Smoking and the Risk of LADA: Results From a Swedish Population-Based Case-Control Study

Smoking increases the risk of type 2 diabetes (1). This effect is exerted mainly through a decrease in insulin sensitivity (2). Knowledge about the impact of smoking on other forms of diabetes is limited. In a study based on the Norwegian HUNT study (3), we showed that smoking is associated with a reduced risk of latent autoimmune diabetes in adults (LADA). These findings fit with previous observations (4,5) in individuals with type 1 diabetes, indicating a reduced risk in the offspring of smoking parents. A suggested mechanism behind a protective effect could be the modulating effect of smoking/exposure to nicotine on immune response and an inflammatory process (6,7). The clinical impact of such a biological mechanism has been debated (6–9); however, the manifestations of at least one autoimmune disease (ulcerative colitis) are known to be ameliorated by smoking.

LADA is a heterogeneous form of diabetes sharing clinical and genetic characteristics with both type 1 and type 2 diabetes (10). The impact of autoimmunity and insulin deficiency on the one hand, and insulin resistance (IR) on the other hand, varies within and between populations (10). Hence, similar to type 2 diabetes, smoking might increase the risk of LADA by aggravating the risk factor of IR. In the current study, our aim was to investigate the role of smoking in the development of LADA using data from a Swedish case-control study with three times as many case patients as in our previous analysis based on the HUNT study.

ESTRID (Epidemiological Study of Risk factors In LADA and type 2 Diabetes) is a Swedish population-based case-control study, which has been ongoing since 2010 (11). ESTRID is a substudy to the ANDIS (All New Diabetes in Scania [http://andis.ludc.med.lu.se]) study, a large-scale study performed in Scania, a county in the south of Sweden (∼1,300,000 inhabitants), with the aim of characterizing all new cases of diabetes according to diabetes type, clinical features, and genetic factors. To ESTRID, we invited all patients with newly diagnosed LADA recorded in the ANDIS study since 2010, together with a random sample of type 2 diabetes case patients (four per LADA case). Six nondiabetic control subjects per LADA case (corresponding to 1 control subject per case patient with LADA or type 2 diabetes) were selected at random from the population register covering Scania, matching the date of participation and the area of residence (incidence density sampling) (12). In 2012, ESTRID expanded the recruitment to the ANDIU (All New Diabetes in Uppsala [http://www.andiu.se/]) study, a sister study to the ANDIS study performed in the County of Uppsala (∼300,000 inhabitants) (Supplementary Fig. 1).

The participation rate was 81% among case patients and 66% among control subjects. In the current study, we included all individuals recruited until July 2015, consisting of 377 LADA case patients, 1,188 type 2 diabetes case patients, and 1,472 control subjects available for tobacco use and other covariates of interest. Ninety-seven percent of participants came from Scania, and 3% came from Uppsala. Ethical approval for ESTRID was obtained from the ethical review board in Stockholm, and all participants consented to take part in ESTRID.

GAD antibodies (GADAs) were analyzed with an ELISA (RSR Ltd.). The GADA was considered to be positive if the serum antibody level exceeded 10 international units (IU)/mL (sensitivity of 84%, and specificity of 98% at a cutoff of 10.7 IU/mL) (13). GADA is the autoantibody with the highest penetration, being present in 70–80% of patients with autoimmune diabetes, and is known as a good predictive marker of insulin dependency in adult patients (14,15). C-peptide level was measured using the IMMULITE 2000 (Siemens Healthcare Diagnostics Product Ltd., Llanberis, U.K.) or the Cobas e 601 analyzer (Roche Diagnostics, Mannheim, Germany) (16). Fasting plasma glucose and C-peptide measurements were used to calculate the HOMA of IR (HOMA-IR) to estimate IR and the HOMA of β-cell function (HOMA-β) to assess the percentage of β-cell function (17). The clinical measurements were conducted only in diabetes patients.

The patients with diabetes received diagnoses in the primary health care facilities of Scania and Uppsala counties. Patients with diabetes with an age at disease onset of ≥35 years were defined as a LADA case patient if the following criteria were met: 1) GADA positive (≥10 IU/mL) and 2) C-peptide level ≥0.2 nmol/L (IMMULITE analyzer) or ≥0.3 nmol/L (Cobas e 601). And they were considered as having type 2 diabetes if they were 1) GADA negative (<10 IU/mL) and 2) had a C-peptide concentration of >0.6 nmol/L (IMMULITE analyzer) or ≥0.72 nmol/L (Cobas e 601 analyzer). The cutoffs were based on recommendations from the manufacturers.

The exposure (smoking) was self-reported in a questionnaire that patients received as near as possible to the time of diagnosis. The questionnaires contained instructions to report lifestyle habits before the onset of disease. There were detailed questions regarding lifetime history of smoking (cigarette and cigar), including smoking status (current, former, and never), number of cigarettes smoked per day, and duration of smoking. In order to analyze the intensity of smoking, we categorized smokers into light smokers (20 cigarettes/day) and heavy smokers (≥20 cigarettes/day). Additionally, the cumulative dose of smoking (pack-years) was assessed in ever smokers. One pack-year was equivalent to smoking 20 cigarettes/day for 1 year, and it was assessed in three categories (never, 0–15 cigarettes, and ≥15 cigarettes) and continuous per 5 pack-years. For case patients, the index date was defined as the date of diagnosis. For control subjects, the index year was the participation date. The exposure to smoking was assessed for the period of time before the index date.

Information on background and lifestyle factors was derived from questionnaires. BMI was calculated as weight in kilograms divided by the square of height in meters. Overweight was defined as a BMI of ≥25 kg/m². Alcohol consumption was assessed in the following four categories: nonalcohol drinkers and alcohol drinkers (0.1–4.9 g/day, 5–14.9 g/day, and ≥15 g/day). Highest achieved education was categorized into the following three levels: low (primary school), medium (upper secondary school), or high (university). Information on average leisure time physical activity during the preceding year was collected by a question with four response options, ranging from inactive/sedentary lifestyle (<2 h/week) to very active. The family history of diabetes was positive if any first-degree relative had any type of diabetes.

Using conditional logistic regression, the impact of smoking on the risk of LADA and type 2 diabetes was investigated by calculating the odds ratio (OR) and its 95% CI. An association was considered significant if the 95% CI for an OR did not include the null value of 1, which corresponds to a P value of 0.05 based on a two-tailed test. The incidence density sampling design of this case-control study (matched for date of participation and residential area) allowed us to interpret ORs as incidence rate ratios (12). All analyses were adjusted for age (continuous), sex, BMI (continuous), alcohol consumption, and family history of diabetes, and were performed with SAS version 9.4. Further adjustment for physical activity and education did not change the ORs (<10%). The Kruskal-Wallis H test was used to assess differences in the median levels of GADA across categories of smoking. We also compared differences in the mean levels of HOMA indices across categories of smoking by using t tests. Also, a linear regression model was used to assess the association between smoking HOMA indices (lnHOMA-IR and lnHOMA-β; logarithmic transformation was applied because of the skewing of variables), adjusted for age, sex, BMI, alcohol consumption, and family history of diabetes.

The mean age was 63 years in type 2 diabetes patients, 59 years in LADA patients, and 58 years in control subjects (Table 1). Among diabetes patients, 62% of patients with type 2 diabetes and 55% of patients with LADA were ever smokers, compared with 51% among control subjects. Compared with LADA patients, type 2 diabetes patients were heavier, less physically active, more likely to be insulin resistant, and had higher levels of HOMA-β. Only 6% of type 2 diabetes patients were treated by insulin, compared with 43% of LADA patients. The median duration from diabetes diagnosis to the time at which patients participated in the study was 5.4 months among type 2 diabetes patients, and 7.7 months among LADA patients.

Smoking was associated with an increased risk of type 2 diabetes; the OR was estimated at 1.50 (95% CI 1.14–1.97) in current smokers (Table 2), and current smokers of ≥20 cigarettes/day had a 73% higher risk (OR 1.73, 95% CI 1.00–2.99) compared with never smokers. Every 5 pack-years of smoking was associated with a 9% increased risk of type 2 diabetes (OR 1.09, 95% CI 1.05–1.13). Long-time heavy smokers (≥15 pack-years) had an almost twofold increased risk of type 2 diabetes (OR 1.90, 95% CI 1.49–2.41). Raising the cutoff (≥30 pack-years) resulted in an OR of 2.05 (95% CI 1.48–2.84). The association appeared to be more pronounced in men (OR for ≥15 pack-years 2.01, 95% CI 1.46–2.77) than in women (OR for ≥15 pack-years 1.69, 95% CI 1.16–2.47).

Heavy smokers compared with never smokers had higher levels of HOMA-β (75.8 vs. 64.4, P = 0.0003); however, there was not a clear difference in HOMA-IR levels between heavy smokers and never smokers (6.10 vs. 6.00, P = 0.9141).

Among LADA patients taken as a whole, there was a tendency toward an increased risk of LADA associated with current smoking (OR 1.26, 95% CI 0.93–1.72), and the risk was more pronounced in long-time heavy smokers (OR for >15 pack-years 1.37, 95% CI 1.02–1.84) (Table 3). A 7% increased risk was seen for every 5 pack-years of smoking in men (OR 1.07, 95% CI 1.01–1.13), but not in women (OR 0.98, 95% CI 0.91–1.05). Further analysis revealed that the increased risk of LADA was restricted to men who were long-time heavy smokers (≥15 pack-years) (OR 1.61, 95% CI 1.09–2.40), but no increased risk was seen in women (OR 1.06, 95% CI 0.67–1.67) or in light smokers (OR for <15 pack-years 0.95, 95% CI 0.70–1.29). In men with ≥30 pack-years of smoking, an OR of 1.93 (95% CI 1.11–3.34) was observed.

We also performed further subgroup analysis of the association between smoking and LADA stratified by family history of diabetes (yes, no), BMI (<25, ≥25 kg/m²), age (<55, ≥55 years), or median level of GADA (<178, ≥178 IU/mL), separately in light and heavy smokers (Fig. 1). The risk of LADA appeared to be unrelated to light smoking (<15 pack-years) in any subgroup analyses. The excessive risk of LADA related to heavy smoking (≥15 pack-years) appeared to be more pronounced in individuals who were younger (OR 1.74, 95% CI 1.04–2.90) and not overweight (OR 1.76, 95% CI 1.02–3.02), and who did not have a family history of diabetes (OR 1.53, 95% CI 1.05–2.23). Stratification by median GADA levels revealed that the results only pertained to LADA with low GADA levels (OR 1.59, 95% CI 1.08–2.35), whereas no clear association was seen with LADA with high GADA levels (OR 1.17, 95% CI 0.78–1.75).

Among LADA patients, heavy smokers, compared with never smokers, had higher levels of HOMA-IR (9.89 vs. 4.38, P = 0.0479) and HOMA-β (55.7 vs. 42.5, P = 0.0204). Moreover, linear regression adjusted for age, sex, BMI, alcohol consumption, and family history of diabetes indicated that LADA patients who were heavy smokers had 47% higher levels of HOMA-IR compared with never smokers (expβ = 0.47483, P = 0.0241). The median level of GADA was lower in heavy and long-time smokers compared with never smokers (75 vs. 250, P = 0.0445).

We confirm that smoking, particularly heavy smoking, is associated with an increased risk of type 2 diabetes (1). We did not observe a beneficial effect of smoking on the risk of LADA. These results are in contrast to our previous findings based on the Norwegian HUNT study (3), and the results were consistent across different strata and subsets of LADA. At first glance, the results from these two studies appear contradictory. However, LADA is a heterogeneous disease (10), with an etiology that includes both IR and autoimmunity. It is possible that smoking has a beneficial effect on autoimmunity, but a detrimental influence on insulin sensitivity, and the net effect may be different in different populations. Hence, the absence of a protective effect by smoking on LADA in ESTRID could be due to masking by the negative effect of smoking on insulin sensitivity (2). In line with this, we did see that heavy smoking was associated with higher levels of HOMA-IR, and also with better HOMA-β and lower GADA levels. There are few and inconclusive studies (18–20) of the effect of smoking/nicotine on β-cell function. However, a biphasic action and dose-dependent effect of smoking/nicotine has been suggested. Some studies (19,20) reported that long-term smoking was associated adversely and dose-dependently with β-cell function, whereas in another study β-cell function differed by smoking status; current smoking was associated with higher β-cell function, and former smoking with lower β-cell function (18). It might be hypothesized that smoking in low doses and with short-term exposure may work differently than high doses and long-term use. HOMA indices are, however, proxy measures for IR and β-cell function, and the results need to be interpreted cautiously (21,22).

The Swedish and Norwegian LADA patients investigated here and previously in the HUNT study were similar with regard to most clinical characteristics (Supplementary Table 1). With regard to genetic factors, LADA is commonly described as a genetic mix of type 1 and type 2 diabetes (10). In contrast to most LADA populations, including the one from Scania (23), LADA patients from the HUNT study do not display an association with the type 2 diabetogenic allele TCFL7 (24). This may indicate that the Norwegian LADA patients are more like type 1 diabetes patients. One speculation is that the stronger the impact of type 2 diabetes risk factors for the development of LADA, the weaker the possibility of detecting a clear protective effect that impacts autoimmunity.

The effect of smoking is also controversial with regard to other autoimmune disorders. It has been shown that cigarette smoking increased the risk of systemic lupus erythematosus (25), rheumatoid arthritis (26), and multiple sclerosis (27), but reduced the risk of type 1 diabetes (4,28) and some inflammatory bowel diseases (25). In LADA, like other autoimmune disorders, different combinations of genes, immune reactions, and environmental factors are involved in the development of disease (10), meaning that the effect of environmental triggers may be different in different subsets of disease. It has also been shown that smoking (or some compounds in cigarette smoke) decreases the level of specific antibodies (6) and can suppress inflammation (9), but contradictorily it can also enhance immune reactions (6,29) and act as a proinflammatory factor (7). This could explain, at least in part, the conflicting results regarding the influence of smoking on LADA, as well as in other autoimmune disorders (25).

The strengths of this study include the population-based design and the large number of LADA patients, almost three times as many compared with the HUNT study. One limitation is the use of retrospective information on smoking. Bias will be introduced if patients recall information on smoking habits in a different way than control subjects, or if early-stage symptoms of diabetes lead to lifestyle changes, including smoking cessation. To minimize this potential bias, we only included information on smoking until the index year, which was 1 year prior to diagnosis. If patients exaggerated their previous smoking habits, this could explain the excess risk observed in ever smokers. Still, in order to replicate the findings in the HUNT study of a 58% reduced risk (heavy vs. never smokers), two-thirds of the LADA patients in ESTRID would have to report heavy consumption instead of no consumption, which does not seem likely. Selection bias would be introduced if the smoking habits of participating control subjects differed from those of the population that generated the case patients (12). We find this unlikely, since the prevalence of smoking among control subjects was comparable to that in reports based on the general population living in Scania County (30). In this context, it is noteworthy that our results for type 2 diabetes were in agreement with reports from numerous prospective studies where information on smoking habits is collected several years before the onset of disease, including results based on HUNT study (1,3). Differences in diagnostic criteria between the HUNT study and ESTRID are unlikely to have contributed to the conflicting results because, in both studies, GADA positivity was used to indicate autoimmunity. However, the cutoff level was higher in the HUNT study than in ESTRID (43 vs. 10 World Health Organization units/mL) (13,31,32), and raising the cutoff level does not change the results.

In conclusion, we did not observe a reduced risk of LADA in smokers in the ESTRID population. Further, heavy smoking may increase the risk of LADA through increasing IR. Our present and previous findings highlight the heterogeneity of LADA ranging from very autoimmune to more type 2 diabetes like and IR patients with clear implications for the total impact of smoking on the risk of disease.

Funding. ESTRID was funded by grants from the Swedish Medical Research Council; the Swedish Research Council for Health, Working Life and Welfare; AFA Insurance Company; and the Swedish Diabetes Association. The ANDIS study is funded by grants from the Swedish Medical Research Council and ALF-Swedish Research Council funding for clinical research. Funding for the ANDIU study was provided by the Swedish Medical Research Council, a strategic Research Grant from the Swedish Government (Excellence of Diabetes Research in Sweden-EXODIAB).

Duality of Interest. No potential conflicts of interest relevant to this article were reported. All authors declare no support from any organization for the submitted work, no financial relationships with any organizations that might have an interest in the submitted work, and no other relationships or activities that could have influenced the submitted work.

Author Contributions. B.R. developed the objective of the study, analyzed the data, and wrote the manuscript. T.A. contributed to data analysis, and reviewed and revised the manuscript. P.-O.C. and L.G. researched the data, and reviewed and revised the manuscript. V.G., M.M., P.S., and T.T. contributed to the discussion and reviewed and revised the manuscript. S.C. researched the data, contributed to developing the objective of the study and interpretation of the results, and reviewed and revised the manuscript. All authors read and approved the final version of the manuscript. B.R. 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.