In the circles where I travel this is well-digested news, but if you're a normal person you may be hearing for the first time about this meta-analysis of trials of a drug marketed as an anti-depressant, called reboxetine. I say "marketed as an anti-depressant" because as it turns out, it is not an anti-depressant.
This is one of those rare occasions on which the U.S. comes out ahead of the EU on drug approvals. Reboxetine was never approved in the United States, but it was sold in Europe. Published data showed it to be the equal of the most popular class of anti-depressants, selective serotonin re-uptake inhibitors such as fluoextine (Prozac). (Actually that isn't saying much -- as you know if you've been reading along time, SSRIs don't actually work very well, and really can only be justified in cases of fairly severe depression. But I digress.)
It turns out that of all the clinical trials that had been conducted on reboxetine, Pfizer had published data on only 26% of patients. After the German equivalent of NICE blackmailed them into giving up the rest of the data, this analysis found reboxetine to be a) completely useless and b) to have a worse side effect profile than SSRIs. (In these trials, SSRIs were only slightly better than placebo, as we would expected now, but again I digress.)
It is likely, though not publicly known, that the drug was not approved in the U.S. because regulators had access to some of this unpublished data. And since 1997, all clinical trials funded by the U.S. government or sponsored by companies seeking approval for a drug have to be registered with the feds; no more hiding unfavorable results from regulators.
But as extensive commentary in BMJ, from multiple perspectives (off limits to non-subscribers, I'm sorry to say) makes clear, that isn't the only problem. Here's my bullet list, for what it's worth:
- Even if regulators have access to results of unpublished trials, doctors don't. The drug companies can still bias the literature and, as long as their products pass the minimal standards for approval, make them look a lot better than they really are. Indeed . ..
- Previous meta-analyses of reboxetine, based obviously only on published results, made it seem jes' fine. That's not the fault of the analysts (for the most part, these authors do catch a mistake), but it's inevitable when we don't have access to all the trials.
- Even when we have the results of trials the authors choose to report, we can still be misled by what they do and do not choose to tell us. If you were lucky enough to be around for my statistics primer, you know that if you make a whole lot of comparisons, some associations will appear just by chance. You can always dredge through your data and find that a drug seems to work for some arbitrarily chosen sub-group of patients, or on some endpoint or other. Registered trials have to specify their end points in advance, but publications are often based on end points chosen after the fact -- and this is not always disclosed.
- Trials can be designed in the first place to with a big fat thumb on the scale. Often the dose of the comparison drug is deliberately made too small. Or people who drop out of the trial -- perhaps because of side effects, or because the drug wasn't working -- are not properly accounted for. Or the assessment of effects is done by people who aren't really blinded to the treatment. Or whatever. Sometimes these flaws are detectable by critics who have access to the original raw data, but usually nobody does. It's proprietary.
- Drugs are approved and marketed based on short-term studies of carefully selected people. But then they are used over the long term, in the real world with all sorts of people. Far too little investment -- and often none at all -- is made in following up to see what really happens when drugs are widely prescribed.
I could go on -- that isn't all of it. But the bottom line is, if companies that stand to make big bucks not only pay for the studies, but also control their conduct, analysis of the data, and what gets published, we're going to have these scandals again and again. The public is tired of hearing about them by now and probably just tunes them out, while the misleading TV ads still have them popping the pills with faith in a miracle cure.
Don't get me wrong. There are pharmaceuticals that have been around for a long time, whose safety and effectiveness and the limitations thereof are well understood, and you and your doctor may reliably agree that it makes sense for you to take them. But too often, you really don't know, and there are far too many nasty surprises. No, you can't trust drug companies, their executives, or for that matter the research scientists who take their money, either to actually do research or just to put their names on work they didn't really do and papers they didn't write. And no, the professors don't get in trouble for it. At least not yet.