WHAT DO RESULTS FROM STUDIES MEAN?
How important do you consider the effect of a prophylactic measure that reduces the incidence of diabetes mellitus after three years from 29% in the control group to 14% in the intervention group? And how important does an intervention appear to you that reduced the average HbA1c from 6.1% to 6.0%? Over 300 participants at three European diabetes conferences were confronted with these and eight other questions of a similar nature in the framework of a Hamburg study (1). If one assesses the first result as useful but the second as insignificant, one finds oneself in good company with the diabetes experts who have been partly trained in "evidence-based medicine". But one has been deliberately misled: both of the outcomes presented are the result of one and the same intervention study on prevention of diabetes by life style modifications and metformin (GLUCOPHAGE etc.) (2).
If a threshold ("cut off") is defined when an end point is recorded - as necessary when defining diabetes mellitus - and if therefore all those who are above this value are classified as ill and those below it as healthy, this can artificially inflate the therapy effect. It is only necessary for a large number of patients in the intervention group to be just below and a large number of the control patients to be just above the cut-off. Whether the minimal real differences that exist in the blood sugar levels are clinically relevant can be doubted.
However, emphasising changes in the incidence of diabetes is the rule in studies of prevention; according to the results of the STOP-NIDDM* study, the incidence of diabetes in healthy subjects with abnormal glucose tolerance is reduced from 42% to 32% after 3.3 years with acarbose (GLUCOBAY) (3). Apart from the numerous other deficiencies of the study, no glucose or HbA1c values are given in the article. The information that there are no significant differences in the course of the study between placebo and acarbose is only provided orally (a-t 2002; 33: 72-3). Efficacy in preventing diabetes is also claimed for the lipase inhibitor orlistat (XENICAL). In the XENDOS study a reduction in new cases of diabetes from 9% on placebo to 6.2% on orlistat is reported (4). Persons with abnormal glucose tolerance are said to benefit. However, data on the course of the blood sugar and HbA1c are lacking here also. By chance? To our query we receive further analyses from the sponsor Roche. According to this, the average fasting blood sugar in the placebo group increased from 83 mg/dl to 87 mg/dl in the placebo group and from 83 mg/dl to 85 mg/dl on orlistat (5).
Apart from other more common practices of dressing up study results, for instance, stating relative rather than absolute risk reductions and spreading result scales, the translation of a continuous metabolic parameter into diagnostic classifications can lead to a false estimation, particularly when corresponding measurement data are concealed.
|Studies on the prevention of diabetes suggest considerable effects by statements about the prime importance of disease incidence, which are often not confirmed when the metabolic parameters are considered.
|In order to be able to evaluate the benefit of preventive measures for diabetes mellitus, studies with clinical end points, for example, complications of diabetes, are essential.