Performance of Zika Assays in the Context of Toxoplasma gondii, Parvovirus B19, Rubella Virus, and Cytomegalovirus (TORCH) Diagnostic Assays

Infections during pregnancy that may cause congenital abnormalities have been recognized for decades, but their diagnosis is challenging. This was again illustrated with the emergence of Zika virus (ZIKV), highlighting the inherent difficulties in estimating the extent of pre- and postnatal ZIKV complications because of the difficulties in establishing definitive diagnoses.

SUMMARY

Infections during pregnancy that may cause congenital abnormalities have been recognized for decades, but their diagnosis is challenging. This was again illustrated with the emergence of Zika virus (ZIKV), highlighting the inherent difficulties in estimating the extent of pre- and postnatal ZIKV complications because of the difficulties in establishing definitive diagnoses. We reviewed the epidemiology, infection kinetics, and diagnostic methods used for Toxoplasma gondii, parvovirus B19, rubella virus, and cytomegalovirus (TORCH) infections and compared the results with current knowledge of ZIKV diagnostic assays to provide a basis for the inclusion of ZIKV in the TORCH complex evaluations. Similarities between TORCH pathogens and ZIKV support inclusion of ZIKV as an emerging TORCH infection. Our review evaluates the diagnostic performance of various TORCH diagnostic assays for maternal screening, fetal screening, and neonatal screening. We show that the sensitivity, specificity, and positive and negative predictive value of TORCH complex pathogens are widely variable, stressing the importance of confirmatory testing and the need for novel techniques for earlier and accurate diagnosis of maternal and congenital infections. In this context it is also important to acknowledge different needs and access to care for different geographic and resource settings.

INTRODUCTION

The present Zika virus (ZIKV) epidemic was first noted through an alert from the Brazilian health authorities notifying the World Health Organization (WHO) of an illness characterized by skin rash in the northeastern states and subsequently signaling an almost 20-fold increase in the incidence of microcephaly in newborns coinciding with rapid spread of ZIKV after incursion into the continent (1–4). Until then, ZIKV infection was generally assumed to be associated with mild and transient disease, estimated to be asymptomatic in approximately 80% of the cases (5). Hence, it is likely that the infection is underdiagnosed or underreported in a setting where disease is endemic (6).

The association of ZIKV infection with congenital neurological disease has since then been the subject of numerous publications. First, establishment of a causal association with neuropathological processes came from a study showing widespread ZIKV infection in the brain of a fetus from a pregnancy that had been terminated due to severe fetal malformation (7). Further evidence for the association came from larger case series, retrospective analysis of notification data from regions with prior outbreaks (8), and replication of the syndrome in animal models (9–13). Although there is general agreement on such an association, many uncertainties remain with regard to the actual risk of fetal infection during pregnancy (14). A systematic review estimated the prevalence of microcephaly at 2.3% (95% confidence interval [CI], 1.0 to 5.3%) (15) though estimates range widely. Little is known about the risk of complications in relation to timing of maternal infection (first trimester versus later exposures), prior (flavivirus) exposure, or the rates of transplacental transmission, fetal infection, and congenital disease once infection is present (Fig. 1). Two recent reports suggest a decreasing risk over the course of pregnancy (16, 17). A recent European study showed an overall congenital anomaly prevalence of any (nongenetic) cause of 1.5/100 total births, but for microcephaly interpretation was hampered by differences in diagnostic criteria (18).

FIG 1.

Seroprevalence, maternal infection risk, fetal transmission risk, and congenital infection risk of the following selected infections: TOXO, PB19, rubella, CMV, and ZIKV. Seroprevalence at childbearing age is shown. IR/year, maternal annual infection rate; FTR, fetal transmission rate; CIR, congenital infection rate, determined as the number of congenital infections per 1,000 live births (TOXO) or per 1,000 pregnancies (RV, CMV, PB19, and ZIKV).

At present, ZIKV diagnostic algorithms are based on the use of reverse transcriptase PCR (RT-PCR) for virus detection and/or serological determination of pathogen-specific IgM and IgG antibodies, supplemented with virus neutralization assays if available. All of the assays have benefits but also known limitations, challenging interpretation at different stages of pregnancy, particularly in relation to the wide diversity of flavivirus background exposures in the regions where ZIKV circulates. Virus genome detection by RT-PCR is considered confirmatory but has a very short detection window as ZIKV viremia is of short duration. Virus may persist for longer periods in other body fluids, with reported persistence up to 120 days for semen (19). In addition, pregnant women may experience prolonged viremia (20, 21), with reported (transient) presence of ZIKV in fetal blood and amniotic fluid (21). Antibody-based testing is severely hampered by cross-reactivity with antibodies from prior flavivirus exposures. Other infections during pregnancy are associated with congenital and subsequent neonatal disease, sometimes referred to as TORCH infections (Toxoplasma gondii [TOXO], other [e.g., varicella zoster and parvovirus B19 (PB19)], rubella virus [RV], cytomegalovirus [CMV], and herpes simplex virus [HSV] with or without syphilis) (22). Diagnosis of fetal infection and linking fetal infections to clinical outcomes require knowledge of infection kinetics, including timing and differentiation of primary from nonprimary infection (i.e., reactivation or reinfection), of maternal and fetal immune responses in relation to pathology, and of the availability of biomarkers predictive of vertical transmission and presence and/or severity of fetal abnormalities (23–25). For instance, TOXO and CMV cause persistent or latent infections, whereas RV, PB19, and ZIKV infection are thought to be primarily self-limiting. Immunocompetent pregnant women with previous infection with TOXO are considered not at risk for congenital abnormalities, whereas for CMV primary as well as subsequent infections are associated with congenital infection and abnormalities, albeit with a lower attributable risk. The recent ZIKV epidemic and its possible association with microcephaly have initiated the discussion to include ZIKV as a novel TORCH pathogen (26).

Although the described maternal infections are important causes of fetal and neonatal morbidity and mortality on a global scale, the overall contribution to fetal and congenital diseases is limited (27–29) due to cumulative low and/or limited a priori risks of maternal infection, vertical transmission, and subsequent congenital infection (Fig. 1).

Generally, the risk of congenital disease following maternal infection is linked to primary infection and timing in pregnancy infection, which is pathogen dependent (25, 30–37). Nonprimary maternal infections may result in fetal transmission, but only in the case of CMV does this contribute significantly to congenital disease (38).

The low attributive risk of TORCH and ZIKV infections to congenital disease has consequences for diagnostic accuracy and the ability to provide information relevant for clinical decision-making. This is further complicated by the high proportion of asymptomatic maternal infections (6, 35, 39–41), challenging timely detection of fetal infection and early neonatal congenital disease, which may remain asymptomatic for years (40, 42). Early diagnosis of fetal disease risk in pregnancy, however, is important particularly when early therapeutic management is available, e.g., for TOXO infection (43). Consequently, diagnostic algorithms should reliably detect maternal infection in a timely fashion, determine (risk of) vertical transmission, and establish or exclude congenital infection. This review assesses diagnostic methods presently used for TORCH infections, their correlation with congenital and/or neonatal disease, and their predictive value in prenatal screening; it will document gaps in methods used; and it will draw implications for diagnostic algorithm development in novel or (re)emerging infections such as Zika virus.

EPIDEMIOLOGY, INFECTION, AND TRANSMISSION RISKS OF SELECTED TORCH PATHOGENS

The risk of infection during pregnancy varies by pathogen and depends upon geographic region, prevalence in the population, and preventive (vaccination) practices.

The seroprevalence of Toxoplasma gondii among women of childbearing age shows a broad range, from <2% in a large Chinese cohort (44) up to 75% in Brazil (45), with a mean estimate of around 40% (46–48). Similar, broad ranges in seroprevalence between 30 and 72% are reported for PB19 and CMV (49–56) although for CMV seroprevalences up to 100% are also reported (57). RV seroprevalence estimates depend on (differences in) vaccination practices (58–60). ZIKV seroprevalence has a geographic distribution, varying from <2% in travelers returning from areas of endemicity or blood donors in settings of nonendemicity (61–63) to up to 39% in healthy individuals living in areas of endemicity (64, 65) and increasing to >60% following outbreaks (66, 67). These wide ranges in background seroprevalence affect the likelihood of primary infection during pregnancy, as well as the interpretation of diagnostic assays, and need to be taken into account in the development of diagnostic algorithms.

To assess the clinical impact of exposure to TORCH pathogens and ZIKV during pregnancy, it is important to consider maternal infection risk (MIR), fetal transmission risk (FTR), and congenital infection risk (CIR) for each pathogen (Fig. 1). MIR estimates, defined as the annual infection rate for selected TORCH infections, range from 0.1 to 0.6% for TOXO (31) to 2 to 7% for PB19 and CMV (68–73), with an epidemic rise up to 10% in PB19. Reliable data for RV MIR is lacking in an elimination setting, but the annual incidence is estimated at 1.3/100,000 pregnancies in the general population (74). A 6.4% IgG seroconversion rate was reported in women with nonimmune RV titers prior to pregnancy (75). Although efforts have been made to calculate the MIR for ZIKV, reliable data are still lacking due to several factors, including diagnostic limitations (e.g., cross-reactivity) (76, 77) and differences and rapid changes in epidemiology.

Fetal transmission risk (FTR), defined as the proportion of transplacental transmission following (primary) infection during pregnancy, is also pathogen dependent and is linked to the timing of infection during pregnancy. The FTR may increase (TOXO, PB19, and CMV) or decrease (RV) during the pregnancy period, with a variable mean FTR estimated to range between 24 and 80% for these pathogens (25, 33, 42, 78, 79). For ZIKV, the FTR is thought to be highest during the first trimester, but more data are needed (16, 17). Perinatal transmission has also been reported for ZIKV (80).

Despite a high FTR, the congenital infection risk (CIR), defined as the number of congenital infections per 1,000 live births (or number of fetal deaths/hydrops fetalis per 1,000 infected fetuses in PB19), is low, ranging from <1/100,000 pregnancies (RV) up to 0.1/1,000 to 20/1,000 pregnancies for TOXO, PB19, and CMV infection (33, 34, 81–89). Overall congenital CMV infection is most prevalent in the developed world (5/1,000 to 20/1,000 live births) (30, 90), followed by infection with TOXO (0.1/1,000 to 5/1,000 live births) (31, 46), and RV (annual incidence of 0.4/100,000) (81). Parvovirus B19 infection is associated with hydrops fetalis (82–84) and fetal death, with an estimated annual incidence of <4/1,000 fetuses (85). Recent studies estimated the contribution of symptomatic ZIKV during pregnancy to ZIKV-associated congenital disease at 7% (17) and evidence of acute infection in pregnancy at less than 4% (77). How this translates to the overall contribution of ZIKV to, e.g., congenital microcephaly prevalence depends on the baseline risks, and these are uncertain (18, 91). A retrospective analysis in French Polynesia estimated a risk of microcephaly associated with ZIKV infection at 9.5/1,000 pregnancies, with an overall risk of microcephaly of 0.2/1,000 neonates (8).

The CIR does not necessarily follow the FTR, with the highest CIR in the first trimester for TOXO, RV, and CMV (25, 34, 35, 42) and the highest CIR in the second trimester for PB19 (36, 37). In addition, for ZIKV, the highest risk is reported in the first trimester (8, 17). Overall, CIR is limited to primary infection, except for CMV, where reinfection or reactivation contributes mainly to congenital CMV disease burden (56, 92). Latent or chronic TOXO infection does not exclude transmission but does not result in CIR in immunocompetent pregnant women (93, 94). In general, the low attributive risk of the reported infections to overall prevalence of congenital disease (27–29) impacts the performance of diagnostic assays.

This implies that the low a priori attributive risk of TORCH and ZIKV to congenital infections needs to be included in every step excluding or confirming maternal, fetal, and/or congenital infection.

MATERNAL DIAGNOSTIC TESTING

Infection Kinetics

Interpretation of diagnostic testing during pregnancy requires knowledge on infection kinetics. These are defined by many factors, including the presence and duration of symptoms, duration of pathogen presence in different body fluids, loads, timing of development of specific antibodies, background antibody prevalence, and relationship between these parameters. Molecular detection in early symptomatic infection is generally considered proof of acute primary infection (17, 40, 59, 95) although this is not true for each pathogen, i.e., CMV. Primary infections typically show IgM and IgG development, determined with serological assays with or without confirmatory testing (96, 97). Reported antibody kinetics differ between selected pathogens. TOXO IgM seroconversion occurs relatively late, between 15 and 30 days (98, 99), while an early IgM rise is observed for PB19 infection, i.e., toward the end of the first week of infection and coinciding with peak viremia (56), and for infection with RV, whose IgM rises within 5 days after rash onset (100). CMV IgM may become detectable between 0 and 3 weeks (101), with peak IgM observed between 1 and 3 months (102). IgM antibodies against ZIKV show an early rise and can be first detected within the first week after clinical symptoms, while IgG antibodies can be detected within the first 2 weeks (103). PB19 IgM can persist up to 3 months following infection (104). ZIKV IgM can persist beyond 3 months (105). Long-term persistence of rubella IgM is reported following vaccination (106–108) due to the natural occurrence of nonspecific IgM (109, 110) and despite attempts to improve assays (111–114). Differentiation of acute infection from latent infection or reactivation/reinfection is important in TOXO and CMV infections as IgM/IgG may coincide, thereby making it difficult to diagnose primary infection if the first consultation yields an IgM/IgG-positive test result (115). In this case, confirmatory testing is needed, e.g., by avidity index (AI) or immunoblotting (IB), using presently available assays (47, 116, 117). Development of CMV-specific IgG with a negative sample collected earlier in pregnancy is considered proof of primary CMV infection; however, in the absence of routine screening this is usually not feasible (118).

Molecular Assay Performance and Limitations

Although PCR assay specificity is high in acute primary infection, the window of PCR positivity may be short, as shown for PB19 (95) (Fig. 2). Limited data from ZIKV showed a similar pattern. A recent external quality assessment (EQA) suggested similar more robust specificity but variable sensitivity between participating laboratories (119). Furthermore, most acute (primary) infections in pregnancy are asymptomatic, and the day of infection is unknown, precluding use of this gold standard test. In acute primary PB19 infection in pregnancy, high viral load is associated with an early positive IgM test (39, 56, 120, 121). Long-term, low-load DNA persistence is observed following PB19 infection (104, 122). In one study, the use of endonuclease treatment before molecular testing differentiated naked DNA persistence from true viremia (123). For ZIKV, rapid degradation of RNA was reported in urine samples (124). In acute maternal PB19 infection, the positive predictive value (PPV) of PB19 PCR is high, but at the time of fetal symptoms, the PPV of PB19 DNA detection in maternal blood is generally low as clinical symptoms in the fetus are usually observed when maternal viremia has ceased. This temporal relation has not been established for other primary infections such as TOXO (125, 126), or CMV (127). Viremia in pregnant women is associated with vertical transmission risk and increased CIR, but the relationship between maternal infection, FTR, and CIR is different for the selected pathogens (128–134). Currently, there is no obvious predictor for transmission risk. For instance, viral load does not differentiate transmitters from nontransmitters in CMV infection (129, 130, 135). Absence of a relationship with maternal disease severity or viral load was also recently described for congenital ZIKV infection (136). Low-viral-load positives may occasionally not show IgM seroconversion (RV and ZIKV) (137–139). Low assay sensitivity was suggested as one of the possible explanations (139). Genotype differences for individual pathogens may impact the sensitivity of assays (140–142), which is important in considering the use of assays in different regions.

FIG 2.

Routine maternal diagnostic methods: sensitivity and specificity, positive predictive value, and negative predictive value median point estimate ± 95% CI. For IgM and IgG data, plus and minus signs indicate allocation of “gray zone” results to the seropositive or seronegative results, respectively. NA, not available. References for the data for each virus are as follows: TOXO (n = 28), 48, 99, 113, 134, 143, 150, 153, 160–163, 165, 170, 171, 175, 177, 179, 183, 189, 195, 199–201, 204, 205, 209, 262, and 302; PB19 (n = 14), 29, 36, 53, 79, 84, 86, 121, 146, 176, 197, 211, 224, 303, and 304; RV (n = 9), 59, 106, 108, 137, 149, 151, 186, 204, and 303; CMV (n = 24), 23, 53, 92, 101, 112, 128, 145, 148, 155, 156, 159, 166, 169, 173, 178, 190, 196, 204, 206, 250, 265, 281, 303, and 305; ZIKV (n = 15), 184, 289, and 306–318.

Serological Assays and Performance

Primary diagnostic assays.

Generally, TORCH immunoassays report relatively high specificity for IgM-positive and IgG-positive samples or for IgG in IgM-negative samples (34, 108, 113, 143), but as the a priori likelihood of a maternal infection with TORCH pathogens is generally low, even a relatively low false positivity rate translates to a low PPV for all pathogens (including ZIKV) except PB19, stressing the need for confirmatory testing (144–147) (Fig. 2). Comparative studies of assays reporting relative performance data tend to overestimate sensitivity and specificity (108, 113, 143, 148). In view of the above, a positive IgM test result always requires confirmation with other assays (36, 101, 113, 149) and follow-up samples. More specific (recombinant) peptide-specific IgM assays may provide solutions, but their performance also needs to be fully evaluated (121, 150).

Confirmatory testing.

The use of confirmatory testing with avidity index measurements (AI), immunoblotting (IB), and virus neutralization testing (VNT) is not consistent between pathogens and also shows variable performance levels (Fig. 2). Testing for AI is common practice for TOXO and CMV diagnostics but not for PB19 (121) or RV (151). The rationale for avidity testing is that the avidity of antibodies increases with time, and a high AI correlates with infection in the more distant past (152–158). The sensitivity of the AI as a confirmatory test depends upon the initial screening platform used, as was shown in one study. A negative initial IgM screening is unlikely to be confirmed (145). IgM positivity combined with a low AI increases the sensitivity and PPV of the combined assays in diagnosing recent infection (159–162). In contrast, a low AI with positive IgG has a relatively high negative predictive value (NPV) (163, 164). A high AI plus positive IgG usually confirms past infection; however, for TOXO infection, AI maturation may never occur (165). A rapid increase in the AI in CMV infection was associated with the false exclusion of recent infection (166), with a higher FTR (167) and CIR (168). Therefore, exclusion of acute infection based on (high) AI requires a predefined time window (101, 114, 169–171), and the size of the window varies depending on pathogen, thresholds, and platforms (172). Different antigens, including recombinant antigens as targets for antibody responses, may improve AI assay performance (170).

Antibody test results may also be confirmed by IB (101, 173) or different complementary assays (174). Epitope-specific IgG IB used in TOXO and PB19 infection could confirm IgM/IgG measurements (175, 176). This is particularly relevant in equivocal outcomes (177) or in (false) negative results with high viral load (146) and may help timing of the infection by correlating IB with virus neutralization in CMV infection (178) or with the AI in TOXO infection (179).

An interesting application is the use of IB in TOXO infection for discrimination of maternal and neonatal antibody responses by comparing patterns of antibody binding to different proteins or peptides in the blood from mother and neonate (180). Limitations of IB include lack of standardization with variable concordance between assays, particularly in acute infection (181, 182), and different diagnostic accuracies of band patterns in the blots (175). The predictive value of IB depends upon the target, and IgM blots often have poorer predictive value than IgG blots (173, 183). Virus neutralization data are primarily available for CMV (96) but are also commonly used in confirmation of ZIKV infection (184), with generally high assay performance.

Limitations of Serodiagnostic Assays

Major limitations for all of the diagnostic methods described include interassay variability (98, 185), use of different cutoffs, differences in classifications of positives (185, 186), low agreement between AI index assays (187), and variability between platforms (114, 165, 171, 172, 188–191). Other serious concerns are the so-called “gray zone” classification, i.e., the area between the negative outcome and the positive outcome of a test (98, 114, 153, 163, 164, 166, 186, 187, 192, 193), which differs considerably between assays (194), and the lack of standardization of cutoff values for the same assay (183). How gray-zone results are interpreted has a profound impact on the sensitivity or specificity of a given assay. This is indicated in Fig. 2, where, for example, the category Tox+ denotes the assignment of the gray zone to the seropositive group and Tox– indicates assignment to the seronegative group (148, 195, 196). This variability stresses the need for common standards for assay development and validation (197–201), use of standard curves (202), and/or (international) standardization, as shown for RV (186, 203). Even when general standards for defining seropositivity are applied, different assays show different performance characteristics which are impacted by the assignment of equivocal results (i.e., gray zone) to the positive or negative outcome (108, 204, 205). Since this mainly affects sensitivity, it increases the NPV, particularly when prevalence decreases (92, 168, 206–208).

FETAL INFECTION DIAGNOSTIC TESTING

In case of suspected fetal infection, molecular detection of virus DNA or RNA in amniotic fluid (AF) or cord blood (CB) is the primary diagnostic option in most cases except in TOXO infection. Limited serological data on AF include IgM determination (209, 210). IgG determination is not informative as IgG is usually of maternal origin. IgM and/or IgA determination in AF or fetal blood (FB) has low diagnostic value (211–215), whereas cell culture isolation (virological confirmation) is more specific, for example, in CMV infection (215). Loads in FB or AF may be 100-fold to 1,000-fold higher than those in maternal blood (129, 157, 216), particularly in symptomatic fetuses (217–219), as shown in PB19, CMV, and TOXO infection. Although (viral) load in primary infection may be high in the fetus (211), its presence is not necessarily associated with symptomatic infection (220–223). These discrepancies possibly reflect different windows of infection detection. But overall, PCR on AF or FB has good specificity and NPV in the fetus (Fig. 3) (210, 224–229). In the presence of maternal viremia, sensitivity has been shown to increase for PB19 (216) and for TOXO with a shorter interval to AF or FB sampling (224, 226) or use of multicopy genes (230). Assay performance may be different between AF and FB, with a reported concordance between 73% and 99% (210, 213). Although (transient) ZIKV was reported in AF and FB in fetuses of women with proven infection during pregnancy (21), there are no quantitative data on FB/AF in ZIKV available at this point (231).

FIG 3.

Routine fetal diagnostics methods: sensitivity and specificity, positive predictive value, and negative predictive value median point estimate ± 95% CI. References for the data for each virus are as follows: TOXO (n = 13), 33, 43, 126, 154, 164, 209, 212, 226–228, 230, 319, and 320; PB19 (n = 5), 79, 213, 220, 222, and 223; RV (n = 1), 320; CMV (n = 10), 89, 129, 169, 210, 222–224, 229, 277, and 320.

POSTPARTUM DIAGNOSTIC TESTING

Postpartum sequelae of fetal infections have been observed for TOXO, RV, CMV, and ZIKV. Although the literature is not consistent on this issue, fetal anemia following PB19 infection may result in severe postpartum sequelae (232). In general, timely postpartum diagnosis is hampered by the low sensitivity of IgM testing (33, 36, 40, 205, 228, 233–235), the presence of maternal antibodies, and the high proportion of asymptomatic CMV-infected or TOXO-infected newborns (31, 128, 224, 228, 236). Consequently, ascertainment of congenital disease typically requires longer-term follow-up, posing challenges to the differentiation from postpartum infection (40, 237, 238).

IgM positivity in cord blood or peripheral blood in newborns at <24 h confirms prenatal infection with RV and CMV when supported by viral load testing (233, 239). IgM assay performance is better when testing is done more selectively in symptomatic neonates, as reported for TOXO (218). Contamination with maternal blood should be excluded within the first 10 days postpartum if the first sample was taken from CB, as reported for TOXO (35).

The majority of IgG detected at birth will be of maternal origin (31, 240) but may be neonatal (241). Generally, maternal IgG is assumed to persist for less than 6 months (242, 243). Therefore, in TOXO infection, persistence of the IgG levels in the infant at 12 months is used to confirm or exclude congenital toxoplasmosis. Comparison of IgG immunoblot patterns in mother and child is used to reduce the period of the uncertainty about IgG origin (205, 244). The use of immunoblotting or other multiantigen assays early postpartum (245) has thus been shown to provide an opportunity to differentiate congenital from noncongenital infection by comparing maternal and neonatal antibody binding patterns (246, 247).

The feasibility of using differences in the AI for this purpose has also been studied in different settings, e.g., in infection with RV, TOXO, and CMV (245, 248–251). In RV infection, slow IgG AI maturation in neonates, in combination with presence of IgM, correlated with congenital infection (251, 252). In TOXO infection, identical neonatal and maternal AI excluded congenital toxoplasmosis (35).

Molecular testing of neonatal blood or urine has generally good specificity (239, 252) (Fig. 4) with higher viral load and longer RNA/DNA persistence in symptomatic babies, particularly in urine or throat samples for selected pathogens within a selected period after birth (239, 253–255) (Fig. 4). Viral load has been used to differentiate congenital from postpartum infection when early samples are available, but it is not clear if these findings can be generalized (256).

FIG 4.

Routine neonatal screening methods: sensitivity and specificity, positive predictive value, and negative predictive value median point estimate ± 95% CI. References for the data for each virus are as follows: TOXO (n = 9), 174, 201, 212, 228, 234, 245, 247, 249, and 321; PB19 (n = 2), 36 and 85; RV (n = 2), 233 and 251; CMV (n = 11), 102, 115, 128, 158, 159, 215, 238, 250, 253, 277, and 322.

EFFORTS AT IMPROVING DIAGNOSTIC ACCURACY

There have been many efforts to improve diagnostic accuracy; however, this has not yet resulted in significant improvements. These efforts include development of recombinant (multi-)proteins and peptide-specific tests using different techniques (e.g., immunoproteomics) (257) in (multiplex) assays to improve sensitivity and specificity (143, 258–260), distinction between primary and postprimary infection (261–263), timing (264), and differentiation of transmitters from nontransmitters (265). For example, recombinant proteins in novel avidity assays reported a PPV of >85% (266) and were better suited for IgG detection in TOXO infection (267) or could serve as proxies for functional antibody measurements, for example, virus neutralization in RV infection (268). Multiplex assays are used for simultaneous detection of different antibodies in TORCH infections (34, 108, 113, 143), which is important for differential diagnostic approaches. However, assays still have the performance limitations of the standard assays described. Microarray-based assays have been developed to improve simultaneous testing of antibodies of different pathogens, including (extended) TORCH (269). Use of dry blood spots in multiplex serological assays allows the use of small volumes and may result in shorter diagnostic delays (270–272). Additional use of novel platforms, such as plasmonic gold chip multiplex immunoassay platforms in TOXO infection, may yield potentially better assay performance (273–275). Cell-mediated immunity (CMI) assay data (gamma interferon release assay [IGRA] or enzyme-linked immunosorbent spot [ELISpot] assay) particularly come from CMV, but these assays are not routinely used. A higher CD4⁺/CD8 proliferative T-cell response was associated with primary infection (276–280), improving assay sensitivity for low IgG avidity (281), but here also variability in assay performance was reported in different settings, e.g., in primary infection and transmitters (282). Since CMI in the neonate is never of maternal origin, it is hypothesized that it might aid in differentiating maternal from fetal ZIKV infection. Rapid point-of-care testing, such as immunochromatography (98), loop-mediated isothermal amplification (LAMP) (141, 283), or digital microfluidic (DMF) diagnostic platforms (284), studied for different pathogens, may further reduce the time to first positive test and increase sensitivity and/or decentralized availability of assays in resource-limited settings. Such developments are also reported for ZIKV (285–288). Alternative, novel methods in AF samples include comparisons of metabolic profiles (metabolomics) of transmitter versus nontransmitter infections (289, 290), cytokine profiles (291), or peptidome prognostic classifiers (229) to differentiate infected from noninfected fetuses or to distinguish symptomatic from asymptomatic infections postpartum. Such developments are particularly important as they may provide early (prenatal) information on the risk of overt clinical congenital disease postpartum. Other nonpathogen-related methods are those comparing differential gene or protein expression levels between fetal cells and maternal cells (292). In analogy with previously developed tests for noninfectious prenatal screening (293, 294), genes associated with neurodevelopment were studied as biomarkers in cell-free RNA transcripts in AF samples (295). Paper-based, cell-free detection of RNA was recently evaluated for rapid point-of-care testing of ZIKV (296).

CONCLUSIONS

Our review of approaches to diagnose acute maternal infection, determine vertical transmission risk, and establish presence or absence of congenital infection has shown similarities but also large variations in approaches between pathogens, risking underexploration of methods for optimal diagnostic algorithms. Present diagnosis of TORCH and ZIKV infections is primarily based on serological testing with a focus on IgM and/or IgG detection, for which a variety of commercial assays is available. These assays show variable levels of performance and may not differentiate between primary and nonprimary infections (115) or persistence and/or may be limited by cross-reactivity (108, 112). A positive serological test thus always requires confirmatory testing, including IgG avidity index determinations, immunoblotting, virus neutralization, and molecular testing (43, 212, 224).

The use of different assays and the lack of (international) standardization hamper the interpretation of, and agreement between, different studies (297), despite availability of (WHO) recommended antigens, primers, and probes (298). Efforts to improve detection of primary infection and determination of its time of occurrence during pregnancy have not yet resulted in reliable biomarkers for fetal or congenital disease risk (299). Even if protocols and/or algorithms are in place, variabilities between assays interfere with unambiguous and timely decision-making (300). Predefined evidence-based approaches with standardized diagnostic assays and algorithms are clearly needed to improve adequate and timely diagnosis of (primary) maternal infections and subsequent postpartum congenital disease (297). This is of particular relevance in settings of low endemicity where suboptimal diagnostic performance may increase the risk of false-positive outcomes. Lessons to draw from this review for novel challenges such as ZIKV are to directly combine methods (52), increase epitope specificity (e.g., avidity, immunoblotting, and virus neutralization), and implement paired mother-fetus and/or mother-child testing, as was recently reported for ZIKV neutralizing antibodies (301). Differences in background exposure to ZIKV and other flaviviruses will have an (age-dependent) effect on cross-reactivity and on interpretation of protein-driven assays, such as IB or microarray analysis. In these instances, CMI might be explored as an alternative method to differentiate maternal from congenital ZIKV infection. Standardization of (validated) reference methods is critical in order to compare different methods and might need reference centers to confirm acute infection. As discussed in this review, diagnostics for the TORCH infections, including ZIKV infection, remain complex although techniques continue to develop. Systematic explorations of the published methods across pathogens is important to improve diagnostic algorithms. In the meantime, it is essential to raise awareness among medical microbiologists and treating physicians about the limitations of the presently applied tests and algorithms and the need to improve protocols for diagnostic testing and clinical decision-making. Given the observed disconnect between the different pathogen specialist fields, we conclude that there is a clear case to be made for an integrated TORCH-ZIKV diagnostic approach.

Biographies

Bettie Voordouw, M.D., Ph.D., M.P.H., was trained at Leiden University, the Netherlands. She has worked since 2016 as a medical microbiologist with a focus on clinical virology, vaccinology, and public health microbiology at the National Institute of Public Health and Environment, Bilthoven. There, she is also a project leader for reference laboratories. Furthermore, since 2017 she has been a senior scientist at the Department of Viroscience, Erasmus Medical Centre Rotterdam, with particular interest in emerging (viral) infections and optimizing diagnostic algorithms. She has previous long-term experience in the (international) regulatory field, public health agencies, academia, and scientific committees, leading assessment teams and being responsible for and/or participating in scientific advisory bodies and for clinical, treatment, and vaccine guidelines. Since 2017 she has been the Dutch alternate National Microbiology Focal Point to the European Center for Disease Prevention and Control (ECDC). Since 2019 she has been a member of the board of the National Working Group of Clinical Virology of the Dutch Society of Medical Microbiology.

Barry Rockx, Ph.D., is Head of the Exotic Viruses working group in the Department of Viroscience, Erasmus Medical Centre, Rotterdam. His main research lines involve studies on the tropism, pathogenesis, and host responses of emerging zoonotic viruses causing hemorrhagic, respiratory, and neurological diseases, including orthohantaviruses and arboviruses in a variety of in vitro and in vivo models. He has authored over 50 peer-reviewed scientific publications and several book chapters and has supervised several postdoctoral fellows and students. He has coordinated several NIH-funded projects and contracts and is currently task leader of projects funded by Horizon 2020 of the European Union and by the Netherlands Organization for Health Research and Development (ZonMw). Fields of expertise include tropism, pathogenesis, and the host responses following emerging virus infection.

Thomas Jaenisch, M.D. Ph.D., is clinical researcher and infectious disease epidemiologist at Heidelberg University Hospital (UKHD). In his research, he focuses on emerging arboviruses of global importance, notably dengue viruses (DENV) and Zika virus (ZIKV). Until recently, he was also responsible for the parasitology reference laboratory at UKHD. Dr. Jaenisch has coordinated in direct sequence two multicenter observational dengue consortia and is now the work package leader in the ZIKAlliance consortium (funded by the European Commission [EC] Horizon 2020 program), where he coordinates the multicenter cohorts in pregnant women and children in Latin America and the Caribbean. He is also responsible for harmonization and data sharing between different Zika consortia, aiming at a pooled analysis of cohort data between the three EC-funded Zika consortia and, in the context of a WHO-moderated effort, with other Zika maternal cohort studies globally.

Pieter Fraaij is a pediatric infectious disease and immunology consultant and head of the subdepartment of Pediatric Infectious Diseases, Immunology, and Rheumatology at the Sophia Children’s Hospital of the Erasmus University Medical Centre in Rotterdam, The Netherlands. In addition, he is employed at the Department of Viroscience at the same institution as a clinical scientist. This unique combination of professional environments has allowed him to set up and implement translational research studies, initiating and coordinating interactions between the scientific and clinical departments. He obtained his Ph.D. at the Erasmus University Rotterdam in clinical, virological, and immunological aspects of highly active antiretroviral therapy in HIV-1-infected children. His current research focuses on the impact of emerging and reemerging viruses and the effects of childhood growth and development on host-virus interactions.

Philippe Mayaud, M.D., M.Sc., is a Professor of Infectious Diseases and Reproductive Health at the London School of Hygiene and Tropical Medicine (Faculty of Infectious and Tropical Diseases) and the Head of the HIV Epidemiology and Intervention Research Programme at the Medical Research Unit/Uganda Virus Research Institute and London School of Hygiene and Tropical Medicine Research Unit in Uganda, which he joined in 2019. His research focuses on epidemiological, clinical, and intervention aspects of HIV/AIDS and sexually transmitted infections, particularly in sub-Saharan Africa. In addition, he has conducted research projects on arbovirus/Zika epidemiology research in Brazil where he held the position of Visiting Professor at the University of Sao Paulo (2015 to 2018).

Ann Vossen, M.D., Ph.D., is Associate Professor of Clinical Virology at the Department of Medical Microbiology, Leiden University Medical Center. Her research focuses on congenital cytomegalovirus infection, including several large cohort studies on disease burden, virus-host interaction, and antiviral treatment. She is actively involved in the European Congenital CMV Initiative. Since 2018, she has been President of the Dutch Society of Medical Microbiology.

Marion Koopmans, D.V.M., Ph.D., is head of the Department of Viroscience and director of the WHO collaborating center for arboviruses and viral hemorrhagic fevers, Erasmus Medical Centre, Rotterdam, The Netherlands. Her research focuses on the global population-level impact of rapidly spreading zoonotic and vector-borne virus infections, with an emphasis on improving preparedness and outbreak response through advanced laboratory studies. She is scientific coordinator of COMPARE, a large Horizon 2020-funded project (20 MEuro) exploring the potential uses of next-generation sequencing techniques for outbreak detection and tracking (www.compare-europe.eu), coprincipal investigator in the FP7-funded PREPARE project (www.prepare-europe.eu) aimed at building a pan-European operational network for rapid and large-scale European clinical research in response to infectious disease outbreaks with epidemic potential, and coprincipal investigator in ZIKALLIANCE. She has coauthored >500 papers that have been cited >20,000 times.