This review evaluated the effectiveness of biphasic, basal and prandial insulin regimes in patients with type 2 diabetes. The authors concluded that greater glycaemic control was possible using biphasic or prandial insulin. The risk of hypoglycaemia was uncertain. This was a largely well-conducted review based on small numbers of variable trials. Some caution is warranted when interpreting the authors' conclusions.

To evaluate the effectiveness of biphasic, basal and prandial insulin regimens in treating patients with type 2 diabetes.

The Cochrane Central Register of Controlled Trials (CENTRAL), MEDLINE, and EMBASE were searched from 1996 to 2008. Search terms were reported. Reference lists of relevant trials and reviews were scanned to identify additional studies. Unpublished data were sought from abstracts of international diabetes meetings and the Internet sites for Current Controlled Trials, the UK National Research Register and the National Institutes of Health.

Randomised controlled trials (RCTs) comparing basal, prandial or biphasic insulin formulations in patients (at least 18 years of age) with type 2 diabetes were eligible for inclusion in the review. Trials were excluded when: the use of oral hypoglycaemic agents was unbalanced across the trial arms; the intervention lasted less than three months; or the comparisons were analogue and conventional insulin within the same regimen. The outcomes of interests were mortality, morbidity, glycaemic control (glycosylated haemoglobin - HbA1_c), fasting glucose, postprandial glucose (PPG), weight gain, hypoglycaemic event rate and quality of life.

In the included trials, the proportion of men ranged between 21 and 73.8%; the age range of patients was 50 to 69 years, and trial length ranged from three to 12 months. Titration methodology, glycaemic targets and definitions of hypoglycaemia varied.

Two reviewers selected trials for inclusion in the review, and disagreements were resolved by consensus or discussion with a third reviewer.

Trial quality was assessed using the Jadad scale, with a maximum score of 5 points. An additional criterion was added to assess the presence of intention-to-treat analysis.

It was unclear how many reviewers performed the quality assessment.

Two reviewers extracted the data on mean changes in the outcome variables, along with 95% confidence intervals (CI). Standard deviations were imputed or calculated, where necessary. Disagreements were resolved by consensus or discussion with a third reviewer.

Data from patients with similar pre-trial insulin exposure were pooled in a meta-analysis to calculate weighted mean differences (WMD) and 95% confidence intervals using the inverse variance method. Heterogeneity was assessed using the I² statistic. Where significant heterogeneity was detected, a random-effects meta-analysis was performed. Sub-group analyses were carried out according to oral hypoglycaemic agent use, insulin type (analogue or conventional), quality score (less than 3 points, 3 points and over), titration reporting, fasting glucose target (less than 7, 7mmol/L and over), and baseline glycaemic control (less than 9% or 9% and over). Publication bias was assessed using funnel plots.

Twenty-two RCTs (n=4,379 patients; sample size range 21 to 472) were included in the review. Methodological quality was considered to be low (the median score was 3 points; range 2 to 5). Twelve trials used intention-to-treat analysis. Only insulin-naive patients were included in the meta-analysis (15 trials). Small data sets precluded the assessment of publication bias.

Biphasic versus basal insulin regime (10 trials; n=2,160 patients): Compared with basal insulin, pooled results showed that biphasic insulin reduced glycosylated haemoglobin (HbA1_c) by 0.45% (WMD 0.45, 95% CI 0.19 to 0.70; five trials; n=1,022 patients), but there was significant variation between trials (I²=66.1%) which could not be explained by oral hypoglycaemic agent use, titration reporting or baseline diabetic control. Compared to biphasic insulin, basal insulin produced a statistically significant reduction in fasting glucose (WMD 0.93 mmol/L, 95% CI 0.21 to 1.65; four trials; n=876 patients), but with high heterogeneity (I²=55.9%) which was not explained by fasting glucose target or baseline diabetic control. The results for postprandial glucose were mixed (nine trials), and there were no statistically significant differences between the regimens for weight gain.

Prandial versus basal insulin regime (seven trials; n=1,216 insulin-naive patients): Compared with basal insulin, pooled results showed that prandial insulin reduced glycosylated haemoglobin (HbA1_c) by 0.45% (WMD 0.45, 95% CI 0.16 to 0.73; seven trials; n=1,156 patients) with significant variation between the trials (I² -70.6%) unexplained by oral hypoglycaemic agent use, baseline diabetic control, titration reporting, use of analogue or conventional insulin, or quality score. Compared with prandial insulin, basal insulin produced a statistically significant reduction in fasting glucose (WMD 2.20 mmol/L, CI 1.70 to 2.70; six trials; n=1,115 patients), but variation between the trials was high (I²=44%) and was not explained by use of oral hypoglycaemic agents, fasting glucose target, baseline diabetic control, titration reporting, use of analogue or conventional insulin, or quality score. Postprandial glucose was reduced after treatment with prandial insulin (five trials) and weight gain was significant increased (WMD 1.86 kg; 95% CI 0.80 to 2.92; six trials; n=1,079 patients). There was high heterogeneity (I²=85.4%) which was not explained by use of oral hypoglycaemic agents, use of analogue insulin, baseline diabetic control, or titration reporting.

Further comparisons were made for basal-bolus versus basal insulin, biphasic versus prandial insulin, and biphasic versus basal-bolus insulin. Biphasic insulin produced similar outcomes when compared with prandial or basal-bolus outcomes (results reported in the paper). There were no trials assessing morbidity and mortality. Pooled analyses for hypoglycaemia and quality of life measures were not possible.

Greater glycaemic control could be achieved in patients with type 2 diabetes when insulin was initiated using biphasic or prandial insulin rather than a basal regimen. The risk of hypoglycaemia was uncertain.

The review question was clear, and this was supported by explicit and potentially reproducible inclusion criteria. The search strategy included some relevant sources of published and unpublished material, although the planned assessment of publication bias was not possible. The review process was carried out with sufficient attempts to minimise errors and bias in the selection and data extraction of trials.

An appropriate validity assessment tool was applied, but it was not clear how many reviewers were involved at this stage. Trial characteristics and results were presented clearly. The decision to pool only data on insulin-naive patients appeared to be appropriate, and attempts were made to explore various sources of variation amongst the included trials. The authors acknowledged the limitations of pooling small numbers of variable short-term trials.

This was a largely well-conducted review, but given the limitations identified and low quality of trials, a degree of caution is warranted when interpreting the authors' conclusions.

A potential conflict of interest was declared in terms of authors connections with drug manufacturers.

Practice: The authors did not state any implications for practice.

Research: The authors stated that rigorously conducted trials are required to assess the long-term outcomes of different insulin regimens, using standardised glycaemic targets, titration protocols, and definitions of hypoglycaemia.

One author was funded by the National Institute of Health Research.

This is a critical abstract of a systematic review that meets the criteria for inclusion on DARE. Each critical abstract contains a brief summary of the review methods, results and conclusions followed by a detailed critical assessment on the reliability of the review and the conclusions drawn.