Study selection, Characteristics, and Quality
The study selection process is shown in Figure 1. We identified 6113 unique publications from the databases and one from other sources. After title and abstract assessment, 98 publications were selected for full-text screening. A total of 34 articles met our eligibility and were included in this review, and 33 were included in meta-analyses.
As shown in Table S2, only two studies33,34 were of moderate quality (NOS = 6) and all other studies were of high quality (NOS 7-9). The detailed characteristics of included studies are presented in Table S3. The sample sizes and study periods of the cohorts ranged from 41 to 72,072, and 1991 to 2017 respectively. All studies but one were conducted in pregnant women from the general population, with variable exclusion criteria related to maternal prepregnancy BMI, history of preterm birth, and maternal comorbidities. One study was conducted in pregnant women with obesity or history of GDM35. Four cohorts were conducted in Singapore36, Malaysia33, Mexico37, and a French Caribbean island38. The other cohorts were conducted in high-income countries located in Europe, North America, and Oceania. Six studies used at least one self-reported outcome, while all other studies used valid assessments including medical records or standardized measurements. For dietary assessment, four33,37,39,40studies used dietary recalls, one36 used both dietary recalls and food diaries, three35,41,42 used non-validated FFQ, and the remainder of studies used a validated FFQ. Studies varied as to the timepoint and period of dietary assessment. Four studies43-46 assessed diet in prepregnancy and other studies assessed diet during pregnancy. Diet quality was assessed using 22 different scores, with versions of aHEI, HEI and MD scores most frequently assessed (Table S4). There was large variability of potential confounders included in the multivariable models across studies (Table S5).