1.3 Data extraction
Two reviewers (CC and CXY) retrieved articles independently about the
titles and abstracts. If there was any contradiction, we would discuss
it with the third author (WDH) and decide it together.
We searched 17 types of cancer including 6,139 articles in total. Then
240 observational articles were collected after reading the
title/abstract. Finally, after careful review of the inclusion and
exclusion criteria, 46 articles of 6 types of cancers were identified
for dose-response meta-analysis. The study selection process and results
from the literature search are presented in Figure S1.
Data from all eligible articles were collected by two authors (CC and
CXY), including the first author’s name, publication year, BW range
(kg), BW measure methods, cancer type, number of cancer case, sample
size, and the baseline status of the study populations such as age, sex,
ethnicity. (Table S2) For dose-response meta-analysis, we extracted some
data that need to be calculated including the first author, publication
years, BW dose, total participants, HRs or ORs, 95% confidence
interval, the logarithm for HRs or ORs, standard error of HRs or ORs
(SE). All of them were inputted into Excel for analysis. If the article
provided more than one HRs or ORs, we chose the most adjusted one to
exclude as many confounding factors as possible. If there were some
missing data, we used appropriate statistical methods to calculate39, 40. 5 studies did not report the number of
participants exposed under different BW categories 14,
35, 41-43, we averaged the total sample sizes of all BW groups (C) to
each group. And 3 studies lacked the number of cases of each BW group
(Nx) 36, 41, 42, so we calculated them
deriving from the HRs or ORs (RRx) and sample sizes
(Mx) at each BW group.
(Nx=RRx×Mr/NrMx;
C=Nx1+……+Nxn;
Mr: sample sizes of reference BW group;
Nr: cases of reference BW group)