The Authors Reply

We thank the Journal for this opportunity to reply to Carozza (1). We appreciate the thoughtful comments of the writer but believe there is misunderstanding as to the intent of our commentary. Our message is not reductionist; rather, we highlighted the need to translate and integrate the rigor of epidemiologic study design into the rapid advances in data sciences, analytic systems, and bioinformatics in order to impact public health (2). Nowhere did we formally propose that epidemiologists “should be trained in computer science and molecular biology” (1, p. 360). Rather, we noted that this approach mandates the need for a variety of skills. Further, we highlighted the exciting opportunities for epidemiologists to launch new multidisciplinary collaborations that reflect these rapidly evolving technological and computing advances (3).

We made it clear that there was no substitute for elegant, well-designed, and pristine descriptive epidemiology, recognizing that classical epidemiologic studies have indeed made seminal contributions to identification of the etiology of most common cancers and to declines in cancer incidence. We agree with the author that the promise of genomic research is yet to be fulfilled, although we must point out that we also included approaches that extend beyond somatic mutations (such as epigenetic changes). We also stressed the need to train cancer epidemiologists in new exposure measurements, critical literature review, appropriate study design to address multilevel analyses, statistical modeling to evaluate complex interactions, and risk communication. Finally, because we recognize that such studies are inherently transdisciplinary, we emphasized the need for effective communication with these diverse scientists and highlighted the importance of multidisciplinary team science. We agree that personal preference is paramount in such training.

Carozza cites the work of Soto and Sonnenschein (4) because they refute the classic hallmarks of cancer. Yet, these same authors state that “the re-emergence of Systems Biology offers an opportunity to overcome the impasse” and that there is a need for “a new methodological outlook where mathematicians will join biologists in having an active participation in the design, exploration and interpretation of these subjects” (4, p. 4).

In his discussion of the complexity of the cancer problem, Weinberg (5) laments the “lack of the conceptual paradigms and computational strategies” for dealing with this complexity. He also stresses the need of integrating and distilling the vast data sets to arrive at a “useful understanding of the behavior of individual cancer cells and the tumors that they form” (5, p. 271).

Finally, we strongly agree with the conclusion in the recent Science commentary by Khoury and Ioannidis that “the combination of a strong epidemiologic foundation, robust knowledge integration, principles of evidence-based medicine, and an expanded translation research agenda can put Big Data on the right course to impact public health” (6, p. 1005).

Thus, we continue to believe that there is a need to train 21st-century epidemiologists in the challenges of integrating big data into classic epidemiologic studies in order to analyze, interpret, and display such data and also to communicate and translate findings. Finally, nowhere in our text did we use military terminology.