Does a prior diagnosis of infectious disease confer an increased risk of latent autoimmune diabetes in adults?

Infections are proposed risk factors for type 1 diabetes in children. We examined whether a diagnosis of infectious disease also confers an increased risk of latent autoimmune diabetes in adults (LADA).

Latent autoimmune diabetes in adults (LADA) is the most common form of autoimmune diabetes with an adult onset 1 and is estimated to account for 3%-14% of all diabetes in adults. 2 Genetically, LADA resembles type 1 diabetes, with the highest genetic risk conferred by certain HLA DR-DQ haplotypes. 3However, the autoimmune process leading to beta cell failure and subsequent insulin deficiency is typically slower in LADA, and at the time of diagnosis there may not be an imminent need for insulin treatment.5][6] Moreover, LADA is likely to result from an interaction between environmental and genetic risk factors, 1 but little is known about such environmental risk factors.
Exposure to infections is among the few environmental factors consistently linked to type 1 diabetes. 7Studies based on birth cohorts, including the TEDDY study, 8 the TRIGR cohort, 9 the Norwegian MIDIA study, 10 BABYDIET 11 and DIABIMMUNE 12 find an increased risk of islet autoantibody seroconversion and type 1 diabetes associated with respiratory infections during the first years of life.In addition, type 1 diabetes is linked to severe early life gastroenteritis, 13 bacterial infections, 9 and to herpes simplex infections. 14evious studies have primarily or exclusively addressed infections and type 1 diabetes with onset during childhood.Whether infections in adulthood may promote autoimmune diabetes with adult onset has not been investigated.
Infections caused by viruses, particularly enteroviruses, have drawn much attention in relation to type 1 diabetes.Enteroviruses are a broad group consisting of many serotypes, which replicate in the oropharynx and gastrointestinal tract and can cause both respiratory and gastrointestinal infections.They have consistently been associated with type 1 diabetes 15 and were found in the pancreas of newly diagnosed individuals. 16The role of enteroviruses in the pathogenesis of type 1 diabetes is thought to be either direct through viral infiltration of the pancreas, or indirect by initiating adverse immune responses. 17,18][10][11][12] HLA alleles drive immune responses against infections and a stronger immune response against enterovirus antigens was observed among individuals at high genetic risk. 19Therefore, it is plausible that genetically susceptible children may be at a particularly high risk of type 1 diabetes after infection.Notably, no association was observed between respiratory infections and neither islet autoimmunity nor type 1 diabetes when not restricted to a high-risk population. 13e to the genetic and pathophysiological similarities with type 1 diabetes, we hypothesised that exposure to infections may also promote LADA.We explored this hypothesis using information from a Swedish case-control study with incident cases of LADA, and patient register information on infectious disease diagnoses up to 10 years prior to LADA diagnosis.Additionally, we examined the impact of HLA risk genotypes on potential associations and compared results for LADA to corresponding results for type 2 diabetes, which has a non-autoimmune pathogenesis.

| Study population
The Swedish Epidemiological study of risk factors for LADA and type 2 diabetes (ESTRID) is a population-based case-control study with incident LADA and type 2 diabetes cases, recruited through the All New Diabetics in Scania (ANDIS) register and biobank. 20

| Diabetes classification
All cases were initially diagnosed with diabetes within the usual healthcare system of Scania.Recruitment to ANDIS occurred at the time of diagnosis, and subsequently, blood samples were drawn and fasting glucose and C-peptide analysed. 21Classification of diabetes was based on fixed criteria.Glutamic acid decarboxylase antibodies (GADA) were measured by ELISA (RSR, Cardiff, UK; sensitivity 0.84, specificity 0.98 22

| Infectious diseases
The study population was linked to Swedish national and regional registers through a unique personal identity number which is assigned to individuals upon birth or immigration.ICD 10th Revision (ICD-10) and corresponding primary care codes (ICD-10-P) were used to identify relevant diagnoses from the National Patient Register 24 and the Skåne Health-care register. 25For sensitivity analyses, we retrieved information from the National Prescribed Drug Register (NPDR) 26 on dispensations of antimicrobial drugs, classified by Anatomical Therapeutic Chemical codes.A description of the registers and the study design is found in the supplemental material (Supporting Information S1: page 3, Figure S1, Tables S1 and S2).
We categorised infectious diseases into respiratory (any/upper/ lower), gastrointestinal, herpetic skin and mucous membrane, other (not included in the above categories), and any (including all the above categories) infectious disease (Table S1).We used the exposure windows 0-1, 1-3, 3-5, and 5-10 years before diabetes diagnosis/matching.These periods were chosen to facilitate interpretations of the results in terms of aetiologically relevant periods and reverse causation while considering statistical power.We also calculated the number of diagnoses (0, 1, ≥2) within each exposure window.
Self-reported information on infectious disease (sinusitis, tonsillitis or other throat infection, teeth or gum, influenza, common cold, gastroenteritis, or urinary tract infection) including number of times over the past 10 years (yes/no) was available from the ESTRID questionnaire and categorised into respiratory, gastrointestinal, urinary tract, tooth or gum, or any infection.

| Covariates
Current BMI (weight/height 2 [kg/m 2 ]) was calculated from selfreported height and weight, which correlated well (r = 0.92) with BMI measured at diabetes diagnosis for the cases. 27Self-reported smoking was categorised into never, former, or current, with current including those who quit smoking within 1 year from diabetes diagnosis/matching.Beverage-specific questions on amount and frequency consumed during the preceding year were used to determine alcohol consumption and assessed in six categories from abstainers to high consumers.The level of physical activity (sedentary, low, moderate, or high activity) was assessed from validated questions 28 on leisure time activity and based primarily on current activity before diagnosis/matching.Educational level was categorised into primary, secondary, or tertiary.
From the patient registers, we extracted information on the total number of healthcare contacts and all prior diagnoses of elevated blood glucose, autoimmune diseases and diseases characterised by non-acquired immunodeficiency (Table S1).

| Genetic information
DNA samples from the ANDIS biobank were analysed to retrieve genetic information.Three single-nucleotide polymorphisms (SNPs; rs3104413, rs2854275, rs9273363) were used to determine HLA genotypes with an accuracy of 99.3%. 29

| Statistical analysis
Differences in characteristics between individuals with LADA and type 2 diabetes were calculated by Student's t test for means (�SD), Kruskal-Wallis test for medians (IQR) and χ 2 test for proportions, and given as two-sided p-values.
To study the association between any episode of infectious diseases and LADA within different exposure windows, we conducted logistic regression analyses conditioned on sampling strata to estimate odds ratios (ORs) with 95% confidence intervals (CI).Each sampling stratum included cases of both LADA and type 2 diabetes as well as controls.Model 1 was adjusted for age and sex and Model 2 additionally for lifestyle factors (BMI, smoking, alcohol consumption, physical activity, educational level).Model 3 was further adjusted for the number of health care contacts, comorbidities (LADA only) and elevated blood glucose, and mutually for infectious disease categories.Estimates hereafter refer to the fully adjusted model.Missing information on lifestyle factors was imputed by assigning the median value and adding an indicator variable for missing covariate information.We repeated our analyses using type 2 diabetes as a negative control outcome. 30Furthermore, we investigated whether the associations in LADA were restricted to individuals with high genetic risk, by stratifying the LADA cases by HLA genotypes into HLA high and HLA low .We also stratified the analyses in LADA based on median autoantibody concentration, to examine whether those with a stronger autoimmune reactivity would be more likely to develop LADA after an infection, resulting in separate estimates for GADA high (GADA ≥250 U/mL) and GADA low (GADA <250 U/mL).
To reduce misclassification, we restricted the analyses to infectious diseases with an accompanying dispensation of an antimicrobial drug.We considered disease severity by restricting the analyses to diagnoses in the inpatient register, meaning that the disease was diagnosed during hospitalisation.Finally, we performed analyses on self-reported infectious diseases.All analyses were performed in SAS 9.4 (SAS Institute, Cary, NC, USA).

| RESULTS
The study population consisted of 597 incident cases of LADA, 2065 incident cases of type 2 diabetes, and 2386 controls.Compared to individuals with type 2 diabetes, individuals with LADA were leaner, had less insulin resistance, a worse insulin deficiency and were more often on insulin treatment (40 vs. 5.9%; Table 1).A higher proportion of individuals with LADA carried a high-risk genotype for type 1 diabetes (Table 1). 31
Restricting the analyses to LADA with HLA high-risk genotypes (Figure 2, Table S4) yielded similar results, as did analyses of LADA individuals with high GADA levels (Figure 2, Table S5).Results were similar in analyses of infectious diseases with a dispensed antibiotic (Figure S2) and in analyses restricted to diagnoses from inpatient care (Figure S3).Having ≥2 diagnoses of an infectious disease within the same exposure window was not associated with a higher risk of LADA.An exception was lower respiratory infections in the year before diabetes diagnosis, where two episodes conferred a 4.7-fold increased risk of LADA (95% CI 1.08, 20.15; Figure S4).None of the self-reported infectious diseases was associated with LADA except gastrointestinal infection (OR 0.78, CI 0.63, 0.95) (Table S6 and S7).

| Type 2 diabetes and infectious disease
The risk of type 2 diabetes was elevated in individuals with a diag-  S8).

| DISCUSSION
We found little evidence to support the hypothesis that exposure to infections increases the risk of a future LADA diagnosis.Our study ascertained different types of infectious diseases including, among others, respiratory, gastrointestinal, and herpetic infections that were previously shown to confer a risk of type 1 diabetes.We used several definitions of infectious disease to reduce potential misclassification, with findings being consistent across definitions.Null results were also consistent across LADA endotypes, that is, for those with high-risk HLA types and higher levels of GADA.
Previous studies indicate that respiratory infections are more prevalent in children who later develop islet autoimmunity or type 1 Note: Data are shown as mean � SD or median (IQR) unless stated otherwise.Clinical and genetic information in case subjects was available for 98% (C-peptide), 85% (HOMA), 97% (insulin treatment), 64% (HLA) and 56% (T2D-GRS).The p value is shown for the difference between LADA and type 2 diabetes.
1][12] Given the genetic and pathophysiological relatedness, similar findings in LADA seem plausible.In contrast, we did not observe an association between LADA and either clinically diagnosed or self-reported respiratory disease.Furthermore, we did not observe an association with either gastrointestinal or herpetic infections, previously linked to type 1 diabetes, 13,14 nor with infections accompanied by an antimicrobial dispensation.The exception was an increased risk associated with lower respiratory infectious diseases diagnosed within 1 year before LADA diagnosis.This could hypothetically reflect that such infections trigger autoimmunity or promote progression to LADA.A more plausible explanation, however, is reverse causation, since hyperglycemia, which commonly precedes any diabetes diagnosis 32,33 is likely to increase susceptibility to infections. 34In line with this, we observed similar associations with type 2 diabetes which does not have an autoimmune pathogenesis.
Interestingly, studies in children link respiratory infections to development of islet autoimmunity 8,10,11 ; islet autoimmunity in LADA may occur as early as 10 years before diagnosis. 5,35In this study, we could retrieve diagnoses of infectious disease up to 10 years before diabetes diagnosis, but we did not observe any increased risk of LADA except for infections occurring within a year before diagnosis, that is, the least relevant period in terms of LADA pathogenesis.These

F I G U R E 1
Odds ratios with 95% CI for the association between any event of infectious disease and the risk of LADA within different categories and exposure windows.Reference is no event.Model adjusted for age, sex, BMI, smoking, physical activity, educational level, alcohol consumption, number of health care contacts within the exposure window, elevated blood glucose and comorbidity prior to diagnosis/ matching, and mutually for the other disease categories.Total number of cases (n = 597) and controls (n = 2386).
findings strengthen our reasoning that lower respiratory infections do not trigger LADA but are rather triggered by preclinical hyperglycemia in LADA.Additionally, research conducted in children suggests that recurring 9,11,12 or persistent 36 infections increase the risk of type 1 diabetes.However, we found little evidence of an elevated risk of LADA in individuals with two or more diagnosed episodes of infection within any exposure window.
A potential explanation for our null findings regarding LADA is that we primarily based our study on diagnoses recorded in patient registers and therefore failed to capture all cases of infectious disease.Previous studies on respiratory infections in children were based on parent-reported infections, hence including those that were never diagnosed by a physician.Common viral infections, as those caused by enteroviruses, are often asymptomatic or only manifest as mild illness. 37Particularly in adults, such symptoms may not be regarded as severe enough to require medical attention.On the other hand, misclassification should be a lesser problem in our study since it is based on physician-diagnosed infections.Moreover, a study on influenza and type 1 diabetes 38 gives some support to the possibility that an association may be stronger with more severe infections.This suggests that with our data, we would be likely to observe an association if there was one.However, our results were consistent both over time and in different analyses.
The lack of association between LADA and infectious diseases may reflect that infections have an age-dependent effect on immune responses.The adaptive immune system in infants and children is functionally distinct from that in later life.There may be a critical time in early childhood when the protective effect of maternal antibodies fades away, and the innate and adaptive immune responses are still naïve.In contrast, the adult immune system has developed a persistent memory against many antigens, resulting in more potent immune responses. 39The development of the immune system over the life course suggests that infectious agents threatening to invade the pancreas or trigger an autoimmune reaction may be more efficiently fought off in adulthood.
Lastly, there may be important differences in the pathogeneses of LADA and type 1 diabetes.Although risk variants within the HLA region confer the strongest genetic risk of LADA, the association is weaker than with type 1 diabetes 40 and subsequently, the autoimmune reaction tends to be milder in LADA, as illustrated by slower insulin-deficiency progression and lesser need for insulin treatment at the time of diagnosis. 1While several studies have linked infections to type 1 diabetes in genetically susceptible children, 8,12,19 we observed no association when restricting the analyses to high-risk HLA carriers.Most importantly, we did not observe significant associations in those with the highest autoantibody levels, who would be more severely autoimmune.
To elaborate on potential underlying mechanisms, we compared results for LADA to those in type 2 diabetes, where the pathogenesis does not include autoimmunity.Interestingly, the risk of type 2 diabetes was elevated in individuals with diagnoses of infectious disease prospective study in Latino elderly 42 but in contrast to findings in a Chinese population. 43

| Strengths and limitations
The strengths of this study include the population-based design, the extensive information on previous infectious diseases through primary and specialist care registers, the analyses adjusted for registerbased comorbidity and a range of lifestyle factors, the possibility to account for genetic susceptibility and autoantibody concentrations, and the comparative analyses in type 2 diabetes.
We based our exposure information on Swedish national and regional registers with essentially complete coverage of all diagnoses in primary, inpatient and specialist outpatient care.We also had access to information on all dispensations of antimicrobial drugs prior to diagnosis/matching through the NPDR.This way, we could create a unique study with measures of infectious disease and antimicrobial dispensations collected before diabetes diagnosis and essentially free from recall bias and differential misclassification, along with detailed information on lifestyle factors, within the largest study with incident cases of LADA to date.However, the study design also has limitations.Immigration or emigration within the study period would lead to incomplete register information but any related gaps in register information is likely to be independent of later status as case or control.
We could only capture infections that had resulted in a diagnosis, potentially missing milder, subclinical, or asymptomatic infections that did not require medical attention, and which may be associated with low-grade autoimmunity.However, sensitivity analyses with self-reported infectious diseases generated similar results.
Some studies suggest that certain serotypes or viral strains may be implicated in beta-cell destruction and islet autoimmunity. 36,44It was not possible, through the registers, to distinguish between viral and bacterial infections, or to identify virus-specific infections.Failure to adequately identify or capture the exposure could potentially lead to the estimates approximating the null.Furthermore, we performed analyses to reduce misclassification of the exposure and to address disease severity, but with similar results as in the main analysis.The possibility remains, however, that there may be an association between infections and LADA of an effect size that we did not have enough power to detect.Notably, previous observations in children indicated relative risks from 1.3 to 2.6 for respiratory infections 8,11 , which is above the upper limit of the confidence interval we estimated for the exposure window 1-3 years prior to diagnosis.
Susceptibility to infections could make individuals more inclined to seek medical care, and for diabetes to be diagnosed.It is also possible that the symptoms of prediabetes would prompt a more thorough examination, and thereby a higher chance of being diagnosed with an infectious disease.Other factors, such as health literacy, may influence the likelihood of seeking care.To minimise such biases, we adjusted the analyses for the number of health care contacts, and mutually for the different infectious disease categories.
The analyses were also adjusted for a diagnosis of elevated blood glucose, to some extent accounting for a prediabetic period.This is the first study investigating the potential role of infections in the incidence of adult-onset autoimmune diabetes and therefore replications are clearly warranted, particularly in adult-onset type 1 diabetes, where autoimmunity is an even more distinct feature in the pathogenesis.Ultimately, studies on prospectively collected biological samples could shed light on the impact of specific virus or bacterial strains.It should also be noted that this study was undertaken in a predominantly Scandinavian population and results may be different in other countries where autoimmune diabetes is less prevalent.
In summary, our research suggests that in contrast to findings in Each LADA case is invited to the ESTRID study along with four randomly selected type 2 diabetes cases.Incidence-density sampled control participants were identified through the Swedish Population Register and matched to LADA (6:1) and type 2 diabetes cases by residency and time of participation.Further details on the study design are given in the Supporting Information S1: page 2, Figure S1.We included all cases of LADA (n = 597), type 2 diabetes (n = 2065), and control subjects (n = 2386) enroled in ESTRID 2010-2019.All participants provided written informed consent and the study was approved by the regional ethical review board in Stockholm (Dr nr 2010-1036-31, 2018-1036-32).

F I G U R E 2
Odds ratios with 95% CI for the associations between having any event of infectious disease and the risk of LADA within different categories and exposure windows, in (A) LADA with high-risk HLA genotypes and (B) high (≥median) GADA concentrations.Reference is no event.Model adjusted for age, sex, BMI, smoking, physical activity, educational level, alcohol consumption, number of health care contacts within the exposure window, elevated blood glucose and comorbidity prior to diagnosis/matching, and mutually for the other disease categories.High-risk HLA was defined as DR3/3, DR3/4, DR4/4, or haplotypes of DR4-DQ8 or DR3-DQ2.Total number of LADA with high-risk HLA (HLA high , n = 243), high GADA (GADA high , n = 301) and controls (n = 2386).

recorded up to 5 F I G U R E 3
years before diabetes diagnosis.Particularly, there was an almost 2.5-fold increased risk associated with gastrointestinal infections 3-5 years before.The fact that a higher frequency of infections was observed up to 5 years prior to type 2 diabetes diagnosis compared to 1 year prior to LADA diagnosis could reflect a longer preclinical phase in type 2 diabetes.32However, a few prospective studies point to an increased risk of metabolic syndrome and insulin resistance associated with Helicobacter pylori infection41 but the link to type 2 diabetes is unclear.Our finding is in line with a Odds ratios with 95% CI for the association between having any event of infectious disease and the risk of type 2 diabetes within different categories and exposure windows.Reference is no event.Model adjusted for age, sex, BMI, smoking, physical activity, educational level, alcohol consumption, number of health care contacts within the exposure window, elevated blood glucose and comorbidity prior to diagnosis/matching, and mutually for the other disease categories.Total number of cases (n = 2065) and controls (n = 238).

childhood onset type 1
diabetes, an infectious disease diagnosis up to 10 years before does not increase the risk of LADA diagnosis, regardless of the type or severity of the infection.AUTHOR CONTRIBUTIONS Contributions to the data collection were made by Sofia Carlsson, Bahareh Rasouli and Jessica Edstorp (ESTRID), Emma Ahlqvist and Tiinamaija Tuomi (ANDIS).Sofia Carlsson and Jessica Edstorp were responsible for conceptualising the research objectives and designing the study.Jessica Edstorp was responsible for drafting the manuscript and analysing the data.Marios Rossides and Tiinamaija Tuomi provided medical expertise.Sofia Carlsson thoroughly revised the manuscript.All authors have read and approved the final manuscript.