Continuation of metformin after introduction of insulin in type 2 diabetes

BMJ 2009; 339 doi: (Published 01 December 2009) Cite this as: BMJ 2009;339:b5199

Some small discrepancies occurred between the data reported in this editorial by Adriaan Kooy (BMJ 2009;339:b4227, doi:10.1136/bmj.b4227) and those reported in the linked research paper by Søren S Lund and colleagues (BMJ 2009;339:b4324, doi:10.1136/bmj.b4324). These discrepancies arose from an error in the editorial department: when invited to write the editorial, Kooy was given the accepted (but unedited) manuscript of the research paper, but we failed to pass on to him the information about the subsequent changes in data introduced by the authors during the editing process. Kooy has become aware of these changes only since publication. All the discrepancies occurred in the second paragraph of the editorial and are explained below. None of the discrepancies affect the validity of any conclusion or point made by Kooy.

The editorial reported a fall in glycated haemoglobin level in the metformin plus insulin group from 8.15% to 6.71% (but this level should have been reported as 6.72%). The difference in the reduction of glycated haemoglobin concentration between the metformin plus insulin group and the repaglinide plus insulin group was therefore P=0.177 (not P=0.125).

The editorial also reported that in the patients with a negative glutamic acid decarboxylase 65 antibody status the difference in glycated haemoglobin level between the two treatment groups was −0.29% (95% confidence interval −0.56% to −0.02%, P=0.034); in fact, the values were −0.27% (−0.55% to 0.00%, P=0.052). Hence, the small difference in favour of metformin plus insulin was of borderline significance. As the interaction of treatment by glutamic acid decarboxylase 65 status was statistically significant (P=0.037), Kooy’s original statement that differences might exist in glycaemic response in the metformin plus insulin group compared with the repaglinide plus insulin group on the basis of autoimmunity is still valid.

Finally the difference in weight gain between the metformin plus insulin group and the repaglinide plus insulin group was reported in the editorial as −2.51 kg (95% confidence interval −4.10 kg to −0.95 kg; P=0.002), whereas the correct lower confidence limit is −4.07.


Cite this as: BMJ 2009;339:b5199

View Abstract