GLYCEMIC RESPONSE TO AN INSULIN DOSE CHANGE - COMPUTER SIMULATOR PREDICTIONS VS MEAN PATIENT RESPONSES

We compared the change in blood glucose (BG) following a common insuli n dose change predicted by a computer simulator (based on experts' kno wledge), with the observed mean response of a group of diabetic subjec ts. Twenty-two insulin-dependent diabetes mellitus (IDDM) and 26 non-I DDM patients, requiring 2-3 injections (total 25-100 units) per day, m easured BG 4 times daily (before meals and bedtime) for 4 days while r eceiving their usual insulin dose, and 4 days the following week recei ving a 15% (greater than or equal to 3 unit) lower dose of breakfast i ntermediate acting insulin. For each patient the resulting BG change a t each time point was calculated from the median BG pre and post dose change. As the BG change in IDDM and NIDDM was not significantly diffe rent, the simulator's prediction at each time point was compared with the mean response of 48 patients. The mean rise in BG (mmol/l) in 48 p atients (+95% confidence interval) vs the simulator's corresponding pr ediction was: pre breakfast 0.3 (-0.3-0.9) vs 0.1; pre lunch 0.6 (0-1. 2) vs 0.5; pre dinner 1.1 (0.3-2.0) vs 1.3; pre bed 1.2 (0.4-2.0) vs 0 .4. This is the first reported test of computer simulator predictions against the mean response of a large group of diabetic patients. There was an expected but considerable variation among patients' responses to a reduced morning insulin dose with a surprisingly large BG elevati on pre bed during the late stages of intermediate insulin's action; ho wever even this value overlapped the 95% confidence interval. We concl ude that this simulator, developed from physicians' predictions, match ed the mean patient glycaemic response sufficiently closely to indicat e its potential usefulness as an educational tool.