Reassessing the causal role of early-life adiposity in breast cancer: could the apparent inverse associations be a manifestation of survival bias?

Abstract

Listen

Translate article

Hao et al.1 present an extensive examination of the role of adiposity in female breast cancer (BC) using comprehensive Mendelian randomization (MR) methods. Previous MR studies of adiposity on BC have not matched the increased risk from adult body mass index (BMI) established by the International Agency for Research on Cancer and other expert reviews.1,2 Hao et al.1 found adult BMI inversely associated with the onset of BC, but identified childhood, rather than adult, BMI as driving the inverse association. Hao et al.’s use of an MR study largely obviates confounding by using genetically predicted exposures. MR studies are open to the other major source bias affecting causal inference, i.e. selection bias, including survival bias arising from only selecting survivors of exposure and outcome. Many cancer studies, as used by Hao et al.,1 typically recruit in late middle age (cases mean age ∼57 years here)3,4 and so inevitably miss the BC deaths that occurred before recruitment, particularly given that BC is the leading cause of natural death until late middle age5 and is the leading cause of death for young women diagnosed with cancer.6 Childhood BMI tracks into adulthood.7 In young to middle-aged women, higher BMI historically was inversely associated with survival.8 Thus, Hao et al.’s study is open to unrecoverable survival bias from missing deaths that occurred before recruitment from genetically predicted BMI, BMI and BC (Figure 1), i.e. suffers from under-ascertainment of exposed cases. Notably, survival bias may also account for some of the other inverse associations of childhood BMI with BC because many studies of childhood/adolescent adiposity on risk are based on recall of childhood weight in cohorts and case–control studies that differentially select individuals surviving to mid-life.