June 20th, 2012

Reality Check: The ORIGIN of Spin in a Randomized Trial

In the ORIGIN randomized trial, involving about 12,500 people who had diabetes or were at risk for it, insulin glargine showed no advantage over standard care in preventing the primary composite cardiovascular endpoints at a median follow-up of 6.2 years (see news coverage on CardioExchange). ORIGIN was a completely negative trial, yet in the article’s abstract the authors reported a benefit of basal insulin in the secondary endpoint of incident diabetes (odds ratio [OR], 0.80; 95% CI, 0.64–1.00; P=0.05) and stated that the treatment “reduced new-onset diabetes.” The discussion section of the paper further asserted that the treatment “slowed progression of dysglycemia.”

A few qualifications are in order:

1. This seemingly positive finding (in incident diabetes) was for only one of several secondary endpoints, and the alpha level needed to detect a difference was not corrected for multiple outcome assessments.

2. Among the 1456 participants without diabetes at randomization, only 44% of insulin glargine recipients and 47% of standard-care recipients underwent both prespecified oral glucose tolerance tests at the end of the study. So, in effect, fewer than half of the patients were assessed on the one, barely positive endpoint of incident diabetes. Plus, the fact that an absolute 3% more of the control group underwent the necessary testing could have resulted in ascertainment bias.

3. Incident diabetes among those 1456 participants was much greater than 10% (30% in the insulin glargine group, 35% in the standard-care group), but the authors reported an odds ratio rather than a relative risk, yielding an overinflation of the purported benefit: The relative risk would have been 0.86, compared with the reported OR of 0.80. (Only when the probability of a disease is low does the odds ratio approximate the relative risk.)

4. The upper boundary of the confidence interval included 1.00.

Despite these realities, Sanofi and the investigators didn’t hesitate to spin the trial findings to the media. The Sanofi press release stated, “…the results also showed that insulin glargine delayed progression from pre-diabetes to type 2 diabetes.” Sanofi’s VP of Medical Affairs said, “ORIGIN shows that it is possible to maintain low and stable HbA1c levels that are close to normal over a long time, and to potentially delay the progression from pre-diabetes to diabetes.” One of the lead investigators took this positive angle: “We know that in this study, insulin lowered blood sugar levels and did so safely and effectively. I can look at a patient and say, ‘Your blood sugar is not well controlled right now, so let’s go ahead and add insulin’.”

There is a temptation for those involved in a trial to present findings as positive, even if the positive aspect is only a morsel. However, scientists should stick to the facts. The first sentence of the abstract of this paper asked whether “the provision of sufficient basal insulin to normalize fasting plasma glucose levels may reduce cardiovascular events.” The results of the trial tell us that the answer is simply “No.”

How did you read the ORIGIN trial? What’s your take on the facts versus the spin?

3 Responses to “Reality Check: The ORIGIN of Spin in a Randomized Trial”

  1. D H N, MD says:

    Excellent points. For another conspicuous example see IST-3, recently published in The Lancet. In this trial, the largest and best of thrombolytics for acute stroke, the primary endpoint of death and dependence was no different between groups and yet the authors brazenly claim victory on the basis of a secondary exploratory ‘ordinal analysis’. Perhaps more shocking is the accompanying editorial that either does not notice, or else plays along.

  2. Steven Greer, MD says:

    Glad more people have the courage to be blunt, as I have been

  3. A more meaningful – and perhaps more striking – headline for the media might have been that the basal insulin glargine and control groups did not differ in either of the coprimary outcomes or any of their components. In other words, insulin did not improve outcomes on the major endpoints. More generally, I agree with the points made in the careful commentary on the Origin results. I do think, however, that this was an important study and that the results provide significant guidelines for clinical practice, as well as for future research. The very fact that the major results were all negative is, in and of itself, a compelling finding that should be emphasized to both the profession and the public.