Obesity and Gynaecological Cancer — an Umbrella Review and Meta-analysis.
Kalliala Ilkka1, Tsilidis Kostas34, Kyrgiou Maria1,2
1Institute of Reproduction and Developmental Biology, Department of Surgery & Cancer, Imperial College, London, UK
2Imperial Healthcare NHS Trust, London, UK
3Department of Hygiene and Epidemiology, University of Ioannina School of Medicine, Ioannina, Greece
4Department of Epidemiology and Biostatistics, School of Public Health, Imperial College London, London, UK
Background:
In order to estimate the strength of current evidence and existence of potential biases in the literature, we performed an umbrella review of meta-analyses studying the association between obesity and gynaecological cancers.
Methods:
We searched PubMed, Embase, and the Cochrane database of systematic reviews, and were able to includedata from12paperscontaining altogether 47 meta-analyses of incidence or mortality from breast, cervical endometrial or ovarian cancer.
For each meta-analysis, we calculated both fixed and random effect summary estimates; 95% prediction intervals for the random effects estimates;used the Cochran Q test and the I2 metric of inconsistency with 95% CI to asses heterogeneity; the Egger’s test and an estimate of a theoretical study of infinite sample size by extrapolating the Egger’s regression line for small study effect;and excess significance by comparing the number of studies that had the power to have significance to the number of studies that showed significant results.
Results:
Of the 47 meta-analyses included, in 18 studies obesity or weight gain showed associationat strong level of significance (P<0.001)in both fixed and random effects models. Mostmeta-analysis, however,showed hints of significant heterogeneity, excess significance, small study effects, or the 95% PI did not exclude the null value. Only 2meta-analyses, between excess weight gain and increased incidence of postmenopausal breast cancer,met all these criteria.
Conclusions:
Most summary meta-analytic estimates showed nominally significant results (p<0.05) for the association between obesity and the risk of developinggynaecological cancer. Only two associations, however, had highly significant results andno hints of biases. These results underline the substantial uncertainty around the nominally significant associations, but do not necessarily mean that only these two associations are genuine — the large observed heterogeneity and observed biases in almost all of theincluded studies do, however, suggest that some of these observed associations could well be inflated or false positive.