Back

Does sugar control arrest complications in type 2 diabetes? Examining rigor in statistical and causal inference in clinical trials.

Ojha, A.; Vidwans, H.; Watve, M.

2022-08-03 endocrinology
10.1101/2022.08.02.22278347 medRxiv
Show abstract

In contrast with type 1 diabetes mellitus (T1DM), in type 2 (T2DM) the success of intensive glucose normalization in arresting diabetic complications is marginal and inconsistent across multiple clinical trials. However, glucose regulation still largely remains the main target of treatment for T2DM in clinical practice. We examine the scientific rigor behind the design, conduct and inferences of 6 major clinical trials targeting glucose normalization and following up for diabetic complications and mortality. We find and discuss multiple flaws in reporting the results, their statistical treatment and clinically useful recommendations. The most serious flaw is the inability to recognize the limitations of statistical inferences when multiple comparisons are involved. Further we show using simulations that when different outcomes are not independent of each other, significance gets overestimated. We also suggested alternative ways to assess the effect of antihyperglycemic treatment, if any. Using more sound and elaborate statistical methods and inferential logic we find no support to the prevalent belief that intensive glucose normalization has any benefit in terms of reducing the frequency of any of the complications. Furthermore, alternative interpretations of the results have not been considered and evaluated in any of the clinical trials or their meta-analysis so far. Because of failure to show consistent significant benefit across multiple trials, we should now treat the hypothesis that glucose normalization prevents complications in T2DM as decisively falsified. This necessitates rethinking about some of the fundamental beliefs about the pathophysiology of diabetic complications and facilitate novel alternative lines of research.

Matching journals

The top 6 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.