400 likes | 579 Views
Birth Weight and Childhood Cancer and Leukemia. Update from the I4C Environmental Working Group on Birth Weight and Childhood Cancer. Ora Paltiel, Hadassah-Hebrew University, School of Public Health, Jerusalem , Israel. Lyon, October 2013. Results. Background- Big babies and leukemia.
E N D
Birth Weight and Childhood Cancer and Leukemia Update from the I4C Environmental Working Group onBirth Weight and Childhood Cancer Ora Paltiel, Hadassah-Hebrew University, School of Public Health, Jerusalem , Israel Lyon, October 2013
Determinants of birth weight SES ethnicity Smoking Gestational diabetes Maternal height Pre-pregnancy BMI Gestational weight gain • Gestational age • Birth length • Child gender • Altitude • Birth order
Beyond BWT – Fetal Growth- BWT corrected for gestational age Birth certificate data of 2,254 children with cancer <5 years old at diagnosis and registered at Texas Cancer Registry 1995-2003 were compared to 11,734 age-matched controls. • Using model diagnostics, the model containing BW corrected-for-gestational age was a better predictor than the model with BW alone
New case for action…. >30% BMI>29 kg/m2 Prevalence of obesity among pregnant women >50% above IOM recommendations
Stated Aim Of I4C study • To investigate the association between birth weight (BW) and other measures of fetal growth and childhood cancer, specifically leukemia, with specific attention to determinants of BW such as maternal obesity, weight gain in pregnancy, pregnancy complications, in a pooled analysis of childhood cancer cohorts.
Specific objectives • To examine the pattern of the association between BW and other measures of fetal growth and the risk of childhood AML, ALL, all leukemia, other cancers: U, linear, threshold effect, and examining BW both as a continuous and categorical variable • To examine these associations in specific age groups: Infant (up to age 1 year); Early childhood (1-4); Later childhood (5-9); Early adolescence (10-14), controlling for and in strata of maternal pre-pregnancy BMI, and weight gain during pregnancy.
Participating cohorts MOBA DNBC ALSPAC CPP JPS TIHS
Original data analysis strategy • Exclude multiple births and Down syndrome • Include all cancer cases besides above • Case-cohort design with subcohort approx 1:10 for cancer/leukemia cases • Cox regression • Report HR and 95% CI • Always adjust for study • STATA
Decisions along the way Not to analyse AML or solid tumor subtypes JPS, choose only from sub-cohort with gestational age adjust all analyses for GA Remove all postnatal exposures from the analysis Report on covariates when reported in >5 cohorts Report on BW in categories; top decile; continuous
TABLE 1: Descriptive characteristics of sub-cohorts in pooled database
Table 2: Distribution of All Cancers, Leukemia and ALL by cohort, gender and age of diagnosis Based on singleton births, Down Syndrome excluded
Main Findings1. Linear association of all outcomes with BWadjusted for GA ALL All leukemia All cancer
BW associations with leukemia ALL Cancer All leukemia
Relationship between BW and CC, Leukemia, and ALL Effect modification? Across gender?- no difference, not shown Across age groups (at diagnosis)? Across study? Across strata of maternal BMI? Across strata of weight gain in pregnancy?
Effect modification- age at diagnosis: HR HR All cancer Leukemia ALL Age at dx. 0-<5 Age at dx. 5-15
No effect modification by maternal anthropometrics (all cancer) Pre-pregnancy BMI • <25kg/m2 • HR 1.33 (1-3.59) • >25 kg/m2 • HR 1.24 (0.8-1.92) (Continuous per kg( Pregnancy weight gain • <16kg • HR 1.29 (0.95-1.77) • >16 kg: • HR 1.27 (0.79-2.03) (Continuous per kg(
Issues raised at and after last I4C • Use of case cohort design; not taking full advantage of the rich data available in the cohorts by sampling non-cases instead of using a cohort approach? • Missing data • Heterogeneity • Small sample size (few events) in some birth weight categories esp. low birth weight • Novelty
Covariates: Example
Adjusted analyses • Cancer hazard ratio is adjusted for gestational age, maternal age, paternal age (rescaled as quadratic), maternal height, first born, and maternal pre-pregnancy BMI. • Leukemia hazard ratio is adjusted for gestational age, maternal age, total pregnancy weight change, maternal pre-pregnancy BMI, first born, and any maternal smoking. • ALL hazard ratio is adjusted for gestational age, paternal age (rescaled as quadratic), total pregnancy weight change, and any maternal smoking. • Non leukemia cancer hazard ratio is adjusted for gestational age, paternal age (rescaled as quadratic), maternal height, total pregnancy weight change, first born, and maternal pre-pregnancy BMI.
New Results- Birth weight and cancer outcomes Reference category
New results 4000
New results 2Hazard ratio is for a one kilogram increase in birth weight
Relationship between BW and CC, Leukemia, and ALL • Is it consistent: • Across age groups (at diagnosis)? • Across study? • Across strata of maternal BMI? • Across strata of weight gain in pregnancy?
Is there effect modification related to maternal pre-pregnancy obesity?
Effect modification: Age at diagnosis, multinomial regression
Coherence with literature?High Birth weight (categorical) and childhood leukemiaCaughey and Michels International Journal of Cancer 2009 1.36 (1.24-1.49) I4C 1.19 (0.74-1.93)
Birth weight per kg and childhood leukemia: Millions of observations 1.15 (1.13-1.23) International Journal of CancerVolume 124, Issue 11, pages 2658-2670, I4C 1.16 (0.81-1.67) Samuelsen (Epidemiology 2009) Norwegian Cancer Registry including 1,842,113 live-born infants born 1967 -1998 : increase in leukemia risk of 29% per 1000g increase in birth weight after adjustment for gestational age
High Birth Weight and Childhood ALL 1.24 (1.16-1.33) I4C 1.2 (0.72-2.02) International Journal of CancerVolume 124, Issue 11, pages 2658-2670,
Summary • Pooled adjusted analysis confirms BW-cancer and BW- leukemia association, but most associations NS. • Possible effect modification by age at diagnosis with stronger findings above age 5 yrs (but small numbers limit robustness) • Heterogeneity among cohorts dealt with in analysis • Maternal obesity and weight gain do not appear to alter the association (except for >3000 gm category) • Comparable to literature (more or less)
Questions remaining • Which cut-off to use (3000, 3500, 4000 g)? • Loss of information on LBW • Weight change≠ weight gain • Effect of child gender?
Major limitations • Small sample size –still! • Limited power for MV analysis • Validity of imputation • Inconsistencies in measures (parity, paternal age) • Uncertainty regarding follow up and cancer ascertainment in some cohorts (CPP)
Gabriella Tikellis • Terry Dwyer • Stan Lemeshow • Gary Phillips • Karen Lamb • Working group members • All participating cohorts • Mothers and babies