Sir,

We thank Yang et al (2015) for their interest in and thoughtful review of our study. We agree with the correspondents that certain limitations should be considered when interpreting findings on sleep duration and breast cancer risk. As we acknowledged in our paper, self-reported sleep duration involving measurement error could lead to misclassification of our main exposure. We agree with Yang et al that misclassification can lead to bias in either direction. However, in our case, we believe the bias is more likely to be towards the null. A validation study by Lauderdale et al (2008) compared self-reported sleep duration to an objective measure (actigraphy), and found that the validity of self-reported sleep varied by the amount of sleep recorded. In general people tended to over-report their sleep duration, but the extent of over-reporting increased as sleep duration decreased. Therefore, short sleepers were at a higher risk of being misclassified as normal or long sleepers, which might have led to an inability to detect an increased risk of breast cancer among such individuals. Of course, regardless of the direction of the bias, misclassification of exposure is an important problem to consider, and we appreciate that Yang et al highlighted this particular element of our report. Moreover, we also agree with the correspondents that it is important to measure sleep at different time points in order to get a better estimate of long-term sleep duration, and to consider other sleep characteristics and sleep-related factors, such as sleep quality and exposure to light at night.

We are aware of the discrepancy between our findings and the melatonin hypothesis, which suggests that short sleep duration is associated with decreased levels of melatonin. Because melatonin is a molecule with anti-oestrogenic effects, decreased levels of melatonin may increase the risk of ER+ tumours (Blask, 2009). We found no association between sleep duration and hormone receptor-positive tumours, which is consistent with the only two studies that examined sleep duration in relation to breast cancer subtypes defined by hormone receptor status. Notably, both studies, an Australian case–control study and a study in the Women’s Health Initiative, showed no relationship between sleep duration and ER+ tumours (Girschik et al, 2013; Vogtmann et al, 2013). In contrast, we found an increased risk associated with short sleep durations for hormone receptor-negative breast cancers. Although we cannot exclude the possibility that our finding is due to bias or chance, we believe that there are biological mechanisms that support this observed association. For example, short sleep and sleep deprivation have been associated with factors that may influence breast cancer risk independent of oestrogen pathways, such as metabolic dysfunction (Gangwisch et al, 2007) and chronic inflammation (Irwin et al, 2006).

Again, we thank Yang et al for this letter and are glad that more studies, such as the population-based case–control study in Jiujiang city mentioned by Yang et al, are using objective measures along with questionnaires to better assess both the quantity and quality of sleep in relation to breast cancer risk and other health outcomes.