Use of birth certificates to check completeness added two new cases to the Lombardy birth defects registry. The quality and the completeness of birth certificates is known to vary considerably . We found a large number of false positives and also massive under-reporting, which shows that birth certificates are not a reliable single source for ascertaining congenital malformations . However their availability allowed us to perform a natural experiment to assess our registration system and identify weaknesses in data collection .
We discuss our birth defects prevalence findings in relation to those from three other birth defect registries: Georgia USA , Hawaii  and Finland . There are several reasons for this. All three are large (national or state) long-established population-based national registries, with a history of quality scientific publications. They also use differing data collection methods and cover geographically disparate areas.
The overall prevalence at birth of the 40 selected congenital defects in our study area in 1999 was 204.9/10,000. This compares with 223.9/10,000 for the birth defects registry of Georgia USA in 1995–1999 , 233.6/10,000 for the registry of Hawaii  from 1995 to 1999, and 103.5/10 000 for birth defects recorded in Finland in 1998 .
Cardiovascular system defects were by far the most common birth defects in our population. Prevalence rates for these defects were grossly similar to rates published previously , except for ventricular septal defects (VSD), for which we found particularly high overall rates (58.3/10,000). The latter are nonetheless fairly comparable with those reported by Florida (41.6/10,000), Georgia (35.6/10,000) and Hawaii (43.4/10,000) . The elevated prevalence of VSD in our study areas is probably due to over-diagnosis: routine use of echocardiography on virtually all newborn infants has resulted in the identification of large numbers of small and clinically unimportant VSDs. Subsequent echocardiography results in the inclusion of clinically unimportant defects (that close spontaneously). Progress in clinical management and more frequent prenatal diagnosis have reduced neonatal mortality, increasing the prevalence at birth of these defects, while generally only severe congenital heart defects diagnosed in utero are aborted . These findings indicate a strong need to standardize both diagnostic and registration criteria for congenital heart malformations, as also suggested by Hofmann .
We found rates of 6.7/10,000, 3.3/10,000 and 7.5/10,000 for transposition of great arteries, endocardial cushion defects, and coarctation of aorta, respectively. These figures compare with lower birth prevalences for transposition of great arteries in Finland (4.0/10,000; endocardial cushion defects not reported), and a higher rate for endocardial cushion defect (5.5/10,000) but lower rates for transposition of great arteries (5.4/10,000) and coarctaction of aorta (5.6/10,000) in Georgia. In cases of complete transposition, the aorta arises from the right ventricle and the pulmonary artery arises from the left ventricle, and there is an associated ventricular septal defect, which in all eight of our registered cases was large. Botto et al.  observed an increase birth prevalence of atrial and ventricular defects from 1968 to 1995 in the population of Atlanta but no change in the rate of grand artery transpositions.
We found low birth prevalence rates for tetralogy of Fallot (2.5/10,000) and hypoplastic left heart syndrome (0.8/10,000) compared with the corresponding figures of 3.3/10,000 and 4.7/10,000 in Finland, 4.5/10,000 and 3.2/10,000 in Georgia, and 3.1/10,000 and 1.5/10,000 in Hawaii. Our rate of 2.5/10,000 is comparable with other European prevalence rates of 2.56/10,000 and 2.45/10,000 for Central-Eastern France and Sweden respectively . The study by Hoffman  reviewed 41 studies on cardiac malformation and found prevalence rates for tetralogy of Fallot ranging from 2.91/10,000 to 5.77/10,000, with a mean of 4.21/10,000.
The birth prevalence of hypospadias varies widely (2.0–39.7/10,000) . We found very high rates of hypospadias in accord with data from Georgia (35.1/10,000)  but in stark contrast to the low figure of 2.6/10,000 from Finland .
The study by Pierik et al.  addressed the highly variable hypospadias rates in Holland, studying a cohort of all newborn in Rotterdam over a two year period. They found a high rate (38/10,000) which is comparable to ours. They concluded that the variation in hypospadias birth prevalence in Holland is due both to differences in case ascertainment and real geographic variation.
The comparability of reported rates for hypospadias is poor, however there may be an important underlying geographic variation which cannot be properly addressed until ascertainment and diagnostic criteria (inclusion or exclusion of minor cases) are standardized . Hypospadias could be related to exposure to antiandrogenic factors that interfere with androgen activity . There is also evidence that increases in these birth defects are linked to increased risk of testicular cancer, both of which may be related to exposure to environmental substances having estrogen-like effects . Genetic factors may also predispose to the development of hypospadias .
Our overall figure for Down syndrome was 8.3/10,000. Other published population-based rates for this syndrome are higher  (12.4–22.2/10,000). Down syndrome can be diagnosed relatively easily prior to birth by measuring alphafetoprotein, human chorionic gonadotropin and unconjugated estriol in fetal serum, detecting a thickened nuchal fold on fetal ultrasound, and by cytogenetic analysis. Thus, it is likely that many fetuses with Down syndrome are aborted electively and that variable abortion rates contribute to the variable prevalence at birth rates. This conclusion is supported by data from the Strasbourg birth defect registry, which reported the low rate of 2.2/10,000 live births in 1998. This rate however became 19.7/10,000 when aborted fetuses were considered . The number of Down syndrome fetuses has increased as the mean age of pregnant women has increased .
With regard to nervous system defects, the prevalence rate for spina bifida in our population was lower (1.7/10,000) than found in other registries (3.2–5.2/10,000). For hydrocephalus the combined male-female rate was 5.8/10,000 in our registry, and varied from 3.2 to 11.6 in the other population-based registries . The reasons for these wide variations in rates for nervous system defects are not known, but it is likely that variations in rates of elective terminations contribute substantially. No cases of anencephalus or encephalocele were found, probably in relation to improved prenatal diagnosis and elective termination. Recent surveillance data on neural tube defects show they are diagnosed prenatally in about 80% of cases . In the US, 20–30% of fetuses with neural tube are terminated . Folate intake also varies between populations, and lack of this nutrient is known to cause such defects .
Facial clefts are a heterogeneous group of easy-to-recognize non-fatal birth defects . They are reported as the most common congenital facial abnormalities, with prevalence at birth estimated at 10/10,000  and in the range 6–17 per 10,000 Caucasian births, 4/10,000 for African-Americans and 17/10,000 for Japanese . Our population-based data for facial cleft (overall birth prevalence 10.8/10,000) are in line with these estimates but considerably lower than those for Georgia, (15.8/10,000) and Finland (24.3/10,000). We found that they were about twice as common in boys than girls, which is consistent with other published data . Facial clefts are considered to be etiologically heterogeneous. A small proportion occur as a part of recognizable pattern of malformations or have a genetic etiology , while epidemiologic data suggest that exogenous factors contribute to these conditions. Maternal factors that have been studied for their influence on cleft risk include smoking, alcohol consumption, medication use, environmental chemicals and nutritional factors, but none appear to explain a significant proportion of the population burden of these anomalies . Geographic differences in birth prevalence for these anomalies probably reflect differences in maternal life style and exposure to environmental causative factors.
The birth prevalence for esophageal atresia (4.2/10,000) was similar in our study to that reported in Finland (3.3/10,000), whereas that of small intestine atresia was higher (1.7/10,000) in our population than in the Finnish (0.7/10,000). Prevalence in Georgia, (1968 to 1989) has been reported as 2.8/10,000 overall, but higher in black infants (3.7/10,000) .
Hirschprung's disease is relatively easy to diagnose, so higher birth prevalence in our population (3.3/10,000) compared to Georgia's (2.1/10,000) and Hawaii's (2.2/10,000) is probably due to genuinely higher birth prevalence of this defect in our population. Although this disease is inherited, environmental factors may be responsible for sporadic cases .
The birth prevalences of unilateral and bilateral renal agenesis have been reported at 10/10,000 and 2.5/10,000 respectively . At the same time, a study on the epidemiology of kidney malformations reported marked variations in the birth prevalence of renal agenesis (0.6–29 per 10,000), which is considered to be mainly due to variations in the diagnostic procedures used . Our overall figure (3.3/10, 000) was somewhat lower than the one for Georgia (4.4/10,000), while Finland reports a very low birth prevalence of this condition (0.7/10,000) in relation to a high rate of terminations of pregnancies with it: the rate for these defects increases to 2.43/10,000 when terminations are considered. We had no cases of bladder exstrophy; birth prevalences of this condition were also low in Finland (0.2/10,000) and Georgia (0.1/10,000).