April 1, 2008

Compared to What? What's Wrong with the Latest Actos Study

The medical media yesterday were touting Dr. Nissen's report to the big cardiology conference about his latest study of Actos. Here's the headline that accompanied the report in Endocrinology Today:

Pioglitazone may slow progression of atherosclerosis in patients with type 2 diabetes.

But take a look at the actual JAMA journal article this presentation was based on. Its title is: Comparison of Pioglitazone vs Glimepiride on Progression of Coronary Atherosclerosis in Patients With Type 2 Diabetes: The PERISCOPE Randomized Controlled Trial.

What should immediately leap out to you is that in this study Actos' effect hardened arteries was being compared with Amaryl, a sulfonylurea drug which is part of a drug family that has long been suspected of worsening heart disease. Not a placebo.

This was a small study. Only 543 patients participated, split into two groups, but 90-some patients in each group dropped out. This left only about 180 subjects in each arm of the study. To their credit, the researches cite the small study size and the unusually high drop out rate as a possible problem.

The study only enrolled people who already had evidence of blocked arteries. It concludes that, compared to those taking Amaryl, patients taking Actos saw a tiny though statistically significant decrease in the thickness of problem arteries. They had improved CRP (a measure of inflammation) and HDL though there was no actual difference between the number who died in both groups during the study.

The group taking Actos had slightly better fasting blood sugars and fasting insulin levels than those taking Amaryl, though of course, as is the case in every drug study you will ever see, their actual fasting blood sugars were high enough to damage all their organs.

A cheery graph shows the average fasting blood glucose of each group. The group average of those taking Actos started at 147 mg/dl and dropped by 8.5 mg/dl by the end of the study.

But this graph ignores the standard deviation of the group. The standard deviation is a measurement of the range within which most of the values in the group cluster. In this study, the standard deviation for fasting glucose was 41. That means that most of the fasting glucose values clustered between 106 mg/dl and 188 mg/dl.

But here's another concern: though standard deviations are presented for the starting values and measurements in the middle of the study, no standard deviation is given for the amount of improvement experienced at the end of they study. That is presented as an average with a confidence interval--the latter being an estimate of how accurate that average might be. This is far less informative than information about the standard deviations of the individuals' improvements might have been. So it is possible that the inclusion in the group of several people who were very strong responders to the drug could have skewed the final result.

Averages presented without their standard deviations should always be treated with great caution. If you average my income, your income, and that of Bill Gates the amount you come up with could be $500,000,000. That's because you get an average by adding all your values and dividing by the number of values. But standard deviation would show you how misleading that average is, because it might be $499,950,000 telling you that values averaged fell in a range plus or minus $499,950,000. This example should show you why averages are meaningless without knowing the standard deviation.

A far better measurement of values in a group is the median, a measurement which lines up all your values in order of size and picks the middle value. The median of your income, my income, and Bill Gates' income is likely to be something like $50,000--a more informative statistic.

More importantly, by pooling group data and using averages to analyze the group, we miss the really important statistic: the actual change per individual. Though the authors present a dizzying description of highly confusing statistical techniques they used, ones I doubt anyone but they themselves could understand, one wonders why they did not simply look at the increase and decrease per individual of the parameters measured and present the median and standard deviations of these individual changes which would have been far more informative.

That said, Dr. Nissen has a very good reputation in the field, so lets give him the benefit of the doubt and conclude, as he did, that Actos did make a very small improvement in the thickness of arteries in the people who took Actos compared to those who took Amaryl.

It is time to ask your doctor for some Amaryl?

Probably not. There were two other statistics buried in this study that should give you pause. Both of them relate to the most troubling side effects of Actos.

We all know that one of the biggest problems with the sulfonylurea drugs is the weight gain they cause. Well, this study showed that people taking Actos gained more than twice as much weight as those taking the sulfonylurea drug, Amaryl. Not only that, the standard deviation of the weight gain in the middle of the study was much larger for those taking Actos, suggesting that some people on Actos gained a lot more weight than the average would suggest.

We know that Actos lowers the insulin resistance by growing baby fat cells and stuffing them full of new fat. These cells, once grown, will never go away, which is why people who gain weight on this drug report that even after they stop taking it, getting rid of the additional weight is extremely hard if not impossible.

The second side effect is worse: There were no bone fractures in the Amaryl group, but fully 3% of the Actos group experienced bone fractures.

It is now well known to those of us who learn about drugs from sources other than manufacturer-funded studies that Actos builds those baby fat cells we mentioned above by hijacking the stem cells that should have turned into new bone. Over time this leads to osteoporosis, and there is no doubt that Actos causes osteoporosis in older women, a group whose thinner bones are more likely to fracture if weakened.

This study lasted only 18 months but there was still a very significant difference in the number of fractures between the two groups. So the high number of fractures--8 in this group of less than 200 people of mixed gender and ages taking Actos is extremely troubling.

Actos is a drug people will be put on not for months, but for years or even decades. Osteoporosis is one of the major killers of older women who break hips and then almost always die. So how Dr. Nissen can dismiss this drug as having a "reasonable safety profile" is beyond me.

And finally, whatever this drug did to artery thickness, there was no significant difference in actual mortality outcome in this study and, indeed, there were more fatal heart attacks in the group taking Actos. This study was short, but the other worrisome side effect of Actos in people who take it for longer time periods is that it produces congestive heart failure in people who did not have it when they started taking the drug. This is probably due to the edema (water swelling) it causes. It will be cold comfort indeed to know that your arteries are some tiny fraction thinner when you discover you have stretched your heart muscle to where it can barely function and that thanks to heart failure, your chances of surviving eight years is one in five--far worse than those for people with cancer.