Scarce evidence is available for prognosis in adult-onset type 1 diabetes (T1D). The aim of this study is to investigate the risk of all-cause mortality, cause-specific mortality, and incident CVD in adult-onset T1D, as compared to type 2 diabetes with comparable age at diagnosis and population controls. We will explore potential modifiable factors including lifestyle and clinical characteristics that contribute to T1D prognosis. In addition, we will estimate the life expectancy associated with different age at T1D diagnosis as compared to population controls. This study will be based on people with diabetes recorded in Swedish National Diabetes Registers, with linkages to other nationwide registers to retrieve data on lifestyle factors, biomarkers, treatment and outcome information.

Introduction Type 1 diabetes (T1D) is traditionally considered as childhood-onset disease while it occurs more often in adults than in children, with a median onset-age of 29 years[1]. There is a scarcity of evidence on mortality (and causes)[2] and cardiovascular diseases (CVD) in adult-onset T1D. Previous studies involving adult-onset T1D mostly included individuals diagnosed before age 30 years[3-8]. Nationwide studies in Sweden found that T1D diagnosed at any age between 0-30 have shorter life expectancy, higher risks of all-cause mortality, CVD mortality, and incident CVD than population controls, and age of T1D onset and HbA1c levels are important risk stratifiers[4,5].

Age at diagnosis is an important factor affecting prognosis of diabetes but evidence on T1D diagnosed after age 30 years is relatively scarce. A study of 35355 incident T1D cases in England and Wales found that the hazard for all-cause mortality in people with T1D was higher than that in people with T2D across all groups of age at diagnosis (from 20-39 years to ≥60 years) but such excess risk was observed in men and not in women[9]. A small Swedish study of 112 adult-onset T1D cases (diagnosed between age 18 and 100 years) indicated no excess mortality in T1D as compared to population controls[10]. In comparison, a recent study of 1573 people with newly diagnosed (at age ≥35 years) adult-onset T1D in Sweden found that these individuals had excess mortality as compared to people with T2D while the CVD incidence was close to that in population controls[11]. It is unclear how mortality and CVD incidence in adult-onset T1D vary by diabetes duration, lifestyle factors, and clinical characteristics (HbA1c, blood pressure, and lipids, etc).

This study aims to investigate the risk of all-cause mortality, cause-specific mortality, and incident CVD in adult-onset T1D, as compared to T2D with comparable age at diagnosis and population controls. We will explore potential modifiable factors including lifestyle and clinical characteristics that contribute to T1D prognosis.

Methods Study population Adult-onset (≥18 years) T1D cases diagnosed in 2006-2020 will be identified from the Swedish National Diabetes Register (NDR: 1996-2020), without conflicting types of diabetes diagnosis (primary diagnosis) in the National Patient Register (NPR; 1995-2021), and with exclusive insulin prescription initiated within the first 6 months of diagnosis and recorded in the National Prescribed Drug Register (NPDR; 2005-2022). We will also include all T2D cases diagnosed at age ≥18 years in NDR in 2006-2020 and without conflicting types of diabetes diagnosis (primary diagnosis) in NPR. The date of diabetes diagnosis will be defined as the date of first diabetes record in NDR (visiting date or self-reported date of diagnosis, whichever comes first), NPR, or the date of first prescription of glucose-lowering drugs in NPDR, whichever comes first. Individuals whose self-reported year of diabetes diagnosis ("debutar") in NDR was earlier than the year of diagnosis as defined above will be excluded. Population controls will be selected from participants who have not any record of diabetes diagnosis in NDR or NPR before the end of follow-up (2021). Each T1D case will be matched for age, sex, county[5] with 50 (the ratio depends on the minimum number of available population controls in each matched stratum) population controls who are still alive at the year of diagnosis in their matched T1D cases. The analysis of CVD and MACE (major adverse cardiovascular events) incidence will exclude individuals with corresponding diagnosis at baseline.

Covariates and prognostic factors Information on sex, year and month of birth, country of birth, and marital status will be obtained from the Total Population Register, while information on education will be obtained from the Longitudinal integration database for health insurance and labor market studies (LISA). NDR provides information on smoking (yes or no), body mass index (BMI), physical activity, HbA1c, blood pressure, lipids profiles, eGFR and albuminuria after diabetes diagnosis. HbA1c within control target will be defined as an HbA1c of <7.0%[12]. Blood pressure within control target will be defined as systolic blood pressure <140 mmHg, and diastolic blood pressure <90 mmHg[13]. A favorable HDL cholesterol level will be defined as HDL >1.0 mmol/l in men and >1.3 mmol/l in women[14]. A low-risk LDL cholesterol level will be defined as LDL <2.6 mmol/l[14]. We will categorize eGFR levels into two groups, namely eGFR ≥60 or <60 mL/min/1.73 m2. Microalbuminuria will be defined as two positive tests from three samples taken within 1 year, with an albumin/creatinine ratio of 3-30 mg/mmol (~30-300 mg/g) or U-albumin of 20-200 μg/ min (20-300 mg/L), and macroalbuminuria as albumin/creatinine ratio >30 mg/mmol (~>300 mg/g) or U-albumin >200 μg/ min (>300 mg/L)[4]. Physical activity is recorded as never, <1 time/week, 1-2 times/week, 3-5 times/week, or daily. Individuals with physical activity >1 time/week will be defined as having regular physical activity. We will obtain information on use of anti-hypertensive drugs (ATC codes: C02, C03, C04, C07, C08, C09) and statins (ATC codes: C10A, C10B) from NPDR. Information on insulin regimens (insulin pump or short-acting insulin combined with long-acting insulin) will be retrieved from NDR and NPDR.

Outcome assessment Outcomes include all-cause mortality, cause-specific mortality (diabetes-related deaths: ICD-10 E10-E14; CVD-related deaths: ICD-10 I00-I99; cancer-related death: ICD-10 C00-C97), incident CVD, and MACE. Information on vital status and causes of death will be retrieved from the Causes-of-Death Register (2006-2021). We will define the composite CVD outcome as the first inpatient record (primary diagnosis and up to 7 contributory diagnoses[3]) of ischemic heart disease (ICD-10 codes: I20-I25), stroke (I60-I64), or heart failure (I50) in NPR or the record of corresponding CVD events as the underlying causes of death in Causes-of-Death Register. MACE will be defined as cardiovascular deaths (ICD-10: I70-I77[15]) recorded in Causes-of-Death Register, or the first inpatient record (primary diagnosis and up to 7 contributory diagnoses[3]) of nonfatal myocardial infarction (ICD-10: I21) or nonfatal stroke (ICD-10: I60-I64)[15-18] in NPR.

Statistical analysis Basic characteristics Basic characteristics will be presented as means or medians for continuous variables (age at diabetes diagnosis/baseline, BMI, HbA1c, blood pressure, HDL, LDL, total cholesterol, eGFR) and as proportions for categorical variables (sex, country of birth, marital status, education, CVD diagnosis at baseline, smoking, physical activity, albuminuria, insulin regimen) in individuals with T1D, T2D, and population controls (if feasible). Differences across groups will be tested using Student's t-test, Kruskal-Wallis test, or χ2 test.

Duration of follow-up will be calculated from the date of diabetes diagnosis (date of diabetes diagnosis of their matched T1D cases for population controls) to the date of the occurrence of outcomes, death, or December of 2021 for CVD (June of 2022 for mortality, depending on the availability of different registers).

Risks of mortality and CVD as compared to population controls We will estimate the cumulative probability (95% CI) of all-cause mortality, cause-specific mortality, incident CVD, and MACE in T1D and population controls over diabetes/follow-up duration (according to age at diagnosis if power allows), by plotting Kaplan-Meier curves (R package "survfit" and "ggsurvfit") We will estimate the hazard ratio (HR) for these outcomes in T1D (overall, onset-age 18-29 years, 30-39 years, and ≥40 years) as compared to population controls in Cox models. We will also estimate the diabetes duration-specific HR (95% CI) in T1D, as compared to population controls, by splitting participants according to duration of diabetes/follow-up (0-5 years, 5-10 years, 10-15 years, and >15 years).

To explore the benefit of having modifiable factors within control level, we will estimate the HR for different outcomes in different T1D subgroups separated according to modifiable factors (the first record before the occurrence of outcomes) including smoking status, level of physical activity, control of HbA1c, blood pressure, lipids, eGFR, or albuminuria status, with population controls as the reference group.

All the Cox models estimating HR for T1D vs population controls will be fitted with diabetes/follow-up duration as the time scale, with adjustment for education, country of birth, and marital status, and with stratification by matching groups.

Risks of mortality and CVD as compared to T2D We will estimate the HR for all-cause mortality, cause-specific mortality, incident CVD, and MACE in T1D as compared to T2D in Cox models, with attained age as the time scale, with adjustment for age and calendar year at diabetes diagnosis, sex, education, marital status, country of birth. We will further adjust for smoking, BMI, physical activity, HbA1c (continuous), blood pressure (continuous), lipids (continuous), eGFR (continuous), albuminuria (continuous), anti-hypertensive drugs, and statins to explore potential factors leading to the excess or reduced risks of different outcomes in T1D as compared to T2D We will also estimate the HR in T1D vs T2D according to different subgroups of age at diabetes diagnosis separately (18-29 years, 30-39 years, and ≥40 years). Individuals with missing data for categorical covariates will be treated as a separate group and those with missing values on continuous covariates were assigned the median value with a binary variable indicating whether the values are imputed or not (A previous NDR study used linear mixed models to impute missing data for biomarkers such as HbA1c[19]. Not sure if we can also perform such imputation).

Trajectory analysis (The first step of analysis) Finally, we will estimate the trajectories of smoking, BMI, physical activity, HbA1c, blood pressure, lipids, eGFR, albuminuria, and insulin regimens over diabetes duration in individuals with T1D and T2D according to age at diabetes diagnosis (18-29 years, 30-39 years, and ≥40 years). The trajectories will be estimated using generalized linear model (GLM, "glm" package in Stata 17.0) adjusted for sex and calendar year at diabetes diagnosis, with logit link function and binomial distribution, and with cluster robust standard errors (SEs)[20] to account for the dependence among measurements in the same individual diabetes duration. Such trajectory analysis will provide clues to factors contributing to the potential difference in mortality/CVD risks between T1D and T2D with comparable age at diagnosis.

Inclusion Criteria:

• people with adult-onset (≥18 years) T1D diagnosed in 2001-2020 from the National Diabetes Register
• all T2D cases diagnosed at age ≥18 years in 2001-2020 recorded in the National Diabetes Register

Everyone with T1D was matched by age, sex, and county to 50 population controls from the Total Population Register.

Exclusion Criteria:

• To avoid diabetes secondary to pancreatic cancer, we excluded people diagnosed with cancer in the digestive system 1 year prior to diabetes diagnosis and with pancreatic cancer as the underlying cause of death.

the source population was everyone born 2001-2020

Our data is register-based data that is pseudonymised and thus subject to General Data Protection Regulation (GDPR) and cannot be shared openly