In preparation of major molecular epidemiological field studies in PNG essential parasite detection techniques were compared under conditions of a laboratory close to the field site and located in a malaria endemic country. The diagnostic requirements were: (i) good performance in the detection of mixed species infections, as all four species concurrently occur in PNG, (ii) recognition of P. malariae and P. ovale variants present in the study area, (iii) high through put capacity and robustness of assay, (iv) quantitative results and (v) reasonable costs. The qPCR assay described here was implemented and validated at the PNG-IMR site in Madang, demonstrating the feasibility of applying state of the art techniques in this context. In the meantime the qPCR assay is routinely implemented for molecular diagnosis in large scale epidemiologic studies at IMR.
As part of test validation in the field, the performance of this qPCR assay for Plasmodium species discrimination was compared to two other PCR-based assays (nPCR and LDR) and to LM. Traditionally, test outcomes for different assays are compared to an established 'gold-standard' in order to calculate sensitivity and specificity estimates and to evaluate the performance of newly developed tests. The classical 'gold standard' for malaria diagnosis has been LM , however, with the appearance of new molecular diagnostic tools, LM has become less suitable for this purpose due to its lower sensitivity than molecular methods . Even though the nPCR developed by Snounou et al  has been extensively used as 'gold-standard' for molecular diagnosis [25, 26], the concept of using a 'gold-standard' for the evaluation of new assays is being questioned by various authors, which alternatively propose the use of 'non-gold standard' approaches [27, 28].
The agreement between qPCR and the other techniques was substantial for P. falciparum, but only moderate for P. vivax, P. malariae and P. ovale. In particular, the agreement between qPCR and nPCR for P. falciparum detection was almost perfect. The lower agreement between PCR-LDR and nPCR, together with the higher prevalence shown by PCR-LDR (47.1% compared to 40.9% by qPCR and 43.8% by nPCR), may indicate false positive results by LDR. This is supported by our pairwise analysis and the agreement of two independent PCR based assays, namely nPCR and qPCR. However, in absence of a suitable diagnostic 'gold standard', it remains unclear if those 33 samples positive for LDR but negative by the two alternative molecular methods, represent a greater sensitivity of LDR or simply false positives. This issue cannot be easily resolved in a study involving 'unknown' samples from the field, potentially infected by four different Plasmodium species.
P. vivax prevalence was higher by nPCR than by both, qPCR and PCR-LDR (73.2% by nPCR, 65.7% by qPCR and 67.5% by PCR-LDR). This again could reflect false postitives by nPCR or lower sensitivity by both other molecular methods. Our observations in qPCR validation using plasmid template suggested that qPCR of P. vivax is lightly compromised by performing a duplex Pf/Pv reaction. nPCR involves a very high number of cycles (55 cycles by nPCR versus 45 cycles by qPCR and 35 cycles by LDR), and therefore is expected to show maximal sensitivity. Despite measurements taken over 45 cycles in qPCR, we followed the consensus rule for considering a sample positive, i.e. a Ct value < 40 . In our samples this led to the loss of 9 samples with Ct values for P. vivax between 40 and 43.6 cycles, which otherwise would have increased the sensitivity of the assay. Further analysis was performed on samples with discrepant results for P. vivax (P. vivax negative samples by qPCR and positive by nPCR). Most of these samples were mixed infections by nPCR and harboured P. falciparum with more than 10,000 copies/μl. Thus competition for amplification at the beginning of the PCR due to P. falciparum high densities may be precluding P. vivax detection . 14/16 of the remaining samples were also negative by LM. Most likely these very low-grade P. vivax infections were missed. The scarcity of the template in case of a very low parasite density is expected to lead to imperfect detection. Prevalence for P. malariae and P. ovale were low with significant differences between assays, even though the agreement between pairwise compaired methods was moderate. Higher prevalence for P. malariae detection by LDR is likely to occur as a result of false positive results, probably occurring due to high background noise of the P. malariae probe used in the assay. Low detection of P. ovale by nPCR (3.8%) is due to the use of a primer pair with sub-optimal amplification of P. ovale sequences present in the study area. Finally, LM measured the lowest prevalence for all four Plasmodium species.
The major advantage of qPCR over the other compared molecular techniques was the quantification of parasite densities. Parasite densities shown as copies of 18 S rRNA template/μL were quantified by converting the threshold cycle (Ct) into template copy number by using the standard curves. When correlating quantification by qPCR with LM counts in samples where both techniques showed positive results, a high correlation for P. falciparum (R2 = 0.8253) and a lower correlation for P. vivax (R2 = 0.5049) was found. But for P. vivax this correlation of parasite densities by qPCR and LM increased when only single infections were taken into account. Therefore, our results suggest a variation in the detection limit in both methods, due to overlooking P. vivax in case of an overwhelming P. falciparum infection. Difficulties in identifying P. vivax by LM arise when this parasite is found at low densities and in mixed infections. The high P. falciparum densities found in the samples identified as mixed infection by qPCR (> 10000 target copies/μl) further supports this explanation. The correlation for P. malariae and P. ovale could not be analysed due to poor detection of both species by LM.
The qPCR assay was found optimal for both tasks, detection of all four Plasmodium species and quantification. The latter could only be analyzed for P. falciparum and P. vivax. Overall qPCR shows substantial agreement with other molecular techniques for the detecting prevalence of P. falciparum and P. vivax, while moderate agreement was observed for P. malariae and P. ovale. It is clear, that sensitivity of our qPCR assay can be increased by simply performing independent reactions of each Plasmodium species. However, this would substantially increase costs. Limiting factors, such as duplex assays, need to be balanced against costs or work load. The specific research objectives of a particular study should guide the choice of experimental procedures.
Overall, the superior performance of PCR based methodologies over LM has been clearly demonstrated by these results and others. In a recent study conducted in Benin, a high number of children (between 27% and 44%) aged 5 or above, who initially had negative RDT tests (most also with negative blood slides), were later found to be infected with P. falciparum using PCR . These undetected submicroscopic infections have an enormous impact for malaria transmission in endemic areas. In a time where malaria erradication has become the primary goal of malaria agendas, the accurate estimation of the burden of malaria infection is imperative to control transmission.