Evaluating the impact of non-random mating: psychiatric outcomes among the offspring of pairs diagnosed with Schizophrenia and Bipolar Disorder

Abstract

Background:

Nonrandom mating has been shown for psychiatric diagnoses, with hypothesized - but not quantified - implications for offspring liability. This national cohort study enumerated the incidence of major psychiatric disorders among the offspring of parent pairs affected with schizophrenia and/or bipolar disorder (i.e., dual-affected pairs).

Methods:

Participants were all Swedish residents alive/born 1968-2013 (n = 4,255,196 unique pairs and 8,343,951 offspring). Offspring with dual-affected, single-affected, and unaffected parents were followed (1973-2013) for incidence of broad psychiatric disorders. Primary outcomes included hazard ratio (HR) and cumulative incidence for schizophrenia and bipolar disorder in the offspring. Additional outcomes included any neuropsychiatric, anxiety, depressive, personality, or substance use disorders. Cumulative incidences of schizophrenia and bipolar disorder were used to inform heritability models for these disorders.

Results:

Hazards were highest within-disorder (e.g., offspring of dual-schizophrenia pairs had sharply raised hazards for schizophrenia [HR = 55.3]), however were significantly raised for all diagnoses (HR range: 2.89-11.84). Incidences were significantly higher for the majority of outcomes, with 43.4-48.5% diagnosed with ‘any’ disorder over follow-up. Risks were retained, with modest attenuations, for the offspring of heterotypic pairs. The estimated heritability of liability for schizophrenia (0.62 [0.55-0.70]) and bipolar disorder (0.52[0.46-0.58]) did not differ significantly from estimates derived from single-affected parents.

Conclusions:

Risks for a broad spectrum of psychiatric diagnoses are significantly raised in the offspring of dual-affected parents, in line with expectations from a polygenic model of liability to disease risk. How these risks may contribute to population maintenance of these disorders is considered.

Introduction

Evidence is strong that nonrandom mating occurs for psychiatric traits(1–4) as it does for features such as height, personality, and IQ,(5–7) with recent epidemiological work illustrating pervasive spousal resemblance within and across major diagnostic groups.(8) Such findings have renewed questions about the impact nonrandom mating may have on the transmission dynamics and population maintenance of psychiatric disorders.(9,10)

The double-predisposition inherent to offspring of dual-affected pairs (both at the level of genetic risk and family environment) has been hypothesized to place this group at heightened risk for psychiatric disorders(9,11) With evidence converging on a shared genetic architecture for many major psychiatric conditions, the potential for these risks to extend beyond a parent’s specific diagnoses has also been highlighted.(12,13) If present, such risks would reflect a complex interplay of factors, with nonrandom mating introducing new - or, alternatively, augmenting known - liability pathways. (9,10)

Though theoretically consistent, study of these hypothesized risks and mechanisms has been limited. Recent simulation work (10) concluded that nonrandom mating at the scale observed in current literature - if fully attributable to assortative mating - could impact heritability and increase the population prevalence of select disorders 1.5-fold over one generation. However, such simulations carry strong assumptions and, while useful for quantifying an upper-limit, real-world assessments are needed to pinpoint true effects.

National register data have previously been leveraged to these ends, with a Danish investigation showing much higher cumulative incidences for both schizophrenia and bipolar disorder (> 25%) among the offspring of homotypic parents (e.g., both partners having the same diagnosis)(11), compared to offspring of single-affected or unaffected pairs. A novel attempt to extend this examination to heterotypic pairs (i.e., one parent with schizophrenia, the other bipolar disorder) was underpowered (despite national sample sizes of 2.7m) and cross-disorder outcomes were not extensively interrogated. In subsequent Danish work, Dean and colleagues surveyed a broad spectrum of cross-disorder outcomes among the offspring of more loosely defined heterotypic pairs (one parent with ‘psychosis,’ co-parent any alternate diagnosis). Their findings suggested heightened risks across the full range of examined conditions,(14) however, differential risks by specific parental diagnoses (e.g., dual-schizophrenia versus dual-bipolar) could not be assessed.

Access to data that permits study of well-defined, homotypic (same-disorder) and heterotypic (differing disorder) pairs is critical for untangling the relationship between dual-parental psychopathology, offspring outcomes, and population disorder maintenance. The Swedish medical registers are unique in scale, offering coverage of psychiatric diagnoses for > 4.2 million parental pairs and 8.3 million offspring since records began in 1969. Using these data, the present study will consider the cumulative incidence (by age 42) of broad psychiatric conditions among the offspring of parental pairs affected with schizophrenia and/or bipolar disorder. To evaluate the impact of nonrandom mating on liability, these full-population incidences will then be used to inform heritability models, testing for variation in liability to schizophrenia and bipolar disorder among the offspring of dual versus single-affected pairs. Consistent with expectations from a polygenic model of inheritance, we hypothesize no difference in disease heritability between the offspring of single versus dual-affected pairs.

Methods and Materials

Approval for the study was granted by the Regional Ethical Review Board in Stockholm, Sweden (reference 2013/862-31/5). All data were de-identified and collected as part of governmental health and resource tracking. Accordingly, the requirement for informed consent was waived.

Swedish National Registers

Data were linked across multiple total population registers via unique civic registration numbers (personnummer). These identifiers are assigned to all Swedish citizens at birth and all immigrants upon arrival in Sweden.(15) Data were derived and aggregated from the following, independent registers:

Demographic data were drawn from The Swedish Total Population Register (TPR), which records basic information (e.g., sex) for all Swedish residents (1968+). The Multi-Generation Register (MGR), which defines the relationship between Swedish persons, was used to create pedigrees for individuals born (1933+) or living in (1960+) Sweden.(16) The Cause of Death Register records all deaths in Sweden since 1952, and was used to define cohort follow-up. The Migration Register was used to identify emigration of cohort members outside of Sweden.

The National Patient Register (NPR) was used to identify both exposure (parental psychiatric diagnosis) and outcome (offspring psychiatric diagnosis) in the present study. To be captured in the register, patients must seek care through either inpatient or select outpatient services, with the NPR containing nationwide coverage of inpatient care data (1969+, with complete psychiatric information 1973+), as well as information on outpatient specialist services (since 2001). Each visit is recorded in the register with corresponding primary discharge diagnoses [International Classification of Diseases (ICD), eighth (ICD-8; 1969-1986), ninth (ICD-9; 1987-1996), or tenth (ICD-10; 1997-present) revisions]. In the present study, these diagnoses have been used to define both exposure (e.g., parental diagnoses) and outcome (e.g., offspring diagnoses). The Prescribed Drug Register (PDR), which records all medications prescribed and dispensed in Sweden [Anatomical Therapeutic Chemical (ATC) Classification System codes], was also used in select outcome definitions..

Study Population

A population-based cohort consisting of all Swedish residents alive/born 1968-2013 was generated, containing 8,343,951 index offspring from 4,255,196 unique parent pairs. These indexes were followed from either 1973 (the point at which the NPR achieves full psychiatric coverage) or their more recent date of birth.

Measures

Exposures

Parental exposure groups were generated, separately, for schizophrenia and bipolar disorder during the study period. Consistent with convention,(17) our definition of schizophrenia (SCZ) required ≥2 discharge diagnoses from the following list of SCZ codes: 295 excluding .5 (ICD-8), 295 excluding F (ICD-9), and F20, F23.1/.2, or F25 (ICD-10).(18) For bipolar disorder (BIP), ≥2 of the following were permissible: 296 excluding .2 (ICD-8), 296 excluding B, and F30/F31 (ICD-10).(19)

Exposure groups were: 1) homotypic dual-affected parental pairs (e.g., both parents with registered SCZ diagnoses), 2) heterotypic dual-affected pairs (e.g., one parent SCZ, the other BIP), 3) single-affected parental pairs (e.g., one parent SCZ, the other no SCZ diagnosis), and f) general population pairs (e.g., neither parent diagnosed during the study period). To capture all diagnostic combinations and permit exploration of sex effects, two samples each were generated for single-affected groups (e.g., one sample in which fathers had the SCZ diagnosis; another in which mothers carried the diagnosis) and heterotypic pairs (e.g., one sample in which fathers had a SCZ diagnosis and their opposite-sex partner BIP, the other sample visa-versa).

Consistent with the high rates of comorbidity inherent to psychiatric samples, no restrictions were imposed on the number of diagnoses permissible for either parents or their offspring. Thus, a SCZ parent with a BIP partner may appear as part of the ‘single-affected’ exposure group in an analysis considering homotypic SCZ pairs, and later be classed as ‘dual-affected’ in analysis considering heterotypic SCZ/BIP pairs. Likewise, their offspring may be classed as an incident case for multiple outcomes (see next section). General population pairs, while limited in relevant analyses to individuals without SCZ and/or BIP, had no further diagnostic restrictions. Utilization of these ‘uncleaned’ samples is common in register-based work, including studies considering offspring psychiatric outcomes,(20) and has been employed here to ensure conservative, real-world estimates of risk.

Due to the length of follow-up in this study (eldest offspring aged 41 years), some individuals may appear both as offspring as well as member of a parental pair. Censoring occurred for diagnosis, migration out of Sweden, or death prior to end of the follow-up period.

Outcomes

Offspring of all parental pairs were followed-up to check for registration of psychiatric diagnoses, including SCZ, BIP, and additional diagnoses of interest. These disorders were grouped into categories and defined operationally, per Supplemental Table 1. An aggregate ‘any disorder’ outcome was also established, reflecting diagnosis of any disorder of interest during the follow-up period. The majority of ICD codes underlying all key outcomes have been subjected to validation studies.(21,22)

Additional Variables

Essential covariates extracted from the TPR included parent/offspring gender and birth year of parents (used to compute parental age at birth).

Follow-Up

Offspring were followed from either 1973 or their more recent date of birth, through to the first of the following: outcome event (registered diagnosis), death, emigration, or study conclusion (December 31, 2013). This follow-up procedure was repeated anew for each disorder of interest, thus timing of the ‘first’ event in the incidence figures may vary within individual by examined outcome.

Statistical Analysis

Hazard Ratios

Hazard ratios (HR) for offspring outcomes were computed for each of the following exposure groups, relative to the offspring of general population indexes: offspring of homotypic parental pairs (both parents with the same diagnosis), offspring of heterotypic parental pairs (one parent with SCZ, the other BIP), and offspring of single-affected parental pairs (only one partner diagnosed). Both single-affected and heterotypic estimates were produced by parent gender. HR estimates were adjusted for the key confounders of parental age (treated categorically as decade of birth) and age at birth of each child (5-year increments from age 20).

Cumulative Incidence

Cumulative incidence estimates were subsequently generated for each disorder of interest, again by parental exposure group (homotypic, heterotypic, single-affected, and general population). These figures represent the incidence of a register-based diagnosis (e.g., the proportion of individuals with this diagnosis) by conclusion of study follow-up (age 41).

Data extraction, formatting, and estimation of FRs/HRs were undertaken in SAS (version 9.4). Cumulative incidences were computed and visualized using R (version 3.4.3).

Heritability

In order to understand if the observed rates of psychiatric disorders in children of dual-affected pairs is consistent with expectations under a polygenic model, we estimated heritabilities for schizophrenia $\left(h_{SCZ}^2\right)$ and bipolar disorder $\left(h_{BIP}^2\right)$ from the HR estimates using equations provided Wray and Gottesman(23) derived from liability threshold model theory.(23–25)

Results

Within-Disorder Risks

Homotypic Parents

Schizophrenia.

The age-adjusted risk of SCZ among the offspring of homotypic pairs (see Table 1), was high relative to the general population (HR = 55.3, 95% CI = 39.9-76.7). These risks were also substantially higher than those observed in the offspring of single-affected partnerships (range= 8.4-8.8, 95% CI = 7.5-9.5; see Table 2). By age 41, the expected cumulative incidence of schizophrenia in the offspring of homotypic pairs reached 18.7% (95% CI = 13.1-24.8), exceeding the incidence rates in all comparison groups (range: 0.5-4.0%; see supplement).

Table 1.

Number of diagnosed offspring by parental diagnostic combination

		Parent Diagnostic Combination
		Homotypic				Heterotypic				Single-Affected
Offspring N	Total N	SCZ	%	BIP	%	mSCZ fBIP	%	mBIP fSCZ	%	mSCZ	%	fSCZ	%	mBIP	%	fBIP	%
All	8343951	265	0.00	537	0.01	195	0.00	240	0.00	18832	0.23	11864	0.14	56992	0.68	33581	0.40
SCZ	27996	36	0.13	6	0.02	12	0.04	17	0.06	653	2.33	268	0.96	521	1.86	254	0.91
BIP	47123	18	0.04	52	0.11	16	0.03	25	0.05	421	0.89	267	0.57	1929	4.09	1113	2.36

Table 2.

Within and cross-disorder hazard ratios for offspring of homotypic, heterotypic, and single-affected pairs, relative to general population

			Offspring HR
Mother	Father	Offspring	Both Affected			Mother Only			Father Only
			HR	Lower	Upper	HR	Lower	Upper	HR	Lower	Upper
SCZ	SCZ	SCZ	55.33	39.90	76.74	8.78	8.13	9.49	8.42	7.51	9.44
		NPSY	4.13	2.89	5.90	2.08	1.94	2.23	2.08	1.93	2.25
		ANX	2.89	2.05	4.08	1.92	1.83	2.02	2.01	1.89	2.15
		DEP	3.37	2.47	4.62	1.82	1.73	1.91	1.96	1.84	2.08
		BIP	11.84	7.46	18.79	3.38	3.07	3.72	4.01	3.56	4.52
		SUB	4.37	3.31	5.78	1.90	1.81	1.99	2.04	1.92	2.16
		PERS	7.18	4.63	11.13	3.14	2.91	3.38	2.97	2.68	3.29
		ANY	4.40	3.57	5.43	2.24	2.17	2.32	2.26	2.16	2.35
BIP	BIP	BIP	23.28	17.64	30.72	6.57	6.27	6.88	6.29	5.93	6.68
		NPSY	4.11	3.23	5.23	2.41	2.33	2.49	1.96	1.87	2.05
		ANX	3.47	2.70	4.45	2.03	1.97	2.09	1.92	1.84	1.99
		DEP	4.91	3.97	6.06	2.27	2.20	2.33	2.10	2.02	2.18
		SCZ	6.46	2.90	14.38	3.16	2.89	3.45	2.73	2.41	3.08
		SUB	4.58	3.69	5.70	2.07	2.01	2.13	1.93	1.86	2.01
		PERS	6.68	4.58	9.74	2.32	2.19	2.45	2.23	2.07	2.39
		ANY	4.43	3.79	5.18	2.30	2.25	2.35	2.08	2.02	2.13
SCZ	BIP	SCZ	23.96	13.61	42.20	8.78	8.13	9.49	2.73	2.41	3.08
		NPSY	2.96	1.86	4.70	2.08	1.94	2.23	1.96	1.87	2.05
		ANX	3.19	2.19	4.65	1.92	.183	2.02	1.92	1.84	1.99
		DEP	3.15	2.18	4.56	1.82	1.73	1.91	2.10	2.02	2.18
		BIP	14.55	8.91	23.75	3.38	2.07	3.72	6.29	5.93	6.68
		SUB	3.90	2.77	5.48	1.90	1.81	1.99	1.93	1.86	2.01
		PERS	6.39	3.71	11	3.14	2.91	3.38	2.23	2.07	2.39
		ANY	4.04	3.16	5.16	2.24	2.17	2.32	2.08	2.02	2.13
BIP	SCZ	SCZ	29.23	18.16	47.03	3.16	2.89	3.45	8.42	7.51	9.44
		NPSY	3.30	2.17	5.02	2.41	2.33	2.49	2.08	1.93	2.25
		ANX	4.68	3.45	6.35	2.03	1.97	2.09	2.01	1.89	2.15
		DEP	4.16	3.04	5.70	2.27	2.20	2.33	1.96	1.84	2.08
		BIP	20.66	13.96	30.58	6.57	6.27	6.88	4.01	3.56	4.52
		SUB	5.26	4.01	6.90	2.07	2.01	2.13	2.04	1.92	2.16
		PERS	4.62	2.56	8.33	2.32	2.19	2.45	2.97	2.68	3.29
		ANY	5.10	4.13	6.28	2.30	2.25	2.35	2.26	2.16	2.35

SCZ = Schizophrenia, NPSY = Any Neuropsychiatric Disorder, ANX = Any Anxiety Disorder, DEP = Any Depressive Disorder, BIP = Bipolar Disorder, PSY = Any Psychotic Disorder, SUB = Any Substance Abuse, PERS = Any Personality Disorder, ANY = Any Major Psychiatric Disorder

Bipolar Disorder.

Among the offspring of offspring of homotypic BIP pairs, adjusted risk was also higher than all comparison populations (HR = 23.3, 95% CI = 17.6-30.7), including single-affected partnerships (range = 6.3-6.6, 95% CI = 5.9-6.9; see Table 2). Over follow-up, expected cumulative incidence of BIP in the offspring of homotypic parents reached 18.8% (95% CI = 13.6-24.3%), representing a significant increase relative to offspring of both single-affected and general population pairs (0.9-5.5%: see Figure 1).

Figure 1. Cumulative incidence of schizophrenia and bipolar disorder among the offspring of homotypic schizophrenia and bipolar parents, relative to comparison groups.

Curves represent the cumulative incidence of the titled disorder among offspring, by parental pair type (e.g., for Schizophrenia: “Gen Pop” = Offspring of general population pairs (neither parent diagnosed with schizophrenia). “Father/Mother Only” = Offspring of pairs in which only the father/mother has a schizophrenia diagnosis; “Both” = Offspring of homotypic pairs (both parents diagnosed with schizophrenia).

Heterotypic Parents

Among the offspring of heterotypic pairs (Table 2), hazards were significantly raised relative the general population for both SCZ (HR range = 23.9-29.2, 95% CI = 13.6-47.0) and BIP (HR range = 14.6-20.7, 95% CI = 8.9-30.6). The cumulative incidence of both disorders likewise exceeded rates in comparison offspring (11.1% SCZ, 17.4% BIP; see Figure 2).

Figure 2. Cumulative incidence of cross-disorder diagnoses among the offspring of homotypic pairs, relative to comparison populations.

Curves represent the cumulative incidence of the titled disorder among offspring, by parental pair type (e.g., for Dual Schizophrenia: “Gen Pop” = Offspring of general population pairs (neither parent diagnosed with schizophrenia). “Father/Mother Only” = Offspring of pairs in which only the father/mother has a schizophrenia diagnosis; “Both” = Offspring of homotypic pairs (both parents diagnosed with schizophrenia).

Cross-Disorder Risks

Table 1 summarizes HRs for a range of psychiatric diagnoses, among the offspring of homotypic, heterotypic, and single-affected SCZ/BIP pairs. Cumulative incidences for these broad diagnoses are presented in Supplemental Table 2, with a selection of curves presented in Figure 2. Disorder-specific summaries are offered below.

Cross-Disorder Risks from Homotypic Parents

Schizophrenia.

Hazard ratios for all examined diagnoses were raised in the offspring of homotypic SCZ pairs (range: 2.9-11.8) relative to single-affected and general population pairs, with confidence intervals overlapping only between offspring of dual and single-affected (fathers) for anxiety disorders (Table 2). Similarly, the cumulative incidence for each examined outcome was significantly higher among the offspring of homotypic pairs (range = 9.6-26.8%) than among comparison populations, excepting again a slight overlap with anxiety disorders (see Supplemental Table 2).

Bipolar Disorder.

Among the offspring of homotypic BIP pairs, hazard ratios for all alternate conditions were high relative the general population (HR range = 3.5-4.6; see Table 2), exceeding also those for single-affected partnerships across the majority of examined conditions (HR range = 1.8-3.0). An overlap in confidence intervals, between dual and single-affected pairs, was observed for SCZ (HR dual-affected = 6.5, 95% CI = 2.9-14.4; HR single-affected range = 2.7-3.2, 95% CI = 2.4-3.5). The cumulative incidence of all examined disorders was highest for the offspring of dual-bipolar pairs, significantly exceeding the incidence in the offspring of single-affected and general population pairs for all conditions with the exception of schizophrenia (see Figure 2; Supplemental Table 2).

Cross-Disorder Risks from Heterotypic Parents

Among the offspring of heterotypic pairs, the HRs for all examined conditions were raised, relative to the offspring of general population pairs (HR range = 3.1-6.4; see Table 2). Cumulative incidences were also significantly higher for these offspring, trending also- in the majority of cases - above the rates seen in the offspring of single-affected pairs (see Supplemental Table 2).

Any Disorder Risks

Table 2 presents the hazard ratios, and Figure 3 the cumulative incidences, for any offspring psychiatric disorder, by parental pair type. Incidences were high among the offspring of both homotypic SCZ and BIP pairs (43.6% and 48.5% respectively). Estimates from heterotypic samples were comparable, with 43.8-49.9% of offspring endorsing an examined disorder, with non-significant variation based on the arrangement of parental diagnoses (e.g., which parent has the SCZ diagnosis).

Figure 3. Cumulative incidence of any major disorder among the offspring of homotypic and heterotypic parents, relative to general population.

Curves represent the cumulative incidence of ‘any disorder’ (defined as any of the defined disorders/disorder groups) among offspring, by parental pair type (e.g., for ‘Mother SCZ, Father BIP’: “Gen Pop” = Offspring of general population pairs (mothers no SCZ diagnosis, fathers no BIP diagnosis), “Heterotypic” = Pairs in which mothers diagnosed with SCZ, fathers BIP).

Heritability Estimation

Heritability estimation required input of the population risks for schizophrenia, which we estimated as 27211/8343951 (number of affected offspring/total number of offspring) = 0.33% for SCZ and 45374/8343951= 0.54% for BIP.

Using the figures for cumulative incidence of SCZ or BIP diagnoses, based on hazard ratios in offspring of dual-affected parents, heritabilities were estimated as 0.62 (0.55-0.70) for SCZ and 0.52 (0.46-0.58) for BIP. These estimates were not significantly different from h² estimates derived from hazard ratios in offspring of single-affected pairs (see Table 3), demonstrating the high HR for offspring of dual-affected parents is consistent with the HR for offspring of single-affected parents under a polygenic model of liability to risk.

Table 3.

Estimates of heritabilities based on incidence up to age 42

Child	Mom	Dad	HR	95% CI	h2	95% CI	SEa h2
SCZ	SCZ		8.42	7.51-944	0.58	0.55-0.61	0.017
		SCZ	8.78	8.13-9.49	0.59	0.57-0.62	0.012
	SCZ	SCZ	55.33	39.90-76.74	0.62	0.55-0.70	0.039
BIP	BIP		6.29	5.93-6.68	0.55	0.53-0.57	0.01
		BIP	6.57	6.27-6.88	0.57	0.55-0.58	0.008
	BIP	BIP	23.28	17.64-30.72	0.52	0.46-0.58	0.031

^aSE = standard error

Discussion

This national cohort study followed the psychiatric outcomes of > 8 million offspring from 4.2 million unique parental pairs, in an effort to develop a targeted understanding of the relationship between serious dual-parental psychopathology and the risk for broad-spectrum psychiatric diagnoses in the subsequent generation. Offspring outcomes were established cumulatively over a period of 41 years, using independently validated ICD-10 diagnoses administered at the time of treatment and documented in the Swedish National Patient Register. Results from this study signal significant, broad risks for psychiatric diagnoses among the offspring of dual-affected pairs, with heightened risks retained in the offspring of mixed pairs and extending to disorders both concordant with, and discordant from, those of their parents.

Overall, risks were highest within disorder (e.g., risk in the offspring of dual-SCZ pairs was highest for SCZ), with hazard ratios ranging up to 55 for SCZ, and 23 for BIP (among the offspring of homotypic and heterotypic pairs). These HRs are interpreted relative the lower cumulative incidence, for SCZ versus BIP, in our general population samples (0.33% vs. 0.54%) - values which are comparable to reported cumulative incidences of SCZ (0.39%) and BIP (0.45%) in an earlier study of Swedish offspring.²⁷ Differences may partially reflect variations in clinical practice over time (the Lichtenstein sample was diagnosed 1974-2002), though prior work suggests psychiatric admissions have remained relatively stable in Sweden, despite sharp declines in duration of hospital stays during the overlapping periods examined.(26)

Our estimates regarding risk within-disorder align closely with those of prior register-based(11,14) and meta-analytic works.(27) Gottesman and colleagues,(11) for example, found increased incidence of both SCZ (27.3%; 95% CI: 18.3-36.2%) and BIP (24.9%; 95% CI: 14.0-35.8%) in the offspring of homotypic pairs - estimates which overlap with the tighter intervals derived in the current investigation (18.7% [13.13-24.75] and 18.8% [13.58-24.26], respectively). To our knowledge, the present study is the first work powered to examine within-disorder outcomes in the offspring of well-defined heterotypic pairs, with significant estimates in this study (SCZ: 8.47-11.14%; BIP: 14.64-17.38%) intersecting the trends signaled in earlier work.(11) Of note, SCZ and BIP were also examined as distinct outcomes in Danish register-based work by Dean at al.,(14) however, their definition of parental exposures (dual-affected parents shared ‘severe mental disorder,’ defined as either affective or non-affective psychosis) would have grouped the heterotypic and homotypic pairs isolated in the current work. That the SCZ incidences derived using these ‘dual-SMD’ samples (range: 7-9%) align more closely with our heterotypic figures can thus be seen as consistent.

In keeping with the liability overlap suggested by twin and family studies, (28,29) the offspring of dual-SCZ parents in our study showed heightened risks for BIP (HR = 11.84 [7.46-18.79]), with risks for SCZ likewise raised among the offspring of dual-BIP pairs (HR = 6.46 [4.58-9.74]). The associated incidences (10.4% BIP; 2.1% SCZ) align closely with those reported in Gottesman et al(11) (10.8% and 4.8%, respectively) - with both studies having wide-intervals around the SCZ estimates. Heightened risks for disorders discordant with those of their parents extended to all examined outcomes, with hazards for all but one (anxiety disorders) significantly raised relative even to single-affected pairs. Here, we show that all such risks are maintained, to a marked degree, in the offspring of heterotypic schizophrenia-bipolar pairs.

That hazards in children of two affected parents are markedly higher than those for children of one affected parent is a pattern expectable under a polygenic model of inheritance - now widely accepted for application to psychiatric disorders.(29) The model can be tested by estimation of population heritability of disease, looking for consistent estimates from calculations based on different hazard or risk ratios. This simple approach ignores contributions of non-genetic factors, such as shared family environment on the estimates of HR - the latter of which has been shown to have a small contribution (SCZ: 4.5% [CI 4.4%-7.4%]; BIP: 3.4% [95% CI 2.3%-6.2%]) in prior analyses of Swedish national data.(30) In that previous analysis, heritabilities were estimated as 64.3% (95% CI: 61.7%-67.5%) for SCZ and 58.6% (95% CI: 56.4%-61.8%) for BIP from a linear mixed model applied to individual data. In a follow-up analysis, heritabilities were estimated using the methodology presented here from the simple summary statistics of risk and increased risk in relatives.(23) The estimates were in good agreement with those from the full statistical analysis, 64% (95% CI: 61%-67%) for SCZ and 56% (95% CI: 54%-58%) for BIP, validating the methodology used in the present work.

Our estimates, which factor in the increased risk of disease in the children of dual and single-affected parents, are comparable to these prior values, for both SCZ and BIP (Table 1). We thus conclude that the increased risk of psychiatric disorders in children of two affected parents follow expected values based on a polygenic model, which implies additive genetic factors on the underlying scale of risk, but non-additivity on the scale of disease. These results are consistent with similar analyses applied to summary data from Danish national registers.(23)

Non-Specificity of Diagnostic Risk from Dual-Affected Pairs

Our results accord with an emerging, polygenic picture of psychiatric liability, and are consistent with the hypothesis that a clustering of genetic risk (e.g., increased variance in disorder liability as a consequence of genetic loading from two affected parents) is expectable in the offspring of two affected parents and will drive increases in incidence.(6,31,32) Recent molecular work implicates common, often overlapping variants,(33–36) as risk factors for a range of major psychiatric conditions and there exists a broad body of literature - from familial, genetic, and meta-analytic studies(14,27,37,38) - attesting to the non-specificity of psychiatric risk among individuals with an affected first-degree relative. Indeed, family history is regarded as among the strongest predictors for disorders including those examined here, with emerging work indicating a potentially stronger role for common variants in cases where a family history is present.(33,35,36,39) Work in SCZ is forefront in this area, with one novel investigation signaling that case individuals with a family history of SCZ may, relative to non-family history cases, be enriched for polygenic risk linked to this and other conditions (e.g., BIP, MDD).(35)

Our findings would support such a transdiagnostic risk dynamic. While hazards varied, to an expectable degree(14), with the molecular relatedness of conditions (e.g., risks were higher for neuropsychiatric disorders than anxiety disorders), the overall portrait is one of broadly heighted liability in the offspring of dual-affected pairs. Indeed, the cumulative incidence for ‘any’ major psychiatric disorder was near to half for all samples - with the highest incidence observed in the offspring of heterotypic pairs in which mothers carried the bipolar diagnosis (49.87%; 41.91-59.33). At the population level, such findings suggest that, if nonrandom mating is to have an impact on the maintenance of psychiatric disorders, the pathways are unlikely to be confined to (or even primarily defined by) transmission within-disorder, but rather will integrate these contributions to cross-disorder liability - reflected, crucially, in not only homotypic but also the more numerous heterotypic pairings.

Of note, genetic modeling work has suggested that nonrandom mating could, if due to assortative mating, increase the prevalence and, potentially, heritability of rare, highly-heritable conditions (e.g., schizophrenia). (40) However, the mechanism of effect for established patterns of nonrandom mating (e.g., assortment vs. convergence) remains unclear. While high, risks are furthermore not decisive and mechanisms of balance (e.g., reduced fecundity in major psychiatric populations) have not been integrated into current transmission models. Even this is not straightforward, however, with sharply reduced fecundity suggested in some major disorders (e.g., schizophrenia), but less-so in others (e.g., major depression).(41) Work considering the degree to which risks shown in the present study are maintained in more common disorder groups (e.g., parents with major depression) will, therefore, be integral to parsing the relevance of established nonrandom mating to cross-generational disorder maintenance. With cross-disorder mating established as common, and the current work highlighting the maintenance of risk in mixed-pairs, consideration of these dynamics will also be essential in such groups.

Strengths and Limitations

The strengths of this study rest in the large-scale, population-based register data utilized, which permitted estimation of target outcomes, including lifetime incidences and heritability, from a single data source. Limitations are those inherent to register data, including the detection bias which may result from dependence on treatment-seeking, hospital-based, largely inpatient samples - a greater concern for select outcomes in the current investigation (e.g., depression, anxiety), as the primary disorders of interest are those more likely to require inpatient management. This self-selection, along with our decision to use naturalistic samples with minimal restrictions on psychiatric comorbidity, means that our incident estimates are appropriately viewed as a minimum for all sample. It is also possible that both the partners and offspring of individuals with major psychiatric disorders are more likely to be in contact with medical services, and thus be documented in the registers – potentially inflating the relative risks seen in the present investigation.

We have made efforts to be strict in our definitions of the primary disorders of interest, requiring at least two diagnoses to minimize the impacts of misdiagnosis on results (estimates derived using single-diagnoses were similar [Supplemental Table 3]). This conservative approach has resulted in smaller numbers for select groups, with care required when interpreting (or generalizing from) these rare outcomes. The validity of the heritability estimates is directly dependent on the quality of the register diagnoses, and therefore shares their limitations. Our estimates of heritability ignored contributions of shared common environment; however, we note these had previously been estimated to be small. Moreover, we note that the increased risks of disorders in children stratified by whether the mother or father is affected are similar, which might not be expected if common family environmental effects are important. This, interpretation, however, carries with it an assumption of traditional gender roles child-rearing and shaping of the home environment, which may differ uniquely in Sweden relative to other nations.

Summary

Risks for a broad spectrum of psychiatric diagnoses are significantly raised in the offspring of dual-affected parents, in line with expectations from a polygenic model of liability to disease risk. Further work should extend this examination to more common psychiatric groups, to untangle the relevance of nonrandom mating to population maintenance of these disorders.

KEY RESOURCES TABLE

Resource Type	Specific Reagent or Resource	Source or Reference	Identifiers	Additional Information
Add additional rows as needed for each resource type	Include species and sex when applicable.	Include name of manufacturer, company, repository, individual, or research lab. Include PMID or DOI for references; use “this paper” if new.	Include catalog numbers, stock numbers, database IDs or accession numbers, and/or RRIDs. RRIDs are highly encouraged; search for RRIDs at https://scicrunch.org/resources.	Include any additional information or notes if necessary.
Antibody
Bacterial or Viral Strain
Biological Sample
Cell Line
Chemical Compound or Drug
Commercial Assay Or Kit
Deposited Data; Public Database		Swedish National Population Registers: doi.org/10.1186/1471-2458-11-450
Genetic Reagent
Organism/Strain
Peptide, Recombinant Protein
Recombinant DNA
Sequence-Based Reagent
Software; Algorithm
Transfected Construct
Other