Familial risk and heritability of intellectual disability: a population‐based cohort study in Sweden

Introduction

Intellectual disability (ID) affects about 2% of children and 1% of adults worldwide (Maulik, Mascarenhas, Mathers, Dua, & Saxena, 2011).

It is defined as a deficit in general cognitive ability below 2 standard deviations from the population mean (IQ score < 70) accompanied by impairment of everyday adaptive abilities. Although familial recurrence of ID has been long recognized, the complex etiology of ID remains poorly understood. Heterogeneity of genetic factors encompasses both common polygenic traits and numerous forms of rare defects, including aberrations and small point mutations. Mild ID is often hypothesized to represent the lowest end of the intelligence spectrum with the polygenic influences similar to those present in the general population, while severe ID is often associated with rare de novo mutations (Reichenberg et al., 2016). The genetic background of ID is further complicated by the presence of several forms of X-linked ID, including expansions of the FMR1 gene responsible for the fragile X disorder, several described X-linked syndromes, as well as numerous distinct forms of nonspecific X-linked ID (Tejada & Ibarluzea, 2020). It has also been hypothesized that majority of polygenic influences responsible for intelligence are located on the X chromosome resulting in higher prevalence of ID in males then females (Turner & Partington, 1991).

In 30%–50% of ID, a genetic cause cannot be identified suggesting complex interplay between genetic and environmental factors (Lee et al., 2019; Wayhelova et al., 2019). The contribution of environmental and genetic factors has been traditionally measure by familial recurrence risk. We have systematically searched PubMed for articles published in English between 1 January 1970 and 30 September 2021 using the terms ‘mental retardation’ or ‘ID’ or ‘mental deficiency’ or ‘retardation’ AND ‘recurrence’ or ‘recurrence risk’ or ‘recurrent risk’ or ‘genetic studies’ or ‘familial’ or ‘twin’ or ‘sibling’. In general, studies of ID are scarce and mostly limited to old reports on ID frequency in siblings of ID individuals from one clinical center (Table S1; Angeli & Kirman, 1975; Becker, Kaveggia, Pendleton, & Opitz, 1977; Costeff & Weller, 1987; Durkin, Kaveggia, Pendleton, Neuhauser, & Opitz, 1976; Opitz, 1979; Opitz, Kaveggia, Durkin-Stamm, & Pendleton, 1978; Turner, Collins, & Turner, 1971). Twin studies of ID (Nichols, 1984; Wilson & Matheny, 1976) involve often too small samples (<10 monozygotic twin pairs) to draw reliable conclusions on the relative importance of genetic and environmental influences. To our best knowledge, there only three studies on distant relatives of ID individuals, based mainly on data from educational institutions and family interviews from 1950s to 1980s (Becker et al., 1977; Bundey, Thake, & Todd, 1989; Reed & Reed, 1965). There are no population-based studies on the RR in relation to families without any affected member.

Risk estimates also play a crucial role in genetic counseling, especially important for families when no specific cause of ID could be identified in an affected child. Although the applicability of estimates determined decades ago is unclear in contemporary medicine, clinicians are forced to apply them in their clinical reasoning due to the lack of more recent data.

Traditionally, three factors are considered to be important for the familial recurrence of ID. First, coexisting medical conditions and underlying ID etiology in affected probands are hypothesized to affect risk in relatives. For example, chromosomal abnormalities and major congenital malformations are often associated with de novo genetic aberrations. At the same time, adverse birth events are treated as an environmental cause of ID in a particular affected individual. Consequently, both conditions are associated with a low risk of ID in siblings (Clarke, 2019). Autism and epilepsy are common comorbidities both in individuals with ID and their relatives (Xie et al., 2019), suggesting similar polygenic genetic factors being responsible for those conditions and affecting the familial risk of ID. Previous studies attempting to stratify small samples on coexisting medical conditions failed to provide conclusive findings (Bundey & Carter, 1974; Opitz et al., 1978).

Second, severe ID is often associated with de novo mutations and low ID frequency in relatives (3%–5%; Angeli & Kirman, 1975, Bundey & Carter, 1974, Durkin et al., 1976), while mild ID may be a polygenic trait with higher risk (19%–26%; Bundey et al., 1989, Herbst & Baird, 1982).

Third, an excess of ID-affected brothers in comparison to sisters of ID probands is often attributed to X-linked inheritance (Herbst & Miller, 1980). However, to date, no epidemiological study has examined X-linked contribution in a nationwide population sample and across relatives, for example, maternal versus paternal half-siblings.

The heritability (the proportion of phenotypic variance attributable to genetic factors) for general cognitive abilities within the normal range for adults has been estimated at 66%–68% (Haworth et al., 2010; Panizzon et al., 2014), but there are no studies evaluating the contribution of genetic and environmental effects in ID.

Thus, the purpose of this study was to evaluate the familial risk and heritability of ID using population-based registers in Sweden to provide modern and comprehensive data for clinical guidelines and etiological research.

Methods Setting and data sources

All individuals in Sweden receive a unique personal identification number, which allows for extensive linkages between administrative, educational and health registers (Ludvigsson, Otterblad-Olausson, Pettersson, & Ekbom, 2009). Registers used in this study are described in Table S2.

The population consisted of individuals born in Sweden between 1 January 1973 and 31 December 2013. Each individual was linked to his/her relatives to create eight subcohorts. Similar to previous family based studies (Mataix-Cols et al., 2015; Svensson et al., 2012), we used a matched study design, where each exposed individual is matched with unexposed reference individuals on confounders. From the population, we identified all pairs of relatives of ID probands and matched them on sex, year and county of birth, with up to 50 corresponding relative pairs consisting of individuals without ID and their relatives (Figure 1). The study was approved by the regional ethics review board in Stockholm, Sweden (2013/862-31). The use of Swedish register data does not require informed consent.

image

Study Design. Unexposed reference relatives (relatives of individuals without Intellectual Disability) are randomly selected from source population and matched on sex, birth year and county of birth with relatives of probands with Intellectual Disability. One individual may appear several times as different type of relative

Intellectual disability

Data on ID was obtained from the National Patient Register (NPR; Ludvigsson et al., 2011), Clinical Database for Child and Adolescent Psychiatry in Stockholm (Pastill; Lundh, Forsman, Serlachius, Lichtenstein, & Landen, 2013), the Habilitation Register (Idring et al., 2012), and The Halmstad University Register on Pupils with ID (HURPID; Arvidsson, Widén, & Tideman, 2015). Chromosomal abnormalities, major congenital malformations, adverse birth events and epilepsy were obtained from NPR and the Medical Birth Register. Autism diagnoses were retrieved from NPR, Pastill and Habilitation Registers (ICD-codes are given in the Table S3). Exact information on severity level and date of ID diagnosis was not available for individuals identified by the HURPID register.

Statistical analysis

We followed individuals from birth to the first record of ID, censoring date due to emigration, death, or end of study (31 December 2013), whichever occurred first. We calculated the incidence rate by dividing the number of individuals with ID with the total time of follow-up among all individuals (referred to as person-years), and expressed this as rate ID per 100,000 person-years. The 20-years risk was calculated as the percentage of affected relatives from the total number of relatives followed over 20 years, separately for all types of kinship. To evaluate the RR, we compared the relatives of ID probands (exposed relatives) to matched relatives of unaffected individuals (relatives of individuals without ID i.e. unexposed relatives, Figure 2). The RRs were estimated as Hazard ratios with 95% confidence intervals (95% CI) obtained from stratified Cox proportional-hazard models.

image

Selection of unexposed reference relatives on sibling example. Siblings pairs are matched by sex, birth year and county of birth of respective siblings in the pair (i.e., individuals 1A and 2A are of the same sex, born in the same year and county, as the 2A and 2B are of the same sex, born in the same year and county). Risk of ID was compared between exposed siblings (1B) and unexposed siblings (2B) graphically represented as arrows on the figure

Schoenfeld residuals were used to check for proportional hazards assumption for each covariate. Because several pairs of relatives from families with more than two affected relatives could be included in analyses, a cluster robust sandwich variance estimator was used to account for the dependence in the data and adjust standard errors. First, we evaluated the association between ID diagnosis in individuals and diagnosis of ID in their relatives. Second, to evaluate the association between the risk and potential etiology of ID, analyses were stratified by medical condition in ID proband representing underlying mechanisms involved in ID (i.e. chromosomal abnormalities, major congenital malformations, adverse birth events, autism, and epilepsy). Third, since mild and severe ID may have distinct genetic architectures, analyses were repeated separately for different levels of ID. We also compared risk in male and female relatives. To investigate the association between ID and line of inheritance, we compared the risk between maternal versus paternal line of relatives (e.g. the risk of ID in an individual having a cousin with ID on maternal line vs. paternal line).

To explore whether case definition (medical diagnosis from in- or outpatient care or affiliation to a special education school) affected our estimates, we performed sensitivity analyses for data obtained separately from the two nationwide registries (NPR and HURPID) and the time of first record of ID (≤2001 only inpatient care; and >2001 inclusion of both in- and outpatient care in the NPR).

Furthermore, we performed sensitivity analyses ignoring the time scale and treating ID diagnosis as binary variable in conditional logistic regression models, by design adjusted for sex, birth year and county of birth. The first model included all individuals and their outcomes regardless of time of follow-up and age at diagnosis. Further, we performed two additional secondary analyses with constant observation time of 10 year and 15 years. We have excluded from this part of analysis individuals who died or emigrated before the respective age (10 years or 15 years). The PROC PHREG statement, procedure for Cox regression model in the SAS software (version 9.4; Cary, NC, USA) was used for the analyses; 95% confidence intervals not including 1 were considered statistically significant.

Heritability

Similar to our previous studies we used a quantitative genetic approach from each family, one relative pair was randomly included in the calculations (Sandin et al., 2017). Individuals with chromosomal abnormalities and major congenital malformations were excluded from the analyses to address common, rather than rare, genetic inheritance for liability to ID, thus disregarding de novo genetic aberration of large effect. Concordance rates were calculated as the proportion of individuals with a relative with ID out of all individuals with ID and relatives of the specific relative type. Tetrachoric correlations were calculated for all levels of ID and after exclusion of individuals with severe and profound ID. Using the so-called liability-threshold model in a structural equation model framework, the risk-liability was assumed to be above an estimated threshold if individual was diagnosed with ID. We estimated additive genetic effects (A) and dominance deviations (D). The overall heritability (A + D) was estimated as the total variance related to genetic influences. The shared environmental effect (C) reflects nongenetic influences behind similarity between siblings. This shared environment was assumed shared to similar extent among twins, full and maternal half-siblings. Further, we calculated the effect of individually unique environment (E). Likelihood ratio tests were used to compare the goodness of fit between the full ADCE-model and reduced models (ACE, ADE, AE and CE). Akaike’s Information Criterion (AIC) and Bayesian Information Criterion (BIC) were used for model comparison. We tested both nonadjusted and adjusted for sex and birth year models. For heritability analyses, we used the R software with OpenMx package.

Results

We identified 37,787 (0.9%) ID probands out of 4,165,785 individuals born in Sweden between 1973 and 2013 (Study Design in Figure 1, ID prevalence per birth year in Figure S1). Among ID individuals, there were 22,998 (59.0%) males and 15,489 (41.0%) females. Mild ID was present in 16,808 (44.5%) probands, moderate in 5,205 (13.8%), severe in 2,323 (6.1%) and profound in 1,292 (3.4%), whereas severity was not specified for 12,159 (32.2%). The relative cohorts included 116 monozygotic twins, 5,125 parents, 3,680 children, 284 dizygotic twins, 46,174 siblings, 12,074 maternal half-siblings, 11,981 paternal half-siblings, 20,911 nephews/nieces, 10,712 uncles/aunts, and 169,432 first cousins matched to relatives of unaffected individuals from the general population. Baseline characteristics are presented in Tables S4 and S5.

Familial risk

All relatives of individuals with ID were at significantly increased risk of ID compared to the relatives of unaffected individuals from the general population (Table 1, Figure 3). The increase was related to the genetic relatedness with ID probands. In monozygotic twins to ID individuals, sharing 100% genes with probands estimates were especially high (RR 256.70, 95% CI 161.29–408.53). Estimates for dizygotic twins (RR 7.04, 95% CI 4.67–10.61) and for the full-siblings (RR 8.38, 95% CI 7.97–8.83) were comparable with overlapping 95% CI. Lower RR were found for maternal (RR 4.56 95% CI 4.02–5.16) and for paternal half-siblings (RR 2.90, 95% CI 2.49–3.37).

Table 1. The 20-years risk, frequency, rate per 100,000 person-years, person-years of follow-up and RR for ID, comparing relatives of ID probands (exposed relatives) with relatives of matched unaffected individuals from general population (unexposed relatives) Relative type Degree of genetic similarity (%) Relatives of ID probands (Exposed relatives) Relatives of matched unaffected individuals from general population (Unexposed relatives) 20-years risk (95% CI) Affected/total (%) Rate of ID per 100,000 person-years Person-years of follow-up 20-years risk (95% CI) Affected/total (%) Rate of ID per 100,000 person-years Person-years of follow-up Monozygotic twins 100 52.46 (39.93–64.99) 82/116(70.7) 4,480 1,830 0.17 (0.02–0.31) 23/5,800 (0.4) 19 124,125 First degree Parents 50 1.05 (0.77–1.33) 170/5,125 (3.3) 92 183,858 0.06 (0.05–0.07) 298/165,431 (0.2) 5 6,057,214 -Mothers 50 1.12 (0.75–1.49) 126/3,174 (4.0) 111 113,347 0.06 (0.04–0.07) 207/102,374 (0.2) 6 3,730,438 -Fathers 50 0.94 (0.51–1.37) 44/1,951 (2.3) 62 70,511 0.06 (0.05–0.08) 91/63,057 (0.1) 4 2,326,776 Children 50 8.70 (0.00–20.21) 170/3,680 (4.6) 887 19,158 2.24 (0.45–4.02) 324/108,743 (0.3) 60 539,201 Dizygotic twins 50 4.35 (0.18–8.52) 26/284 (9.2) 503 5,172 1.09 (0.78–1.39) 186/14,200 (1.3) 71 261,071 Full siblings 50 6.28 (6.00–6.56) 4,122/46,174 (8.9) 422 975,637 0.75 (0.73–0.76) 20,732/1,895,993 (1.1) 51 40,725,035 Second degree Maternal half-siblings 25 4.85 (4.34–5.36) 810/12,074 (6.7) 319 254,311 1.14 (1.07–1.21) 2,511/163,291 (1.5) 74 3,379,123 Paternal half-siblings 25 3.23 (2.80–3.67) 521/11,981 (4.3) 212 245,183 1.06 (0.99–1.12) 2,410/177,679 (1.4) 68 3,525,432 Nephews/nieces 25 3.48 (2.33–4.63) 277/20,911 (1.3) 186 149,200 1.40 (1.19–1.61) 1,889/476,847 (0.4) 65 2,928,117 Uncles/aunts 25 1.22 (1.01–1.43) 282/10,712 (2.6) 79 356,341 0.40 (0.38–0.43) 1,883/227,986 (0.8) 24 7,811,891 Third degree First cousins 12.5 1.49 (1.42–1.57) 3,804/169,432 (2.2) 100 3,812,638 0.80 (0.79–0.81) 59,539/5,016,486 (1.2) 52 114,282,417 image

Relative risk of intellectual disability (ID)

Medical conditions and adverse birth events

To evaluate the association between ID risk and potential etiology of ID, we have compared the risk in siblings to ID probands with siblings of unaffected individuals, stratified by the medical conditions in ID proband. The lowest risk was observed among siblings of individuals with ID caused by chromosomal abnormalities (RR 5.53, 95% CI 4.74–6.46). There was no substantial difference in risk estimates related to major congenital anomalies, adverse birth events, autism and epilepsy (Table 2). Similar patterns were observed among other types of relatives (Tables S6 and S7).

Table 2. The 20-years risk, frequency, rate per 100,000 person-years, person-years of follow-up and RR for ID, comparing relatives of ID probands (exposed relatives) with full siblings of matched unaffected individuals from general population (unexposed siblings) Medical conditions/events in ID probands in hierarchical ordera Full siblings of ID probands (Exposed siblings) Full siblings from general population (Unexposed siblings) Relative risk (95% CI)b 20-years risk (95% CI) Affected/total (%) Rate of ID per 100,000 person-years Person-years of follow-up 20-years risk (95% CI) Affected/total (%) Rate of ID per 100,000 person-years Person-years of follow-up Chromosomal abnormalities 4.79 (4.16–5.43) 359/6,620 (5.4) 239 150,151 0.60 (0.56–0.63) 2,717/275,719 (1.0) 43 6,387,357 5.53 (4.74–6.46) Major congenital malformations 5.59 (4.91–6.28) 599/8,083 (7.4) 376 159,333 0.40 (0.00–0.96) 3,592/330,159 (1.1) 54 6,645,388 7.04 (6.35–7.80) Adverse birth events 8.35 (7.37–9.33) 608/6,180 (9.8) 533 114,129 0.91 (0.86–0.96) 2,859/253,519 (1.1) 60 4,798,047 9.17 (8.18–10.28) Autism 5.12 (4.37–5.86) 518/6,666 (7.8) 404 128,107 0.77 (0.72–0.81) 2,967/281,856 (1.1) 55 5,420,599 7.50 (6.74–8.35) Epilepsy 6.40 (5.21–7.59) 233/2,550 (9.1) 416 56,027 0.66 (0.60–0.72) 1,105/105,342 (1.0) 47 2,375,336 9.38 (7.97–11.04) None 8.35 (7.37–9.33) 1,805/16,075 (11.2) 491 367,891 0.91 (0.86–0.96) 7,492/649,398 (1.2) 50 15,098,308 9.98 (9.34–10.66) Analysis stratified by medical conditions/events in ID probands representing mechanism involved in the ID etiology. Severity of ID

Risk estimates in siblings of ID individuals were inversely related to level of intellectual impairment in affected probands (Table S8). The highest estimates were noted among siblings to individuals with mild (RR 9.15, 95% CI 8.55–9.78) and moderate ID (RR 8.13, 95% CI 7.28–9.08) with lower risks being observed among siblings of individuals with severe (RR 6.80, 95% CI 5.74–8.07) and profound ID (RR 5.88, 95% CI 4.52–7.65). A similar pattern was observed in the second and third-degree relatives (Table S9).

Sex and linage of relationship

Generally, male relatives of ID individuals were more likely to be diagnosed with ID compared to female relatives of ID probands, (Table S10) with the exception of fathers who were less likely to have ID than mothers (RR 0.56, 95% CI 0.40–0.79). No significant differences were observed for twins. Maternal line was associated with higher risk estimates in comparison to the paternal line (Table S11) with statistically significant results for half-siblings (RR 1.49, 95% CI 1.34–1.68) and cousins (RR 1.35, 95% CI 1.24–1.47).

Sensitivity analyses

The main findings reported above did not change substantially when analyzed separately on data from NPR and HURPID, and for probands with first record of ID before or after 2001 (Tables S12 and S13). There were no substantial differences between results obtained from the Cox regression and logistic regression model without taking into account the time of follow-up (Table S14).

Heritability

The concordance rate for monozygotic twins was 73.2%, with much lower estimates observed for dizygotic twins (9.1%), full siblings (8.1%), maternal (6.5%) and paternal halfsiblings (4.2%). The unadjusted ID tetrachoric correlation was estimated to be 0.97 (95% CI 0.94–0.99) for monozygotic and 0.43 (95% CI 0.28–0.57) for dizygotic twins; 0.45 (95% CI 0.44–0.47) for full siblings; 0.32 (95% CI 0.29–0.36) for maternal half-siblings; and 0.25 (95% CI 0.20–0.29) for paternal half siblings (Table S15). Similar estimates were obtained after exclusion of severe/profound ID (Table S16). In quantitative genetic analysis, reduced models resulted in reduction of model fit for ACE, ADE, and AE-models with p-values >.20 in comparison to ADCE-model. AIC and BIC preferred the AE-model, thus we present results from the most parsimonious AE-model (Table S17). In the AE-model, the heritability (A) was estimated at 0.95 (95% CI 0.93–0.98), and nonshared environment (E) at 0.05 (95% CI = 0.02–0.07). Adjustment for sex and birth year resulted in similar estimates (Table S18).

Discussion Main findings

In this nationwide population-based study, we evaluated familial risk and heritability of ID. The sample of ID probands is 100 times bigger compared to the largest previous study on risk among siblings from British Columbia (Herbst & Baird, 1982). The use of several types of relatives allows evaluating the importance of genetic and environmental risk factors in the etiology of ID.

There are four main findings from our investigation. First, our 20-years risk and frequency of ID in full-siblings are comparable to old reports (Angeli & Kirman, 1975; Becker et al., 1977; Bundey & Carter, 1974; Bundey et al., 1989; Costeff & Weller, 1987; Durkin et al., 1976; Herbst & Baird, 1982; Laxova, Ridler, Bowen-Bravery, & Opitz, 1977; Turner et al., 1971; Turner & Partington, 2000). Now, we provide estimates with narrow confidence intervals also for halfsiblings and other relatives. Moreover, our study is first to report RR for ID comparing relatives of ID probands with relatives of unaffected individuals. High RR for monozygotic twins reflects ID characteristics with 0.9% prevalence and high heritability. Similar RR were reported for autism (Sandin et al., 2014), a disability with similar prevalence (1%–2%) and high (85%) heritability (Sandin et al., 2017).

Second, similar to the study by Bundey et al. (1989), we found no evidence for differences in familial risks related to comorbid medical conditions and adverse birth events, two probable biological mechanisms of ID. It is possible, that some of the birth complications traditionally considered as environmental risk factors reflect the response to genetic abnormalities in the fetus (Bolton et al., 1994; Simonoff, Bolton, & Rutter, 1996).

Third, our findings of an inverse relation between ID severity and the familial risk estimates are in accordance with older studies on siblings (Angeli & Kirman, 1975; Bundey & Carter, 1974; Durkin et al., 1976). We have observed a similar pattern of inheritance in second-degree relatives. A previous study by Reichenberg et al. (2016) showed a normal distribution of intellectual abilities among military conscripts who had siblings with severe ID. But those results are limited by the fact that individuals with medical diagnosis of ID are exempted from military assessment in Sweden and were not likely to be covered by the study. Our results showed a gradual risk decline in relation to level of intellectual impairment with shared etiological links across different levels of severity suggesting that the arbitrary cut-off for classifying ID severity into binary measure seems to be an oversimplification.

Forth, our data provide compelling evidence for the role of sex in familial risk of ID. In our study, we not only confirmed the higher risk for males compared to females in siblings of ID probands, but we also showed that similar sex differences were present in more distant relatives (Table S10). Higher prevalence of ID among males in comparison to females has been reported in population studies (Maulik et al., 2011) but studies in relatives produced ambiguous results, probably due to small sample size (Table S1). Our study confirms that, the sex differences observed in the general population are present in the relatives of ID probands, with males being more predisposed to ID. Matrilineallity was related to a higher risk in half-siblings and cousins of ID individuals in accordance with X-linked ID inheritance pattern.

Previous quantitative genetic methods focused on heritability of general (Kovas et al., 2013; Plomin & Deary, 2015) or low cognitive abilities within the normal range (Reichenberg et al., 2016). Heritability of ID as medical diagnosis has not been estimated before. The concordance in our study was as high as 73% for monozygotic, but relatively low of 9% for dizygotic twins. However, concordance is a measure on an absolute scale and depends on the disease prevalence in the population (Tenesa & Haley, 2013). With ID prevalence of 0.9%, genetic liability may be more accurately measured on a transformed scale by tetrachoric correlations where the liability for a disease is assumed to follow a standard normal distribution. The threshold on the liability distribution is estimated from the observed prevalence of the disease. Individuals are assumed to have a disease if they are above this threshold.

Further, severe ID is often linked to distinct genetic etiology caused by de novo mutations rather than polygenic traits associated with mild ID. In our study, tetrachoric correlations calculated separately for all ID levels and after exclusion of individuals with severe ID resulted in almost identical estimates, suggesting a low contribution of those phenotypes in ID heritability at the population level.

Heritability of ID was estimated to be 95%, considerably higher than 66%–86% reported for cognitive abilities in adulthood (Haworth et al., 2010; Panizzon et al., 2014). Those findings further highlighting importance of familial factors in diagnostics and etiological studies on ID. For future studies on the heritability of ID It is important to consider ‘maternal-effects in etiology’, previously described in quantitative genetic studies on general cognitive abilities (Devlin, Daniels, & Roeder, 1997). Substantially higher RR and 20-years risk estimates for maternal compared with paternal half-siblings were found in our study, suggesting that maternal factors may play an essential role also in clinically diagnosed ID.

Strengths and limitations

This study has several strengths, including the sample size, nationwide coverage, and access to both health and educational records. The ID prevalence of 0.9% in our study is close to 1% provided by the meta-analysis (Maulik et al., 2011). Swedish pediatric healthcare is publically funded with universal access to both primary and nonprimary healthcare (Wettergren, Blennow, Hjern, Soder, & Ludvigsson, 2016). A large number of relatives of ID probands allowed us to provide precise estimates of familial risk for several types of kinship. Th

留言 (0)

沒有登入
gif