Reply

We thank Drs. Perera and Rundle for their interest in our recently published research on PAH¹-DNA adducts assessed in peripheral blood in relation to breast cancer incidence among Long Island women (1). In their letter to the editor², Drs. Perera and Rundle make three comments. First, they mention their research using PAH-DNA adducts in peripheral blood and claim that their findings are similar to ours. However, we are not aware that their peripheral blood results have been published, nor are they described sufficiently for us to comment on whether the two studies are similar.

Second, Drs. Perera and Rundle² describe the results of their already published research on DNA adduct levels as measured in breast tissue (2), work that we also cited and described in our recent publication (1). Drs. Perera and Rundle assert that their cases and controls arise from the same source population. The study included cases and controls selected from a single teaching hospital, which as described by Rothman and Greenland (3), is a design where “the source population is often not identifiable.” Moreover, merely coming from the same source population as the cases does not make a control group representative of that population, as Drs. Perera and Rundle imply.

Third, Drs. Perera and Rundle² suggest that PAH-DNA adducts as measured in breast tissue may be a better measure than adducts assessed from peripheral blood. They appear to base their assertion on comparing the odds ratio of 2.56 (95% confidence interval 1.05, 6.24) for detectable adducts in tumor tissue observed in their hospital-based study of ∼200 subjects (2), versus the estimate of 1.35 (95% confidence interval 1.01, 1.81) for detectable adducts observed in peripheral blood in our population-based study of ∼1000 subjects (1). They further suggest that PAH is a risk factor for breast cancer. We believe that these assertions are premature, because of the paucity of literature that has been published to date on this issue.

Currently, there are several interpretations that are consistent with the epidemiologic studies that have been reported to date that address the PAH-DNA adduct and breast cancer hypothesis (1, 2, 4). First, the odds ratio from the two studies (1, 2) cited in the paragraph above could be described as compatible (primarily attributable to the wide confidence interval estimates from the hospital-based study). This first interpretation would suggest that PAH-DNA adducts may be associated with breast cancer incidence, regardless of the method used to assess the body’s exposure dose.

Alternatively, as recently described by several authors (5, 6), early reports of a strong link to human disease based on small numbers of subjects can be misleading; when replicated in larger studies, the estimates of effect often move closer towards the null. These authors are describing a common pattern in genetic association studies, yet it is also a pattern that is consistent with the few publications on the issue of PAH-DNA adducts and breast cancer, with smaller studies reporting higher estimates of effect (2, 4) and our large study reporting an estimate closer to the null (1). This alternative interpretation would suggest that these early reports may be false positives.

To resolve these issues, confirmation is needed using multiple assessment methods, larger sample sizes, and stronger study designs. Only through multiple repetitions and improved study methods can we determine whether PAH-DNA adducts are indeed a risk factor for breast cancer or just another false positive result.