivax mono-infections [
23,
24,
27]. Severe complications due to
P. knowlesi malaria in those co-infected patients in non-endemic countries such as Thailand and Indonesia were less likely since low parasite density of
P. knowlesi was observed [
23,
24], and
P. knowlesi was responsive to chloroquine treatment in cases of mixed infections with
P. vivax malaria. In addition, severe adverse events from unnecessary primaquine treatments were not experienced among co-infected patients [
27].
The present study had some limitations. First, high heterogeneity among the included studies was observed, although subgroup analysis was performed; therefore, the results of the pooled prevalence needed to be interpreted carefully. Second, a low number of included studies were used for pooled analysis; therefore, the pooled prevalence might not have been precisely estimated. Third, studies reporting on P. knowlesi/P. malariae as determined by microscopy were not included in the present study since the number of misidentifications could not be estimated.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.