BOSTON ( TheStreet) --I've long warned investors to be wary of companies that use dubious scientific analysis to spin positive results from what are really failed clinical trials.

Call it post-hoc data mining or retrospective data dredging, but by either name, the purpose of these all-too-common tactics is usually to deflect investors' attention away from bad news.

Apricus Biosciences ( APRI) and Cel-Sci ( CVM) issued separate press releases in the past week which relied heavily on post-hoc data mining to hype the prospects of experimental -- and deeply flawed -- drugs in their respective pipelines. Knowing how to spot the red flags in these types of announcements can save you a lot of pain and losses down the road.

The headline on Apricus' announcement set off alarm bells immediately: "Apricus Biosciences Reports Reanalysis of Its U.S. Phase III Trials for MycoVa Showing Drug is Effective in Mycological Cure Resulting in Eradication of Nail Fungus." Emphasis mine.

Re-analysis? What happened to the first analysis of the phase III trials of MycoVa?

Apricus doesn't say, but Google and SEC filings are a big help. In August 2008, Apricus (then known as NexMed) announced that its partner Novartis decided against seeking U.S. approval for MycoVa (then known as NM100060) based on disappointing results from two double blind, placebo-controlled phase III clinical trials.

One year later, in July 2009, Apricus/NexMed announced that Novartis was terminating its partnership for NM100060/MycoVa based on results from a third phase III study which compared MycoVa to Loceryl, a topical nail laquer marketed in Europe. This European phase III study failed although Apricus was spinning the results even back then, claiming that a "post hoc" analysis showed MycoVa worked better than Loceryl in patients with mild toenail fungus. Unfortunately for Apricus, the study enrolled patients with mild and moderate toenail fungus.

Three failed phase III studies, one dropped Big Pharma partnership and two years later, Apricus is trying to convince investors that there's life left in MycoVa. This time, Apricus claims that a re-analysis of the two U.S. clinical trials found MycoVa cures toenail fungus better than placebo in patients who do not also have athlete's foot.

That's a classic data dredge. The MycoVa studies in "all comer" patients with mild to moderate toeail fungus failed as designed originally, so Apricus just ignores the data and finds a subgroup of patients (those who have toenail fungus but not athlete's foot) where MycoVa allegedly works. Victory declared.

Except, of course, that the studies do not prove at all that MycoVa works in toenail fungus, non-athlete's foot patients because the studies weren't designed with this goal in mind. What looks good in a press release will almost certainly be dismissed as pseudo-scientific nonsense by regulators, if Apricus ever follows through on a promise to seek approval for MycoVa based on this dubious "re-analysis."

Moving on to Cel-Sci, which announced last week that a "meta analysis" of four phase II clinical trials involving 120 patients with head and neck cancer found a "highly statistically significant" reduction in total cholesterol following treatment with the company's experimental drug Multikine.

I must admit to confusion reading the Cel-Sci announcement a few times because Multikine is being developed as a treatment for head and neck cancer. In fits and starts, Cel-Sci is now trying to conduct a phase III study. Who cares if Multikine lowers cholesterol, especially since Cel-Sci has never before discussed the possibility of developing Multikine as a cholesterol-lowering drug? (I'm also putting aside the nontrivial issue of explaining how Multikine might lower cholesterol, since Cel-Sci doesn't offer any.)

Cel-Sci's claim about Multikine lowering cholesterol is just more hyer-promotional data mining and doesn't pass the smell test. According to regulatory filings, Multikine was studied in previous clinical studies enrolling over 200 patients, so how does Cel-Sci explain a meta-analysis that includes four studies and 120 patients? How were those studies and patients selected and why were others left out?

A 2005 phase II study of Multikine in head and neck cancer published in the Journal of Clinical Oncology was almost certainly included in the meta-analysis, but that study makes no mention of any cholesterol-lowering effects of the drug. In fact, the study publication doesn't mention cholesterol at all.

Cel-Sci claims treatment with Multikine led to a statistically significant reduction in cholesterol, but what was their drug compared to for that conclusion to be reached? None of the phase II studies of Multikine conducted by Cel-Sci included a control arm -- meaning all the patients received Multikine. How then, can these studies demonstrate a statistically significant reduction in cholesterol? Cel-Sci doesn't say.

Perhaps Cel-Sci split the Multikine-treated patients into two groups -- those taking statin drugs for high cholesterol and those who were not. Again, Cel-Sci doesn't explain. Of course, it doesn't really matter because we run full circle back to the data-mining problem -- the studies weren't originally designed to split patients into statin and non-statin users.

If the cholesterol-lowering finding from this meta-analysis of Multikine studies is reality challenged, why would Cel-Sci bother with it? The answer goes back to what I warned up top -- positive results derived from post-hoc data mining are usually a smokescreen seeking to cover up other problems.

In Cel-Sci's case that means problems with the ongoing phase III study of Multikine, which is taking forever to get off the ground. The number of press releases issued by Cel-Sci announcing the start of the phase III trial appears to exceed the number of patients actually enrolled. Cel-Sci's problems run deeper still because lackluster Multikine data from the phase II studies gives the drug little chance of succeeding in the phase III study.

Yet Cel-Sci's needs to keep hope alive, lest gullible retail shareholders lose interest and the disbursement of management's annual paychecks become threatened.

Desperate companies like Apricus and Cel-Sci use data dredging as a necessary survival tool. Smart investors recognize the danger signs and steer clear.

Post script: As I was putting this column together, I saw news that Transdel Pharmaceuticals ( TDLP) filed for bankruptcy this week and sold off its remaining assets. Transdel was also a data-mining offender.

Fair warning.

More companies guilty of data-mining: Osiris Therapeutics ( OSIR), Agenus ( AGEN - Get Report) (formerly Antigenics), Introgen Therapeutics (defunct, another bankruptcy), and Cell Therapeutics ( CTIC - Get Report), which famously re-analyzed a failed lung cancer trial claiming its drug Opaxio (formerly Xyotax) only worked for women.

--Written by Adam Feuerstein in Boston.

>To contact the writer of this article, click here: Adam Feuerstein.

>To follow the writer on Twitter, go to

>To submit a news tip, send an email to:

Adam Feuerstein writes regularly for TheStreet. In keeping with company editorial policy, he doesn't own or short individual stocks, although he owns stock in TheStreet. He also doesn't invest in hedge funds or other private investment partnerships. Feuerstein appreciates your feedback; click here to send him an email.