Background
Globally, there are 85.3 million pregnancies occurring annually in
Plasmodium falciparum-endemic areas [
1,
2]. Malaria infection in pregnancy has significant, direct, negative health effects on the mother and the neonate, such as severe maternal anaemia and related maternal mortality, low birth weight from both preterm delivery and intra-uterine growth restriction, and postnatal mortality [
3]. A key risk factor for these complications is placental malaria, defined as the sequestration of
P. falciparum-infected erythrocytes in the placenta. The ability to accurately identify placental malaria before and at delivery is critical for both research and programmatic purposes.
It is often difficult to diagnose malaria in pregnant women because parasitized erythrocytes sequestered in the placenta are often absent from the peripheral circulation. Microscopy of peripheral and placental blood, polymerase chain reaction (PCR), and rapid diagnostic tests have been used but vary in sensitivity and specificity. Traditionally, histopathology has been the gold standard for the diagnosis of placental malaria (reviewed in [
4,
5]). The quality of histopathology is very dependent on obtaining high-quality specimens, since formalin pigment has similar optical properties to malaria pigment and can lead to misclassification [
6]. Unfortunately, residual formalin pigment is common in under-resourced areas due to the use of inexpensive, low-quality fixative.
Here PCR, immunohistochemistry (IHC) and histopathology are compared as diagnostics for
falciparum malaria using dried blood spots and formalin-fixed placental tissue from a longitudinal cohort study conducted in Kinshasa, Democratic Republic of Congo (DRC) in 2005–2006 [
7,
8]. Because the placental tissue had abundant formalin pigment, two separate microscopists examined them. The aim of this analysis is to examine the concordances of the different tests using both the classical contingency table [
9] and the latent class analysis (LCA) approaches [
10]. LCA can be used to estimate the sensitivity and specificity for independent diagnostic tests, when all tests are regarded as imperfect.
Discussion
Under ideal circumstances, placental histopathology is considered the gold standard for diagnosis of placental malaria [
4,
5]. However, in this study, both histopathologists noted an abundance of formalin pigment, which makes accurate diagnosis difficult. Because of this, the correlation between the two blinded pathologists was poor. Here, alternative diagnostic approaches were evaluated, including IHC and PCR. Because of the absence of a gold standard, sensitivity and specificity were evaluated in several different ways.
The extent of formalin pigment in this cohort was a surprise considering the use of buffered formalin and careful storage conditions. A possibility is that low-quality formalin was used in the first step of the hospital processor during the paraffin embedding process (formalin is often used to continue fixation of recently grossed specimens). Researchers should be aware of this potential pitfall. Other ways to avoid formalin pigment are to use fresh neutral buffered formalin, small biopsy size with >10:1 ratio of formalin to sample in container, prevent desiccation, and prompt processing or prompt transfer to 70 % ethanol after 24–48 h of fixation.
Correlation between methods was first measured using Cohen’s kappa [
18,
19]. Using this method, the findings of the two microscopists did not correlate at all. There was some correlation between one of the pathologists (P2), and both IHC and PCR. Similarly, there was some correlation between IHC and PCR. This is consistent with the finding that none of the tests had sensitivities >50 % compared to IHC or PCR.
Some of the discordances are expected, since different tests have different operator characteristics. For example, the 50 % sensitivity of IHC compared to PCR could have been due to the fact that PCR detects parasite densities that are lower than the threshold for microscopic detection. Alternately, the discordance could have also been the result of timing: the PCRs were done at the last antenatal visit and could have cleared by the time of delivery. On the other hand, PCR only had a 29 % sensitivity compared to IHC. This could be because parasitized erythrocytes sequester in the placenta [
20]; thus detectable placental infections are often associated with undetectable peripheral infections.
The current study was carried out in the context of Kinshasa, DR Congo in 2005–6. As a point of comparison, in 2009–2011, the prevalence of malaria parasitaemia in children six to 59 months old in Kinshasa was an average of 6.4 %, with higher levels reported in less urban areas [
21]. The prevalence of malaria in the current study population varied from 21 % prevalence in early pregnancy to 9 % at delivery, which is low compared to previous studies of malaria in pregnancy in endemic areas [
3,
22]. The relatively low proportion of positive samples at delivery can largely be explained by two-dose IPTp and treatment of microscopic malaria infection throughout the study.
Because it is ultimately important to be able to compare different diagnostics, LCA was used to compare the four tests in an agnostic fashion. LCA has previously been used to compare PCR and histopathology among smear-negative pregnant women in Malawi [
23]. LCA does not assume a gold standard but constructs one based on the results of the individual tests [
10]. LCA assumes that diagnostic tests are conditionally independent, given the true disease status of the subject. Under the assumption of the model, the joint likelihood for the diagnostic tests can be developed and parameters are estimated by maximum likelihood using an expectation-maximization type procedure [
17]. While LCA assumes test independence, it is robust to violations of this assumption [
24]. Based on the LCA, all tests were relatively specific (>95 %) but IHC and PCR had the highest sensitivities (68 and 56 %, respectively). Since IHC can detect parasites when they sequester in the placenta but do not circulate, while PCR detects extremely low parasitaemias, the combination of both methods might provide a means of detecting the greatest number of infections.
Authors’ contributions
SHL, JMMK and AKT were responsible for acquiring the samples and data in Kinshasa, DRC. JMMK was also responsible for histopathology realized in Kinshasa. AJB, RS and DJS performed the immunohistochemistry. SMT did the PCRs. AM and SJR evaluated the placental histopathology. YL, JBG and SRM did the data analyses and wrote the manuscript. All authors read and approved the final manuscript.
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (
http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (
http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.