Association of Depression With Past-Month Cannabis Use Among US Adults Aged 20 to 59 Years, 2005 to 2016

Key Points Question Did the association of depression with past-month cannabis use among US adults change from 2005 to 2016? Findings In this repeated cross-sectional study of 16 216 adults, those with depression increased their rates of cannabis use significantly faster than those without depression. In 2005 to 2006, individuals with depression had 46% higher odds of any cannabis use and 37% higher odds of near-daily cannabis use, while in 2015 to 2016, individuals with depression had 130% higher odds of any cannabis use and 216% higher odds of daily cannabis use. Meaning In this study, an increasing number of adults with depression used cannabis during the study period.


Introduction
Cannabis is among the most widely used psychoactive substances in the United States. 1 During the past 2 decades, the prevalence of adult cannabis use has steadily increased [2][3][4] ; from 2002 to 2017, past-month cannabis use increased by 98%, and the prevalence of daily or near-daily cannabis use increased by 40%. 1 Simultaneously, the perceived risks associated with cannabis have decreased, 5,6 perceived availability has increased, 3 and state marijuana laws have become more permissive, with 34 states now permitting medical marijuana use and 11 states permitting recreational use. 7 As a result, concern is increasing regarding potential consequences of cannabis use, including cannabis use disorder, vehicle crashes, social impairment, and mental disorders. [8][9][10] A particular concern is the potential adverse effects of cannabis on major depression, which is consistently associated with cannabis use. 2,[11][12][13][14][15][16] In the United States, major depression is among the most common reasons for seeking health care, 17 affecting more than 17 million adults and resulting in considerable impairment. 18, 19 Existing evidence indicates that cannabis may worsen depressive symptoms, [20][21][22] particularly if used regularly. 20,21,23 However, much of the public views cannabis as helpful for treating depression. In a national survey of US adults, nearly 50% reported their belief that cannabis is beneficial for anxiety or depression, while only 15% believed cannabis increases the risk of these conditions. 24 Depression is among the most commonly self-reported reasons for cannabis use, [25][26][27][28] and nearly 25% of adults with mood or anxiety disorders report using cannabis to self-medicate. 29,30 The belief that cannabis may alleviate symptoms of mood disorders may be partially because of the proliferation of misleading information from media and advertising. Indeed, some of the most popular sources of information about cannabis, including the internet, social media, and the marijuana industry, 31 have repeatedly been shown to present false or misleading information about the health effects of cannabis. [32][33][34] Among the most common health claims made in online advertising for recreational cannabis dispensaries is depression treatment. 35 These messages may be increasing in frequency, while media messaging about marijuana has become more positive over time and includes less information about risks. 34 Given these changes in cannabis use and media presentation of cannabis as beneficial for health conditions including depression, understanding time trends in the association of cannabis use with depression has become important. Therefore, we investigated the prevalence of any past-month cannabis use and daily or near-daily past-month cannabis use among US adults with and without depression. We then investigated trends in cannabis use and depression over time and trends in the association between any and daily or near-daily cannabis use and depression from 2005 to 2016.
We hypothesized that depression would be associated with an increased likelihood of cannabis use (any and daily or near daily) and that this association would have strengthened over the study period.

Setting, Participants, and Procedures
Deidentified data for these analyses come from the National Health and Nutrition Examination Survey (NHANES). 36 The NHANES is an annual, nationally representative cross-sectional survey of the US civilian population that uses multistage area probability sampling. The stages of sample selection are as follows: (1) selection of primary sampling units (PSUs); (2) segments within PSUs (ie, a block or group of blocks containing a cluster of households); (3) households within segments; and (4) at least 1 participant within each household. [36][37][38] The 2007 to 2010 NHANES survey cycles oversampled major US demographic subgroups, including Hispanic and non-Hispanic Black persons and low income White persons. 39 Sample weights and poststratification adjustments during analysis are used to account for oversampling and to control for nonresponse, providing study estimates reflecting US Census Bureau population distributions. Additional information describing the complex sampling weight methodology for the NHANES is presented elsewhere. [36][37][38] In the first step of the NHANES survey, trained interviewers conducted interviews in participants' homes. In the second step, participants were asked to attend a mobile examination center (MEC) for physical examinations and additional questionnaires, which generally occurred within 1 to 2 weeks after the in-home interview. 40 Informed consent was obtained from all participants, and all data collection protocols were approved by the National Center for Health Statistics research ethics review board. The environment, equipment, and specimen collection within MECs were standardized. As compensation for participating, all participants were given a cash payment as well as reimbursement for any transportation and child or elder care. 40 The proportion of participants who attended the MEC examination following the at-home interview between 2005 to 2006 and 2015 to 2016 was very high, ranging from 95.7% to 97.3%. 41 Because NHANES data are released in biannual groupings, the current study merged and year. Cannabis use was covered among participants aged 20 to 59 years in all study periods, so the sample for analysis included all NHANES respondents aged 20 to 59 years. Demographic covariates were assessed during the home visit by an interviewer, while depression and cannabis use were assessed using the audio computer-assisted personal self-interview questionnaire in the MEC. 42 Using this format, respondents hear questions through headphones and read questions on a computer screen. Participants then respond to the questions at their own speed using a touch screen. Because the data are publicly available and deidentified, the current study was exempt from review by the Columbia University Medical Center institutional review board. This report was prepared using the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.

Measures Outcomes
Two measures were created to examine past-month cannabis use, both of which dichotomized responses to the question, "How many times have you used cannabis in the past 30 days?" Any pastmonth cannabis use was defined as using cannabis at least 1 time vs 0 times in the past 30 days. Daily or near-daily past-month use was defined as using cannabis at least 20 times vs less than 20 times in the past 30 days, consistent with prior studies. 1,43

Main Exposure
Probable depression was identified using the Patient Health Questionnaire-9 (PHQ-9). This is a validated and widely used measure of depression in clinical research. 44 Consistent with previous research, major depression was dichotomized as absent vs present with scores of less than 10 and at least 10, respectively. 44

Statistical Analysis
A new sample weight variable for the 6 combined data sets from 2005 to 2016 was created by dividing each MEC sample weight by the total number of data sets, then summing them. 45 Prevalence for all outcomes and demographic characteristics were calculated for each survey year.
Unadjusted and covariate-adjusted logistic regression models were used to assess linear time trends To determine whether associations differed by survey year, the logistic models were rerun including an interaction term between survey year and depression. In all analyses, Taylor series estimation methods were used to obtain standard error estimates. Statistical significance was set at α < .05 for all analyses, and tests were 2-tailed. All analyses were conducted using SAS version 9.4 (SAS Institute).
To check robustness of results, analyses were rerun using a redefined depression variable to potentially indicate severe major depression. This was dichotomized as a PHQ-9 score of at least 20 vs less than 20. 44

Results
The final study sample consisted of 16 216 US adults aged 20 to 59 years, of whom 7768 (weighted percentage, 48.9%) were men, 6809 (weighted percentage, 66.4%) were non-Hispanic White participants, 9494 (weighted percentage, 65.6%) had at least some college education, 11 602 (weighted percentage, 62.4%) had an annual family income of less than $75 000, and 9812 (weighted percentage, 63.5%) were married or living together. Sample characteristics of included participants by survey year are presented in Table 1.

Any Past-Month Cannabis Use From 2005 to 2016
The prevalence of any past-month cannabis use increased from 12

Daily or Near-Daily Past-Month Cannabis Use From 2005 to 2016
The prevalence of daily or near-daily past-month cannabis use increased

Depression and Any Past-Month Cannabis Use
Overall, the crude OR indicating the association of depression with any past-month cannabis use was

Depression and Daily or Near-Daily Past-Month Cannabis Use
The crude odds ratio for depression and daily or near-daily past-month use was 2.

Change in the Association Between Depression and Cannabis Use From 2005 to 2016
The interaction results of depression and survey year indicated that the associations between depression and any cannabis use as well as daily or near-daily cannabis use increased significantly

Sensitivity Analyses
After redefining the primary exposure variable as positive if the PHQ-9 score potentially indicated severe major depression (ie, score Ն20), results were virtually unchanged. The aOR indicating the association between any past-month cannabis use and severe depression was 2.22 (95% CI, 1.32-3.74; P = .003), while the aOR for daily or near-daily past-month use and severe depression was 3.26 (95% CI, 1.62-6.57; P = .001). The tests for interaction between severe depression and survey year were not significant for the association between severe depression and any cannabis use (χ 2 1 = 3.84; P = .05) but showed a significant increase in the association between severe depression and daily or near-daily cannabis use over time (χ 2 1 = 7.71; P = .006).

Discussion
This study used nationally representative data to examine trends in the association between depression and cannabis use in US adults aged 20 to 59 years using the NHANES. Overall, there were 3 major findings, as follows: (1) the prevalence of any past-month cannabis use and daily or neardaily cannabis use increased from 2005 to 2016, while the prevalence of depression remained stable; (2) individuals with depression had approximately double the odds of using cannabis compared with people without depression; and (3)  presenting cannabis as beneficial to health. [32][33][34][35] These results could also be interpreted as indicating that an increasing proportion of individuals who use cannabis are developing depression. However, if this were true, we would expect to see that as the prevalence of cannabis use increased, the prevalence of depression increased as well. However, depression stayed relatively stable during the study period, not supporting the latter interpretation.
These findings are also consistent with a 2020 article using data from the National Survey on Drug Use and Health (NSDUH). 50 Although the NSDUH is a common source of information on substance use in the US, other sources are needed to confirm this survey's findings. In presenting information from a different nationally representative data set, the NHANES, the current study ensures that these results are not because of idiosyncrasies of the NSDUH methodology and association between depression and any cannabis use and depression and daily cannabis use in the most recent time period (ie, 2015-2016). These findings may suggest that cannabis use among people with depression is increasing at a faster rate and, subsequently, have a current association of greater magnitude than has been previously reported. Future studies should examine this association using other data sets, presenting time-stratified ORs to further confirm these findings.

Limitations and Strengths
The results of this study should be considered in light of some limitations. prevalences, data permitting. Fourth, owing to insufficient information regarding past-year or pastmonth cigarette use (the time frames of the cannabis and depression variables), we were unable to control for cigarette use. Future research should measure tobacco use in the same time frames as depression and cannabis and include tobacco as an additional control covariate. Finally, the metric for depression, the PHQ-9, is a screening scale rather than a diagnostic instrument. Future research should aim to identify depression using a clinical diagnosis.
The current study also has multiple strengths that warrant mention. Data are nationally representative and used clustered, stratified sampling to ensure representativeness of estimates.
The data therefore allowed for the valid measurement of prevalence, time trends, and the association between cannabis and depression among US adults over time, a unique advantage to national survey data. Moreover, despite not having a formal diagnosis, the current study used an extensively validated screening scale for major depression, 5 which has been used widely in research studies and is simple and short enough (9 questions) for use in large national surveys.

Conclusions
In this study, we examined the association between depression and cannabis use in US adults aged 20 to 59 years from 2005 through 2016. We observed a significant association between depression and any past-month cannabis use, which strengthened when looking at depression and daily pastmonth cannabis use. From 2005 to 2016, cannabis use increased in the general population. However, there was a particular increase in cannabis use among people with depression, with approximately 30% of individuals with depression using any cannabis and 15% using cannabis near-daily from 2015 to 2016. While further research to understand the mechanisms underlying the increasingly strong association of depression and frequent cannabis use is needed, the study findings highlight a current need for information campaigns around cannabis and depressive disorders. In addition, clinicians should be aware of the changing trends and the association between cannabis use and depressive symptoms when treating patients. This is particularly important in reference to frequent cannabis use, which is associated with greater risk of harm and potential worsening of depressive symptoms.