All micro- and macrovascular complications were weakly associated with exposure to high HbA1c [HbA1c ≥64 mmol/mol (8%)] at index to some extent, but this was only significant for renal complications [HR (95% CI) of 1.10 (1.02–1.18)] and CAD [HR (95% CI) of 1.18 (1.04–1.34)]. When modeling glycemic exposure as the time-dependent observed HbA1c, a significantly increased risk for all microvascular complications was observed for HbA1c ≥64 mmol/mol (8%). For the macrovascular events, this model was inconclusive, and the results were not significant. These results are largely in line with other findings. For renal complications, Gerstein et al. [
17] found an association between an increase in HbA1c at index and nephropathy, and Klein [
3] found an association between an increase in HbA1c at index and both gross proteinuria and renal failure. For diabetic foot, an increased risk was expected for the model using the HbA1c value at index in concordance with earlier findings [
3,
18]; however, these studies found associations with either amputation [
3] or sensory neuropathy [
18], while we defined a composite endpoint covering a broad spectrum of peripheral arterial problems. However, our findings do suggest that the most recent HbA1c measurement (i.e., time-dependent HbA1c) is a better predictor for diabetic foot than the HbA1c at index. For CAD, the results regarding HbA1c at index are in concordance with the literature [
10] where an HR (95% CI) of 1.05 (1.00–1.10) was reported for a 1% increase in HbA1c also using the HbA1c at index. On the other hand, no significant effect was found for the time-dependent HbA1c, suggesting that the HbA1c at index is a better predictor for CAD than the most recent HbA1c measurement. The results for cerebrovascular disease were inconclusive, which is in line with other findings [
10] where an HR (95% CI) of 0.96 (0.88–1.06) for a 1% increase of HbA1c at index was reported. The EMA-based results for the microvascular events showed similar results as the time-dependent HbA1c. Retinopathy specifically showed a clear dose response for EMA, and associations were stronger than in the time-dependent HbA1c model. For the macrovascular event of coronary artery disease, a significantly increased risk was observed in the highest risk category [EMA ≥64 mmol/mol (8%)]. Again, the results for cerebrovascular disease were inconclusive. An association between the increase in updated mean HbA1c and the risk has been reported for diabetic foot [
1], retinopathy [
8,
19], renal failure [
17], CAD [
1,
2] and cerebrovascular disease [
1]. With the exception of cerebrovascular disease, our results reflected previously reported findings. Absence of an association between EMA and the risk of cerebrovascular disease may be explained by the fact that we defined cerebrovascular disease as a composite endpoint including stroke, TIA, CVA and cerebral infarction, while Stratton et al. [
1] reported an HR (95% CI) of 1.14 (1.01–1.27) for a 1% increase in updated mean HbA1c for stroke only. For the microvascular events, the analyses using glycemic burden showed the clearest dose response. Overall, the HRs were higher, but the categories used were different from the previous models and not directly comparable as the GBY categories of this cumulative outcome were based on the distribution of the number of patients rather than clinically recognizable cutoff points for HbA1c. Again, the effects for retinopathy were stronger than for the other microvascular events: the HRs ranged from 1.41 to 2.60 against 1.30–2.15 for diabetic foot and 1.20–1.61 for renal failure. Maple-Brown et al. [
13] used the AUC as a measure for chronic glycemic exposure and found that it was a good predictor for microvascular complications. For the macrovascular events, CAD (GBY >0 to ≤1 and GBY >3) showed a significant association. Cerebrovascular disease showed a significant, albeit small, dose response with HRs ranging from 1.20 to 1.51.