Data and study characteristics
The literature search generated 466 results in PubMed, 354 in Scopus, 271 in Web of Science and 909 in Elsevier (Fig. 1). After removing duplicates, 1202 articles were left for screening. Following screening of titles and abstracts, 295 studies were retained for more detailed evaluation. The most common reason for exclusion was the unavailability of data for analysis. Other reasons for exclusion including experimental studies, cohort studies, and reviews. As a result, 46 articles were selected in the study for full data extraction (5, 7-9, 23-25, 27, 28, 31-67).
Description of included studies
Of the 46 included studies, six studies (13%) were published from 1988 to 1999, followed by 11 studies (24%) from 2000 to 2010 and 29 studies (63%) from 2011 to June 2019 (Table 1). The majority of studies were conducted in Malaysia Borneo (31/46; 67.4%), followed by Peninsular Malaysia (13/46; 28.3%) and in both areas (2/46; 4.3%). In term of sampling strategy, 33 studies derived from hospital data, 12 studies from population data and only one study from the combination of both hospital and population data.
In term of malaria diagnosis method, 28 studies utilised both microscopy and PCR but in 18 of them, the samples tested for PCR were chosen from the microscopy positive cases for malaria species confirmation. Of those 18 studies tested for PCR, ten studies (7-9, 35-37, 40, 53, 55, 59) used all the microscopy positive cases, six studies (5, 32, 46, 48, 50, 51) used more than 70% of the cases, one study (47) used for 2.5% randomly selected cases, and one study (38) used for 19% of microscopy positive cases for either two Plasmodium spp. namely P. malariae or P. knowlesi. Furthermore, of the 28 studies, ten studies (12, 24, 25, 28, 31, 34, 41, 52, 61, 63) conducted PCR in the whole samples regardless of the microscopy results in order to trace the sub-microscopic infections. In addition, 18 studies utilised only one method of detection for malaria. Among them, 15 studies (33, 42, 43, 45, 49, 54, 56-58, 60, 62, 64-67) and three studies (27, 39, 44) were using microscopy and PCR, respectively.
In overall, the median sample size for cases were 308 (range 31-18993) for microscopy and 261 (range 47-4257) for PCR. Most microscopy measurements used Giemsa-stained thick and thin smears (n=38), while the remaining studies (n=8) used only thick smear. In term of PCR method, majority of studies used conventional nested PCR (n=22), followed by three studies used multiplex PCR and two studies used real-time PCR. Other studies (n=4) reported combination of different PCR methods; conventional/multiplex/real-time PCR, loop-mediated isothermal amplification (LAMP) assays (37), conventional/multiplex PCR assays (28), conventional/real-time PCR assays (12), and multiplex/real-time PCR assays (27).
Trend of malaria cases by species
The cumulative number and trend of PCR confirmed malaria species by year of sampling is show in Figure 2. Overall, detection of Plasmodium species was highest in 2006 for P. vivax and P. falciparum. The study also revealed that P. knowlesi cases rose steadily throughout the year and peaked in 2017 (5). The trend of malaria cases by PCR was low with less than 30 cases reported in 2017 and no mixed infection reported.
Sensitivity and specificity of detection methods
The performance of microscopy and PCR in detecting Plasmodium spp. are shown in Figure 4. In total, 21 studies were included; 18 studies were undertaken in Malaysia Borneo (5, 7, 24, 25, 28, 32, 38, 40, 41, 46-48, 51, 53, 55, 59, 61, 63), one study was in Peninsular Malaysia (34) and two studies were in both Peninsular Malaysia and Malaysia Borneo (8, 9). Overall, the summary estimate of sensitivity by microscopy was highest for P. knowlesi (35% [95% CI: 34 – 36]), followed by P. malariae (25% [95% CI: 24 – 26]) and P. vivax (14% [95% CI: 14 – 15]), and lowest for P. falciparum (11% [95% CI: 10 - 11]). Nevertheless, the summary estimate of specificity by microscopy was similar in all species ranged 40% - 43%.
Similar to microscopy, the summary estimate of sensitivity by PCR was highest for P. knowlesi with 56% (95% CI: 55 – 57). Whereas, the summary estimate of sensitivity by PCR for P. vivax, P. falciparum and P. malariae were 14% (95% CI: 13 – 15), 11% (95% CI: 10 – 11) and 1.6% (95% CI: 1.4 – 1.9), respectively. On the other hand, the specificities of summary estimate for species-specific were less than 35% with P. knowlesi, P. vivax, P. falciparum and P. malariae were 34.5% (95% CI: 34 – 35), 34% (95% CI: 33 – 35), 16.5% (95% CI: 16 – 17) and 2.3% (95% CI: 2 – 3), respectively.