This is an observational single-center study with prospective assessment of all-cause mortality and CV events in a cohort of individuals with type 1 diabetes. The research design and the study population have been previously described [10]. All subjects with type 1 diabetes (n = 843) attending the Diabetes Outpatient Clinic of the Azienda Ospedaliero-Universitaria Pisana from January 1, 2001 through December 31, 2009 because of their usual screening for diabetic complications were considered eligible. Type 1 diabetes was diagnosed based on age at onset < 36 years and immediate requirement of insulin therapy with unbroken need for insulin after the first year since diagnosis [11]. Pregnant women, participants of not-white ethnicity, those with type 1 diabetes for less than one year, those on dialysis or with renal transplantation, those with a prior history of viral hepatitis or cirrhosis of any etiology as well those with significant alcohol intake (≥ 2 alcohol units per day in men and ≥ 1 alcohol unit per day in women), those with active cancer, and three individuals for whom valid information on vital status could not be retrieved were excluded. Thus, a total of 774 individuals were recruited into the study. The Ethics Committee of the University of Pisa approved the study protocol, consent procedures, and data analysis plan.

Information about onset and duration of diabetes (DD), smoking habits, physical activity, current insulin treatment, concomitant blood pressure- and lipid-lowering therapies were collected at baseline together with the assessment of the presence and severity of micro- and CV complications, as previously reported [10]. Body weight, height, and waist circumference (WC) were obtained for body mass index (BMI) calculation. Blood pressure (BP), taken after 5-min rest in a sitting position, was calculated as the average of at least two consecutive measurements obtained about 5-min apart. Hypertension was defined as systolic BP > 140 mmHg and/or diastolic BP > 80 mmHg and/or the use of any antihypertensive drug.

In all subjects, urine samples were obtained, and blood samples were drawn at study entry after an overnight fast for determination of serum creatinine, HbA1c, total- and HDL-cholesterol, triglycerides, alanine aminotransferase (ALT), aspartate aminotransferase (AST), gamma-glutamyl transferase (GGT), uric acid, fibrinogen, and urinary albumin to creatinine ratio (ACR). Finally, all participants underwent a screening for diabetic complications as previously described [10]. Estimated glucose-disposal rate (eGDR), a proxy of insulin resistance, was calculated based on WC, presence of hypertension and HbA1c, as previously described [12].

The Fatty Liver Index (FLI) was calculated as proposed by Bedogni et al. [13] based on triglycerides, GGT, BMI, and WC:

$$}\, = \,\left( }^}*}\left( }} \right)} + }0.}*} + }0.}*}\left( }} \right)} + }0.0}*} - }.}}}} \right)/\left( }\, + \,}^}*}\left( }} \right)} + }0.}*} + }0.}*}\left( }} \right)} + }0.0}*}}.}}}} \right)}*}00.$$

FLI was categorized as follows: i. FLI < 30: no fatty liver, ii. FLI 30–59: intermediate status, and iii. FLI ≥ 60: hepatic steatosis [13].

To assess to which extent FLI could be associated to overall risk independent of other risk factors we have calculated the Steno Type 1 Risk Engine (ST1-RE) [14]. The ST1-RE is a risk model for composite CV outcome (risk of first fatal or nonfatal CV event: coronary heart disease—CHD—ischemic stroke, heart failure, and peripheral artery disease) in subjects with type 1 diabetes. It is based on age, sex, DD, systolic BP, LDL-cholesterol, HbA1c, albuminuria, glomerular filtration rate, smoking, and exercise [14].

To the same purpose, we also calculated the EURODIAB Prospective Complication Study Risk Engine (EURO-RE) score [15]. The score, based on age, HbA1c, waist-to-hip ratio, ACR and HDL-cholesterol levels, is used to calculate the risk of major CHD, stroke, end-stage renal failure, amputations, blindness, and all-cause death in type 1 diabetes. The performance of both models has been previously validated in independent prospective cohorts.

For both scores, three groups have been defined: low CV score (10-year risk < 10%), intermediate score (10-year risk 10–19%), and high score (10-year risk ≥ 20%).

For each patient we searched and recorded major CV events up to December 31, 2017, and all-cause death up to October 31, 2018 [10, 12]. Vital status was available for all participants and was verified by interrogating the Italian Health Card Database (http://sistemats1.sanita.finanze.it/wps/portal/). Data on the incidence of all CV outcomes and coronary events were available for 736 participants (95.1% of the whole cohort) and were obtained, upon data anonymization, in collaboration with the Regional Health Agency of the Tuscany Region through hospital discharge registers. International Classification of diseases, Ninth Edition, Clinical Modification codes was used to identify major CV outcomes (i.e., first event of myocardial infarction, coronary revascularization, stroke, carotid revascularization, ulcer, gangrene, amputation, and peripheral revascularization) and coronary artery events (i.e., first event of myocardial infarction or coronary revascularization) (Additional file 1: Table S1). All events occurred between the date of enrollment and the end of follow-up, or the date of death were considered as incident.

Data are expressed as median (interquartile range, IQR) and/or mean ± SD for continuous variables, and number of cases and percentage for categorical variables. Continuous variables were compared by Student’s t-test or one-way ANOVA for normally distributed ones. Wilcoxon Sum-of-Ranks (Mann–Whitney) U test or Kruskal–Wallis tests were used for variables with skewed distribution. Pearson χ2 or the Fisher exact tests were applied to categories. For post-hoc comparisons, Scheffe’s test or Tamhane’s test, Mann–Whitney U test, and χ2 test were used for normally distributed, not-normally distributed, and categorical variables, respectively. The Spearman’s rank-order correlation was run to determine the strength and the direction of associations between two variables measured on ordinal scale.

Logistic regression analyses were used to examine the association between FLI and risk of prevalent microvascular complications after adjustment for diabetes-related variables and other potential confounding factors. Four logistic regression models were performed as follows: the first model was unadjusted; model 1 was adjusted for age and sex; model 2 was adjusted for age, sex, DD, HbA1c, smoking habits, hypertension, treatment with lipid-lowering agents and prior CV events; model 3 was like model 2 additionally adjusted for eGDR; model 4 like model 3 further adjusted for all other microvascular complications. Covariates were chosen as potential confounding factors based on their significance in univariate analyses or based on their biological plausibility.

Crude mortality rates and incidence of outcomes were described as events per 1,000 patient-years (PYs), with 95% exact Poisson Confidence Intervals (CI). Time to all-cause death or to each first outcome was plotted according to FLI categories as Kaplan–Meier (K-M) curves and comparisons were made using the log rank test. Associations between FLI categories and outcomes were tested by Cox regression analyses. The proportional hazard assumptions were checked, and none have been violated. Univariate and multivariate Cox proportional hazard models were used to identify the effect of FLI independently of key covariates, i.e., the ST1-RE (model 1), or ST1-RE and prior CV events (model 2). Models 1 and 2 were also run including the EURO-RE instead of ST1-RE. Furthermore, all models were re-run to include risk engines as continuous variables instead of categorical ones. Finally, Cox regression analyses were performed including age, sex, DD, HbA1c, smoking, LDL-cholesterol, HDL-cholesterol, ACR, eGFR, hypertension (model 1) and prior CV events (model 2) instead of ST1-RE or EURO-RE. To test the robustness of the associations between FLI (stratified as FLI < 60—ref.—and FLI ≥ 60) and outcomes, we have also calculated the Hepatic Steatosis Index (HSI = 8 × [ALT/AST ratio] + BMI [+ 2, if female; + 2, if the presence of diabetes]) [16] and added this index in the regression models that included ST1-RE and prior CV events as covariates. Results of Cox regressions are expressed as Hazard Ratio (HR) and 95% Confidence Interval (CI). All statistical analyses have been performed using SPSS package 25.0 version (IBM SPSS, Chicago, IL).