Background
Malaria is a public health problem for many countries around the world. Some 3.2 billion people are at risk [
1,
2] and in 2015 there were 214 million cases leading to 438,000 deaths [
2]. Parasites from the genus
Plasmodium (
Plasmodium falciparum,
Plasmodium vivax,
Plasmodium malariae,
Plasmodium ovale and
Plasmodium knowlesi) are the aetiological agents for the disease [
1,
3]. Malaria is considered to be one of the severest public health problems in Colombia as more than 90% of cases occur in 7% of all Colombia’s municipalities, rural areas (85%) being the most affected [
4].
Plasmodium vivax represents about 70% of reported cases, whilst the rest are attributed almost exclusively to
P. falciparum [
5].
Plasmodium malariae infections usually do not surpass 1% [
6]. Accordingly, there has not been report of cases of malaria throughout 2015, as stated by the Colombian Public Health Surveillance System’s epidemiological bulletins [
7].
Microscope examination of thick blood smear (TBS) is the conventional gold standard for malaria in routine diagnosis, given its low cost and easy implementation in remote areas. Nevertheless, the amount of time spent on each sample, infrastructure maintenance, training and the ability of the personnel involved are components that heavily compromise the method’s sensitivity and the reproducibility of the results [
8‐
10]. TBS sensitivity is 10–30 parasites per μl of blood, this being around 0.001% of infected red blood cells. However, this technique requires trained personnel, particularly when parasitaemia is low or in cases of mixed infection [
11]. Molecular techniques relaying on polymerase chain reaction (PCR) and rapid diagnostic tests (RDTs) have been developed to cope with the drawbacks akin to microscopy examination. RDTs represent a cheap alternative to microscopy diagnosis. However, reports of cross-reactivity and less-than-desirable performances regarding mixed infections hinder its potential and, therefore, it has been considered inferior to microscopy in such scenarios [
12,
13]. According to some studies, HRP-2 malaria RDT and microscopy have been less sensitive than PCR and especially show limited detection thresholds in situations with low parasitaemia [
14‐
16]. Microscopy and RDTs cannot reliably detect low-density infections [
17].
Conversely, PCR-based diagnostics can identify infections below the threshold of detection for microscopy and RDTs [
17]. Such techniques are adaptable to individual emergency diagnosis, possess high sensitivity and specificity, and are capable of detecting low parasitaemia (about 5 parasites/μl of blood) [
18,
19]. Recently, PCR has been regarded as a new gold standard for malaria diagnosis [
17]. Prevalence by microscopic observation is underestimated by around 50.8% when compared to PCR [
20]. Similarly, many studies show a significant share of positive infections, which have been overlooked by microscopy standard diagnostics [
21‐
27]. Nested PCR (nPCR) shows higher sensibility than conventional and multiplex PCR diagnostics for malaria. Samples with <3000 parasites/µl of blood parasitaemia, which had positive results by the nPCR, were negative when analysed by conventional and multiplex approaches, using the same primer sets [
19,
28].
A seasonal outbreak of malaria cases has been observed since 2013 in the Colombian Amazon region [
29]; in 2015 such a rise was higher compared to previous years, doubling throughout 2016 [
30,
31]. Problems of public order, the irregularity of malaria surveillance campaigns and Plasmodium resistance to existing anti-malarial drugs may account for this increase in malarial burden, as has been previously stated [
32,
33]. Of the aforementioned factors, drug resistance is linked to accurate diagnosis, as misidentification of malaria species and degree of mixed infection inevitably lead to treatment with erroneous or incomplete medication schemes, exerting selection pressure on resistance phenotypes. This is particularly feasible for the Colombian Amazon region, a triple frontier with the Peruvian and Brazilian Amazon where the circulation of resistant
P. falciparum and
P. vivax phenotypes has been reported along borders [
33‐
35].
Molecular diagnosis of a sample of symptomatic patients during the previously mentioned outbreak surprisingly revealed high prevalence values for single and mixed
P. malariae infection according to PCR diagnostics [
36], thus confirming previous suspicions that
P. malariae prevalence may have been underestimated [
22,
23,
37]. The present study represents an evaluation of microscopy observation of TBS for malaria detection and species identification, comparing this to PCR diagnosis. This work also involves the diagnosis of mixed infections and the identification of un-expected
Plasmodium species, such as
P. malariae. The results of this work constitute a wider and more rigorous approach towards updating the epidemiological landscape and provide a critical perspective with regard to cost-effectiveness of current diagnosis in the Colombian Amazon trapezium, an area of unstable risk and endemic transmission.
Methods
Study population
The samples analysed in this study came from the municipalities of Leticia (41,326 population) and Puerto Nariño (8162) in Colombia’s Amazon department (data taken from Amazonas Department Development Plan 2012–2015) [
36]. The study area covered 53 settlements on the banks of the Amazon and Loretoyacu Rivers located on Colombia’s frontier with Brazil and Peru [
36].
Sample size calculation
This was a cross-sectional study. Sample size was calculated considering the estimated prevalence values from several studies performed in geographically similar populations [
22,
23,
38,
39], as well as a previous work performed in the Colombian Amazon, which was regarded as a pilot survey [
36]. A 1.5% prevalence was assumed as the largest sample size, taking into account all aspects to be evaluated. Accordingly, a 0.75% significance level and 95% confidence interval were chosen to avoid sample-size bias [
40]. A total of 989 samples were required to fulfil the minimum sample size calculated, consistent with the information obtained when using the EPIDAT 3.1 software (Dirección Xeral de Saude Pública, Organizacion Panamericana de la Salud, Galicia, Spain).
Sample collection
Inclusion criteria for obtaining samples from patients who were symptomatic for malaria were headache, fever during the previous 8 days, sweating, vomiting, and diarrhoea, and residing in the southern area of Colombia’s Amazon region (in and around Puerto-Nariño and Leticia). The blood samples used in this investigation were collected by personnel from the Fundación Instituto de Inmunología de Colombia (FIDIC) from July 2015 to April 2016. Each participant had a TBS test whilst blood spots on Flinders Technology Associates (FTA) cards were stored for subsequent detection of Plasmodium spp. by PCR.
Ethics, consent and permissions
Each participant signed an informed consent form after having received detailed information regarding the project’s objectives, and filled in a questionnaire regarding sociodemographic characteristics; the consent form and questionnaire for minors (under 18 years old) were filled in and signed by a parent or tutor and supervised by witnesses. This project was approved by the Universidad del Rosario’s School of Medicine and Health Sciences’ research ethics committee (resolution CEI-ABN026-000161).
Microscopy
Each TBS slide was stained with methylene blue phosphate and the cover slip was stained with 10% Giemsa (Merck, Darmstadt, Germany) for 15 min; it was then observed in immersion oil (Olympus CX21 microscope, Tokyo, Japan) for
Plasmodium spp. parasite forms [
41]. Parasite count was based on 200 leukocytes. A reference value of 8000 leukocytes was assumed for reporting parasitaemia per cu mm. A sample was considered negative when no parasite form was observed in more than 200 microscope fields observed [
42]. Diagnosis was performed by personnel trained in TBS preparation, reading and reporting.
Genomic DNA (gDNA) samples were extracted from each drop of blood collected on the FTA cards using a Pure Link Genomic DNA mini kit (Invitrogen), according to manufacturer’s specifications. The samples were eluted in a final volume of 50 µl buffer containing 10 mM Tris–HCl and 0.1 mM EDTA at pH 9.0. Extraction was verified by conventional PCR on all samples with primers directed towards a segment of the human
β-
globin gene to guarantee the presence of gDNA (Additional file
1: Table S1) [
43]. For each reaction 1 µl of genomic DNA was used as template.
Detecting Plasmodium spp. by PCR
Plasmodium spp. were identified by nested PCR in samples proving positive for human
β-
globin PCR. Specific primers for parasite 18S ribosomal subunit RNA (SSRNA) were used, following a previously described protocol with some modifications [
9] (Additional file
1: Table S1). The PCR mix contained 1× buffer, 3.8 mM MgCl
2, 1.4 mM dNTPs, 0.2 µM primers, 1 U/µl Taq polymerase (BIOLASE DNA Polymerase, Bioline), 2 µl of genomic DNA and molecular grade water (21 µl final volume). Amplification conditions were: 95 °C × 5 min, followed by 25 cycles at 94 °C × 1 min, 58 °C × 2 min and 72 °C for 2 min, with a final extension step at 72 °C × 5 min.
The corresponding PCR products were amplified again, using them as templates for a second PCR for type-specific identification of
Plasmodium spp. (
P. falciparum, P. vivax and
P. malariae) using specific primers for each species (Additional file
1: Table S1). PCR mix conditions for the second PCR were: 1× buffer, 4 mM MgCl
2, 2.5 mM dNTPs, 0.25 µM primers, 0.5 U/µl Taq polymerase and molecular grade water (20 µl final volume).
Two microlitre amplification product from the first PCR was used as template. Amplification conditions were 94 °C × 5 min, followed by 35 cycles at 94 °C × 30 s, 58 °C × 1 min and 72 °C × 4 min and a final extension cycle at 72 °C × 4 min.
gDNA samples from P. falciparum and P. vivax species were used as positive controls. Regarding P. malariae, a pGem-T plasmid (Promega) with the fragment of interest cloned within was used. Ultra-pure distilled water (GIBCO) was used as negative control. All products were analysed by horizontal electrophoresis (100 V, 30 min) on 2% agarose gels stained with SYBR safe (Invitrogen) and visualized on a MiniBIS Pro image analyser (DNR Bio-Imaging Systems).
Sequencing mixed infections
Given the high prevalence found for co-infection by Plasmodium spp., 30 samples were randomly selected for sequencing by an ABI-3730 XL sequencer (Macrogen, Seoul, South Korea) to confirm such mixed infections.
Statistical analysis
STATA software (Stata 12.0, Statacorp, Texas, USA) was used for obtaining descriptive statistics and determining raw values of molecular diagnosis’ performance indicators, such as sensitivity, specificity, predictive values, and related operating characteristics. Respective calculations were done regarding TBS diagnosis as a reference test. Performance indicator values have been corrected for imperfect gold standard using EPIDAT 3.1 software (Dirección Xeral de Saude Pública, Organizacion Panamericana de la Salud, Galicia, Spain), bearing previously reported sensitivity and specificity values for TBS in mind, based on other diagnostic techniques [
38,
44].
Discussion
For nearly 50 years in malaria-endemic areas in Colombia, diagnosis has been made by microscope observation of Giemsa-stained TBS [
34]. The prevalence values given by TBS in the present outbreak agree with those reported in previous independent studies and by the Colombian Public Health Surveillance System; in such surveys
P. vivax represented 70% of infection, whilst the remaining 30% were attributable almost exclusively to
P. falciparum [
5,
7,
36]. Likewise,
P. malariae was regarded as sporadic, having lower than 5% prevalence [
45]. Nevertheless, molecular diagnostics provided a very different epidemiological landscape, where
P. malariae was relevant regarding both single and mixed infections. Such prevalence values agreed with what had been observed for populations from geographically related regions of the Amazon region where this diagnostic test has been used [
22,
23,
36,
46]. The dramatic differences between both diagnostic tests feasibly highlighted the characteristic drawbacks of TBS: its reliance on observable parasitaemia and microscopist experience for high sensitivity and specificity, in addition to involving a risk of underestimating parasitaemia, reporting false negatives and committing errors in the identification of infecting species [
15]. Consequently, such results question the usefulness of TBS when retrieving epidemiological information related to sudden outbreaks in malaria-endemic areas, despite its well-known low cost and easy implementation.
Surprisingly, nested PCR was the only diagnostic test capable of identifying
P. malariae infection. The corresponding samples were in turn diagnosed by TBS as negative or simple infection caused by either
P. vivax or
P. falciparum, the predominant and regular species in the target region. Lack of quartan malaria detection by microscopy may have been related to TBS limitations per se as
P. malariae is characterized by sustaining low infection rates and low parasitaemia [
47,
48]. Similarly, the common loss of cells’ distinctive characteristics in samples treated for TBS can also account for overlooking
P. malariae infection, given that it hampers accurate species identification [
15,
22,
23,
27,
48].
Plasmodium malariae maximum parasite counts are usually low compared to those in patients infected with
P. falciparum or
P. vivax due to its longer developmental cycle (72 h for
P. malariae versus 48 h for
P. vivax and
P. falciparum), lower number of merozoites produced per erythrocyte cycle, and its preference for developing in older erythrocytes; the combination of the foregoing is a trigger for the earlier development of an immune response by a human host [
49].
The high share of sub-microscopic infections due to
P. malariae reported in this work raises important questions about how individuals became infected in the first place and how long they have been bearing quartan malaria infection. The latter is relevant considering that this parasite’s blood stage persists for extremely long periods; it is often believed that it lasts for the whole life of a human host [
49]. The former is important as populations in Colombia’s Amazon region co-exist with New World primates which could be a possible natural reservoir for
P. malariae due to their striking resemblance to the zoonotic parasite
Plasmodium brasilianum [
36,
50,
51], which is now commonly thought to be an anthroponosis from
P. malariae parasites [
51].
As parasite exchange between monkeys and humans is a well documented phenomenon, the risk of primate reservoirs acting as source for outbreaks in the human population is latent. Documented chimpanzee infection with human
P. malariae is thought to contribute to continuous parasite exchanges in Africa [
52]. The preceding, combined with the characteristic low parasitaemia and long-lasting persistence of this parasite, could provide an explanation for the outbreak observed in terms of recrudescence and imported infections from nearby areas. The imported infections should be carefully considered in this particular case, taking into account that the Colombian Amazon region shares a border with both Brazil and Peru [
33‐
35].
Many unnoticed quartan malaria parasites in mixed infections have been reported as only single infections by TBS. Such difference has usually been attributed to the fact that mixed-species infection generally implies the predominance of one species, the others having very few parasite forms [
36,
53]; this gives an advantage to PCR as TBS has higher detection thresholds [
54].
Erroneous identification of
P. malariae is frequently due to haemolysis during Giemsa staining, added to morphological similarity amongst
Plasmodium spp. during their growth stages [
22,
23]. Particularly regarding
P. malariae, this alters ring forms thus limiting routine diagnosis [
55]. It is normally difficult to distinguish between
P. malariae and
P. falciparum parasite forms; nevertheless, in studies in South America,
P. malariae is usually confused with
P. vivax [
22,
23,
48]. This could also account for the large amount of mixed
P. vivax/P. malariae and
P. falciparum/P. malariae infections which, in the present study, were rather classed as simple infections caused by just
P. vivax or
P. falciparum by TBS. Regardless of TBS’ inherent limitations, microscopists might have had insufficient training in recognition of
P. malariae parasite forms. Equally, the personnel would benefit from the use of the microscopy observation of thin blood smear more extensively, given that in Colombia and other malaria-endemic regions it is used only as confirmatory analysis [
22,
27]. Although thin blood smear has lower sensitivity, it better preserves the morphology of the parasite’s cells [
15].
In Colombia, the prevalence of sub-microscopic infections has been observed to vary from 3 to 20%, having greater occurrence in regions where
P. vivax is the predominant species [
56]. Such a figure constitutes a worrying factor when the relationship between malaria diagnosis and treatment are taken into consideration. One possible scenario relates to favouring
Plasmodium-resistant phenotypes due to treatment failure linked to improper diagnosis. This is particularly plausible for the
P. falciparum/P. vivax mixed infections reported in this work, given that in Colombia, amodiaquine, followed by sulfadoxine–pyrimethamine constitute the first and second lines of treatment for falciparum malaria, respectively, whilst malaria caused by
P. vivax is usually treated with chloroquine and primaquine schemes [
34]. Therefore, underestimation of
P. vivax infections might allow the thriving of vivax malaria phenotypes due to an incomplete elimination of liver hypnozoites, whilst underestimating
P. falciparum infections might lead to treatment failure given its already reported resistance to chloroquine.
The present study was aimed at comparing the performance of the TBS technique to PCR diagnosis for detecting malaria in populations from the Colombia’s Amazon region. It was found that molecular diagnosis had a high sensitivity for detecting malaria in general and for malaria caused by
P. vivax, as well as having a high NPV within the study population. These results coincided with those from previous work reporting 75–98% sensitivity for PCR regarding the identification of
Plasmodium spp. [
38,
39,
57,
58], together with 98–100% estimations for detecting
P. vivax [
38,
59,
60]. Similarly, PCR estimated higher prevalence values for the species evaluated and for certain types of co-infection, such increases having been observed in previous studies for both simple and mixed infections [
22,
23,
27,
47,
48]. This result highlights PCR’s potential for confirming a clinical suspicion of malaria, in spite of being expensive and not available in health centres having limited resources [
61]. This study has thus confirmed the importance of PCR-based diagnosis as the norm in future studies concerning
P. malariae epidemiology [
19,
36,
48,
53].
Authors’ contributions
CHN conceived and designed the experiments, CHN, JRC and PACA performed the experiments, CHN, JRC, PACA, RS, MC, CARC, TQ, LSS, MEP, and MAP analysed the data; CHN, JRC, MC, MEP, and MAP wrote the paper. All authors read and approved the final manuscript.