Determination of glycemic differences between groups treated with continuous subcutaneous insulin infusion (CSII) or conventional insulin therapy (CIT) was central to the major objective of the study. Assessment of glycemia was based on 24-h inhospital profiles of plasma glucose; pre- and postprandial and bedtime (seven time points) diurnal profiles performed monthly on outpatient samples; and glycosylated hemoglobin (HbA1) measured bimonthly at each center. The correlation between plasma glucose determinations in the central laboratory and in local laboratories was 0.988. Significance of differences between treatments was by analysis of variance and least-squares regression. At baseline, mean inhospital plasma glucose and HbA1 concentrations and insulin dosages were identical in the groups randomized to CSII or CIT. A prompt decrement of indices of glycemic control during CSII was observed such that mean decrements sustained over the 8-mo treatment period in home and in hospital plasma glucose profiles and HbA1 relative to values obtained during CIT (P less than 0.0001). The likelihood of CSII-treated patients achieving glycemic indices within the normal range was increased. The standardization of the mean and the M-value calculated from inhospital glucose profiles during CSII and CIT at 4 and 8 mo indicated that there was less plasma glucose fluctuation during CSII. The method of pooling standardized local HbA1 measurements from the six centers appeared to be an adequate substitute for centrally performed HbA1 determinations. Advantages of inhospital plasma glucose measurements in terms of accuracy and ability to obtain nocturnal samples contrasted with the likelihood of increased realism and superior correlation with HbA1 in home-obtained samples.(ABSTRACT TRUNCATED AT 250 WORDS)