Sunday, June 7, 2009

BARI 2D: When It Comes to Diabetes Outcomes, 100% Compliance and 0% Variation is Impossible (and a handy summary for readers too)

The Disease Management Care Blog saw the "BARI 2D" Study Group's results come over the wire from the New England Journal of Medicine. The results should be of some interest to the disease management community because it deals with concurrent morbidities of diabetes mellitus and coronary artery disease in populations with chronic illness. While it's been pretty clear that conservative non-surgical treatment of stable heart disease among all patients is safe, we didn't know - until now - if the same was true among persons with established diabetes.

In addition, this trial gives us some interesting benchmark data.

First, BARI-2D. 2,368 persons being treated at 49 different centers underwent a cardiac catheterization and, depending on the severity of their blockages, were then were non-randomly assigned to having open heart surgery (coronary artery bypass grafting or 'CABG') or 'non-invasive'percutaneous coronary intervention' ('PCI'). Then - that's right, after the catheterization - patients were randomly assigned to getting their planned CABG or being switched to medical treatment only without a CABG.... or, their PCI or being switched to medical treatment only without a PCI.

The medical treatment consisted of pills or insulin that were aggressively concocted to achieve optimum control of blood sugar, cholesterol and blood pressure. However, that's not all. The medical treatment was also randomly assigned. The two options were pills alone versus pills for the blood pressure, cholesterol but the substitution of insulin for the treatment of the diabetes. An additional wrinkle was the types of pills used for the diabetes: they had to include an insulin 'sensitizer' such as metformin or a thiazolidinedione.

You can see the complex allocation strategy here.

The results? The ability of medical therapy to compete with PCI held up over 5 years: there was no difference in outcomes. However, among the persons that were originally assigned a CABG, it turned out the CABG was better than pills when it came to the combined end point of death or non-fatal stroke or heart attack: 22.4% vs. 30.5% (death rates by themselves were not different). What's more, being assigned to a CABG and getting the surgery plus receiving the pills for the diabetes did statistically better than getting insulin. That combined endpoint rate was 18.7% versus 32%.

And now for the benchmark data. Recall this was a carefully performed clinical research trial involving clearly defined endpoints: the physicians worked much harder than in usual clinical practice to get their patients to assigned levels of blood pressure control, cholesterol levels and average blood glucose. In general:

the A1c (a measure of sugar control) was about (there were subtle differences in the various groups, so the DMCB is rounding things) 7.2%, plus or minus (i.e., standard deviation) 1.3%,

the LDL was about 80 mg% plus or minus 25 and

the blood pressure was about 125 systolic plus or minus 17 over 70 diastolic plus or minus 10.

Keep in mind that the current national diabetes guidelines, HEDIS measures and pay for performance programs key on keeping the A1C below 7%, the LDL less than 100 and the blood pressure less than 130/80. According to these data, substantial numbers of individuals (recall there is a plus or minus around the averages) receiving exquisitely close monitoring and highly aggressive therapy in this state of the art reseach initiative failed to reach outcomes targets. The DMCB can't tell the percent of patients failing to achieve recommended levels of control, but they must be substantial, especially for the A1c and the systolic blood pressure: the standard deviations extended well above recommended levels.

Implications? The insights about treating stable heart disease in persons with diabetes are important enough that they should probably be included in the information that patients could use when they are being coached in the self-management decision-making in conjunction with their physicians.

According to this well performed study, if a person with diabetes and symptomatic but stable coronary heart disease isn't bad enough to require a CABG, pills work fine: a PCI offers no additional advantage. However, if the disease is bad (involves all three vessels of the heart for example), a CABG results in fewer heart attacks over the next five years; unfortunately, that doesn't mean patients are going to live longer. Without a meaningful survival benefit, some persons with diabetes may elect to forego a CABG and, given a patient's individual values, that may be - depending on the circumstances - reasonable.

In addition, having a CABG may also mean that it would be better to manage the diabetes with insulin sensitizing pills rather than insulin.

In addition, if a clinic or a disease management initiative or even a 'medical home' has a population of persons with diabetes and known heart disease, having an average A1c of 7.2 ain't bad. What's more, achieving a mean LDL of 80 and mean blood pressure of 125/70 are commendable, but the 'spread' of patients around those averages, despite the best medical therapy, will still put substandial numbers above an LDL of 100 and a blood pressure that exceeds 130/80.

This has implications for the assessment of quality in health care settings and suggests HEDIS-based assessments of quality and P4P need to explicitly recognize that 100% compliance and zero variation is a pipe dream. The DMCB wonders, for example, if a primary care physician achieves outcomes measures that replicate those found in this study, they should get full credit (and all their P4P bucks) even if the standard deviations put substatial numbers of patients outside the official definition of meeting quality.


AlanHP said...

or non-randomly assigned to post-cath treatment? You say both in the third paragraph of the post.

Jaan Sidorov said...

The patients' cardiologists did the cath. They decided on the invasive therapy needed: CABG or PCI. Once patients were bucketed into one of those two groups, THEN they were randomly assigned to the original plan (CABG or PCI) or being switched to medical therapy.

It's awkward grammar I know.....

Thanks for the feedback!!