Metabolic phenotypes, vascular complications, and premature deaths in a population of 4,197 patients with type 1 diabetes

Abstract

Objective: Poor glycemic control, elevated triglycerides, and albuminuria are associated with vascular complications in diabetes. However, few studies have investigated combined associations between metabolic markers, diabetic kidney disease, retinopathy, hypertension, obesity, and mortality. Here, the goal was to reveal previously undetected association patterns between clinical diagnoses and biochemistry in the FinnDiane dataset.

Research design and methods: At baseline, clinical records, serum, and 24-h urine samples of 2,173 men and 2,024 women with type 1 diabetes were collected. The data were analyzed by the self-organizing map, which is an unsupervised pattern recognition algorithm that produces a two-dimensional layout of the patients based on their multivariate biochemical profiles. At follow-up, the results were compared against all-cause mortality during 6.5 years (295 deaths).

Results: The highest mortality was associated with advanced kidney disease. Other risk factors included 1) a profile of insulin resistance, abdominal obesity, high cholesterol, triglycerides, and low HDL(2) cholesterol, and 2) high adiponectin and high LDL cholesterol for older patients. The highest population-adjusted risk of death was 10.1-fold (95% CI 7.3-13.1) for men and 10.7-fold (7.9-13.7) for women. Nonsignificant risk was observed for a profile with good glycemic control and high HDL(2) cholesterol and for a low cholesterol profile with a short diabetes duration.

Conclusions: The self-organizing map analysis enabled detailed risk estimates, described the associations between known risk factors and complications, and uncovered statistical patterns difficult to detect by classical methods. The results also suggest that diabetes per se, without an adverse metabolic phenotype, does not contribute to increased mortality.

FIG. 1.

Multivariate metabolic profiles of patients with type 1 diabetes. The figure depicts the distribution of men and women on the self-organizing map that was constructed based on the listed biochemical variables. The self-organizing neural network algorithm places those patients that have similar biochemical profiles close to each other and those that have differing profiles far apart on the map. The bar plots illustrate the averaged profile for patients that reside on a given hexagonal region. •, Sets of 10 men; ○, sets of 10 women: this was done to avoid excessive clutter from individual markers for each patient.

FIG. 2.

Self-organizing map colorings of clinical features for men with type 1 diabetes. The map in Fig. 1 can be colored according to the characteristics of the local residents within each hexagonal unit. The color scale indicates the deviation from population mean with respect to the random fluctuations that could be expected by chance. The numbers on selected units tell the local prevalence (binary variables) or mean value (continuous variables) for that particular region. For plot A, which illustrates time-adjusted mortality, the random fluctuations could not be estimated using the standard procedure, hence the pseudocolors are different to avoid direct comparisons with the rest of the colorings. The P values below the plots indicate the probability of observing equivalent regional variability for random data. *The metabolic syndrome included variables that were also self-organizing map inputs, hence the P value is only suggestive. Colorings for women are available in online appendix 4.

FIG. 3.

Colorings of biochemical variables on the self-organizing map. The map colorings were produced with the same procedure as in Fig. 2. However, empirical P values are not available for the biochemical variables that were included in the self-organizing map training data. Colorings for women are available in online appendix 5.

FIG. 4.

Metabolic phenotypes and risk of premature death. The relative risk of death for men and women was estimated against the expected sex-specific mortality in Finland. A–E: Five model phenotypes were constructed based on observations from Figs. 1–3. The models do not represent distinct clusters in the dataset, but they summarize the characteristics of patients around the highlighted area to make the discussion in relation to Figs. 2 and 3 easier. F: A high-risk region was highlighted for detailed comparisons of DKD categories (results are listed in text).