Genetic epidemiology is an observational science that studies naturally occurring informative relationships to clarify the sources of the familial transmission of traits and disorders. One of the foundational questions in genetic epidemiology is the source of parent-offspring resemblance. Although other approaches are available [e.g. children of twins (McAdams et al., Reference McAdams, Neiderhiser, Rijsdijk, Narusyte, Lichtenstein and Eley2014) and assisted conception designs (Lewis et al., Reference Lewis, Rice, Harold, Collishaw and Thapar2011)], the paradigmatic method applied to this question has been the study of biological parents and their adopted-away offspring. However, this method is becoming increasingly difficult to implement due to declining rates of adoption (Wikipedia contributors, 2019). Furthermore, while elegant, the adoption design has several limitations including non-random placement of adoptees, unrepresentativeness of adoptive parents and frequent difficulty in obtaining data on biological fathers (Cadoret, Reference Cadoret1986).
The increasing availability of high-quality population-based data in national registries has expanded the designs available to the genetic epidemiologist. In the last several years, our group has proposed two such methods, the first of which is the triparental family (Kendler et al., Reference Kendler, Ohlsson, Sundquist and Sundquist2015b). In this design, we identified for the total Swedish population individuals who had three kinds of parents: a biological mother who reared them, a biological father who never lived with them and a step-father who reared them (Kendler et al., Reference Kendler, Ohlsson, Sundquist and Sundquist2015b). Thus, these three parents provided to their offspring, respectively, rearing + genes, genes only and rearing only.
In the second method, the multiple parenting relationships design, we identified from the Swedish population four kinds of informative parents with multiple children with whom they had different genetic and/or rearing relationships (Kendler et al., Reference Kendler, Ohlsson, Sundquist and Sundquist2019). These types had children for whom they provided: (i) genes (G) plus rearing (R), G only and R only; (ii) G + R and G only; (iii) G only and R only; and (iv) G + R and R only.
In this report, we describe and implement a third such informative design: Maternal Half-Sibling Families with Discordant Fathers (MHSFDF). Such families contain at least two maternal half-siblings (i) whose fathers are discordant for the disorder of interest, (ii) who differ in age by ⩽7 years, (iii) who each have lived with their shared biological mother for ⩾15 years while growing up, and (iv) have resided with their affected biological father for ⩽ one year. The key contrast in these families is the risk for the disorder of interest in the two half-siblings. These siblings share the same mother who reared them in the same household and community. Critically, they differ in one being sired by an affected and the other by an unaffected father.
We ascertained a Swedish national sample of these maternal half-sibling families with fathers discordant for drug abuse (DA) and alcohol use disorder (AUD) – classical externalizing syndromes – and major depression (MD): a paradigmatic internalizing disorder. We examine in these half-sibling pairs both their differential risk for DA, AUD and MD and a range of other potential risk indices with the goal of clarifying the nature of the genetically transmitted paternal risk for externalizing and internalizing psychopathology.
Methods
This study utilized several different Swedish population-based registers with national coverage, linking them using each person's unique identification number. To preserve confidentiality, this ID number was replaced by a serial number. We secured ethical approval for this study from the Regional Ethical Review Board located at Lund University (no. 2008/409).
From the Multigenerational register, we selected all maternal half-sibling pairs born in Sweden between 1960 and 1990 with a maximum 7-year age difference. Using information from the Population and Housing Census and the Total population register, we included in the database the number of years, during ages 0–15, that the siblings resided in the same household as their biological mother and biological father. Furthermore, in the database we included, among the half-sibling pairs, information on the following variables: AUD, DA, MD, attention deficit hyperactivity disorder (ADHD), conduct disorder (CD), concussion, academic achievement (AA), years of education, pre-term birth, small for gestational age, birth weight, gestational age, Apgar-score (1 min) and birth length. Among the biological fathers and biological mothers, we included information about AUD, DA and MD. For a definition of all variables see online Supplementary Appendix 1.
The focus of interest in the following analyses was pairs in which the two biological fathers were discordant for AUD (or DA or MD). For each of the three phenotypes (AUD, DA, and MD) we created four samples to study the effect of AUD (DA, MD) in biological father on several outcomes measured in the child.
Due to limitations in registration coverage, we had to construct these four different samples. In the first sample, we selected sibling pairs who each have lived in the same household as their shared biological mother for at least 15 years while growing up and have resided maximum 1 year in the same household with an affected (AUD/DA or MD) biological father. This restriction was made to limit the environmental influence of the affected biological father on the children. We also required that both siblings in the pair were residing in Sweden at age 16. In total 73 108 pairs lived in the same household as their shared biological mother for at least 15 years.
The second sample was identical to sample 1, with the exception that we required that the child was registered in the National School Registry that contains AA (a grade point average) for all students at the end of grade nine (usually age 16) from 1988 to 2014. In total 36 589 maternal half-sibling pairs met these criteria.
The third sample was also almost identical to sample 1 except that the child had to have the number of years of education registered in the Swedish registers. In this sample, we also excluded pairs where at least one in the sample was registered for AUD (DA or MD) so that we could rule out the possibility that the impact of the paternal disorder on educational attainment was mediated through transmitting the disorder to their child. In total 71 759 maternal half-sibling pairs met these criteria.
In the fourth sample we required that both siblings in the pair were registered in the Medical Birth Register. As the outcome variables were measured at birth, we made no restrictions on the number of years the siblings had resided with their biological mother or biological father. In total there were 57 059 half-sibling pairs where we had information from the medical birth register.
For a description of the overlap between the four samples, see online Supplementary Appendix Table S2.
Due to the distribution of the outcome measures as well as when in time they were measured, we were required to use three different types of regression models. The main exposure variable of interest in all analyses was AUD, DA, or MD in the biological father. For binary outcome measures where the onset could occur at different ages (AUD, DA, MD, ADHD, CD, and concussion) we used stratified Cox proportional hazards models, with a separate stratum for each half-sibling pair. The use of a separate stratum allowed us to adjust for a range of unmeasured factors among the half-siblings attributable to the fact that they share the same biological mother. These types of models are also referred to in the literature as within-cluster analyses, fixed effects regression, or conditional regressions. Follow-up time in a number of weeks was measured from age 15 of the offspring (for ADHD, CD and concussion follow-up time was measured from date of birth) until of first registration for AUD (DA or MD), death, emigration or end of follow-up (year 2015), whichever came first. For concussion, we ended the follow-up time at age 18. In all models we controlled for year of birth and sex of a child. In all models the proportionality assumption was fulfilled. The corresponding hazard ratio refers to the risk-increase in the child with an affected father compared to their half-sibling with an unaffected father.
For binary outcome measures with onset at birth [pre-term birth, small for gestational age, Apgar-score (1 min)], we used stratified logistic regression models, with a separate stratum for each half-sibling pair. In all models, the effect of AUD (DA, MD) in the biological father was controlled for by maternal age, sex of child and birth order. In none of our analyses did we control for features of the biological mother as she was the same individual for both half-siblings. The corresponding odds ratio refers to increased odds for the child with an affected father compared to their sibling with an unaffected father.
For continuous outcome measures (AA, years of education, birth weight, gestational age, and birth length), we used stratified linear regression models, with a separate stratum for each half-sibling pair. In the analysis of AA and years of education the effect of AUD (DA, MD) in the biological father was controlled for by year of birth and sex in a child. In the analysis of birth weight, gestational age, and birth length, the effect of AUD (DA, MD) in the biological father was controlled for by maternal age, sex of child and birth order. The corresponding β-coefficients refers to absolute increase for the half-sibling with an affected v. unaffected father.
In a sensitivity analysis for variables included in samples 1–3, instead of restricting cohabitation with the affected biological father, we included all families and controlled in an additional model for the number of years that the offspring resided with an affected father. Note that this father could either be the biological father or the biological father to the half-sibling. For variables in the medical birth register (sample 4) we only had information on smoking status in the mother at gestational week 10–12 and her cohabitation status with the biological father at birth of the child for a sub-set of the sample. In a sensitivity analysis we controlled for these variables in two separate models.
All statistical analyses were performed using SAS 9.4 (SAS Institute, 2012).
Results
Samples
Details of the four inter-related samples of our discordant maternal half-sibling families are provided in Table 1.
Descriptive statistics on maternal half-sibling families with fathers discordant for AUD, DA and MD
Cross-generational transmission of AUD, DA and MD
The number of informative pairs of maternal half-siblings identified for these analyses (sample 1 in Table 1) varied from a low of 4800 for DA to a high of 15 274 for AUD. The mean year of birth was very similar for the unaffected and affected fathers. However, for AUD, DA and MD, the offspring of the unaffected fathers were on average 3 years younger than those of the affected fathers, suggesting that age should be controlled for in subsequent analyses. The raw prevalence of the disorder of interest was, for all three syndromes, considerably more common in the offspring of the affected v. unaffected fathers.
The results of our Cox proportional hazard models, controlling for the child's year of birth and sex, are seen in sample 1 section of Table 2 and bolded. For AUD, DA and MD, the HR [and 95% confidence intervals (CIs)] for the offspring of the affected v. unaffected father were, respectively, 1.72 (1.61–1.84), 1.55 (1.41–1.70) and 1.51 (1.40–1.64). We also examined cross-disorder parent-offspring transmission (Table 2). The results suggested substantial sharing of genetic risk between AUD and DA and more modest sharing between these two substance use disorders and MD.
Prediction of risk in offspring from maternal half-sibling families with fathers discordant for AUD, DA and MD
a Stratified Cox regression models, controlled for year of birth of child and sex of child. Numbers are hazard ratios and 95% CIs.
b Stratified linear regression models, controlled for year of birth of child and sex of child. Numbers are β coefficients and 95% CIs.
c Stratified logistic regression models, controlled for maternal age, sex of child and birth order. Numbers are odds ratios and 95% CIs.
d Stratified linear regression models, controlled for maternal age, sex of child and birth order. Numbers are β coefficients and 95% CIs.
Putative indices of genetic risk assessed in childhood and adolescence
We examined, in the MHSFDF, several classes of potential indices for genetic risk that were assessed from birth through young adulthood. As outlined in Table 1, we required four different samples to evaluate these indices. Starting with those measured in sample 1, rates of ADHD, CD and concussions before the age of 18 –a rough index of impulsivity – were all significantly higher in the offspring of fathers affected v. unaffected with DA and AUD, with consistently stronger effects seen for DA. By contrast, none of these indices was significantly associated with MD in the affected v. unaffected father.
We next examined AA and educational attainment which were assessed, respectively, in samples 2 and 3 (Table 1). Significant reductions were seen in both measures in the offspring of the affected v. unaffected fathers with stronger associations seen with fathers with DA. By contrast, no significant differences in these variables were seen in the half-siblings whose fathers did v. did not have MD.
Putative indices of genetic risk assessed in the birth registry
The final set of six putative risk indices that we examined came from the Medical Birth register, assessed in sample 4 (Table 1). Controlling for maternal age, sex of child and birth order, we examined a range of pregnancy outcomes in the half-siblings of fathers discordant for our three disorders. Offspring of fathers with v. without DA had the worst outcomes, with significant findings for all six indices. That is, they had higher risks of being small for gestational age and preterm had lower APGAR scores, and were significantly lighter, younger and shorter at birth. Offspring of fathers with v. without AUD and MD were at somewhat lower risk, differing significantly on, respectively, four and three of these indices. Further, examining the magnitude of deviation of the high v. low-risk offspring on these measures, for all but birth length, the order was DA > AUD > MD.
We examined three important possible confounds. First, mothers were significantly more likely to smoke during the pregnancy carrying the child of the father affected with AUD, DA and MD compared to when they were pregnant with the child from the unaffected father. Therefore, we controlled for maternal smoking status during pregnancy for the sub-sample with available data for AUD, DA and MD in online Supplementary Appendix Tables S3–S5. No appreciable changes were seen in the observed associations. Second, residing while pregnant with the affected v. with the unaffected spouse could impact on pregnancy outcomes for a variety of reasons ranging from levels of social support, passive smoke exposure, quality of maternal nutrition or frequency of prenatal care. We controlled for cohabitation with the biological father in analyses seen in online Supplementary Appendix Tables S3–S5. No impact of cohabitation was seen on any of the pregnancy outcomes examined. Third, for fathers affected with DA or AUD, could their substance use at or around the time of conception have, either via biological or psychological pathways, an impact on the pregnancy outcomes of their child? We examined this by eliminating from our sample affected fathers with a registration for AUD or DA for a 3-year period centered on the time of conception. As seen in online Supplementary Appendix Table S6, this had minimal effects on our results.
An alternative design controlling for contact with the affected biological father
In our main design, we only examined maternal half-sibling families who had very limited contact with the affected biological father. We modified our design to examine all eligible families and controlled for the duration of the cohabitation with the affected father in predicting risk for AUD, DA and MD in the offspring of the affected v. unaffected fathers (Table 3). The increase in our number of informative families was substantial, ranging from 56% for AUD to 184% for DA. The results were similar to those obtained with our more restricted MHSFDF sample.
Prediction of risk in offspring from maternal half-sibling families with fathers discordant for AUD, DA and MD controlling for years of cohabitation with the affected father
a Stratified Cox regression models, controlled for year of birth of child, sex of child and number of years in the same household as the affected father. Numbers are hazard ratios and 95% CIs.
Discussion
We sought to introduce and implement a novel genetic epidemiological design to assess parent-offspring genetic transmission which utilized maternal half-sibling families with discordant biological fathers controlling for environmental contact with the affected father. This design is appealing because reared together maternal half-siblings share the same mother differing only in their paternal genetic endowment. They also would typically share the same family, school and community, although being different ages, these environments will not be identical for the two half-siblings. Unlike adoption samples, no legal proceedings or parental screenings are involved. Other designs that assess parent-offspring transmission examine resemblance between parents and offspring. This design is contrastive because in a single design we examine the difference in risk between two matched offspring in the same family.
We had five major findings. First, we replicated prior evidence for the cross-generational genetic transmission of risk for AUD (Goodwin, Reference Goodwin1981; Sigvardsson et al., Reference Sigvardsson, Bohman and Cloninger1996; Kendler et al., Reference Kendler, Ji, Edwards, Ohlsson, Sundquist and Sundquist2015a), DA (Cadoret et al., Reference Cadoret, Troughton, O'Gorman and Heywood1986, Reference Cadoret, Yates, Troughton, Woodworth and Stewart1995; Kendler et al., Reference Kendler, Sundquist, Ohlsson, Palmer, Maes, Winkleby and Sundquist2012) and MD (Cadoret, Reference Cadoret1978; Wender et al., Reference Wender, Kety, Rosenthal, Schulsinger, Ortmann and Lunde1986; Kendler et al., Reference Kendler, Ohlsson, Sundquist and Sundquist2018). It is of interest to compare our results with those found using the adoption or triparental family design in Sweden which were based on the same diagnostic methods used in this report (Table 4). For AUD, our results are similar to those previously found. For DA and MD, MHSFDF produce, respectively, evidence for somewhat weaker and stronger genetic effects than seen previously. Second, consistent with prior twin studies (Kendler et al., Reference Kendler, Prescott, Myers and Neale2003, Reference Kendler, Aggen, Knudsen, Roysamb, Neale and Reichborn-Kjennerud2011, Reference Kendler, Lonn, Maes, Lichtenstein, Sundquist and Sundquist2016a), we found cross-generational evidence for a close sharing of genetic risk factors for AUD and DA, as expected for two externalizing disorders, with more modest genetic transmission between MD and these two substance use disorders.
A comparison of the magnitude of parent-offspring genetic transmission of AUD, DA, criminal behavior and MD estimated in MHSFDF, triparental families and adoption studies
a In this design, the not-lived with father contributes genes to their offspring but never lived with or near them and so provides no rearing.
b Odds ratios, otherwise hazard ratios are presented.
Third, in accord with prior twin samples (Kendler et al., Reference Kendler, Aggen, Knudsen, Roysamb, Neale and Reichborn-Kjennerud2011; Edwards and Kendler, Reference Edwards and Kendler2012), we demonstrated a cross-generational association between paternal genetic risk for DA and AUD and risk in offspring for ADHD and CD. Of note, no such link was seen for MD. The pattern was the same for concussions before age 18 – a plausible rough index of impulsivity (Beidler et al., Reference Beidler, Covassin, Donnellan, Nogle, Pontifex and Kontos2017; Edwards et al., Reference Edwards, Witt, Dawson, Harkey and Bui2018). We previously examined, in both adoptive and not-lived-with samples in Sweden, the association between DA in fathers and ADHD in reared apart biological offspring (Kendler et al., Reference Kendler, Ohlsson, Sundquist and Sundquist2016b). The estimate across samples was HR = 2.10 (1.92–2.29), close to our finding here: 2.04 (1.74; 2.41).
Fourth, genetic risk for AUD and DA, but not MD, predicted poorer AA at age 16 and fewer years of education. Our results are consistent with prior evidence both from genetic epidemiologic methods (Latvala et al., Reference Latvala, Kuja-Halkola, D'Onofrio, Larsson and Lichtenstein2016) and from polygenic risk score analysis (Walters et al., Reference Walters, Polimanti, Johnson, McClintick, Adams, Adkins, Aliev, Bacanu, Batzler, Bertelsen, Biernacka, Bigdeli, Chen, Clarke, Chou, Degenhardt, Docherty, Edwards, Fontanillas, Foo, Fox, Frank, Giegling, Gordon, Hack, Hartmann, Hartz, Heilmann-Heimbach, Herms, Hodgkinson, Hoffmann, Jan, Kennedy, Alanne-Kinnunen, Konte, Lahti, Lahti-Pulkkinen, Lai, Ligthart, Loukola, Maher, Mbarek, McIntosh, McQueen, Meyers, Milaneschi, Palviainen, Pearson, Peterson, Ripatti, Ryu, Saccone, Salvatore, Sanchez-Roige, Schwandt, Sherva, Streit, Strohmaier, Thomas, Wang, Webb, Wedow, Wetherill, Wills, Boardman, Chen, Choi, Copeland, Culverhouse, Dahmen, Degenhardt, Domingue, Elson, Frye, Gabel, Hayward, Ising, Keyes, Kiefer, Kramer, Kuperman, Lucae, Lynskey, Maier, Mann, Mannisto, Muller-Myhsok, Murray, Nurnberger, Palotie, Preuss, Raikkonen, Reynolds, Ridinger, Scherbaum, Schuckit, Soyka, Treutlein, Witt, Wodarz, Zill, Adkins, Boden, Boomsma, Bierut, Brown, Bucholz, Cichon, Costello, de, Diazgranados, Dick, Eriksson, Farrer, Foroud, Gillespie, Goate, Goldman, Grucza, Hancock, Harris, Heath, Hesselbrock, Hewitt, Hopfer, Horwood, Iacono, Johnson, Kaprio, Karpyak, Kendler, Kranzler, Krauter, Lichtenstein, Lind, McGue, MacKillop, Madden, Maes, Magnusson, Martin, Medland, Montgomery, Nelson, Nothen, Palmer, Pedersen, Penninx, Porjesz, Rice, Rietschel, Riley, Rose, Rujescu, Shen, Silberg, Stallings, Tarter, Vanyukov, Vrieze, Wall, Whitfield, Zhao, Neale, Gelernter, Edenberg and Agrawal2018) suggesting an inverse relationship between genetic risk for AUD and educational attainment.
Finally, using the Medical birth registry, we found significant differences in pregnancy outcomes in the offspring of affected v. unaffected fathers. The differences were most marked for those at genetic risk for DA and least for MD. We evaluated three plausible confounds: maternal smoking during pregnancy, contact with affected father (Shah et al., Reference Shah, Gee and Theall2014), and heavy paternal substance use around conception [(Shah, Reference Shah2010) but see Passaro et al., Reference Passaro, Little, Savitz and Noss1998; Mutsaerts et al., Reference Mutsaerts, Groen, Buiter-Van der Meer, Sijtsma, Sauer, Land, Mol, Corpeleijn and Hoek2014)]. None substantially attenuated the findings. While our confounder analyses make these explanations less likely, we cannot, with observational data, confidently rule them out.
Our results provide insights into the nature of the genetically transmitted risk from fathers to children for externalizing and internalizing disorders. The impact of genetic risk for externalizing disorders was broader, influencing in childhood through early adulthood risk for (i) psychiatric disorders (ADHD, CD), (ii) impulsive behavior (higher concussion risk), (iii) worse AA, and (iv) poorer educational attainment. In our analyses, the genetic risk for MD was narrower, having no significant effect on these outcomes. Finally, we found tentative evidence for an impact of paternal genetic risk for DA and AUD on pregnancy outcomes although risk for MD also had an effect albeit less generalized and weaker.
Limitations
These results should be interpreted in the context of several potential methodological limitations. While ascertaining cases of DA, AUD and MD from registry data have important advantages, especially independence from subject cooperation and accurate recall, it also has significant limitations. For DA and AUD, there are surely false negatives for individuals who abuse substances but avoid medical or police attention. However, the validity of our detection of these syndromes is supported by evidence for strong associations of cases detected from different registries. The mean OR for case detection of DA across our five relevant registries was 52 (Kendler et al., Reference Kendler, Sundquist, Ohlsson, Palmer, Maes, Winkleby and Sundquist2012) and for AUD across four registries was 33 (Kendler et al., Reference Kendler, Ji, Edwards, Ohlsson, Sundquist and Sundquist2015a). For MD, our prevalence rates were approximately half of that found in an epidemiologic survey in next-door Norway (Kringlen et al., Reference Kringlen, Torgersen and Cramer2001) and in a large interview-based study in the Swedish twin registry (Kendler et al., Reference Kendler, Gatz, Gardner and Pedersen2006), consistent with estimates from Western countries that around half of individuals suffering from MD present for treatment (Kessler et al., Reference Kessler, Berglund, Demler, Jin, Koretz, Merikangas, Rush, Walters and Wang2003).
We made several assumptions in our design that merit examination and we performed sensitivity analyses to examine four of them. First, we narrowed age differences between the offspring of the affected and unaffected fathers from <8 years to ⩽4 years, thereby increasing similarity of their rearing experiences. Analyses of this sample produced estimates similar to those in Table 2, although the impact of genetic risk for DA was modestly attenuated (online Supplementary Appendix, Table S6). Second, we required that affected biological fathers live with their offspring for ⩽1 year, but some might live near-by and interact frequently. We re-analyzed our sample requiring that affected biological fathers not live in the same neighborhood as their children (online Supplementary Appendix Table S6). This produced little change in our findings. Our main analyses limited cohabitation with the affected father but placed no restrictions on the unaffected father. We modeled the impact of years living with the unaffected father on our measures of transmission. As seen in online Supplementary Appendix Table S6, this produced a modest decrease in cross-generational transmission for AUD, a modest increase for DA and no change for MD. Third, the uterine and rearing experiences of children are not identical across pregnancies particularly as mother's age. To address possible systematic biases that might have impacted on our analyses in this way, we examined cross-generational transmission controlling now for birth order in the half-sib pairs. At most modest changes were observed as the HRs for AUD, DA and MD changed respectively from 1.72 (1.61–1.84) to 1.71 (1.60; 1.83), 1.55 (1.41–1.70) to 1.55 (1.41–1.69) and 1.51 (1.40–1.64) to 1.59 (1.38; 1.61).
In this design, we able to examine, in adequate numbers, only fathers of maternal half-siblings thereby limiting our ability to examine sex effects in cross-generational transmission. We therefore explored whether father to son and father to daughter transmission differed in any of our analyses. As seen in online Supplementary Appendix Table S7, four of the nine results were statistically significant, three of which found stronger father→son than father→daughter transmission. These results are consistent with prior evidence from twin-sibling analyses that DA demonstrates modest higher within- than across-sex transmission (Kendler et al., Reference Kendler, Maes, Sundquist, Ohlsson and Sundquist2013).
Given that our analyses (Table 3) containing all fathers controlling to length of contact with offspring produced very similar results to our more restricted original sample, the question arises of what we would find if we repeated such analyses without the inclusion of length of contact as a covariate. As seen in online Supplementary Appendix Table S8, these results differ very little for those seen in Table 3. Contrary to results we have obtained for our three phenotypes in adoption (Kendler et al., Reference Kendler, Sundquist, Ohlsson, Palmer, Maes, Winkleby and Sundquist2012, Reference Kendler, Ji, Edwards, Ohlsson, Sundquist and Sundquist2015a, Reference Kendler, Ohlsson, Sundquist and Sundquist2018), triparental (Kendler et al., Reference Kendler, Ohlsson, Sundquist and Sundquist2015b) and multiple parenting relationships design (Kendler et al., Reference Kendler, Ohlsson, Sundquist and Sundquist2019), while not the focus of this method, our MHSFDF design provides little evidence in favor of parent-offspring familial-environmental transmission of AUD, DA and MD.
Finally, our results were specific to maternal half-sibling families and may or may not extrapolate to my typical family constellations.
Conclusions
The MHSFDF is an elegant design to clarify cross-generational genetic transmission by comparing maternal half-siblings who differ only in their paternal genetic endowment, sharing the same mother, and most aspects of the family, school and community. This contrastive design clearly demonstrates the genetic transmission from parent to offspring of AUD, DA and MD and the close genetic link between the two substance abuse syndromes. We extend this design to show a genetic relationship between paternal DA and AUD, but not MD, and ADHD, CD and poor educational performance in their offspring. In these families, the offspring of the affected father had poorer pregnancy outcomes and this effect was strongest for paternal DA and weakest for paternal MD. In the presence of registry information, maternal half-sibling families represent an important source of information to clarify the magnitude of the cross-generational transmission of psychiatric and substance use disorders and the nature of the potential mediating mechanisms. By applying multiple genetic-epidemiological designs, each with different assumptions and potential methodological weaknesses, we can, if we obtain similar results, be increasingly confident of their validity (Munafo and Davey, Reference Munafo and Davey2018).
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S0033291719000874
Author ORCIDs
Kenneth S. Kendler, 0000-0001-8689-6570
Acknowledgements
This project was supported by grants R01DA030005 and R01AA023534 from the National Institutes of Health, the Swedish Research Council (2014–02517, 2014–10134 and 2016–01176) and ALF funding awarded from Region Skåne.
Conflict of interest
The authors have no conflicts of interest to declare.
Ethical standards
The authors assert that all procedures contributing to this work comply with the ethical standards of the relevant national and institutional committees on human experimentation and with the Helsinki Declaration of 1975, as revised in 2008. We secured ethical approval for this study from the Regional Ethical Review Board of Lund University (No. 2008/409).
