November 2006. It enrolled 208 patients with cancer concentrated primarily in their liver, what’s known as hepatocellular carcinoma. An estimated 19,000 people are affected by this disease in the U.S. each year, although it’s much more common in other parts of the world, especially Asia. These patients were randomly assigned to get the Light Sciences treatment, or the physician’s choice of therapy if they went into the control group. Patients enrolled at 33 sites in Asia and Europe. They entered the study with a life expectancy of about nine months for the control group. The primary goal of the trial is the gold standard in cancer research—showing that the Light Sciences drug helps patients live longer.
When I profiled Keltner back in July 2008, he was already in suspense. Not everyone had enrolled in the study, but it was designed so that when 142 people in total had died, a proper statistical analysis could be done to show with certainty that the drug/device combo was helping people live longer. Based on enrollment trends and life-expectancy, Keltner told me then that he expected the results in early 2009.
But that time came and went, and even in March 2010, this study still can’t claim to have reached the threshold of 142 deaths required to perform the analysis. When I spoke to Keltner on Friday, he said the study was fully enrolled by May 2009, and that he now projects the full 142 patients will have died by May of this year. After another month of statistical analysis, he expects Light Sciences to issue a statement on what the data says.
It would be tempting for Keltner to inject big-time bias, by trying to imply that this trial is lasting longer because the drug must be working. If Light Sciences were publicly traded, you can bet your entire net worth that hedge funds would be running all kinds of statistical models and betting fortunes for and against this result. But a lot of surprising things can happen in clinical trials. It could be that doctors didn’t follow study protocol, and snuck in lots of patients who had a healthier prognosis than they were supposed to. It could be that because patients enrolled at a clinical trial site, they all got better than average care, making the historical life expectancy figure inaccurate. Nobody will really know for sure if the drug is offering an advantage until study monitors take the blind off—which is meant to protect the study from biases.
Keltner did offer one clue about how
