Subjects were defined as having DM if they answer ‘yes’ to the question ‘Have you ever been told you have sugar/diabetes?’ or had a fasting blood glucose level $126 mg/dl . Of the 719 patients with DM, 418 answered the question about whether they take insulin and 116 reported that they do take insulin. Of those, nine reported that they began using insulin at age #20 years, the majority being likely to have type 1 DM, although a few may have had type 2 DM. Thus, we estimate that 1.5% of patients with DM had type 1 DM, and because of this low number, we analysed all subjects with DM together. There was no difference in any of our analyses if the nine patients of age #20 years were excluded. The study included 151 pregnant women . Of them, eight women had diabetes. There was no difference in the use of marijuana by DM. Because of the low number in the diabetes category, we included them in the analysis. A series of sensitivity analyses excluding the pregnant women showed no difference. Plasma glucose and whole blood haemoglobin A1c were measured at the University of Missouri Columbia School of Medicine Department of Child Health, Diabetes Reference Laboratory, Columbia, Missouri, by David Gold stein, MD, director.Subjects were classified as obese/non-obese according to the BMI level using a cut-off of 30 kg/m2 . We analysed data related to DM, age, gender, race/ ethnicity, education level, family history of DM, physical activity, BMI, cigarette smoking, cocaine use, alcohol use, total serum cholesterol, high-density lipoprotein cholesterol, low-density lipoprotein cholesterol, triglycerides, serum 25-hydroxy vitamin D , HbA1c,vertical outdoor farming fasting plasma glucose level, C reactive protein level and the serum levels of less robust inflammatory markers count and uric acid that have been previously used in NHANES III analysis.
Physical activity was assessed using self-report to several questions . For the physical activity variable, subjects were classified as inactive if they did not report engaging in any of the following activities during the previous month: walking, jogging, bike riding, swimming, aerobics, dancing, calisthenics, gardening, lifting weights or other physical activity outside their occupation. Physical activity was classified as moderate or vigorous intensity based on metabolic equivalent intensity levels. Individuals were considered to fulfil national recommendations for physical activity if they reported five or more episodesper week of moderate-intensity physical activity or three or more episodes per week of vigorous-intensity physical activity.Descriptive statistics were used to characterise the subjects . To test the statistical difference between the groups, we used c2 test for categorical variables and two-sided t tests for continuous variables. A p value of <0.05 was considered significant. Univariate and multivariate logistic regression analyses were used to determine the relationship between DM and marijuana use. We used multivariate logistic regression to adjust for confounding variables and reported the OR and the 95% CI. Variables considered as possible confounders in the multivariate analysis were age, gender, race/ ethnicity, BMI, education level, cigarette smoking, alcohol use, physical activity, serum total cholesterol, HDL cholesterol, LDL cholesterol, triglycerides, vitamin D, CRP, ferritin, fibrinogen, WBC count and uric acid. In order to confirm that marijuana use was associated with DM and not due to confounders, we analysed how each potential confounder changed the OR of having DM. Variables that changed the OR by $10% were considered as confounders and included in the multivariate model. We performed stratified analysis to test for effect modification. For effect modifier variable, multivariate logistic regression model was constructed for each subgroup.
In addition, to help adjust for selection bias, we analysed the data using the propensity score matching and estimated the average treatment effect for the treated, bootstrap SE and t statistics. We added the propensity score to the logistic regression model as inverse weight, blocks that satisfy the balancing property and quartiles. Data were analysed using SAS and the survey module of STATA . Sample weights, provided by the National Center for Health Statistics, were used to correct for differential selection probabilities and to adjust for non-coverage and non-response.Among NHANES III participants aged 20e59 years, there were 6667 non-marijuana users, 3346 past marijuana users, 557 light current users and 326 heavy current users. As shown in table 1, current and past marijuana users tended to be <40 years old, be male, had a BMI of <30 kg/m2 , smoked cigarettes and used alcohol and cocaine more frequently compared to non-marijuana users. Compared to non-marijuana users, past users tended to be white and to have a college education, while current users included more white and black subjects and were more likely to have a high school education or less. Non-marijuana users, past and current marijuana users had a similar percentage of family history of DM but significantly different percentage of physical activity levels , with past and current marijuana users being more active than non-marijuana users. As shown in supplement table 1, marijuana users had a lower adjusted prevalence of DM, but not hypertension, stroke, myocardial infarction or heart failure compared to non-marijuana users. The unadjusted prevalence of DM for non-marijuana users, past marijuana users, current light marijuana users and current heavy marijuana users was 6.3%, 2.9%, 1.9% and 3.0%, respectively, and there was a statistically significant difference between the groups . For subjects without DM , 46.4% were marijuana users and 53.6% were non-marijuana users . For subjects with DM , 26.9% were marijuana users and 73.1% were non-marijuana users . The difference in % of marijuana users between those with and without DM was highly significant .
As shown in table 1, all marijuana users had a higher prevalence of serum HDL cholesterol >40 mg/dl, total cholesterol <240 mg/dl and triglycerides <200 mg/dl compared to non-users . Current marijuana users had a higher prevalence of LDL cholesterol <160 mg/dl . All marijuana users had a higher prevalence of CRP <0.5 mg/dl . Past users, but not current users, had a lower prevalence of vitamin D level <70 nmol/l compared to non-users . All marijuana users had a higher prevalence of plasma HbA1c <6.0% . Serum glucose levels and BMI were lower in all marijuana user groups compared to non-marijuana users . We then examined the variation of markers of inflammation with marijuana use . Serum CRP and fibrinogen were significantly lower in past marijuana users compared to current and non-marijuana users suggesting lower inflammation in past marijuana users. In contrast, serum ferritin levels were higher in past and current heavy users, and lower in light users,rolling grow table compared to non-users. Serum uric acid levels were higher in past and lower in current users compared to non-users. WBC count was higher among current users relative to non-users and past users. In order to confirm that marijuana use was associated with a decreased prevalence of DM and not due to confounders, we analysed how each potential confounder changed the OR of having DM. Variables that changed the OR by $10% were considered as confounders . Table 2 shows the unadjusted as well as the cumulative effect of the confounders, including race/ethnicity, physical activity and those variables that showed changes of $10% in the OR of having DM among all marijuana users relative to non-users in a series of regression models. Of note, race/ethnicity and physical activity did not change the OR by $10%, but we included them in the model because they are known risk factors. The interaction effect of the marijuana use and age was significant in the model indicating that age is an effect modifier . Stratified analysis by age group found an association between marijuana use and DM among subjects aged >40 years and no association among subjects aged #40 years . The association of DM and marijuana was significant in both the overall and older age group even after adjusting for social variables , laboratory variables , inflammatory marker and the comorbidity variable to the previous model. Using the propensity score matching, we found similar results showing a lower prevalence of DM among marijuana users relative to non-users. The average treatment effect for the users ¼0.024, bootstrap SE¼0.005 and t¼4.46, p<0.05 . When we added the propensity score to the logistic regression model, marijuana users still had lower odds of DM than non-users . Adding it as inverse weight, yielded an OR¼0.52 . We also added it as blocks and found an OR¼0.53 . Adding it as quartiles yielded an OR¼0.51 . All still revealed a lower odds of DM with marijuana use. For age group 41e59 years, adding the propensity score as quartiles to the model, we found an OR¼0.55 , whereas for age group 20e40 years, OR¼0.88 . We examine whether DM as diagnosed by self-report as compared to laboratory evidence of hyperglycaemia was correlated with different prevalence of marijuana use. As shown in the supplement table 2, there was no difference in marijuana use among those with DM by self-report and those with DM who were included based on an elevated fasting glucose . Patients with DM by self-report who were hyper glycaemic at the time of sampling had a statistically similar rate of marijuana use as those whose DM was well controlled at the time of sampling , although there was a trend for patients with a history of DM by self-report who were euglycaemia at the time of sampling to be associated with a lower rate of non-marijuana use.
Those with DM by self-report and those with DM who were included based on an elevated fasting glucose had similar rates of the type of marijuana use . Additionally, for subjects who did not have DM by self-report and did not have an elevated fasting glucose level but had an elevated HbA1c , their prevalence of non-marijuana use was similar to the prevalence of non-marijuana use among subjects with DM . We then examined the prevalence of all marijuana users among subjects with different fasting glucose levels. As shown in figure 1, the highest prevalence of marijuana users was found in those with the lowest glucose levels. As the glucose levels increased, the prevalence of marijuana users decreased. For subjects with DM , the prevalence of marijuana users was 23.6%. Similarly, the highest prevalence of marijuana users was found in those subjects with the lowest plasma HbA1c values . As the HbA1c levels increased, the prevalence of marijuana users decreased. Furthermore, we analysed the data using logistic regression to assess the odds of having DM, an elevated glucose value or an elevated HbA1c for the categories of marijuana use. The OR for all marijuana users to have DM was 0.42 , which was statistically significant . Relative to non-marijuana users, past marijuana users had an OR of having DM of 0.44 , current light marijuana users had an OR of 0.29 and current heavy marijuana users had an OR of 0.47 , all were statistically significant from non-marijuana users . Relative to non-marijuana users, marijuana users had significantly lower odds of having glucose level of >125 mg/dl and HbA1c level >7.0% .Our analyses of adults aged 20e59 years in the NHANES III database showed that participants who used marijuana had lower prevalence of DM and had lower odds of DM relative to non-marijuana users. We did not find an association between the use of marijuana and other chronic diseases, such as hypertension, stroke, myocardial infarction and heart failure. This could be due to the smaller prevalence of stroke, myocardial infarction and heart failure in the examined age group. We noted the lowest prevalence of DM in current light marijuana users, with current heavy marijuana users and past users also having a lower prevalence of DM than non-marijuana users. The finding that past marijuana users had lower odds of prevalent DM than non-users suggests that early exposure to marijuana may affect the development of DM and a window of time of marijuana exposure earlier in life could be a factor to study. Similarly, our findings of a significant association between marijuana use and DM was only found in those aged $40 years suggest that the possibility of some protection from marijuana use may require many years before they become manifested.