Technology assessment of molecular pathology testing for the estimation of prognosis for common cancers

Abstract

OBJECTIVES:
To conduct a systematic review and meta-analysis assessing the prognostic value and test performance (analytic validity, clinical validity, clinical utility, and harms) associated with 11 prognostic molecular pathology tests. Many of these tests are indicated for prediction of therapeutic responses, but this review focuses on their potential use as prognostic factors. Our overarching question was whether there is direct evidence that the addition of these molecular pathology tests changed physician decision making and improved outcomes for adult patients. We evaluated the following tests: microsatellite instability assessment by polymerase chain reaction (PCR) for colorectal cancer (CRC), MLH1 promoter methylation for CRC, KRAS mutation testing for CRC, BRAF mutation testing for CRC, Oncotype DX Colon mRNA expression for CRC, Oncotype DX Breast mRNA expression for breast cancer, MammaPrint mRNA expression for breast cancer, ALK cytogenetics for lung cancer, EGFR mutation testing for lung cancer, KRAS mutation testing for lung cancer, and UroVysion cytogenetics for urinary bladder cancer.

DATA SOURCES:
PubMed®, the Cochrane Library, and EMBASE® (through November 2013); reference lists of pertinent review articles and included studies, test developers’ Web sites, ClinicalTrials.gov, the Food and Drug Administration (FDA) Web site, Health Services Research Projects in Progress, the European Union Clinical Trials Register, the College of American Pathologists (CAP), and data submitted by companies that developed the tests.

REVIEW METHODS:
Two investigators independently selected, extracted data from, and rated risk of bias of each study. For clinical validity, we conducted meta-analyses to estimate weighted summary hazard ratios when three or more studies reported an eligible outcome. For the other review questions, we did not find sufficient data to conduct meta-analyses. Therefore, we synthesized the information qualitatively. We graded strength of evidence based on established guidance.

RESULTS:
Evidence from multiple studies supports associations between test results and prognosis, with added value beyond known independent prognostic factors, for MammaPrint, Oncotype DX Breast, KRAS mutation testing for lung cancer, BRAF and KRAS mutation testing for CRC, and microsatellite instability for CRC for at least one of our included outcomes (i.e., risk of recurrence, cancer-specific survival, or overall survival). Although UroVysion is marketed as a diagnostic rather than a prognostic test, limited evidence from two small studies (total N=168) rated as low or medium risk of bias supported associations between test result and prognosis for risk of recurrence. We found no studies that directly assessed the impact of a test of interest on both physician decision-making and downstream health outcomes to establish clinical utility. We attempted to construct an indirect chain of evidence to answer the overarching question, but we were unable to do so. Even in the cases where the tests seemed to add value in determining prognosis (i.e., evidence of clinical validity), we found no evidence that using the test was related to improved outcomes for patients. However, for impact of test use on treatment decisions, we found moderate strength of evidence that Oncotype DX Breast leads to changes in treatment decisions. Although the decision changes were observed in both directions for individual patients, studies consistently showed an overall shift to less-intensive treatment recommendations as a result of using Oncotype DX Breast, with fewer recommendations for chemotherapy (and less exposure to potential harms of chemotherapy).

LIMITATIONS:
We were not able to directly address the prognostic value of these tests for the Medicare population. There were several studies that included patients from the Medicare age group and although the impact of these tests were not analyzed with specific reference to this population, we did not come across any evidence that suggested that the prognostic value of these tests would be different for the Medicare population.

CONCLUSIONS:
Modest evidence supports added prognostic value (i.e., clinical validity) for over half of the tests evaluated, and that Oncotype DX Breast leads to changes in treatment decisions, but we found no evidence to determine whether using the tests to estimate prognosis leads to improved outcomes for patients.