Forty-eight trials, with 85 treatment arms (n=30,588 patients), were included in the review. Intention-to-treat analyses were used in 38 trials. Methodological quality scores on the 5-point Jadad scale ranged from 1 point to 4 points; the median score was 3 points. Twenty studies were allocated Jadad scores of 2 points or less. Predefined titration steps of insulin doses were used in 41 trials.
Basal insulin (29 RCTs, 38 arms, n=17,588): The proportion of patients with glycated haemoglobin levels of less than 7% was 41.4% (95% CI 35.6 to 47.4), with substantial heterogeneity (I2=95.7%). Meta-regression analyses found that first insulin treatment (46.1%, 95% CI 40.1 to 52.3), lower insulin dose (42.8%, 37.0 to 48.8) and the use of two oral drugs (45.9%, 95% CI 39.2 to 52.7) were also associated with beneficial responses. The median number of hypoglycaemic events per patient per 30 days was 0.5 (interquartile range (0.39 to 0.62) and weight gain was 1.75kg (1.2 to 2.1).
Biphasic insulin (20 RCTs, 26 treatment arms, n=9,237): The proportion of patients with glycated haemoglobin levels of less than 7% was 46.5% (95% CI 40.8 to 52.3), with substantial heterogeneity (I2=89.3%). Higher insulin doses were the only variable that was significantly associated with beneficial response (52.8%, 95% CI 45.9 to 59.6). The median number of hypoglycaemic events per patient per 30 days were 0.37 (interquartile range 0.07 to 0.70) and weight gain was 3kg (1.7 to 4.0).
Prandial insulin (eight RCTs, nine arms, n=1,605): The proportion of patients attaining glycated haemoglobin levels of less than 7% was 39.6% (95% CI 28.6 to 51.3), with substantial heterogeneity (I2=94.6%). Due to the small number of trials, meta-regression analyses were performed only on baseline glycated haemoglobin, final insulin dose and concomitant oral drug use, but none of these were associated with the response observed in glycated haemoglobin levels. The median number of hypoglycaemic events per patient per 30 days was 0.67 (interquartile range 0.44 to 1.0) and weight gain was 2.3kg (1.87 to 3.8).
Basal-bolus insulin (eight RCTs, 12 arms, n=2,114): The proportion of patients attaining glycated haemoglobin levels of less than 7% was 53.9% (95% CI 43.5 to 64), with substantial heterogeneity observed across trials (I2=92.1%). There were non-significant trends found (using meta-regression analyses) for final insulin dose and concomitant use of oral drugs associated with the response. The median number of hypoglycaemic episodes per patient per 30 days was 0.88 (interquartile range 0.35 to 1.3) and weight gain was 2.75 kg (0.78 to 1.3)