Baseline characteristics

Demographics are provided in Table 1, including a subdivision by subsequent CAD status. From 3517 FinnDiane Study participants, 539 experienced a CAD event during the 19-year follow-up. At baseline those who subsequently developed CAD vs. those who did not, were older, had longer diabetes duration, higher BP, higher BMI and waist height ratio, higher triglycerides, total and LDL cholesterol concentrations, lower HDL cholesterol concentrations, and were more likely to be in the KDIGO category high or very high. Additionally, they were more likely to have the metabolic syndrome, lower eGDR scores, to be on antihypertensive and lipid-lowering drugs, and more likely to have a history of smoking. Insulin pumps were used by 3.5 vs. 6.3% of those who did vs. did not develop CAD respectively (p = 0.02).

Table 1 Baseline clinical characteristics with and without incident coronary artery disease during follow-upComparison of insulin resistance by eGDR formulae’s lowest quartile and metabolic syndrome

There was little overlap in the number of insulin-resistant individuals when defined as being in the lowest quartile in each of the eGDR formulae (Additional file 1: Fig. S1). The frequency of the metabolic syndrome in participants considered insulin-resistant, i.e., the lowest quartile of eGDR by the Williams, Duca, and Januszewski eGDR formulae, was 64, 71, and 64%, respectively. In comparison, for those in the highest quartile of eGDR, the corresponding frequencies were 17, 16, and 17%.

As a continuous score calculated with the three assessed formulae, the eGDRs were significantly (p < 0.001), albeit weakly, correlated: The Spearman correlation coefficient was 0.42 for Williams vs. Duca scores; 0.10 for Williams vs. Januszewski; and 0.10 for Duca vs. Januszewski.

eGDR, the metabolic syndrome, and KDIGO risk categories

Baseline characteristics by KDIGO categories are provided in Additional file 1: Table S2. Worsening kidney disease was associated with higher rates of CAD, male sex, longer diabetes duration, younger age of diabetes onset, higher HbA1c concentrations, and worse traditional risk factors, such as adiposity, BP, and lipids. For all three formulae, the eGDR decreased with worsening KDIGO risk category, and, in addition, the percentage of individuals with metabolic syndrome increased with higher KDIGO category.

Associations between baseline eGDR quartiles and subsequent CAD

Kaplan–Meier curves for incidence of CAD (Additional file 1: Fig. S2) by eGDR quartiles showed increasing separation of curves over follow-up time, with different patterns of spread between formulae, reaching statistical significance for all eGDR formulae: eGDR by Williams and Januszewski, both p < 0.001, and eGDR by Duca, p = 0.015. Similarly, metabolic syndrome status curves separated significantly (p < 0.001) over time, with higher CAD rates in those with vs. without the metabolic syndrome at baseline (Additional file 1: Fig. S3).

Risk for CAD based on baseline eGDR score and by kidney disease severity

At all the proposed cut-offs for eGDR calculated by the Williams and Januszewski formulae, we found an association with the CAD incidence (Table 2). The strengths of the associations varied depending on the cut-off, but was stronger for scores based on the Williams formula at all cut-offs. For the Duca formula, only the lowest quartile of its eGDR was associated with incidence of CAD. When restricting the follow-up time to maximum of 15 years, the hazard ratio (HR) for CAD incidence decreased linearly with an increasing eGDR score for all three formulae (Fig. 1A–C), indicating that any increase in eGDR (improvement in insulin sensitivity) was cardioprotective. In Fig. 1D, the HRs are defined per score percentile and therefore allow for a direct comparison of the strength of the association for all three formulae. When using the C-index, the Williams-derived eGDR discriminated individuals with regards to CAD incidence either better or at least equally well compared to the other formulae. Only in the lowest score percentiles, the HRs based on Januszewski were higher, but their 95% confidence intervals (CI) overlapped with the Williams formula, e.g., in the 0.28 percentile Januszewski-based HR was 9.29 [6.46, 13.36], whereas the Williams-based HR was 7.69 [5.78, 10.24].

Table 2 Hazard ratios and 95% confidence intervals for incident coronary artery disease in the full cohortFig. 1

Cohort-wide hazard ratios for incident coronary artery disease by estimated glucose disposal rate formulae. Williams (A), Duca (B), Januszewski (C). D Compares all three formulae and shows the hazard ratio per score percentile

Additional file 1: Figures S4, S5, and S6 show the HR for the 15 years incidence of CAD, separately for each KDIGO category. In all three categories, when using the eGDR score by Williams, the HR for CAD decreased linearly with increasing insulin sensitivity.

Comparison of baseline eGDR scores with other risk factors for subsequent CAD

As shown in Fig. 2 and Additional file 1: Tables S3–S6, age and diabetes duration showed the highest C-index for the association with incident CAD for the whole cohort and for each KDIGO risk category. In the full cohort, the eGDR score by Williams had a higher C-index compared to the Januszewski score (0.69, 95% CI [0.67, 0.72] vs. 0.62 [0.60, 0.65]) and the Duca score (0.53 [0.51, 0.56]). This was observed in the low KDIGO category (Additional file 1: Table S4), however in the moderate (Additional file 1: Table S5) as well as the pooled high and very high KDIGO categories (Additional file 1: Table S6), there were no significant differences between the eGDR scores among the three formulae as the 95% CIs overlapped. However, in the KDIGO categories high and very high, the C-index for Januszewski was nominally higher than the C-index for Williams, but the CIs overlapped (0.58, [0.54, 0.63] vs. 0.57 [0.52, 0.61]).

Fig. 2

C-indexes for cardiovascular risk factors including estimated glucose disposal rate scores. For all individuals and separately for individuals in Kidney Disease Improving Global Outcomes risk categories low, moderate and high combined with very high