Hansen: Meta-analysis of Adolescent Marijuana Use Studies

6 min read

Considerable studies have been conducted in search of variables which will explain drug abuse, and present markers of individuals especially at risk to drug dependency. The purpose of these studies is to identify variables that have high correlations with drug use. This process eliminates variables with low statistical correlations and identifies the few variables that are likely to explain drug use in a particular statistical sample. Several thousand of these correlations between various independent variables and school-age drug use have been quantified in hundreds of different social science research projects.

A review and summary of these thousands of correlations would be extremely valuable to analysts converting such basic research into policy or scheduling decisions. In the early 1990’s the Congressional Office of Technology Assessment contracted with a professor at the Bowman Gray School of Medicine at Wake Forest University for this sort of report on “Drug Abuse in Schools: Contributing Factors and Preventive Interventions.”(38) The report is described by OTA in a 1994 publication:

“OTA commissioned a review of the survey research literature on school-aged substance use that compiled, classified, and examined 9,930 statistical analyses from 242 separate studies. This is by far the most extensive systematic examination of this body of research conducted so far. Most of the studies dealt with school-based populations, but some focused on school-age army recruits, dropouts, children of alcoholics, and individuals involved in clinics. The studies reported statistical relationships between substance use and its postulated causes. Statistical findings from the study reports were sorted into 11 major categories and 50 subcategories, and then analyzed to identify strong, moderate, and weak statistical relationships, as well as those that had been insufficiently studied.”(39)

Use of tobacco, alcohol, and marijuana accounted for 82 percent of the analyses in the report, prepared by William B. Hans

“In part, the focus on use rather than abuse reflects much of the philosophy that has guided research and educational funding. Namely, for youth, any use of illegal substances (including alcohol and tobacco) has been considered abuse by some. Use has received the most extensive attention perhaps in part because use among youth is easier to define than abuse but there are also multiple methods for defining use. These include any use versus no use definitions as well as intensity, frequency, quantity and problematic measures. The point at which use becomes abusive escapes definition.

“At a minimum, use is a precursor to abuse although abuse does not necessarily follow from use . . . Nearly half of the youth who experiment with cigarettes go on to develop long-term smoking habits. The use of alcohol and marijuana by youth who use but have no definable chronic problems contribute to the highway death toll, violence, and crime. The public costs of these alone support the study and prevention of even low levels of use.

“This report addresses the causes of school-aged substance use as examined through the use of survey research methods.”(40)

As described by OTA the analyses were sorted into 50 subcategories, and the average correlations in each sub category were determined.

Four categories of evaluation were established. First, the most important variables are those with a relatively strong statistical relationship, and these should be used in explanatory models regarding experimentation and progression of use behaviors. The next category includes variables with moderate relationships, and these variables provide secondary explanations. The third category is variables with weak relationships.

“These are variables with no empirical support for playing a causal role in the onset of substance use or transition from experimentation to abuse.”(41)

Finally, there are some variables that have not been sufficiently studied.

How will the correlations of variables in different studies be compared?

“The average magnitude of correlations for each category will be examined. Previous research has suggested that correlations with absolute values above .300 have clear meaning and contribute important information to understanding the nature of such behaviors as substance use. On the other hand, correlations of less than .200 have minimal importance. Correlational values in between are of moderate importance.”(42)

Meta-analyses are difficult to interpret because of the possible variations in the manner individual studies define and measure their variables.(43) This paper reports unweighted mean correlation coefficients, and most of the variables produced consistent measures in the meta-analysis (standard deviations, for example).(44)

For the Hansen meta-analysis, primary correlations were defined as over .30, secondary correlations were between .20 and .30, and correlations under .20 were considered tertiary and of minimal importance.

“The four variables that dominate as correlates of and possible contributors to substance use are: 1) prior and concurrent use of substances, 2) substance use by peers and friends, 3) perceived peer attitudes about substance use, and 4) offers to use substances. The prominence of prior and concurrent use is consistent with the reinforcing nature of substance use itself. The prominence of the other three variables emphasizes the importance of the social environment in contributing to and reinforcing substance use among school-age youth.”(45)(emphasis added)

Section one above discusses contemporary standards for evaluating the reinforcing nature of a substance, which are based on self-administration of the drug in animal models. The evidence in this petition asserts that marijuana does not have a significant dependence liability. The Hansen report supports the assertion that it is not the drug, but the set and setting, which have the greatest influence on school-age use of marijuana. Consequently, negative effects of marijuana on adolescents can not be an indication or predictor of the substance’s effects on adult users.

“The first variable clearly speaks to the behavioral and neuropharmacological nature of substance use, suggesting that substance use is patterned behavior that is highly influenced by neurological reinforcement. The remaining three variables all suggest that a major cause of substance use among school-age youth is the social environment. That is, theories of substance use and abuse must include social processes as central explanations of the onset and development of patterns of consumption. Social norms (including issues about prevalence and acceptability), the process of social development, the ready availability of substances in the social environment, and the structure of social groups must all be considered as central elements for explaining substance use.”(46) (emphasis added)

The role of risk perception, for example, is integral to the social environment; it is a component of the “set” in drug, set, and setting. However as an explanatory variable for marijuana use, it is secondary to prior use of alcohol and tobacco.

Here is a summary of the variables that have secondary explanatory value, many of which can be considered corollaries of the primary correlates above.

“An additional fifteen variables were judged to be of secondary importance. These included (a) self-efficacy (assertiveness, resistance skills and susceptibility to peer pressure), (b) deviance, (c) attitudes toward drug use, (d) beliefs about psychological consequences of substance use, (e) age and grade, (f) beliefs about social consequences of substance use, (g) intimacy, (h) intentions, (i) school bondedness, (j) others’ attitudes about drug use, (k) independence, (l) achievement values, (m) parental attitudes about drug use, (n) non-structured activities, and (o) peer group characteristics.”(47)

One of the great contributions of the Hansen meta-analysis is the following list of variables which can be considered somewhat irrelevant to explaining substance use among school-age youths.

“This analysis leaves 38 variables with minor or no justifiable role in causing substance use onset or development. Among these are variables that have been extensively examined in research projects, including (a) substance use by parents, (b) personality traits, (c) intelligence (d) social personality traits, (e) parental relations, (f) affect, (g) participation in structured activities, and (j) self-esteem. Based on the observations that each of these variables were included in at least 100 correlational analyses, the popularity of these issues can be judged to be high. Each contribute in only a minor way to understanding substance use. The practical value of including these in theories of substance use is minimal and, without compelling empirical findings that support strong theoretical arguments, should be removed from theories of substance use onset and development.

“Ten of the remaining tertiary variables had 50 or more analyses reported, suggesting that researchers considered them of relative importance. Included in this list were (k) general values, (l) school performance, (m) stress management skills, (n) non-peer, non-family attitudes about drug use, (o) church attendance, (p) availability, (q) academic expectations, (r) drug use by extended relatives, (s) drug use by siblings, and (t) socioeconomic status. Given the frequency with which these variables have been examined, these variables also seem to play little or no role in explaining substance use onset or development.”(48)

The variables with insufficient study to draw conclusions from were religious affiliation, motivation to comply, self-management skills, exposure to moral codes, media influences, and values specific to substance use. It is interesting to note that when called upon to explain recent increases in teenage marijuana use, both Donna Shalala and Lloyd Johnston attributed the rise to the media influences of grunge, rap and rock music stars (see section 4).

Hansen did several confirmatory analyses, and one specific to marijuana shows that regardless of the factors that shape them, peer attitudes have an overwhelming influence on the decision to use marijuana. The top seven variables with the highest mean correlations with marijuana use are (a) Peer attitudes about Drug Use (.677); (b) Drug Use by Peers (.381); (c) Offers (.377); (d) Deviance (.344); (e) Previous Drug Use (.335); (f) Beliefs about social consequences (.314); and (g) attitude toward drug use (.291).(49)

The .677 correlation coefficient between marijuana use and peer attitudes is inconsistent with the other measures. According to Dr. Hansen, this particular variable was “unusual” because it only relied on 18 correlation coefficients from 5 studies. Over half the correlations came from one study. A reanalysis based on equally weighted studies (compensation for differing study sizes) produced a mean correlation coefficient of .53 with a standard deviation of .21, still higher than all the others.(50) The high standard deviation suggests “that in some cases, the particular method of measurement differed or the sample was unusual.”(51) In this case, the .67 figure is an anomaly, a piece of data that sticks out of the distribution. It is easy to pluck this figure out of context and claim that peer attitudes explain more about teenage marijuana use than any other variable, especially since this would confirm the NIDA hypothesis that risk perception is the key to discouraging teen marijuana use. A closer examination of the data suggests that the some odd data has skewed the distribution in this particular sub-category distribution.

Hansen’s paper is an early but valuable attempt to provide a statistically-based review of quantified social science data regarding the use of drugs by school-aged youths. Future reviews will refine Hansen’s analytical techniques and build on his findings.(52) The meta-analysis discussed above was only one aspect of the paper OTA requested from Hansen; future reviews of this material will have a closer focus on the material than the scope of OTA’s request allowed in this review.