ICUs were faster to adopt tight glycemic control than to deadopt it

Glycemic control gradually tightened in U.S. ICUs after the publication of a trial supporting tight control, but practice changed little when a subsequent trial found harm from tight control, a new study notes.


Glycemic control gradually tightened in U.S. ICUs after the publication of a trial supporting tight control, but practice changed little when a subsequent trial found harm from tight control, a new study notes.

To assess the practice impact of the 2001 van den Berghe trial, which supported tight glycemic control, and the 2009 NICE-SUGAR trial, which found higher mortality with tight control, researchers conducted an interrupted time-series analysis of more than 300,000 patients treated in 113 ICUs between 2001 and 2012. Results were published online by JAMA Internal Medicine on March 16.

Before publication of either trial, 17.2% of patients had tightly controlled glucose on day 1 of ICU admission (between 80-110 mg/dL or 4.4-6.1 mmol/L), while 3% had hypoglycemia (<70 mg/dL or 3.9 mmol/L), and 40.2% had hyperglycemia (≥180 mg/dL or 10 mmol/L). After publication of the first trial, the percentage of patients with tight control gradually increased (1.7% per quarter) as did the rate of hypoglycemia (2.5% per quarter), while hyperglycemia dropped (−0.6% per quarter). After the NICE-SUGAR trial was published, there was an immediate drop in hypoglycemia (−22% in the first quarter), but no subsequent change, and there was no change in rates of tight glycemic control or hyperglycemia.

The study authors concluded that publication of evidence for tight glycemic control led to slow steady adoption of this practice but that the later evidence of harm from tight glycemic control resulted in little deadoption. The findings show the importance of seeking confirmatory research before implementation of a new practice and observing new practices for potential harms. “There is an urgent need to understand and promote the deadoption of ineffective clinical practices,” wrote the authors.

An accompanying commentary agreed on the need for a model of deadoption or undiffusion, which lacks even an agreed-upon name, the commentary author noted. This study and others on the topic (published in the same issue of the journal) show that “it will often be prudent to phase in an innovation practice gradually, waiting whenever possible for reports of real-world effectiveness studies and syntheses of multiple well-conducted controlled studies in systematic reviews or meta-analyses,” the commentary said.