use and Mortality
Sidney, Jerome E. Beck, Irene Tekawa, Charles Quesenberry, and Gary Friedman
Stephen Sidney, Irene S. Tekawa, Charles P. Quesenberry, Jr, and Gary D. Friedman
are with the Division of Research, Kaiser Permanente Medical Care Program (Northern
California Region), Oakland, Calif. Jerome E. Beck is with the School of Public
Health, University of California, Berkeley.
American Journal of Public Health, April 1997, Volume 87 Issue 4, p585
The purpose of this study was to examine the relationship of marijuana use to
The study population comprised 65 171 Kaiser Permanente Medical Care Program enrollees,
aged 15 through 49 years, who completed questionnaires about smoking habits, including
marijuana use, between 1979 and 1985. Mortality follow-up was conducted through
Compared with nonuse or experimentation (lifetime use six or fewer times), current
marijuana use was not associated with a significantly increased risk of non-acquired
immunodeficiency syndrome (AIDS) mortality in men (relative risk [RR] = 1.12,
95% confidence interval [CI] = 0.89, 1.39) or of total mortality in women (RR
= 1.09, 95% CI = 0.80, 1.48). Current marijuana use was associated with increased
risk of AIDS mortality in men (RR = 1.90, 95% CI = 1.33, 2.73), an association
that probably was not causal but most likely represented uncontrolled confounding
by male homosexual behavior. This interpretation was supported by the lack of
association of marijuana use with AIDS mortality in men from a Kaiser Permanente
AIDS database. Relative risks for ever use of marijuana were similar.
Marijuana use in a prepaid health care-based study cohort had little effect on
non-AIDS mortality in men and on total mortality in women. (Am J Public Health.
is the most commonly used illegal drug in the United States. Over 65 million Americans
(31% of the US population aged 12 and older) are estimated to have used marijuana(n1);
its mean retail sales value in the United States is approximately $10 billion.(n2)
Despite its longstanding popularity and increasing use among youth in recent years,(n1,
n3) we still know little about long-term health risks associated with marijuana
use. Harvard policy analyst Mark Kleiman recently concluded that "aside from
the almost self-evident proposition that smoking anything is probably bad for
the lungs, the quarter century since large numbers of Americans began to use marijuana
has produced remarkably little laboratory or epidemiological evidence of serious
health damage done by the drug."(n4 (p253)) Similar appraisals of the health
effects of cannabis were offered in the two most comprehensive reviews from the
1980s.(n5, n6) More currently, Hall and coauthors concluded that while there are
no well-established health or psychological effects of chronic cannabis use, the
following were considered to be probable major adverse effects: respiratory diseases
associated with smoking as the method of administration, including chronic bronchitis
and premalignant histopathological changes in the lung; development of a cannabis
dependence syndrome; and subtle forms of cognitive impairment.(n7(p16))
only other large-scale study of marijuana use and mortality was performed in a
cohort of 45 540 male Swedish conscripts, aged 18 through 20 years at baseline
and followed for 15 years.(n8) In this study, the relative risk (RR) for mortality
associated with marijuana use (more than 50 times) was 1.2 (95% confidence interval
[CI] = 0.8, 1.9) after adjustment for social background.
report here the findings of a study of the relationship of marijuana use to mortality
in a cohort of over 65 000 members of a large prepaid health plan. Data on marijuana
use in this cohort were collected before the "war on drugs" escalated
in the latter half of the 1980s, which may have resulted in underreporting of
illegal drug use.(n9) Mortality is one of several health outcomes being studied;
other endpoints include cancer incidence and outpatient utilization for respiratory
illnesses and injuries. We hypothesized that marijuana use would be associated
with increased risk of respiratory disease and injury.
cohort of 65 171 men and women aged 15 through 49 years (mean age, 33 years) completed
detailed self-administered research questionnaires on tobacco, marijuana, and
alcohol use from mid-1979 through 1985. The subjects were undergoing multiphasic
health checkups in the San Francisco (until 1980) and Oakland Kaiser Permanente
facilities. Mortality was followed through December 31, 1991, for a mean length
of 10.0 years.
marijuana smoking was defined by admission to smoking currently and more than
six times ever. Former marijuana smoking was defined by denial of current smoking
but admission to having smoked more than six times ever. Nonsmoking was defined
as never having smoked. Experimenters were defined as those admitting to having
ever smoked from one through six times. Ever users included current and former
users but excluded experimenters. Smoking duration was expressed as total years
of use. Smoking frequency was expressed as less than once per month, once or twice
per month, once or twice per week, and daily or almost daily.
were classified as current, former, or never smokers and users of alcoholic beverages
on the basis of their questionnaire responses.(n10, n11) Current and former smokers
were categorized by frequency (number of cigarettes per day) and duration (years)
of smoking. Current alcohol users were categorized by usual numbers of drinks
consumed per day.
was ascertained through 1991 by computer-matching study cohort members with the
Kaiser Permanente Medical Care Program membership file as of 1992 and extracting
a list of subjects who were no longer members. From this list we accepted as confirmed
deaths those ascertained in previous research studies. The mortality status of
the remaining study subjects who were no longer members was ascertained by computer-matching
names and other demographic data with the California death file, using the California
Automated Mortality Linkage and Information System (CAMLIS).(n12) Death certificate--specified
underlying causes of death (International Classification of Diseases, 9th rev.--ICD-9)
were used for coding. Centers for Disease Control and Prevention criteria were
used to code acquired immunodeficiency syndrome (AIDS) prior to the introduction
of specific disease codes associated with the human immunodeficiency virus (HIV)
of subjects who left California were investigated by linking pertinent Social
Security numbers to a Pension Benefits Information (Tiburon, Calif) database that
included mortality data from the state of California Center for Health Statistics,
the Social Security Administration, the Department of Defense, the Civil Service
Commission, and the Railroad Retirement Board. While Social Security numbers were
available for about two thirds of the study cohort, only 27 of 1215 deaths (2.2%)
were ascertained by out-of-state search. Since causes of death were unavailable
for out-of-state deaths, these deaths were included in analyses of total mortality
but excluded in subcategory mortality analyses.
overall age-specific mortality rates of this group were about three quarters as
large as the corresponding 1987 United States rates,(n14) a discrepancy we attribute
to the probable better health and predominantly employed status of our insured
population and to our inability to ascertain mortality in subjects without Social
Security numbers who had left California.
programs were used for statistical analyses.(n15) Cox proportional hazards models
were used to examine the joint effect of sociodemographic characteristics and
use of marijuana, tobacco, and alcohol on mortality risk; estimates of relative
risks and associated 95% confidence intervals were obtained from these models.(n16)
Age-squared terms were entered into Cox proportional hazards models to determine
whether there was a nonlinear relationship between age and mortality and were
included when significant. Interactions between marijuana and tobacco use and
between marijuana and alcohol use were tested in the selected models (total mortality,
AIDS mortality [men only], non-AIDS mortality [men only], and mortality from injuries/poisonings)
by including cross-product terms in our proportional hazards models. None of the
interactions were statistically significant (P < .05).
characteristics of the sample are shown in Table 1.
cohort consisted of 38% nonusers, 20% experimenters, 20% former users, and 22%
current users. The percentage of ever users was highest in the 20- through 29-year-old
age group. Ever use of marijuana was more common among men than among women and
was highest among Whites. Never-married men and women were about twice as likely
to be ever users as their married counterparts. Sociodemographic patterns were
generally similar for current marijuana use.
marijuana users were twice as likely as never users to be current tobacco cigarette
smokers and nearly 2.5 times as likely to be alcohol drinkers. The percentage
of current smokers was 21% for never marijuana users, 31% for experimenters, 32%
for former users, and 42% for current users. The corresponding percentages of
those consuming one or more drink per day were 16%, 27%, 31%, and 39%. While few
marijuana users were nonusers of alcohol, a substantial proportion of ever marijuana
users (25% of men, 30% of women) and current marijuana users (22% of men, 28%
of women) were nonsmokers of tobacco cigarettes and occasional (less than once
per month) drinkers.
compared risks of mortality associated with ever and current use relative to never
or experimental use of marijuana. There were 807 deaths among men and 408 deaths
among women in this cohort. We performed analyses for total mortality, AIDS (men
only), neoplasms, circulatory disease, injury or poisoning, "other causes"
of mortality, and total non-AIDS mortality (men only) (Table 2).
Use in Relation to Mortality
men, ever use of marijuana was associated with a significantly increased risk
of total mortality (28%) and AIDS mortality (80%) and a nonsignificant (P >
.05) increase (11%) in risk of non-AIDS mortality. Relative risks associated with
ever marijuana use for these mortality categories were similar or higher in nonsmokers/occasional
drinkers (a group in which marijuana use could be evaluated without uncontrolled
confounding by cigarette and substantial alcohol use). Of note was the nearly
significant 47% increase in "other causes" of mortality, examination
of which revealed higher proportions of deaths from infectious diseases and from
alcohol and drug abuse in ever users than in never users/experimenters. Current
marijuana use was also associated with a significantly increased risk in men of
total mortality (33%) and AIDS mortality (90%).
women, there were no significant increases or decreases in mortality risk associated
with ever or current marijuana use. Current use was associated with a nearly significant
86% increase in mortality from injury or poisoning, which could not be attributed
to any specific category of injury.
risks associated with marijuana use among nonsmokers/occasional drinkers were
generally similar to those for the complete cohort, suggesting that increased
risks in the complete cohort were not an artifact resulting from incomplete control
of the effects of cigarette smoking or alcohol use. The results of an analysis
of mortality excluding subjects who died within the first 5 years of follow-up
(data not shown) were similar to the overall results shown in Table 2, suggesting
that the overall results were uncompromised by the possibility that serious illness
occurring before the multiphasic health checkup affected subjects' decision to
of use in current marijuana users was not consistently related to the risk of
AIDS mortality in men or to total mortality in women, and had an inverse tendency
in relationship to total and non-AIDS mortality in men (data not shown). A continuous
duration-of-use variable was not significant when added to the full models for
each mortality outcome.
use at least once a week was associated with slightly higher relative risks of
mortality than less frequent use. The addition of frequency of use improved the
fit of the model (P < .05) only for total mortality in men (RR = 1.25, 95%
CI = 0.97, 1.62, for total mortality among those who used less than once a week
and RR = 1.46, 95% CI = 1.19, 1.79, among those who used at least once a week,
relative to nonusers/experimenters).
vast majority of AIDS deaths (172/207 = 83%) occurred among never married men.
Current marijuana use was nearly twice as high in never married as in married
men (Table 1), raising the question of whether analytic control for marital status
was insufficient to adjust for confounding lifestyle factors, particularly male
address this question, the study cohort was linked to the Northern California
Kaiser Permanente Medical Care Program AIDS Database, which revealed 214 men with
a diagnosis of AIDS after determination of their marijuana use status. The prevalence
of current marijuana use at the time of the checkup (56%) in these AIDS patients
was substantially higher than the prevalence in unmarried men in the total study
these 214 AIDS patients, current marijuana use was associated with a nonsignificant
decrease in relative risk for total mortality (RR = 0.78, 95% CI = 0.47, 1.30)
and for AIDS mortality (RR = 0.71, 95% CI = 0.41, 1.23). Assuming that most of
the unmarried men who developed AIDS were homosexual or bisexual, these findings
supported the hypothesis that the prevalence of marijuana use was higher in homosexual
and bisexual men in the cohort, a group at high risk for AIDS mortality. Therefore,
male homosexual behavior, a critical confounding variable, could not be controlled
for in complete cohort mortality analyses.
Risks of Tobacco, Alcohol, and Marijuana Use
relative risks of total mortality in men and women, and of AIDS and non-AIDS mortality
in men, associated with current cigarette smoking, consumption of three or more
drinks per day, and current marijuana use are shown in Table 3
Except for AIDS mortality, the risks associated with marijuana use were lower
than those for tobacco cigarette smoking. Compared with consumption of three or
more drinks per day, marijuana use was associated with a higher risk of total
mortality and AIDS mortality in men and a lower risk of total mortality in women.
main overall findings were an increased risk of total mortality associated with
marijuana use in men but not in women. The increased risk of total mortality in
men was explained by the strong relationship between marijuana use and AIDS mortality.
Marijuana use was unassociated with non-AIDS mortality in men.
question of the effect of marijuana use on AIDS mortality is an important one.
Marijuana use has been advocated as a therapeutic adjunct to ameliorate the nausea
and loss of appetite commonly associated with the wasting syndrome in AIDS.(n17)
We have provided substantial evidence that the increased risk of AIDS mortality
in the total study cohort probably resulted from uncontrolled confounding by homosexual
behavior. Other studies have reported a substantially higher prevalence of marijuana
use in homosexual and bisexual men, supporting the hypothesis that marijuana use
is a marker for homosexuality or bisexuality.(n18-n20)
are several other potential explanations for the increased risk of AIDS in marijuana
users. Marijuana smoking might theoretically place AIDS patients at increased
risk of infection because of its irritative effects on the respiratory system
or because of infectious contaminants (e.g., fungi) in marijuana. Other potential
explanations include marijuana as a marker of high-risk sexual behavior or intravenous
drug use; initiation of marijuana use as a result of having HIV or AIDS, rather
than preceding the disease; and possible immunosuppressive properties of marijuana.
use of alcohol and nonmedical psychoactive drugs, including marijuana, is associated
with risky sexual behavior such as unprotected intercourse,(n20) but methodological
limitations have made it impossible to determine causality.(n21) Marijuana use
may serve to a certain extent as a marker of intravenous drug use. However, the
relative risk of AIDS mortality associated with marijuana use did not diminish
when the analysis was limited to men who were nonsmokers of tobacco and occasional
alcohol drinkers, a subgroup unlikely to contain many parenteral drug users. Additional
evidence against marijuana as a marker for parenteral drug use was the finding
of only one case of infective endocarditis in Kaiser Permanente hospitalization
records of the AIDS decedents.
lack of increased mortality during the first 5 years of follow-up suggests that
therapeutic use of marijuana at baseline for AIDS-related symptoms has little,
if any, explanatory effect on the association between marijuana use and AIDS.
Furthermore, the majority of AIDS patients initiated marijuana use long before
the onset of clinical disease; nearly two thirds (65%) of AIDS patients reported
initiation before 1976, when HIV infection in the San Francisco Bay area was either
nonexistent or negligible.(n22)
marijuana and its psychoactive cannabanoids possess known immunosuppressive qualities,
there is no consensus as to whether typical doses result in clinical immunosuppression
in humans.(n23) Marijuana use has been associated with a higher prevalence of
seropositivity for HIV in some cross-sectional studies of homosexual and bisexual
men,(n20, n24) but it has not been shown to be an independent predictor of seroconversion,(n25)
nor does it increase the risk of AIDS in seropositive men.(n24)
nearly significant increase in mortality risk from injury or poisoning for female
current marijuana users was consistent with our hypothesis that marijuana use
is a risk factor for death due to injury. Marijuana is known to decrease psychomotor
performance; some studies have implicated its use in motor vehicle crashes.(n7
(pp43-50)) Marijuana use is also strongly associated with alcohol use, another
major risk for accidental death. There were too few deaths to meaningfully study
the other main hypothesis, that marijuana use would be associated with increased
respiratory disease mortality. Another study performed on a subgroup of this cohort
showed that daily or near-daily marijuana users who were not tobacco cigarette
smokers had a 19% higher risk of outpatient visits for respiratory disorders than
nonusers of both substances.(n26)
major limitations of this study include its reliance on self-report for ascertainment
of marijuana use status; the inability to study changes in marijuana use status
during follow-up; a lack of lengthy follow-up into the geriatric age range (maximum
follow-up, 12.5 years; maximum age reached, 63 years); a lack of information regarding
other illegal drug use; and potential underascertainment of mortality (noted earlier).
Estimates of marijuana use were similar to those obtained during this period by
the National Household Survey on Drug Abuse, the most authoritative source of
illegal drug use information for US adults.(n27) The lack of longitudinal data
regarding use status is common to many cohort studies. It seems unlikely that
"ever" marijuana use status would have changed substantially over time,
because relatively few adults in this cohort are likely to have initiated marijuana
use during follow-up in a period (the 1980s) when there was a marked secular decline
in self-reported marijuana use in the United States.(n1) It is possible that relationships
between marijuana use and mortality might be found with longer-term follow-up
or later in life. It is likely that if information on subjects' use of other illegal
drugs had been available, adjustment for other drug use would have lowered the
relative risk: estimates for marijuana use.
noted earlier, relatively few adverse clinical health effects from the chronic
use of marijuana have been documented in humans.(n7 (p16)) The criminalization
of marijuana use may itself be a health hazard, since it may expose the consumer
to violence and criminal activity.(n28) While reducing the prevalence of drug
abuse is a laudable goal, we must recognize that marijuana use is widespread despite
the long-term, multibillion dollar War on Drugs. Therefore, medical guidelines
regarding its prudent use should be established, akin to the commonsense guidelines
that apply to alcohol use. Unfortunately, clinical research on potential therapeutic
uses for marijuana has been difficult to accomplish in the United States, despite
reasonable evidence for the efficacy of tetrahydrocannabinol (THC) and marijuana
as antiemetic and antiglaucoma agents and the suggestive evidence for their efficacy
in the treatment of other medical conditions, including AIDS.(n7 (pp185-262))
summary, this study showed little, if any, effect of marijuana use on non-AIDS
mortality in men and on total mortality in women. The increased risk of AIDS mortality
in male marijuana users probably did not reflect a causal relationship, but most
likely represented uncontrolled confounding by male homosexual behavior. The risk
of mortality associated with marijuana use was lower than that associated with
tobacco cigarette smoking.
research was supported by grant R01 DA06609 from the National Institute on Drug
Abuse. Dr Friedman is supported in part by grant R35 CA49761 from the National
Cancer Institute. The collection of data on alcohol use was supported by a grant
from The Alcoholic Beverage Medical Research Foundation (Baltimore, Md).
authors acknowledge Christianna Williams, Steve Wilson, and Marianne Sadler for
computer programming and Leo Hurley for consultation regarding the use of the
Kaiser Permanente AIDS database.
Preliminary Estimates from the 1994 National Household Survey. Advance Report
No. 10. Rockville, Md: Substance Abuse and Mental Health Services Administrations;
What America's Users Spend on Illegal Drugs. Washington, DC: Office of National
Drug Control Policy;1991.
Johnson LD, O'Malley PM, Bachman JD. National Survey Results on Drug Use from
the Monitoring the Future Study, 1975-1994. Vol. 1. Rockville, Md: National Institute
on Drug Abuse; 1995. DHHS Publication NIH 95-4026.
Kleiman MAR. Against Excess: Drug Policy for Results. New York, NY: Basic Books;
Report of an ARF/WHO Scientific Meeting on the Adverse Health and Behavioral Consequences
of Cannabis Use. Toronto, Ont, Canada: Addiction Research Foundation; 1981.
National Academy of Science, Institute of Medicine. Marijuana and Health. Washington,
DC: National Academy Press; 1982.
Hall W, Solowij N, Lemon J. The Health and Psychological Consequences of Cannabis
Use. Canberra, Australia: Australian Government Publishing Services; 1994. National
Drug Strategy, Vol. 25
Andreasson S, Allebeck P. Cannabis and mortality among young men: a longitudinal
study of Swedish conscripts. Scand J Soc Med. 1990;18:9-15.
Sidney S. Evidence of discrepant data regarding trends in marijuana use and supply,
1985-1988. J. Psychoactive Drugs. 1990;22:319-324.
Friedman G, Sidney S, Polen M. Smoking habits among multiphasic examinees, 1979
to 1984. West J Med. 1986;145:651-656.
Klatsky AL, Armstrong MA, Friedman GD. Risk of cardiovascular mortality in alcohol
drinkers, ex-drinkers and nondrinkers. Am J Cardiol. 1990;66:1237-1242.
Arellano M, Peterson G, Petitti DB, Smith R. The California automated mortality
linkage system. Am J Public Health. 1984;74:1324-1330.
Vermund S. Rising HIV-related mortality in young Americans. JAMA. 1993;269: 3034-3035.
Health United States, 1989. Hyattsville, Md:National Center for Health Statistics;
1990. DHHS Publication PHS 90-1232.
SAS Users Guide: Basic Version. 5th ed. Cary, NC:SAS Institute Inc; 1985.
Cox DR, Oakes D. Analysis of Survival Data. New York, NY: Chapman & Hall;
Grinspoon L, Bakalar JB. Marijuana, the Forbidden Medicine. New Haven, Conn: Yale
University Press; 1993.
Stall R, Wiley J. A comparison of alcohol and drug use patterns of homosexual
and heterosexual men: the San Francisco Men's Health Study. Drug Alcohol Depend.
Skinner WF, Otis MD. Drug Use among Lesbian and Gay People: Design, Insights,
and Policy Issues from the Trilogy Project. In: Proceedings of the Research Symposium
on Alcohol and Other Drug Problem Prevention among Lesbians and Gay Men. Sacramento,
Calif: California Department of Alcohol and Drug Programs; October 1992:34-60.
Ostrow DG. Substance use and HIV-transmitting behaviors among gay and bisexual
men. In: Battjes RJ, Sloboda Z, Grace WC, eds. The Context of HIV Risk among Drug
Users and Their Sexual Partners. Rockville, Md: National Institute on Drug Abuse;
1994. NIDA Research Monograph 143.
Leigh BC, Stall R. Substance use and risky sexual behavior for exposure to HIV:
issues in methodology, interpretation, and prevention. Am Psychol. 1993;48:1035-1048.
Jaffe HW, Darrow WW, Echenberg DF, et al. The acquired immunodeficiency syndrome
in a cohort of homosexual men. Ann Intern Med. 1985;103:210-214.
Hollister LE. Marijuana and immunity. J Psychoactive Drugs. 1992;24(2):159-164.
Kaslow RA, Blackwelder WC, Ostrow DG, et al. No evidence for a role of alcohol
or other psychoactive drugs in accelerating immunodeficiency in HIV-1 positive
individuals: a report from the multicenter AIDS cohort study. JAMA. 1989;261:3424-3429.
Ostrow DG, Difranceisco WJ, Chmiel JS, Wagstaff DA, Wesch J. A case-control study
of human immunodeficiency virus type 1 seroconversion and risk-related behaviors
in the Chicago MACC/CCS cohort, 1984-1992. Am J Epidemiol. 1985; 142:875-883.
Polen MR, Sidney S, Tekawa IS, Sadler M, Friedman G. Health care use by frequent
marijuana user who do not smoke tobacco. West J Med. 1993;158:596-601.
National Household Survey on Drug Abuse: Main Findings 1985. Rockville, Md: National
Institute on Drug Abuse; 1985. DHHS publication ADM 88-1586.
Duke SB, Gross AC. America's Longest War: Rethinking Our Tragic Crusade against
Drugs. New York, NY: Tarcher/ Putnam; 1993:22-42.
for reprints should be sent to Stephen Sidney, MD, Division of Research, Kaiser
Permanente Medical Care Program, 3505 Broadway, Oakland, CA 94611.