Reply to Comment on: “Night shift work and risk of breast cancer in women: the Generations Study cohort”

We thank Kogevinas for his interest in our paper on night-shift work and breast cancer risk in women. 1 He raises four points: i.That our question on night work mixes evening and nightshift work, and this would result in the "exposed" group including non-exposed evening workers.This, of course, depends on one's definition of "exposure".Interest in nightshift work and breast cancer springs from the hypothesis that exposure to light at night may increase cancer risk. 2 In 2007, the International Agency for Research on Cancer concluded that there was limited evidence in humans for the carcinogenicity of shift work that involves night work, but overall shift work that involves circadian disruption is probably carcinogenic. 2Our analyses included hours 10pm-7am as "exposure" since these involve light at night and may influence the circadian rhythm.The exact times that count as "night" are somewhat arbitrary, if only because nightfall varies by season and latitude, and bedtimes vary between individuals.However, we did present the results for <7 and 7 +hours worked per night (Table 2) and found no raised risk with working 7+hours per night.Working 7+hours between 10pm and 7am could not be considered primarily evening work.ii.That our study includes mostly middle-aged participants and may have misclassified some women who worked night shifts in an earlier period.At recruitment, women in our study had a mean age of 40.2 years, which was considerably younger than the mean ages (between 54 and 59 years) in the five case-control studies in the pooled analysis 3 cited by Kogevinas as evidence for an association, and much younger than the mean ages in the recent pooling of prospective studies (Million Women Study: 69 years; UK Biobank: 56 years; EPIC-Oxford: 58 years). 4To limit the burden on respondents, we did not collect the lifetime history of night shifts but only night work that occurred, or at least ended, during the 10 years before recruitment.Some have suggested that any increased risk associated with night-shift work may diminish soon after exposure ceases, 3,5,6 and recent, rather than historic, night shifts may be the more appropriate exposure measure.Thus, we analysed night work in the last 10 years, for which period earlier night work would not be the exposure of interest.iii.That we only included long-duration workers if they had survived and entered the period 10 years before recruitment, and this could bias towards the null analysis by duration.We found no association with duration in our study, and as we acknowledged in our paper, our data for durations beyond 10 years of night work were limited (see response to (ii)).iv.That the least biased results in our study are certain ones cited by Kogevinas that approached statistical significance.