Abstract

In the past five years, technological advances in the field of precision medicine in oncology have led to the emergence of a number of commercially available services for tumor profiling. There is an increased acceptance of these services that can help in finding treatment options based on molecular alterations as individual case reports have shown outstanding response to targeted therapies. Tumor profiling must seek to augment, rather than replace, existing local testing with the objective of helping more patients to find appropriate treatment options. The opportunity has arrived to integrate these services into the clinical setting, specifically in cases of refractory and rare tumors. The promise of precision medicine is clear but conclusive demonstration of how each of the different services delivers on the promise is not uniformly forthcoming. The selection of which service is most appropriate must be based on demonstrated analytical validation, quality standards, and data on clinical benefit and utility. The aims of this paper are to evaluate three tumor tissue profiling services commercially available and to propose to the physician how molecular profiling information can be used in clinical treatment decisions today.

1. Introduction

Precision medicine in oncology promises to find the right drug for the right patient at the right time ¹. The high expectations and remarkable results of this field have led to the emergence of a number of commercial services that offer tumor profiling in the routine clinical setting. In the past decade, our increased understanding of many disease-specific molecular alterations in different tumor types has been translated into unprecedented improvements in patients’ outcomes ^{2, 3, 4}. The precision medicine approach has been widely accepted into contemporary clinical oncology not only because of the medical and economic sense in finding the right drug for each patient and avoiding those that are not likely to work, but also because of the emergence of many new targeted therapies. In the past, pharmaceutical companies did not limit patient populations based on molecular alterations and used a ‘one size fits all’ approach to drug development. With the advent of targeted therapies, clinical trials run by pharmaceutical companies have now begun to focus on niche subgroups of patients who carry a specific molecular alteration. Most clinics use commercially available companion diagnostic tests to screen for any specific potential alterations and direct patients to the respective treatments. Some institutions have also established molecular screening tools to identify patients suitable for clinical studies. But validation of next generation sequencing for clinical use, for example, requires a huge investment for university and local funding bodies and can only be achieved in few centers. Commercially available tumor profiling services can complement this local tumor testing and in particular help to find the right treatment options for all those patients for whom no clinical trial options could be found.

Also must offer a broader panel of biomarkers and should meet the highest quality standards to ensure a treatment option can be seeked out. It is important that any services use technologies that are appropriate and accurate. Another key element is the predictive evidence that underlies a treatment association; this should always be based on established and published clinical data. Using treatment associations based on preclinical- or animal models or on scientific assumptions carries a high risk of failure ⁵. This kind of association can only be used as the basis for highly experimental treatments for which a good clinical rationale has failed.

All physicians who treat cancer patients in whom standard of care options have failed will find themselves with a therapeutic dilemma where the next treatment decision will not have sufficient scientific evidence to be included in standard of care guidelines. This point will often be reached earlier in patients suffering with rare cancers, where standard treatment options will be limited and many combinations have been tried without generating sufficient evidence to be included in the treatment guidelines ⁶. In treating such patients, oncologists must base their decisions on an empirical approach weighing the possible risks against the potential benefits of using off-label therapies if clinical trials are not an option ^{7, 8}. In spite of the lack of evidence supporting off-label drug use, and with the expectation that the response rates decrease with each successive line of cancer treatment, there remains a relevant margin of therapeutic benefit when compared against palliative treatments given after all standard therapies have been exhausted ^{9, 10, 11, 12}. The selection of appropriate treatments that can improve the survival remains an unmet clinical need and precision medicine is the best way to achieve clinical benefits in these patients.

Precision medicine approaches promise to direct each individual patient to the right drug, but it is critical that the drugs are also accessible, thus allowing redistribution of available chemotherapy resources to those patients who are more likely to benefit. The ability of a precision medicine approach to deliver on this promise must be measured by clinical utility, showing how often the information provided to the physician impacts on a treatment decision.

The field of precision medicine in oncology has developed at a rapid pace since the beginning of this decade. This success has been possible thanks to technological advances, meaning that new diagnostic technologies are more readily accessible to physicians, but also advances in bioinformatics aiding in translating biomarker results to actionable associations. However, the rapid emergence of a number of commercially available services, which deliver tumor profiling services, along with the prevalence of molecular testing at local academic institutes, means that the field is becoming rapidly crowded with different options. There is a risk that inadequate regulation in this field will lead to confusion and frustration of physicians and in disappointment of cancer patients who are being sold on the idea of precision medicine but find that the particular offering they used cannot deliver on the promise. A precision medicine approach makes rational sense in oncology and it is important to carefully evaluate commercial platforms to ensure that the services used provide sufficient evidence of quality, utility and clinical benefit. Only if all requirements are met the promises of precision medicine can be fulfilled and the patients protected from exploitation or research purposes in routine clinical practice. Here we provide a fact-based evaluation of published evidence supporting the three commercial available services focused on tumor tissue profiling.

2. Materials and Methods

An expert panel of multidisciplinary cancer specialists analyzed and compared available tumor profiling platforms in Spain: Caris Molecular Intelligence^®, FoundationOne^® and OncoDEEP^®. An extensive literature review and collection of information from websites of each platform was completed and deeply examined and discussed in a review meeting. As a result one draft consensus document was prepared that was submitted to the review of two external reviewers. No formal systematic literature search was performed.

3. Results

Many clinics already use a local lab for the assessments of individual biomarkers or small panels. The use of commercial services may bring a number of complementary advantages over local testing, including

• New technologies can be incorporated more quickly once sufficient utility and validation have been demonstrated because of the higher case numbers.

• Global standardization and quality assurance from independent quality organizations, in particular College of American Pathologists (CAP), Clinical Laboratory Improvements Amendment (CLIA), New York State Department of Health (NYS-DOH), consistency of results can be monitored.

• Centralization and high-throughput of analyses allow for quick building of clinical evidence databases with internal comparability.

The analytical comparison of the tumor profiling service is based on three aspects of the offerings:

1.1. Which technologies are used as part of each service and how have they been validated?

1.2. What evidence is used to drive the predictive association between biomarker results and treatments?

1.3. What are the demonstrated quality standards for each service?

A number of technologies are currently used in molecular profiling. These include immunohistochemistry (IHC), Sanger sequencing, next generation sequencing (NGS), fluorescent or chromogenic in situ hybridization (FISH and CISH), and polymerase chain reactions. Most local testing today comprise of NGS using commercially available gene panels to look at hotspots in approximately 50-70 genes, and immunohistochemistry or in situ hybridization testing using companion diagnostic kits or validated monoclonal antibodies established in the treatment guidelines. Hotspot panels are useful to identify pre-specified mutations occurring in limited areas of genes of interest. They focus on a narrow subset and tend not to analyze the entire coding regions or detect all known alterations or classes of alterations. Deeper analyses using full exon coverage could help to complement testing already performed to find further treatment opportunities for patients. Other newer technologies are often available locally but only have application in the research setting and do not have application in the clinical setting due to lack of validation at the present time. The usefulness of most available local NGS panels is for the selection of patients for current clinical trials together with the well defined molecular aberrations with regulatory approved targeted therapies.

Commonly available commercial profiling services include IHC (looking at protein expression), CISH/FISH (looking at amplification, gene copy number variations or gene fusion), NGS (looking for point mutations, insertions/deletions, gene coy number variations) and pyrosequencing (looking for epigenetic changes to the DNA such as promoter methylation). All commercially available services should be analytically validated. A summary of tissue requirements, technologies used and other technical aspects of the three commercially available services has been collated from the respective company websites and included in this review are presented in Table 1. Information was taken from the respective company websites accessed on 05 October 2016 ^{13, 14, 15}.

Caris Molecular Intelligence^® is based on a multiplatform approach that seeks to use all clinically relevant information to find treatment options. Testing currently includes 12-15 IHC tests, 592 genes assessed with full exon coverage by NGS (including copy number variations in 442 genes and an assessment of total mutational load), 10 genes with RNA sequencing to find gene fusions, and other tests such as CISH, pyrosequencing and microsatellite instability performed in certain indications.

FoundationOne^® is based on a single platform approach based on NGS alone. The development and validation of the approach has been published ¹⁶. Testing currently includes a panel of 315 genes assessed with full exon coverage plus introns from 28 genes often rearranged or altered in cancer (including an assessment of total mutational load).

OncoDEEP^® is based on a multiplatform approach that seeks to use all clinically relevant information to find treatment options. Testing currently includes 5-8 IHC tests (including 2-5 tests of protein phosphorylation), 150 genes assessed using a hotspot panel by NGS, gene translocations in 32 genes, and copy number variations in 53 genes and other tests such as pyrosequencing and microsatellite instability performed in certain indications.

Tumor profiling can provide insights into the tumor’s fundamental oncogenic mechanisms based on biomarker abnormalities in well-understood cancer signaling pathways. This information helps to provide a more complete picture of an individual patient’s tumor and supports clinical judgment in the selection of the best treatment options for an individual patient. Published literature that demonstrates a predictive association between a biomarker and treatment outcome strongly supports the use of biomarker information in a treatment decision. Evidence can be graded according to the United States Preventative Services Taskforce guidelines, which takes both study design and study validity into consideration ²⁰. Typically, the data for associations based on protein evidence associated with chemotherapies has been graded as Level 1 or Level 2 and have been demonstrated retrospectively. Many of the data linking signaling pathway alterations to targeted therapy response or resistance are based on individual case reports at most, or more likely, based on mechanistic associations. These may be sufficient to use as the basis for experimental treatments but should be used with caution as the likelihood of successful outcome is unknown. A list of the treatment associations available in each commercial platform is provided in Table 2.

Caris Molecular Intelligence^®: 96% of associations which appear on the front page of the Caris report are based on Level 1 or Level 2 evidence. Extrapolations based on level 3 clinical evidence or preclinical data are presented as experimental opportunities and provide the basis for the Clinical Trial Connector. The Caris report provides associations for a comprehensive range of chemotherapies, hormone therapies, targeted therapies, immunotherapies and investigational agents.

FoundationOne^®: FoundationOne^® is a comprehensive and fully informative genomic profile that can reveal all classes of actionable alterations, including those in cancer-driving genes that are rarely or never tested for in solid tumors. The FoundationOne^® report often reveals alterations that may lead to additional treatment options for physicians and their patients to consider. The testing is focused on providing associations to targeted therapies and investigational agents. Predictive associations with platinum agents (based on BRCA1/2 mutations) and immunotherapies (through tumor mutational load) are also provided. The level of clinical evidence underlying the proposed associations is not clear and the power of associations is not ranked.

OncoDEEP^®: The OncoDEEP^® report is focused on targeted therapies with these associations being driven by the results of NGS and protein phosphorylation assays. Predictive associations on selected chemotherapies, hormone therapy, immunotherapies and investigational drugs are also provided. The level of clinical evidence underlying the proposed associations is not clear and the power of associations is not ranked.

It is important that both in vitro diagnostics (IVDs) and laboratory developed tests (LDTs) are performed to high analytical standards so that health care providers do not seek unnecessary treatments, delay needed treatments, or expose patients to inappropriate therapies. External quality assessment of both local laboratory testing and the commercially available services by independent bodies is required in order to ensure that all testing can be relied upon. In order for a service to be marketed in the US, external certification by the Clinical Laboratory Improvements Amendments (CLIA) and the College of American Pathologists (CAP) is required. Further, stricter validation is needed from the New York State Department of Health (NYSDOH) to test samples from New York. Together, these three provide quality standards for internal processes as well as for individual tests for biomarkers. Internationally, ISO15189 accreditation can be obtained to achieve quality control in internal processes and is compulsory for all medical laboratories operating in Belgium. A review of adherence to guidelines over the first five years of implementation of ISO15189 in 17 BELAC-accredited laboratories in Belgium found 421 non-conformities, including 19 major findings ²¹. None of the commercially available services included in this review has been cleared or approved by the United States Food and Drug Administration (FDA). The FDA has determined that such clearance or approval is not necessary, although it is likely that regulation will follow in 2017. A summary of certifications and accreditations by each laboratory providing services described in this review are provided in Table 3.

Caris Molecular Intelligence^®: The Caris laboratory is CLIA, CAP and NYSDOH certified/approved for all testing performed. The Caris Laboratory also holds state licensures from California, Florida, Maryland, Pennsylvania and Rhode Island. In addition, Caris is the only biolaboratory that has received the European ISO15189: 2012 certification in the United States. The service also holds a CE mark. Caris routinely uses commercially available in vitro diagnostic (IVD) kits to perform testing as part of the CMI service. If no kit is available, Caris develops a laboratory-developed test (LDT) to perform the testing.

FoundationOne^®: The Foundation medicine laboratory is CLIA, CAP and NYSDOH certified/approved for all testing performed, and also holds a CE mark.

OncoDEEP^®: All testing conducted as part of OncoDEEP is performed by the Institute of Pathology and Genetics (IPG) Gosselies. This laboratory holds ISO15189 accreditation, including accreditation for the Ion Torrent genetic sequencing performed as part of OncoDEEP®. All other aspects of the testing platform are not mentioned in the ISO15189 accreditation ²².

Evidence-based guidelines on the use of genetic tests in clinical practice require a systematic assessment of their usefulness, which, following a commonly used framework proposed in1998 by a U.S. Task Force on Genetic Testing is commonly referred to as clinical utility ²³. Clinical utility can broadly refer to any use of test results to inform clinical decision-making. A summary of the impact of the information provided by the various services under review and the number of profiled patients based on the report is presented in Table 4 and detailed below.

Caris Molecular Intelligence^®: In CMI evaluations where the planned treatment option was recorded prior to receipt of the report (n=137), a comparison to the actual treatment given could be performed. The treatment decision was changed upon receipt of the report in 120 of 137 (88%) cases ^{25, 26, 27, 28, 32}. Three hundred and forty eight of 473 patients (74%) profiled as part of CMI evaluations in routine clinical practice globally were treated in line with the report ^{25, 26, 27, 28, 29, 30, 31, 32, 33, 34}. This figure is lowest in evaluations of gastric cancer (28 of 46 patients; 61%) and pancreatic cancer (30 of 55 patients; 55%). This reflects the aggressive nature of the disease and rapid progression in many cases rendered the patients no longer fit to receive treatment after profiling. The figure is also impacted by the interim nature of results from a prospective study in Austria, which is still enrolling and where many patients still remain untreated, as they had not yet progressed on prior therapy at the time of the snapshot.

FoundationOne^®: A study of decision impact using FoundationOne^® demonstrated that the treatment decision was switched in 36 of 132 cases (27%) ³⁸. In studies using FoundationOne® at key academic centres in the United States and Brazil, only 285 of 1,174 profiled patients (24%) received a sequencing guided treatment ^{38, 39, 40, 41, 42, 43}.

OncoDEEP^®: There is no published record of clinical utility for the OncoDEEP® service.

Another aspect of clinical utility to be considered is the impact of the profiling-guided treatment decisions on the patient outcomes. Measures of benefit include progression-free survival ratio, response or disease stabilization (SD≥6 months/PR/CR), time to next treatment, and overall survival. A summary of the demonstrated clinical benefit is presented in Table 5 and detailed below.

Caris Molecular Intelligence^®: Caris has published 12 case reports that show the impact of successful treatment selection in patients across tumor types. In addition, Caris has conducted a number of physician-led prospective and retrospective evaluations of the service ^{25, 26, 27, 28, 29, 30, 31, 32, 33, 34}. A recent independent review of the clinical data for Caris Molecular Intelligence® has been published by key opinion leaders from the UK, Belgium and Italy ²⁴. Of the 322 evaluable patients in these evaluations, 131 (40%) had clinical benefit as defined by the treating physician. An ongoing prospective observational study is being run which is following over 3,000 profiled patients for overall survival ³⁵. The study, which started in 2009, presented a first snapshot based on 1180 patients with different types of solid tumors. Patients were divided into two cohorts, based on whether they followed the report or not. The ‘Matched’ cohort consisted of 534 patients, while the ‘Unmatched’ cohort had 493 patients. Both groups were balanced in terms of participant’s demographics and tumor types involved. As shown in Figure 1, patients in the Matched cohort had a significantly longer overall survival than patients in the unmatched cohort, with a median increase in OS of 422 days (HR=0.68, p=0,001). According to the data from this study, patients in the Matched cohort received less lines of therapy than patients in the Unmatched cohort (median treatment lines after receiving the Caris report of 3.2 versus 4.2 respectively).A paper describing the 9-month median overall survival benefit observed in ovarian cancer patients treated in line with the report was recently published by Herzog and colleagues ³⁶.

In early 2015, Caris established a Centers of Excellence (COE) Network to track and integrate longitudinal clinical outcomes with comprehensive molecular data. Treatment data has been retrospectively collected from the electronic medical records of over 4,700 patients across all solid tumor types who received Caris Molecular Intelligence^® profiling as part of their clinical care ³⁷.

Time to next therapy (TNT) was defined as the length of time between the start date of the first treatment received after sample collection and the start date of the second treatment received after sample collection. Given the lack of reliable information about progression in electronic medical records, TNT is used as a surrogate for progression-free survival in this cohort of patients. The robustness of the treatment data captured in the electronic medical records allows for a complete and accurate measurement of TNT.

To assess the influence of Caris Molecular Intelligence^® on TNT, two patient cohorts were defined based upon the first treatment administered after the time of specimen collection. The Matched cohort (n=3,011) received only drugs associated with benefit in the CMI report. The Unmatched cohort received at least one drug, which was associated with lack of benefit in the CMI report (n=1,718). TNT is censored after 1 year. Patients in the matched cohorts had a significantly longer time to next therapy (TNT) than patients in the unmatched cohort (median TNT 248 days vs. 215 days, HR=0.85; p=0.00018).

FoundationOne^®: Foundation Medicine has actively published case reports showing outstanding clinical benefit in patients with different solid tumor who received guided therapy, 46 individual case reports are currently listed on the company’s website. Of the 260 patients treated in a FoundationOne® study, 69 (27%) had clinical benefit ^{38, 39, 40, 41, 42, 43}. The strongest data to date comes from a single center study conducted at MD Anderson in Texas ⁴¹. Of the 339 successfully profiled patients, 317 (93.5%) had at least one potentially actionable alteration. Over a third of patients (36.5%; 122 of 339) received a treatment guided by the profiling results. Of the evaluable 118 patients who received matched therapy, 45 were directly matched while 73 were indirectly matched to the targeted treatment in question. A further 66 patients received an unmatched therapy. Twenty-three (19%) patients in the matched treatment arm had clinical benefit (SD≥6 months/PR/CR) compared to only 5 of the 64 patients (8%) in the unmatched cohort. The median time to treatment failure was significantly longer (2.8 months versus 1.9 months, p=0.001) and a trend towards longer overall survival was observed (9.3 months versus 7.2 months). Interestingly, the clinical benefit does not appear to translate directly to an OS benefit. In the directly matched subgroup of the matched therapy cohort, 6 of 45 patients (13%) had clinical benefit with a median overall survival of 11.3 months. By comparison, most of the patients with clinical benefit were in the indirect matched subgroup (23%, 17 of 73) but had a shorter median overall survival of 7.3 months.

OncoDEEP^®: There is no published evidence of clinical benefit for the OncoDEEP® service.

4. Conclusions

Our review of different molecular profiling services agrees that this new approach has high potential clinical value and, if used wisely can lead to remarkable benefits for patients. One key question to consider is the identification of which patients would need and benefit most of molecular profiling. There are currently several settings in which molecular tests are well justified: First, for patients with diseases for which novel targeted agents are approved. It is important to demonstrate the presence of such targets to give them access to the respective drug. Examples are ALK and EGFR genomic analysis in NSCLC, for which tests are usually performed at local institutions because validated test kits are commercially available.

A second indication is the screening of patients that may be candidate to participate in certain clinical studies of novel agents that need to be tested on patients with tumors holding an specific molecular profile Today, specialized clinics achieve to include a large number of patients into clinical studies; this is only possible when they perform molecular profiling that screens for mutations that can justify inclusion into the respective drug study. Usually broader genomic profiling is performed for this purpose and can be implemented by specialized local laboratory or facilitated by a profiling service like the ones we reviewed.

The third clinical use is for patients for whom the testing of standard biomarkers did not yield an actionable result and participation in a clinical study would not be the first choice. After the standards of care failed, a broad search can be started to find drugs that are potentially active for their disease. These can be drugs that would usually not be considered and the search should be performed across all classes of drugs, ie. includes conventional cytotoxic chemotherapies, hormone therapies, and targeted therapies.

In this task, our review support that Caris CMI offers the most complete multiplatform profile with which favorable clinical outcome data were produced. These clinical outcome data clearly show that with the right approach to clinical use, the broader practice is well justified as the clinical benefit has been demonstrated. This review has the limitation that the information about the use of platforms for molecular profiling is continuously updated, and most of this new information is not formally published and indexed. The analyzed commercial platforms were those available in Spain, but are the most widely spread too. We consider that the general scheme of analysis described can be applied for the evaluation of other platforms.

From the patient perspective pre-test medical information must be clear, transparent and honest, as not all the patients tested will get a clear benefit. Limitations and barriers in the post-test access to the therapy should also be considered, with special mention to a couple of specific and common situations: the rapid deterioration of the patient if the test is ordered in a too-late stage of the disease, and barriers in the access and reimbursement of non-licensed drugs for the specific patient’s disease, specially in target therapies.

We believe that it is time for multiplatform profiling to be introduced into general clinical practice. For this it is important to offer them at institutions where experienced oncologists know how to best interpret the data from a molecular profile, where there is access to experts who can help characterize unexpected findings, and where a multidisciplinary exchange is part of the common clinical practice. Currently available services as those reviewed here should complement and augment existing research institute capabilities and could be used for select patients who may benefit. The institutional approval and reimbursement of multiplatform profiling must be considered in certain well defined clinical settings such as for instance rare tumors, carcinoma of unknown primary, triple negative breast cancer or third line colorectal cancer. Only ‘proven’ services should be reimbursed, they should meet most stringent quality criteria and they should have produced real life data on the benefit they can achieve.

Introduction of molecular profiling into clinical practice must also consider that patients need to be informed adequately by the treating oncologists in order to set the expectations correctly.

Acknowledgements

The authors acknowlegde to Laboratorios Leti (Madrid, Spain) the support for the organization of the experts meeting.

References