Hanefeld et al. (1) assert that the conclusion (“no evidence for an effect on mortality or morbidity”) from our systematic review on the effects of α-glucosidase inhibitors for patients with type 2 diabetes was biased. Furthermore, they claim to have found evidence for such an effect based on their own meta-analysis. We disagree with both statements.
First, we would like to underline that the solid basis of our results is a systematic review and that meta-analyses were only applied when this was methodologically sound. The extensive search for all possible trials investigating α-glucosidase inhibitor monotherapy yielded only one study with prospectively collected data on morbidity or mortality (2), so a meta-analysis could not be done with these end points; therefore, we concluded that no evidence for an effect on mortality and morbidity could be found (which is essentially different from “evidence for no effect”). In the above-mentioned study, it was reported that for the entire treatment group (α-glucosidase inhibitors given both as monotherapy and as additional therapy), no effects of acarbose on cardiovascular end points were found.
This makes it quite remarkable that this particular study (2) was not included in the MERIA (MEta-analysis of Risk Improvement under Acarbose) study (3). Hanefeld et al. assert that this meta-analysis shows a beneficial effect of acarbose on the occurrence of myocardial infarctions. If it had been included in the MERIA study, it would have been the study with the second longest duration, it would have nearby doubled the number of patients, and it would have been the only study with a sound method of collecting end points. This points to the fact that the sole use of a manufacturer’s database is not a reliable method for the selection of studies for a meta-analysis and that an extensive systematic review is necessary to reduce the risk of selection bias.
Other differences between the conclusions of MERIA and our Cochrane review can be explained by differences in inclusion and methodological robustness. Three of the seven studies in MERIA were also included in our Cochrane review, but no reliable data on cardiovascular outcomes could be obtained. The four other publications were excluded from our review, mainly because no data on α-glucosidase inhibitor monotherapy were available or accessible. Moreover, it should be noted that there was no quality assessment of the studies included in MERIA. Other serious concerns about the MERIA study were expressed in our previous letter and remain largely unresolved (4).
In conclusion, there is currently no evidence for an effect on cardiovascular morbidity and mortality of monotherapy with α-glucosidase inhibitors in patients with type 2 diabetes. In the near future, indirect evidence may be derived from another Cochrane review on the effects of α-glucosidase inhibitors for patients with glucose intolerance (5). We are pleased that the authors of the main studies in this field already have assured their cooperation.
