Association Between Incident Type 2 Diabetes and Opium Use: Mediation by Body Mass and Adiposity

Abstract

Opiates can affect glucose metabolism and obesity, but no large prospective study (to our knowledge) has investigated the association between long-term opium use, body mass index (BMI; weight (kg)/height (m)²), and incident type 2 diabetes mellitus (T2DM). We analyzed prospective data from 50,045 Golestan Cohort Study participants in Iran (enrollment: 2004–2008). After excluding participants with preexisting diseases, including diabetes, we used adjusted Poisson regression models to estimate incidence rate ratios (IRRs) and 95% confidence intervals (CIs) for T2DM in opium users compared with nonusers, using mediation analysis to assess the BMI-mediated association of opium use with incident T2DM. Of 40,083 included participants (mean age = 51.4 (standard deviation, 8.8) years; 56% female), 16% were opium users (median duration of use, 10 (interquartile range), 4–20) years). During follow-up (until January 2020), 5,342 incident T2DM cases were recorded, including 8.5% of opium users and 14.2% of nonusers. Opium use was associated with an overall decrease in incident T2DM (IRR = 0.83, 95% CI: 0.75, 0.92), with a significant dose-response association. Most (84.3%) of this association was mediated by low BMI or waist circumference, and opium use did not have a direct association with incident T2DM (IRR = 0.97, 95% CI: 0.87, 1.08). Long-term opium use was associated with lower incidence of T2DM, which was mediated by low body mass and adiposity.

Abbreviations

BMI

body mass index

CI

confidence interval

GCS

Golestan Cohort Study

IQR

interquartile range

IRR

incidence rate ratio

SD

standard deviation

T2DM

type 2 diabetes mellitus

Diabetes is one of the fastest-growing public health problems in the 21st century. In 2019, approximately 463 million adults had diabetes mellitus, and this number is expected to increase by 700 million by 2045 (1). The World Health Organization estimates that approximately 50% of diabetes-related deaths occur prematurely (before age 70 years) (2). Accounting for nearly 90% of all diabetes cases (1), type 2 diabetes mellitus (T2DM) is strongly associated with excess body weight, and up to 90% of incident T2DM is estimated to be attributable to adiposity (3). Among the burden of disease related to excess weight, T2DM is the leading cause of years lived with disability and the second-leading cause of mortality worldwide (4).

Opium, the dried latex of the opium poppy (Papaver somniferum) (5), is abused either as minimally processed raw opium or after being processed into other opiates such as heroin. Opium was also an important antidiabetic drug in Europe prior to the discovery of insulin (6). However, data concerning the effects of opiates on both body weight and glucose homeostasis are complicated and inconclusive (7–9). Opium and other opiates contain a variable mixture of substances, including several opioid alkaloids, especially morphine. Morphine, as well as endogenous opioids, can exert diverse effects on the neuroendocrine system, including glucose and energy balance (5, 7, 10). Studies have shown that blood glucose level has a complicated response to opioids, depending on the duration and dose of use and the prevailing metabolic state, such as obesity and hyperglycemia (11). Opioid alkaloids can induce obesity by stimulating food intake, especially intake of palatable foods (12, 13). However, some investigators (5) (but not all (14, 15)) have reported a loss of appetite and weight loss in long-term users. As a result, the association between use of opiates and incidence of T2DM is not clear. The best way to understand the complex association between opiates, adiposity, and T2DM directly is to have information on all 3 factors in a longitudinal study where the temporality of exposures can be established and confounders can be adequately controlled for.

In 2018, more than 30 million people in the world were estimated to have abused opium-derived substances (opium or heroin) in the past year, and 28 million more were estimated to have misused pharmaceutical opioids (16). Opium use has been associated with mortality in the general population (17), and we have previously shown that among individuals with diabetes this association is even more pronounced (18). As a follow-up, we decided to study the direct relationship between opium use, adiposity, and T2DM incidence, since it can complicate the effect of the opium-T2DM interaction on mortality. The Golestan Cohort Study (GCS) has recruited more than 50,000 adults, including 8,500 regular opium users with more than 12 years of follow-up. The availability of measured body mass index (BMI) and data on new-onset T2DM provided us with a unique opportunity to investigate the association between opium use, body mass, and the incidence of T2DM.

METHODS

Study design and population

The GCS is a population-based prospective study conducted in northeastern Iran. The GCS has been described in detail previously (19). In brief, residents aged 40–75 years without a history of upper gastrointestinal cancer were eligible to take part in the study. During the period 2004–2008, a total of 50,045 adult participants from the city of Gonbad-e Kavus (20%) and 326 villages in Golestan Province (80%) were recruited, and they have been followed up regularly since then. In 2011–2012, a subsample of the participants (n = 11,418) were reevaluated with regard to the baseline characteristics, along with biochemical assays, including fasting blood glucose (the repeated-measurement phase).

For the current study, we excluded all people who had diabetes at baseline (n = 3,548) to limit the outcome to new-onset T2DM occurring during the course of follow-up. We also excluded participants with preexisting conditions known to increase the likelihood of opium use to alleviate pain or discomfort and impact adiposity, including heart disease, stroke, cancer, renal failure, chronic obstructive pulmonary disease, asthma, tuberculosis, and liver disease (n = 6,414). The final analytical sample included 40,083 participants.

Baseline assessment

Trained interviewers collected data on demographic characteristics and medical history, including age, sex, ethnicity, marital status, urbanicity of residence, education, lifestyle, history of diseases, and medication use. The GCS questionnaire also included detailed questions about opium and tobacco use. Opium in this area mainly includes 2 types: teriak (dried latex of the poppy) and shireh (a filtrate product of boiling teriak dross residue after smoking, with or without adding teriak, in water). Opium can be taken orally or by smoking. We asked about all periods of opium use during a person’s life, including dates of starting and stopping, types, routes, and doses in nokhods (a local unit equal to 0.2 g) (20). Ever use of opium or tobacco at least once a week for 6 months was considered regular opium or tobacco use (17, 18). Current opium or tobacco users included those who used the substance at baseline, and former users included those who had quit using the substance more than 1 year before enrollment. To better characterize the participants’ lifetime exposure, we used periods of use (in years) and daily amounts (in nokhods) to build a cumulative scale called “nokhod-years” (similar to pack-years, i.e., daily nokhods multiplied by years of use) (17). We categorized the participants into 7 categories: never users, past users, and 5 groups of current users based on nokhod-years of use (<1.0, 1.0–4.9, 5.0–14.9, 15.0–29.9, and ≥30.0). The validity of self-reported opium use was assessed through comparison with urinary levels of opium metabolites (i.e., codeine and morphine), and the data showed high sensitivity and specificity (93% and 89%, respectively) (21). In addition, the data had excellent reliability using a reinterview after 2 months, with a kappa (κ) statistic greater than 0.9 (21). Tobacco use included tobacco smoking (i.e., cigarettes, cigars, pipes, and water pipes) and nass (a local smokeless tobacco product) chewing. Blood pressure was measured, and high systolic or diastolic blood pressure (≥140 mm Hg and ≥90 mm Hg, respectively) or use of antihypertension medication was considered hypertension. Body weight, height, and waist and hip circumferences were measured using standard protocols. BMI was calculated as weight (in kilograms) divided by squared height (in meters). Wealth score was calculated from appliance ownership and other variables, using multiple correspondence analysis, which explained 86.8% of the variation in the data, and was categorized into quartiles (from 1, the lowest category, to 4, the highest category) (22). There were no missing values for opium-use history, and for other covariates, the numbers of missing values were small (BMI and waist and hip circumferences: <0.01%; systolic and diastolic blood pressure: 0.08%).

Follow-up and outcome assessment

At enrollment, participants were asked to provide their home addresses, cell-phone numbers, and 2 other phone numbers of family members, close friends, or neighbors (19). Trained interviewers, using structured questionnaires, actively followed all participants annually by phone. If a participant was not accessible after 7 calls during 2 consecutive weeks, other phone numbers or a home visit by the GCS team were considered. During 480,293 person-years of follow-up—a median of 12.1 (interquartile range (IQR), 11.6–13.2) years—only 396 (0.99%) participants were lost to follow-up, and 5,360 deaths (13.4% of the participants, including 4,804 participants without T2DM) were recorded. During follow-up, data on health status, any physician-diagnosed diseases, medications, and admission to hospitals or clinics were collected. If a death was reported, the cause of death was determined using collected medical documents and verbal autopsy, if needed. Self-reported T2DM was defined on the basis of a physician’s diagnosis or antidiabetic drug use. Based on a previous study, the sensitivity and specificity of self-reported T2DM were 61.5% and 97.6%, respectively, using fasting plasma glucose in a random sample (n = 3,811) of our participants (23). For the present report, follow-up continued until the incidence of T2DM, death, loss to follow-up, or January 1, 2020, whichever came first.

Statistical analysis

Descriptive statistics are presented as percentages, mean values with standard deviations (SDs), or median values with 25th–75th percentile ranges (also called the IQR), whichever is appropriate. We used Poisson regression models to estimate incidence rate ratios (IRR) and 95% confidence intervals (95% CI) for T2DM in opium users compared with nonusers and across types, routes, and categories of cumulative use (nokhod-years). We report IRRs adjusted for potential confounders, including age at enrollment (years; continuous), sex (male/female), ethnicity (Turkmen/non-Turkmen), urbanicity of residence (rural/urban), marital status (married/unmarried), education (none/primary/postprimary), tobacco smoking (never/past/current), nass chewing (never/past/current), alcohol drinking (never/ever), systolic blood pressure (mm Hg; continuous), diastolic blood pressure (mm Hg; continuous), antihypertension drug use (no/yes), and wealth score (quartiles) (see Web Figure 1, available at https://doi.org/10.1093/aje/kwad166). The models further adjusted for BMI to estimate BMI-adjusted IRRs and for waist and hip circumferences to investigate the role of adiposity in the association between opium use and incident T2DM. To decompose the total effect of opium use on the incidence of T2DM into direct and indirect (BMI-mediated) effects, we performed mediation analysis using the “paramed” command in Stata (24). The direct and indirect effects together form the total effect. The indirect effect is the part of the total effect that can be explained by a mediator (BMI), while the direct effect is the residual part of the total effect. In this method we are assuming that the baseline BMIs of never users are equal to what the opiate users would have experienced if they had not initiated opiate use (the counterfactual framework). The estimation of direct and indirect effects further requires assuming no unmeasured mediator-outcome, exposure-outcome, or exposure-mediator confounding. Up to 90% of the incidence of T2DM is attributable to obesity (3); therefore, there are few confounders in the mediator-outcome association. We adjusted the mediation models for all known exposure-mediator and exposure-outcome confounders, as shown in Web Figure 1. Finally, since there was a significant interaction between opium use and BMI in the association with incident T2DM (P = 0.05), we included this interaction term in the mediation analyses. The mediated proportion was computed as the log of the indirect effect divided by the log of the total effect.

Five sensitivity analyses were conducted: 1) exclusion of ever tobacco and/or alcohol users, because these habits were highly associated with opium use and adjustment might not fully control their confounding effects; 2) exclusion of participants without formal education and/or a low wealth score (less than median), because we wanted to evaluate obesity/T2DM due to opium use, not poverty or low socioeconomic status; 3) restriction of incident T2DM to persons who used antidiabetes drugs; 4) use of waist and hip circumferences instead of BMI to evaluate whether the visceral obesity component of BMI is affected by opium use; and 5) competing-risk analysis (25) considering mortality as a competing risk for incident T2DM and evaluation of the association between opium use and the incidence of T2DM. Subgroup analyses and assessment of multiplicative interaction were performed for potential effect modifiers, including baseline age, sex, residence, ethnicity, education, wealth score, tobacco use, and history of hypertension.

We used the repeated-measurement data for quantitative bias analysis and to correct for the potential misclassification of self-reported T2DM. The gold standard for T2DM (i.e., confirmed T2DM) was defined as either a fasting blood glucose concentration greater than or equal to 126 mg/dL or physician-diagnosed T2DM in the repeated-measurement phase. We determined the predictors of confirmed T2DM using logistic regression and estimated the probability of T2DM in all participants, which was used as an inverse probability weight to correct the Poisson regression results.

P values less than 0.05 or 95% CIs not including 1 were considered to indicate statistical significance. Statistical analyses were conducted using Stata statistical software, version 12 (StataCorp LLC, College Station, Texas).

Ethical standards

The study protocol was approved by the ethical review committees of the Digestive Disease Research Institute of Tehran University of Medical Sciences, the US National Cancer Institute, and the International Agency for Research on Cancer. All participants provided written informed consent.

RESULTS

The mean age of the 40,083 included participants was 51.44 (SD, 8.8) years, and 56.1% were female. Table 1 shows the characteristics of the included participants by opium use status. Among these participants, 6,354 (15.9%) were opium users, of whom 5,571 (87.7%) were current users. Opium users, compared with nonusers, were more likely to be male (75.5% vs. 37.9%), rural (rather than urban) residents (88.8% vs. 78.8%), current or former tobacco smokers (56.0% vs. 11.1%), current or former nass users (30.5% vs. 3.3%), and ever alcohol drinkers (11.2% vs. 1.9%), and they had a lower wealth score (61.9% below median vs. 47.9% below median). The mean BMI and waist circumference in opium users were significantly lower than those in nonusers (23.46 (SD, 4.7) vs. 27.03 (SD, 5.3) and 88.93 (SD, 13.3) vs. 95.65 (SD, 13.3) cm, respectively) (Table 1). As Figure 1 shows, BMI was highly correlated with waist circumference in both opium users (r = 0.90, P < 0.001) and nonusers (r = 0.88, P < 0.001). Web Table 1 shows the mean BMI values and SDs among all subgroups of participants. Of the 6,354 opium users, 5,544 (87.3%) used only teriak, 540 (8.5%) used only shireh, and 270 (4.3%) used mixed/other types. Routes of consumption included smoking only (n = 4,426; 69.7%), ingestion only (n = 1,570; 24.7%), and mixed/other routes (n = 358; 5.6%). Opium users started using the drug at a median age of 39 (IQR, 30–47) years, had used it for a median duration of 10 years (IQR, 4–20), and used a median amount of 0.6 g (IQR, 0.2–1.2) per day.

Table 1.

Baseline Characteristics of Participants According to Opium Use in the Golestan Cohort Study, 2004–2008

Participant Characteristic a	Non–Opium Users (n = 33,729)		Opium Users (n = 6,354)
	No.	% b	No.	%
Age, yearsc	51.20 (8.7)		52.73 (9.1)
Body mass indexc,d	27.03 (5.3)		23.46 (4.7)
Waist circumference, cmc	95.65 (13.3)		88.93 (13.3)
Hip circumference, cmc	100.09 (9.2)		95.41 (8.1)
Systolic blood pressure, mm Hgc	127.71 (24.0)		123.12 (24.9)
Diastolic blood pressure, mm Hgc	77.72 (13.8)		73.80 (14.7)
Sex
Female	20,936	62.1	1,558	24.5
Male	12,793	37.9	4,796	75.5
Urbanicity of residence
Rural area	26,570	78.8	5,641	88.8
Urban area	7,159	21.2	713	11.2
Ethnicity
Turkmen	25,374	75.2	4,982	78.4
Non-Turkmen	8,355	24.8	1,372	21.6
Marital status
Married	29,882	88.6	5,732	90.2
Unmarried	3,847	11.4	622	9.8
Formal education
None	23,391	69.4	4,104	64.6
Primary school	5,760	17.1	1,320	20.8
Post–primary school	4,578	13.6	930	14.6
Tobacco smoking
Never smoker	29,996	88.9	2,799	44.1
Past smoker	1,487	4.4	1,416	22.3
Current smoker	2,246	6.7	2,139	33.7
Nass chewing
Never user	32,646	96.8	4,417	69.5
Past user	220	0.7	334	5.3
Current user	863	2.6	1,603	25.2
Alcohol drinking
Never drinker	33,100	98.1	5,643	88.8
Ever drinker	629	1.9	711	11.2
Wealth score
First quartile	8,717	25.8	2,463	38.8
Second quartile	7,458	22.1	1,469	23.1
Third quartile	8,646	25.6	1,418	22.3
Fourth quartile	8,908	26.4	1,004	15.8

^a P < 0.0001 for all characteristics in opium users versus nonusers.

^b Percentages shown in the table may not total 100 because of rounding.

^c Values are expressed as mean (standard deviation).

^d Weight (kg)/height (m)².

Figure 1.

Correlation between body mass index (BMI; weight (kg)/height (m)²) and waist circumference among opium users (r = 0.9, P < 0.001) (A) and nonusers (r = 0.88, P < 0.001) (B) in the enrollment phase of the Golestan Cohort Study, 2004–2008.

During the follow-up period, 5,271 new cases of T2DM were reported; among those cases, 1,307 participants (24.8%) used antidiabetes drugs. An additional 71 participants used such medications without reporting a diagnosis of T2DM, so there were a total of 5,342 new cases of T2DM in the cohort (a cumulative incidence of 13.3%).

Table 2 shows the cumulative incidence of T2DM in opium users (8.48%) and nonusers (14.24%)—539 of 6,354 participants and 4,803 of 33,729 participants, respectively. The incidence of T2DM tended to be lower in opium users of all subgroups defined by age, sex, residence, ethnicity, marital status, education, tobacco smoking, nass use, alcohol drinking, wealth score, hypertension, and BMI category.

Table 2.

Cumulative Incidence of Type 2 Diabetes and Crude Incidence Rate Ratios for Type 2 Diabetes in Opium Users Compared With Nonusers in the Golestan Cohort Study, 2004–2020

	Incidence of Diabetes
	Non–Opium Users (n = 33,729)				Opium Users (n = 6,354)
	No		Yes		No		Yes
Participant Characteristic	No.	% a	No.	%	No.	%	No.	%	Crude IRR	95% CI
All participants	28,926	85.8	4,803	14.2	5,815	91.5	539	8.5	0.60	0.54, 0.65
Age group, years
<50	15,463	86.1	2,500	13.9	2,733	91.5	255	8.5	0.61	0.54, 0.70
≥50	13,463	85.4	2,303	14.6	3,082	91.6	284	8.4	0.58	0.51, 0.65
Sex
Female	17,544	83.8	3,392	16.2	1,378	88.5	180	11.6	0.71	0.61, 0.83
Male	11,382	89.0	1,411	11.0	4,437	92.5	359	7.5	0.68	0.60, 0.76
Urbanicity of residence
Rural area	23,033	86.7	3,537	13.3	5,181	91.9	460	8.2	0.61	0.56, 0.68
Urban area	5,893	82.3	1,266	17.7	634	88.9	79	11.1	0.63	0.50, 0.79
Ethnicity
Turkmen	22,115	87.2	3,259	12.8	4,608	92.5	374	7.5	0.58	0.53, 0.65
Non-Turkmen	6,811	81.5	1,544	18.5	1,207	88.0	165	12.0	0.65	0.55, 0.76
Marital status
Married	25,661	85.9	4,221	14.1	5,253	91.6	479	8.4	0.59	0.54, 0.65
Unmarried	3,265	84.9	582	15.1	562	90.4	60	9.7	0.64	0.49, 0.83
Formal education
None	19,997	85.5	3,394	14.5	3,765	91.7	339	8.3	0.57	0.51, 0.64
Primary school	4,965	86.2	795	13.8	1,205	91.3	115	8.7	0.63	0.52, 0.77
Post–primary school	3,964	86.6	614	13.4	845	90.9	85	9.1	0.68	0.54, 0.85
Tobacco smoking
Never smoker	25,591	85.3	4,405	14.7	2,515	89.9	284	10.2	0.69	0.61, 0.78
Past smoker	1,320	88.8	167	11.2	1,307	92.3	109	7.7	0.69	0.54, 0.87
Current smoker	2,015	89.7	231	10.3	1,993	93.2	146	6.8	0.66	0.54, 0.82
Nass chewing
Never user	27,928	85.6	4,718	14.5	3,998	90.5	419	9.5	0.66	0.59, 0.73
Past user	200	90.9	20	9.1	315	94.3	19	5.7	0.63	0.33, 1.17
Current user	798	92.5	65	7.5	1,502	93.7	101	6.3	0.84	0.61, 1.14
Alcohol drinking
Never drinker	28,380	85.7	4,720	14.3	5,167	91.6	476	8.4	0.59	0.54, 0.65
Ever drinker	546	86.8	83	13.2	648	91.1	63	8.9	0.67	0.48, 0.93
Wealth score
First and second quartiles	14,017	86.7	2,158	13.3	3,625	92.2	307	7.8	0.59	0.52, 0.66
Third and fourth quartiles	14,909	84.9	2,645	15.1	2,190	90.4	232	9.6	0.64	0.56, 0.73
Hypertension
No	17,708	88.3	2,348	11.7	3,963	92.8	310	7.3	0.62	0.55, 0.70
Yes	11,218	82.0	2,455	18.0	1,852	89.0	229	11.0	0.61	0.54, 0.70
Body mass indexb
<25.0	11,900	93.5	823	6.5	4,039	94.8	223	5.2	0.81	0.70, 0.94
25.0–29.9	10,196	85.0	1,796	15.0	1,302	88.0	178	12.0	0.80	0.69, 0.94
≥30.0	6,827	75.8	2,183	24.2	474	77.5	138	22.6	0.93	0.78, 1.11

Abbreviations: CI, confidence interval; IRR, incidence rate ratio.

^a Percentages shown in the table may not total 100 because of rounding.

^b Weight (kg)/height (m)².

Web Table 2 shows the characteristics of participants according to opium use and incident T2DM. Mean BMIs were higher in diabetic participants than in nondiabetic participants among both opium users (26.30 (SD, 5.2) vs. 23.20 (SD, 4.6), respectively) and nonusers (29.80 (SD, 5.4) vs. 26.58 (SD, 5.1), respectively) (P < 0.0001 for both groups). A similar pattern was seen with waist circumference in diabetic and nondiabetic participants. Web Table 3 shows the predictors of incident T2DM in the GCS. BMI, waist circumference, hypertension, female sex, and non-Turkmen ethnicity were associated with a higher risk of incident T2DM, and current tobacco smoking, current nass use, and hip circumference were associated with a lower risk.

The IRRs for T2DM in opium users compared with nonusers in crude and adjusted models were 0.60 (95% CI: 0.54, 0.65) and 0.82 (95% CI: 0.74, 0.91), respectively (Table 3). When the results were further adjusted for BMI, the association disappeared, with an IRR of 0.94 (95% CI: 0.85, 1.04). Results for waist- and hip-circumference–adjusted models were similar. This pattern was seen in all sensitivity analyses (Table 3) and all defined subgroups (Figure 2). The inverse association between opium use and incident T2DM and its attenuation after adjustment for BMI/waist circumference were confirmed in participants with higher socioeconomic status and those without a history of tobacco and/or alcohol use. When we considered mortality as a competing risk factor for incident T2DM, the results confirmed that opium use was associated with a lower risk of incident T2DM, with an adjusted subhazard ratio of 0.77 (95% CI: 0.70, 0.86). We determined predictors of confirmed T2DM in the repeated-measurement phase (Web Table 4) and used these predictors to correct for the potential misclassification of self-reported T2DM. The inverse probability weight–corrected association between opiate use and incident T2DM was strengthened (IRR = 0.64, 95% CI: 0.58, 0.72). We also evaluated the associations between types (teriak and shireh) and routes (smoking and oral) of opium use and T2DM. IRRs were 0.83 (95% CI: 0.74, 0.94) for opium smoking, 0.68 (95% CI: 0.55, 0.83) for opium eating, 0.82 (95% CI: 0.73, 0.91) for teriak use, and 0.60 (95% CI: 0.42, 0.86) for shireh use.

Table 3.

Results of Sensitivity Analysis for the Association Between Opium Use and Risk of Incident Type 2 Diabetes Mellitus in the Golestan Cohort Study, 2004–2020

				Model
				Crude		Adjusted a		BMI b -Adjusted c		Waist- and Hip-Adjusted d
Definition of Population	Total No. of Persons	Definition of Incident Diabetes	No. of Persons	IRR	95% CI	IRR	95% CI	IRR	95% CI	IRR	95% CI
All participants	40,083	Self-report or antidiabetes drug use	5,342	0.60	0.54, 0.65	0.82	0.74, 0.91	0.94	0.85, 1.04	0.91	0.82, 1.00
Never users of tobacco or alcohol	31,137	Self-report or antidiabetes drug use	4,539	0.76	0.66, 0.87	0.82	0.72, 0.94	0.95	0.83, 1.08	0.88	0.76, 1.01
Educated, wealthy participants	9,086	Self-report or antidiabetes drug use	1,261	0.65	0.54, 0.78	0.86	0.69, 1.06	0.96	0.78, 1.20	0.96	0.77, 1.18
All participants	40,083	Antidiabetes drug use	1,378	0.41	0.33, 0.50	0.70	0.56, 0.88	0.83	0.67, 1.05	0.75	0.60, 0.94

Abbreviations: BMI, body mass index; CI, confidence interval; IRR, incidence rate ratio.

^a Adjusted for age, sex, ethnicity, marital status, urbanicity of residence, education, tobacco smoking, nass chewing, alcohol drinking, systolic and diastolic blood pressure, antihypertension drug use, and wealth score, if applicable.

^b Weight (kg)/height (m)².

^c Adjusted for all of the above variables plus BMI.

^d Adjusted for all of the above variables plus waist and hip circumferences.

Figure 2.

Association of opium use with the risk of incident type 2 diabetes mellitus among subgroups of participants in the Golestan Cohort Study, 2004–2020. The graph shows results from multivariable adjusted models without (A) and with (B) adjustment for body mass index (BMI; weight (kg)/height (m)²). Red lines indicate that when the adjusted models (panel A; adjusted for age, sex, ethnicity, marital status, urbanicity of residence, education, tobacco smoking, nass chewing, alcohol drinking, systolic and diastolic blood pressure, antihypertension drug use, and wealth score, if applicable) were further adjusted for BMI (panel B), the associations disappeared. All interactions between the subgroups in the BMI-adjusted models were nonsignificant (P > 0.05). CI, confidence interval; IRR, incidence rate ratio.

Table 4 shows the dose-response associations between cumulative opium use (nokhod-years) and incident T2DM, BMI, and waist circumference. The associations were stronger in current users than in past users, and among current users, higher nokhod-years of opium use were associated with lower BMI and waist circumference and decreased risk of incident T2DM in crude and adjusted models (P for trend < 0.001). The inverse dose-response associations were also present between cumulative opium use and BMI and waist circumference measured 5 years later in the repeated-measurement phase (P for trend < 0.001; Web Table 5). Similar to the whole cohort, in this repeated-measurement subsample, incident T2DM was observed among 6.5% of current opium users, 9.7% of past users, and 10.5% of never users (Web Table 6). Measured BMI remained mainly unchanged between the baseline and repeated-measurement phases of the study (Web Table 6).

Table 4.

Associations of Cumulative Opium Dose (Nokhod-Years) With Incident Type 2 Diabetes Mellitus, Body Mass Index, and Waist Circumference in the Golestan Cohort Study, 2004–2020

	Type 2 Diabetes				Body Mass Index a				Waist Circumference, cm
	Crude		Adjusted b		Crude		Adjusted		Crude		Adjusted
Opium Use Status and Level	IRR	95% CI	IRR	95% CI	β c	95% CI	β	95% CI	β	95% CI	β	95% CI
Never users	1.00	Referent	1.00	Referent	1.00	Referent	1.00	Referent	1.00	Referent	1.00	Referent
Past users	0.74	0.59, 0.91	0.93	0.74, 1.16	−2.72	−3.09, −2.35	−0.94	−1.29, −0.59	−4.06	−5.00, −3.12	−1.73	−2.65, −0.81
Current users, nokhod-years
<1.0	0.82	0.64, 1.06	0.95	0.73, 1.23	−1.73	−2.18, −1.27	−0.68	−1.10, −0.26	−2.47	−3.64, −1.31	−1.19	−2.30, −0.08
1.0–4.9	0.65	0.52, 0.81	0.82	0.66, 1.03	−2.92	−3.27, −2.57	−1.47	−1.80, −1.14	−5.50	−6.40, −4.60	−3.56	−4.42, −2.69
5.0–14.9	0.62	0.50, 0.77	0.82	0.66, 1.01	−3.56	−3.89, −3.23	−1.82	−2.13, −1.51	−7.28	−8.12, −6.43	−4.49	−5.31, −3.68
15.0–29.9	0.51	0.40, 0.66	0.73	0.57, 0.94	−4.12	−4.47, −3.77	−2.17	−2.50, −1.84	−8.42	−9.31, −7.53	−5.31	−6.18, −4.44
≥30.0	0.50	0.43, 0.59	0.77	0.65, 0.91	−4.30	−4.51, −4.08	−1.98	−2.20, −1.75	−8.10	−8.65, −7.54	−4.48	−5.07, −3.89

Abbreviations: CI, confidence interval; IRR, incidence rate ratio.

^a Weight (kg)/height (m)².

^b Adjusted for age, sex, ethnicity, marital status, urbanicity of residence, education, tobacco smoking, nass chewing, alcohol drinking, systolic and diastolic blood pressure, use of antihypertension medication, and wealth score. P values for trend were less than 0.001 in all models.

^c Linear regression coefficient.

Figure 3 shows the results of the mediation analysis carried out to decompose the total association between opium use and T2DM incidence into direct and indirect (BMI-mediated) effects. There was no evidence for a direct effect of opium use on the incidence of T2DM (IRR = 0.97, 95% CI: 0.87, 1.08). However, the indirect (BMI-mediated) association between opium use and incident T2DM was significant (IRR = 0.85, 95% CI: 0.83, 0.88). As a result, opium use was associated with a T2DM incidence that was decreased by almost 17% (95% CI: 8, 25), and 84.3% of this association was mediated through low BMI. When we used waist circumference instead of BMI, the results were similar; IRRs and 95% CIs for direct and indirect effects were 0.93 (95% CI: 0.84, 1.03) and 0.87 (95% CI: 0.85, 0.89), respectively (Web Figure 2).

Figure 3.

Results of a mediation analysis conducted to decompose the total effect of opium use on the incidence of type 2 diabetes mellitus (T2DM) into direct and indirect (body mass index (BMI)–mediated) effects in the Golestan Cohort Study, 2004–2020. Incidence rate ratios (with 95% confidence intervals in parentheses) and P values are shown. BMI was defined as weight (kg)/height (m)².

DISCUSSION

In this study, individuals with long-term opium use (median duration of use, 10 years) had significantly lower mean BMI (23.5 vs. 27.0) and waist circumference (88.9 cm vs. 95.7 cm) than nonusers. Furthermore, long-term opium use was associated with an approximately 17% lower incidence of T2DM, which was explained mainly by the lower body mass and waist circumference. We observed dose-response relationships between opium use and BMI, waist circumference, and incident T2DM.

Results of previous population-based studies of the association between opiate or opioid use and obesity and T2DM are varied. A study carried out in the United States showed that BMI in opioid-dependent individuals was lower than in the general population (adjusted BMI = 25.6 vs. 28.9) (26). A population-based study conducted among 5.7 million adults who underwent cardiac surgery also showed lower crude rates of obesity and T2DM in opioid users (n = 11,359) (27). Another study among 28,691 coronary surgery patients showed a lower rate of T2DM prevalence in chronic opium users than in nonusers (30% vs. 40%) (28). In contrast, some studies did not find any association (14) or even showed a higher BMI (29) or fasting blood glucose/T2DM level (15, 23) in opioid/opiate users. Another study, a randomized cluster household survey in Iran (n = 5,895), showed that opium users had a 20%–40% higher risk of T2DM in the crude analysis, but this relationship disappeared after adjustment for confounders (30). All of these comparisons were based on cross-sectional analyses. Cross-sectional analyses cannot determine whether opium use is the cause or a consequence of obesity or T2DM (i.e., reverse causality). Some people with chronic diseases may use opium or opioids because of pain symptoms or their beliefs about opium’s beneficial effects; for example, one study showed that obesity was strongly associated with long-term opioid use, probably due to musculoskeletal pain (31). In addition, painful diabetic neuropathy may result in long-term opioid use (32). Even an earlier cross-sectional analysis in the GCS showed that opium use was correlated with T2DM (23), but the researchers acknowledged the possibility of reverse causality. The prospective design of the current study permitted modeling and mediation analyses showing that the incidence of T2DM was lower in opium users mainly due to the reduced BMI, which was the most important predictor of T2DM (3).

In our study, long-term opium users, as well as tobacco users, had significantly lower BMI than nonusers. Our results are in line with a study that found a lower incidence of overweight and obesity in drug-dependent men during a 3-year follow-up period (33). Another study showed an inverse association between obesity prevalence and drug dependence, especially opioid dependence (26). The authors suggested that cravings for foods and drugs might compete or compensate each other for the same reward area in the brain (26). Research also suggests an increase in appetite and overeating during recovery from drug abuse; during abstinence, patients may experience overeating, especially of pleasurable foods, to compensate for decreased drug-related hedonic responses (26). Similar patterns are also reported in tobacco smokers. In one study, the risk of T2DM was 22% higher among recent quitters than among current smokers, and this excess risk was directly proportional to weight gain (34).

In the brain, opioids can stimulate food intake and induce obesity (35). In contrast, a loss of appetite and weight loss in chronic opium users has been reported (5). Neuroadaptation can partly explain some discrepancies in short-term and long-term opioid effects on appetite and body weight. Long-term activation desensitizes opioid receptors (tolerance) and produces effects opposite those of short-term activation (5, 36). Therefore, neuroadaptation in long-term opioid users may result in reduced appetite and hedonic response to pleasurable foods, and eventually weight loss. Interestingly, opioid system adaptation after administration of pleasurable foods has been reported (37).

In addition to their impact on weight, opioids have complex effects on glucose metabolism. Opioids may have both hyperglycemic and hypoglycemic effects (11). Giugliano et al. (11) suggested that the normal response to physiological doses of β-endorphin, an endogenous opioid, is the suppression of insulin release and hyperglycemia to redirect glucose during periods of glucose need (e.g., stress) from insulin-dependent tissues to insulin-independent tissues (e.g., the brain). However, laboratory-based studies suggested that opium addiction can produce impaired glucose tolerance similar to that in T2DM (11, 38). Several studies that suggested a higher risk of hyperglycemia/T2DM in long-term opioid users were body-weight–matched studies (11, 38, 39). Weight/BMI matching between opium users and nonusers can obscure the chronic effects of opium use on BMI and T2DM incidence.

To our knowledge, our study is the only large prospective study to have evaluated the association between long-term opium use, obesity, and the incidence of T2DM. We adjusted the models for many potential confounders, such as age, sex, education, urbanicity of residence, tobacco smoking, nass use, alcohol drinking, and socioeconomic factors. Opium may be used as a remedy for chronic diseases in the region; thus, to limit the possibility of reverse causality and confounding, we excluded all participants with a history of heart disease, stroke, cancer, renal failure, chronic obstructive pulmonary disease, asthma, tuberculosis, and liver disease. Our results were robust in several sensitivity analyses and among several subgroups based on age, sex, residence, ethnicity, education, tobacco use, and wealth. Our study had several potential limitations as well. The diagnosis of T2DM in our study was based on self-reporting, with sensitivity and specificity of 61.5% and 97.6%, respectively (23). Using data with a T2DM definition based on fasting blood glucose in a subsample of participants, we performed a quantitative bias analysis to explore how the potential misclassification of self-reported T2DM might have affected our results. These results showed a strengthened association, which indicated that our estimates of opium-T2DM associations might be underestimated. When we used antidiabetes drug use as a definition of T2DM, the results remained largely unchanged. Since tobacco use was more prevalent (5 times more prevalent for tobacco smoking and 10 times more prevalent for nass chewing) in opium users than in nonusers, it could potentially confound the association between opium use and T2DM. However, our findings did not change when we restricted the analyses to never tobacco users. Poor health care in opium users may lead to lower self-reporting of new T2DM cases. We tried to overcome this potential bias by adjusting for education, urbanicity of residence, and wealth score. Our findings also were consistent in subgroups based on these main socioeconomic indicators. Finally, in our mediation analysis, we assumed that never users of opiates constitute a reasonable counterfactual for opiate users, by assuming that the lower BMI among the latter is caused by opiate use. To test this assumption, we would ideally need to show similar BMIs between the two groups, before they started opiate use and many years before the cohort was formed. While such measurement was not available (as in most similar cohort studies), we believe that there is no reason to suspect a systematic difference in BMI between the users and nonusers before starting opiates.

The health complications of long-term opium use are well-documented. Chronic opium users have an increased risk of death from cardiovascular disease and cancer compared with nonusers (17, 40), and the International Agency for Research on Cancer has classified opium as a group 1 carcinogen (41). We previously showed that the coexistence of opium use and T2DM aggravated mortality as compared with each of these risk factors alone (18). Our current study shows that the interaction between opium use and T2DM is further complicated by the fact that opium users lose weight, similar to what is seen among tobacco users, and this weight loss is in turn associated with lower risk of T2DM. Better understanding of the health consequences of complex exposures, such as opium use, requires careful evaluation of confounders and mediators, especially in longitudinal studies, while cross-sectional analyses and case-control studies may be inappropriate methods for evaluating these associations.