Hypospadias and environment: A registry-based case-control study

Abstract

Introduction Monitoring carried out by Remera (“Registre des Malformations en Rhone-Alpes”) since 2009 shows a significantly increasing prevalence of hypospadias. As endocrine disruptors are suspected risk factors for hypospadias and could explain the observed trend, we analyzed the registry data to determine whether factors such as pesticide exposure are involved, using multivariate statistical methods, machine learning and spatial data mining. Materials and methods Cases and controls were identified in the registry. Cases (n = 479) were male children born in Rhone, Isere, or Loire between 2009 and 2015 with an isolated hypospadias (CIM 10 code Q54). Controls (n = 1142) were male children born in the same period and region, with a unique and isolated malformation. Malformations were selected as controls if and only if no correlation with endocrine disruptors exposure is known and if they are independent from genital abnormalities. Multivariate generalized linear models (binomial regressions) were used to estimate adjusted odds ratios (OR) and P-values (P) after univariate regressions and backward selection of variables of interest. Cross-validation, c-statistic and Hosmer–Lemeshow test were used for model validation. Spatial interpolations were computed with inverse distance weighting and k nearest neighbours methods. A statistic was introduced to test the null hypothesis H0: “Hypospadias prevalence is constant on all the region”. Two clustering algorithms were performed either to survey high-risk (e.g., polluted) zones or as post hoc analysis after rejecting H0. Algorithms were based on Kulldorff or hierarchical clustering approach. Results Mothers living in a rural municipality (with more than 40% of utilized agricultural area) at the beginning of their pregnancy had an increased risk to bear boys with hypospadias (OR = 1.42, P = 0.02). The risk also increased if the baby was conceived between May and October (OR = 1.35, P = 0.01). In that case, the first months of pregnancy matched with agricultural spreading of insecticides, fungicides and several herbicides. Moreover, prevalence increased each year (OR = 1.09, P = 0.005) from 2009 to 2015. Risk decreased with parity (OR = 0.79, P = 0.0002). No correlation was shown between assisted reproductive technology and hypospadias (P = 0.4). Hypospadias risk was higher when the boy was born before 32 weeks after last menstrual period (severe prematurity: OR = 2.4, P = 0.002). Neither maternal/paternal age nor medically assisted procreation appeared as risk factors (P > 0.2), which could exclude subfertility as a risk factor. Conclusion Our study based on more than 1500 cases and controls, with a good statistical power and without identified selection bias, confirms the previous observation of a link between the risk of hypospadias and exposure to agricultural pollutants in early pregnancy. As prevalence was not homogeneous across the studied region (P