1. Tony on March 15, 2012 11:37 AM writes...

I am really not looking forward to the reading of 329 possible side effects after every damn ambien commercial..

2. Eve on March 15, 2012 11:57 AM writes...

I think that would do a better job of sending you to sleep than the Ambien. Without the hallucinations.

3. Anon on March 15, 2012 12:30 PM writes...

Maybe this will keep MDs from prescribing multiple drugs as though each drug targets a pathway that [is the only] cause of a symptom?

4. SBC on March 15, 2012 1:18 PM writes...

What clinical trials can measure 329 end points?

5. gretchen on March 15, 2012 2:08 PM writes...

And how many "side effects" are there from eating an apple? Or drinking a cup of tea?

6. NorthwesternChemist on March 15, 2012 3:26 PM writes...

Looks like a really poor signal to noise ratio.

7. paperclip on March 15, 2012 6:04 PM writes...

#1, better than 329 possible side effects after every Cialis commercial.

8. Arthur on March 15, 2012 6:06 PM writes...

On average, 50% of Americans already don't adhere to treatment, primarily out of fear of potential side effects of the medicine prescribed for them. Listing even more extremely rare side effects on the drug label, and Americans may very well respond by abandoning modern medicine altogether.

9. TJMC on March 15, 2012 11:06 PM writes...

I suspect the percent incidence of some of these new AE's are pretty low - maybe well below 0.5%? But this is not a surprising capability with consolidation of eHealth records, semantic structuring, etc.

We were advising firms about this coming capability quite a while ago. In fact, the underlying tech and approach have been trialed in more focused ways years ago, looking for "signals" that were weaker than formal post-approval trials could yield. I was wondering then what the FDA would then do when the rigor was proven and easily in reach.

Like all new tools, they offer both risks as well as opportunites - for firms as well as patients. Well beyond boring ads. Consider this - there are ways to link not just molecules, but substructures to the AE incidences...

10. Rich Rostrom on March 16, 2012 1:02 AM writes...

Are drug interactions always bad?

I think I have read of serendipitous synergy between drugs - though I believe only between drugs in related fields.

It would be interesting if there was an unexpected and significant benefit of interaction between two drugs used for unrelated purposes.

11. simpl on March 16, 2012 4:01 AM writes...

The methodology of patient matching could be applied afterwards to any study looking for side-effects and having information on concomitant medication. It could usefully be designed into phase IV studies. The larger number of side-effects should help patients to put them into better perspective: to accept their existance, thus be observant, control them early, and switch if necessary.
Would someone with experience like to comment on the Altman quote “Adverse events are incredibly valuable clues to what these drugs are doing in the body” Is this to be taken as a platitude, or is it a major way to find 2nd generation drugs?

12. Morten G on March 16, 2012 6:15 AM writes...

Won't this make it easier to have a new drug approved? "The current standard of care has this list of potential side effects. Our new drug only has this much shorter list."

And like Rick Rostrom said, can't these tools be used to tease out positive drug interactions?

13. Bernard Munos on March 16, 2012 12:12 PM writes...

I tried to download the drug interaction database, which came in the supplemental info. Excel truncated it at 1,048,576 rows, the maximum it allows! I can hardly imagine the elation of entering the final row in that spreadsheet.

14. emjeff on March 20, 2012 7:05 AM writes...

Typical data-mining nonsense. Fols, if you have a big ol' pile of data and start looking for correlations, you will find them. It is cheap, shoddy, "look I have software that does statistics" science like this that has made the field of epidemiology a laughingstock...

15. Northwest AJ on March 21, 2012 1:41 AM writes...

You know what I'd LOVE to see? A database correlating drug responses with other "ground conditions" in the patient's life, drug or non-drug. There are also foods that keep liver enzymes occupied, for one (grapefruit comes to mind) and genetic backgrounds, sleeping/eating habits and a number of other things all seem to have some influence... In general items such as "People who respond to drug Y usually don't respond to drug X, but often respond to drug Z" are observed by doctors but rarely quantified.

(Yes, this is a personal soapbox of mine. I won't normally take a drug without looking up the Wikipedia entry and hitting at least a few of the journal articles linked from it; and this means that my success rate with prescriptions is quite a bit higher than average as I tend to look up causality of side effects as much as possible and request meds that will have fewer side effects in *me* based on my history with similar. But I'm always annoyed when I run into sparse data. Yes, five percent of the trial group developed glowing red eyes, but WHICH five percent? That stuff is important!)

Re: adverse effects being a hint as to what a drug is doing in the body - well, they can help to indicate whether it's condition-appropriate and if not, what is (for example, when similar conditions respond better to different drugs and/or a side effect appears in people who *don't* have the problem the drug is meant to address). They can also help point out what receptors/molecules it's acting on, where (especially for side effects that occur in a defined population group - say, menstruating women, or people with Parkinson's disease, or...)

I guess this is my long way of saying that being able to better predict side effects should actually IMPROVE treatment adherence.