* FDA says device not proven, could cause harm
* Agency staff say Mela needs a new clinical trial
* FDA advisers to review device on Thursday
* Mela shares down as much as 56 pct
* Analysts see new trial, Mela may need to raise money (Adds analyst comments, details; updates stock movement)
WASHINGTON, Nov 16 (Reuters) - U.S. health regulators said Mela Sciences Inc's (MELA.O) experimental device to help diagnose deadly skin cancer could cause harm because of the potential for misdiagnosis, wiping out more than half of the company's market value.
FDA staff pointed to numerous problems with Mela's study of the device, called MelaFind, including a significant lack of data, and urged a new clinical trial. But analysts saw major stumbling blocks ahead and were doubtful the company had enough resources to conduct another study.
"The FDA review team has significant concerns this device has not been studied adequately for its current indications for use and therefore puts the health of the public at risk," the U.S. Food and Drug Administration said in documents released on Tuesday.
Irvington, New York-based Mela's shares were down 53 percent to $3.01 in heavy volume trading on Nasdaq. They had touched a low of $2.80 earlier in the day.
The documents were released ahead of an FDA advisory panel meeting Thursday to weigh whether to approve the device for use in detecting melanoma. While the FDA will make the final approval decision, it often follows the advice of its panelists.
The concerns raised by the FDA staff are "problematic at its best" and investors should be warned, WBB Securities analyst Steve Brozak said.
"This now comes down to two main questions -- does Mela have enough resources to address the issues and, if not, what is the salvage value of the technology," Brozak added.
As of Sept. 30, Mela had about $35.8 million in cash and cash equivalents and said it was enough to fund operations for at least 12 months.
However, the company had also said in a regulatory filing that it might require additional funds to "achieve significant commercialization of MelaFind."
Boenning & Scattergood analyst Greg Chodaczek said the company would need to raise money to do a new trial.
Failing to raise capital could also prompt Mela to look for a partner, Chodaczek said, noting that bigger firms with diagnostics business such as General Electric Co's (GE.N) healthcare unit GE Healthcare and Siemens AG (SIEGn.DE) might be interested in the technology.
Chodaczek expects MelaFind to get approved by the first quarter of 2011 and sees sales of $9.8 million in 2011 and $42.1 million in 2012.
MELA DISPUTES FDA'S VIEW
The company has already seen various delays, with the FDA asking for more data and later postponing the meeting of outside advisers now scheduled for Thursday.
In its own documents, Mela outlined 13 areas of disagreement with the FDA. The company has said its noninvasive computerized system can help physicians diagnose early melanoma when more information is needed in deciding whether a skin biopsy is necessary.
Mela said the FDA staff believes that "the fatal risk of missing melanomas (two of 127) is not worth the marginal benefit of a clinically meaningless reduction in biopsy ratio compared to dermatologists."
The company said this view was "completely without context."
Boenning's Chodaczek said that while the device missed some melanoma cases, it still has significant benefits compared with dermatologists, who miss more.
"You are comparing the device with the best of the best dermatologists in the country ... who miss about 15 percent to 20 percent of the melanoma cases," he told Reuters.
Brozak, who has a "sell short" rating on the stock, said the shares were heavily overpriced, but does not see Tuesday's sell-off as an overreaction.
More than 10 million shares, which is 10 times the company's 10-day average volume, had changed hands by Tuesday afternoon.
The stock had fallen 38 percent this year, before Tuesday's losses.
FDA posted the documents on its website at link.reuters.com/kev65q. (Reporting by Susan Heavey and Esha Dey; editing by John Wallace and Gerald E. McCormick)