New analytical approach that combines genetic and clinical data to give individualized prognosis of cancer recurrence

Researchers at Duke University have developed a new analytical approach that combines genetic and clinical data to give cancer patients an individualized prognosis of their cancer recurrence. The information could prove critical in deciding how aggressively to treat the disease following surgery.

The researchers reported their findings using breast cancer as a test case in the April 26, 2004, online early edition of the Proceedings of the National Academy of Sciences.

"Currently, it is primarily traditional clinical information alone that aids in understanding a patient's risk profile," said Mike West, Ph.D., Arts & Sciences professor of statistics and decision sciences at Duke and lead author on the study. "However, the resulting predictions typically lump patients into broad categories. Access to detailed genomic information now provides the opportunity to move far beyond this -- toward customized risk predictions and prognoses more widely, for the individual patient."

Nevertheless, most previous studies that have focused on developing genomic-based predictors of cancer recurrence risk have only broadly defined patients as high versus low risk, leaving considerable room for error about an individual's true chance of recurrence, the researchers said. The power of the Duke team's approach to improve such predictions lies in the combined use of multiple sources of clinical and genomic data, they said.

In their case study of breast cancer, the Duke team developed methods that utilize diverse information including traditional clinical variables, such as lymph node and estrogen receptor status, and multiple, complex patterns of gene activity, or "genetic fingerprints," of a patient's tumor. They integrated these data to formulate unique predictions about individual patients' recurrence risk. While the Duke study focused on patients with breast cancer, the approach is applicable more broadly to any form of cancer and can incorporate any type of information relevant to disease outcome, said the researchers.

"Cancer is an extremely heterogeneous disease," said co-author Joseph Nevins, Ph.D. "Therefore, every cancer has its own distinct properties. Our approach allows us to capture characteristic patterns underlying those different disease states and to utilize that information to make informed predictions about a patient's risk of recurrence that can then be applied to make the best treatment decisions." Nevins is director of the Center for Genome Technology of the university's Institute for Genome Sciences and Policy, and James B. Duke professor of molecular genetics and microbiology at Duke University Medical Center.

The study also highlights the importance of a multidisciplinary approach -- including clinical, genomic, statistical and computational scientists -- for bringing genomic discoveries and technologies to clinical application, which are core goals of the Duke IGSP, said West.

The new study builds on the Duke team's earlier research, published in the May 10, 2003, issue of The Lancet, indicating that genetic profiles could accurately discriminate breast cancer patients at very high risk of the cancer spreading to the lymph nodes -- a critical factor for long-term survival -- from those at low risk.

In the current study, the researchers demonstrated their method in a group of 158 breast cancer patients seen at the Koo Foundation Sun Yat-Sen Cancer Center in Taipei, for which primary tumor biopsies had been collected and banked between 1991 and 2001.

The researchers measured the activity of more than 12,500 genes in the patients' tumor samples using DNA microarray, or gene chip, technology. Gene chip technology works by measuring the relative abundance of messenger RNA (mRNA) -- molecules that serve as templates for the synthesis of proteins that carry out the myriad functions of the cell. The resulting gene expression profiles provide "snapshots" of the tumor state at the time of surgery. The Duke researchers analyzed these gene expression profiles to identify clusters of genes, or "metagenes," with related characteristics.

The team then used this genomic information along with standard clinical factors to organize patients into categories representing different risk groups with respect to cancer recurrence. Their statistical method, utilizing so-called "classification trees," successively splits patients into ever smaller groups having similar risk profiles. Such an analysis is repeated multiple times according to different classification schemes. Those many solutions are then summarized into precise estimates of each patient's unique cancer recurrence risk.

Models combining clinical and genomic information performed far better than those built with genomic data alone, the team found. That result suggests that new genomic information should be used to augment, rather than replace, more traditional clinical predictors of risk, said the researchers.

"The framework allows us to utilize all of the information available -- both traditional variables and gene expression data -- to sort through the disease heterogeneity and get much closer to a personalized prediction of disease outcome. There's no reason clinical and genomic predictors must be an 'either/or' proposition," Nevins said.

Importantly, the researchers added, their analysis also provides a measure of the confidence with which recurrence risk can be estimated for a given patient. Such measures of confidence are critical, they said, because some patients' recurrence risk might be less clear than others, information that might determine how heavily patients' should weigh their personalized risk prediction.

"If relevant clinical and genomic prognostic factors provide conflicting information about a patients risk status, it is critical that the resulting uncertainty be properly represented and communicated to adequately inform -- and indeed caution -- the clinical interpretation of a patient's estimated risk, and that efforts be made to dissect and understand the underlying basis for that conflict," West said.

The newly developed predictive approach is now being further extended and evaluated in additional patient groups. Once complete, said Nevins, the test would be ready to incorporate into clinical practice in a limited way. For instance, patients might use the information about their relative risk to decide between a more or less aggressive therapy.

The method has other uses as well, West noted. For example, the analysis could provide information about patients' responsiveness to particular treatment methods, measures that could be usefully applied in the evaluation of clinical trials and selection of the most effective treatment regimens.

The Duke project involves collaborators Jennifer Pittman, Ph.D., Erich Huang, M.D., Holly Dressman, Ph.D., Cheng-Fang Horng, Andrea Bild, Ph.D., and Edwin Iversen, Ph.D., all of Duke, as well as Skye Cheng, M.D., Mei-Hua Tsou, M.D., Chii-Ming Chen, M.D., and Andrew Huang, M.D., of the Koo Foundation Sun Yat-Sen Cancer Center in Taipei.

Comments

The opinions expressed here are the views of the writer and do not necessarily reflect the views and opinions of News Medical.
Post a new comment
Post

While we only use edited and approved content for Azthena answers, it may on occasions provide incorrect responses. Please confirm any data provided with the related suppliers or authors. We do not provide medical advice, if you search for medical information you must always consult a medical professional before acting on any information provided.

Your questions, but not your email details will be shared with OpenAI and retained for 30 days in accordance with their privacy principles.

Please do not ask questions that use sensitive or confidential information.

Read the full Terms & Conditions.

You might also like...
Researchers uncover why the TP53 gene is especially prone to mutations in cancer