Scientists offered new hope in the fight against a difficult-to-treat skin cancer on Wednesday as an early-stage clinical trial showed an experimental drug dramatically shrank tumors.

Another study also found Roche's established cancer drug Avastin had promise in treating melanoma, but it failed to show a statistically significant improvement in survival.

Paul Chapman at Memorial Sloan-Kettering Cancer Center in New York said 70 percent of patients with a particular gene mutation saw their tumors shrink when given the new drug PLX4032.

To put that into context, chemotherapy is about 13 to 15 percent, and that's where we have been stuck with conventional treatments, Chapman told reporters at the ECCO-ESMO cancer congress in Berlin.

Of the 27 patients evaluated in the small Phase I study, two responded completely and signs of their disease disappeared.

Preliminary findings with the new drug were first reported in June but the new data from more patients has increased confidence in the approach.

Alexander Eggermont, president of the European Cancer Organization, described the trial as simply spectacular and showed the benefits of targeting treatment.

The news would transform melanoma work into a very exciting field instead of a graveyard, he said.

PLX4032, which Roche is developing with privately-held U.S. biotech company Plexxikon, is the latest in a growing group of gene-targeted drugs that oncologists believe will drive future cancer treatment.

It works by blocking the activity of the cancer-causing mutation of the BRAF gene, which is implicated in 50 to 60 percent of melanomas.

AVASTIN MISSES

The failure of Avastin -- a $5 billion-a-year seller in colon, lung and breast cancer -- confounded expectations of doctors in Berlin, since a late-breaking abstract published earlier had reported a meaningful survival benefit.

But a last-minute review of the clinical data changed the outcome and researchers said the Phase II study in fact missed its goal, despite an encouraging trend.

Medical experts said larger trials might yet prove Avastin worked in melanoma, noting that the drug also initially failed to show significance in breast cancer.

Despite the disappointment, principal investigator Steven O'Day of the Angeles Clinic and Research Institute in California said the data collected on Avastin were very encouraging and warrant continued investigation.

The failure of Avastin -- a $5 billion-a-year seller in colon, lung and breast cancer -- confounded expectations of doctors in Berlin, since a late-breaking abstract published earlier had reported a meaningful survival benefit.

But a last-minute review of the clinical data changed the outcome and researchers said the Phase II study in fact missed its goal, despite an encouraging trend.

Medical experts said larger trials might yet prove Avastin worked in melanoma, noting that the drug also initially failed to show significance in breast cancer.

They also said there was a potential for combining Avastin with the new gene-targeted drug in future.

Despite the disappointment of the Avastin study, lead investigator Steven O'Day of the Angeles Clinic and Research Institute in California said the data collected on Avastin were very encouraging and warrant continued investigation.

The 214-patient study found the median overall survival in the Avastin arm was 12.3 months, against 9.2 months in the control arm, but there was a 19 percent likelihood that this result happened by chance.

To prove statistical significance, clinical trials must show a likelihood of 5 percent or less that a finding is due to chance.