Comorbidities and mortality in persons with type 1 diabetes: the vantage point of multiple time scales via Bayesian intensity model

Abstract

Introduction Type 1 diabetes is an incurable autoimmune condition occurring typically before the age of 30 years in which the ability of pancreas to produce insulin in compromised resulting in too high levels of blood sugar. Diabetes often causes other health problems over time. Calendar year and age of diagnosis of type 1 diabetes as well as diabetes duration are considered as important predictor factors for mortality and morbidity in persons with type 1 diabetes. One of the common comorbidities in type 1 diabetes is the chronic kidney disease, which increases the risk for both cardiovascular death and end-stage renal disease (ESRD). ESRD is a life-threatening complication resulting in a poor prognosis for patients. Cardiovascular death and renal failure are therefore competing risks in patents with type 1 diabetes. So far, researchers have focused on the competing risks approach to account for mortality when evaluating the ESRD risk. Our aim is to provide an overall picture on the changes in the progression of type 1 diabetes to ESRD and/or death without ESRD, when studying the risk on multiple time scales simultaneously, and accounting for the interplay of the relevant time scales. Methods This study is based on the cohort formed in the register-based FinDM (Diabetes in Finland) study, in which incident type 1 diabetes patients were identified from the Finnish drug reimbursement register (Social Insurance Institution) and from the Finnish Hospital Discharge Register (National Institute for Health and Welfare). This study focused on those diagnosed with type 1 diabetes between 1972 and 1991 before age of 30 years (n = 11,810). Follow-up data from the date of entitlement until end of 2014 on ESRD (Finnish Registry for Kidney Diseases) and death (Statistics Finland) were obtained through a computerized register linkage. We tracked the individual's risk for ESRD or death without ESRD on the relevant time scales, age, calendar time and diabetes duration, by applying nonparametric Bayesian intensity model. The model allows incorporation of multiple time scales and multiplicative effects of different time scales. We used the informative graphical displays of the hazard, such as 3-dimesional plots and heatmaps, to present and interpret the results. Results After a mean follow-up of 26.7 years, a total of 338,492.8 person-years accumulated and 844 persons (7.1%) developed ESRD, yielding a crude incidence rate of 2.5 per 1000 person-years. Deaths of persons with ESRD (n = 440) comprised almost 19% of all of the 2345 deaths. The crude mortality rate was 5.7 and 86.2 per 1000 person-years among type 1 diabetic patients without and with ESRD, respectively. Estimation of multiplicative effect of the diabetes duration and attended age by means of Bayesian model depicted both time scales as determinants of the risk for ESRD as well as the risk of death without ESRD. After attended age of 20 years, the risk of ESRD increased steeply, reached a plateau around 30 years, and showed a slight decrease after age of 40 years. The risk of ESRD associated with diabetes duration incorporated two levels, the lower one followed by the higher one, with the change point at 15 years. The risk of death without ESRD increased gradually with increasing age, whereas a decrease in risk was seen for increasing diabetes duration in those diagnosed with type 1 after age of 8 years. Calendar time was associated with improved survival of type 1 diabetic patients without ESRD and decreased risk for ESRD. Conclusions Overall prognosis of type 1 diabetes has improved over time. Attended age and duration of diabetes, both important determinants of the morbidity and mortality in patients with type 1 diabetes, have a multiplicative effect. The strongest joint effect of two time scales on the risk of ESRD was seen for those attending 30-40 years after diabetes duration over 15 years.