In the January 2018 issue of Cancer Epidemiology, Biomarkers & Prevention, Yuan and colleagues concluded that their meta-analysis “confirmed the positive association between night shift work and the risks of several common cancers in women.” (1). However, the following methodologic shortcomings restrict the strength of the article:
(i) Absence of definition of night shift work leading to exposure misclassification: Night shift work is not defined in the methodology, and the authors used indiscriminately studies reporting on “sleeping mostly during the day,” sleeping duration, ever working as a nurse or an airplane cabin attendant, and working night shifts with or without rotation. In addition to misclassification, using job title as a proxy for night shifts increases the possibility of confounding by coexposures (e.g., cosmic radiation for airplane crew).
(ii) Methodological flaws: Inclusion of articles providing overlapping data, despite the Materials and Methods stating otherwise, e.g. for breast cancer: Truong 2014 (ever/never night work) and Menegaux 2013 (≥4.5 years of night work); Schernhammer 2006 and Wegrzyn 2017 (same data for NHS-1 study presented in both articles); Pukkala 2012 (presents data included in Rafnsson 2003 and Linnersjö 2003). Inclusion of results from conference abstracts that were not further published (e.g., Wegrzyn 2017, Schernhammer 2014, Marino 2008). Nonsystematic selection of results from reviewed studies: several results from the same article are used to calculate a meta-estimate (e.g., Lie 2013, with risks for estrogen receptor–negative and progesterone receptor–positive breast tumors), whereas for some articles, only one result is used (e.g., Carter 2014, where only rotating shift work results are used, but not those for fixed night work).
(iii) Inclusion of at least one inappropriate study: For digestive cancer the cited article with the largest weight (Tsai 2014) reports the “risk” of nonadherence to colorectal screening among shift workers, and not the risk of colorectal cancer.
These methodologic deficiencies, and the lack of analysis of reasons for heterogeneity (certain cancer sites), explain the divergence in conclusions between the Yuan and colleagues' meta-analysis and reports from multidisciplinary expertises conducted by the International Agency for Research on Cancer (2), and more recently by the ANSES (3), which concluded that there was supportive evidence of an association between night work and breast cancer risk, but insufficient data to support an association with risk of other cancers. We believe that meta-analyses with more stringent methods are needed for asserting the existence of a positive association between night shift work and cancers.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.