my undergrad thesis lol
Cannabis Use and Suicidal Ideation: Prospective Longitudinal Analysis of Twin Data
Abstract
Adolescence and young adulthood are developmental periods marked by increased risk of suicidality as well as cannabis use. Previous literature suggests that individuals who use more cannabis during these periods are at increased risk of suicidality, including suicide attempts and ideation. However, the causal nature of this association is uncertain. To clarify causality and the directionality of this association, the present study analyzed associations between cannabis and suicidal ideation using data from three prospective, longitudinal studies of twins (total N = 3,762, 51.94% female). Results indicated that twins who reported greater cannabis use in adolescence and adulthood were significantly more likely to endorse suicidal ideation, and that this association survived when controlling for shared genetic and environmental vulnerabilities. However, follow-up longitudinal analyses indicated that greater cannabis use did not significantly predict increased risk for suicidal ideation across time, whereas suicidal ideation did predict significantly greater cannabis use during early and mid-adolescence. These findings suggest that the association between cannabis and suicidal ideation likely does not reflect a causal effect of cannabis use on ideation, but instead either (a) confounding by unmeasured twin-specific risk factors, or (b) pre-existing distress and experience with suicidal ideation predisposing youth to greater cannabis use, perhaps as a means of self-medication. Based upon these findings, an emphasis on treatment of mental health issues as prevention for substance use is recommended.
Introduction
Suicide is a major public health problem. It is the second leading cause of death for adolescents and younger adults (ages 10 to 34), claiming the lives of approximately 47,500 people in the United States in 2019 alone (Suicide, 2019). The suicide rate, as defined by the number of deaths per 100,000 individuals in a population, has also been increasing over the past two decades—in 1999, the rate was 10.5 per 100,000; by 2019, the rate was 14.2 (Suicide, 2019). Although the 2020 CDC report indicates that suicide rates decreased overall by 3%, rates still increased significantly for the population aged 25 to 34 (Curtin & Hedegaard, 2021). Beyond the individual, suicide also impacts families and communities. Known as “the ripple effect” of suicide, it is estimated that approximately 115 people are exposed to a single suicide, including individuals of varying closeness, from emergency responders and medical professionals to friends, family, schools, and workplaces (Sandler, 2018). Finally, a recent literature review indicates that suicide risk has also increased during the ongoing COVID-19 pandemic, due to financial strain and the toll of stressful events and illness on mental health (Ivbijaro et al., 2021).
One factor that may also increase risk of suicide is cannabis use. Cannabis is one of the most widely used federally illegal substances among adolescents, with 37% of US high school students reporting lifetime use (Jones et al., 2020). Additionally, frequent cannabis use has been increasing over time, from 8.9% of 17-year-olds in 1991-2000 to 10.4% in 2011-18 (Hamilton et al., 2019). It is possible that these increases may explain part of the increase in suicidality observed over the same period, as research has repeatedly found associations between cannabis use and suicide in the general population using a variety of different study designs, including cross-sectional surveys, psychological autopsies, longitudinal cohort studies, and analyses of national registry data.
Associations Between Cannabis Use and Suicide
A recent US national survey reported that increased past-year heavy, daily, and non-daily cannabis use was significantly associated with increased risk of past-year suicidal ideation, planning, and attempts; this association was particularly strong in women (Han et al., 2021). Other cross-sectional studies have reported similar associations in populations outside of the US, including low- and middle-income countries (Borges et al., 2017; Carvalho et al., 2019; Chabrol et al., 2008; Maharajh & Konings, 2005).
Psychological autopsy studies directly examine the possible antecedents to completed suicides for which information is available. These studies can provide additional insight into cannabis use and death by suicide. An investigation of medicolegal autopsy reports of non-natural deaths among illegal drug users found that cannabis-related cases increased from 24% in 2000 to 32% of cases in 2009 (Delaveris et al., 2014). Additionally, medicolegal investigation of deceased illegal drug users found that cases who used only cannabis (detected post-mortem) had higher relative percentage of suicide; also, violent suicide methods (e.g., guns, falling, trains) were most common in cases for which only cannabis was detected post-mortem (Eksborg & Rajs, 2008). Together, this body of work suggests that cannabis is a relatively commonly used substance implicated as a toxicological antecedent of suicide.
Several longitudinal studies have converged on the finding that earlier and more frequent cannabis use is associated with increased risk of suicide. Using a New Zealand birth cohort, Ferguson et al. (2002) found that cannabis use frequency in adolescence was associated with suicidal ideation and attempts in young adulthood, especially for early cannabis users (ages 14 to 15). A recent retrospective longitudinal study replicated these findings: cannabis use during adolescence was associated with suicidal ideation and attempts in young adulthood, particularly for early cannabis users and high frequency users (Hengartner et al., 2020). These findings were also replicated using a Norwegian birth cohort (Pedersen, 2008). Other longitudinal studies have not found a significant association. For example, a longitudinal study using a Swedish military service cohort found that the association between frequent cannabis use and suicide disappeared when controlling for confounding factors, like IQ, alcohol consumption, parental psychotropic medication, smoking, other drug use, psychiatric diagnosis, and other psychosocial measures (Price et al., 2009). Another study using a Finnish birth cohort did not find a statistically significant association between cannabis use and suicide diagnostic codes/deaths, although a significant association with self-harm was identified (Denissoff et al., 2021).
Meta-analyses of longitudinal studies suggest that cannabis use in adolescence is associated with increased risk of depression, anxiety, suicidal ideation, and suicide attempts in young adulthood (Borges et al., 2016; Gobbi et al., 2019). However, these meta-analyses also include important caveats. Namely, Borges et al. (2016) identifies a lack of controlled studies that can address other risk factors that may confound the association between cannabis use and suicidality, as noted in an older review of studies on cannabis-related mortality (Calabria et al., 2010). Gobbi et al. (2019) also discusses the issue of controlled studies and premorbid depression: “…since the individual patient data from the majority of studies were not reanalyzed using causal inference methods, strong causal association cannot be made with respect to the relationship between cannabis and later depression, suicide, or anxiety” (p. 431).
Causal Considerations
Although several studies present evidence for a correlational relationship between frequent cannabis use and suicidality, there is a need for causally informative designs to rule out the possibility that these associations arise solely due to confounding by shared vulnerability factors, including shared genetic and environmental influences (e.g., socioeconomic status, systematic oppression, social welfare policies). It is important to control for confounds because there may be multiple etiological explanations for the association between cannabis use and suicide. Firstly, cannabis use may increase suicidal thoughts and behaviors. This relationship could be stronger or weaker depending on the frequency and potency of cannabis use. Second, baseline levels of suicidal thoughts and behaviors could increase cannabis use. For example, individuals might use cannabis to alleviate pre-existing distress (e.g., suicidal ideation). Third, there might be confounding variables that lead to both increased cannabis use and increased suicide risk. For example, high stress levels associated with low socioeconomic status or systematic oppression could increase risk for both frequent cannabis use and increased suicidal thoughts and behaviors.
To date, many researchers have attempted to account for these possibilities by adjusting for potential confounds in their statistical models. However, this approach has limitations, in that researchers can only adjust for confounds that they think to measure and include in models, and including too many confounds risks “controlling away” real effects. Twin studies provide an avenue for circumventing these limitations, as twins share environmental influences if reared in the same household and monozygotic twins additionally share 100% of the same genes (McGue et al., 2010). Thus, models that compare twins within a family (often labeled “discordant” or “co-twin control models”) can be used to compare genetically and environmentally similar individuals who also happen to differ in cannabis use frequency and suicidal ideation. If a causal relationship does exist between frequent cannabis use and suicidality, possible mechanisms may involve acute effects of cannabis and/or psychosocial effects of cannabis use. Acute effects of cannabis include momentary disinhibition, paranoia, anxiety, and psychosis; these effects could potentially increase suicide risk (Karila et al., 2014). Additionally, cannabis may hinder academic, vocational, and social performance, potentially increasing risk for stress and suicidal ideation. Indeed, stress generation theory suggests that individuals already struggling with mental health and substance use may contribute to a cycle of negative life events, thus generating more distress and contributing to existing problems (Alloy et al., 2010).
Evidence from Twin Studies
Thus far, twin studies of cannabis and suicidality have found significant associations, even after controlling for shared environmental and genetic risk. Retrospective data from a discordant twin study showed an association between frequent cannabis use and suicidal thoughts and behaviors, as well as depression at the individual-level, controlling for covariates such as sex, sample or study effects, and birth year (Agrawal et al., 2017). The authors also found that monozygotic twins who reported more frequent cannabis use were more likely to report suicidal ideation than their lesser-using co-twin (Agrawal et al., 2017). Results of another study using co-twin control models found that any level of cannabis use was modestly associated with suicidal ideation and unplanned suicide attempts (Delforterie et al., 2015). This association held even when controlling for other psychiatric disorders and substance use.
Although these twin studies provide suggestive evidence that an association between cannabis use and suicidal ideation survives when controlling for confounds, there are important limitations related to methodology. For example, reliance on cross-sectional data precludes temporal ordering, which is necessary to infer direction of causation. Retrospective longitudinal data is also limited, insofar as retrospective measures may lead to biased estimates due to recall errors and other well-documented sources of bias. The present study aimed to address these limitations by examining associations between frequency of cannabis use and suicidal ideation in a large twin sample assessed from adolescence into adulthood using prospective, repeated measurements. In addition to leveraging the twin data to test for evidence of a causal relationship between cannabis and suicidal ideation in adolescence and young adulthood, we also conducted longitudinal analyses that aimed to clarify temporal ordering and rule out reverse-causation.
Consistent with prior twin research, we hypothesized that cannabis use would be associated with suicidal ideation cross-sectionally, controlling for shared genetic and environmental confounds. We further hypothesized that suicidal ideation increases risk of cannabis use, such that prospective associations of suicidal ideation predicting cannabis use would be stronger than prospective associations of cannabis use predicting suicidal ideation.
Methods
Participants and Procedures
Participants were a combined twin sample (N = 3,762) from three longitudinal studies at the Minnesota Center for Twin and Family Research (MCTFR). Twin pairs were identified from Minnesota birth records. To be eligible for study participation, twins had to reside within a day’s drive of Minneapolis, live with at least one biological parent, and have no physical or mental conditions that would interfere with completing of a day-long, in-person assessment. The “Enrichment Sample” (ES) cohort (n = 998) and “Younger” cohort (n = 1,512) were first assessed with their parents at age 11, and the “Older” cohort (n = 1,252) was first assessed with their parents at age 17. Follow-up assessments were conducted approximately every 3 years into young adulthood (see Table 1 for the timing of assessments across all three cohorts). This study uses data from early adolescence until the peak of cannabis use in young adulthood. Detailed overviews of the MCTFR, twin samples, inclusion/exclusion criteria, and procedures and assessments are provided in previous articles (Iacono et al., 2006; Keyes et al., 2009; Wilson et al., 2019).
The Institutional Review Board of the University of Minnesota approved these studies at each wave. After the study protocol was explained, caregivers provided permission for their minor children to participate, and children provided written assent. Informed consent was obtained from the participants during each assessment wave conducted when participants were 18 years of age or older.
Measures
Cannabis use frequency. Frequency of cannabis use over the past year was assessed at ages 11, 14, 17, 20, and 24 years using either a computerized substance use (CSU) inventory, the SAM of the CIDI (Robins et al., 1987, 1988), or both measures. An ordinal scale was derived from this measure where cannabis use frequency may be rated as follows: 0 = None, 1 = Less than once per year to less than once per month, 2 = 1-3 times per month, 3 = 1-4 times per week, 4 = Nearly every day to once a day, and 5 = Two or more times per day. We then derived indices of cumulative cannabis use frequency by averaging these index scores across assessment waves.
Suicidal Ideation. Twins were interviewed regarding symptoms of major depressive disorder (MDD) at ages 11, 14, 17, 20, and 24, which included a single item assessing thoughts of suicide. At ages 11 and 14, these interviews were conducted using the DICA-R. For subsequent assessments (ages 17, 20, and 24), the Structured Clinical Interview for DSM-III-R and IV (Spitzer et al., 1987) was used. Symptom endorsement was coded based on twins’ self-report and parent-report using a best-estimate approach at ages 11, 14, and 17, and using twin report only at ages 20 and 24. Because suicidal ideation was assessed in the context of a larger clinical interview, participants were only asked about thoughts of suicide if initial gateway criteria for MDD were met (i.e., they must have met criteria for depressed mood or loss of interest or pleasure [i.e., anhedonia] for most of the day, for at least most of the days, for 2 weeks). Like cannabis use, cumulative measures of suicidal ideation were derived so that twins were given as score of “1” if they ever experienced suicidal ideation prior to a specific assessment wave, and a “0” if they did not.
Statistical Analyses
We examined associations between adolescent cannabis use and suicidality in adolescence and young adulthood using (1) individual-level models, comparable to linear regression models in a sample of singletons, and (2) co-twin control models, which control for shared familial confounds. A significant limitation of individual-level analyses is that they do not control for genetic and environmental confounds, which limits causal inference. Co-twin control analyses decompose each individual twin’s cannabis exposure into between-pair and within-pair effects (McGue et al., 2014). The between-pair effect is represented by the twin pair’s mean cannabis use frequency, and the within-pair effect is represented by the difference between the twin pair’s mean and an individual twin’s score. The between-pair effect in this parameterization largely captures genetic and shared environmental propensity toward suicidal ideation among reared-together twins, independent of individual twins’ cannabis use status, and a significant between-pair effect suggests that differences in suicidal ideation are consistent with a preexisting liability toward cannabis use. Using co-twin control analyses, twins discordant for cannabis use frequency can thus be compared for suicidal ideation to derive effect estimates that control for these shared sources of risk. Within twin-pair effects were of primary interest because significant associations at the within-pair level would be consistent with a causal relationship between cannabis and suicidal ideation, rather than one that is wholly attributable to shared genetic or environmental risk.
A significant within-pair effect would implicate three possible explanations: greater cannabis use causes suicidal ideation, suicidal ideation causes greater cannabis use, or the association arises primarily due to unmeasured twin-specific factors not accounted for in our co-twin control models. Although it is impossible to rule out the third possibility using observational data, longitudinal assessments can be used to adjudicate between the first two by examining temporal ordering. Thus, we conducted prospective analyses testing whether individual-level differences in cannabis use predicted subsequent suicidal ideation (controlling for baseline ideation), and vice versa.
All analyses were conducted using R Studio Version 2022.02.0 using the “lmer” and “glmer” functions from the “lme4” package. Sex, zygosity, cohort membership, and age at time of outcome assessment were included as covariates in all models. We additionally controlled for either cumulative cannabis use frequency or cumulative history of suicidal ideation in models examining temporal ordering.
Results
Descriptive analyses of cannabis use frequency and suicidal ideation indicated that out of the 3,762 participants with cannabis use data, 1,612 (42.85%) participants reported at least some cannabis use between ages 11 and 24 and 320 (8.52%) reported severe use, which we defined as reporting cannabis use nearly every day or more. A total of 533 out of 3,762 (14.17%) of participants reported suicidal ideation at least once during the study period.
Cross-Sectional Analyses
We tested for cross-sectional associations between frequency of cannabis use and suicidal ideation at the individual level. Within-wave analyses revealed a significant association at each time point (see Table 3), with the highest odds ratio (OR) occurring at the age 14 assessment. Analyses testing for associations between cumulative cannabis use (i.e., average frequency of use between ages 11 and 24) and suicidal ideation indicated similar results, suggesting that each 1-unit increase in our cumulative cannabis use measure was associated with 2.73 greater odds of reporting suicidal ideation at some point during the study period.
At the co-twin control level, the full sample (including both dizygotic and monozygotic twin pairs) was analyzed first. We conducted these analyses using cumulative cannabis use frequency and suicidal ideation variables to maximize the number of discordant twins and reduce the possibility of Type II error. Full sample results for co-twin control analyses indicated a significant within-twin effect, wherein greater cannabis use frequency was associated with increased risk for suicidal ideation (OR = 2.07, 95% CI = 1.47-2.93). Analyses were repeated using monozygotic twins only. For this group, cumulative within twin-pair effects were also significant (OR = 2.66, 95% CI = 1.50-4.70), consistent with a potentially causal effect of cannabis use on suicidal ideation, above and beyond shared genetic and environmental factors.
Prospective Longitudinal Analysis
We tested whether greater cannabis use prospectively predicts increased suicidal ideation by conducting individual-level analyses examining cumulative cannabis use frequency at each wave and odds of suicidal ideation at subsequent waves, controlling for previous suicidal ideation. Results indicated that cumulative cannabis use frequency at each wave did not significantly predict suicidal ideation at the next wave once previous suicidal ideation was accounted for in the models (see Table 4).
We tested whether suicidal ideation prospectively predicts increased cannabis use by conducting individual-level analyses examining suicidal ideation at each wave and cumulative cannabis use at subsequent waves, controlling for previous cannabis use. Because cannabis use frequency is a continuous variable, results are reported as regression coefficients (the average increase in cannabis frequency associated with reporting suicidal ideation). Results indicated that suicidal ideation in early adolescence significantly predicted increases in cannabis use frequency, although this association became nonsignificant in subsequent waves (see Table 5).
Discussion
The present study aimed to establish the nature of the association between cannabis use and suicidal ideation, a common mental health problem with potentially devastating consequences, particularly if the ideation is acted upon. The results of our analyses support our hypotheses twofold: cannabis use and suicidal ideation are associated cross-sectionally controlling for shared genetic and environmental risk, and greater suicidal ideation predicts increased cannabis use over time during adolescence.
Our cross-sectional findings are consistent with those of other twin studies, which have reported associations between cannabis use and suicidality (e.g., ideation, attempts) in general survive when controlling for shared genetic and environmental influences, in addition to controlling for other psychiatric disorders and substance use (Delforterie et al., 2015; Lynskey et al., 2004). On the individual level, within-wave analyses found that the strongest association occurred at age 14. One possible interpretation for this is that cannabis use at age 14—a relatively early age of cannabis use initiation—is more likely to occur against a backdrop of other risk factors that could also contribute to suicidal ideation. It is also possible that the physiological effects of cannabis use that may contribute to suicidal ideation are more pronounced in younger users.
Previous literature has suggested a time-sensitive association, wherein earlier and more frequent cannabis use during adolescence may pose a higher risk for suicidality in young adulthood (Ferguson et al., 2002; Hengartner et al., 2020; Pederson, 2008). The present study enriches these findings by suggesting that suicidal ideation during early adolescence may predict greater cannabis use during mid- and late adolescence. In the context of our co-twin control results, our findings suggest that the association between cannabis use and suicidal ideation may emerge primarily because adolescents experiencing suicidal ideation are more likely to engage in early cannabis use, rather than because cannabis use leads to future suicidal ideation.
Strengths and Limitations of the Present Study
Because the Minnesota Twin and Family Study was designed explicitly to examine relationships between substance use and mental health, the present study has several strengths. In addition to the use of twin data that permitted controlling for shared genetic and environmental vulnerabilities, participants also completed repeated measures of key study constructs at multiple time points from adolescence into early adulthood, permitting inferences regarding the directionality of observed associations. Our findings indicated a significant association between greater cannabis use and suicidal ideation, which persisted after controlling for shared genetic and environmental vulnerabilities. In addition, prospective analyses revealed that suicidal ideation predicted greater cannabis use during early adolescence, rather than vice versa. Our findings thus confirm those of past twin studies with coarser, retrospective measures of cannabis use and suicidality, while also extending these results by showing the directionality of the effect and identifying developmental periods where this association appears strongest.
The present study also has some limitations. Firstly, the reliance on self-report measures of past-year cannabis use frequency could bias our results, as participants could feasibly misremember, misjudge, or misconstrue their cannabis use. Future studies assessing cannabis use on a more frequent basis or objectively (e.g., via toxicology) could help address this concern. Additionally, questions ascertaining suicidal ideation were gated behind more general depression symptom questions in our cohorts. Participants who experienced suicidal ideation in the absence of cardinal symptoms of depression (e.g., low mood, anhedonia) may thus be miscategorized. Our analyses controlled fully for shared genetics in monozygotic twins and partially in dizygotic twins, as well as general shared environmental influences (e.g., shared home, school, and neighborhood). However, these analyses do not control for various twin-specific (i.e., nonshared) factors potentially influencing the association between cannabis use and suicidal ideation. For example, twin-specific measures of personality, victimization, and other traumatic experiences could be influencing both cannabis use and suicidal ideation. Finally, generalizability of results is limited by the lack of sociodemographic diversity in the sample. Although representative of Minnesota during the 1970s to 1990s—predominantly white—the sample is no longer representative of the increasing sociodemographic diversity in Minnesota and across the US.
Implications of the Present Study
Results of this study have potential implications for evidence-based recommendations for clinicians and policymakers, as well as future research directions. Rather than being a risk factor for poor mental health, cannabis use (among other commonly used substances) may instead be better viewed as a coping mechanism used by youth already struggling with suicidal ideation and related mental health problems. Sometimes referred to as “self-medication,” youth may use cannabis to deal with problems such as suicidal ideation, instead of more effective interventions (e.g., psychotherapy, pharmacotherapy). In addition to careful assessment of suicidal ideation, clinicians should also be aware that adolescent cannabis use may be an indicator of serious mental health concerns, including suicidal ideation. In such cases, prioritizing treatment of mental health is likely to be more effective for adolescent well-being than a more limited focus on reducing cannabis use.
References
