The more we learn about the Zika virus disease, the more conclusive becomes the association between maternal infection and babies born with microcephaly (smaller-than-normal heads), and to a lesser extent with Guillain-Barré syndrome (GBS), an autoimmune neurological disorder. The research cliché -- correlation does not equal causation -- is certainly instructive in this situation, as there is no established direct and specific link between maternal Zika virus infection and these two relatively rare disorders. Nonetheless, the increasing incidence (particularly for microcephaly), the severity, and the implication of the Zika virus, which is mosquito borne, a vector with a wide geographic distribution, spell tremendous concern and urgency to explain the link.
The questions on the minds of researchers and epidemiologists are: What is the nature and extent of the association between maternal Zika virus infection and these adverse pregnancy outcomes (microcephaly and GBS)? Is it a causal link? Is the Zika virus acting alone or jointly with other factors (co-factors) to mediate this disease process? In other words, might the Zika virus be necessary but not sufficient to cause microcephaly and GBS? The latter is certainly a plausible hypothesis given that a substantial number of Zika-infected mothers apparently had babies with neither of these adverse outcomes. Equally important, is exploring the prospect that this may be a spurious or confounded association -- that is, it may be explained by other factors related to both maternal Zika virus infection and the adverse pregnancy outcomes. This, however, does not seem very likely (especially for microcephaly), given the evidence to date.
Other than a few case studies, the emerging research evidence is mostly from case-control studies, and to a much lesser degree, from longitudinal or cohort studies. The longitudinal design is prospective in nature, beginning with the primary risk factor or exposure -- that is, maternal Zika virus infection -- and following the pregnant women forward in time to observe for the adverse outcomes (fetal microcephaly and GBS). The case-control study design is retrospective in terms of data collection -- that is, the adverse outcomes (microcephaly and GBS) have already occurred, and the researchers are going back in time, interviewing mothers, and reviewing healthcare records to determine what factor(s) can explain the presence or absence of the adverse pregnancy outcomes. Key limitations of this design are recall bias and incomplete or nonexistent medical records. Notably, both the case-control and longitudinal designs are observational or non-experimental in nature, and hence, do not directly allow for determining cause and effect.
Indeed, the "gold standard" research design for exploring cause and effect is a true experiment, often referred to as an RCT (Randomized Controlled Trial), but this design is neither appropriate nor ethical in this context, given the available evidence of a strong association between maternal Zika virus infection and major adverse pregnancy outcomes.
Therefore, given our reliance on data from non-experimental studies, it is critical that the quality of the evidence be rigorously evaluated to determine if causality is suggested or can be inferred. The commonly used model for evaluating research evidence is the Bradford Hill criteria put forth by Sir Austin Bradford Hill in 1965, and it can be operationalized as follows, in this context.
Researchers will examine the strength of the association between maternal Zika virus infection status and these adverse pregnancy outcomes, by determining the likelihood (odds ratio or relative risk) of the disorder being present in babies whose mothers were Zika-infected compared with mothers who were not. Thus far, the risk (particularly for microcephaly) seems significantly higher for mothers who had Zika virus infection.
The consistency of this association will be checked to determine if similar findings have been observed in different places and populations, at different times, and from different study designs. The evidence in this regard, is limited and mixed, a finding that can implicate co-factors in the causal process.
The next consideration is specificity of the association -- that is, is there a direct one-to-one relationship between maternal Zika virus infection and the adverse pregnancy outcomes (primarily microcephaly)? This criterion is generally not required to establish causation, and in its strict sense, has limited application in this context, recognizing that there are other known causes of microcephaly. However, it may be, that these adverse pregnancy outcomes are specific to a set of factors -- that is, the Zika virus and co-factors -- suggesting multiple causality or multifactorial causation.
Another and indeed the most important criterion (for establishing causality) is the temporality of the association between maternal Zika virus infection and these adverse pregnancy outcomes (particularly microcephaly) -- the "chicken and the egg" conundrum; which came first? If the maternal Zika virus infection caused the congenital microcephaly, then it had to precede it. For the case-control studies, maternal Zika virus infection status is being determined by self-reports, medical records, and in some instances, biological markers (such as specific antibodies). The evidence thus far, for the assessed cases, seems to support that the maternal Zika virus infection preceded the development of microcephaly, but this does not rule out the implication of co-factors (personal, environmental, and biological, including the stage of pregnancy at which infection occurred).
A further criterion is the biological gradient of the association, or the dose-response relationship, and while this may have limited application in this context, it can be explored by checking to see if the degree of severity of symptomology (for maternal Zika virus infection) is positively correlated with the probability of having a baby with one of the adverse outcomes. For longitudinal studies, relevant biomarkers such as viral load would be helpful, in this regard.
Yet another criterion is the biological plausibility of the association, and this can be combined with the coherence and analogy of the association, which together means that based on the body of knowledge in medicine and related fields, does it make scientific sense that maternal Zika virus infection can cause microcephaly and GBS? And the answer is yes for microcephaly, in particular, as it is known to be caused by other infectious organisms. Also, a couple of recent case reports have directly implicated the Zika virus in the pathogenesis of fetal microcephaly.
And finally, we cannot escape the ubiquitous criterion of statistical significance (p<.05), which is often erroneously considered as independently indicative of the quality of the research evidence, when in fact, by itself, it tells us nothing about the importance and utility of the evidence to decision-making. Statistical significance is purely probabilistic, and only helps to determine the role of chance or random error in explaining the results. Accordingly, the validity, clinical significance, and reproducibility of research evidence depend on much more than statistical methodology, and in this regard, the Bradford Hill criteria can be helpful toward considering the totality of the evidence.
It seems safe to say that the evidence to date strongly implicates the Zika virus in the multifactorial causation of microcephaly, in particular. The challenge will be to identify the specific co-factors with which the Zika virus interacts to mediate this disease process.
Rossi A. Hassad, PhD, MPH, is an epidemiologist and professor at Mercy College, in Dobbs Ferry, N.Y. He is a member of the American College of Epidemiology, and a Fellow and Chartered Statistician of the Royal Statistical Society, U.K.