Characteristics of cases and controls
The characteristics of cases and controls are shown in Table 2. Compared with controls, cases had higher HbA1c levels, a longer duration of diabetes, a higher frequency of insulin deficiency (17% of cases vs 1% of controls), and a higher percentage of insulin use (89% of cases vs 24% of controls) at baseline. Cases had comparable fasting glucose levels to controls, but higher rates of positive islet antibodies (GAD, IA2, IAA, ZnT8). The quality of the optimal matching on the continuous age and BMI variables was excellent: 90.9% of age matches were within 5 years, with a median difference of 0.5 years. For BMI, 90.7% of matches were within 5 kg/m2, with a median difference of 1.21 kg/m2.
Islet autoantibody profile of participants with baseline insulin deficiency
All participants with insulin deficiency (n = 63) were taking insulin and therefore designated as IAA positive. Excluding IAA, the most common positive islet autoantibody was GAD (44.4% in participants with insulin deficiency, 6.7% in those without insulin sufficiency). These findings persisted after the removal of cases involving patients who died (n = 21) and their respective controls (n = 86).
Odds of SH and failure to achieve optimal glycaemic control for participants with baseline insulin deficiency or islet autoantibodies
Baseline insulin deficiency was associated with a higher OR (reported as OR [95% CI]) of SH and failure to achieve an HbA1c level of <6.0 (42 mmol/mol) which persisted after adjusting for age, BMI, diabetes duration and exclusion of those who died during the study (unadjusted OR 35.6 [95% CI 14.2, 89.6], p < 0.0001; fully adjusted and after removing deaths, OR 23.2 [95% CI 9.0, 59.5], p < 0.0001). This outcome was also associated with a positive IAA and baseline insulin use (unadjusted OR 4.5 [95% CI 3.4, 6.0], p < 0.0001; fully adjusted and removing deaths, OR 3.8 [95% CI 2.7, 5.3], p < 0.0001), positive GAD (unadjusted OR 3.6 [95% CI 2.4, 5.4], p < 0.0001; fully adjusted and removing deaths, OR 3.9 [95% CI 2.5, 6.0], p < 0.0001), positive IA2 (unadjusted OR 10.6 [95% CI 3.4, 33.4], p < 0.0001; fully adjusted and removing deaths, OR 16.7 [95% CI 3.9, 71.6], p = 0.0001) and positive ZnT8 (unadjusted OR 4.0 [95% CI 1.4, 11.4], p = 0.01; fully adjusted and removing deaths, OR 3.9 [95% CI 1.2, 12.4], p = 0.02; Table 3).
We analysed whether the presence of multiple autoantibodies might be significant (Table 4). The presence of one autoantibody compared with zero autoantibodies was associated with a significant increase in the odds of developing the adverse outcome (unadjusted OR 4.0 [95% CI 3.0, 5.3], p < 0.0001; adjusted OR 3.3 [95% CI 2.4, 4.6], p < 0.0001; fully adjusted and removing deaths, OR 3.4 [95% CI 2.4, 4.7], p < 0.0001), although all participants on baseline insulin were considered positive for IAA. The presence of two or more autoantibodies compared with zero autoantibodies was associated with even higher odds of the adverse outcome (unadjusted OR 12.4 [95% CI 7.1, 21.6], p < 0.0001; adjusted OR 9.2 [95% CI 5.2, 16.5], p < 0.0001; fully adjusted and removing deaths, OR 9.9 [95% CI 5.4, 18.0], p < 0.0001).
We also performed a sensitivity analysis excluding controls (n = 183) or cases (n = 16) not well-matched for age or BMI; these results were similar to those of the full sample.