DNA microarrays have been used increasingly in research and clinical applications; however, drawing conclusions from the results have been problematic because of difficulty with accurate reporting of results, experimental reproducibility, and identification and interpretation of relevant information. In their 2008 review, Walker and Hughes point out that studies utilizing DNA microarrays for molecular profiling have been aimed at stratifying the severity of disease, identifying disease subtypes, predicting prognosis, and predicting benefit from adjuvant therapy. They state that, while collectively these studies do provide evidence of improvements in understanding of disease and patient care, the high costs and doubts about reproducibility have limited clinical use of these new tools. The analysis of the high volume of data is more complicated than anticipated. Some inconsistencies include those between data from different experimental platforms, as well as between laboratories using the same experimental platform. They concluded that the “original excitement surrounding DNA microarrays has given way to a feeling in the scientific community that expectations have exceeded reality. There have been difficulties in applying the technology, in achieving reproducible results and in the management of high throughput data.”
Analytic Validity (Technical performance, i.e., reproducibility)
Fresh Frozen Tumor Sample
In 2008, Dumar and colleagues analyzed performance characteristics of the Pathwork test in a cross-laboratory comparison study of 60 poorly and undifferentiated metastatic (77%) and primary (23%) tumors. Three academic and one commercial laboratory received archived frozen tissue specimens for procurement and processing at their individual sites. Steps performed by each of the four laboratories included tissue handling, RNA extraction, and microarray-based gene expression assays using standard microarray protocol. The resulting microarray data generated at each laboratory were sent in a blinded fashion to Pathwork Diagnostics for generation of similarity scores for each type. Reports of the similarity scores were sent back (blinded) to the pathologists at the four laboratories for their use in generating an interpretation. Data were compared among the four laboratories to determine assay reproducibility. Correlation coefficients were between 0.95 and 0.97 for pathologists’ interpretations of the similarity scores, and cross-laboratory comparisons showed an average 93.8% overall concordance between laboratories in terms of final tissue diagnosis. A detailed summary of the data is available online at <http://www.accessdata.fda.gov>.
Formalin-fixed, paraffin-embedded (FFPE) Tumor Sample
Analytical performance characteristics of the Pathwork test for FFPE were analyzed in a cross-laboratory comparison study of 60 poorly and undifferentiated metastatic (45%) and primary (35%) tumors. Each of the 15 tumor tissue types were represented by four specimens each, with the exception of breast (n=3) and soft tissue sarcoma (n=5). Samples were distributed among three laboratories for procurement and processing at their individual sites. Data were compared among the three laboratories to determine assay reproducibility. Correlation coefficients were between 0.92 and 0.93 for pathologists’ interpretations of the similarity scores, and cross-laboratory comparisons showed an average 82.1% overall concordance between laboratories in terms of final tissue diagnosis. A detailed summary of the data is available online at: <http://www.accessdata.fda.gov>. Additional analyses of the analytic performance of the test have produced similar results.
Clinical Validity (Sensitivity and specificity)
Fresh Frozen Tumor sample
The clinical validation study for the Pathwork Tissue of Origin test that was submitted to the FDA involved a comparison of the gene expression profiles of 25 to 69 samples to each of the 15 known tumors on the Pathwork panel (average 36 specimens per known tumor). The specimens included poorly differentiated, undifferentiated, and metastatic tumors. A similarity score was given to 545 specimens and then compared to the available specimen diagnosis. Based on the 545 results, the probability that a true tissue of origin call was obtained, when a similarity score of 30 or more was reported, was 92.9% (95% CI: 90.3–95.0), and the probability that a true negative tissue call was made, when a similarity score of five or less was reported, was 99.7% (95% CI: 99.6–99.8%). Overall, the Pathwork performance comparing the profiles of the 545 specimens to the panel of 15 known tumor types showed a positive percent agreement of 89.4% (95% CI: 86.5-91.8%), negative percent agreement of 99.6% (95% CI: 98.6–100%], non-agreement of 6.2% (95% CI: 4.4–8.6%), and indeterminate of 4.4% (95% CI: 2.8–6.5%).
Monzon and colleagues conducted a multicenter blinded validation study of the Pathwork test. The specimens included poorly differentiated, undifferentiated, and metastatic tumors. A total of 351 frozen specimens and electronic files of microarray data on 271 specimens were obtained, with 547 meeting all inclusion criteria. A similarity score was given to the specimens, which was then compared to the original pathology report that accompanied the specimen. Overall, the Pathwork performance comparing the profiles of the 547 specimens to the panel of 15 known tumor types showed an overall agreement of 87.8% (95% CI: 84.7–90.4%) with the reference diagnosis. Sensitivity and specificity were 87.8% (95% CI: 84.7–90.4%) and 99.4% (95% CI: 98.3–99.9%), respectively, with the original pathology report acting as the reference standard. The authors acknowledged that since there was no independent confirmation of the original pathology, using the pathology reports as the reference standard could introduce errors into the study results. Agreement differed by site: 94.1% for breast, 72% for both gastric and pancreatic. Performance differences between tissue sites were statistically different (chi-squared=42.02; p=0.04; degrees of freedom [df]=28; n=547). Rates of agreement between test result and reference diagnosis varied by site: 88%, 84.4%, 92.3%, and 89.7% for Clinical Genomics facility, Cogenics, Mayo Clinic, and the International Genomics Consortium, respectively, but these differences were not statistically significant.
Formalin-fixed, paraffin-embedded (FFPE) tumor sample
The clinical validation study for the Pathwork Tissue of Origin Test Kit-FFPE that was submitted to the FDA involved a comparison of the gene expression profiles of 25 to 57 samples to each of the 15 known tumors on the Pathwork panel (average 31 specimens per known tumor). The specimens included poorly differentiated, undifferentiated, and metastatic tumors. A similarity score was given to 462 specimens and then compared to the available specimen diagnosis. Based on the 462 results, the probability that a true tissue of origin call was obtained when a similarity score was reported was 88.5% (95% CI: 85.3-91.3%), and the probability that a true-negative tissue call was made when a similarity score of 5 or less was reported was 99.8% (95% CI: 99.7–99.9%). Overall, the Pathwork performance comparing the profiles of the 462 specimens to the panel of 15 known tumor types showed a positive percent agreement of 88.5% (95% CI: 85.3-91.3%), negative percent agreement of 99.1% (95% CI: 97.6–99.7%], non-agreement of 11.5% (95% CI: 8.7–14.7%). Further details of these data are available online at <http://www.accessdata.fda.gov>.
Few other studies have analyzed the clinical validity of using microarray gene expression technology. One study used microarray technology (i.e., CupPrint, Agendia, Amsterdam, the Netherlands) that used formalin-fixed paraffin-embedded tumor samples. The study analyzed 495 genes in 84 patients with tumors of known origin and 38 patients with cancer of unknown primary to assess the potential contribution to patient management. Sixteen of the patients with cancer of unknown primary had their primary site of tumor origin identified by standard laboratory techniques. Molecular testing identified the correct site of tumor origin in 94% of cases of cancers of unknown primary and 83% of the tumors of known origin. Ferracin and colleagues published a report of microRNA profiling using 101 FFPE tumor samples from primary cancers and metastases. Forty samples, of 10 cancer types, were used to build a cancer-type-specific microRNA signature. This signature was then used to predict the primary site of metastatic cancer. Overall accuracy was 100% for primary cancers and 78% for metastatic cancers in the cohort sample. The signature was then applied to a published set of 170 samples where the prediction rates were consistent with the cohort results.
Clinical Utility (Impact on patient outcomes)
No clinical trials have been conducted that would provide direct evidence of the clinical utility of the Pathwork Tissue of Origin test, nor has the clinical application of gene expression profiling to direct patient management and tumor site-specific therapy been demonstrated in prospective studies.
One small study using microarray technology (not Pathwork) on formalin-fixed paraffin-embedded tumor, retrospectively analyzed the gene expression profile of tumors from 21 patients with cancer of unknown primary. The clinical relevance and implications of the results on patient management were reviewed. In the 21 patients, standard methods had failed to determine a primary tumor origin. Results of gene expression profiling were reviewed in the context of tumor histology and clinical suspicion of tumor origin. Gene expression profiling confirmed the clinical suspicion in 16 of 21 cases, with a clinical/gene expression profile inconsistency in four of 21 and a pathological/gene profile inconsistency in one patient. The authors concluded that the use of gene expression profiling would have influenced patient management in 12 of 21 of the cases.
In June 2009, final data collection for the primary outcomes was completed in a study aimed at the identification of the tissue of origin in patients with metastatic tumors of unknown primary site. As of October 2011, no results have been published in the peer-reviewed literature.
An October 2011 search of the National Cancer Institute and ClinicalTrials.gov databases returned no ongoing Phase II or III studies investigating the use of molecular gene expression profiling with microarray technology in patients with cancer of unknown primary.
Practice Guidelines and Position Statements
National Comprehensive Cancer Network (NCCN) guidelines for the workup of an occult primary malignancy address the use of molecular methods in the classification of tumors. They conclude that there is insufficient data to confirm whether gene expression profiling can be used in choosing treatment options that would improve the prognosis of patients with occult primary cancers. Therefore the panel does not recommend the testing as a part of routine evaluation of a cancer of unknown primary origin.
Limited data have been published on the clinical impact of the Pathwork test. Without knowledge of how this test would affect clinical practice and clinical health outcomes (clinical utility) for patients diagnosed through the use of this test, the experimental, investigational and unproven coverage statement remains unchanged. A trial where patients with a cancer of unknown primary were randomized to receive treatment based on the results of the Pathwork Tissue of Origin test or based on standard diagnostic procedures would be useful to determine the clinical utility of the Pathwork test.
Target Now™ Test
At the 100th Annual Meeting of the American Association for Cancer Research (AACR) in April 2009, lead researcher Dr. Daniel Von Hoff reported a pilot study of molecular profiling of tumors (Target Now), conducted in patients with advanced cancers that were progressing despite several previous treatments. This study found that molecular profiling helped identify therapies that ultimately had an impact on the disease. Shrinkage of tumors was shown in 47% of patients. Also, 27% of the profiled patients showed improvement in progression-free survival, compared with that seen with the previous therapy, and there is a suggestion of improved overall survival in these patients. Although Dr. Von Hoff said this is a promising result, he acknowledged that the study was small (66 patients) and that, because patients acted as their own controls, a larger randomized trial is needed.
Although GEP (e.g., the Target Now Test) is promising, there remain problems with accuracy and reproducibility of results, and with identification and interpretation of information obtained. A March 2009 report states Massachusetts General Hospital had decided to make gene testing standard in cancer treatment; it is believed to be the first hospital in the United States to do so. However, the report also states that doctors acknowledge that it is unclear whether screening patients for an expanded library of tumor defects will actually save money on drugs, or whether it will translate into longer lives. No peer reviewed clinical studies were identified that would change this conclusion. Therefore, GEP using microarray technology is not supported by evidence in the peer-reviewed medical literature that:
- Permits conclusions on the effect of GEP using microarray analysis on health outcomes.
- Demonstrates an improvement in net health outcome through use of GEP using microarray analysis.
- Demonstrates that improvement attainable by GEP using microarray analysis is attainable outside investigational settings.
A search of peer reviewed literature through January 2012 identified no new clinical trial publications or any additional information that would change the coverage position of this medical policy.
Disclaimer for coding information on Medical Policies
Procedure and diagnosis codes on Medical Policy documents are included only as a general reference tool for each policy. They may not be all-inclusive.
The presence or absence of procedure, service, supply, device or diagnosis codes in a Medical Policy document has no relevance for determination of benefit coverage for members or reimbursement for providers. Only the written coverage position in a medical policy should be used for such determinations.
Benefit coverage determinations based on written Medical Policy coverage positions must include review of the member’s benefit contract or Summary Plan Description (SPD) for defined coverage vs. non-coverage, benefit exclusions, and benefit limitations such as dollar or duration caps.