History of cannabis smoking and subjective cognitive complaints in older women

Abstract

Introduction:

Cannabis use has been increasing in the United States (U.S.), yet its potential long-term effects on neurocognitive outcomes remain unknown. We aimed to examine the association of the history of cannabis smoking with subjective cognitive complaints (SCC) in older women.

Methods:

This prospective cohort analysis included 15,378 older women (age ≥ 65 years) of the U.S. National Institute of Environmental Health Sciences’ Sister Study. Participants reported their history of cannabis smoking at enrollment (2003–2009) and completed the Eight-item Interview to Differentiate Aging and Dementia (AD8) as an SCC screener at the cohort’s 2nd (2011–2014), 3rd (2014–2016), and 4th (2017–2019) follow-ups. We used multivariable joint models to assess the association.

Results:

3,973 (25.8%) women reported ever smoking cannabis, mostly in their early adulthood. Compared with never cannabis smokers, the multivariable odds ratios (ORs) for ever smokers were 1.27 (95% confidence interval (CI): 1.13, 1.43) at the 2^nd follow-up, 1.28 (95% CI: 1.14, 1.44) at the 3^rd follow-up, and 1.30 (95% CI: 1.11, 1.52) at the 4^th follow-up. Associations were stronger for regular than occasional cannabis smokers. For example, at the 2^nd follow-up, the OR was 1.61 (95% CI: 1.31, 1.98) for regular smokers and 1.19 (95% CI: 1.04, 1.35) for occasional smokers. Results were overall consistent in subgroup and sensitivity analyses.

Conclusions:

This study suggests a potential association between a history of cannabis smoking and SCC in older women, calling for further research on cannabis use and cognitive outcomes in the context of aging.

Introduction

Cannabis is one of the most commonly used drugs worldwide, with about 219 million individuals reporting use in 2021[1]. In the United States (U.S.), 40 states, three territories, and the District of Columbia have legalized medical cannabis use, and most of them have also legalized its non-medical (i.e., recreational) use[2]. Increasing legalization, coupled with substantial increases in societal acceptance, will likely further drive cannabis use in the coming years. Nonetheless, the potential health impact of cannabis use, particularly its long-term consequences on the cognitive health of susceptible populations, is far from understood[3]. Increased understanding of such health concerns has substantial public health and policy implications.

Although the evidence on the potential risks of cannabis use in aging brains is limited, it has well-documented acute adverse effects on the developing brains of teenagers and young adults[4-7], including neuropsychological deficits[8-10], neuroanatomic changes[11], and neurotransmitter malfunctioning[12]. While these adverse effects may not persist into late adulthood and abstinence may abate these adverse effects to a certain extent[13-15], they may set the stage for a higher susceptibility to cognitive decline in the context of aging. Subjective cognitive complaints (SCC) are self-reports of cognitive decline, which are an early marker of mild cognitive impairment and dementia[16]. Therefore, we investigated the association between the history of cannabis smoking, mostly in early adulthood, and SCC among U.S. women 65 years or older in a nationwide cohort with repeated assessments of self-reported cognitive symptoms.

Methods

Study participants

The National Institute of Environmental Health Sciences’ Sister Study is a nationwide cohort of women from all 50 U.S. states and Puerto Rico, designed to investigate risk factors for breast cancer and other chronic diseases[17]. Briefly, in 2003–2009, 50,884 women aged 35 to 76 years enrolled in the study by providing comprehensive information about their demographics, diet, lifestyle, health behaviors, environmental exposures, and health history. Since enrollment, participants have been followed annually with a short health update survey and with a detailed follow-up survey every 3 years, 1^st during 2008–2012, 2^nd during 2011–2014, 3^rd during 2014–2016, and 4^th during 2017–2019. Subjective cognitive screening was conducted at the Sister Study’s 2^nd, 3^rd, and 4^th follow-ups, using the Eight-item Interview to Differentiate Aging and Dementia (AD8). The AD8 was designed to assess recent changes in cognitive function in older adults, and it has been commonly used in community-based longitudinal studies of adults 65 years or older[18-22]. The current study was limited to 16,521 participants who were at least 65 years old at the 2^nd Sister Study follow-up survey when the AD8 was first administered and who provided valid exposure data on cannabis smoking at enrollment (Fig. 1). Of these, 15,421 individuals completed at least one AD8 screener during the follow-ups. After excluding 43 participants missing on covariates, the final analysis included 15,378 women. In the present analysis, we used the 2^nd follow-up when the AD8 was first administered as the baseline to minimize bias due to left truncation. The Sister Study was approved by the Institutional Review Boards at the National Institutes of Health (approval number 02EN271), and all study participants provided written informed consent.

Fig. 1.

Flowchart of study participation. AD8, eight-item interview to differentiate aging and dementia.

History of cannabis smoking

Participants were asked at study enrollment: “Have you ever smoked marijuana?”, and for those who answered yes, they were further asked “How old were you the first time you smoked marijuana?”, “In total, how many years did you smoke marijuana?”, and “During the years that you smoked marijuana, on average how often did you smoke it?”. We defined women who reported that they had ever smoked cannabis as ever cannabis smokers. Based on the years of cannabis use and usage frequency, we further classified them into occasional cannabis smokers who smoked less than once per month or for less than 6 months versus regular cannabis smokers with at least once per month for 6 months or longer[23]. We also classified 1,477 women who reported < 1 year of use but “Don’t know” on use frequency as occasional smokers. Most women who smoked cannabis reported their first experience in their late teens or early adulthood, with only 12.4% reporting first use after age 40 years (online suppl. Table 1). In the analyses, we defined the age of first cannabis smoking as ≤ 20 years, 21–24 years, and ≥ 25 years. Further, few cannabis smokers (6.9%) reported more than 10 years of use and women who started cannabis smoking at an older age reported fewer years of use.

Subjective cognitive complaints

The AD8 is a brief screening tool to assess intraindividual changes in cognitive and functional symptoms that may indicate an increased risk of mild cognitive impairment or dementia[24, 25]. The measure contains 8 questions asking about changes in memory, orientation, judgment, and function over the past years[24]. The score is the sum of the items reported to have changed over time, with a range of 0-8 points. The AD8 was developed to differentiate cognitively normal older adults from those with mild cognitive impairment or very mild dementia[24], and it has been widely used for SCC screening in population-based epidemiological studies[18, 26, 21]. We chose the commonly used cutoff of ≥ 2 to define SCC[27, 26, 28, 21, 29].

Covariates

We considered the following covariates at the 2^nd follow-up primarily based on the literature[30, 31]: age, race and ethnicity (non-Hispanic White, non-Hispanic Black, and other, including American Indian or Alaska Native, Asian, Native Hawaiian or Other Pacific Islander, and all those of Hispanic ethnicity regardless of race), education (no college, some college to Bachelor’s degree, and Master’s degree or higher), marital status (married or living as married versus not married, including never married, widowed, divorced, and separated), household income (< $50,000, $50,000–$99,999, and ≥ $100,000), body mass index (BMI: kg/m², < 25.0, 25.0–29.9, and ≥ 30), smoking status (never, former, and current), alcohol use (never/former versus current), physical activity (hours of all physical activity per week, in quartiles), self-reported health status (excellent or very good, good, and fair or poor), family history of Alzheimer’ disease’s (no versus yes), and chronic diseases (never, prevalent, and diagnosed but timing unknown), including hypertension, diabetes, depression, lung diseases, and cardiovascular diseases.

Statistical analysis

We used Sister Study Data Release 10.1 for the current study. In the descriptive analysis of population characteristics at baseline (i.e., the 2^nd follow-up), we calculated mean ± standard deviation (SD) for continuous variables and numbers and percentages for categorical variables. We used t-test, one-way analysis of variance (ANOVA), and χ² test to assess differences in baseline characteristics across groups, as appropriate. Participants were followed from baseline to their date of death, last follow-up, or the 4^th follow-up, whichever came first. Compared to those who completed AD8 at follow-ups, participants who were lost to follow-up or died were more likely to be older, of racial/ethnic groups other than Non-Hispanic White, and unmarried or not living as married, to have a lower education level, and more likely to report poorer health and chronic diseases (online suppl. Table 2). To account for potential bias due to dropout or death, we conducted joint model analyses to assess the longitudinal associations between cannabis smoking and SCC[32]. The principle of the joint model is to define two sub-models, a longitudinal sub-model for the repeated measurements and a survival sub-model for time-to-event outcomes, which are linked using a common latent structure[33].

Specifically, we first fitted a generalized linear mixed model to assess the association between cannabis smoking and SCC over the course of the follow-up. In this mixed-effect sub-model, we included an interaction term between the cannabis smoking variable and time to account for the individual-specific variation in the SCC over time, incorporating a subject-specific random intercept. Then, we fitted a survival sub-model for the time to dropout or death to account for the competing risks after adjusting for the same covariates as the longitudinal sub-model. The corresponding joint models were fitted using R package “JMbayes”[34]. Based on the jointly fitted mixed sub-model, we calculated the odds ratios (ORs) and 95% confidence intervals (CIs) for SCC associated with cannabis smoking over the years of follow-up, with never smokers as the reference. The OR estimated the association between cannabis smoking and SCC later in life by comparing the odds of having SCC versus not between cannabis smokers and never smokers.

In the joint model analyses, we adjusted for three sets of covariates, first demographics of age, race, education, and marital status (model 1), then adding smoking, alcohol drinking, BMI, and physical activity (model 2), and finally further for self-reports of family history of Alzheimer’s disease, overall health status, and afore-mentioned chronic diseases (model 3). Although few study participants reported a clinical diagnosis of dementia or Parkinson’s disease, we conducted a sensitivity analysis that excluded them. To further assess whether cigarette smoking and alcohol drinking might have confounded the association, we conducted additional sensitivity analyses by adjusting for pack-years of cigarette smoking and drinks per week for current alcohol drinkers. Finally, we conducted subgroup analyses by race and ethnicity (non-Hispanic White versus all others) and by U.S. Census region (Northeast, Midwest, South, and West). In all statistical testing, we used a two-tailed α of 0.05. We fitted joint models using R 4.1.3 (The R Foundation for Statistical Computing) and conducted all other statistical analyses using SAS 9.4 (SAS Institute Inc., Cary, NC, U.S.).

Results

Of the 15,378 eligible participants, 11,405 (74.2%) did not report any cannabis smoking. Of the 3,973 who ever smoked, 2,994 (75.4%) smoked cannabis occasionally, and 979 (24.6%) smoked it regularly. Compared with those who never smoked cannabis, ever smokers were younger, and they were more likely to be from the West and Northeast regions of the U.S., not married, had higher educational attainment, household income, ever smoked tobacco, and were currently drinking alcohol, and were less likely to be overweight or obese (Table 1). Women who smoked cannabis were generally comparable to never smokers in reporting their overall health and the diagnostic history of most chronic diseases, except for more reports of depression and lung diseases and fewer reports of hypertension. Further separation of regular from occasional use showed consistent patterns (online suppl. Table 3).

Table 1.

Characteristics of study participants by cannabis smoking status in the Sister Study ^a

Population characteristics	Never use (N=11,405)	Ever use of cannabis (N=3,973)	P value
Age, mean (SD), years	71.4 (4.2)	69.4 (3.6)	< 0.01
65–69 years, n (%)	4,701 (41.2)	2,497 (62.8)
70–74 years, n (%)	4,143 (36.3)	1,108 (27.9)
≥ 75 years, n (%)	2,561 (22.5)	368 (9.3)
Census region, n (%)			< 0.01
Northeast	1,820 (16.0)	811 (20.4)
Midwest	3,259 (28.6)	741 (18.6)
South	3,781 (33.2)	1,151 (29.0)
West	2,341 (20.5)	1,258 (31.7)
Puerto Rico	199 (1.7)	10 (0.3)
Missing	5 (0.04)	2 (0.1)
Race and ethnicity, n (%)			< 0.01
Non-Hispanic White	10,178 (89.2)	3,556 (89.5)
Non-Hispanic Black	559 (4.9)	248 (6.2)
Others b	668 (5.9)	169 (4.3)
Education, n (%)			< 0.01
No college	2,356 (20.7)	443 (11.2)
Some college to Bachelor’s degree	6,540 (57.3)	2,160 (54.4)
Master’s degree or higher	2,509 (22.0)	1,370 (34.5)
Marital status, n (%)			< 0.01
Married or living as married	7,577 (64.4)	2,369 (59.6)
Not married c	3,828 (35.6)	1,604 (40.4)
Household income, n (%)			< 0.01
< $50,000	4,807 (42.1)	1,352 (34.0)
$50,000–$99,999	4,256 (37.3)	1,556 (39.2)
≥ $100,000	2,068 (18.1)	1,027 (25.8)
Missing	274 (2.4)	38 (1.0)
BMI, n (%)			< 0.01
< 25.0 kg/m2	4,395 (38.5)	1,756 (44.2)
25.0–29.9 kg/m2	4,030 (35.4)	1,272 (32.0)
≥ 30 kg/m2	2,980 (26.1)	945 (23.8)
Smoking status, n (%)			< 0.01
Never	6,741 (59.1)	1,180 (29.7)
Former	4,443 (39.0)	2,561 (64.5)
Current	221 (1.9)	232 (5.8)
Alcohol use, n (%)			< 0.01
Never/former	3,752 (32.9)	724 (18.2)
Current	7,653 (67.1)	3,249 (81.8)
Physical activity, n (%)			< 0.01
Quartile 1	2,937 (25.7)	1,001 (25.2)
Quartile 2	2,706 (23.7)	997 (25.1)
Quartile 3	2,641 (23.2)	1,004 (25.3)
Quartile 4	3,121 (27.4)	971 (24.4)
Overall health status, n (%)			0.21
Excellent or very good	7,193 (63.1)	2,578 (64.9)
Good	2,943 (25.8)	965 (24.3)
Fair or poor	1,000 (8.8)	342 (8.6)
Missing	269 (2.4)	88 (2.2)
Family history of Alzheimer’s disease, n (%)			0.33
No	9,413 (82.5)	3,313 (83.4)
Yes	1,705 (14.9)	556 (14.0)
Missing	287 (2.5)	104 (2.6)
Hypertension, n (%)			< 0.01
Never	4,646 (40.7)	1,942 (48.9)
Prevalent at the 2nd follow-up	5,972 (52.4)	1,801 (45.3)
Diagnosed but timing unknown	787 (6.9)	230 (5.8)
Diabetes, n (%)			0.01
Never	9,767 (85.6)	3,478 (87.5)
Prevalent at the 2nd follow-up	1,379 (12.1)	428 (10.8)
Diagnosed but timing unknown	259 (2.3)	67 (1.7)
Depression, n (%)			< 0.01
Never	8,089 (70.9)	2,335 (58.8)
Prevalent at the 2nd follow-up	2,208 (19.4)	1,093 (27.5)
Diagnosed but timing unknown	1,108 (9.7)	545 (13.7)
Lung diseases d, n (%)			< 0.01
Never	8,156 (71.5)	2,633 (66.3)
Prevalent at the 2nd follow-up	2,603 (22.8)	1,091 (27.5)
Diagnosed but timing unknown	646 (5.7)	249 (6.3)
Cardiovascular diseases e, n (%)			0.02
Never	9,084 (79.6)	3,237 (81.5)
Prevalent at the 2nd follow-up	1,428 (12.5)	431 (10.8)
Diagnosed but timing unknown	893 (7.8)	305 (7.7)

BMI, body mass index; SD, standard deviation. ^a Cannabis smoking was assessed at study enrollment; all study participant characteristics were up to the 2^nd follow-up, except for census region, race and ethnicity, and education, which were assessed at study enrollment. ^b Others included American Indian or Alaska Native, Asian, Native Hawaiian or Other Pacific Islander, and all those of Hispanic ethnicity, regardless of race. ^c Not married included never married, widowed, divorced, and separated. ^d Lung diseases included chronic obstructive pulmonary disease and asthma. ^e Cardiovascular diseases included stroke, transient ischemic attack, heart failure, angina, chronic ischemic heart disease, other acute ischemic heart disease, and myocardial infarction.

The raw data on the prevalence of SCC showed higher prevalence among cannabis smokers at each follow-up, most evidently for regular smokers (online suppl. Table 4). Joint model analyses showed that cannabis smoking was significantly associated with the odds of having SCC throughout the follow-up (Fig. 2). The adjusted ORs were 1.27 (95% CI: 1.13, 1.43) at the 2^nd follow-up, 1.28 (95% CI: 1.14, 1.44) at the median time of the 3^rd follow-up, and 1.30 (95% CI: 1.11, 1.52) at the median time of the 4^th follow-up (Fig. 2A). The corresponding β coefficients and standard errors were 0.240 ± 0.059, 0.250 ± 0.059, and 0.263 ± 0.079, respectively. The association appears to be stronger for regular cannabis smokers than occasional users (Fig. 2B). For example, at the 2^nd follow-up, compared with never cannabis smokers, the ORs were 1.61 (95% CI: 1.31, 1.98) for regular smokers versus 1.19 (95% CI: 1.04, 1.35) for occasional smokers, corresponding to β coefficients of 0.478 ± 0.105 and 0.171 ± 0.066, respectively. Notably, the magnitudes of the ORs remained stable throughout follow-up. As for age of first cannabis smoking, we found the largest ORs for participants who reported first use at age 21–24 years, followed by those younger than 20 and those older than 25; however, their 95% CIs overlapped substantially (online suppl. Fig. 1). Detailed coefficient estimates derived from the joint model analyses are provided in online supplemental Tables 5 and 6. Using marginal standardization based on the joint models, the marginal risk of SCC was modestly higher in ever cannabis smokers than never smokers (online suppl. Fig. 2). Sensitivity analyses showed similar results after excluding self-reported cases of dementia and Parkinson’s disease (online suppl. Fig. 3) and when adjusting for quantitative measures of smoking and alcohol drinking (online suppl. Fig. 4).

Fig. 2.

Adjusted odds ratio (OR) and 95% confidence interval (CI) of subjective cognitive complaints by cannabis smoking status and years of follow-up. (Panel A) Ever versus never cannabis smoking; (panel B) Regular, occasional, versus never use of cannabis.

We further examined the association of ever cannabis smoking and SCC by subgroups of race and ethnicity and census region. A statistically significant association was evident among non-Hispanic whites (Fig. 3A) but not among the combined group of women with other races and ethnicities (Fig. 3B). Notably, the sample size for other races and ethnicities was small, and the composition was heterogeneous (online suppl. Table 7). Among women of other races and ethnicities, the percentages of those using cannabis were notably different. In the subgroup analysis by census region, the association was most pronounced for women from the West region of the U.S. (online suppl. Fig. 5).

Fig. 3.

Subgroup analyses of cannabis smoking and subjective cognitive complaints by races and ethnicities. (Panel A) Non-Hispanic White; (panel B) All other races and ethnicities. CI, confidence interval; OR, odds ratio.

Discussion

With increasing legalization and societal acceptance, cannabis use in the U.S. has been rising over the past decades. In 2021, more than 52.5 million people aged 12 or older (18.7%) used cannabis at least once in the past year, and the prevalence of cannabis use is highest in adolescence and early adulthood[35]. Further, the potency of cannabis has increased by more than threefold over the past two decades[36, 37], making it imperative to assess the potential health consequences of cannabis use in U.S. populations.

In this nationwide cohort of older U.S. women, nearly all women who had ever smoked cannabis reported their first use before age 40 years. Although we did not specifically assess whether their use was for recreational or medical purposes, we speculated that cannabis smoking reported at study enrollment was predominantly for recreational purposes, given individuals’ age of first use and the fact that smoking is the most common way of recreational cannabis use. We found that cannabis smoking was associated with a higher risk for SCC in later adulthood. The association was more pronounced among women with a history of regular use than those who smoked cannabis occasionally. Moreover, this association remained relatively stable throughout the follow-up period. Our findings raise the concern that regular cannabis use in early adulthood may be associated with cognitive decline later in life and predispose individuals to a higher risk of dementia. Given the increasing use of cannabis in the U.S., our findings, if confirmed, may have significant public health and policy implications.

To date, biological and mechanistic evidence on the neurocognitive consequences of cannabis use has focused mainly on developmental or maturing brains[4, 38, 5, 6, 39, 7]. Animal experimental studies found that exposure to cannabinoids in adolescence induced structural brain changes and compromised synaptic plasticity, affecting the fronto-limbic systems crucial for higher brain functions[40-43]. These changes may be the result of an overactivated endocannabinoid system, which plays a pivotal role in brain maturation[40]. During adolescence, overactivation of the endocannabinoid system by delta-9-tetrahydrocannabinol from cannabinoids may change brain neuroplasticity, disrupt brain maturation, and induce neurocognitive changes[42, 44]. Therefore, animal experimental evidence implicates adverse effects of exposure to cannabinoids on neurodevelopment.

Human studies on cannabis use and cognitive functioning provided mixed but generally suggestive evidence of harm in adolescents and young adults. A recent meta-analysis of 69 cross-sectional studies showed that cannabis use was associated with modestly impaired performance on learning and delayed memory, executive functioning, speed of processing, and attention during adolescence and early adulthood; however, these associations tended to diminish following sustained abstinence[45]. Evidence from longitudinal studies supports an association between regular cannabis use and poor cognitive performance in this age group[10, 46-49]. Extending into later life, several cohort studies further found that habitual or long-term cannabis use was associated with poor cognitive performance in mid-adulthood[13, 50, 51]. However, there are important inconsistencies. For example, in twin studies, cannabis use in adolescence was not associated with cognitive function in early adulthood[52-54]. Notably, all studies discussed above assessed cognition in early-to-mid adulthood when human cognitive health remains resilient. As such, findings may not be readily extended to late adulthood when human brains are more susceptible to degeneration.

In older adults, cannabis has also been increasingly used over the past decades for medical and other reasons[55, 56]. Although findings are not consistent, medical benefits of cannabis have been proposed for alleviating symptoms of pain, sleep disturbances, depression and anxiety, and motor dysfunction of patients with various diseases[57, 58]. Empirical data on cannabis use and cognition in older adults are limited but possible adverse effects have been suggested. A cross-sectional study using the UK Biobank showed that cannabis use, particularly longer duration and frequent use, may be related to smaller grey and white matter volumes in older adults[59]. Three large U.S. studies have examined the cross-sectional association between cannabis use in the past year and cognitive deficits in adults over age 50 years[60-62]. In the Health and Retirement Study, frequent cannabis use in the past year was associated with short-term memory deficit[61]. In the National Survey of Drug Use and Health, individuals who used cannabis in the past year were 38% more likely to report subjective memory complaints than those who did not use[62]. Finally, the National Epidemiologic Survey on Alcohol and Related Conditions-III reported a modest association of past-year cannabis use with SCC[60]. Therefore, emerging data extend the concerns about the potential adverse effects of cannabis use on cognition to middle-aged and older adults; however, the evidence to date is overwhelmingly cross-sectional. To our knowledge, our study is the first to investigate whether cannabis smoking in early adulthood, when use is most common for recreational purposes, may have adverse cognitive consequences in late life, when the human brain is susceptible to degeneration. Our findings, while preliminary, clearly support the possibility that cannabis smoking in early adulthood is associated with a higher risk of cognitive complaints later in life.

Strengths of this study included the large nationwide cohort, the prospective study design, repeated AD8 assessments, and the comprehensive approach to controlling for missing data to minimize bias due to dropout and death. The present study also has limitations. First, we used the short AD8 screener for SCC without any cognitive testing. Future studies need to confirm our findings and further evaluate cannabis use and cognitive outcomes using objective neuropsychological testing and clinical evaluations. Second, information on cannabis smoking was self-reported, consequently the information is subject to recall and reporting errors. Further, cannabis use was assessed before the first legalization of recreational cannabis in any of the U.S. states in 2012; participants might have underreported their cannabis smoking. Bias may arise if such reporting errors are correlated with AD8 assessments in subsequent follow-ups. Third, we only assessed cannabis smoking but not other types of cannabis use, and our exposure assessment did not differentiate recreational from medical use. Therefore, our findings may not be readily generalizable to other types and purposes of cannabis use. Importantly, the increasing legalization of cannabis use has also brought about changes in consumption patterns and user behaviors, and the potency of cannabis has been increasing over the years. All of these should be considered and thoroughly investigated in future health research. Finally, participants of the Sister Study are predominantly non-Hispanic white (89.3%) and college-educated (81.8%) women. The results may not be readily generalizable to men, women of other races and ethnicities, or individuals with lower educational levels.

In conclusion, our findings suggested that cannabis smoking in early adulthood may be associated with a higher risk of SCC in late adulthood in this nationwide large cohort of U.S. women 65 years or older. These findings, if confirmed, will have medical and public health implications, highlighting the need for further research.

Data Availability Statement

Data used in the present analysis are available from the National Institute of Environmental Health Sciences’ Sister Study. To access the data, investigators should submit an analytic proposal for approval (https://sisterstudy.niehs.nih.gov/English/coll-data.htm).