Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2024 Jul 31.
Published in final edited form as: Autism Res. 2021 Oct 19;15(1):171–182. doi: 10.1002/aur.2629

Evaluating the interrelations between the autism polygenic score and psychiatric family history in risk for autism

Diana Schendel 1,2,3,4, Thomas Munk Laursen 1,3, Clara Albiñana 1,3, Bjarni Vilhjalmsson 1,3, Christine Ladd-Acosta 5, M Danielle Fallin 6,7, Kelly Benke 7, Brian Lee 4,8,9, Jakob Grove 1,10,11,12,13, Amy Kalkbrenner 14, Linda Ejlskov 1,3, David Hougaard 15, Jonas Bybjerg-Grauholm 15, Marie Bækvad-Hansen 15, Anders D Børglum 1,10,11,13, Thomas Werge 1,16,17, Merete Nordentoft 1,18, Preben Bo Mortensen 1,3,19, Esben Agerbo 1,3,19
PMCID: PMC11289736  NIHMSID: NIHMS2006100  PMID: 34664785

Abstract

Psychiatric family history or a high autism polygenic risk score (PRS) have been separately linked to autism spectrum disorder (ASD) risk. The study aimed to simultaneously consider psychiatric family history and individual autism genetic liability (PRS) in autism risk. We performed a case-control study of all Denmark singleton births, May 1981-December 2005, in Denmark at their first birthday and a known mother. Cases were diagnosed with ASD before 2013 and controls comprised a random sample of 30,000 births without ASD, excluding persons with non-Denmark-born parents, missing ASD PRS, non-European ancestry. Adjusted odds ratios (aOR) were estimated for ASD by PRS decile and by psychiatric history in parents or full siblings (8 mutually-exclusive categories) using logistic regression. Adjusted ASD PRS z-score least-squares means were estimated by psychiatric family history category. ASD risk (11,339 ASD cases; 20,175 controls) from ASD PRS was not substantially altered after accounting for psychiatric family history (e.g., ASD PRS 10th decile aOR: 2.35 (95% CI 2.11-2.63) before vs 2.11 (95% CI 1.91-2.40) after adjustment) nor from psychiatric family history after accounting for ASD PRS (e.g., ASD family history aOR: 6.73 (95% CI 5.89-7.68) before vs 6.32 (95% CI 5.53-7.22) after adjustment). ASD risk from ASD PRS varied slightly by psychiatric family history. While ASD risk from psychiatric family history was not accounted for by ASD PRS and vice versa, risk overlap between the two factors will likely increase as measures of genetic risk improve. The two factors are best viewed as complementary measures of family-based autism risk.

Keywords: autism spectrum disorder, family history, genetic risk factors, polygenic risk score, case-control studies

Lay Summary

Autism risk from a history of mental disorders in the immediate family was not explained by a current measure of genetic risk (autism polygenic risk score) and vice versa. That is, genetic risk did not appear to overlap family history risk. As genetic measures for autism improve then the overlap in autism risk from family history versus genetic factors will likely increase, but further study may be needed to fully outline the components of risk and ‘cross talk’ between these key related factors. Meanwhile, the two factors may be best viewed as complementary measures of autism family-based risk.

INTRODUCTION

Autism spectrum disorder (ASD) is a neurodevelopmental disorder affecting 1-3% of persons and marked by deficits in social communication and restricted, repetitive behaviors (American Psychiatric Association, 2013; Baio et al., 2014; Schendel & Thorsteinsson (2018)). Early observations of greater than expected familial aggregation of ASD, including high twin concordance (Bailey et al., 1995), suggested genetic components in ASD etiology. Currently, a family history of ASD is among the strongest of autism risk factors and family-based studies have further revealed elevated ASD risk from a family history of psychiatric disorders overall, as well as from specific psychiatric disorders besides ASD (Hansen et al., 2019; Xie et al., 2019). These findings support the role of heritable psychiatric factors in ASD etiology, although both environmental and genetic factors likely contribute to risk (Bai et al., 2019).

With molecular genetic approaches such as genome-wide association studies (GWAS) we can now estimate individual genetic liability for ASD using a polygenic risk score (PRS) which aggregates the risk effects of common genetic variants associated with ASD detected in large GWAS studies (Grove et al., 2019). Further, genetic correlation analyses using common variants genome-wide indicate genetic overlap between ASD and other psychiatric disorders (Cross-disorder Group of the Psychiatric Genetics Consortium, 2013a; Cross-disorder Group of the Psychiatric Genetics Consortium, 2013b; Robinson et al., 2016), such as ADHD, depression and schizophrenia, as well as cognitive phenotypes (Grove et al., 2019).

Despite clear evidence linking psychiatric family history or individual genetic liability to ASD and genetic overlap between ASD and other psychiatric conditions, integration of psychiatric family history and individual genetic factors in ASD studies has not been carried out. Generally speaking, what we might expect is that genetic factors account for a proportion of the risk associated with psychiatric family history, i.e., genetic risk overlaps family risk. Our study aim was to perform population-based integration of psychiatric family history and individual ASD genetic liability in autism risk estimation and investigate the interrelations of these key elements in autism risk. We addressed the research questions: 1) To what extent does the estimate of ASD risk associated with an elevated ASD PRS reflect the effect of psychiatric family history on ASD risk (i.e., is there confounding of risk from PRS by psychiatric family history)?; 2) To what extent does consideration of ASD PRS affect the estimate of ASD risk associated with different types of psychiatric family history (i.e. is there evidence for mediation of risk from psychiatric family history by individual genetic liability)?; and 3) Does the risk from ASD PRS vary across different types of psychiatric family history (i.e., is there interaction of risk between individual genetic liability and psychiatric family history)?

METHODS

The study sample was derived from iPSYCH (Lundbeck Foundation Initiative for Integrative Psychiatric Research), a nation-wide, Danish case-cohort study of select psychiatric disorders (Pedersen et al., 2018). The iPSYCH study population was identified in the Danish Civil Registration Service and comprised: all singleton births in Denmark from May 1981 through December 2005; living in Denmark on the one-year birthday; and known mother (n=1 472 762). iPSYCH cases comprised persons in the study population who received specific ICD-10 psychiatric diagnoses including ASD (F84.0, F84.1, F84.5, F84.8, F84.9) at any time through 2012 (7-year minimum), yielding a total of 16 146 ASD cases. The iPSYCH subcohort sample (controls) comprised a random sample of 30 000 persons from the study population (2.04%).

For the current study, iPSYCH ASD cases and controls were restricted to persons of parents born in Denmark (84% of ASD cases; 84% of controls) and, of these, had an ASD PRS (91% of ASD cases; 88% of controls). An additional 272 ASD cases were excluded from the control group (the subcohort random selection process could, by chance, select ASD cases). Finally, 7.9% of controls and 8.2% of cases with non-European ancestry were excluded as genetic outliers (Martin et al., 2017) (see principal component analysis (PCA) methods below) yielding 20 175 controls and 11 339 ASD cases included in the study. The sample restriction to persons of parents born in Denmark enhanced completeness of registry-based family linkages and family morbidity histories which would be highest in multi-generation, Denmark-born families. ASD cases and controls were linked to parents and full siblings via the Civil Registration Service using each person’s unique identifier and parental identity (Pedersen et al., 2011).

Psychiatric diagnoses of all persons were obtained via linkage with the Psychiatric Central Research (Mors et al., 2011) and National Patient Registers (Lynge et al., 2011) to which diagnoses are reported from all inpatient admissions to hospitals and wards since 1969 and both in- and outpatient admissions since 1995. The International Classification of Diseases Eighth Revision (ICD-8) was used for reporting from 1969 through 1993 and replaced by the Tenth Revision (ICD-10) in 1994. Information on child sex, dates of birth, death and immigration, birth weight, gestational age and parental ages at birth was obtained from the Medical Birth Register (Knudsen & Olsen, 1998) and Civil Registration Service.

Psychiatric family history was based on reported diagnoses for a parent (reported before birth of the index person) or full sibling (reported before end of study follow-up on 1 April 2017) and assigned into 1 of 8 mutually exclusive, hierarchical categories (eTable 1). Selection of the specific mental conditions for the psychiatric family history hierarchy was based on the conditions’ relatively high ASD family history risks, genetic correlations and co-occurrence with ASD and, from a practical analytic standpoint, to limit the number of levels in the hierarchy. The top disorder in the hierarchy was ASD family history by virtue of its top rank in autism family history risk; in this study ASD family history largely reflected frequent ASD diagnosis in siblings and not parents. The order of the hierarchy below ASD was a product of balancing diagnoses more likely to be made in parents versus siblings across the selected conditions which have rather similar parent/sibling family history risks for ASD and genetic correlations with ASD. Thus, the hierarchy was: 1) ASD (predominant sibling diagnosis), 2) Schizophrenia (predominant parent diagnosis), 3) Other disorders of psychological development or intellectual disability (ID) (predominant sibling diagnoses), 4) ADHD (predominant sibling diagnosis), 5) Affective disorder (predominant parent diagnosis), 6) Adult personality disorder (predominant parent diagnosis), 7) Other psychiatric disorder, and 8) No psychiatric family history. Within a category there could be other psychiatric diagnoses in the family besides the key diagnosis, e.g., in the ASD category family members could have other psychiatric diagnoses besides ASD. We observed, however, that distributions of psychiatric diagnoses by family history category were very similar between ASD cases and controls (eTables 2–3).

To control for genetic population structure in the GWAS data, we performed a principal components analysis (PCA) on the iPSYCH case-cohort GWAS sample using the autoSVD algorithm available in the R package bigsnpr (Privé et a., 2018). We computed 10 principal components (PCs), and used the first 2 PCs to account for population structure throughout genetic analysis, as the population was very homogeneous. We then computed the Gnanadesikan-Kettenring pairwise distance of the first 10 PCs (Maronna & Zamar, 2002) and classified as European ancestry the individuals who had a log-distance from the center < 3. We also filtered out individuals in pairs with a pi coefficient greater than 0.2, corresponding to second-degree relatedness or more. The non-european ancestry and relatedness filters removed 9 490 individuals (3 425 cases and 6 065 controls) from the ASD sample (N = 40 715) used to generate the PRS.

The ASD PRS was derived using BOLT-LMM20 for risk prediction. Although primarily intended for GWAS, BOLT-LMM can and has been used in multiple studies for deriving polygenic risk scores; to our knowledge the methods and corresponding results have been described in at least 6 papers by other investigators (Loh et al., 2015; Zhang et al., 2020; Weissbrod et al., 2021; Yang et al., 2011; Speed & Balding, 2014; Loh et al., 2018; Loh et al., 2020; Albiñana et al., 2021), including the original BOLT-LMM papers by Loh et al. (Loh et al., 2015; Loh et al., 2018) and a third more recently (Loh et al., 2020). In addition, we recently described and evaluated the BOLT-LMM PRS method that we used (Albiñana et al., 2021). In brief, one can obtain the prediction weights using a BOLT-LMM command-line option (--predBetasFile; not reflected in the user manual but can be found on the software’s --helpFull flag), and use them as variant weights for genomic prediction (polygenic scores). These prediction weights are derived using a Bayesian regression and are conceptually different from the GWAS effect estimates as they account for linkage disequilibrium, thus no clumping or thresholding steps are necessary to calculate the PRS. The difference is explained in the Supplementary Text of Loh et al., 2015 in section 1.1.3. “Distinction between estimated effect sizes in association tests vs. prediction”.

The sample of 31,225 (11,257 cases and 19,968 controls) unrelated individuals of European ancestry was used for training, The BOLT-LMM model was fit for 544,955 SNPs, which correspond to the “hard-called” SNPs present in the iPSYCH2012 genotyping microarray (Pedersen et al., 2018), using a minor allele frequency filter of 1% and <10% missing values. Because the goal of using BOLT-LMM was to obtain variant weights, it was not necessary to include more SNPs in the model as this would not have translated into an increased prediction accuracy of the PRS (Loh et al., 2018). BOLT-LMM does an internal cross-validation step for parameter selection similar to selecting a p-value threshold. For ASD, the infinitesimal model was selected. Genotype waves, 2 PCs, sex, and age were added as covariates for BOLT-LMM. To avoid overfitting, we ran BOLT-LMM using a 10-fold cross-validation scheme on the European ancestry unrelated sample, fitting iteratively the model for 9/10th of the data and projecting the weights into the remaining 1/10th test set. The final PRS was defined as the weighted sum of the prediction betas with the test set genotypes. Finally, we averaged over the prediction models to calculate PRS for all individuals that had been dropped for PRS derivation due to ancestry, relatedness or due to being a case for a different disorder. A standardized PRS (z-score) was estimated for each individual based on the PRS mean and standard deviation of the cohort sample.

We examined PRS both by decile and as a continuous measure (PRS z-score) in multiple analyses. We calculated adjusted odds ratios (adjOR) and 95% confidence intervals (95% CI) for ASD by decile of PRS (not considering psychiatric family history) and by psychiatric family history category (not considering ASD PRS) using logistic regression adjusting for the first 5 PCs, year of birth, sex, parental ages at birth, and having a sibling (Yes/No) (full model). Separately, ORs by PRS decile were additionally adjusted for psychiatric family history and ORs by psychiatric family history were additionally adjusted for PRS. To test for interaction between PRS and psychiatric family history we estimated separate adjORs for ASD per 1 standard deviation increase in PRS (continuous score) for each psychiatric history group and used a likelihood ratio test (chi-square with 1 degree of freedom) of the difference in the adjOR of PRS for ASD in each family history group compared to ‘no psychiatric family history’. We calculated the R2 statistic on the liability scale and the Nagelkerke R2 to estimate the proportion of variance in ASD explained by the ASD PRS and psychiatric family history.

We estimated ASD PRS z-score adjusted or least-squares means (LS-Mean)(Searle et al., 1980) of cases and controls by psychiatric family history category adjusting for the first 5 PCs, year of birth, sex, parental ages at birth, and having a sibling (Yes/No). We tested for differences in ASD PRS z-score LS-Means by psychiatric family history category via pair-wise comparisons (with Bonferroni correction) across all possible psychiatric family history combinations in ASD cases and controls.

In secondary analyses, we assessed pattern specificity of mean ASD PRS by psychiatric family history in two ways. First, we examined mean body mass index (BMI) PRS (negative control) to assess variation in polygenic scores of a non-psychiatric phenotype by psychiatric family history in ASD cases and controls. We used a BMI PRS derived by Grove et al., 2019 and used the same analytic methods and covariates as for ASD PRS to compute BMI z-scores and LS-Means by psychiatric family history category for ASD cases an controls. Second, we computed ASD PRS z-score LS-means by diabetes family history (negative control) in ASD cases and controls to assess variation in ASD PRS with a non-psychiatric family history. We further assessed variation in case and control ASD PRS distributions by psychiatric family history by 1) computing adjORs for ASD by PRS decile stratified by psychiatric family history category and 2) calculating the ratio of % ASD cases-to-% controls by PRS decile in each psychiatric family history category. We also assessed variation in mean ASD PRS z-score LS-Means by family history category among ASD cases by sex and by co-occurring ID status in view of a potentially lower rate of psychiatric family history, thereby potentially lower ASD genetic liability, in ASD cases with versus without co-occurring ID which may also vary by sex (Robinson et al., 2014).

Study approval

iPSYCH was approved by the Danish Scientific Ethics Committee, the Danish Health DataAuthority, the Danish Data Protection Agency, Statistics Denmark and the Danish Neonatal Screening Biobank Steering Committee. The Danish Scientific Ethics Committee, in accordance with Danish legislation, has, for this study, waived the need for informed consent in biomedical research based on existing biobanks.

RESULTS

ASD cases were more likely than controls to be male, born after 1990, have parents that were 30 years of age or older at birth of the child, not have any full siblings, have co-occurring intellectual disability, an ASD PRS above the 5th decile and to have a parent or full sibling diagnosed with any psychiatric condition (36.2% versus 19.8%), as well as each specific type of psychiatric family history, especially ASD (Table 1).

Table 1.

Characteristics of ASD cases and controls

ASD (N=11,339) Control (N=20,175)
N % N %
Sex
Male 8 930 78·7 10 205 50·6
Female 2 409 21·3 9 970 49·4
Birth year
1981-84 444 3·9 2 508 12·4
1985-87 653 5·8 2 491 12·4
1988-90 1 097 9·7 2 627 13·0
1991–93 1 687 14·9 2 758 13·7
1994–96 2 330 20·6 2 796 13·9
1997–99 2 208 19·5 2 719 13·5
2000–02 1 810 16·0 2 522 12·5
2003–05 1 110 9·8 1 754 8·7
Maternal age
< 30 years 6 611 58·3 12 201 60·5
30–39 years 4 543 40·1 7 711 38·2
40+ years 185 1·6 263 1·3
Paternal age
< 30 years 4 523 39·9 8 584 42·6
30–39 years 5 752 50·7 10 251 50·8
40+ years 1 064 9·4 1 340 6·6
Number of full siblings
0 1 255 11·1 1 513 7·5
1+ 10 084 88·9 18 662 92·5
Co-occurring intellectual disability
Yes 1 845 16·3 199 1·0
No 9 494 83·7 19 976 99·0
Psychiatric family History
ASD 1 021 9·0 346 1·7
Schizophrenia 369 3·3 418 2.1
Developmental/intellectual disability 643 5·7 572 2.8
ADHD 389 3·4 366 1·8
Affective disorder 489 4·3 680 3·4
Adult personality disorder 287 2·5 376 1·9
Any other psychiatric condition 911 8·0 1 240 6·2
No psychiatric history 7 230 63·8 16 177 80·2
ASD PRS decile
1 867 7·7 2 284 11·3
2 921 8·1 2 230 11·1
3 971 8·6 2 181 10·8
4 1 010 8·9 2 141 10·6
5 1 118 9·9 2 034 10·1
6 1 178 10·4 1 973 9·8
7 1 187 10·5 1 965 9·7
8 1 252 11·0 1 899 9·4
9 1 322 11·7 1 830 9·1
10 1 513 13·3 1 638 8·5
Open in a new tab

As shown in Table 2, the risk for ASD increased with increasing ASD PRS decile (no adjustment for psychiatric family history), achieving a 2.4-fold increased risk (adj OR 2.35, 95% CI 2.11-2.63) with an ASD PRS in the 10th decile. The risk for ASD was increased by 6.7-fold (95% CI 5.89-7.68) with an ASD family history and 1.7- to 2.7-fold across all other family history categories (no adjustment for ASD PRS). The risks associated with the ASD PRS were only slightly decreased after adjustment for psychiatric family history (e.g., change of risk from PRS in the 10th decile from adj OR 2.35 (95% CI 2.11-2.63) to 2.14 (95% CI 1.91-24.0)) which indicates that the PRS effect was scantly confounded by family history. Further, the risks from psychiatric family history were only slightly attenuated after adjusting for the ASD PRS (e.g., change of risk from an ASD family history from adj OR 6.7 (95% CI 5.89-7.68) to 6.32 (95% CI 5.53-7.22)) indicating that the effects of family history were only marginally mediated through the polygenic liability for autism. As shown in Table 3, the separate adjORs associated with ASD PRS (per one standard deviation increase in PRS) did not vary markedly across the different family history categories and none of the risk estimates per category differed significantly from the risk in persons with no psychiatric family history, indicating there was no interaction between the two factors. The amount of variance in ASD explained by the ASD PRS and psychiatric family history (R2 on the liability scale and Nagelkerke R2) increased somewhat above the amount explained by covariates only (Table 2).

Table 2.

Risk for ASD associated with ASD PRS and psychiatric family history

Model: covariatesa + ASD PRS covariates + ASD PRS + psychiatric family history
ASD PRS decilea Odds ratio (95% confidence limits) Odds ratio (95% confidence limits)
1 referent referent
2 1·11 (0·99-1·25) 1·11 (0·98-1·25)
3 1·18 (1·05-1·33) 1·15 (1·02-1·29)
4 1·25 (1·12-1·40) 1·19 (1·01-1·34)
5 1·45 (1·30-1·63) 1·38 (1·23-1·55)
6 1·56 (1·39-1·74) 1·49 (1·33-1·67)
7 1·60 (1·42-1·79) 1·53 (1·36-1·72)
8 1·74 (1·55-1·94) 1·65 (1·47-1·85)
9 1·96 (1·75-2·19) 1·82 (1·62-2·04)
10 2·35 (2·11-2·63) 2·14 (1·91-2·40)
Model: covariates + psychiatric family history covariates + ASD PRS + psychiatric family history
Psychiatric family historyb Odds ratio (95% confidence limits) Odds ratio (95% confidence limits)
No psychiatric history referent referent
ASD 6·73 (5·89-7·68) 6·32 (5·53-7·22)
Schizophrenia 2·14 (1·84-2·50) 2·10 (1·80-2·45)
Developmental/intellectual disability 2·68 (2·37-3·04) 2·65 (2·34-3·01)
ADHD 2·36 (2·02-2·76) 2·32 (1·98-2·71)
Affective disorder 1·97 (1·73-2·25) 1·96 (1·72-2·23)
Adult personality disorder 1·95 (1·65-2·31) 1·94 (1·64-2·30)
Other psychiatric disorder 1·73 (1·57-1·90) 1·70 (1·54-1·88)
Model: covariates only covariates + ASD PRS covariates + psychiatric family history covariates + ASD PRS + psychiatric family history
R2 on liablity scale 0·11 0·12 0·14 0·15
Nagelkerke R2 0·17 0·18 0·22 0·23
Open in a new tab
a

Covariates: sex, first 5 genetic ancestry principal components (PRS), birth year, parental ages at birth, have a sibling (Y/N)

Table 3.

Risk of ASD from ASD PRS (continuous measure) within each category of psychiatric family history

PRS Odds ratioa (95% confidence limit) p-valueb
Psychiatric family history
  ASD 1.14 (1.01-1.30) 0.11
  Schizophrenia 1.15 (0.99-1.34) 0.19
  Developmental/intellectual disability 1.15 (1.02-1.30) 0.12
  ADHD 1.29 (1.11-1.51) 0.87
  Affective disorder 1.34 (1.18-1.51) 0.47
  Adult personality disorder 1.28 (1.08-1.52) 0.94
  Other psychiatric disorder 1.24 (1.13-1.36) 0.59
  No psychiatric family history 1.27 (1.24-1.31) not applicable
Open in a new tab
a

PRS Odds ratio adjusted for main effect of psychiatric family history, sex, first 5 genetic ancestry principal components (PRS), birth year, parental ages at birth, have a sibling (Y/N)

b

H0: PRS Odds ratio at each psychiatric family history category = PRS Odds ratio with no psychiatric family history vs H1: PRS Odds ratio at each psychiatric family history category ≠ PRS Odds ratio with no psychiatric family history

Overall, the ASD PRS z-score LS-mean was higher in cases than controls (adj z-score LS-mean for controls 0.09 (95% CI 0.05-0.13); for ASD cases 0.32 (95% CI 0.28-0.37); difference in means −0.24 (95% CI −0.26 - −0.21)). Figure 1 (Panel A) and eTable 4 present the ASD PRS z-score LS-means per family history category. After Bonferroni correction, among controls the only pair-wise differences at p<0.05 were higher ASD PRS z-score LS-means with an ASD family history versus all other family history groups except with a schizophrenia family history. Similarly, among ASD cases the only pair-wise differences at p<0.05 after Bonferroni correction were higher ASD PRS z-score LS-means in cases with an ASD family history versus other family history groups except the ADHD or schizophrenia family history groups. Comparing ASD cases versus controls there was also substantial overlap in ASD PRS z-score LS-means depending on family history. As shown in the Figure 1, Panel B, there were no differences in ASD PRS z-score LS-means at p<0.05 after Bonferroni correction between controls with an ASD family history versus any of the ASD case groups regardless of family history, including cases with an ASD family history, and few differences between controls with schizophrenia, ADHD or learning disorders family histories and ASD cases of different family histories (except ASD history).

Figure 1.

Figure 1.

Open in a new tab

Least-squares means of ASD PRS z-scores by psychiatric family history category in ASD cases and controls

When we substituted the ASD PRS with BMI PRS, the BMI PRS z-score LS-means (eFigure 1) were similar across all ASD case and control psychiatric family history groups. In contrast, the ASD PRS z-score LS-means in ASD cases with or without a diabetes family history were virtually the same and markedly higher than the corresponding control means (eFigure 2).

As shown in eFigure 3, an ASD risk increase with higher ASD PRS decile was generally observed across persons in each of the psychiatric family history categories although less markedly in the ASD category. With an ASD family history (compared to other family history categories) there was less increased risk with increasing ASD PRS decile and less variation in ASD risk by decile, i.e., less difference in the ASD PRS distribution between ASD cases with an ASD family history and controls with an ASD family history. There was a fairly consistent ratio of % ASD cases-to-% controls across PRS deciles in each family history category, except the ASD family history where the ratio declined by decile (eFigure 4). The latter pattern was due to a relatively more marked increase in the % controls than % cases by decile.

Autistic females or autistic persons with ID tended to have lower ASD PRS means than autistic males or autistic persons without ID across most psychiatric family history groups, although the sample sizes were relatively small and the corresponding 95% confidence intervals for each mean were wide (eFigure 5).

DISCUSSION

While we might expect that genetic risk may account to some degree for the association between psychiatric family history and ASD, after systematic investigation we observed little statistical evidence of overlap or interdependence in autism risk between the two factors, using current PRS as the measure of genetic risk. There was scant evidence for confounding (genetic risk as measured by the autism PRS was not substantially altered after accounting for psychiatric family risk), interaction (ASD risk associated with the autism PRS risk did not vary by type of psychiatric history in the family) or mediation (the risk for ASD by psychiatric family history category was not substantially altered after accounting for genetic risk as measured by the PRS). In the absence of interaction (Vanderweele et al., 2010) the latter suggests that the ASD PRS did not act as an intermediate factor between psychiatric family history risk and ASD (i.e intermediate in the sense of a measure of the transmission of common variants associated with mental disorders (including ASD risk alleles) from parents to offspring and on the pathway between parental genetic profile and final ASD status in offspring). The lack of change in ASD risk from either factor after statistical adjustment for the other was reflected in the similar patterns of ASD case and control PRS distributions across different psychiatric family histories assessed in a variety of primary and secondary analyses. The highest mean ASD PRS level was in ASD cases with an ASD family history, but in fact there were was no difference in mean ASD PRS levels between ASD cases with an ASD family history and controls with an ASD family history and few differences in mean ASD PRS levels between controls with ASD, schizophrenia, ADHD or learning disorder family histories and ASD cases across all family history groups (except ASD family history). The magnitude of risk for ASD associated with each type of family history was as high or somewhat higher than the risk associated with an ASD PRS in the highest decile. Considering all results together, the findings of little overlap may reflect both the limitations of our current measures of the genetic (i.e., common variant polygenic risk) and family history risk architecture of ASD and our limited understanding of the very specific elements which gives rise to autism risk from these two sources.

Despite established links between either psychiatric family history or individual genetic liability to autism, after systematic search we found no studies to date that simultaneously considered both psychiatric family history and individual genetic factors in autism risk estimation or investigated the specific nature of overlap or interdependence of these key elements in autism. In the current results, ASD PRS and psychiatric family history appeared, statistically, to represent two sources of autism risk. The current ASD PRS accounted for little of ASD risk from an ASD family history despite the significant gradation of risk for ASD recurrence in a family depending on the degree of genetic relatedness between affected family members (Hansen et al., 2019). In this study we also observed elevated ASD risks across all psychiatric family history groups suggesting that there may be components of family risk for ASD arising from each type of psychiatric history, not just ASD family history. The ASD PRS, however, accounted for little ASD risk associated with other types of psychiatric family history despite the genetic overlap of ASD with other psychiatric conditions such as schizophrenia, ADHD, cognitive ability and major depression (Grove et al., 2019). Thus, on the one hand, psychiatric family history may comprise components of ASD risk apart from enhanced individual common variant genetic liability alone, at least as measured by the current ASD PRS. Theoretically, this is not unexpected. For example, for Type 2 diabetes the variance explained by family history in a liability threshold model was partitioned into both genetic and environmental components, including genetic and non-genetic components of shared environmental effects (Cornelis et al., 2015).

On the other hand, it is unlikely to be completely true that ASD PRS and psychiatric family history are fully separate contributors to ASD risk. The low level of risk overlap we observed between the two factors is at least partly a reflection of the small amount of variance in ASD currently captured by each measure (reflected in the low values shown in Table 2 from the models for R2 on the liability scale and Nagelkerke R2). That is, despite their consistent, significant associations with ASD both measures – psychiatric family history and ASD PRS - are still rather ‘blunt instruments’ as a means of detecting potential ASD risk in individuals. It is likely that further development of ASD PRS (comprised of low-risk, common variants) and other genetic metrics (e.g., high-risk, rare variant burden) will capture more individual genetic liability for ASD and thereby permit more refined genetic risk mapping against psychiatric family history risk. Meanwhile, even in families with no psychiatric history in first degree relatives, an ASD PRS in the top decile was associated with a 2.3-fold increased risk for ASD. Thus, ASD PRS is a measure of ASD risk even in the absence of a psychiatric family history while ASD risk from a psychiatric family history only affects the minority of persons who have such a history (20% of persons in this study).

In previous studies both schizophrenia PRS and schizophrenia or psychiatric disorder family history increased schizophrenia risk and together the two factors explained more of the variance in schizophrenia (7.8% (Agerbo et al., 2015), 8.9% (Lu et al., 2018) on the liability scale) than either factor alone. In a study of common cancers, Do et al. (2012) observed that the low correlation in risk between family history and PRS made it appear that the risks from these two factors were relatively independent. Do et al. (2012) argued that rather than one measure replace the other, PRS and family history are best viewed as complementary tools for understanding an individual’s predisposition to disease. In the same vein, in a more recent study of individual risk prediction (for a variety of complex diseases) conditional on both PRS and family history, the combination greatly improved the accuracy of polygenic risk scores, with a particularly large improvement in diverse populations (Hujoel et al., 2021).

The overlap in ASD PRS means of ASD cases and controls when they have ASD or select other psychiatric family histories may partly reflect misclassification of some controls who, with longer follow-up, may be identified as cases. The overlap may also reflect controls with the broad autism phenotype (Sasson et al., 2013; Warrier et al., 2019) which may share genetic underpinnings with ASD. Finally, the overlap may reinforce other evidence for genetic overlap between ASD and other psychiatric disorders (Grove et al., 2019). As shown in our secondary analyses, the pattern of mean ASD PRS by psychiatric family history was distinct from the pattern of mean BMI PRS by psychiatric family history or mean ASD PRS by diabetes family history.

We assessed variation in mean ASD PRS by psychiatric family history by sex or ID status in ASD cases and there tended to be lower means in autistic females or autistic persons with ID in most, but not all psychiatric family history groups, although larger sample sizes are needed to confirm the pattern.

The study strengths include a large, population-based sample with unbiased, prospective collection of diagnostic information for all study participants and their family members. Although diagnostic misclassification is possible, validation of select diagnoses (e.g. schizophrenia, single episode depression, dementia) including childhood autism has been carried out with good results (Mors et al, 2011; Lauritsen et al 2010). Genetic information was obtained from iPSYCH which applies rigorous quality control and data management protocols for genomic data. PRS were derived using state-of the-art methods (Loh et al., 2018) and multiple analyses were based on both PRS deciles and a continuous measure (standardized PRS z-scores). We conducted several secondary analyses to assess stability and consistency of the primary findings with regards to ASD which suggested that the ASD case versus control patterns of ASD PRS and psychiatric family history were distinct from patterns with other non-psychiatric PRS or family histories. The sex- and ID-stratified analyses lacked precision, however, due to small sample sizes and we lacked in-depth clinical information to stratify the ASD case group into more homogeneous phenotypes. Selection of the specific mental conditions pulled out for the psychiatric family history hierarchy was informed by knowledge of their family history risks, genetic correlations and co-occurrence risks with ASD but the precise order was somewhat arbitrary. Notably, however, the precise order may not matter with regard to PRS: mean PRS scores in cases across the family history groups in the hierarchy were quite similar or mean PRS scores in controls across the family history groups were quite similar (greatest difference with an ASD family history) and the risk for ASD from PRS did not depend on psychiatric family history group.

In conclusion, both individual genetic liability for autism from common variants captured by current PRS and psychiatric history in first degree relatives appear to contribute to autism and, statistically, appear to represent different components of family-based risk. Although the latter is unlikely to be completely true, the results inform our understanding of what these two factors are, or are not, currently measuring. Acknowledging the current ‘bluntness’ of our measures, it would appear nevertheless that family history, not unexpectedly, may be tapping a reservoir of family risk beyond inherited common gene variants for ASD – possibly even non-genetic sources of risk. And current ASD PRS may be tapping a reservoir of family genetic risk in common variants beyond those attributed to psychiatric disorder diagnoses in the family. Fuller understanding of the underlying risk architecture of ASD will require enhanced metrics of autism genetic liability (e.g., rare variant burden), further understanding of the components comprising psychiatric family history risk and consideration of other non-genetic risks. Clinical risk prediction efforts for ASD may advance by careful attention to the sourcing of risk and ‘cross-talk’ among these two fundamental factors which may be best viewed, at least at the present time, as complementary measures of family-based autism risk.

Supplementary Material

ASD PRS and psych family hx_Supple material

ACKNOWLEDGEMENTS

The study was supported by the Lundbeck Foundation (iPSYCH, Grant numbers R102-A9118 and R155-2014-1724) and by the National Institute of Environmental Health Sciences of the National Institutes of Health under Award No. R01ES026993. Data handling and analysis on the GenomeDK HPC facility was supported by the NIMH (grant no. 1U01MH109514-01). High-performance computer capacity for the handling and generation of PRS data on the GenomeDK HPC facility was provided by the Center for Genomics and Personalized Medicine and the iSEQ center at Aarhus University, Denmark. Research conducted using the Danish National Biobank resource is supported by the Novo Nordisk Foundation. The funding sponsors had no role in the study design; in the collection, analysis, and interpretation of data; in the writing of the report; and in the decision to submit the paper for publication. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Footnotes

CONFLICT OF INTEREST

The authors have no potential conflict of interest from financial or personal relationships with other people or organizations to disclose.

REFERENCES

  1. Agerbo E, Sullivan PF, Vilhjálmsson BJ, et al. Polygenic risk score, parental socioeconomic status, family history of psychiatric disorders, and the risk for schizophrenia: A Danish population-based study and meta-analysis. JAMA Psychiatry. 2015;72(7):635–641. doi: 10.1001/jamapsychiatry.2015.0346 [DOI] [PubMed] [Google Scholar]
  2. Albiñana C, Grove J, McGrath JJ, Agerbo E, Wray NR, Bulik CM, Nordentoft M, Hougaard DM, Werge T, Børglum AD, Mortensen PB, Privé F, Vilhjálmsson BJ. Leveraging both individual-level genetic data and GWAS summary statistics increases polygenic prediction. Am J Hum Genet. 2021. Apr 30:S0002-9297(21)00145-2. doi: 10.1016/j.ajhg.2021.04.014. Epub ahead of print. [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. American Psychiatric Association. Diagnostic and statistical manual of mental disorders. 5th ed. 2013, Arlington, VA: American Psychiatric Association. [Google Scholar]
  4. Bai D, Yip BHK, Windham GC, et al. Association of genetic and Environmental factors with autism in a 5-country cohort [published online ahead of print, 2019 Jul 17]. JAMA Psychiatry. 2019;76(10):1035–1043. doi: 10.1001/jamapsychiatry.2019.1411. [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Bailey A, et al. , Autism as a strongly genetic disorder: evidence from a British twin study. Psychol Med, 1995. 25(1): p. 63–77. [DOI] [PubMed] [Google Scholar]
  6. Baio J, Wiggins L, Christensen DL, et al. Prevalence of Autism Spectrum Disorder among children aged 8 years - Autism and Developmental Disabilities Monitoring Network, 11 sites, United States, 2014. MMWR Surveill Summ 2018; 67(No. SS-6):1–23. DOI: 10.15585/mmwr.ss6706a1 [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Cornelis MA, Zaitlen N, Hu FB, Kraft P, Price AL. Genetic and environmental components of family history in type 2 diabetes. Hum Genet. 2015; 134(2): 259–267. doi: 10.1007/s00439-014-1519-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Cross-Disorder Group of the Psychiatric Genomics Consortium, Lee SH, Ripke S, et al. Genetic relationship between five psychiatric disorders estimated from genome-wide SNPs. Nat Genet. 2013;45(9):984–994. doi: 10.1038/ng.2711 [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Cross Disorder Group of The Psychiatric Genetics Consortium. Identification of risk loci with shared effects on five major psychiatric disorders: a genome-wide analysis. Lancet, 2013. 381(9875): p. 1371–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Do CB, Hinds DA, Francke U, Eriksson N. Comparison of family history and SNPs for predicting risk of complex disease. PLoS Genet. 2012;8(10):e1002973. doi: 10.1371/journal.pgen.1002973. Epub 2012 Oct 11. [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Grove J, Ripke S, Als TD, et al. Identification of common genetic risk variants for autism spectrum disorder. Nat Genet. 2019; 51(3):431–444. doi: 10.1038/s41588-019-0344-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Hansen SN, Schendel DE, Francis RW, et al. Recurrence Risk of Autism in Siblings and Cousins: A Multinational, Population-Based Study. J Am Acad Child Adolesc Psychiatry. 2019;58(9):866–875. doi: 10.1016/j.jaac.2018.11.017 [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Hujoel MLA, Loh P-R, Neale BM, Price AL. Incorporating family history of disease improves polygenic risk scores in diverse populations. bioRxiv 2021.04.15.439975; doi: 10.1101/2021.04.15.439975. [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Knudsen LB, Olsen J. The Danish Medical Birth Registry. Dan Med Bull. 1998;45(3):320–323. [PubMed] [Google Scholar]
  15. Lauritsen MB, Jørgensen M, Madsen KM, Lemcke S, Toft S, Grove J, et al. Validity of childhood autism in the Danish Psychiatric Central Register: findings from a cohort sample born 1990–1999. J Autism Dev Disord 2010; 40(2):139–48. [DOI] [PubMed] [Google Scholar]
  16. Loh P-R, Tucker G, Bulik-Sullivan BK, et al. Efficient Bayesian mixed model analysis increases association power in large cohorts. Nat Genet. 2015. Mar; 47(3): 284–290. Published online 2015 Feb 2. doi: 10.1038/ng.3190 [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Loh PR, Kichaev G, Gazal S, Schoech AP, Price AL. Mixed-model association for biobank-scale datasets. Nat Genet. 2018. Jul;50(7):906–908. doi: 10.1038/s41588-018-0144-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Loh PR, Kichaev G, Gazal S, Schoech AP, Price AL. Mixed-model association for biobank-scale datasets. Nat Genet. 2018;50(7):906–908. doi: 10.1038/s41588-018-0144-6 [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Loh PR, Genovese G, McCarroll SA. Monogenic and polygenic inheritance become instruments for clonal selection. Nature. 2020. Aug;584(7819):136–141. doi: 10.1038/s41586-020-2430-6. Epub 2020 Jun 24. [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Lu Y, Pouget JG, Andreassen OA, Djurovic S, Esko T, Hultman CM, Metspalu A, Milani L, Werge T, Sullivan PF. Genetic risk scores and family history as predictors of schizophrenia in Nordic registers. Psychological Medicine 2018; 48: 1201–1208. 10.1017/S0033291717002665. [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Lynge E, Sandegaard JL, Reboli M. The Danish National Patient Register. Scand J Public Health 2011; 39(7 Suppl): 30–33. [DOI] [PubMed] [Google Scholar]
  22. Maronna RA, Zamar RH. Robust estimates of location and dispersion for high-dimensional datasets, Technometrics 2002; 44: 307–317, DOI: 10.1198/004017002188618509 [DOI] [Google Scholar]
  23. Martin AR, Gignoux CR, Walters RK, et al. Human demographic history impacts genetic risk prediction across diverse populations. Am J Hum Genet 2017; 100(4):635–649. doi: 10.1016/j.ajhg.2017.03.004. Epub 2017 Mar 30. [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Mors O, Perto GP, Mortensen PB. The Danish Psychiatric Central Research Register. Scand J Public Health 2011; 39(7 Suppl): 54–57. [DOI] [PubMed] [Google Scholar]
  25. Pedersen CB. The Danish Civil Registration System. Scand J Public Health 2011; 39: 22–25. [DOI] [PubMed] [Google Scholar]
  26. Pedersen CB, Bybjerg-Grauholm J, Pedersen MG, Grove J, Agerbo E, Bækvad-Hansen M, Poulsen JB, Hansen CS, McGrath JJ, Als TD, Goldstein JI, Neale BM, Daly MJ, Hougaard DM, Mors O, Nordentoft M, Børglum AD, Werge T, Mortensen PB. The iPSYCH2012 case-cohort sample: new directions for unravelling genetic and environmental architectures of severe mental disorders. Mol Psychiatry 2018. Jan;23(1):6–14. doi: 10.1038/mp.2017.196. Epub 2017 Sep 19. [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Privé F, Aschard H, Ziyatdinov A, Blum MGB. Efficient analysis of large-scale genome-wide data with two R packages: bigstatsr and bigsnpr. Bioinformatics. 2018. Aug 15;34(16):2781–2787. doi: 10.1093/bioinformatics/bty185. [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Robinson EB, Samocha KE, Kosmicki JA, et al. Autism spectrum disorder severity reflects the average contribution of de novo and familial influences. Proc Natl Acad Sci U S A. 2014. Oct 21;111(42):15161–5. doi: 10.1073/pnas.1409204111. Epub 2014 Oct 6. [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Robinson EB, et al. , Genetic risk for autism spectrum disorders and neuropsychiatric variation in the general population. Nat Genet, 2016. 48(5): 552–5. [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Sasson NJ, Lam KS, Parlier M, Daniels JL, Piven J. Autism and the broad autism phenotype: familial patterns and intergenerational transmission. J Neurodev Disord. 2013; 5(1):11. Published 2013 May 2. doi: 10.1186/1866-1955-5-11 [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Schendel DE, Thorsteinsson E. Cumulative incidence of autism Into adulthood for birth cohorts in Denmark, 1980-2012. JAMA. 2018. Nov 6;320(17):1811–1813. doi: 10.1001/jama.2018.11328. [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Searle SR, Speed FM, Milliken GA. Population marginal means in the linear model: An alternative to least squares means. American Statistician 1980; 34: 216–221. [Google Scholar]
  33. Speed D, Balding DJ. MultiBLUP: improved SNP-based prediction for complex traits. Genome Res. 2014. Sep;24(9):1550–7. doi: 10.1101/gr.169375.113. [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Vanderweele TJ, Vansteelandt S. Odds ratios for mediation analysis for a dichotomous outcome. Am J Epidemiol. 2010. Dec 15;172:1339–48. doi: 10.1093/aje/kwq332. Epub 2010 Oct [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Warrier V, Toro R, Won H, et al. Social and non-social autism symptoms and trait domains are genetically dissociable. Commun Biol. 2019;2:328. Published 2019 Sep 3. doi: 10.1038/s42003-019-0558-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Weissbrod O, Kanai M, Shi H, Gazal S, Peyrot W, Khera A, Okada Y, The Biobank Japan Project, Martin A, Finucane H, Price AL. Leveraging fine-mapping and non-European training data to improve trans-ethnic polygenic risk scores. medRxiv 2021.01.19.21249483 doi: 10.1101/2021.01.19.21249483 [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Xie S, Karlsson H, Dalman C, et al. Family History of Mental and Neurological Disorders and Risk of Autism. JAMA Netw Open. 2019;2(3):e190154. Published 2019 Mar 1. doi: 10.1001/jamanetworkopen.2019.0154 [DOI] [PMC free article] [PubMed] [Google Scholar]
  38. Yang J, Lee SH, Goddard ME, Visscher PM. GCTA: a tool for genome-wide complex trait analysis. Am J Hum Genet. 2011. Jan 7;88(1):76–82. doi: 10.1016/j.ajhg.2010.11.011. [DOI] [PMC free article] [PubMed] [Google Scholar]
  39. Zhang Qianqian, Prive F Vilhjalmsson, Speed D. Improved genetic prediction of complex traits from individual-level data or summary statihstics. bioRxiv 2020.08.24.265280 doi: 10.1101/2020.08.24.265280. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

ASD PRS and psych family hx_Supple material
NIHMS2006100-supplement-ASD_PRS_and_psych_family_hx_Supple_material.docx (386.9KB, docx)

RESOURCES