WASHINGTON — Diabetes doesn't kill inpatients; high blood sugar does. That was the underlying theme of a consensus conference sponsored by the American Association of Clinical Endocrinologists, American College of Endocrinology, and the American Diabetes Association.
In separate talks at the meeting, Dr. Anthony P. Furnary and Dr. Irl B. Hirsch presented some of the accumulating evidence supporting the notion that the “diabetic disadvantage” in morbidity and mortality—particularly with regard to cardiovascular outcomes—can be largely mitigated by normalization of glucose levels while patients are in the hospital.
“Diabetes per se is not a risk factor for increased mortality, length of stay, deep sternal wound infection, or postoperative complication rates in cardiac surgery patients. [Hyperglycemia] is the true risk factor,” said Dr. Furnary, a cardiothoracic surgeon at Providence Heart and Vascular Institute and Providence St. Vincent Medical Center, Portland, Ore.
He presented the latest data from the Portland Diabetic Project, a prospective, nonrandomized interventional study of the relationship between inpatient glucose levels and hospital outcomes in patients with diabetes undergoing cardiac surgery. The study began in 1987. In 1992, the group instituted the Portland Protocol, a finely tuned set of orders for insulin infusions for use in the operating room, in the intensive care unit, and on the wards (www.portlandprotocol.org
Of the 5,619 diabetic patients who underwent open heart surgery from 1987 through the end of 2005, 91% underwent coronary artery bypass grafting (CABG). Glucose levels were measured every 30–120 minutes throughout the patients' stay, and the average glucose from the first 3 perioperative days was calculated. This average, termed “3-BG,” was used to assess overall glycemia for each patient.
Glucose targets for the insulin infusion protocol have been gradually ratcheted down over the years, from 150–200 mg/dL in 1992 to 70–120 mg/dL in 2005. At first the infusion was used only in the ICU, but in 1995 its use was expanded into the operating room and onto the non-ICU floors as well. For the 210 diabetic patients who underwent open heart surgery at the Portland hospital in 2005, the daily average 3-BG across all three hospital settings was 121 mg/dL.
During 1987–2005, inpatient CABG mortality was 1.6% for the 2,886 CABG patients with a 3-BG less than 200 mg/dL, compared with 4.4% for the 1,552 with 3-BG greater than 200 mg/dL. When broken down by glucose sextile, mortality ranged from 0.7% for those with 3-BG less than 150 mg/dL to 2.5% with 3-BG 175–200 mg/dL, up to 14% for those whose blood sugars averaged more than 250 mg/dL during their first 3 perioperative days.
In a multivariate analysis, the highly significant impact of 3-BG on CABG mortality was independent of epinephrine use. After adjustment for other preoperative risk factors such as age, ejection fraction, and renal failure, the insulin infusions independently reduced mortality by 60%, Dr. Furnary reported. Indeed, mortality among CABG patients in the Portland Diabetic Project has dropped steadily over time, whereas mortality among nondiabetic patients hasn't changed. Now the mortality for both groups averages 0.9%, compared with 3.4% among diabetic CABG patients nationwide, he said.
Rates of other outcomes are also being found to be strongly related to 3-BG levels. Deep sternal wound infections have occurred in just 0.6% of patients with 3-BG less than 150 mg/dL, compared with 1.1% with 3-BG 175–200 mg/dL and 3.7% with 3-BG greater than 250 mg/dL. Compared with 3-BG below 175 mg/dL, deep sternal wound infections are more than three times more likely among patients with levels above that value. Rates of other types of infection, postoperative transfusion, and new-onset atrial fibrillation were also found to be independently affected by 3-BG. Length of stay is also significantly related to 3-BG, by approximately 1 day for every 77-mg/dL increase.
In all, the Portland data thus far point to the need for change in current hospital practice. “It begs us to do something different, to not just report if a patient is diabetic or not,” Dr. Furnary commented.
Dr. Hirsch, medical director of the University of Washington Diabetes Care Center, Seattle, agreed. He summarized data from other recent studies of non-ICU populations that come to the same conclusion, including one of 1,253 patients with acute MI in whom in-hospital mortality did not differ between diabetic and nondiabetic patients, but was significantly greater among those with hyperglycemia on admission (defined as plasma glucose greater than 198 mg/dL) in both diabetic (10% vs. 5%) and nondiabetic (24% vs. 6%) patients (Am. Heart J. 2005;150:814–20).
According to Dr. Hirsch, two major trials published in 2005 that also support the same conclusion have been misinterpreted as negative. One was a multinational study of 1,253 patients with type 2 diabetes and suspected MI randomized to either a glucose/insulin/potassium (GIK) infusion for 24 hours followed by a home insulin prescription, GIK infusion followed by standard glucose control, or routine metabolic management. Although there were no differences in survival at 2 years by treatment group, an epidemiologic analysis showed that fasting blood glucose at baseline and during the study strongly predicted mortality (Euro. Heart J. 2005;26:650–61).