Familial Aggregation and Heritability of Pyloric Stenosis FREE

Pyloric stenosis is a severe and potentially fatal condition in which apparently healthy infants, typically 2 to 8 weeks old,^1- 5 develop an inability to pass food from the stomach into the duodenum. The condition is characterized by hypertrophy of the pylorus smooth circular muscle layer. Mortality was high (50%)⁶ until successful treatment by pyloromyotomy was developed by Ramstedt⁷ in 1911. Today pyloric stenosis is the most common condition requiring surgery in the first months of life.⁸ Among white individuals, pyloric stenosis is relatively common,^{2- 3,5,8- 10} with an incidence of 1.5 to 3 per 1000 live births.^9- 10

Other than male sex,^2- 3,9- 11 the most consistently reported risk factors for pyloric stenosis are a family history of pyloric stenosis^{2- 4,8,10,12- 21} and being a firstborn child.^2- 4,11,22 The epidemiological features of skewed sex distribution and clustering of cases within families suggest a genetic component to the etiology. However, it remains to be determined whether aggregation in families is mostly due to genes, maternal factors operating during in utero development, or a common family environment.

In this large cohort study, we examined the familial aggregation of pyloric stenosis from monozygotic twins to fourth-generation relatives in the entire population of Denmark. We furthermore evaluated familial aggregation according to sex of cohort member, sex of relatives, and maternal and paternal contributions. We also estimated disease heritability.

Familial aggregation of pyloric stenosis was analyzed in a large cohort consisting of all children born in Denmark between 1977 and the end of 2008. The cohort was identified based on information from the Danish Civil Registration System. Since April 1968, the Danish Civil Registration System has registered sex, date and location of birth, identity of parents, and updated information on vital status and emigration using a unique personal identification number assigned to each Danish resident.²³ This number permits accurate linkage of individual-level information between the nationwide registries in Denmark.

The Danish Family Relations Database established at Statens Serum Institut, Copenhagen, is based on parent-child links registered in the Civil Registration System.²⁴ Most individuals born in Denmark since 1950 have links to their parents. Therefore, the Danish Family Relations Database can identify parents, siblings, and half-siblings residing in Denmark for nearly all persons born in Denmark since 1950. Using the parental links of parents, it is also possible to identify grandparents and, thereby, cousins and half-cousins. Ninety-four percent of children born after 1990 have a known grandparent, whereas 64% of children born between 1977 and 1989 have a known grandparent.

Twins were identified as individuals having the same mother and being born on the same day (±1 day on either side of midnight). We distinguished between monozygotic and dizygotic twin pairs by information from the Danish Twin Registry.²⁵ In this registry, the zygosity of the twins was determined based on a questionnaire on physical similarity. It has previously been documented that this method is very reliable, with less than 5% misclassification.²⁶

Information on pyloric stenosis was obtained by linkage with the Danish National Patient Registry, a nationwide registry of all hospital discharge diagnoses and operations performed since 1977.²⁷ Pyloric stenosis cases were defined as children who, in their first year of life, had a pyloromyotomy according to the Danish Classification of Surgical Procedures codes up to December 1995 (International Statistical Classification of Diseases, Eighth Revision codes 41840, 41841, 44100) and the Nordic Classification of Surgical Procedures codes after January 1996 (International Statistical Classification of Diseases, 10th Revision codes KJDH60, KJDH61) (see http://www.sst.dk).²⁷

The familial aggregation of pyloric stenosis was evaluated by (incidence) rate ratios (RRs). The RR is the ratio between the rate of pyloric stenosis among children with an older affected relative (exposed) and the rate of pyloric stenosis among children whose known older relatives of the same type are all unaffected (nonexposed). For example, the RR for siblings was estimated as the rate of pyloric stenosis in children who had an older sibling with pyloric stenosis compared with the rate of pyloric stenosis among those having only unaffected older siblings. We compared only individuals with the same type of familial relationship to reduce bias due to incomplete registration of more distant family members in early birth cohorts and to further adjust the RRs for the effect of having a specific relative. We considered only older relatives as exposure to ensure that affected pairs contributed only once. For family history, first-degree relatives were defined as older siblings, second-degree relatives as older half-siblings, third-degree relatives as older cousins, and fourth-degree relatives as older half-cousins.

Except in the analyses of twins, RRs were estimated using log-linear Poisson regression adjusted for sex, age at pyloric stenosis (0-<1, 1-<2, 2-<3, 3-<4, 4-<6, and 6-<12 months), and calendar period (2-year intervals from 1977 to 2008). The follow-up period was from birth to age 1 year, death, emigration, pyloric stenosis diagnosis, or the end of 2008, whichever occurred first. In additional analyses, we furthermore adjusted for number of older relatives (eg, in the analysis of having affected siblings, we adjusted for number of older siblings; in the analysis of having affected half-siblings, we adjusted for number of older half-siblings).

Because data on zygosity from the Danish Twin Registry were available in aggregated tables, RRs for twins were approximated by risk ratios. Risk ratios for twins were estimated as the proband-wise concordance rate of pyloric stenosis divided by 1 minus the proband-wise concordance rate of not having pyloric stenosis. The 95% confidence intervals (CIs) were estimated by log-linear binomial regression.

Comparisons of 2 RRs that were both estimated by Poisson regression were performed using Poisson regression as follows: We estimated RRs according to family history stratified by sex of the cohort member by analyzing sex interactions. The homogeneity of RRs according to sex was tested by interaction tests. Furthermore, RRs according to family history by sex of affected relatives were estimated by including the effect of having both a male and a female relative in the same model, and the RRs were compared by testing for homogeneity of the 2 sex-specific RRs. The same approach was applied when testing for homogeneity of RRs for maternal and paternal relatives. All tests were likelihood ratio tests.

Comparisons of 2 RRs that were not both estimated by Poisson regression were performed using inverse variance–weighted linear regression; ie, by a meta-analytic approach wherein the logarithm of the estimated RR was analyzed with the inverse of the square of the standard error of log(RR) as the weight.²⁸ Thus, this approach was used when evaluating homogeneity of the RR for having an affected monozygotic twin vs the RR for having an affected dizygotic twin and when evaluating homogeneity of the RR for having an affected dizygotic twin vs the RR for having an affected sibling. The difference between the estimated RRs is expressed as the ratio between the RRs, with a 95% CI to reflect the power of the comparisons. The power of the study to detect the observed difference between the RR for having an affected monozygotic twin vs the RR for having an affected dizygotic twin is 68%. The power of the study to detect the observed difference between the RR for having an affected dizygotic twin vs the RR for having an affected sibling is 12%. It should be noted that the post hoc power is reflected in the reported CIs for the ratios.

Log-linear Poisson regression, log-linear binomial regression, and inverse variance–weighted linear regression were all performed within the framework of generalized linear models.²⁹ Accordingly, estimation was conducted by PROC GENMOD in SAS software, version 9.1 (SAS Institute Inc, Cary, North Carolina), using the Poisson, binomial, and normal distributions, respectively. All tests were 2-sided, with P<.05 indicating statistical significance.

Heritability analyses were based on twins of known zygosity. Using the liability threshold model, heritability was estimated by structural equation modeling techniques.³⁰ In the model, liability is considered a latent underlying trait consisting of both genetic and environmental factors. The genetic factors are divided into additive (A) and dominant (D) factors, whereas the environmental factors are classified as shared (C) and nonshared (E). Four component models were run to estimate variance components for factors ACE, ADE, AE, and CE. Based on the Akaike Information Criterion³¹ and the likelihood ratio test, the AE model, which consists of an additive genetic component and a nonshared environmental component, was determined to be the best-fitting model. Estimation of heritability is therefore based on this model.

The study was approved by the Danish Data Protection Agency. According to Danish law, pure registry studies are exempted from ethics committee evaluation and do not require consent from study participants.

The cohort of 1 999 738 children was followed up for the first year of life (1 948 616 person-years), during which 3362 children (1.7 per 1000 person-years) had surgery for pyloric stenosis. Of these, 2741 (81.5%) were boys, resulting in a male-to-female ratio of 4.4 to 1. The overall rate of pyloric stenosis was 2.7 per 1000 person-years for boys and 0.7 per 1000 person-years for girls. The cohort consisted of 1 939 276 singletons and 60 462 twins. Of the singletons, 995 773 were boys and 943 503 were girls, and of the twins, 30 827 were boys and 29 635 were girls. The rate of pyloric stenosis was 3.1 per 1000 person-years for twins and 1.8 per 1000 person-years for singletons. The proportions of children with known siblings, half-siblings, cousins, and half-cousins were 47.9%, 17.7%, 53.0%, and 12.6%, respectively. In the cohort, 98 children had at least 1 older family member with a history of pyloric stenosis: 15 twins (6 monozygotic, 3 dizygotic, and 6 of unknown zygosity), 43 siblings, 9 half-siblings, 28 cousins, and 3 half-cousins.

Table 1 shows the RRs of pyloric stenosis in the cohort according to family history of pyloric stenosis by type of relative. The overall RR of pyloric stenosis for twins was 67.7 (95% CI, 40.5-113). The RR was 182 (95% CI, 70.7-467) in monozygotic twins, 29.4 (95% CI, 9.45-91.5) in dizygotic twins, and 69.1 (95% CI, 30.7-155) in twins of unknown zygosity. Among children with an affected monozygotic twin, 46% were themselves diagnosed as having pyloric stenosis. In dizygotic twins, 7.7% with an affected twin had pyloric stenosis. The RR of pyloric stenosis was 18.5 (95% CI, 13.7-25.1) for siblings (first-degree relatives), 4.99 (95% CI, 2.59-9.65) for half-siblings (second-degree relatives), 3.06 (95% CI, 2.10-4.44) for cousins (third-degree relatives), and 1.60 (95% CI, 0.51-4.99) for half-cousins (fourth-degree relatives).

The overall pyloric stenosis rates per 1000 person-years among children with affected relatives were 25.7 for those with affected siblings, 8.4 with affected half-siblings, 4.6 with affected cousins, and 2.6 with affected half-cousins. The RR for monozygotic twins was significantly higher than the RR for dizygotic twins, with a ratio between RRs for monozygotic twins and dizygotic twins of 6.19 (95% CI, 1.41-27.09; P = .02). The RRs for dizygotic twins and siblings were not significantly different, with a ratio between RRs of 1.59 (95% CI, 0.49-5.15; P = .44).

We also estimated the RRs according to family history of pyloric stenosis by sex of relatives and sex of cohort members (Table 2). Siblings of affected boys and girls had RRs for pyloric stenosis of 16.7 (95% CI, 11.7-24.0) and 29.8 (95% CI, 17.2-51.6), respectively, revealing a nonsignificant difference (P = .19). For half-siblings, the RRs were 5.86 (95% CI, 2.91-11.8) and 5.46 (95% CI, 2.59-11.5) (P = .69), and for cousins, the RRs were 2.83 (95% CI, 1.84-4.36) and 3.95 (95% CI, 1.88-8.31) (P = .49), respectively. The RRs for male and female siblings of affected children were 18.8 (95% CI, 13.4-26.3) and 17.9 (95% CI, 8.85-36.3), respectively (P = .91). Thus, the increased risk for brothers of affected siblings is the same as for sisters of affected siblings. The same was true for male and female half-siblings of affected children (RRs, 5.60 [95% CI, 2.79-11.3] and 2.84 [95% CI, 0.40-20.4], respectively [P = .49]) and for male and female cousins of affected children (RRs, 3.09 [95% CI, 2.05-4.67] and 2.91 [95% CI, 1.20-7.05], respectively [P = .90]).

Table 3 shows the RRs of pyloric stenosis according to family history of pyloric stenosis in maternal and paternal relatives. We found no difference in the risk of pyloric stenosis for maternal and paternal second- and third-degree relatives. The RR for half-siblings who shared the same mother was 5.04 (95% CI, 1.88-13.5) compared with 5.01 (95% CI, 2.07-12.1) for half-siblings who shared the same father (P = .98). Among third-degree relatives, the RRs were 2.54 (95% CI, 1.44-4.50) for maternal cousins and 3.62 (95% CI, 2.21-5.93) for paternal cousins (P = .35). There were too few cases of fourth-degree relatives with affected paternal and maternal half-cousins to perform meaningful analyses.

In additional analyses restricted to male relatives, we found no pronounced differences compared with the results in Table 1 and Table 3: the RR among boys having an affected brother was 17.5 (95% CI, 11.8-26.0), the RR among boys having an affected paternal half-brother was 5.45 (95% CI, 1.74-17.1), and the RR among boys having an affected paternal male cousin through an uncle was 4.64 (95% CI, 2.19-9.81). Parallel analyses in female relatives were not feasible due to small numbers.

Heritability analyses identified the AE model as the best-fitting variance component model. The heritability estimate obtained by this model was 0.87 (95% CI, 0.72-0.95) (eTable 1).

We performed several additional analyses to assess the robustness of the results. We estimated the RR for the first-, second-, third-, and fourth-degree relatives by comparing concordance rates instead of incidence rates as in the main analyses and obtained similar results (eTable 2). We also tested a wider case definition, allowing for not only children with an operation code but also children with only a diagnosis code for pyloric stenosis. This extended case definition included 4062 children, of whom 82.8% had an operation code. The RRs in this analysis were only slightly lower than the RRs in Table 1 (eTable 3).

We also performed stratified analyses based on children born before (1900 cases) and after (1467 cases) 1990 because for birth cohorts born after 1990, there is almost complete registration of grandparents and, thereby, cousins (eTable 4). To evaluate the possible consequences of the change in classification of surgical procedures, we also performed stratified analyses for children born before (2575 cases) and after (792 cases) 1996 (eTable 5). For both analyses, the RRs in the 2 strata were similar.

We performed analyses with additional adjustment for number of older relatives and observed similar results (eTable 6).

This nationwide study documented strong familial aggregation of pyloric stenosis, with a nearly 200-fold increase among monozygotic twins and 20-fold increase among siblings. Familial aggregation of pyloric stenosis was pronounced even in more distant relatives. The aggregation was observed on both the maternal and paternal side of the families and irrespective of sex of cohort member and sex of relative. Heritability was estimated to be 87%.

Previous studies in twins^{3- 5,12,20- 21,32- 34} comprise mostly case reports. In some studies, monozygotic twins have been concordant for the disease at a frequency not much higher than that of dizygotic twins,^2,4- 5,33 whereas other studies have indicated a higher concordance rate in monozygotic than dizygotic twins.^{3- 4,10,12,21,35} However, these studies appear too small to be conclusive. In our nationwide study, a child with an affected monozygotic twin (RR = 182) was at 6-fold higher risk of pyloric stenosis than a child with an affected dizygotic twin (RR = 29.4).

Studies dealing with aggregation of pyloric stenosis among siblings are available^{2,10,12- 13,18,36- 37} but are all smaller in sample size. A combined reanalysis² of 4 such studies^{12,18,36- 37} found 51 affected infants (5.8%) among 885 siblings. This analysis documented a similar risk as a previous reanalysis¹⁰ that, by using pooled data on affected siblings from several family studies and assuming a population prevalence of 0.3% for pyloric stenosis, found the RR for siblings to be 18.^10,15 This estimate agrees closely with the RR of 18.5 for siblings observed in our study.

Dizygotic twins and siblings represent a similar degree of inheritance and the aggregation in dizygotic twins and siblings was not found to be statistically different in our study. Thus, dizygotic twins who shared an intrauterine environment during the same pregnancy do not seem to be at higher risk of pyloric stenosis than other siblings, suggesting that pyloric stenosis risk is not strongly influenced by in utero environmental exposures.

Studies of half-siblings with pyloric stenosis can be informative in evaluating maternal and paternal contributions to development of pyloric stenosis. However, reports on this topic have been few and small in size.^2,12- 14,18 We found that half-siblings had the same increased risk of pyloric stenosis irrespective of whether they were related by sharing the same mother or father. If the intrauterine milieu were a strong risk factor for this condition, we would expect a higher risk association for half-siblings who shared the same mother and uterus but not the same father. Furthermore, if only postnatal factors (most often related to the mother in this early risk period) played an important role, one would expect siblings and maternal half-siblings (as well as cousins and half-cousins) to have a more similar risk than we observed.

Two studies evaluating a possible increased risk among cousins of pyloric stenosis cases have been conflicting.^13,38 We observed a statistically significant 3-fold increased risk in cousins of both sexes. To our knowledge, there are no reported studies on fourth-generation relatives. We found a 60% increased risk in half-cousins. Although this increase was not significant, it suggests aggregation in even very distant relatives.

Our findings argue for a hereditary component of pyloric stenosis: (1) predominance in boys; (2) familial aggregation in first-, second-, and third-degree relatives; (3) high concordance rate in monozygotic twins; (4) similar degree of aggregation in dizygotic twins and siblings; (5) difference in risk for siblings vs maternal half-siblings (as well as cousins vs half-cousins); and (6) heritability of 87%. However, the condition does not follow classic mendelian modes of inheritance. A multifactorial threshold model has previously been discussed that suggests that pyloric stenosis is caused by polygenic inheritance of genes that are modified by sex and environmental factors.^10,12 The model assumes that the “liability” to pyloric stenosis is determined by the additive effect of numerous genetic and environmental factors and that the condition is expressed when an individual's liability exceeds a critical threshold value. If pyloric stenosis is inherited as a multifactorial threshold trait, the risk for relatives of affected girls is expected to exceed the risk for relatives of affected boys. Furthermore, male relatives are expected to be at increased risk compared with female relatives. We did not observe such differences.

Discordance in the development of pyloric stenosis in identical twins causes other factors to be considered. Within the monozygotic group, we found 54% of affected twin pairs to have only 1 affected individual. In combination with the decrease in incidence in the last decades,^{1- 2,9,39- 41} this suggests that, while carrying the gene(s) increases the propensity for pyloric stenosis development, environmental factors appear to be needed for the disorder to manifest.^10,12,21 Possible postnatal environmental (nongenetic) factors include feeding practice,⁴² maternal smoking,^2,41 infant sleeping position,^2,9 and postnatal use of macrolides.^43- 44 Along these lines, we speculate whether our finding of a higher absolute risk in twins compared with singletons could be explained by differences in exposure to these factors.

Our study had several strengths. First, it was based on a large, population-based cohort of 2 million children. Second, family history was recorded prior to and independent of a diagnosis of pyloric stenosis, which ruled out differential misclassification. The cohort was based on information from the Danish Civil Registration System, which has been proven close to complete.²³ The validity of specific data in the Danish National Patient Registry varies by procedure type, but surgical diagnoses are likely to be both accurate and well recorded.⁴⁵ Pyloric stenosis is a life-threatening condition if the child is not treated, which makes it unlikely that some cases escaped attention. Finally, health care in Denmark is free to all of its citizens, making it unlikely that socioeconomic differences should dictate treatment practice or lead to a differential recording of cases in the national registries.

The study also had limitations. First, we had a relatively small number of cases, leading to wide and in some situations overlapping CIs. Registration of pyloric stenosis in the Danish National Patient Registry started in 1977, so information on surgeries prior to 1977 is unavailable, limiting our ability to examine pyloric stenosis occurrence in parents. However, we examined the risk of pyloric stenosis for twins, siblings, half-siblings, cousins, and half-cousins, for which this limitation would be of lesser importance. Children born prior to 1990 had incomplete links to grandparents and, thereby, cousins. We took this incomplete registration into account by performing analyses only among children with a registered relevant relative. Furthermore, we found no differences in separate analyses involving children born before and after 1990.

In conclusion, pyloric stenosis strongly aggregates in families. The very high concordance rate among monozygotic twins and the strong aggregation even in more distant relatives argue for an important genetic contribution to its etiology. The similar importance of having affected maternal and paternal half-siblings, together with our observation of a similar risk for dizygotic twins and siblings, indicates that intrauterine environmental factors may have little role in causing pyloric stenosis. Thus, with a heritability estimate of 87%, it seems that familial aggregation is primarily explained by shared genes that may affect responses to postnatal factors. The high rates for twins and siblings should be considered in counseling families with affected children.