Ten-Year Survival of Children With Congenital Anomalies: A European Cohort Study

Initially, 21 population-based EUROCAT registries agreed to participate in the EUROlinkCAT project.¹⁷  Three registries were unable to obtain linked data within the given time frame, whereas the data linkage in 3 other registries was not considered of sufficient quality.¹⁸  An additional registry (Belgium: Antwerp) was not included in this analysis because it did not provide death data beyond infancy and for some specific CAs because of their country’s restrictions on releasing small numbers. Survival of children from the only EUROCAT registry in Eastern Europe (Ukrainian OMNI-Net) was considerably lower compared with all other registries. Because childhood mortality is higher in Eastern than in Western Europe,¹⁹  and because OMNI-Net was the only registry from Eastern Europe, it was decided to limit the analysis to Western European registries.

All live-born children with a major CA born between January 1, 1995, and December 31, 2014, recorded in the 13 registries were linked to mortality sources up to the child’s 10th birthday or to December 31, 2015, whichever was earlier. Given observed increases in survival for births from 2005 to 2014 compared with 1995 to 2004 (Santoro M et al, submitted for publication on July 20, 2021) and improved linkage quality in the later decade,¹⁸  this study restricted the analysis to births between 2005 and 2014 (2007–2014 for the Valencian Region; 2008–2014 for Emilia Romagna) (Table 1) to provide the most up-to-date survival estimates.

The EUROCAT registries have ethics permissions and procedures for routine surveillance, data collection, and transmission of anonymized data to a central database, according to national guidelines. Twelve CA registries sought local ethics approvals or other permissions to link their data with local mortality sources; 1 registry (Norway) obtained permission to use data they had already linked.

Registries linked their CA data to either national/vital statistics or to mortality records only. Linkage to national and vital statistics provided information on the vital status of all linked children (dead or alive) and, hence, a measure of successful linkage. Conversely, only registered deaths could be ascertained from mortality records; that is, children without death certificates were assumed to be alive, although it could have been a linkage failure. A detailed description of the linkage process and results is provided elsewhere.¹⁸  Data were only included in this paper from those registries where the linkage success was >85% for all years; for 5 registries, it was ≥99%.

All major CAs were coded using the International Statistical Classification of Diseases and Related Health Problems, 10th Revision²⁰  or 9th Revision, and categorized by CA group and subgroup (the organ system affected and the individual disorder) following the EUROCAT guidelines.²¹  Children with only minor anomalies, defined in EUROCAT as those with lesser medical, functional, or cosmetic consequences for the child (eg, clinodactyly), were not included.²²  For each CA subgroup, all children with the specified anomalies were included and those with an isolated anomaly only were identified, resulting in 2 groups for analysis: “All” and “Isolated.” An isolated CA was defined as a structural CA in one organ system only or as part of a known sequence (eg, renal agenesis with pulmonary hypoplasia). A child classified as having an isolated anomaly may be included in >1 anomaly subgroup within the same organ system (eg, esophageal atresia and anal atresia). The EUROCAT hierarchical computer algorithm for classification of major CAs was used^23,24  without a manual clinical review of the identified potential multiple CAs.

This article is focused on relatively common structural CAs in live births (live birth prevalence ≥1 per 10 000) and Down syndrome as the most common chromosomal anomaly (Supplemental Table 4).¹⁷ 

In addition to the standardized EUROCAT variables,¹⁷  a common data model was developed to standardize the local variables obtained from linkage. This enabled centrally written syntax scripts for checking the linkage quality and for the analysis of mortality data to be run by all registries.¹⁷  The statistical analysis consisted of 2 stages. First, the probability of survival at specific ages (7 days, 28 days, 3 months, 6 months, 1, 2, 3, 4, 5, 6, 7, 8, 9, and 10 years) was calculated for each CA subgroup using Kaplan-Meier survival analysis, which takes into account the censoring of the data that occurred mainly because of the end of follow-up being December 31, 2015. This analysis was performed on individual case data by the registries locally using the centrally written syntax script. Registries then uploaded the Kaplan-Meier survival estimates with 95% confidence intervals (CIs), the number at risk, and the number of deaths at each age for each CA subgroup to the Central Results Repository at Ulster University, United Kingdom, using a secure Web platform. No individual case data were uploaded. The second stage involved pooling these data by meta-analysis to produce projectwide estimates.

The Kaplan-Meier survival estimates were combined in a random-effects meta-analysis to estimate the overall survival for each CA subgroup by modifying a method proposed by Combescure et al.²⁵  Combescure proposed the random-effects meta-analysis of survival curves by using the DerSimonian and Laird multivariate procedure²⁶  on arc-sine transformations of the conditional survival probabilities with a continuity correction of 0.25. However, when analyzed by individual CA subgroups, low numbers of cases in each registry and relatively low death rates for certain CA subgroups resulted in 100% survival for all registries for certain age years. By applying the method above, the model estimated a decrease in survival at these ages despite no deaths occurring, resulting in an underestimation of the overall survival. We therefore applied 3 adaptations. Firstly, instead of using the fixed continuity correction of 0.25 within the arc-sine transformation, a variable continuity correction equal to 1/n (the number of children alive at the start of the period) was used. This allowed the continuity correction to shrink with increasing sample sizes, while simultaneously reducing the overweighting of high survivals when sample sizes are small, which occurs because of the multivariate meta-analytic technique. This reduced the bias introduced into the country-level estimates when their samples sizes were ≥6 children. Secondly, data were excluded from the analysis if there were <6 children alive with the specified anomaly in a registry at a certain age. This was required because even the variable 1/n continuity correction still introduced bias for sample sizes <6. Thirdly, if no deaths occurred in any of the registries after a certain age, the overall survival for the remaining ages was imputed as the survival rate for the previous time period. This is a logical assumption because no deaths had been observed. In scenarios where there were no deaths in any registry during specific ages (eg, ages 3 and 4), but deaths did occur in later time periods, the meta-analyses were run on a reduced number of time points to limit the prevalence of the “no death” time periods. In these scenarios, instead of the 9 yearly time points,^2–10  average survival was calculated between ages 1 to 5 and 6 to 10. This preserved the use of all the data but reduced the number of time points in which continuity corrections would introduce significant bias. All meta-analyses were performed using R software.

Sensitivity analyses were performed excluding each registry in turn to determine if the overall survival estimates differed significantly.

Four independent categories of children with Down syndrome were analyzed: those without associated CHD or digestive system anomaly, those with only a CHD, those with only a digestive system anomaly, and those with both (other, less common associated CAs were not considered). The 10-year survival estimates for each registry were analyzed using a random-effects meta-analysis comparing the 3 Down syndrome groups with an associated CHD and/or digestive system anomaly with the group without any of these CAs.

Ten-year survival estimates with 95% CI for Isolated and for All groups were plotted for selected CAs. No formal statistical tests were performed because the All group included children in the Isolated group.

Stata v16 (StataCorp LLC, 2019) was used for the above comparisons.

Table 1 shows that 13 registries from 9 countries covering a population of 4 218 786 births from 2005 to 2014 provided survival data for 96 263 live births with a major CA.

Table 2 shows pooled survival estimates (with 95% CI) from ages 1 week to 10 years for children in the Isolated group (n = 77 054) in 32 specific CA subgroups. Overall, 10-year survival of children with any isolated CA was 96.9% (95% CI: 96.0%–97.7%). As expected, the highest mortality was within the first year of life; survival did not substantially decline after the first year for most CA subgroups. There was considerable variation in survival among individual CA subgroups. Ten-year survival varied from 51.6% (95% CI: 44.9%–59.4%) for hypoplastic left heart (HLH) to 99.8% (95% CI: 99.6%–100.0%) for cleft lip with or without cleft palate. Overall, 10-year survival across Europe was >90% for all but 5 isolated CA subgroups analyzed (27 of 32).

Table 2 also shows survival for children with Down syndrome (n = 4011) with or without CHD or digestive system anomaly. Compared with the highest 10-year survival in children with Down syndrome without associated CHD or digestive system anomaly of 97.6% (95% CI: 96.5%–98.7%), survival was significantly lower when Down syndrome was associated with any CHD but not digestive system anomaly (P < .001), with any digestive system anomaly but not CHD (P = .018), and with both CHD and digestive system anomaly (P < 0.001).

In the sensitivity analysis, the pooled survival estimates were robust to the exclusion of data from individual registries for most specific CAs (within ±2.5%, but mostly within ±1%), except for severe microcephaly (Wales registry: 5.7%) and HLH (Finland registry: 3.6%) (Supplemental Table 5).

Table 3 shows 10-year survival estimates for children in the All group and the proportion of children with isolated CAs in all live births and deaths with the specified CAs. Of a total of 4214 deaths, 49% occurred in children with CHD and 30% in children with severe CHD; 5.9% of deaths occurred in children with diaphragmatic hernia. Tables 2 and 3 show that 5.4% of all deaths were in children with Down syndrome (226 of 4214). In Table 3, for “any CA,” although 80% of live births were in children with isolated CAs, only 47.5% of deaths occurred in these children, indicating the higher survival in children with an isolated CA (96.9%, Table 2) compared with All CA (94.8%, Table 3). The proportions of children with isolated CAs in all live births and all deaths differed by CA severity, being more similar for severe CAs (eg, HLH, atrioventricular septal defect, diaphragmatic hernia). This indicates that, for these anomalies, the death is most likely related to that specific CA, whereas in other CAs with large differences in proportions, the death is related to the associated anomalies and karyotype defect.

Survival curves are shown for children with more common selected CAs from different organ systems in the Isolated and All groups, clearly demonstrating the higher survival for children in the Isolated group (Fig 1).

This linked cohort study using population-based data on live births from 2005 to 2014 (n = 96 263 out of a birth population of 4 218 786) among 13 CA registries in Western Europe provided survival estimates for children up to age 10. The pooled 10-year survival was >90% for the majority of isolated CA subgroups (27 of 32), with considerable variation in survival among specific CAs of different severity. Presence of associated anomalies considerably reduced survival in children with specific CAs. For children with Down syndrome, the 10-year survival was significantly higher for children without associated CHD or digestive system anomalies, compared with children with these anomalies.

Survival of children born with specific isolated CAs was higher in our study for spina bifida, CHDs, orofacial clefts, esophageal atresia, anorectal atresia or stenosis, diaphragmatic hernia, abdominal wall defects, and limb reduction defects compared with published population-based studies from Europe, the United States, and Australia,^27–33  and 2 systematic reviews.^14,34  The higher survival is expected, as first, the published studies cover earlier birth cohorts and survival has improved over time partly because of improvements in prenatal diagnosis and consequent increases in terminations of pregnancy for more severe CAs (https://eu-rd-platform.jrc.ec.europa.eu/eurocat/eurocat-data/prevalence_en; 45.1 and 52.1 per 10 000 births in 1995–2004 and 2005–2014, respectively); and second, some studies were not restricted to isolated anomalies.

True differences in survival among different countries are also likely. A recent smaller Australian study analyzing 1- and 5-year survival for births from 2004 to 2009 reported comparable survival estimates for all isolated CAs and for some specific subgroups such as severe CHD and diaphragmatic hernia.³⁰  Despite relatively high survival estimates at 4 weeks, 1 year, and 5 years for children born with isolated CAs reported in our paper, overall, they are still much lower compared with the average survival in the general European population of children at corresponding ages.^35–37 

Five- and 10-year survival of children with Down syndrome was also higher in our study compared with earlier studies^28,38–40  and comparable to more recent ones.^30,41,42  Presence of associated CHD in children with Down syndrome is an acknowledged risk factor for reduced long-term survival,^14,39–43  although a significant improvement in survival of these children over time was recently reported.¹⁴  This European study also reports a significantly reduced survival of children with Down syndrome associated with CHD and/or digestive system anomalies compared with those without.

This study has several strengths. We used high-quality data from specialist population-based registries of CAs that were linked to official mortality data sources, including data from national and vital statistics for 11 of 13 registries. Standardized approaches to data collection, coding, and classification in EUROCAT registries were enhanced by standardization of linked mortality data, creation of standardized syntax scripts, and generation of combined data sets and analytic results. This enabled the establishment of a large cohort of children with CAs in 13 regions of 9 Western European countries, increasing statistical power for the analysis of specific CAs and, thereby, the reliability of our findings. We developed a novel meta-analytic approach of analyzing survival data from several small samples to reduce bias arising as a result of the use of more standard techniques, which rely on the asymptotic properties of estimates from larger samples.

We were not able to include as many registries as originally planned because of barriers to gaining ethical approval, low linkage quality, or lack of survival data beyond 1 year. Despite relatively high linkage success, lack of 100% linkage in all registries may have resulted in an overestimate of the survival because of missed deaths. An overestimate of the pooled survival for children with severe microcephaly revealed by the sensitivity analysis after exclusion of the Wales registry may be because of a less stringent definition of severe microcephaly in Wales (<5th percentile instead of EUROCAT definition of <−3 SD).^44,45  Higher survival of children with HLH in Finland may be because of higher prenatal detection rates, resulting in improved survival after the full implementation of the national ultrasound screening program from 2010.^46,47 

The classification of CA into isolated and multiple CA was computer-based only, without manual expert review of the potential multiple CA cases, which could have resulted in some isolated CAs being misclassified as multiple CAs.²³ 

No formal comparison of 10-year survival among the isolated and nonisolated CAs could be performed because the Isolated and All groups were not mutually exclusive.

The participating registries do not cover all births in Western European countries, but we consider our results to be representative of Western Europe.

The accuracy of estimated long-term survival of children born with specific CAs is ensured by the use of common protocol for data collection, standardization, quality control, and registry-specific statistical analyses, as well as the development of the novel meta-analytic approach. Reliable information on long-term survival of children born with specific CAs is of major importance for counseling parents facing a prenatal diagnosis of CA, families living with a child affected by a CA, and for the health professionals involved in their care. The timely diagnosis of associated anomalies is essential for parental counseling because of their association with reduced survival. The geographical coverage should be widened in future European studies to produce findings that are more representative and generalizable for all of Europe.

We thank EUROlinkCAT contributors to this paper for their work on the project (data linkage and standardization, running syntax scripts): Drs L. Renée Lutke and Nicole Siemensma-Mühlenberg (University Medical Center Groningen, Groningen, Netherlands); Sandra Moreno Marro, Laia Barrachina Bonet and Laura García Villodre (Foundation for the Promotion of Health and Biomedical Research in the Valencian Region, Valencia, Spain); Dr Sonja Kiuru-Kuhlefelt and Tuuli Puroharju (THL Finnish Institute for Health and Welfare, Helsinki, Finland); Drs Gianni Astolfi, Aurora Puccini, Annarita Armaroli (Center for Clinical and Epidemiological Research, University of Ferrara, Ferrara, Italy); Nathalie Bertille and Makan Rahshenas (INSERM, Paris, France); Professor Sue Jordan (Swansea University, Wales, United Kingdom); Professor Elizabeth Draper (University of Leicester, Leicester, United Kingdom); Professor Jenny Kurinczuk (University of Oxford, Oxford, United Kingdom). We also thank Mr Hugh Claridge for the project management.

CA

congenital anomaly

CHD

congenital heart defect

CI

confidence interval

EUROCAT

European Surveillance of Congenital Anomalies

HLH

hypoplastic left heart