Birth prevalence for congenital limb defects in the northern Netherlands: a 30-year population-based study

Background Reported birth prevalences of congenital limb defects (CLD) vary between countries: from 13/10,000 in Finland for the period 1964–1977 to 30.4/10,000 births in Scotland from 1964–1968. Epidemiological studies permit the timely detection of trends in CLD and of associations with other birth defects. The aim of this study is to describe the birth prevalence of CLD in the northern Netherlands. Methods In a population-based, epidemiological study we investigated the birth prevalences of CLD for 1981–2010. Data were collected by the European Surveillance of Congenital Anomalies in the northern Netherlands (EUROCAT-NNL). We excluded malpositions, club foot, and dislocation/dysplasia of hips or knees. Trends were analysed for the 19-year period 1992–2010 using χ2 tests, as well as CLD association with anomalies affecting other organs. Results The birth prevalence of CLD was 21.1/10,000 births for 1981–2010. There was an overall decrease in non-syndromic limb defects (P = 0.023) caused by a decrease in the prevalence of non-syndromic syndactyly (P < 0.01) in 1992–2010. Of 1,048 children with CLD, 55% were males, 57% had isolated defects, 13% had multiple congenital anomalies (MCA), and 30% had a recognised syndrome. The upper:lower limb ratio was 2:1, and the left:right side ratio was 1.2:1. Cardiovascular and urinary tract anomalies were common in combination with CLD (37% and 25% of cases with MCA). Digestive-tract anomalies were significantly associated with CLD (P = 0.016). Conclusions The birth prevalence of CLD in the northern Netherlands was 21.1/10,000 births. The birth prevalence of non-syndromic syndactyly dropped from 5.2/10,000 to 1.1/10,000 in 1992–2010.


Background
Limb defects seen in childhood are mainly congenital and occur when a part of or the entire limb fails to form normally during pregnancy. Reduction defects may be disabling limb defects due to the failure of several elements to form properly [1]. Less disabling limb defects are polydactyly, defined as complete or partial supernumerary digits, and syndactyly, fusion of two or more digits [2]. Disruptive events appear to be the most common cause of congenital limb defects (CLD) [3]. During the gestational period, disruptive events, such as amniotic band or vascular disruptions, may cause amputation or hypoperfusion of the developing limbs [4]. Various CLD are due to prenatal exposure to different teratogens [5], the best-known example of which is thalidomide, which caused a wide range of CLD, especially intercalary reductions and preaxial defects, in the 1960s [6]. To prevent further tragedies, several international registries of congenital defects were established. The European Surveillance of Congenital Anomalies (EUROCAT) network of registries in thirty-seven countries and the International Clearinghouse for Birth Defects Surveillance and Research are two such registries that have the goal of monitoring birth defects [7,8]. Monitoring CLD birth prevalences (BP) permits estimates of how common CLD are in the general population, early detection of risk factors for CLD and its associations with other congenital anomalies, and comparison of standard data collections.
There is no up-to-date, detailed information on CLD in the Dutch population available. Therefore, the aim of this study is to describe the epidemiology of CLD in a population-based study in the northern Netherlands for the period 1981-2010.

Data source
The CLD cases have been collected by EUROCAT in the northern Netherlands (EUROCAT-NNL) since 1981. Children or foetuses with one or more major congenital defects and whose mothers lived in the northern provinces (Groningen, Drenthe or Friesland) at the time of delivery were eligible for registration.
Cases were ascertained according to EUROCAT's Central Registry guidelines [26]. EUROCAT-NNL registers foetuses irrespective of gestational age, spontaneous abortions, terminations of pregnancies (foetuses of ≤ 24 weeks' gestation) following prenatal diagnosis because of a congenital malformation, stillbirths (foetuses of ≥ 24 weeks' gestation), live births, and children diagnosed before 11 years of age.
Cases are reported by general practitioners, midwives and physicians [27]. Hospital registries are also actively and regularly searched by the EUROCAT-NNL personnel to find eligible children/pregnancies. Various sources including hospital files, obstetric and pathology records are searched for case assessment (type of malformation, chronic diseases and dates of screening procedures). When new information becomes available for an already registered case, the case is updated in the EUROCAT-NNL database until the child reaches the age of 11 years. The paediatric cardiology centre (part of the University Medical Centre Groningen) covers all births in the EUROCAT-NNL registration area and supplies systematic lists with cases and diagnostic details to the registry [28]. For all reported cases, results of genetic tests are downloaded from the genetics department, if these results are available. Abnormal karyotype reports are recorded from prenatal and postnatal samples [28].
Since 1992, parents or guardians are asked to give informed consent for registration of their child and for the use the data for research purposes. The response rate is 80%. Up to 1992, no parental approval was required to register cases.
This study was approved by the Medical Ethical Committee, University Medical Centre Groningen, the Netherlands (number M12.118639).

Study population
All children and foetuses with CLD, diagnosed before or after birth, were included in this study. Children with only minor CLD, such as clinodactyly, camptodactyly, brachydactyly of the fourth and fifth fingers, trigger finger, syndactyly of the second and third toes, sandal gap, and short big toe were not included, because EUROCAT does not register minor CLD. We also excluded malpositions, club foot and dislocation/ dysplasia of hips or knees (dislocation of patella) from our analysis because these musculoskeletal anomalies are common birth defects and their inclusion in the calculations of total birth prevalence for CLD would have given an inflated birth prevalence. The ICD-codes were thoroughly checked against the descriptions of CLD and rectified if necessary. To gain insight into CLD trends, we included children with isolated CLD and MCA (non-syndromic CLD). Because the registration method changed in 1992 with the introduction of informed consent, the trend analysis was conducted for the period 1992-2010. Localisation of the CLD (left/right side) was only studied in live births with isolated CLD or MCA, because of possible lack of information on localisation in stillbirths and spontaneous abortions or termination of pregnancies. Syndromic CLD were excluded from the laterality analysis because they have characteristic patterns [31]. We determined the most frequent anomalies in other organ-systems that occurred in combination with CLD (MCA cases). A clinical geneticist reviewed the cases that were suspected of having monogenic or genetic causality based on the description of the CLD, associated anomalies, and/or family history.

Statistical analyses
BPs were determined by dividing the number of affected cases by the total number of births (live births, stillbirths, spontaneous abortions/termination of pregnancies)     in the EUROCAT-NNL region. To visualise trends, a three-year moving average prevalence was calculated. The χ 2 test for trend was used to analyse changes over time in BP and to determine whether a type of CLD was preferentially associated with a congenital anomaly affecting another organ system. Only MCA cases that had one type of CLD were included in this analysis. The association between the number of CLD and the number of anomalies in other organ-systems was tested for trend using the χ 2 test. If χ 2 assumptions (expected cell counts) were not met, the exact method was used. Two-tailed values of P < 0.05 were considered statistically significant. PASW Statistics 18.0 for Windows (SPSS Inc., 2009, Chicago, IL, www.spss.com) was used for the analyses.

Birth prevalence and study population
From 1981-2010, 1,048 cases with CLD were recorded among 497,751 births in the northern Netherlands, yielding a BP of 21.1/10,000 ( Table 1). The prevalence for transverse reduction defects was 3.9/10,000 births and for longitudinal reduction defects 2.4/10,000 (equal rates for preaxial and postaxial: 1.3/10,000).
Of the 1,048 cases, 823 (79%) were live-born children; 181 (17%) were spontaneous abortions, stillbirths or infants who died shortly after birth; and 44 (4%)   were termination of pregnancies. Of all 1,048 cases, 578 (55%) were males and 4 were of undetermined gender. More males (455/823; 55%) were also registered in the live births. An overview of the data is given in Figure 1.
In this study, termination of pregnancy was performed in 44 cases. Isolated CLD occurred in 2 of the 44 cases (split hands and feet; mixed reduction defects of the lower limb), MCA in 15 (CLD with one or more other major non-CLD defects: CNS and neural tube defects (n = 5), urinary (n = 5), digestive system (n = 4), cardiovascular (n = 3), respiratory system (n = 3)), and recognized conditions in 27 (chromosomal (n = 14), other recognized condition (n = 7), and monogenic (n = 6)).    Only cases with multiple congenital anomalies (including a limb defect) were included because genetic abnormalities or syndromes have particular associations [17]. General categories and examples of anomalies occurring with CLD were given. b Numbers do not always add up due to multiple CLD in some cases or due to lack of information on subcategories of CLD (e.g. preaxial, postaxial). Percentages are calculated from the total number of children with multiple congenital anomalies and CLD (n = 135). c In addition to one or several major non-limb defects, 24 cases (17.8%) also had a malposition, or a clubfoot, or hip dysplasia/dislocation (polydactyly n = 1, syndactyly n = 8, reduction defects n = 9, other CLD n = 12).

Trend analyses for the period 1992-2010
A significant decrease of the BP rate over time was found for non-syndromic CLD as a group and for non-syndromic syndactyly in particular (CLD: χ 2 = 5.2, P = 0.023, syndactyly: χ 2 = 6.8, P = 0.009) (Figure 2). The decrease in non-syndromic syndactyly from 5.2/ 10,000 to 1.1/10.000 births was also found to be responsible for the decrease of non-syndromic CLD as a total group (CLD syndactyly excluded: χ 2 = 1.5, P = 0.215). A significant decrease was also noticed in the heterogeneous group of "other CLD" (χ 2 = 4.8, P = 0.028). When we included recognised conditions in the analysis, no trend was identified for syndactyly (χ 2 = 1.5, P = 0.218), but there was a significant decreasing trend present for CLD as a group (χ 2 = 9.3, P = 0.002).

Associated anomalies
CLD were most common with cardiovascular anomalies ( Table 4): 19 ventricular septal defects (of which 8 cases had polydactyly), 7 atrial septal defects, 6 tetralogy of Fallot, and 4 coarctation of the aorta. Urinary tract anomalies were also frequent: 12 bi-or unilateral renal agenesis (of which 7 cases had reduction defects), 5 hydronephrosis, and 5 cystic kidney. Anomalies of the central nervous system were found in fewer cases with CLD: 10 cases of hydrocephaly (of which 5 had syndactyly) and 4 of microcephaly. More males were affected in cases with genital anomalies (n = 17, uniformly spread between CLD) of which most were hypospadias (n = 14). We only found a significant association between digestive anomalies and CLD (P = 0.016). Reduction defects were more likely to occur in combination with digestive tract anomalies, whereas syndactyly and "other CLD" were less likely to occur with such anomalies (Table 5). We found no significant association between the number of CLD and the number of anomalies in other organ-systems (χ 2 = 3.4, df 1, linear-by-linear association P = 0.067).

Discussion
We aimed to describe the epidemiology of CLD in the northern Netherlands. From 1981-2010, the birth prevalence of CLD was 21.1/10,000 births, which falls between the BPs found in two other European registries: Finland (13/10,000 for 1964-1977) and Scotland (30.4/10,000 for 1964-1968) [9,10]. However, there were differences in the inclusion criteria of all three registries.
More recent BPs for CLD are available on the official website of the EUROCAT network of over thirty-seven national registries [32]. These registries include live births, stillbirths and terminations of pregnancies, which allow their BPs for types of CLD to be compared with the Netherlands (NNL). Six registries have reported complete data for the period 1981-2010 (Dublin, Ireland; Odense, Denmark; Paris, France; Hainaut, Belgium; Emilia-Romagna and Tuscany, Italy).
Our BP for polydactyly (8.4/10,000) was close to the figures reported for Emilia-Romagna and Hainaut (8.14 and 8.55; Table 6), but much lower that the BP in Paris (13.8). We found a BP for syndactyly of 4.7/10,000, which is comparable to that reported for Emilia-Romagna (4.5) and Paris (5.1). Our BP for reduction defects (6.9) was close to that reported by Hainaut (7.3; Table 6).
The total BP of each EUROCAT registry includes the club foot and hip dysplasia/dislocation, which hampers direct comparison with the total BP determined in our study (21.1/10,000). However, we can compare the summed BPs for polydactyly, syndactyly and reduction defects. Our calculated BP (20/10,000) is similar to those calculated for Emilia-Romagna (18.0), Tuscany (18.6), and Odense (22.2) ( Table 6).
The differences in BP between countries are most likely to be the result of variations in coding methods, ascertainment, notification and inclusion criteria and, until a consistent and compatible system (such as that of EUROCAT) is universally adopted, there will be no way of determining whether BP variation has an environmental or other cause.

Localisation and laterality of limb defects
We found upper limbs were more commonly affected, which agrees with the literature [9,13,23,33]. Left-sided limb defects prevailed in our study, as in other studies [9,14,24], but others have reported more right-sided defects [16,34] or both sides being equally affected [35]. Longitudinal and multiple reduction defects were more frequently localised on the right side [34].

Gender distribution
Our male:female ratio for CLD was 1.2:1 in both total births and in live births, compared with a male:female ratio of 1.1:1 for live births in the northern Netherlands (1981-2010) [36]. This male excess has been reported before, but its aetiology remains unexplained [17,19,37].

Trends
In the northern Netherlands, the BP of non-syndromic syndactyly has shown a significant decrease since 1992. EUROCAT-NNL has a reliable and well-established network that notifies the registry of new cases shortly after birth, ensuring that the parents of almost all children with congenital anomalies in the region are contacted for registration, thus we do not feel many cases will have been missed. In addition, the diagnosis of syndactyly as part of a genetic syndrome is more commonly detected now than in the early 1990s. Excluding cases with genetic syndromes from our time-trend analysis may have influenced our finding of a drop in cases with nonsyndromic syndactyly over time. Another reason for the decline might be a change in parental attitude towards giving informed consent for registering syndactyly, which might now be seen as a rather minor anomaly. The introduction of the informed consent procedure in 1992 may also have influenced the number of registrations compared to the period before 1992. In a post-hoc analysis, no other significant trends were found for the period 1981-1991. Finally, not all CLD cases may have been reported yet. By including data from recent years in our study, the decline may be due to the fact that some of the cases still need to be registered. Therefore, we cannot fully rule out that the decline observed is a registration artefact.

Associated birth defects and recognised conditions
In comparison with polydactyly, syndactyly and "other CLD", we saw reduction defects more often in combination with congenital anomalies affecting the central nervous system, digestive system (the only significant association in current study, Table 5), genital or urinary system. Associations of reduction defects with these kinds of anomalies have been reported previously [17,19,24]. However, in our study, reduction defects were not the CLD type that occurred most frequently in combination with cardiovascular anomalies, as seen in other studies [17,19,24]. We found polydactyly (equally preaxial and postaxial) occurred most often in combination with cardiovascular anomalies, especially with ventricular septum defects. Another study reported polydactyly occurring mostly with central nervous system anomalies [38]. They also found significant associations of polydactyly with recognised syndromes: trisomy 13, Meckel, and Down syndrome [38]. We only found trisomy 13 occurring often in our cases with postaxial polydactyly of the upper limb.
In our NNL study, amniotic bands were the second most frequently recognised condition after trisomy 13, and they also occurred more often in combination with transversal limb defects and syndactyly, as reported in the literature [34]. We also found syndactyly occurring often with trisomy 18 (mostly syndactyly of lower limbs), Apert, and Poland syndrome, also reported previously [39].
Longitudinal preaxial reduction defects were found to be the most common reduction defects occurring in cases with other congenital anomalies [17,34]. Preaxial defects, for instance, can occur in the VATER association and in many genetic conditions, while transverse defects are usually isolated single defects in a family and may be caused by disruptive events in early pregnancy [17,34]. We also found radius aplasia/dysplasia most frequently in cases with VATER/VACTERL, trisomy 18, and Holt-Oram syndrome.
To our knowledge, there are no genetic predispositions that are specific to the Dutch population that could have influenced the birth prevalences of CLDs in this study. In addition, the tables on the official EUROCAT website did not reveal any particular recognized condition specific for the Netherlands compared to the other six registries (Dublin, Ireland; Odense, Denmark; Paris, France; Hainaut, Belgium; Emilia-Romagna, and Tuscany, Italy) [32]. These observations, and the fact that our birth prevalences of CLD are comparable to those in other EUROCAT registries, imply that our birth prevalences are relatively good estimates of world-wide prevalences. However, studies on larger populations with CLD and with the same inclusion criteria would allow for better generalizability.

Study strengths and limitations
The EUROCAT-NNL database contains specific and detailed information about cases with birth defects. All the cases were verified and corrected for any miscoding of the CLD. Updated information, if available, was retrieved from the notifying hospital to clarify doubtful coding or diagnosis.
Our data will be of clinical relevance to clinicians treating children with CLD and to their parents, not only the parents of live-born children who were unaware of their child's defect prior to delivery, but also the parents who know they are expecting a child with a CLD. The latter group is becoming more and more significant with all the possibilities for information now offered by prenatal diagnostics.
A limitation of our study may be the relatively small number of children with MCA included in the analysis for associations. The response rate for the parental consent for the registration of their child is fairly high (80%). Nevertheless, we cannot exclude the fact that underestimation of the true BP of CLD may have occurred. This type of underestimation, due to the voluntary participation of parents or notifying sources, is also likely to be present in other birth defects registries (e.g., Tuscany and Emilia-Romagna, Italy; Paris, France; United Kingdom; Austria; Switzerland; Spain) [28]. A further limitation may be the classification system used. Although there are several classifications of CLD [1,40,41] available, they all pose difficulties. The International Federation of Societies for Surgery of the Hand has adopted the morphological classification of Swanson et al. [40]. However, this classification for CLD has not been accepted by all surgeons [14], and difficulties based on this classification have been reported [20]. Furthermore, Swanson's classification is a more morphological classification, and does not conform to the aim of our study, which was to provide a descriptive overview of CLD. Stoll et al. [42] proposed a purely descriptive classification of CLD, which complied with the purposes of this article. However, Stoll et al.'s classification did not allow for CLD that were included in the "other CLD" category (e.g. arthrogryposis). We therefore used a classification based on ICD-9 and ICD-10 [29,30].

Conclusions
We established a prevalence of 21.1 children with CLD per 10,000 births in NNL after a thorough coding and correction procedure, which makes this figure reasonably reliable. The decreasing trend observed in non-syndromic syndactyly in 1992-2010 may be real or the result of a registration artefact. CLD occurred frequently in combination with cardiovascular and urinary tract anomalies, and were significantly associated with digestive-tract anomalies.