The authors of the Cochrane systematic review carefully analyzed all available studies that fulfilled the criteria of randomized clinical trials of at least 12 weeks’ duration (1). With the exception of one study (2), all registered mortality and morbidity as secondary objectives. Glycemic control was the primary objective in 40 of 41 of these trials. Thus, the major legitimate conclusion of this careful analysis was that “AGIs [α-glucosidase inhibitors] have clear beneficial effects on glycemic control” mainly through their dose-dependent effect on postprandial hyperglycemia.
However, the authors also state as one of their main conclusions that they “found no evidence for an effect on mortality or morbidity.” Although this statement may be mathematically correct, it is misleading as it purports to be based on a solid analysis of the data from the 41 studies. This is not the case in their meta-analysis. Most of the selected trials had a treatment period of ≤24 weeks; many were of 3-month duration only and were therefore not designed and powered to investigate hard clinical end points such as morbidity or mortality. This is well reflected by the fact that, as reported by the authors, information on morbidity or mortality could only be retrieved in 3 of the 41 trials. While one study showed a significant treatment effect regarding cardiovascular events, the others presented only general statements without providing any detail. It is well known that sample sizes of individual clinical trials are often too small to detect clinically important effects reliably and that this is one of the reasons why meta-analysis is employed (3,4). However, hard end points such as cardiovascular mortality are going to be very rare in short-term duration studies unless compensated for by a huge population sample. Therefore, including short-term duration studies in their meta-analysis dilutes the cases of cardiovascular mortality. That biases the interpretation of the data analyzed.
The MERIA (MEta-analysis of Risk Improvement under Acarbose) analysis of seven placebo-controlled, long-term, randomized studies examining the effect of acarbose on cardiovascular-related mortality and morbidity showed a reduction of cardiovascular events in patients with type 2 diabetes (5). This analysis is based on all available acarbose studies with a minimum treatment duration of 52 weeks from a database including individual patient data. Because of this, publication and selection bias were already ruled out, as discussed in the response (6) to the criticism raised by van de Laar and Lucassen. Unfortunately, the same criticism voiced previously is repeated in their meta-analysis without taking the detailed response into consideration. In summary, the MERIA analysis showed a beneficial effect on cardiovascular complications in patients with established type 2 diabetes, a finding which is in accordance with the results from the STOP-NIDDM trial in subjects with impaired glucose tolerance (7).
We fully agree with the authors’ statement that prospective trials with the primary objective of investigating cardiovascular events and mortality are required to confirm the beneficial effect of acarbose on cardiovascular events in these high-risk populations. However, the combined data from the STOP-NIDDM trial and the MERIA meta-analysis are highly suggestive of the preventive effects of acarbose on cardiovascular complications in subjects with glucose intolerance.
