Laird Harrison

May 13, 2014

CT texture analysis can predict the survival of patients with metastatic melanoma who are receiving antiangiogenic therapy, a new study shows.

"We're trying to find ways to predict how someone is going to do with metastatic melanoma early in their therapy," said Andrew D. Smith, MD, PhD, an assistant professor of nuclear medicine and body imaging at the University of Mississippi in Jackson.

"If they're not doing well on the therapy, there's no point continuing with it," he told Medscape Medical News.

He presented the results at the American Roentgen Ray Society 2014 Annual Meeting in San Diego.

Researchers and clinicians typically use change in tumor size, measured with the Response Evaluation Criteria in Solid Tumors (RECIST), to evaluate the effectiveness of melanoma therapies, Dr. Smith explained.

However, melanoma tumors often change very little, leading to a RECIST category of "stable disease." But the disease is often not stable in these patients. "About half of them do poorly, meaning the drug will fail," he reported. In contrast, some people do extraordinarily well.

To separate these groups from each other, the researchers tried to determine whether antiangiogenic drugs change a tumor's blood vessels. "Presumably, if it does not change the vascularity, the drug is not working. If it is changing, you can quantify it," Dr. Smith said.

In a phase 2 trial of bevacizumab (Avastin), the researchers evaluated change in vascularity in the melanoma tumors of 44 patients. To do this, they looked at changes in mean positive pixels in contrast-enhanced CT images.

Three of the 44 patients fit the RECIST category of "partial response," 30 fit "stable disease," and 11 fit "progressive disease."

The researchers analyzed the images before and after applying various spatial filters at different anatomic scales using TexRAD software.

They used statistical algorithms to assess the association between a patient's overall survival and image findings, texture measurements, and baseline levels of serum lactate dehydrogenase (which is associated with melanoma survival).

In a multivariate analysis of 24 patients with stable disease and a mean overall survival of 1.6 years, change in mean positive pixels at spatial filter 4 was deemed to be a significantly strong predictor of overall survival (P = .004).

The risk for mortality was 10 times higher in patients with a decrease in mean positive pixels of at least 10% than in those with a decrease of less than 10%.

Percent size change from baseline and baseline lactate dehydrogenase were also associated with overall survival.

A prognostic index that took into account mean positive pixels, lactate dehydrogenase, and size was used to categorize patients with stable disease. The 15 patients in the favorable group had a median overall survival of 2.4 years, and the 9 patients in the unfavorable group had a median overall survival of 1.0 years

The index was 97.9% accurate in predicting overall survival at 18 months. In contrast, a model that evaluated only change in the size of target lesions on the initial post-therapy CT and baseline lactate dehydrogenase level was 76.9% accurate.

Preliminary Results

Although promising, the results of this trial are only preliminary. "I view this as a pilot study. It would be nice to have 2000 patients. It's a rare disease and it has a dismal diagnosis, so we measured just a small patient population," Dr. Smith said.

If larger trials validate mean positive pixels as an imaging biomarker, it could be used to save patients from suffering through adverse effects from therapies that aren't helping them and the expense to the healthcare system, he explained.

In addition, mean positive pixels could speed up the work of clinical trials by helping investigators understand results more quickly, he said.

The approach looks promising, said Perry J. Pickhardt, MD, professor of radiology at the University of Wisconsin in Madison and gastrointestinal section editor at the American Journal of Roentgenology.

"Detailed analysis of textural features of tumors on CT, which cannot be discerned by visual review of the images, appears to add useful additional information beyond existing biomarkers," he told Medscape Medical News in an email.

"Texture analysis may help address limitations in the current size-based approach to treatment response," added Dr. Pickhardt. "Further work along these lines is needed."

Dr. Smith reports that he has a research grant from Pfizer to study metastatic renal cell carcinoma, and is president and founder of Radiostics, a company that interprets oncologic trials. Dr. Pickhardt has disclosed no relevant financial interest.

American Roentgen Ray Society (ARRS) 2014 Annual Meeting: Abstract 173. Presented May 9, 2014.


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.