Patricia F. Fitzpatrick Dimond Ph.D. Technical Editor of Clinical OMICs President of BioInsight Communications

Improving the Predictability of Gene Expression Profiling Assays

More than 1.2 million new cases of colorectal cancer (CRC) occur globally each year, resulting in about 600,000 deaths. CRC remains the third most commonly diagnosed cancer in both men and women in the United States; estimated new cases in 2015 will reach 132,700.

Unfortunately, the majority of CRC cases continue to elude early detection. About 50% of CRC patients are diagnosed at advanced tumor stages. Such patients have poor prognoses. Earlier diagnoses, and better prognoses, may be attained by means of early screening for CRC. Implementing such screening has become one of the greatest public health challenges over the last 50 years.

A means of reliably predicting gene expression profiles (GEPs) for those who will develop CRC has proven hard to develop. And questions about the ultimate clinical utility of predictive tests for Stage II patients who require chemotherapy remain, even as companies continue to introduce new tests.

Historically, CRCs have been considered a single disease entity that shares the same cause, clinical characteristics, and treatment outcomes. But analysis of precursor lesions and hereditary forms of the disease has revealed CRC to be a complex and heterogeneous disorder that occurs relatively rarely in a hereditary form. The disease more commonly occurs sporadically.

Colorectal tumors arise through a gradual series of histological changes, the so-called adenoma-carcinoma sequence, with each change accompanied by a genetic alteration in a specific oncogene or tumor suppressor gene. Mutations in the adenomatous polyposis coli (APC) gene are not only responsible for familial adenomatous polyposis (FAP), but also play a rate-limiting role in the majority of sporadic colorectal cancers.

FAP arises as a result of a germline mutation on chromosome 5q21. Individuals who inherit a mutant APC gene have a high likelihood of developing colonic adenomas. The risk is estimated to be more than 90%. Characterized by hundreds of adenomatous colorectal polyps, the disease inevitably progresses to colorectal cancer, and it occurs in patients aged 35 to 40, on average.

About 75% of patients with CRC have sporadic disease with no apparent evidence of having inherited the disorder; the remaining 25% of patients have a family history of CRC that suggests a hereditary contribution, common exposures to carcinogens among family members, or a combination of both.

To date at least three distinct molecular pathways to CRC have been described: the conventional suppressor pathway characterized by mutations in APC and exemplified by FAP; the serrated pathway characterized by aberrant cytosine-guanosine island methylation; and the hereditary microsatellite instability pathway found in Lynch syndrome, with different genes being mutated or altered in carcinomas arising via the same genetic pathway.

Gene Expression Profiles

About 20% of patients with Stage II CRC experience a relapse after surgery. Current clinical-pathologic stratification factors do not allow clear identification of these high-risk patients. Over a dozen different GEP tests have been developed for use as prognostic markers in Stage II colon cancer. These assays are intended to help identify patients with Stage II colon cancer who are at high risk for recurrent disease and are candidates for adjuvant chemotherapy.

To date, no gene expression test for evaluation of prognosis in Stage II colon cancer has been cleared for marketing by the FDA. These tests are offered as laboratory-developed assays in clinical laboratory improvement amendment (CLIA)-licensed laboratories operated by each company and currently do not require FDA premarket review as a result of enforcement discretion.

Several tests have undergone evaluation and continue to be evaluated in multiple validation studies, including ColonPRS® and Signal ColonPRS® (Signal Genetics), Coloprint® (Agendia), Genefx Colon® (Precision Therapeutics), OncoDefender™-CRC (Everist Genomics), and Oncotype DX® colon cancer test (Genomic Health).

Independent validation studies ranging in size from 33 to 1,436 patients have been reported on some of these assays. Depending on the manufacturing process, GEPs may be marketed as either in vitro devices or laboratory-developed tests (LDTs). In the latter case, GEPs are used solely within a particular lab and are not distributed or sold to any other labs or healthcare facilities. As a result, GEPs marketed as LDTs may enter the U.S. market with analytical validation under laboratory regulations imposed by CLIA but without evidence of clinical validity or utility.

Two examples of CLIA LDTs are Oncotype Dx, a 12-gene RT-PCR assay, and Coloprint, an 18-gene microarray. Both tests are intended to predict the development of distant metastasis of patients with Stage II colon cancer and facilitate the identification of patients who may be safely managed without chemotherapy.

Oncotype Dx was developed by evaluating 761 colon cancer-related genes in 1,851 patients with resected colon cancer in four large development studies. These development studies led to the definition of the test’s 12-gene panel (which includes 7 cancer-related genes and 5 reference genes) as well as the Recurrence Score® (RS) algorithm. The test, which was prospectively validated in the QUASAR validation study, generates a score that predicts the risk of recurrence for individuals with Stage II or Stage III colon cancer.

Investigators at the Mayo Clinic and other institutions prospectively evaluated the impact of RS results on physician recommendations regarding the use of adjuvant chemotherapy in T3 stage (according to the tumor, node, and metastasis [TNM] system), mismatch repair-proficient (MMR-P) Stage II colon cancer patients. For each patient, the investigators compared the physician’s recommended postoperative treatment plan of observation, fluoropyrimidine monotherapy, or combination therapy with oxaliplatin before and after the RS.

Of 141 T3 MMR-P patients eligible for the primary analysis, treatment recommendations changed for 63 (45%; 95% confidence interval: 36–53%), with intensity decreasing for 47 patients (33%) and increasing for 16 (11%). Recommendations for chemotherapy decreased from 73 patients (52%) to 42 (30%).

Comparison to Traditional Assessments

Compared with traditional clinicopathological assessment, the results of the study showed that incorporation of the RS result into clinical decision making changed treatment recommendations for 45% of T3 MMR-P Stage II colon cancer patients in this prospective multicenter study. Use of the RS assay, the investigators concluded, may lead to overall reduction in adjuvant chemotherapy use in this subgroup of Stage II colon cancer patients.

ColoPrint was validated in several independent cohorts of Stage II CRC patients at multiple institutions, with the aim being to enroll 785 eligible patients to validate the performance of ColoPrint in estimating three-year relapse rate. In a pooled data set from 416 patients in a study comparing Coloprint assessments with those obtained through clinical risk factors described in the National Comprehensive Cancer Network (NCCN) 2013 Guidelines for Colon Cancer, ColoPrint could distinguish between patients at high and low risk for relapse.

Clinicians say that although information is emerging about use of GEP assays to inform the decision about use of adjuvant chemotherapy in patients with Stage II colon cancer, studies to date have not provided the type of information needed to address major uncertainties.

Insurance companies consider these tests investigational. “It seems unlikely that the 12-gene expression test for predicting colon cancer recurrence risk in individual patients could guide clinical decision making,” says one major medical insurance company. “The differences between recurrence risk categories established in the validation study were not sufficiently discriminative, and the associated confidence intervals overlapped considerably.”

For example, an insurer might say, an RS of 33 corresponds to a recurrence risk of 16%, which could be low risk, intermediate risk, or high risk when confidence intervals of the category mean estimates are considered. To help guide interpretation, Genomic Health has established RS category ranges for low (<30), intermediate (30–40) and high (≥41) risk categories.

Insurance companies also point out that the validation study lacks information on how the test adds to current methods of predicting risk, and on how extensively and accurately patients are reclassified by RS. Relevant issues concern quality of classification by current predictors and comparisons to known recurrence outcomes.

Other researchers investigated whether four well-documented gene expression risk scores could improve patient stratification compared the TNM staging system from the American Joint Committee on Cancer.

Microarray-based versions of risk scores were applied to an independent cohort of 688 Stage II/III tumors from the Pan-European Trials in Alimentary Tract Cancers trial, a Phase III randomized trial that yielded pathologic samples evaluable for immunohistochemistry and molecular testing from 1,404 patients with Stage II/III cancer. Prognostic value assessments were made for relapse-free survival (RFS), survival after relapse (SAR), and overall survival (OS).

All four risk scores showed a statistically significant association (single test, P < .0167) with OS or RFS in univariate models, but with hazard ratios below 1.38 per interquartile range. Three scores were predictors of shorter RFS, one of shorter SAR. Each risk score could, the authors said, only “marginally improve” an RFS or OS model with the known factor’s T stage, N stage, and microsatellite instability status. A combination of the risk scores, they said, might provide more robust information, but they added that RFS and SAR predictors might need to be different.

Identification of patients with Stage II and III colon cancer who will benefit from adjuvant treatment remains a major clinical challenge. Many patients treated with surgery alone experiences relapse; some patients receiving chemotherapy would be cured with surgery alone.

The use of GEP testing in conjunction with clinicopathological parameters has clear potential for improved treatment decisions based on risk assessment, but physicians have concluded that their clinical implementation awaits prospective validation.

Patricia Fitzpatrick Dimond, Ph.D., has been a long-time contributor to GEN and currently serves as the publication’s Technical Editor. She is also president of BioInsight Communications. During her career in the biotechnology industry, she was VP of strategic development and corporate communications at Coley Pharmaceutical Group (now Pfizer), where she developed and managed the company’s investor relations and communications programs. ([email protected])


This article was originally published in the March 2015 issue of Clinical OMICs. For more content like this and details on how to get a free subscription to this digital publication, go to

Previous articleGenes from Dad Used More Than Genes from Mom
Next articleAbbVie Acquires Oncology Drug Firm Pharmacyclics for $21B