Counterpoint: Intensive Glucose Control and Mortality in ACCORD—Still Looking for Clues
2010; American Diabetes Association; Volume: 33; Issue: 12 Linguagem: Inglês
10.2337/dc10-1658
ISSN1935-5548
Autores Tópico(s)Diabetes Treatment and Management
ResumoEarly cessation of the intensive glycemic treatment arm of the Action to Control Cardiovascular Risk in Diabetes (ACCORD) trial because of a 22% excess of all-cause mortality compared with the standard arm (1) came as an unpleasant surprise to many observers. Since then, considerable discussion of the significance of this result has ensued along with various theories as to why it occurred. In the companion article, a respected statistician, John Lachin, proposes yet another hypothesis: that the finding was due to the play of chance (2). In response, this article will address three related issues: 1 ) Dr. Lachin's hypothesis; 2 ) a review of findings of post hoc analyses of the ACCORD data seeking support or lack of support for the other main theories; and 3 ) my own opinions on why continued study of the ACCORD data are needed, including a reformulation of one of the original theories regarding the cause of excess mortality. ### The play-of-chance hypothesis “How often have I said to you that when you have eliminated the impossible, whatever remains, however improbable, must be the truth?” asked Sherlock Holmes of Dr. Watson in The Sign of the Four (1890). Dr. Lachin's exposition of the theory that the apparent excess of mortality in ACCORD was due to the play of chance is very welcome. This possibility has not previously been given enough attention. The excess of deaths with the intensive glycemic treatment strategy was indeed not large, a 22% relative increase from 1.1 to 1.4% of participants affected per year (1). In absolute terms, it was 257 versus 203 deaths or an excess of 54 among 10,251 study participants. Had this been the primary end point, a direct comparison would have shown the level of statistical significance generally regarded as conclusive. But it was not the primary end point and repeated measurements …
Referência(s)