Matthews D R, Edge J A, Dunger D B
Diabetes Research Laboratories, Radcliffe Infirmary, Oxford, UK.
Diabet Med. 1990 Mar-Apr;7(3):246-51. doi: 10.1111/j.1464-5491.1990.tb01379.x.
A simple, unbiased insulin-varying glucose clamp program is described. The aim of the program was to utilize a continuously updated array of data to predict insulin requirements for normoglycaemia. In assessing Type 1 diabetes the quantity of insulin required for maintenance of basal euglycaemia can be more clinically informative than other clamp methods. We present a method which uses an iterative computer program to predict changes in insulin infusion rate required for glucose clamping. After initial parameter estimation, the program uses no fixed algorithm but makes predictions according to previous blood glucose responses to infusion rates. The program has flexible data entry, graphic display, and running statistics including mean infusion data, mean glucose levels, and their respective standard deviations. Data for 26 consecutive overnight clamp studies have been analysed. The median coefficient of variation of glucose values at the end of the clamp was 4.1% (range 1.4-12.0%). The mean bias during the last 2 h was 0.20 +/- 0.24 mmol l-1. Cross-correlation showed that insulin had its maximal effect on the rate of decline of glucose after 15 min, and the nadir of glucose occurred 45 min after a change in insulin infusion rate.
