Incidence of LADA and Type 2 Diabetes in Relation to Tobacco Use and Genetic Susceptibility to Type 2 Diabetes and Related Traits: Findings From a Swedish Case-Control Study and the Norwegian HUNT Study

Latent autoimmune diabetes in adults (LADA) is a common, heterogenous form of diabetes distinguished by a type 1 diabetes–like autoimmune reaction against the insulin-producing β-cells, and a type 2–like phenotype often including obesity and insulin resistance (1). Genetically, LADA is closely related to type 1 diabetes, with a high risk conferred by the HLA gene complex. A genetic overlap with type 2 diabetes exists but is less evident (2). For example, associations with variants of the type 2–associated TCF7L2 gene have been observed, primarily in individuals with glutamic acid decarboxylase antibody (GADA) titers below median (3). The hybrid nature of LADA suggests that environmental factors having either a triggering or promoting effect on autoimmunity, and/or factors known to increase insulin resistance, could enhance the risk. To date, lifestyle risk factors are scarcely investigated in LADA, and neither is their potential interaction with genetic susceptibility (1). However, we recently observed an increased risk of LADA in smokers and users of Swedish smokeless tobacco (snus), using data from two observational studies and a Mendelian randomization study (4), and the excess risk was comparable to that seen in type 2 diabetes (5). Moreover, smoking was associated with insulin resistance in LADA, which is in line with observations in type 2 diabetes (6).

Previous studies on smoking and genetic susceptibility in type 2 diabetes have primarily been carried out in Asian populations, where some have revealed interaction between smoking and type 2–associated single nucleotide polymorphisms (SNPs) (7,8). Because the genetic landscape of type 2 diabetes includes many common genetic variants, each with minor effects (9), susceptibility to type 2 diabetes may be better captured by combining several variants into a genetic risk score (GRS). Interaction between such GRS and smoking has been observed in Asian studies (10,11), but not in a study of European-ancestry populations (12).

Our aim is to investigate whether GRS of type 2 diabetes (T2D-GRS), impaired insulin secretion (IS-GRS), and insulin resistance (IR-GRS) are associated with LADA and to explore potential interaction between tobacco use and GRS in relation to LADA and type 2 diabetes.

The Swedish case-control study Epidemiological Study of Risk Factors for LADA and Type 2 Diabetes (ESTRID) collects data on case subjects with LADA and type 2 diabetes from a diabetes register and biobank (All New Diabetics in Scania [ANDIS]) (13) covering >90% of the health care system in the county of Scania, Sweden. In addition, ESTRID enrolls incidence-density sampled control individuals without diabetes through the Swedish Population Register. All participants fill out a health and lifestyle questionnaire, administered close to diagnosis of case subjects. Genetic information was available for case subjects but not control subjects. We therefore include population-based control subjects from the Epidemiological Investigation of Rheumatoid Arthritis (EIRA) study (14), for whom we have both genetic and questionnaire information. These control subjects are free from rheumatoid arthritis and diabetes at inclusion.

We include all case subjects with LADA (n = 594 [n = 345 with genetic information]) and type 2 diabetes (n = 2,045 [n = 1,122]) enrolled in ESTRID 2010–2019, and all control subjects aged ≥35 years in EIRA 2006–2014 (n = 3,068 [n = 1,537]) with information on tobacco use. All participants gave informed consent, and ethical approvals were obtained from the regional ethical review board in Stockholm. The study design is illustrated in Supplementary Figure 1.

Case subjects were diagnosed within the health care system of Scania, and fasting blood samples were drawn at registration. GADA concentrations were measured using an enzyme-linked immunosorbent assay (ELISA; RSR Limited, Cardiff, U.K.). At the chosen cutoff (≥10.7 U/mL), sensitivity and specificity were 0.84/0.98. Values above 250 U/mL were censored. LADA case subjects were aged ≥35 years at diagnosis, GADA positive (≥10 U/mL), and with C-peptide of ≥0.20 nmol/L (IMMULITE 2000; Siemens Healthcare Diagnostics Product Ltd., Llanberis, U.K.) or ≥0.30 nmol/L (Cobas e601; Roche Diagnostics, Mannheim, Germany). Except for C-peptide, used to distinguish LADA from type 1 diabetes, this is consistent with the criteria put forward by the Immunology of Diabetes Society (15), who proposes a minimum age cutoff of 30 years to exclude type 1 diabetes. We arbitrarily chose to use age 35 years to minimize such misclassification. Case subjects with type 2 diabetes were aged ≥35 years, GADA negative, with C-peptide of >0.60 nmol/L (IMMULITE) or >0.72 nmol/L (Cobas).

Genotyping in ANDIS was done on frozen DNA samples using iPlex (Sequenom, San Diego, CA) or TaqMan assays (Thermo Fisher Scientific, Carlsbad, CA) at the Clinical Research Centre in Malmö, Sweden (13). DNA samples from the EIRA biobank were analyzed with the Illumina Global Screening array or an Infinium Illumina 300 K immunochip custom array (Illumina, San Diego, CA).

The Trøndelag Health Study (HUNT) (16) invited the entire population ≥20 years of age in the Nord-Trøndelag region, Norway, to an extensive health survey on up to four occasions: HUNT1 (1984–1986), HUNT2 (1995–1997), HUNT3 (2006–2008), and HUNT4 (2017–2019). Participants responded to questionnaires and underwent clinical examination and blood sampling. Eligible for inclusion in the current study were all participants with baseline data in HUNT1 to HUNT3 (n = 94,489). Excluding those with diabetes at baseline (n = 2,481) and those without smoking information (n = 13,466), the analytical sample consisted of 78,542 unique individuals and 1,696,503 person-years (Supplementary Figure 1). The study was approved by the Norwegian Data Protection Authority, Oslo, and the Regional Committee for Medical and Health Research Ethics, Trondheim, and all participants provided informed consent.

Identification of case subjects with diabetes was done through self-report (17). GADA was assessed in blood samples collected at follow-up (median 5 years after diagnosis), analyzed by either immunoprecipitation radioligand assay (Novo Nordisk, Bagsværd, Denmark; HUNT2 and HUNT3) (18) or ELISA (RSR Limited, Cardiff, U.K.; HUNT4) with sensitivity and specificity of 0.64/1.00 and 0.84/0.98, respectively.

To harmonize with ESTRID, case subjects aged ≥35 years were classified as LADA (n = 245 [n = 219 with genetic information]) if they were GADA positive (≥0.08 antibody index [World Health Organization; ≥43 U/mL] in HUNT2 and HUNT3, ≥10 U/mL in HUNT4) and as type 2 diabetes (n = 3,726 [n = 3,421]) if GADA negative. The classification excluded C-peptide, since measurements were not from the time of diagnosis.

DNA was extracted and genotyped at the Norwegian University of Science and Technology Genomic Core Facility, Trondheim, by HumanCoreExome, Illumina Inc. (San Diego, CA).

We reproduced previously published GRS of type 2 diabetes (64 SNPs), insulin secretion (16 SNPs), and insulin resistance (5 SNPs) (13). Conferred genotypes were obtained as hard-call dosages with the threshold set to 0.5. The GRS were calculated based on the number of risk alleles in each score weighted by their effect sizes, as reported in previous genome-wide association studies (GWAS) (13) using Plink software (19). Unavailable SNPs (n = 11 [17.2%] in T2D-GRS and n = 1 [6.3%] in IS-GRS in ESTRID; n = 1 [1.6%] in T2D-GRS in HUNT) were replaced by proxy SNPs with r² >0.8, except one with r² of 0.61 in the T2D-GRS (ESTRID only). The resulting GRS were z-score standardized when used as continuous exposures.

GRS were treated as continuous variables and measured as increase per SD and as categorical or binary variables based on tertiles. Smoking and snus use were self-reported and analyzed both categorically (smoking/snus use habit; heavy use [≥15 pack-/box-years]) and for continuity (number of pack-/box-years in ever users; current intensity). One pack/box-year was defined as smoking one pack of cigarettes (20 cigarettes) or using seven boxes of snus per week for a year. Total tobacco use was defined as being either a smoker or snus user, and the total number of tobacco-years was the sum of pack-years and box-years. We set the index date for tobacco use to one year prior to diagnosis/participation in ESTRID.

Self-reported educational level (primary school, upper secondary school, or university) and alcohol intake were included as covariates. In ESTRID, beverage-specific responses about the amount and frequency consumed during the preceding year were available. Corresponding questions in HUNT related to frequency of intake during the past 2 weeks up to the past year. Alcohol intake was divided into six and four categories, respectively, in ESTRID and HUNT, ranging from abstainers to high consumers. We also included BMI as a continuous covariate and added a quadratic term. In ESTRID, self-reported weight and height were used for the calculation of BMI (weight [kg]/height [m²]), which correlates well (r = 0.92) with BMI measured at diagnosis (20). In HUNT, BMI was based on weight and height measured at the clinical examination. Adjustment for waist-to-hip ratio (waist [cm]/hip [cm]), available for a subgroup in HUNT, physical activity, and family history of diabetes did not markedly alter the estimates and were not included in the final models. Exposure variables and covariates were updated, if possible, with the information at each survey for HUNT participants who participated in more than one survey.

Differences in baseline characteristics were calculated by Student’s t test for means (± SD), Kruskal-Wallis test for medians (interquartile range), and χ² test for proportions, and expressed as two-sided P values.

We conducted logistic regression conditioned on age and sex to estimate odds ratios (ORs) with 95% CIs for case-control data. On the cohort data, we performed Cox regression with age as the underlying timescale to estimate hazard ratios with 95% CIs. Participants contributed with person-years up to diabetes diagnosis, death, emigration out of the county, or end of follow-up, whichever came first. Associations between GRS and LADA/type 2 diabetes were adjusted for age and sex in the main model (model 1), and separate estimates were given for LADA_high (GADA ≥ median) and LADA_low (GADA < median). In sensitivity analyses, GRS were adjusted for HLA (model 2) and additionally, in HUNT, for the first 10 principal components and batch number (model 3). The association with tobacco use, as well as interaction analyses, were adjusted for BMI, educational level, and alcohol intake in addition to age and sex, and analyses of snus use were further adjusted for smoking. To account for the association between BMI and insulin resistance, the interaction between smoking and IR-GRS was further stratified by BMI (<30 or ≥30). For those with missing information on any of the covariates, we performed imputation. In ESTRID, this was based on the median value, and, in HUNT, we first used the value from a previous survey, if available, and, secondly, the median value. We included indicator variables for missing covariate information.

The exposure in the models of additive interaction was the combination of binary variables of tobacco use and GRS, resulting in four combinations, with those with the lowest category of tobacco use and GRS serving as the reference group. Additive interaction was defined as departure from additivity of effects and estimated as attributable proportion (AP) due to interaction with 95% CI, representing the proportion in the doubly exposed case subjects that can be attributed to the interaction between the two exposures.

Multiplicative interaction was examined by including product terms for binary tobacco use and GRS. By fitting these models, we derived estimates contrasting the risk of LADA and type 2 diabetes associated with tobacco use in those with high versus low/intermediate genetic susceptibility. A P value of <0.05 for the product term coefficient indicates a statistically significant difference in the association across strata of GRS. The strongest type of interaction is one which is present on both the additive and multiplicative scales (21). When possible, estimates from both studies were meta-analyzed using a fixed-effects model, resulting in relative risks (RR_pooled) for the combined risk. Corresponding APs were obtained by pooling the study-specific APs (AP_pooled) based on their respective weights, and we calculated P values for the pooled coefficients of the product terms using the log-transformed estimates and CIs. All analyses were performed in SAS 9.4 (SAS Institute, Cary, NC).

The analyzed data sets are available from the corresponding author upon reasonable request (ESTRID) and with permission from the HUNT Study by applying to the HUNT Study data access committee.

The two study populations consisted of 839 individuals with LADA, 5,771 with type 2 diabetes, and 3,068 control subjects (ESTRID), along with 1,696,503 person-years at risk (HUNT). Those with LADA were, on average, less insulin resistant (HOMA-IR) and had worse β-cell function (HOMA-B) and lower C-peptide concentrations than individuals with type 2 diabetes. They were also more likely to have a high-risk HLA genotype and to be treated with insulin (Table 1). From here onward, we refer to the fully adjusted models, unless otherwise stated.

The T2D-GRS, IS-GRS, and IR-GRS were approximately normally distributed in the two study populations (Supplementary Figure 2). The T2D-GRS was associated with type 2 diabetes (RR_pooled 1.35 [CI 1.31, 1.40] per SD increase) and LADA (RR_pooled 1.15 [CI 1.05, 1.26] per SD increase) (Fig. 1). The IS-GRS (RR_pooled 1.20 [CI 1.16, 1.23] per SD increase) and IR-GRS (RR_pooled 1.13 [CI 1.10, 1.17] per SD increase) were associated with type 2 diabetes, but not with LADA (Fig. 1). Study-specific estimates were similar and did not change markedly with adjustment for HLA, principal components, and genotyping batch (Supplementary Tables 1 and 2). When LADA was stratified by median GADA concentrations, the T2D-GRS and IS-GRS were associated with LADA_low but not with LADA_high, and there was no association with IR-GRS (Supplementary Table 3). Study-specific estimates pointed in the same directions (Supplementary Tables 4 and 5). Smoking, snus, or tobacco use were associated with higher risk of LADA and type 2 diabetes (Supplementary Figure 3).

There was interaction between tobacco use and IR-GRS in the risk of LADA, which was evident in current and heavy smokers and current and heavy total tobacco users (Table 2, Fig. 2C, and Supplementary Table 6). The combination of current heavy smoking and a high IR-GRS conferred an RR_pooled of 2.01 (CI 1.30, 3.10) with significant additive (AP_pooled 0.67 [CI 0.46, 0.89]) and multiplicative interaction (p = 0.003) (Table 2). Current, heavy total tobacco users with a high IR-GRS had an OR of 2.59 (CI 1.54, 4.35) with AP estimated at 0.52 (CI 0.21, 0.83) and a p for multiplicative interaction of 0.034 (Supplementary Table 6). Study-specific analyses revealed additive interaction between heavy smoking and IR-GRS in the risk of LADA (Supplementary Tables 7 and 8). The results did not differ markedly when stratified by BMI (<30 or ≥30) (Supplementary Table 9). Furthermore, there was no significant interaction between snus use and IR-GRS (Supplementary Table 6).

Additive interaction was observed between the T2D-GRS and ever heavy smoking (AP_pooled 0.38 [CI 0.09, 0.67]), ever (AP 0.72 [CI 0.28, 1.15]) and current (AP 0.67 [CI 0.14, 1.21]) heavy snus use, and heavy total tobacco use (AP 0.49 [CI 0.20, 0.78]) in ever users (Table 2, Fig. 2A, and Supplementary Table 6). A joint exposure to heavy tobacco use and a high T2D-GRS compared with no heavy tobacco use and a low T2D-GRS in ever users resulted in an OR of 2.60 (CI 1.62, 4.17) for LADA, with evidence of additive (AP 0.49 [CI 0.20, 0.78]) and multiplicative (P = 0.038) interaction (Supplementary Table 6). Study-specific analyses of heavy smoking and T2D-GRS yielded similar results, although only significant in the Swedish data (Supplementary Tables 7 and 8). Smoking, snus use, and total tobacco use did not interact with IS-GRS in LADA (Table 2, Fig. 2B, and Supplementary Table 6).

In type 2 diabetes, there was no evidence of interaction between smoking, snus use, or total tobacco use and any of the GRS, neither on the additive nor the multiplicative scale, except for moderate additive interaction between current, heavy smoking and a high IR-GRS (AP_pooled 0.18 [CI 0.03, 0.33]) (Fig. 2D–F and Supplementary Tables 10 and 11). Study-specific results indicated interaction with IR-GRS in ESTRID but not in HUNT (Supplementary Tables 12 and 13). The interaction between smoking and IR-GRS did not differ when stratified by BMI (Supplementary Table 14).

We used data from two Scandinavian population-based studies and showed that a GRS of type 2 diabetes, but not impaired insulin secretion or insulin resistance, was associated with incidence of LADA. We also found that genetic susceptibility to type 2 diabetes or insulin resistance strengthened the association between tobacco use and LADA risk. The strongest evidence of such interaction was observed between smoking and genetic predisposition to insulin resistance, which was seen in both cohorts and visible on both the additive and multiplicative scale. Genetic susceptibility to type 2 diabetes, insulin resistance, or insulin secretion had limited influence on the association between tobacco use and incidence of type 2 diabetes.

We confirmed findings from GWAS (22) and candidate gene studies (23,24) linking genetic predisposition to type 2 diabetes to an elevated risk of developing LADA, and this was primarily seen in individuals with GADA titers below median. LADA was not associated with genetic susceptibility to impaired insulin secretion or insulin resistance, which is consistent with previous findings from the ANDIS study, where the GRS were examined in relation to severe autoimmune diabetes, including LADA and adult-onset type 1 diabetes (13).

This study is the first investigating the interaction between tobacco use and genetic susceptibility to type 2 diabetes and related traits in LADA. Genetically, LADA is a heterogeneous condition with an etiology that encompasses susceptibility to both type 1 and type 2 diabetes. Our findings indicate that smoking primarily increases the risk of LADA in individuals with a genetic predisposition to type 2 diabetes or insulin resistance. Interestingly, we recently also observed interaction between smoking and HLA genotypes associated with type 1 diabetes in relation to LADA (4). This supports the hypothesis that in LADA, the influence of environmental factors increasing insulin resistance is contingent on genetic predisposition. Whether smoking also interacts with LADA- and type 1 diabetes–associated loci outside the HLA region remains to be investigated.

In contrast to our findings in LADA, the association between smoking and type 2 diabetes seen across strata of genetic susceptibility and the limited evidence of interaction indicate that tobacco use and genetic susceptibility are independently associated with type 2 diabetes risk. Our results correspond with previous findings in populations of primarily European descent (12) but are in contrast to findings in Asian populations (10,11). These discrepancies may be due to differences in genetic risk variants, pathways, and minor allele frequencies across populations (25,26). In general, observational studies have, so far, provided inconsistent evidence regarding interaction between lifestyle factors and type 2 diabetes polygenic risk scores in type 2 diabetes (27).

Adverse effects of smoking and nicotine exposure on insulin sensitivity have been documented in experimental studies (28,29). In line with this, we recently showed that smoking, snus, and total tobacco use are positively associated with HOMA-IR in both LADA and type 2 diabetes (4). Against this background, it is interesting that the strongest evidence of interaction was observed between smoking and IR-GRS in LADA. Genetic susceptibility to insulin resistance also seems to increase the risk of LADA only in individuals who are simultaneously exposed to tobacco. In addition, there was some indication of additive interaction between heavy smoking and IR-GRS in type 2 diabetes. This suggests that inherited and acquired insulin resistance may synergistically accelerate the onset of LADA and possibly also type 2 diabetes. The stronger interactions observed with IR-GRS in LADA, and, to a lesser extent, in type 2 diabetes, indicate that specific pathways involved in insulin resistance may play a significant role in these interactions. To the best of our knowledge, this study is the first to address interaction between a trait-specific GRS and tobacco use in type 2 diabetes.

The strengths of this study include the large number of incident case subjects with LADA and type 2 diabetes, the population-based design, detailed lifestyle and clinical information, the use of GRS that capture overall type 2 diabetes genetic risk as well as specific diabetes-related traits, and a follow-up time of up to 35.5 years in the cohort study. We also had the opportunity to replicate the results in two independent cohorts.

Recall bias is a concern in ESTRID, since case subjects may retrospectively report tobacco habits differently from control subjects. However, the association between smoking and type 2 diabetes in ESTRID was similar to that observed in HUNT and numerous previous prospective studies (5). Moreover, self-reported smoking correlates well with gold standard measures of blood cotinine concentrations (30).

We could not separate LADA from adult-onset type 1 diabetes in HUNT, because of the lack of C-peptide measurements from diagnosis and limited information on treatment. This implies that some individuals with adult-onset type 1 diabetes are included in the LADA group. We based the LADA classification on the presence of GADA, for which >90% of case subjects are positive (31). Individuals positive only for other autoantibodies will be misclassified as having type 2 diabetes, but this number is likely small and will have minor influence on the results. However, with the specificity of the GADA assay, it follows that some case subjects with type 2 diabetes will be incorrectly categorized as having LADA. Such misclassification could explain why the T2D-GRS was associated with LADA_low and not with LADA_high. It is, however, unlikely to be driving the observed interactions in LADA, given that these interactions were almost absent in type 2 diabetes.

The GRS used in this study were developed based on published GWAS from European cohorts, which justifies their application on our two study populations that are of primarily European ancestry. More recent GWAS have revealed previously unknown type 2 diabetes–associated loci, not captured by the current T2D-GRS, but we included the loci with the largest effect sizes. Moreover, there are methods for calculating a GRS that may better predict the outcome than a weighted sum approach. However, by replicating these GRS and the method used to calculate them, we can compare our findings with previous findings in subgroups of adult-onset diabetes (13). Our results in LADA were on par with estimates in severe autoimmune diabetes, reassuring the validity of the data and the use of proxy SNPs.

The T2D-GRS includes genetic variants associated with diverse facets of the disease, whereas IS- and IR-GRS pinpoint more specific traits (for further information on included SNPs, we refer to the GWAS catalog [https://www.ebi.ac.uk] and previous association studies). Thus, by incorporating two trait-specific GRS, we could approach potential mechanisms underlying the observed gene-environment interactions. In fact, it has been proposed that the application of GRS targeting specific biological mechanisms may be better aimed at revealing interaction effects due to the etiological complexity of type 2 diabetes (27). It can also be noted that genetic variants in the IR-GRS have been linked to smoke-induced insulin resistance (IRS1) (32) and significantly higher HOMA-IR in obese carriers (PPARG) (33).

A limitation is that we could not adjust the GRS for population genetic structures or genotyping array in the case-control data. However, when adjusting for these factors in the cohort data, it did not change the estimates for the associations between the GRS and LADA and type 2 diabetes. Moreover, estimates were similar for the associations in both data sets, strengthening reliability in the genetic data. Finally, it is important to acknowledge that these results were retrieved in European populations from relatively homogenous environments. They cannot be generalized to non-European populations because there may be differences in genetic makeup and risk factors.

Our results suggest that smoking prevention could potentially inhibit or prolong the development of LADA, especially in individuals with genetic predisposition to insulin resistance or type 2 diabetes. In contrast, tobacco use seems to be a risk factor for type 2 diabetes independent of genetic predisposition. Our findings highlight the importance of tobacco use prevention to reduce the individual and societal burden of LADA, a chronic disease. Replications of the findings are warranted, also in type 1 diabetes, to further promote the discussion on possible underlying mechanisms. Nevertheless, our findings contribute to the sparse body of knowledge about the interplay between environmental and genetic factors in the etiology of LADA.

Acknowledgments. The authors thank the participants in ESTRID, EIRA, ANDIS, and HUNT as well as administrative personnel, nurses, and research team members from all the studies.

Funding. The ESTRID study was funded by grants from the Swedish Research Council (2018-03035); FORTE (2018-00337); the Novo Nordisk Foundation (NNF19OC0057274); and the Swedish Diabetes Foundation (DIA2022-735). EIRA was funded by the Swedish Research Council; FORTE; the Swedish Rheumatic Foundation; the AFA Insurance Company; and Stockholm County Council. ANDIS was financially supported by the Swedish Research Council (project grants 2020-02191 and 2015-2558 and infrastructure grants 2010-5983, 2012-5538, and 2014-6395), Linnaeus grant 349-2006-237, a strategic research grant 2009-1039 (EXODIAB), Swedish governmental funding of clinical research (ALF), and The Swedish Foundation for Strategic Research (IRC15–0067). B.R. is the recipient of a post-doctoral fellowship from the Novo Nordisk Foundation (grant NNF17OC0027580). The Trøndelag Health Study (HUNT) Study is a collaboration between HUNT Research Centre (Faculty of Medicine and Health Sciences, Norwegian University of Science and Technology [NTNU]), Trøndelag County Council, Central Norway Regional Health Authority, and the Norwegian Institute of Public Health. The genetic investigation of the HUNT Study is a collaboration between investigators from the HUNT Study and University of Michigan Medical School and the University of Michigan School of Public Health. The K.G. Jebsen Center for Genetic Epidemiology is financed by Stiftelsen Kristian Gerhard Jebsen, Faculty of Medicine and Health Sciences, NTNU, and Central Norway Regional Health Authority. The genotyping was financed by the National Institutes of Health, University of Michigan, The Norwegian Research Council, and Central Norway Regional Health Authority and the Faculty of Medicine and Health Sciences, NTNU. The genotype quality control and imputation have been conducted by the K.G. Jebsen Center for Genetic Epidemiology, Department of Public Health and Nursing, Faculty of Medicine and Health Sciences, NTNU.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. Contributions to the data collection were made by J.E., B.R., and S.C. (ESTRID); E.A. and T.T. (ANDIS); L.A. (EIRA); and V.G., E.P.S., and B.O.Å. (HUNT). S.C. was responsible for conceptualizing the research objectives, designing the study, and thoroughly revising the manuscript. D.M.A. developed the GRS. J.E. and D.M.A. calculated the GRS in the study. J.E. developed the objectives of the study and was responsible for drafting the manuscript and analyzing the data. All authors approved the final version of the manuscript. J.E. 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 at the following meetings: 56th Annual Meeting of the European Diabetes Epidemiology Group (EDEG), Hersonissos, Greece, 2–5 April 2022; 56th Annual Meeting of the Scandinavian Society for the Study of Diabetes, Reykjavik, Iceland, 12–13 May 2022; the 58th Annual Meeting of the European Association for the Study of Diabetes (EASD), Stockholm, Sweden, 19–23 September 2022; and the 17th International Diabetes Epidemiology Group (IDEG) Symposium, Porto, Portugal, 2–5 December 2022.