Dear editor, We thank Heck et al. for their comment on our recent study of neonatal phototherapy and childhood cancer. 1 Heck et al. suggest that the association between phototherapy and the risk of childhood cancer may be explained by shared risk factors, such as preterm birth and low birth weight. To support this point, the authors performed a matched casecontrol study of Danish children with cancer between 1977 and 2013, and found that hyperbilirubinemia with or without phototherapy was not a risk factor for childhood cancer in analyses stratified by gestational age and birth weight. The authors use these findings as evidence that our results were most likely affected by unmeasured confounders.We commend Heck et al. for their attempt to clarify the relationship between neonatal phototherapy and childhood cancer. Confounding-by-indication is a well-recognized problem of observational studies in which factors associated with cancer cannot be measured or accounted for. The underlying problem is that these factors are an indication for phototherapy, raising the possibility that the increase in risk of childhood cancer attributed to phototherapy is in reality due to these other factors. The optimal way of eliminating confounding-byindication would be a randomized trial, but this approach is not possible due to the potential for harm from withholding phototherapy in patients with severe hyperbilirubinemia.While we appreciate Heck et al.'s attempt to address the role of confounding, we are compelled to point out that their results cannot be used to support their claim. The foremost problem is that the main exposure measure in Heck et al.'s analysis is a combination of phototherapy and untreated hyperbilirubinemia, rather than phototherapy specifically. As many jaundiced infants do not receive phototherapy, it is likely that their main exposure measure is heavily weighted by patients with untreated jaundice. Thus their results are representative of the effect of hyperbilirubinemia, not phototherapy which is the exposure of interest. It is essential to define the exposure as phototherapy specifically, otherwise the analysis merely reflects the association between jaundice overall and risk of cancer.To illustrate the impact of incorrectly defining the exposure, we reanalyzed our data combining phototherapy with untreated jaundice. Table 1 demonstrates that phototherapy is associated with childhood cancer between 4 and 11 years in our data, especially in term infants, but that untreated jaundice is not. When we combine the two exposures, the association with cancer disappears or weakens. The data behave in this manner due to the absence of a relationship between untreated jaundice and cancer. It is essential to analyze phototherapy separately in order to avoid attenuating the associations towards the null.Moreover, Heck et al. show that risk estimates for most cancers are greater in preterm infants than in term infants, a finding consistent with our results for untreated jaundice and phototherapy, or untreated jaundice a...