Association of Birth Weight With Type 2 Diabetes and Glycemic Traits

Key Points Question Is birth weight associated with type 2 diabetes and glycemic traits? Findings This mendelian randomization study found that a 1-SD decrease in birth weight due to the genetic risk score was associated with a higher risk of type 2 diabetes among European and East Asian populations. In addition, a 1-SD decrease in birth weight was associated with a 0.189-SD increase in fasting glucose concentration, but not with fasting insulin, 2-hour glucose, or hemoglobin A1c level. Meaning A genetic predisposition to lower birth weight was associated with an increased risk of type 2 diabetes and increased fasting glucose, suggesting potential mechanisms through which perturbation of the antenatal and early-life environment affect predisposition to diabetes in later life.


Study-level data Birth weight SNP selection and data sources
To explore the causal effect of birth weight (BW) on T2DM (T2DM) in 49 individual studies from the CHARGE-BIG study, we calculated a genetic risk score using 7 SNPs for BW identified from previous EGG consortium. Genetic variants associated with BW at a genome-wide significant level (P < 5 × 10 −8 ) were obtained from EGG consortium with up to 153,781 individuals. 1

Genetic risk score calculation
To estimate the genetic predisposition to low BW, a genetic risk score (GRS) was calculated on the basis of the 7 above-mentioned SNPs (eTable 1). 1 We assumed that each SNP in the panel acts independently in an additive manner, and the genetic score was calculated by using a weighted method. Each SNP was weighted by its relative effect size (β-coefficient) obtained from the reported meta-analysis data. 1 We firstly created a weighted score using the equation: weighted score = β1×SNP1 + β2×SNP2 + … + βn×SNPn, where β is the β-coefficient for each individual SNP, and n is number of SNPs. To reflect the number of BW-decreasing allele, we rescaled the weighted score using the following equation: weighted genetic risk score = weighted score × (total number of SNPs / sum of the β-coefficients).

Meta-analysis and between-study heterogeneity
We assessed between-cohort heterogeneity via Cochrane's Q-statistic and I 2 -statistics. [2][3][4] Histograms of the distribution of I 2 were shown in the associations of BW with T2DM and glycemic traits, as well as the associations of the BW related GRS with glycemic traits. For the proposed cut-off of I 2 >0.25, it means there is non-negligible heterogeneity between studies for associations. If there is non-negligible heterogeneity, we used random-effects meta-analysis throughout, otherwise, fixed effects metaanalysis were used. 5

Standard errors and inference for the instrumental variable (IV) estimator
After meta-analysis, we used the IV estimators to quantify the strength of the causal association of BW and T2DM. 6 The IV estimator, which is identical to that derived by the widely used two-stage least squares method, 7 was calculated as the β of the regression coefficients of GRS-T2DM and GRS-BW: The standard error was calculated via the delta method as = Based on these estimates, we appealed to standard-normal asymptotics, with the resulting Wald test statistic and 95% confidence intervals given as 1.96 The p-value for the : 0 was derived from the standard normal distribution. For the association with T2DM, the 95% CI estimates were back-transformed through the antilog and exponentiation, respectively.
For comparing the IV estimate and the conventional estimate _ , we considered the difference MR analyses assume that the chosen SNPs do not exert pleiotropic effects on the outcomes by operating through biological pathways independent of the exposure. However, in MR analyses, a SNP may influence the outcome via other factors if the SNP acts upon the other factors through the exposure itself. 10 The inclusion of SNPs that contribute through a pleiotropic pathway could bias estimates. This can be difficult to test when examining BW, which is influenced by many different environmental factors and physiological mechanisms. In this study, we used MR-Egger regression to detect the presence of pleiotropy. 13 In brief, the approach is based on Egger regression, which has been used to examine publication bias in the meta-analysis literatures. Using the MR-Egger method, the SNP's effect upon the exposure variable is plotted against its effect upon the outcome, and an intercept distinct from the origin provides evidence for pleiotropic effects. Additionally, the slope of the MR-Egger regression can provide pleiotropy-corrected causal estimates. An important condition of this approach is that the SNP-exposure association must be independent of the SNP's direct effects upon the outcome, which may not always be satisfied in cases where all pleiotropic effects are attributed to a single confounder. Nonetheless, the MR-Egger method can provide unbiased estimates even if all the chosen SNPs are invalid. 13 In addition, the weighted median approach was used to examine causal effect and pleiotropy. 14 Using this method, MR estimates are ordered and weighted by the inverse of their variance. The weighted median approach offers some important advantages over MR-Egger because it improves precision. Therefore, the inverse-variance weighted, MR-Egger and weighted median methods were considered as sensitivity analyses for MR investigations with multiple genetic variants. 12,14,15 For analyses of both study-level data and summary-level data, the effect size for each meta-analysis was reported in the main results as the effect on a one-standard-deviation (1-SD) change in BW or glycemic quantitative traits, because this metric is more interpretable than an arbitrary difference. Analyses were performed using Stata version 12 (StataCorp) and R version 3.2.3 (R Project for Statistical Computing). The threshold of statistical significance for T2DM as the primary outcome was P < 0.05. The threshold of significance for the analysis of glycemic traits as secondary outcomes was P < 0.01 (0.05/4 = 0.01).

Standardization of MR Estimates
Data (β values) from the CHARGE-BIG study were standardized so that the association of BW with T2DM risk could be uniformly expressed in terms of standard deviations. For BW, 1-SD was assumed to correspond to 543 gram, according to the pooled SD from the EGG consortium. 1 The EGG consortium reported estimates of variants in units of standard deviations of BW, while the MAGIC consortium did not. Therefore, β values from the MAGIC consortium were also standardized so that the association of BW with T2DM risk and glycemic traits could be also uniformly expressed in terms of standard deviations.
For fasting glucose, two-hour glucose and HbA1c from the MAGIC consortium, 1-SD was assumed to correspond to 13.1 mg/dl, 10.1 mg/dl and 0.535%, respectively, according to the pooled SD of studies included in a previous report from the MAGIC consortium. 17 1-SD of 0.44 for log-transformed fasting insulin from Europeans was used. 18

Calculation of absolute risk increases
To estimate the absolute risk increase based on calculated odds ratio estimated for T2DM, the United States population level estimate of the incidence of T2DM by the Center for Disease Control and Prevention was used (7.8/1000 participant years of follow up). 19 The absolute risk increase associated with T2DM was then calculated using this formula: ARI = (OR-1)*AI where ARI is the absolute risk increase, OR is the odds ratio and AI is the absolute incidence in events per 1000 participant years.

eFigure 1 Schematic representation of a Mendelian randomization approach
MR can be used to test the hypothesis that exposure (birth weight) causes outcome (T2DM and glycemic traits). Three assumptions of MR: 1. Genetic variants are associated with birth weight. 2. Genetic variants are not associated with confounders. 3. Genetic variants influence T2D and glycemic traits only through the birth weight, not through other pathways.

eFigure 2 Genetic association with birth weight
LISA/GINI: LISA/GINIplus. Results were standardized to a 1-SD decrease in birth weight due to genetic variants. The genetic risk score for low birth weight was selected as instrumental variable. The lower genetic risk score was associated with higher birth weight. Linear regression models were used to test the association of genetic risk score with birth weight, after adjustment of sex, gestational age and principal components for population stratification region in each study. We pooled β coefficients (1-SD of 543 gram in BW) across 23 studies using random-effect meta-analysis due to the heterogeneity between studies (I 2 =79.2%, P <0.001).

eFigure 3 Genetic association with risk of T2DM
Logistical regression models were used to test the association of genetic risk score with risk of T2DM, after adjustment of gender, gestational age and principal components for population stratification region in each study. We pooled β coefficients across 33 studies using random-effect meta-analysis due to the heterogeneity among studies (I 2 =64.0%, P <0.001).

eFigure 4 Association of birth weight with risk of T2DM
ARIC-AA: ARIC (African Ancestry) ARIC-EA: ARIC (European Ancestry) Logistic regression was used to test the association of birth weight with risk of T2DM after adjustment of sex, ethnicity, region, and other baseline covariates if available (age, BMI, smoking status, physical activity, total energy intake, and alcohol intake) in each study. We pooled OR (1-SD of 543 gram in BW) across 11 studies using random-effect meta-analysis due to the heterogeneity among studies (I 2 = 95.2%, P <0.001). Results were standardized to a 1-SD decrease in birth weight due to genetic risk score. SD is 543 gram from EGG consortium. We pooled OR across 8 studies that provided both GRS-BW and GRS-T2DM data using random-effect metaanalysis due to the heterogeneity among studies (I 2 =60.7%, P = 0.013).     rs1801253 G/C −0.041 (0.007) 3.6E-09 Results are from inverse variance, fixed-effects meta-analysis of all available study samples of European ancestry. The effect allele for each SNP is labelled on the positive strand according to HapMap. The beta value is the change in trait z score per birth weight-lowering allele from linear regression, adjusted for sex and gestational age (where available), assuming an additive genetic model. To obtain the equivalent birth weight effect in grams, we multiplied by 484g, the median birth weight standard deviation of European studies in 2. There was little detectable heterogeneity between studies (all P > 0.01). *Results are unadjusted for maternal genotype or birth length, but only in samples where maternal genotype or birth length is available (for direct comparison with the model that is adjusted for maternal genotype or birth length, respectively.) The β value is the change in z score per birth weight-lowering allele from linear regression, adjusted for sex and gestational age (where available), assuming an additive genetic model. To obtain the equivalent birth weight effect in grams, we multiplied by 484 g, the median birth weight standard deviation of European studies. Reference: EGG consortium, 2013, Nature Genetic eTable 7. Genetic association of birth weight genetic variants with glycemic traits (data from summary results). Sixty loci associated with birth weight (P < 5 x 10 −8 ) in European ancestry meta-analysis of up to 143,677 individuals and/or trans-ancestry meta-analysis of up to 153,781 individuals. BW was z-score transformed separately in males and females after excluding non-singletons and premature births and adjusting for gestational age where available. a, Effects (beta values) are aligned to the BW-raising allele. EAF was obtained from the trans-ancestry meta-analysis, except for PLAC1, for which the EAF was obtained from the European ancestry meta-analysis due to lack of X chromosome data from the non-European studies. Chr., chromosome; bp, base pair; EAF, effect allele frequency; SE, standard error. The lead SNP (or proxy, EUR r2>0.6) at the 60 BW loci was queried in publicly available GWAS meta-analysis datasets or in GWAS result obtained through collaboration79. Results were available for 53 of those loci and the extracted z-score (allelic effect/SE) was aligned to the BW-raising allele. Questionnaires were collected at baseline and biennially thereafter, to update information on lifestyle factors and the occurrence of chronic diseases. In the current analysis, we used 1990 as baseline in the NHS, when the earliest complete dietary data were collected. Our analysis included 13,000 women whose genotype data were available. All of the participants were Caucasians and were free of cancer at baseline. The study protocol was approved by the institutional review boards of Brigham and Women's Hospital and Harvard School of Public Health.

HPFS
The HPFS was initiated in 1986, and was composed of 51,529 male dentists, pharmacists, veterinarians, optometrists, osteopathic physicians, and podiatrists, aged 40-75 y at baseline. The male participants returned a baseline questionnaire about detailed medical history, lifestyle, and usual diet.
Questionnaires were collected at baseline and biennially thereafter, to update information on lifestyle factors and the occurrence of chronic diseases. In the current analysis, we used 1990 as baseline in the HPFS, when the earliest complete dietary data were collected. Our analysis included 8,000 men whose genotype data were available. All of the participants were Caucasians and were free of cancer at baseline. The study protocol was approved by the institutional review boards of Brigham and Women's Hospital and Harvard School of Public Health.

DCH
The present study includes a case-cohort sample of 1,812 cases who developed T2DM before 31 Dec 2006, and 1,633 randomly selected controls, nested within the population-based Diet, Cancer and Health cohort. Diet, Cancer and Health is a Danish prospective cohort study originally aimed at investigating the associations between dietary habits, lifestyle, and cancer development. The participants were recruited during 1993-1997. A total of about 160,725 individuals were invited by mail, and 57,053 were enrolled into the study cohort. The participants are men and women born in Denmark, living in the greater Copenhagen or Aarhus areas, aged 50-64 years, and with no previous cancer diagnosis.

GOYA
The GOYA (Male) cohort is a longitudinal case-cohort (obese, non-obese) study comprising a randomly (1%) selected control group and all extremely overweight men identified among 362,200 Caucasian men examined at the mean age of 20 years at the draft boards in Copenhagen and its surrounding areas during 1943-1977. Obesity was defined as 35% overweight relative to a local standard in use at the time (mid 1970's), corresponding to a BMI ≥31.0 kg/m2, which proved to be above the 99th percentile. All of the obese and 50% of the random sampled controls, who were still living in the region, were invited to a follow-up survey in 1992-94 at the mean age of 46 years, at which time the blood samples were taken and genotyping were performed for a total of 673 extremely overweight and 792 controls. With a sampling fraction of 0.5% (50% of 1%), the controls represent about 158,000 men among whom the case group was the most obese.

Raine Study
The Western Australian Pregnancy Cohort (Raine) Study is a prospective pregnancy cohort where 2,900 mothers where recruited between 1989 and 1991. Recruitment took place at Western Australia's major perinatal centre, King Edward Memorial Hospital, and nearby private practices. Women who had sufficient English language skills, an expectation to deliver at King Edward Memorial Hospital, and an intention to reside in Western Australia to allow for future follow-up of their child were eligible for the study. The Raine Study is known to be one of the largest successfully prospective cohorts richly phenotyped at multiple time points over pregnancy, infancy, childhood adolescence, and young adult. The mothers completed questionnaires regarding their children and the children had physical examinations at ages 1, 2, 3, 6, 8, 10, 14, 17, 20 and 22 years.

COPSAC REGISTRY
The COPSAC Registry is part of the Danish birth register comprising childhood asthma cases with data available on birth related traits plus genomic profiling. This cohort contributes in the birthweightgene variant associations necessary for the current study.

COPSAC-CHILDREN
The COPSAC cohort comprises birth weight measures and data combined from two Danish birth cohorts: the COPSAC2000 and the COPSAC2010 studies. The COPSAC2000 cohort is a prospective clinical birth cohort study of 411 children of asthmatic mothers whereas the COPSAC2010 birth cohort is a population based longitudinal clinical study of 700 pregnant women and their offspring.

MESA
The Multi-Ethnic Study of Atherosclerosis (MESA) is a study of the characteristics of subclinical cardiovascular disease (disease detected non-invasively before it has produced clinical signs and symptoms) and the risk factors that predict progression to clinically overt cardiovascular disease or progression of the subclinical disease. MESA researchers study a diverse, population-based sample of 6,814 asymptomatic men and women aged 45-84 at baseline. Thirty-eight percent of the recruited participants were white, 28 percent African-American, 22 percent Hispanic, and 12 percent Asian, predominantly of Chinese descent. Participants were recruited from six field centers across the United States and followed-up five times with an average time period of follow-up of 2 years between each visit. Data from four visits (exam1 to exam5) was used for the analysis. The tenets of the Declaration of Helsinki were followed and institutional review board approval was granted at all MESA sites. Written informed consent was obtained from each participant.

MDC
The Malmö Diet and Cancer study is a population-based cohort with 30,446 adults (62% women; 45-73 years) recruited at baseline in 1991-1996. In a cardiovascular sub-cohort 6103 adults were randomly selected from the parent Malmö cohort. 5040 adults with genotype information and who provided valid dietary information were eligible for the current analysis.

WGHS
The WGHS is a prospective cohort of initially healthy U.S. women. Study participants were health professionals who were age 45 years and older and free of major chronic disease including cancer and cardiovascular disease at study entry (1992)(1993)(1994)(1995). A total of 23,294 had confirmed self-reported European ancestry and genotyping information. There were a total of 12,768 WGHS participants who did not qualify for one or more of the following conditions: were a preterm birth, one of a multiple birth, had missing information for either of these variables, or had diabetes or hypertension at baseline.

PREDIMED-Valencia
The PREDIMED-Valencia study was initiated in 2003 including 1,094 participants. Valencia is one of the field centers participating in the PREvencion con DietaMEDiterranea (PREDIMED) trial. Eligible participants were community-dwelling persons (55-80 years for men; 60-80 years for women) who fulfilled at least one of two criteria: type2 diabetes or 3 or more cardiovascular risk factors. Here we used an observational cohort design and we included 1,023 PREDIMED-Valencia participants with valid genotype data for the analyzed SNPs. Prevalence of T2D was analyzed at baseline. Validated FFQ questionnaires were used at baseline. Weight and BMI were directly measured. The Institutional Review Board of the University of Valencia approved the study protocol. were lost to follow-up and 1,634 children (82.6%) returned for ocular examination at age 11 years. Of the 1,634 children, we further excluded those that were non-myopic (n = 585), with missing data on age of onset of myopia (n = 14) and those with age of onset of myopia at age 11 (n = 107). Thus, the final sample size for the current analysis was 928 (n = 761 Chinese, n = 113 Malays and n = 54 Indians and others). Informed written consent was obtained after the nature of the study was explained to the parents. The tenets of the Declaration of Helsinki were observed and approved by the Singapore Eye Research Institute Ethnics Committee.

STRIP
The STRIP study is a prospective, randomized lifestyle intervention project that began in 1990-1992 when 1,062 infants aged 5 months were recruited to a dietary intervention trial with the main aim of replacing saturated fat with unsaturated fat in the child's diet. The intervention was continued until the age of 20 years and during the follow-up children's diet and other lifestyle factors, growth and biological risk factors have been closely monitored with repeated measurements.

CLHNS
The Cebu Longitudinal Health and Nutrition Survey (CLHNS) is a community-based birth cohort study that originally enrolled 3,327 pregnant women in 1983-84 (3,080 singleton live births), and has since followed them and their offspring to the present. In 2005, 1,895 healthy Filipino mothers and 1,775 of their offspring remained in the study and from whom DNA and measurement of biomarkers were collected. Trained field staff conducted in-home interviews and collected anthropometric measurements and comprehensive environmental data at each visit (data available online at http://www.cpc.unc.edu/projects/cebu/). Blood samples, which were used for biomarker measurement and DNA extraction, were obtained in 2005. Our analysis included 1,798 CLHNS mothers with both genotype and phenotype data. Informed consent was obtained from all CLHNS participants, and the study protocol was approved by the University of North Carolina Institutional Review Board for the Protection of Human Subjects.

RS
The Rotterdam Study I (RS) is a prospective cohort study ongoing since 1990 in the city of Rotterdam in the Netherlands. All inhabitants aged 55 years and over of the Ommoord district in the city of Rotterdam were invited to participate. In the current analyses we included 6,291 participants for whom we had genetic data available. The participants were all examined at baseline. They were interviewed at home (2 h) and then had an extensive set of examinations (a total of 5 h) in a specially built research facility in the centre of their district. These examinations were repeated every 3-4 years in characteristics that could change over time. The Rotterdam Study has been approved by the institutional review board (Medical Ethics Committee) of Erasmus Medical Center and by the review board of The Netherlands Ministry of Health, Welfare and Sports.

NEO
The NEO study was designed for extensive phenotyping to investigate pathways that lead to obesityrelated diseases. The NEO study is a population-based, prospective cohort study that includes 6,671 individuals aged 45-65 years, with an oversampling of individuals with overweight or obesity. At baseline, information on demography, lifestyle, and medical history have been collected by questionnaires. In addition, samples of 24-h urine, fasting and postprandial blood plasma and serum, and DNA were collected. Genotyping was performed using the Illumina HumanCoreExome chip, which was subsequently imputed to the 1000 genome reference panal. Participants underwent an extensive physical examination, including anthropometry, electrocardiography, spirometry, and measurement of the carotid artery intima-media thickness by ultrasonography. In random subsamples of participants, magnetic resonance imaging of abdominal fat, pulse wave velocity of the aorta, heart, and brain, magnetic resonance spectroscopy of the liver, indirect calorimetry, dual energy X-ray absorptiometry, or accelerometry measurements were performed. The collection of data started in September 2008 and completed at the end of September 2012. Participants are currently being followed for the incidence of obesity-related diseases and mortality.

CHOP
The Childhood Obesity Project (CHOP) was conducted as a European multicenter, double-blind, randomized clinical trial that enrolled healthy infants born between October 2002 and July 2004. Formula-fed infants (n = 1,090) were randomly assigned to receive higher protein (HP)-or lower protein (LP)-content formula (within recommended amounts) in the first year of life; breastfed infants (n = 588) were enrolled as an observational reference group. Weight, length, weight-for-length, and BMI were determined at inclusion and at 3, 6, 12, and 24 months of age. The primary endpoints were length and weight at 24 months of age. Anthropometric data have been followed since then biannually up to the age of 6 years and at age 8 and 11 years in this ongoing cohort. Comprehensive nutritional,

WHI
The Women's Health Initiative (WHI)is a large, multi-centre study designed to study major causes of morbidity and mortality in postmenopausal women. The WHI includes a clinical trial (CT) and an observational study (OS) cohort. Women meeting eligibility criteria (age 50-79, postmenopausal, minimum life expectancy 3 years) were recruited at 40 US clinical centres between 1 September 1993 and 31 December 1998. The study sample included 161,808 participants enrolled in the WHI Observational Study and in the three overlapping clinical trials (hormone therapy, dietary modification, and calcium plus vitamin D) prospectively followed for an average of 12 years or until earliest of treated T2DM, death, loss to follow-up, or end of study. Written informed consent was obtained from all study participants before study enrollment, and each of the trials was approved by the Institutional Review Boards of the 40 participating institutions.

YH3 & YH2000
The Young Hearts (YH) project is a prospective study investigating the development of biological and behavioural risk factors for cardiovascular disease in an adolescent population in Northern Ireland. Briefly, in 1989-1990, a 2% representative sample of school children aged 12 and 15 years in Northern Ireland (YH1, n=1,015) was collected. The original 12-year-old population was followed up in 19921993 (YH2) with complete data collected on 225 boys and 230 girls (90% response rate). Between 1997 and 1999, all original YH participants were invited to participate in the third screening phase (YH3, age 2125 years, n= 489), and a blood sample for DNA extraction was taken at that time.
A further cross-sectional survey, the Young Hearts 2000 (YH2000), was carried out in 2000. Approximately 2,000 boys and girls aged 12 and 15 years (500 in each of the four age-sex groups) were recruited through post-primary schools.