1.2 Clarifying the research question

When evaluating a set of environmental factors detected in a given population as an environmental mixture, a critical step is the identification of the research question of interest. The discussion of the different methodologies presented in the aforementioned NIEHS workshop concluded that we do not have an optimal approach, but that each method performed well under a specific research question. Here are some of the most common questions that we may want to address:

1. Do we have recurrent patterns of exposures?

With several factors at play, it is often of interest to understand whether specific components of the mixture are clustered into smaller subgroups, based on similar characteristics, shared sources, or other features.

2. What is the overall effect of the mixture on a given outcome?

From our previous example, we may be interested in evaluating the overall effects of parabens exposure on the risk of diabetes. We are not really interested in the specific role of each compound but only on the cumulative effect of the several components.

3. Who are the bad actors? What are the chemical-specific effects within the mixture?

Let’s assume that we have identified a potentially harmful effect of our mixture on the outcome of interest, and we therefore want to reduce the levels of exposures in our population. If question 1 has identified common patterns due to shared sources, we could simply target these sources, disregarding the actual effects of these chemicals. Alternatively, we could try to identify which component of the mixture is responsible for the effect observed in question 2. In our parabens example, if we had observed a positive association we may want to further investigate what parabens compounds are driving the association between the mixture and the outcome.

4. Is there any interaction between chemicals in predicting the outcome?

When more than one mixture component contributes to a given mixture-outcome association, it is reasonable to expect that some kind of interaction between the two will be present and we might want to formally identify and estimate the role of these joint interactive effects.

In general, one might have one or more research questions in mind, or simply want to evaluate the mixture in an exploratory way. No matter what, it will always be recommended to explore different techniques and thoroughly compare and validate results using different approaches.