Our results revealed significant spatial heterogeneity in the incidence of hypospadias in a region of northern France during the period from 1999 to 2012. We identified two spatial clusters. These results are consistent with previous research performed in northern England, North Carolina, and Nova Scotia.[27–29] When comparing the detected spatial clusters, we found a significant difference in their socio-ecological pattern. Lastly, these findings remained significant after exclusion of patients with a known potential confounding factor (i.e. potential bias for spatial analysis).
Strengths of the study. Because most of the spatial analysis studies are based on massive administrative datasets, they usually lack clinical information, especially specific information related to the disease of interest such as – in our situation – familial history of hypospadias. In our work, we took into account of clinical information about potential confounding factors associated with a higher risk of hypospadias, and we performed a systematic approach for spatial analysis.
Furthermore, the registry included all types of hypospadias - even minor types not requiring surgery (not listed in hospital episodes statistics) and those diagnosed after the child had left the maternity unit (not listed in maternity based birth defect monitoring system).
The geographical unit is an important parameter in spatial analysis. We used a French local administrative unit called a canton (170 cantons over 12,414 km2); this is quite small, relative to the literature studies. Other studies have used larger geographical units, such as the location of the referring hospital (159 hospitals across six South American countries), the county (18 Canadian counties over 55,284 km2), and the region (four Chinese regions over 9,500,000 km2), which limits the ability to interpret environmental factors.
Limitations of the data. In the absence of a French national registry of hypospadias cases, we chose to perform a single-center registry study based on a single referral surgeon for hypospadias in our region’s university hospital. However, the fact that surgeon also consulted in three general hospitals across the region broadened our study’s coverage.
Our data showed that 975 cases of hypospadias were observed among the 393,341 male newborns recorded over the 13-year study period. Thus, the calculated prevalence and annual incidence rate were 24/10,000 male births, and 17.67/10,000 male births, respective. According to the literature, the estimated overall prevalence of hypospadias is around 18.61/10,000 total births (i.e. male and female births) in Europe and 15.41/10,000 total births (i.e. male and female births) in France. These values suggest that we might have underestimated the regional prevalence of hypospadias; some patients might have been seen by a urologist in a general hospital elsewhere in the region. However, the breakdown in the types of hypospadias was in line with the literature data.
With regard to the statistical methodology, we chose to study high-incidence areas only because our objective was to find ecological factors associated with a higher risk of hypospadias. In principle, missing patients could produce “false positive” low-incidence areas or “false negative” high-incidence areas. Hence, the analysis of high-risk area is less biased by missing patients than the analysis of low-risk areas. Unfortunately, missing patients will inevitably constitute a source of bias in spatial analysis unless the data come from a truly exhaustive registry in a geographically isolated region.
With regard to geographical data, we used the zip code at the time of delivery or at the time of the first consultation in our hospital (usually within 12 months of birth for patients with hypospadias) (Table 1). Ideally, we would have analyzed the mother’s zip code before pregnancy and during the first trimester of pregnancy; however, this information was not available retrospectively. Miller et al. recently evaluated birth defects and residential mobility during pregnancy in the USA. They found that 22% of pregnant women moved during pregnancy, and that 51% of these women moved within the same county. There was no difference in residential mobility between mothers of children with a birth defect and control mothers. In their review of mobility during pregnancy, Bell et al. reported that pregnant women typically moved a short distance only (less than 10 km), which would tend not to greatly change their environmental exposure.
Interpretation. Our results evidenced spatial heterogeneity, spatial autocorrelation, and presence of spatial clusters of hypospadias incidence that remained statistically significant after the exclusion of cases with potential confounding factors. This result reinforces the conclusions of previous spatial analyses of hypospadias - none of which took account of potential confounding factors.[27–29, 32]
In 2007, Abdullah et al. reported on significant spatial clustering among 577 cases of hypospadias in northern England (based on hospital episode statistics). Using data from a birth defect monitoring program, Winston et al. showed significant spatial clustering among 995 cases of hypospadias in North Carolina; the researchers took account of data on maternal age, maternal race/ethnicity, maternal educational level, smoking status, parity, and diabetes but not the family history and other potential confounding factors. Recently, Lane et al. published a comparison of the spatial distribution of four different congenital anomalies, of which two were endocrine-mediated (hypospadias and cryptorchidism) and two were not (clubfoot and gastroschisis). Lane et al. detected significant spatial autocorrelation and clustering for hypospadias and cryptorchidism congenital anomalies but none for clubfoot and gastroschisis.
Our comparisons of spatial cluster with regard to ecological variables revealed significant differences between high-incidence cantons and neutral cantons. The first spatial cluster was characterized by a rural land use pattern, with a higher proportion of rural (and agricultural) land cover, and a lower deprivation index than neutral cantons (i.e. less deprived). The second cluster had a more urban and industrial pattern, with higher proportion of artificialized land cover and a higher deprivation index (i.e. more deprived) than neutral cantons. It should be noted that the region’s main city (Lille, where our university hospital) fell outside this “urban” cluster.
Abdullah et al. showed an association between hypospadias and a lower deprivation index but not with the UK wards’ urban/rural status. In their study of North Carolina, Winston et al. found a high-risk area with > 5% crop cover. Lane et al. also mentioned that their hotspots in Canada were associated with intense agricultural activity but did not underpin this comment with statistical results. In China, Li et al. assessed 3,426 cases of hypospadias recorded in a hospital-based birth defect monitoring system. The researchers found that the prevalence rate was higher in urban areas than in rural areas but that it was increasing more rapidly in rural areas. However, Li et al. did not specify how each area had been classified as either rural or urban.
In the present study, there were few differences between the spatial clusters with regard to clinical data. The proportion of preterm births was lower for hypospadias cases from the “rural” cluster (North-West) than for cases in the “urban” and most deprived cluster (Center-East) and the neutral cantons.