eTable. Odds ratios for attempted and completed suicide among relatives of probands with ADHD compared to matched controls, adjusted for comorbid psychiatric disorder
Ljung T, Chen Q, Lichtenstein P, Larsson H. Common Etiological Factors of Attention-Deficit/Hyperactivity Disorder and Suicidal BehaviorA Population-Based Study in Sweden. JAMA Psychiatry. 2014;71(8):958-964. doi:10.1001/jamapsychiatry.2014.363
Copyright 2014 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.
The prevention of suicidal behavior is one of the most important tasks for mental health clinicians. Although a few studies have indicated an increased risk of suicidal behavior among individuals with attention-deficit/hyperactivity disorder, the development of more effective ways of identifying and modifying the risk is hampered by our limited understanding of the underlying mechanisms for this association.
To explore whether attention-deficit/hyperactivity disorder and suicidal behavior share genetic and environmental risk factors.
Design, Setting, and Participants
Matched cohort design across different levels of family relatedness recorded from January 1, 1987, to December 31, 2009. We identified 51 707 patients with attention-deficit/hyperactivity disorder (through patient and prescribed drug registers) in Sweden and their relatives by linking longitudinal population-based registers. Control participants were matched 1:5 on sex and birth year.
Main Outcomes and Measures
Any record of suicide attempt or completed suicide defined by discharge diagnoses of the International Classification of Diseases.
Individuals with attention-deficit/hyperactivity disorder (probands) had increased risks of attempted and completed suicide, even after adjusting for comorbid psychiatric disorders (odds ratio [OR] = 3.62 [95% CI, 3.29-3.98] and 5.91 [95% CI, 2.45-14.27], respectively). The highest familial risk was observed among first-degree relatives (attempted suicide: OR = 2.42 [95% CI, 2.36-2.49] among parents of probands with ADHD and OR = 2.28 [95% CI, 2.17-2.40] among full siblings of probands with ADHD; completed suicide: OR = 2.24 [95% CI, 2.06-2.43] and OR = 2.23 [1.83-2.73], respectively), whereas the risk was considerably lower among more genetically distant relatives (attempted suicide: OR = 1.59 [95% CI, 1.47-1.73] among maternal half siblings, OR = 1.57 [95% CI, 1.45-1.70] among paternal half siblings, and OR = 1.39 [95% CI, 1.35-1.43] among cousins; completed suicide: OR = 1.51 [95% CI, 1.08-2.10], OR = 2.02 [95% CI, 1.47-2.79], and OR = 1.51 [95% CI, 1.36-1.67], respectively). These familial aggregation patterns remained similar across sex, after excluding relatives with attention-deficit/hyperactivity disorder and probands with suicidal behavior, and after excluding probands and relatives with severe comorbid disorders.
Conclusions and Relevance
Attention-deficit/hyperactivity disorder is associated with an increased risk of both attempted and completed suicide. The pattern of familial risks across different levels of relatedness suggests that shared genetic factors are important for this association. This is an important first step toward identifying the underlying mechanisms for the risk of suicidal behavior in patients with attention-deficit/hyperactivity disorder and suggests that individuals with attention-deficit/hyperactivity disorder and their family members are important targets for suicide prevention and treatment.
Attention-deficit/hyperactivity disorder (ADHD) is a highly prevalent disorder that has been recognized as an important public health problem worldwide.1 In addition to having higher rates of psychiatric comorbidities such as affective disorders,2,3 schizophrenia,4,5 disruptive behavior disorders,6,7 and substance use disorders,8 patients with ADHD have recently been shown to have an increased risk of suicidal behavior.6,9- 12 Developing more effective ways of identifying and modifying the risk of suicidal behavior is therefore a clinical priority, but it is hampered by our limited understanding of the causes of these behaviors in patients with ADHD. Research has suggested that the overlap between ADHD and suicidal behavior may be attributable to ADHD drug treatment13 or co-occurring psychiatric disorders,14 but to our knowledge, no prior studies have explored the extent to which ADHD and suicidal behavior share genetic and environmental risk factors.
Family, twin, and genomic studies suggest that susceptibility to ADHD is largely genetic in origin,15 and the familial clustering of suicidal behavior is partly attributable to genetic factors,16- 18 which may indicate that the association between ADHD and suicidal behavior might be due to shared genetic factors. A genetic explanation would also be expected if, for example, impulsivity is a genetically influenced intermediate phenotype behind both ADHD and suicidal behavior.19,20 On the other hand, environmental risk factors (eg, sexual and physical abuse) may also represent important risk factors for offspring with ADHD21 and suicidal behavior.22,23
Taken together, most previous research exploring the association between ADHD and suicidal behavior has used small selected clinical samples and focused primarily on completed suicide,12 and to our knowledge, no prior study has used genetically informed data. Therefore, our objective was to explore the extent to which ADHD shares genetic and environmental risk factors with suicidal behavior (both completed and attempted suicide) using a large population-based family sample. The use of family data from longitudinal registers allowed us to study the occurrence of suicide in individuals with ADHD and the familial risk across different levels of genetic relatedness. We conducted sensitivity analyses to rule out 2 plausible alternative explanations. First, we explored whether the results were consistent even after excluding relatives with ADHD and probands with suicidal behavior. This analysis was one approach to deal with the possible adverse effect of ADHD medication (eg, atomoxetine hydrochloride).13 Second, given that prior research has suggested that ADHD may increase the risk of completed suicide through comorbid conditions such as affective disorder, schizophrenia, and substance use disorder,14 we explored the familial overlap between ADHD and suicidal behavior after exclusion of such coexisting conditions.
The study was approved by the Ethics Committee at Karolinska Institutet. The requirement for informed consent was waived because the study was register based and the included individuals were not identifiable at any time.
We linked several Swedish nationwide registers using the unique personal identification number. The National Patient Register (NPR) includes records of psychiatric inpatient care since 1973 with complete national coverage and psychiatric outpatient care since 2001. All discharge diagnoses were recorded according to the International Classification of Diseases eighth revision (ICD-8) (1973-1986), ICD-9 (1987-1996), and ICD-10 (1997-2009). The Swedish Prescribed Drug Register was established in 2005 and provides complete national information on all dispensed prescribed pharmaceuticals. The register comprises drug classification according to the Anatomical Therapeutic Chemical number, the date of prescription, and possible generic substitution. The register is almost complete, with missing data for less than 0.3% of the population.24 The Multi-Generation Register links everyone born since 1932 (and living in Sweden at any time during 1961-2009) to their biological parents, which made it possible to identify parents, full siblings, half siblings, and cousins. The Total Population Register provides information on sex, birth year, country of birth until 2009, and migration status between 1969 and 2009. The Cause of Death Register provides information on the underlying and multiple contributory causes of death. The register covers nearly all deceased persons registered in Sweden at the time of their death from 1952 onward.
Patients with a discharge diagnosis of ADHD were identified from the NPR (ICD-9 code 314; ICD-10 code F90) by having at least 1 record of inpatient or outpatient care between January 1, 1987, and December 31, 2009. We also identified individuals treated with medication for ADHD (methylphenidate hydrochloride, atomoxetine, amphetamine sulfate, or dextroamphetamine sulfate) between 2005 and 2009 and classified them as patients with ADHD.
Probands were selected as patients with ADHD between ages 3 and 40 years. A recent validity check indicated high specificity for the register-based diagnoses of ADHD in the Swedish registers.25 Previous research has also shown that the comorbidity patterns of ADHD remain similar regardless of using a definition that requires 1 or at least 2 records of an ADHD diagnosis26 and whether ADHD cases were identified through ICD diagnoses or ADHD medication.27,28 In addition, the facts that ADHD is the only licensed indication for methylphenidate and atomoxetine use in Sweden and that the authority to prescribe medication for ADHD in Sweden is restricted to physicians specializing in ADHD treatment indicate that a prescription of ADHD medication is a valid indicator of ADHD diagnosis.29 Suicidal behavior was defined as any record of suicide attempt or completed suicide (including both definite and uncertain diagnoses); we identified suicide attempts from the NPR (ICD-8 and ICD-9 codes E950-E959, E980-E989; ICD-10 codes X60-X84, Y10-Y34) and completed suicides from the Cause of Death Register (ICD-8 and ICD-9 codes E950-E959, E980-E989; ICD-10 codes X60-X84, Y10-Y34). To reduce possible misclassifications, suicidal behavior was allowed only in individuals aged 12 years or older.30
Consistent with previous register-based research,31 we used a hierarchical approach with information from the NPR to define schizophrenia (ICD-8 and ICD-9 code 295; ICD-10 code F20), bipolar disorder (ICD-8 codes 296.1, 296.3, 296.8; ICD-9 codes 296A, 296C-E, 296W; ICD-10 codes F30-F31), depression (ICD-8 codes 296.2, 298.0, 300.4; ICD-9 codes 296B, 300E; ICD-10 codes F32-F34), and anxiety disorders (ICD-8 code 300 except 300.4; ICD-9 code 300 except 300E; ICD-10 codes F40-F42, F44-F45, F48). The NPR also provided information on conduct disorder (ICD-9 code 312; ICD-10 code F91), antisocial personality disorder (ICD-8 code 301.7; ICD-9 code 301H; ICD-10 code F60.2), borderline personality disorder (ICD-10 code F60.3), and substance use disorder (ICD-8 codes 303-304; ICD-9 codes 303-304, 305A, 305X; ICD-10 codes F10-F19).32 Family education level was assessed by highest attained education among the parents, categorized into 3 groups: elementary education (≤9 years), secondary education (10-12 years), and higher education (≥13 years).
We used a matched cohort design to estimate the risk of suicidal behavior within probands (individuals diagnosed as having ADHD) and among relatives of probands with ADHD. First, we evaluated the association between ADHD and suicidal behavior by comparing the risk of suicidal behavior among probands with ADHD with 5 unaffected control participants. Individuals were matched on sex and birth year as an attempt to eliminate confounding, reduce misclassification of exposure, and ensure equal time at register follow-up. Second, we used a family design in which we estimated the risk of suicidal behavior among relatives of probands with ADHD. For each type of relationship (parents, full siblings, maternal and paternal half siblings, and cousins), we matched each proband with ADHD and his or her exposed relative to 5 unaffected control participants and their corresponding relatives on sex and birth year. Several pairs could be descended from 1 proband (eg, if a proband has several siblings), and each individual in the population could appear in multiple relative groups (eg, parent, siblings). Control participants were chosen among individuals who were alive, living in Sweden, and not diagnosed as having ADHD at the year of the proband’s first ADHD diagnosis. We compared the risk estimates separately for first-degree relatives (eg, offspring and parents), second-degree relatives (eg, half siblings), and third-degree relatives (eg, cousins) to assess the genetic and environmental contribution to the overlap between ADHD and suicidal behavior. An increased risk of suicidal behavior among higher-degree relatives of probands with ADHD would suggest familial effects for the overlap. Additionally, comparing the risk estimates among maternal half siblings and paternal half siblings provides information about shared environmental factors. Basically, if the observed familial association is primarily explained by shared environmental influences, we would expect higher risk estimates among maternal half siblings than among paternal half siblings because offspring predominantly live with their mothers when parents separate.33
We conducted 2 sensitivity analyses. First, we estimated the risk of suicidal behavior in relatives of probands with ADHD after excluding relatives with ADHD and probands with suicidal behavior.34 These exclusions adjust for the possibility that ADHD medication itself might increase or decrease the risk of suicidal behavior. Second, we estimated the risk of suicidal behavior in relatives of probands with ADHD after excluding individuals with substance use disorder, depression, anxiety, conduct disorder, bipolar disorder, schizophrenia, antisocial personality disorder, and borderline personality disorder in probands, control participants, and relatives. This was done to explore the extent to which an observed familial association between ADHD and suicidal behavior was explained by comorbid psychiatric conditions.
We used conditional logistic regression to estimate odds ratios (ORs) of suicidal behavior with 95% confidence intervals. Analyses were performed with SAS version 9.3 statistical software (SAS Institute, Inc) along with a robust sandwich estimator to supply standard errors corrected for the dependence between repeated observations within families.
From the Swedish registers, we identified 51 707 individuals (69.8% males) who met our criteria for ADHD. Among all probands, 17 349 (33.6%) had a comorbid disorder. The distributions of psychiatric disorders and sociodemographic variables for cases and control participants are presented in Table 1. Individuals with ADHD had a lower family education level compared with control participants. With respect to the severity of the suicidal behavior, individuals with ADHD were more likely to use poisoning and hanging as a method of suicidal behavior compared with control participants. However, control participants were slightly younger than probands at the time of suicide.
Individuals with ADHD had an increased risk of both attempted and completed suicide compared with matched control participants (OR = 8.46 [95% CI, 8.07-8.87] and OR = 12.22 [95% CI, 8.67-17.22], respectively), even after adjusting for comorbid psychiatric disorders (OR = 3.62 [95% CI, 3.29-3.98] and OR = 5.91 [95% CI, 2.45-14.27], respectively) (Table 2). The risk of attempted suicide among individuals with ADHD differed significantly by sex (χ2 = 1271.0; P < .001); the adjusted estimate was 2.93 (95% CI, 2.60-3.29) for males and 5.41 (95% CI, 4.60-6.36) for females (Table 2). In contrast, we did not observe a statistically significant sex difference for completed suicide (χ2 = 2.56; P = .11).
Percentages of suicidal behavior for specific groups of individuals exposed and unexposed to a relative with ADHD are given in Table 3.
The OR for attempted suicide was 2.42 (95% CI, 2.36-2.49) among parents of probands with ADHD and 2.28 (95% CI, 2.17-2.40) among full siblings of probands with ADHD. Half siblings and cousins of probands with ADHD were also more likely to attempt suicide compared with relatives of control participants (OR = 1.57-1.59 among half siblings and OR = 1.39 among cousins). The difference in risk for first-degree relatives (ie, full siblings) and second-degree relatives (ie, half siblings) was statistically significant (χ2 = 99.54; P < .001), suggesting familial effects for this overlap. However, the difference in the risk of attempted suicide for maternal and paternal half siblings was not statistically significant (χ2 = 0.628; P = .43), which indicates low support for shared environmental factors. Moreover, the estimates from separate analyses for males and females were similar in magnitude and showed the same pattern of familial risks (data not shown).
Relatives of probands with ADHD had a significantly increased risk of completed suicide. The familial aggregation indicated genetic influences because first-degree relatives of individuals with ADHD were more likely to have completed suicide (OR = 2.23-2.24) than second-degree relatives (OR = 1.51-2.02) and third-degree relatives (OR = 1.51). The influence of shared environmental factors for the familial aggregation of ADHD and completed suicide was limited owing to similar estimates among maternal and paternal half siblings (χ2 = 2.209; P = .14).
First, we investigated the risk of attempted and completed suicide among relatives after excluding relatives with ADHD and probands with suicidal behavior. The familial risks were slightly lower, but a similar pattern was observed across different family relationships (Table 3). Second, we repeated the family analyses after excluding probands, control participants, and relatives with comorbid psychiatric disorders. The pattern of familial risks of attempted suicide remained similar after these exclusions (eTable in Supplement). Unfortunately, the power to detect a genetic overlap between ADHD and completed suicide was not sufficient.
In this large longitudinal nationwide sample of Swedish families, we found that ADHD was strongly associated with attempted and completed suicide in males and females, even after adjusting for comorbid psychiatric disorders. The pattern of familial risks across different levels of genetic relatedness suggests that shared genetic factors contribute to the overlap between ADHD and suicidal behavior. One potential explanation for these results is pleiotropic genetic effects31 for ADHD and suicidal behavior, possibly reflecting genetic variants associated with impulsivity, a trait dimension that is highly heritable,32 is a core component of ADHD, and is strongly associated with suicidal behavior.10,14 Our findings lend support for future studies investigating genetic and neurocognitive factors associated with suicidal behavior in ADHD, needed to identify novel biological markers for improved identification of individuals with ADHD at high risk for suicidal behavior. This is an important target given that ADHD shows considerable clinical heterogeneity in terms of both severity and developmental outcomes.
Although previous studies have emphasized the importance of genetic influences on ADHD, attempted suicide, and completed suicide,17,33,34 our study is the first, to our knowledge, to show that ADHD and suicidal behavior share genetic risk factors. The finding of genetic factors does not exclude the possibility of environmental influences, nor does it exclude specific genetic and environmental factors on ADHD and suicidal behavior or the influence of nonshared confounders and measurement errors. It is also important to note that this finding needs to be replicated in other settings using research methods with different strengths and weaknesses in validity, such as multivariate twin studies and cross-disorder genome-wide association studies.
In addition to this main finding, our study contributes 3 other important findings. First, we showed that the observed familial aggregation pattern remained when tested separately for attempted and completed suicide. This is important given that most studies have investigated the association between ADHD and completed suicide.6 We found that the proportion of attempted suicide was 6.6% for parents and 3.4% for siblings of probands with ADHD. This suggests that relatives of patients with ADHD should be screened for suicidal behavior given that untreated suicide attempts may lead to completed suicide and substantial adversity for family members.35
Second, we demonstrated that the familial risk of suicidal behavior remained similar after excluding probands and control participants with suicidal behavior as well as relatives diagnosed as having ADHD (ie, when including only those who have never been medicated for symptoms of ADHD), indicating that harmful effects of ADHD medication or imitation of suicide within families are unlikely explanations for the observed overlap between ADHD and suicidal behavior. This is particularly important given that previous research has suggested that stimulant or nonstimulant (atomoxetine) ADHD medication may increase the risk of suicidal behavior.13,36,37 Nevertheless, additional research is needed on the relationship between medication use and suicide risk in ADHD populations.
Third, we found that the risk of suicidal behavior in individuals with ADHD was attenuated but remained substantial after adjusting for comorbid psychiatric conditions. This indicates that the association between ADHD and suicidal behavior is only partly explained by coexisting diagnoses. In contrast, a few studies have suggested that ADHD might only be a risk factor for suicidal behavior in combination with other psychiatric conditions.6,14 The conflicting results may be explained by the fact that they used a small sample, which reduces the possibility to detect a statistically significant independent association between ADHD and suicidal behavior. There are many examples of widespread pleiotropic effects of genetic risk variants across a broad range of diagnostic categories, including schizophrenia, bipolar disorder, autism, major depressive disorder, and ADHD.38,39 Our finding that the familial aggregation pattern remained similar after adjusting for comorbid psychiatric disorders does not rule out such pleiotropic effects, but it suggests that at least part of the genetic overlap is specific for ADHD and suicidal behavior.
The strengths of this study include the use of nationwide longitudinal data, which made it possible to study the entire Swedish population until the end of 2009. Further, the family-based design allowed us to explore the genetic and environmental contribution to the overlap between ADHD and suicidal behavior. The use of register data also eliminates the risk of recall bias and reduces the risk of misclassification. However, although recent validation checks indicate low numbers of false-positive diagnoses of ADHD in Swedish registers,25,29 it was not possible to classify probands with ADHD according to the DSM-IV ADHD subtypes because these specific diagnoses were not recorded across the registers. Thus, future research needs to consider symptom data of ADHD to explore whether the familial risk of suicidal behavior is driven by the impulsivity component of ADHD. The estimated risk of suicidal behavior among individuals with ADHD might be biased if ADHD medication by itself is associated with suicidal behavior. However, the robust association after excluding relatives diagnosed as having ADHD in the familial risk analyses did not support strong bias of this type.
The ascertainment of probands with ADHD was predominantly based on ICD-10 diagnoses and prescribed medication unique for ADHD treatment. The ICD-10 definition of ADHD is stricter compared with that in DSM-IV, and national guidelines for medication to treat ADHD state that medication should be reserved for cases in which other supportive interventions have failed, indicating that our proxies for ADHD are most likely underestimating the incidence of ADHD. Thus, generalizations to cases of less severe ADHD symptoms should be made with caution.
Moreover, although mortality statistics with regard to suicides have been validated with Swedish data,40 suicidal behavior might be underestimated owing to misclassification in the registers. However, because suicide is such a rare outcome, possible misclassifications would most likely not affect our estimates. Register data also have limitations in the forms of left truncation and right censoring, which might lead to misclassification of outcome and exposure. For example, the outpatient register was not started until 2001. Possible random misclassification would most likely only lead to underestimation of the familial risk. We handled these limitations by matching on birth year to ensure equal follow-up time. Some comparisons for completed suicides were based on small numbers, resulting in large confidence intervals. However, separate analyses for attempted and completed suicide resulted in a similar pattern of familial risks.
Finally, comparing the risk of suicidal behavior in maternal and paternal half siblings made it possible to investigate the effect of shared environmental influences. This comparison was based on the assumption that maternal half siblings are more similar with regard to shared environmental exposures than paternal half siblings, supported by the fact that the majority (91%) of children among divorced parents in Sweden continue to live with their mothers.33 Nevertheless, if the difference in shared environment between maternal and paternal half siblings is small, we will have limited power to identify shared environmental influences. However, the point estimates for maternal and paternal half siblings were basically identical. Future research should use twin studies or other family-based designs to more thoroughly examine the importance of shared environmental factors.
We have found an increased risk of both completed and attempted suicide among relatives of individuals with ADHD. The pattern of familial risks suggests that this association is partly due to shared genetic factors. These findings represent an important first step toward identifying the underlying mechanisms for the risk of suicidal behavior in patients with ADHD and suggest that individuals with ADHD and their family members are important targets for suicide prevention and treatment.
Corresponding Author: Therese Ljung, PhD, Department of Medical Epidemiology and Biostatistics, Karolinska Institutet, PO Box 281, SE-17177 Stockholm, Sweden (firstname.lastname@example.org).
Submitted for Publication: October 16, 2013; final revision received January 24, 2014; accepted February 25, 2014.
Published Online: June 25, 2014. doi:10.1001/jamapsychiatry.2014.363.
Author Contributions: Dr Ljung had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Ljung, Lichtenstein, Larsson.
Acquisition, analysis, or interpretation of data: All authors.
Drafting of the manuscript: Ljung.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Ljung.
Obtained funding: Lichtenstein, Larsson.
Administrative, technical, or material support: Ljung, Lichtenstein.
Study supervision: Lichtenstein, Larsson.
Conflict of Interest Disclosures: None reported.
Funding/Support: This work was supported by grant 061817-01A1 from the Eunice Kennedy Shriver National Institute of Child Health and Human Development, grant 2012-1678 from the Swedish Council for Working Life and Social Research, and grants 2010-3184 and 2013-2280 from the Swedish Research Council.
Role of the Sponsor: The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.