Effects of Smoking Versus Nonsmoking on Postprandial Glucose Metabolism in Heavy Smokers Compared With Nonsmokers

Tobacco smoking is the leading preventable cause of death in the Western world (1). There is overwhelming evidence of the harmful effects of smoking on virtually every organ system in the human body, with well-described and detrimental effects on the pulmonary and cardiovascular systems. During the past decade, accumulating evidence of cigarette smoking constituting a strong and independent risk factor for the development of type 2 diabetes has appeared (1–3). Smokers have an ∼30–50% increased risk of developing type 2 diabetes, with heavy smokers (individuals defined as smoking >20 cigarettes/day) having an up to 54% increased risk (1). This risk elevation is still present after adjustment for confounding factors (a pooled relative risk of 1.39), emphasizing the independent nature of smoking as a risk factor for type 2 diabetes (1). Although the epidemiological evidence for an increased risk of type 2 diabetes among smokers thus is substantial, the underlying mechanisms remain unclear. When tobacco is consumed, temporarily high nicotine levels can be found in the bloodstream, giving rise to acute and, if used over long period of times, chronic effects on the body (1,4). Chronic tobacco smoking (often defined as >1 year of continuous use) has been found to decrease glucose tolerance, lower insulin secretion, and increase insulin resistance (IR), whereas its effects on fasting plasma glucose seem negligible (5). Furthermore, chronic smoking increases central fat deposits (6–8), sympathetic activity, and cortisol levels (9,10), which in turn may promote visceral fat deposits (11) and increase inflammatory markers (12) and oxidative stress (13) and, thus, contribute to the increased IR found in several large-scale studies (14). The acute effects of smoking seem similar to the chronic effects, as intravenous infusion of nicotine has been shown to worsen IR in patients with type 2 diabetes during euglycemic-hyperinsulinemic clamp experiments (15). Reduced insulin action reduces glucose use in peripheral tissues and thus may elevate plasma glucose levels (16). Accordingly, smoking has been shown to acutely impair glucose tolerance and raise insulin levels during an oral glucose tolerance test (OGTT) (17). In addition to this, acute smoking may interfere with gastric and gallbladder motility, thereby influencing postprandial glucose uptake from the gut (18,19). Combined with the finding of nicotinic cholinergic receptors on pancreatic β- and α-cells (20), this could point to a direct interaction between nicotine and glucose homeostasis, both chronically and acutely. However, there is still a need for more knowledge about the direct effects of tobacco smoking on postprandial glucose homeostasis including the effect of tobacco smoking on the incretin hormones, glucose-dependent insulinotropic polypeptide (GIP), and glucagon-like peptide 1 (GLP-1), which play a major role in postprandial glucose homeostasis (21) and in the development of type 2 diabetes (22–24).

Here we investigated whether and how tobacco smoking influences postprandial glucose homeostasis in heavy smokers without diabetes during a 4-h mixed-meal test and compared the results with the postprandial glucose homeostasis in matched nonsmokers.

The study was approved by the Scientific-Ethical Committee of the Capital Region of Denmark (registration no. H-1–2013–042) and registered with ClinicalTrials.gov (clinical trial reg. no. NCT02497651). The study was performed in accordance with the principles of the Declaration of Helsinki (Seventh Revision, 2013). All experiments were carried out at Clinical Metabolic Physiology, Steno Diabetes Center Copenhagen, Gentofte Hospital, University of Copenhagen, Hellerup, Denmark. Informed consent was obtained before any study-related procedures were carried out.

An overview of subject demographics is presented in Table 1. Twelve healthy, heavy-smoking men and 12 healthy nonsmoking men matched for age and BMI were studied. The subjects were all normoglycemic and had no known first- or second-degree relatives with type 2 diabetes. None of the subjects were treated with drugs likely to affect glucose homeostasis or the secretion of gastrointestinal hormones. The smokers each smoked >20 cigarettes per day and had done so for at least 1 year. The smokers had a larger weekly intake of alcohol compared with the nonsmokers (Table 1) but showed no biochemical signs of increased liver damage. Of the 12 smokers, one smoker was excluded after completion of the study, because the measurement of gastrointestinal hormones indicated a failure to fast before an experimental day. The cigarettes used by the smokers were from various standard brands, with normal contents of nicotine (1 mg), charcoal (10 mg), and carbon monoxide (10 mg) per cigarette (one smoker used “light” cigarettes, containing 0.5 mg nicotine/cigarette).

The 24 subjects were included in this interventional crossover study based on an initial screening visit (for more information, see the inclusion flowchart in the Supplementary Data). After inclusion, the smokers underwent two mixed-meal tests in randomized order separated by a minimum of 24 h and a maximum of 1 month: one without concomitant cigarette smoking (referred to as the nonsmoking test day), and one with concomitant cigarette smoking 10 min before and 10 min after ingestion of the meal (referred to as the smoking test day). Participants smoked their regular brand of cigarette in the experiment. The control group underwent one similar mixed-meal test, but without cigarette smoking. All meal tests were initiated the morning (start ∼8:00 a.m.) after an overnight (10-h) fast (including abstinence from food, tobacco, medication, and liquids), with subjects in a 45° recumbent position in a hospital bed. Before the start of the test, the subjects were asked to empty their urinary bladder. At baseline, a cannula was inserted into a cubital vein for the collection of blood samples. The forearm was kept warm (50°C) by a heating blanket for arterialization of the venous blood (25). After an initial 30-min baseline period (i.e., at time 0 min), the subjects ingested the meal: a 350-mL chocolate drink (150 kcal, 18.5 g of carbohydrates, 5.8 g of lipids, 5.8 g of protein/100 mL; Nutricia Denmark). Acetaminophen (1,500 mg) suspended in 50 mL of water was added to the liquid meal for the evaluation of gastric emptying according to the acetaminophen absorption test (26). Heart rate and blood pressure were monitored every 15 min, and gallbladder volume was measured by ultrasonography five times during the meal test (at baseline and 20, 40, 80, and 240 min after ingestion of the meal). Every 30 min, the subjects filled out visual analog scales (VASs) about appetite, thirst, and general well-being. Blood samples were obtained at times −30, −15, 0, 10, 20, 30, 50, 70, 90, 120, 150, 180, and 240 min. For the analysis of gastrin, cholecystokinin (CCK), GIP, GLP-1, and glucagon, blood was added to chilled tubes containing EDTA, aprotinin (500 kallikrein inhibitor units/mL blood; Trasylol; Bayer, Leverkusen, Germany), and a specific dipeptidyl peptidase 4 inhibitor (valine-pyrrolide, final concentration 0.01 mmol/L; Novo Nordisk, Bagsværd, Denmark). For the analyses of insulin and C-peptide, blood was added to plain tubes for coagulation (20 min at room temperature). For the analysis of acetaminophen, blood was added to heparin-containing tubes. All samples were centrifuged for 20 min at 1,200g and 4°C, and serum/plasma was transferred to storage tubes on ice. Plasma samples for gastrin, CCK, GIP, GLP-1, and glucagon analyses were stored at −20°C, and serum samples for insulin and C-peptide analyses as well as heparinized samples for acetaminophen analysis were stored at −80°C. For bedside measurement of plasma glucose, blood was collected in fluoride-coated tubes and centrifuged immediately at 7,400g for 2 min at room temperature. After the 4-h meal test, the subjects delivered a urine sample and were offered a standardized solid ad libitum meal. Diuresis as well as the consumption of food and water was noted.

Plasma glucose concentrations were measured bedside using the glucose oxidase method (Model 2300 STAT PLUS Analyzer; YSI Incorporated, Yellow Springs, OH). Serum insulin and C-peptide concentrations were measured using a two-site electrochemiluminescence immunoassay (Roche/Hitachi Modular Analytics System; Roche Diagnostics). Serum acetaminophen was measured by the Vitro ACET slide, as described previously (26). Plasma concentrations of gastrin, CCK, GIP, GLP-1, and glucagon were all measured by radioimmunoassays, as previously described (27–30).

The sample size of the study was calculated based on a 0.05 level of significance, an 80% power level, an estimated difference in plasma GLP-1 of 15%, and appropriate SDs from previous literature (31,32), leading to a minimum of 8–11 participants in each group in order to achieve study objectives. Data are reported as the mean followed by the mean difference (MD), 95% CI and level of significance unless otherwise stated. The area under the curve (AUC) was calculated by the trapezoidal method and is presented as the total and/or baseline-subtracted AUC (bsAUC) for the entire 4-h period, unless otherwise stated. The time from ingestion of the acetaminophen-containing meal to each subject’s peak serum acetaminophen concentration (T_max) was used as an indirect marker of the gastric emptying rate because T_max is inversely proportional to the gastric emptying rate. Gallbladder volume was calculated by the following formula: length × width × height × π/6 (33). To evaluate IR, HOMA, insulinogenic index, and disposition index (DI) were used. HOMA-IR was calculated as (fasting plasma glucose × fasting plasma insulin)/22.5. The insulinogenic index was calculated as AUC_insulin/AUC_glucose. The DI was calculated as insulinogenic index/HOMA-IR. The overall feeling of appetite was evaluated from the composite appetite score (CAS) derived from the VAS scores. The VAS measurements were carried out in accordance with studies validating the technique (34). The CAS was calculated as (hunger + prospective food consumption − satiety − fullness)/4. Differences among the 3 study days regarding baseline characteristics and AUCs were evaluated using multiple t tests and mixed modeling including subject identification as a random factor and a random covariance structure based on smoking status, BMI, and/or age, respectively, if needed according to log-likelihood ratios. When comparing the 2 test days of the smoking group, a Student t test analysis was used. A two-sided P value of <0.05 was chosen to indicate significant differences. To avoid false-positive results due to multiple comparisons, we performed Tukey post hoc tests using the estimated marginal means method in R (R Foundation [www.r-project.org]). Statistical analysis was carried out using the nlme (version 3.3), multcomp (version 1.4), and emmeans (version 1.1) packages in R (version 3.3.1; R Foundation), GraphPad Prism version 6.0 for Windows/Mac (GraphPad Software, San Diego, CA), and Microsoft Excel version 15.31 (Redmond, WA).

Excursions for each of the 3 test days are presented in Fig. 1A. No significant differences in baseline plasma glucose values were observed between the two groups. Regarding the smoking group, fasting plasma glucose levels on the smoking test day were slightly higher compared with the nonsmoking test day (5.3 vs. 5.1 mmol/L; MD 0.2 [95% CI 0.1, 0.5]; P = 0.04). During the first 2 h after ingestion of the meal, the smokers exhibited a tendency toward a larger increment in plasma glucose on their nonsmoking day compared with the matched control subjects (bsAUC_{0–120 min} 116 vs. 59 mmol/L/min; MD 57 [95% CI 4.6, 109]; P = 0.08). The peak value (Table 2) and overall excursions of plasma glucose were significantly lower throughout the entire test when the smokers were smoking compared with when they were not (bsAUC 35 vs. 105 mmol/L/min; MD −70 [95% CI −119, −21]; P = 0.03).

Excursions of plasma acetaminophen levels are shown in Fig. 1B. We found a trend toward a longer T_max when the smokers smoked compared with their nonsmoking test day (110 vs. 96 min; MD 14 [95% CI 1, 28]; P = 0.08). A trend toward a shorter T_max was observed among smokers (on their nonsmoking day) compared with control subjects (96 vs. 119 min; MD −23 [95% CI −2, −44]; P = 0.10).

Gallbladder volume over time is presented in Fig. 1C. When smoking, the smokers exhibited a tendency toward slower gallbladder emptying; however, no significant differences between the smokers’ 2 test days or the control subjects in postprandial gallbladder motility were found (Table 3).

Mean heart rate over time is presented in Fig. 1D and in Table 3. The smokers’ heart rates were not significantly different on their smoking test day compared with their nonsmoking test day, nor did it differ significantly from the control group on any of the test days. The subjects’ blood pressure values (diastolic as well as systolic) did not fluctuate during the meal test, nor did they differ significantly between the two groups or between the smokers’ 2 test days (Table 3).

Time courses for plasma/serum glucagon, insulin, and C-peptide concentrations are presented in Fig. 1E–G. Compared with the control group, fasting plasma concentrations of glucagon in the smoking group were significantly elevated on the nonsmoking test day (Table 2). The elevated fasting plasma glucagon concentrations remained significant when analyzing the overall mean fasting glucagon concentration from the 2 test days of the smoking group against the control group (8.9 vs. 6.1 pmol/L; MD 2.8 [95% CI 0.01, 5.6]; P = 0.049). However, the postprandial response of glucagon was similar in the two groups and within the smoking group. No significant differences between the groups or between the smokers’ 2 test days were observed in the responses of either insulin or C-peptide after the ingestion of the mixed meal (Table 2). No significant differences in HOMA-IR were found between the groups or within the group of smokers (Table 3). The DI of the control subjects was significantly higher compared with that of the smoking group, among whom there were significant differences between the 2 test days (Table 3).

Plasma gastrin and CCK concentrations over time are presented in Fig. 1H and I. No significant differences in gastrin concentrations were observed between the smokers and the control subjects. A significant increase in the postprandial gastrin response (bsAUC) was observed on the day the smokers smoked compared with their nonsmoking test day (1,534 vs. 1,223 pmol/L/min; MD 311 [95% CI 100, 524]; P = 0.02). CCK concentrations did not differ between the smokers’ 2 test days; however, the smokers exhibited a trend toward a larger postprandial response in CCK on their smoking test day compared with control subjects (bsAUC 421 vs. 290 pmol/L/min; P = 0.10).

Plasma concentrations of GIP and GLP-1 over time are presented in Fig. 1J and K. No significant differences in GIP levels were recorded (Table 2). No significant differences in fasting or postprandial GLP-1 values were observed between the groups or within the smoking group.

Sensations of hunger, satiety, fullness, prospective food consumption, CAS, comfort, level of nausea, and thirst are presented in tables and figures as the change from baseline over time in the Supplementary Data. Overall appetite, summarized as the CAS, decreased during the first 30 min after ingestion of the meal, and then rose steadily during the remaining 3.5 h on all 3 test days. Although a tendency toward a larger appetite in the smoking group was observed, no significant differences among the 3 test days were detected in the overall model in the categories directly related to food intake. If analyzed separately, the smokers’ hunger and CAS values on the nonsmoking test day were significantly larger than the values for the control group during the last 2.5 h of the experiment, indicating a faster buildup of appetite in chronic smokers. In terms of nausea, the control subjects reported an increase in nausea level after meal intake compared with the smokers, who reported a mean decrease in nausea levels on both days after the meal.

There were no significant differences in the subjects’ consumption of the ad libitum meal among the 3 days (Supplementary Data). No differences were observed in either water intake or diuresis.

Here, we investigated the effects of smoking on postprandial glucose metabolism and show that heavy smokers are characterized by fasting hyperglucagonemia and exaggerated postprandial glucose excursions compared with matched nonsmoking control subjects and that smoking in association with a meal, contrary to our expectations, decreases postprandial glucose excursions in heavy smokers.

We hypothesized that smoking and the ensuing elevation in plasma nicotine concentrations would render the heavy smokers more insulin resistant, and, thus, compromise their postprandial glucose tolerance. In line with this, a tendency toward lower plasma glucose responses during the first 2 h after ingestion of the meal was seen in the control group compared with the smokers. The exact mechanisms behind this difference cannot be derived from the current study, but the trend of slower gastric emptying observed in the control group (MD 23.3 min in T_max of acetaminophen) could be part of the explanation.

Significantly higher fasting glucagon concentrations characterized the smoking group. This is interesting, because hyperglucagonemia represents a classic pathophysiological phenomenon of type 2 diabetes (35,36) and is an important contributor to type 2 diabetic hyperglycemia (37). Furthermore, it has been shown that hyperglucagonemia in the fasting state may represent one of the earliest signs of the development of type 2 diabetes in obese people with otherwise normal glucose tolerance (22). Thus, the elevated concentrations of glucagon found in the smoking group, could—at least partly—contribute to the smokers’ elevated risk of the development of type 2 diabetes. The exact mechanisms behind the hyperglucagonemia in the group of smokers cannot be deduced from our study. Because chronic smoking has been shown to be associated with increased visceral fat deposits, steatosis-induced hepatic glucagon resistance and compensatory glucagon secretion from the pancreas may, hypothetically, contribute (38).

We also hypothesized that smoking in conjunction with the mixed meal would have a detrimental effect on the smokers’ postprandial glucose homeostasis compared with their nonsmoking test day, as seen in a previous study evaluating the effect of smoking during OGTTs (17). However, contrary to our expectations, we observed that smoking in conjunction with the meal increased postprandial glucose tolerance compared with an identical test day without any cigarette smoking. The difference in plasma glucose values does not seem to be a result of altered pancreatic function because no differences between the respective test days were observed regarding the postprandial response of insulin, C-peptide, or glucagon. The unaltered insulin response stands in contrast to the study by Frati et al. (17), where acute smoking caused an increment in insulin levels after an OGTT. The lack of difference in postprandial glucagon levels between the smokers’ 2 test days suggests that the fasting hyperglucagonemia in the heavy smokers most likely is an effect of chronic smoking. Similarly, our data suggest that smoking does not have a direct effect on postprandial plasma responses of the incretin hormones GIP and GLP-1. This is in line with other studies showing no effect of smoking cessation in heavy smokers on postprandial responses of GLP-1, GIP, and other gut hormones (39,40). Interestingly, the gut hormone CCK and its cellular pathways have been suspected to interact with and mediate the effects of nicotine, and studies in rats have shown nicotine and other derivatives of smoking to acutely increase CCK levels (41,42). A nonsignificant increment in postprandial CCK levels was detected in the control group when compared with the smokers on their smoking test day. The importance of this potential difference is not known, as no differences in gallbladder motility or gastric emptying between these 2 test days were detected.

A likely explanation of the smoking-induced improvement in glucose tolerance may be smoking-induced deceleration of gastric emptying, as suggested by our acetaminophen data. It has previously been shown that smoking can decrease gastric emptying of solid foods (18), although the opposite effect also has been described (19). Plasma concentrations of gastrin were significantly elevated during the smoking test day. Gastrin is known to enhance smooth muscle contractions in the antral part of the ventricle as well as in the pyloric sphincter and to increase the secretion of gastric acid, both leading to the deceleration of gastric emptying (43,44). Smoking has previously been found to affect gastrin levels, perhaps as a compensatory mechanism for the decreased gastric acid secretion often seen during smoking (45,46). It is thus possible that the reduced gastric emptying on the smoking test day is partly mediated by increased gastrin secretion.

One may speculate that the smokers experience an increased stress response during their nonsmoking test day because they are deprived of their normal dose of nicotine and that this stress response is responsible for the increased plasma glucose concentrations seen on this day. Thus, one could regard the nonsmoking test day as the actual intervention day for the heavy-smoking group. This view would better explain the apparent paradox that the gastric emptying data from the smoking group resemble data from the control group the most on the smoking test day. However, this view is not supported by the smokers’ heart rates, which if anything were slightly lower during the nonsmoking test day.

We did not include a smoking test day for the control group for ethical reasons and practical issues, which could problematize the interpretation of the results (e.g., how to ensure proper inhalation among nonsmokers without aberrant reactions like coughing, nausea, and possible vomiting). Other limitations to the study include the relatively small sample size of 23 participants, limiting the statistical power.

In conclusion, our chronic heavy smokers were characterized by higher fasting plasma concentrations of glucagon, which may play a role in the elevated risk of type 2 diabetes among smokers established from epidemiological studies. In contrast to our hypothesis, smoking in conjunction with a mixed meal seems to decrease postprandial glucose excursions, likely via smoking-induced deceleration of gastric emptying.

Acknowledgments. The authors thank the participants, without whom the study would have been impossible. The authors also thank Inass Al Nachar, Sisse Marie Schmidt, and the Medico-Technical Department at Gentofte Hospital for invaluable help.

Duality of Interest. Within the past 36 months, T.V. has served on scientific advisory panels and/or speaker’s bureaus for and served as a consultant to and/or received research support from Amgen, AstraZeneca, Boehringer Ingelheim, Bristol-Myers Squibb, Eli Lilly, Merck Sharp & Dohme, Novo Nordisk, and Sanofi. Within the past 36 months, F.K.K. has served on scientific advisory panels and/or speaker’s bureaus for and served as a consultant to and/or received research support from Amgen, AstraZeneca, Boehringer Ingelheim, Eli Lilly, Merck Sharp & Dohme, Novo Nordisk, Sanofi, and Zealand Pharma. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. M.F.G. planned and executed the study as well as wrote applications for funding, conducted the clinical experiments, researched the data, performed statistical analyses, and wrote the manuscript. J.I.B. and A.L. planned the study and helped with the clinical experiments and the statistical analyses. A.F. and T.V. planned the study. J.F.R. processed and analyzed CCK and gastrin data. J.J.H. provided radioimmunoassay analyses of glucagon, GLP-1, and GIP. F.K.K. conceptualized the study, wrote applications for funding, planned the study, and wrote the manuscript. All authors contributed to discussion and critically reviewed the manuscript. F.K.K. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this study were presented in abstract form at the 76th Scientific Sessions of the American Diabetes Association, New Orleans, LA, 10–14 June 2016.