Response to Welschen et al. and Kleefstra et al.
I should take responsibility for the differences between the Cochrane Library analysis and the review that appeared in Diabetes Care (1). In the Cochrane Library analysis, no meta-analysis was done for the reasons given in the letter written by Kleefstra et al. (2). However, in the initial submission to this journal, which utilized the same approach as in the Cochrane Library analysis, the authors concluded that the “level of available evidence for the effect of SMBG on glycemic control in patients with type 2 diabetes who are not using insulin is at present only moderate.” The decision to cite “moderate” evidence was based on their statement, “we did not want to depend on statistical significance only because the studies were rather small. Therefore, findings were considered consistent if more than one of the studies reported the same direction of the effect on the outcome measure.” Thus, regardless of statistical significance in individual studies and in the absence of a meta-analysis, studies going in the same direction could constitute “moderate” evidence.
The reviewers of the initial manuscript recommended rejection. However, I obtained other opinions because I felt that this topic was an important one for our readership. One of the subsequent statistical reviewers argued strongly that there was no reason why a meta-analysis of A1C levels could not be carried out on the data in the randomized clinical trials (RCTs) in spite of the fact that the initial values were different among studies and since initial A1C levels were similar in the control and SMBG groups in each study. This is the genesis of the statistically significant difference of 0.39% in A1C levels between the control and SMBG groups in the six RCTs (3–8) in the recent review in Diabetes Care (1). Statistical significance was found in only two (6, 7) of these six RCTs, however.
I would like to point out that in one of them, there were 48 and 40% drop-out rates in the SMBG and control groups, respectively (7). If the nearly half of the SMBG group that failed to complete the study were enriched in those who were showing the least response, the results could be due to self-selection. In the second statistically significant study (6), a difference in counseling between the two groups does not allow the lowered A1C levels to be ascribed to SMBG alone.
In my view, the available evidence does not show that SMBG in diabetic patients not taking insulin leads to lower A1C levels.
