Accuracy of outcome definitions in Mendelian randomization of
maternal health
Qian Yang,1,2 Maria Carolina
Borges1,2
1 MRC Integrative Epidemiology Unit at the University
of Bristol, Bristol, UK
2 Population Health Sciences, Bristol Medical School,
University of Bristol, Bristol, UK
Dear Dr Papageorphiou,
We recently read the article by Dr Ardissino and colleagues entitled
‘Genetically predicted body mass index and maternal outcomes of
pregnancy: A two-sample Mendelian randomization study’ [1], where 11
outcomes were investigated. To conduct Mendelian randomization (MR)
analyses, this study extracted associations of selected genetic variants
with those outcomes from publicly available GWAS (genome-wide
association study) summary data from FinnGen (the sixth release, total
N=147,061 women) – a national-wide network of Finish Biobank [2].
We noticed that “postpartum depression” included in Ardissino et al
was inconsistent with the commonly used definition of postnatal
depression occurring within a year of delivery [3,4]. FinnGen
defined this outcome based on the International Classification of
Diseases 10th Revision (ICD-10) codes (ICD-10 F32, F33
and F53.0) among women with at least one episode of delivery (ICD-10
O80-O84), without considering the time interval between delivery and
diagnosis of depression. Therefore, cases of “postpartum depression”
could be ascertained at any time after giving birth and, therefore,
could be unrelated to pregnancy. As a consequence, findings for
“postpartum depression” in Ardissino et al should be interpreted with
caution due to the unspecific outcome definition.
The increasing availability of publicly available or accessible data
from GWAS consortia (e.g. Early Growth Genetics) and large biobanks
(e.g. UK Biobank and FinnGen), combined with the creation of automated
pipelines (e.g. MR-Base [5] used by Ardissino et al), has supported
an rapid increase in publications using two-sample MR. Though such a
combination has great potential to promote open science and advance
health research, including in maternal-child health, we cautioned that
detailed understanding of procedures used to generate GWAS summary data
underlying MR analyses is of major importance to obtain reliable
evidence and interpretable findings.