Pfizer's (NYSE:PFE) epilepsy drug Neurontin has had generic competition for a few years, but it's still plaguing the pharmaceutical giant.

An article published in yesterday's New England Journal of Medicine claims that Pfizer ran 20 clinical trials to test the drug's ability to treat other afflictions, like migraines and pain, but published only 12 of them.

On one hand, you can see how the company wouldn't want to release information about failed studies. It's not like we require Google (NASDAQ:GOOG) to release every failed search algorithm it comes up with, or make Procter & Gamble (NYSE:PG) disclose every rejected ad campaign.

But we're talking about people's health, here. Determining whether a drug is successful in treating a disease or condition requires knowledge of failed experiments, not just the ones that work. Without the negative information, doctors and patients cannot reach proper, informed decisions about the effectiveness of the drug.

Suppose I told you I had flipped a coin 100 times and gotten 80 heads (it's possible). If you didn't know that I had unsuccessfully tried this lots of times, and had filtered the results when reporting to you, you'd think that coin was something special, when it actually wasn't.

In Pfizer's case, things only get worse. Of the 12 studies that were published, eight had their primary measurement of success changed before being reported. Now, there are legitimate reasons to change a primary endpoint, but it's usually a big red flag for doctors, investors, and the Food and Drug Administration, and all of them are right to be wary of such clinical trial results. Changing one or more primary endpoints on two-thirds of the studies makes that red flag as big as New York.

Pfizer, Eli Lilly (NYSE:LLY), Merck (NYSE:MRK), and others have all paid settlements or fines to the Department of Justice for marketing their drugs to doctors for indications that weren't approved by the FDA. Publication of clinical trial results for additional indications is one way to get around the marketing rules, but it only works if doctors can trust the results.

Pharmaceutical companies had better get it together, or their revenues will suffer.