Book reviews, art, gaming, Objectivism and thoughts on other topics as they occur.

Jul 30, 2007

Government vs. Diabetics

I was a little too busy to blog earlier today, but I just had to say something about this particular piece of stupidity. The article is a bit long, but notice the careful dance of how the numbers are reported.

The first two numbers have no bearing on any data whatsoever: they are votes by government panelists about whether the data (data we have not seen yet) are "convincing". Then we have a theoretical, large number, 2,200, which could be the number of people subjected to "serious side effects" from use of Avandia. What constitutes "serious" in this case? No answer is given.

Next number? An even larger approxmation of the number of people in the U.S. that use Avandia: 1 million. Oh the horror. They must be dropping like flies. (Would someone tell me why the graphic says that there are 3 million prescriptions if only 1 million people actually use the drug?)

Next number? A bunch of studies that report on the next number, which seems huge: a 43% increase in risk of heart failure. But hang on a second, that's a relative number . . . if the percentage chance of heart failure was only 1% before, that means that your risk of heart failure on Avandia is only *gasp* 1.43% now. Oh no!

It's not until the last couple of paragraphs (if anyone could tolerate wading that deep into this overly-long and verbose article) that we have some hard actual numbers. 4100 new cases of diabetes a day. 810 deaths, 60% of those are from heart failure. Is this the before-Avandia or after-Avandia heart failure rate? (Diabetics are already more likely to experience heart failure than other people.) Since these are new, untreated cases, we can probably assume that this is the before-Avandia rate.

Okay, let's do some math here. 810 deaths divided by 4100 cases = .19 and some change: 19%. But hang on, we have a problem here . . . are these deaths within the total population of diabetes cases or deaths within the population of new diabetes cases?! THE ARTICLE DOES NOT MAKE THIS EXPLICIT. Let's assume that, every day, 810 people die from diabetes, period. Note that the article doesn't give an estimate for total diabetes cases in the U.S., either. Hmm. We can't use their graphic for an approximation, because we have that weird 1/3 ratio thing going with drug takers vs. prescriptions.

Lucky for me I know how to use Google, so I found this little number here. And we have some more numbers:

Diabetes was the sixth leading cause of death listed on U.S. death certificates in 2000. This ranking is based on the 69,301 death certificates in which diabetes was listed as the underlying cause of death. Altogether, diabetes contributed to 213,062 deaths.


Let's do more math! 69,301 deaths divided by 365 is . . . 189 and some change. Where the heck did this 810 number come from?! Even if you take all the deaths in which diabetes is simply listed as a contributory factor, that's still only (213,062 divided by 365) 583 and some change. Granted, these were the numbers for 2000, and the CDC documentation says that diabetes is probably underreported as a cause or contributory cause of death. Okay. Let's assume that 810 deaths a day, out of a population of 18.2 million diabetes sufferers, is correct.

So, back to our earlier math: 810 divided by 18.2 million . . . no, wait, let's be fair. Let's do chance of death per year, and use the most generous number we have. So 213,062 divided by 18.2 million is: .01 and some change. That looks a lot like 1% to me . . . and remember that only 60% of that 1% is due to heart failure.

Result: my intial estimate of your increased risk of heart failure (1.43% vs 1%), which seemed AMAZINGLY LOW, was actually TOO HIGH.

And for this the government panelists are considering banning a drug that helps one million people control their diabetes?!

6 comments:

Rational Jenn said...

Brilliant!

I loathe articles like that, which throw around numbers to make it sound like someone in the research department did their job, but don't actually SAY anything meaningful or even sometimes coherent.

I'd also be interested in the breakdown of diabetes deaths and complications by Type 1 versus Type 2. They are, in many respects, completely different diseases. How do the numbers for one group skew the numbers for another?

Anyway, yay for critical thinking skills and I sure wish more people had them and displayed them!

Jennifer Snow said...

Thanks, Jenn! I hope I didn't make some ridiculous error in there that invalidates my entire point.

Nancy said...

The numbers as you present them don't seem to tell the whole story. Using 213,062 for the number of deaths in one year among 18.2 million diabetics, the rate of death is 1170 per 100,000 people. This is considerably higher than the 816.7 deaths per 100,000 in the total US population in 2004 (about 43% higher, coincidentally). However, according to Eric A. Orzeck, MD, clinical associate professor at Baylor College of Medicine, diabetes is not mentioned on up to 75% of death certificates for people known to have diabetes. So we really have no way of knowing the death rate among diabetics other than it's higher than among people without diabetes.

Let's go back to the 213,062 deaths in one year among the 18.2 million diabetics. If heart failure was involved in 60% of deaths, then about 127,837 were due to heart failure. If we assume this is the baseline (although as you point out, this may already include the increase from Avandia) and we add a 43% increase in the rate of heart failure, then that would make for 182,807 deaths, or almost 55,000 extra deaths due to heart failure in a year. That's an extra 302 deaths due to heart failure per 100,000 diabetics each year. Since the US average is 816.7 deaths per 100,000, identifying something that increases the death rate by more than 300 in a particular group -- especially a group with an already elevated death rate -- seems to be something that deserves attention.

But what about the 1 million people that Avandia helps? According to Public Citizen's testimony before the FDA Advisory Committee Meeting on July 30, 2007, "There is no evidence of any uniquely beneficial clinical outcome for Avandia." (see http://www.citizen.org/publications/release.cfm?ID=7536). Avandia does not appear to improve clinical outcomes compared to other medications and treatments that do not have the risks associated with Avandia.

Jennifer Snow said...

I'm not claiming to be an expert on how beneficial/detrimental Avandia is, my point is that the numbers in the article are worse than useless: they are actually misleading. Intentional? Sloppy? Does it make any difference?

Saying that Avandia doesn't have a *unique* benefit means that there are other drugs that do the same thing. *All* drugs have side effects, so it's a simple fact that the drugs other than Avandia have *some* negative effects, just *different* ones.

I think people should decide for themselves whether they think the risks are appropriate. Someone that is heart-health conscious and takes care of themselves may not care about the added risk, because they know they can tolerate that more easily than, say, an increase in strain on their liver. It is better to have lots of drugs with different potentialities than to restrict the supply down to one or two drugs that are "safer" for "most people".

Jennifer Snow said...

P.S. the article does not say "deaths per 100,000", it says "deaths per day", which is what I based my numbers on.

Mike N said...

Great post! Keep up the good work.