Low clinical adoption of tumor genomic profiling: cause for concern?

The way people with cancer are diagnosed, counseled, and treated has shifted dramatically in the precision oncology era. Precision medicine is broadly defined as “prevention and treatment strategies that take individual variability into account”¹. Practically speaking, these efforts encompass genetically profiling both tumor and germline tissues in an effort to tailor care for an individual patient. Success in precision oncology, however, has been mixed. Most clinical gains have been limited to tumor-specific advances, such as the utilization of BCR-ABL-targeted therapy for chronic myeloid leukemia (CML)². Other incremental successes include the recent immunotherapy indication for microsatellite instability-high (MSI-high) colorectal cancers (CRCs), and the development of mutation-directed therapies, such as osimertinib for the T790M EGFR mutation in metastatic non-small cell lung cancer^3,4.

Precision oncology has also produced novel indications for therapies, such as a germline-based approach to therapy for individuals with advanced ovarian cancer who possess pathogenic germline BRCA1/2 mutations (olaparaib) and a tumor agnostic indication for MSI-H or mismatch repair machinery-deficient metastatic tumors (pembrolizumab)^5,6. It remains unclear if these examples will become a universal trend in oncology, or whether tissue histology will continue to be the primary driver of most therapeutic decisions. The tumor agnostic approach has been promoted through dramatic success stories, but in practice has been profoundly underwhelming when rigorously investigated in prospective randomized clinical trials^7,8.

In a recent publication in the Journal of Medical Economics, authors from Foundation Medicine and consultants reported that the clinical uptake of tumor genomic profiling has been lower than anticipated⁹. They conclude that this deficiency represents an opportunity to better tailor treatment strategies for patients with cancer. This study has some important methodological limitations. For instance, the frequency measurement of tumor genomic profiling was restricted to the metastatic setting, which would not capture profiling performed on a primary tumor. Methodological issues aside, we question the conclusion that low uptake is characterized as a missed opportunity, since this notion rests on the unproven presumption that increased uptake of broad genomic profiling in a general cancer population is directly linked with improved clinical outcomes.

It is worth reviewing the ongoing challenges confronting the successful implementation of precision oncology, which might justify the finding of low uptake of genomic profiling. One major issue is the shifting definition of what constitutes a “targetable” mutation. Should a mutation be considered “targetable” if it is found in a tumor type for which no high quality prospective data exists regarding efficacy of the matched intervention? Or, should a mutation only be considered “targetable” if the utilization of the targeted therapy results in a meaningful improvement in patient outcomes, as demonstrated through well-run, prospective, randomized trials? We argue that testing and intervention should only be broadly adopted under the latter scenario. This more stringent expectation should not diminish the current value of well-established targeted therapies, nor should it preclude ongoing efforts to investigate novel therapeutic approaches in the context of clinical research. It would, however, bring pause to the widespread adoption of unproven therapies outside of research trials.

A formidable challenge to precision oncology is the fact that most tumors lack “targetable” mutations, even when utilizing imprecise definitions. It is unclear why head and neck cancers were included in this analysis, since there are no approved genomically-matched targeted therapies, and current guidelines do not recommend genomic testing. Human papilloma virus (HPV) testing is recommended, but this is performed by immunohistochemistry for p16, and subsequent treatments involve chemotherapy and radiation. Cetixumab is approved for head and neck cancer, but its use is not tied to any genomic sequencing results. In a study of 10,000 tumors across 11 cancer types, only 37% of tumors in a large prospective study possessed targetable mutations, and only 11% of patients were enrolled in clinical trials based on a genomic profile¹⁰. The likelihood of identifying a targetable mutation also varies enormously by tumor type. While 78% of tumors from patients with melanoma will possess a targetable mutation, only 21% of patients with CRC possess targetable mutations. Even more importantly, this latter proportion likely over-estimates the proportion of targetable mutations in CRC¹¹.

To understand if, and to what extent, patients derive benefit from genomically-matched targeted therapy, it is important to consider the original data behind each guideline. For metastatic CRC, national guidelines recommend considering cetuximab with standard chemotherapy backbones for patients with wild-type KRAS¹². The recent FDA approval of the FoundationOne CDx assay was driven, in part, by its ability to profile KRAS in CRC¹³. The recommendation for KRAS testing initially stemmed from a retrospective sub-group analysis on patients with tumor tissue available for KRAS testing. For the enriched population of patients with KRAS-wild type tumors, the addition of cetuxiumab led to an increase in median progression free survival (PFS) of less than 5 weeks, with no improvement in overall survival (OS)¹⁶. Patients treated with cetuximab, though, did experience an absolute 18% increase in Grade 3/4 adverse events. Letters to the editor in response to the original study described the benefit as “clinically irrelevant”, and the conclusions as “overly sanguine”.^14,15 Years later, the impact of adding cetuximab to first-line therapy for metastatic CRC remains marginal. A 2017 systematic review from the Cochrane CRC Group demonstrated that cetuximab added to standard chemotherapy for first line therapy of metastatic CRC, even when limited to KRAS WT tumors, does not improve OS as compared to standard leucovorin, fluorouracil, and irinotecan (FOLFIRI), although a pooled analyses of multiple trials across different lines of therapy does reach statistical significance¹⁷. Furthermore, cetuximab-containing regimens cost upwards of $30,000 (USD) and increases a patient’s risk of financial toxicity as compared to FOLFIRI (∼ $9,000 USD)¹⁸. Chawla et al.⁹ include KRAS profiling as a benchmark to strive for in metastatic CRC. Perhaps physicians are less likely to perform genomic profiling in this scenario, because the genomically-targeted therapy has marginal benefit and carries risks of clinical and financial toxicity. Indeed, only a minority (37%) of patients with metastatic CRC in this study had tumor molecular profiling performed. Not surprisingly, an even smaller percentage of patients with CRC (6%) received targeted therapy.

Physicians may also be reticent to perform tumor molecular profiling due to a lack of uniform standards for the processing of samples, handling of bioinformatics data, and reporting of variants back to ordering physicians. Only a minority of molecular pathology laboratories assess tumor samples for copy number variants (CNVs) or gene fusions, and many groups do not report variant allele frequencies (VAFs)¹⁹. The VAF limitation, in particular, impacts the ability of physicians to interpret the importance of a specific genetic mutation in the context of intra-tumor heterogeneity. For example, one could understand how targeting a mutation that is present in 15% of tumor cells may be less successful than targeting the BCR-ABL fusion protein in CML^19,20. Additionally, bioinformatics pipelines that fail to flag pathogenic mutations of potential germline origin also represent a missed opportunity to counsel patients and their family members regarding interventions that reduce the risk for a second primary malignancy. These issues highlight a growing need for uniform standards regarding the sequencing of paired tumor/normal controls as well as for bioinformatics pipelines that correctly identify somatic mutations in tumor tissue without subtracting pathogenic germline mutations. On the other hand, more data may compound the difficulties in interpretation by clinicians. Many treating physicians are uncomfortable interpreting molecular pathology reports, and less than 25% felt “very confident” in their ability to make treatment recommendations based on genomic information²¹.

Perhaps the most distressing drawback of current genomic profiling techniques is a lack of concordance between clinical-grade laboratories. The same patient sample, if sent to multiple Clinical Laboratory Improvement Amendments (CLIA)-licensed laboratories for sequencing, will often yield disparate results²². Practically speaking, a patient’s clinical care may differ based only on the molecular pathology group that performed the genomic profiling. This inconsistency may lead to poorly informed clinical decisions, inappropriate clinical trial enrollment, and skewed interpretations of efficacy for targeted therapies being investigated in molecularly-inspired clinical trials²².

Low rates of reimbursement and the administrative burden of recouping the costs of testing might contribute to reduced frequency of tumor molecular profiling. Single institution studies have demonstrated that third party payers reject the majority (55%) of claims submitted for NGS panels containing more than 50 genes²³. This places an undue administrative burden on physicians, especially those with higher volume practices, and may discourage uptake of tumor molecular profiling. It remains to be seen how the recent FDA approval of the FoundationOne CDx assay will alter coverage by third party payers beyond the Centers for Medicare & Medicaid Services (CMS)¹³.

In summary, low clinical adoption of tumor genomic profiling may not be a cause for concern, but an appropriate response to the current limitations of precision oncology. If the promises of precision oncology are to be fulfilled, genomic profiling must be performed in a rational, rigorous manner. Currently, less than 5% of patients with cancer enroll on clinical trials for myriad reasons²⁴, including, but not limited to, delays and failures of private health insurers to cover routine clinical costs²³, institutional barriers²⁵, funding deficiencies²⁴, overly restrictive inclusion criteria²⁶, and a lack of available studies²⁷. Therefore, every effort should be made to improve enrollment to thoughtfully-designed molecularly inspired clinical trials, and to reserve non-research related genomic profiling for indications in which a clear and clinically meaningful improvement in survival has been demonstrated in randomized trials.

Transparency

Declaration of funding

Supported by NIH K12CA139160 (R.F.S), AACR # 16-40-15-SWEIS (R.F.S.), 2017 ASCO-Conquer Cancer Foundation - Young Investigator Award (R.F.S). MWD is funded by a Physician Scientist Training Award from the Damon Runyon Cancer Research Foundation.

Declaration of financial/other relationships

The authors have no other relationships to disclose. JME peer reviewers on this manuscript have no relevant financial or other relationships to disclose.

Acknowledgements

None reported.