Comprehensive genomic profiling in routine clinical practice leads to a low rate of benefit from genotype-directed therapy

Carcinogenesis is a multi-step process propelled by genomic alterations that leads to dysregulation of signaling pathways, which consequently gives rise to qualities that enable tumor proliferation and dissemination [ 1]. Our understanding of the molecular processes underlying malignancies has translated to targeted therapies, which have transformed the clinical management of some cancers. Indeed, the landscape of systemic therapy in certain malignancies is evolving from its dependence on nonselective cytotoxic therapies to one that includes the utilization of selective inhibitors [ 2]. Since the first breakthrough in molecular targeted therapy with imatinib, whose action against BCR-ABL kinase produced robust responses in chronic myeloid leukemia (CML) [ 3], various oncogenic drivers of the proliferative phenotype have been uncovered and translated into therapies. The use of trastuzumab, a humanized monoclonal antibody against human epidermal growth factor receptor 2 (HER2) is standard of care for HER2 overexpressed breast cancers [ 4] and more recently approved for HER2 overexpressed gastric cancer, which represents the first targeted therapy in this malignancy [ 5]. While these well-studied and validated alterations are routinely targeted in clinical practice, patients with other alterations in frequently mutated pathways may also benefit from targeted therapies.

Today, comprehensive genomic profiling (CGP) of tumors can provide insight into clinically relevant genetic alterations (CRGAs), with goals of guiding clinical decision-making and augmenting therapeutic options. The accessibility of this technology facilitates the shift towards precision medicine by identifying specific patient populations that are most likely to derive benefit from a particular therapy. However, the proportion of patients that end up receiving genotype-directed therapy (GDT) represents the minority of total patients profiled, despite a large proportion having actionable genetic alterations [ 6]. In addition, only a small fraction of patients profiled in such studies derive benefit from the treatment [ 7]. The reasons for the marked divergence in patients profiled to have actionable mutations and those that ultimately receive GDT have not been systematically studied.

Herein we present a single institution study of the outcomes of CGP in the clinical management for these patients, with focus on exploring the reasons for which patients with actionable mutations did not receive GDT.

The concept of individualizing a patient’s treatment based on a specific alteration in their tumor is attractive to oncologists and patients alike. This approach is already validated in a number of malignancies and there are now many FDA-approved therapies that require testing for a predictive biomarker [ 10, 11]. In these cases, where there is a validated drug-biomarker pair, the outcome with biomarker-guided therapy is usually very rewarding, leading to improvement in survival. The success of this approach in some situations has generated tremendous enthusiasm for splitting tumors into groups based on driver alterations and treating based on these results. Testing for driver genomic alterations has become routine clinical practice with the proliferation of a number of commercially available platforms for analyzing patients’ tumor specimens in an attempt to uncover vulnerabilities in the tumor that may be susceptible to new targeted therapies. However, based on the evidence we present here, as well as many other studies, it appears that the widespread uptake of genomic profiling is ahead of the evidence supporting its benefit.

The rate of benefit from GDT in our study, defined as stable disease or a partial response to therapy was 5% in patients with stage IV malignancies. This low proportion is in the range of what has been reported in other series. Table  6 shows a summary of similar real-world single-center experience with GDT in routine clinical practice. These studies all showed a high rate of potentially actionable alterations but a low rate of patients receiving matched therapy and an even lower rate of patients who derive benefit from this process [ 7, 12– 17]. If one uses all patients entering the testing algorithm who are hoping for a “home-run” targeted therapy as the denominator, the proportion of patients who do achieve some disease control from therapy is between 2 to 8% across multiple studies.

Apart from the institutional series described above, the prospective SHIVA trial is a study that is often cited as an example of the limitations of GDT [ 18]. This study randomized patients who had one of a defined set of alterations to received genomically matched targeted therapy versus investigator’s choice chemotherapy. There was no difference in progression-free survival in the patients who received matched therapy versus those who received standard chemotherapy. Despite this apparent negative result, this study was by no means definitive and the main limitation was that the targeted drugs used were not validated against the purported targets and in a heavily pretreated population; the weak drug-target pairs were unlikely to succeed [ 19].

Contrast this study to the study of erlotinib versus chemotherapy for the first-line treatment of patients with lung cancer harboring an EGFR activating mutation. This study showed remarkable superiority of erlotinib in this setting with a hazard ratio for survival of 0.16 (95% CI 0.1 to 0.26) [ 20]. This clearly illustrates the principle that a validated drug-biomarker pair can lead to excellent outcomes. It also highlights the inherent complexity of tumors and the wide variety of genomic alterations that introduce bias in treatment selection. Many cancers do not have clear driver mutations, and targeting a random genomic alteration will have no effect of the natural history of the disease.

More recently, the ProfiLER trial, a multi-institutional prospective study from France, has reported data on 1826 patients. Approximately half (51%) had at least one actionable mutation, and 35% were deemed to be eligible for matched therapy when utilizing a molecular tumor board. Among those, 6% initiated a recommended matched therapy. The rate of benefit, which included complete response, partial response, and stable disease, was 2.4% of the total population (44% of those who initiated therapy) [ 21].

Our study has a number of important limitations. This was not a prospective study and the implementation of GDT was physician-dependent and not standardized. Many institutions now have molecular tumor boards that systematically review profiling results and attempt to match patients to clinical trials or the best available treatment based on the results. Even in a supervised setting like this, the rate of matched therapy is still low. The benefit derived from matched therapy, however, may be higher as shown in one single-center experience [ 14]. This is likely due to a more scrutinized approach in interpreting the genomic data, and selecting the appropriate patients whose tumors have shown a more indolent biology (e.g. most patients in the ProfiLER study who initiated GDT had gynecologic, colorectal and breast malignancies; many known to behave in indolent manner).

Another weakness of our study was the inconsistency in the timing of sending CGP. In some cases, testing was requested early in the disease course and the results were available while patients were still on standard-of-care therapy or in remission. Presumably the testing results could be retrieved and acted upon if the patients developed disease progression. However, the molecular alterations at the time of progression may evolve and the results of prior testing may no longer be reliable [ 22]. On the other hand, many patients had testing done very late in their disease course and did not have the time to have the results implemented. In general, we found that patients who had rapid disease progression or poor performance status due to their malignancy (28% of our cohort) did not have the opportunity to benefit from the results of GDT and testing these patients did not alter the outcome.

Lack of easy access to phase 1 clinical trials was also a major barrier to GDT in our study – our center had few available phase 1 trials at the time of this study. This led to many patients receiving off-label therapy that was felt to be efficacious at the discretion of the treating oncologist. Community oncology practices alike generally have limited access to phase 1 therapeutic trials making our experience reflective of the wider oncology practice. However, the use of “off-label” GDT should not be the norm since the activity of a drug targeting a particular mutation is different for distinct tumor types (e.g. BRAF inhibitors are very efficacious in BRAFV600E mutated melanoma, but lack the same activity in BRAFV600E mutated colon cancer).

So how do we move forward? Based on the evidence we provide in our study and the other studies cited here, a haphazard approach to CGP is very unlikely to help patients. Tumor profiling should be done in a deliberate way with systematic analysis of the results to guide patients into appropriate clinical trials, avoiding those who are unlikely to benefit, such as patients with rapid disease progression or those in whom standard therapy is being successfully applied. This means that the optimal time for requesting CGP may differ with distinct histologically defined tumor types. The aspect of biologic complexity of tumors and intratumor heterogeneity should also be considered, as these patients are unlikely to benefit from a targeted approach unless the target is a driver alteration. Moreover, the financial burden of CGP on both the healthcare economy and individual patients should be kept in mind. In one study of 209 patients in a community practice, the total cost was reported at $1.21 million, and 17% of patients were responsible for the full cost after exhausting financial coverage and support options [ 23].

The ongoing NCI-MATCH trial is an example of a basket trial that is meticulously approaching this problem. The drug-biomarker pairs have passed a minimum bar of validation and this study will provide further evidence as to whether some of the pairs will in fact benefit patients. The ASCO-TAPUR trial is another example that utilizes a molecular tumor board composed on a group of experts convened by ASCO that provide an informed decision regarding the proposed treatment. On the other hand, SPECTAcolor (Screening Patients for Efficient Clinical Trial Access in advanced colorectal cancer) is an example of a large European collaboration that aims to screen patients with colorectal cancer to improve access to molecularly defined clinical trials [ 24].

In this study, the routine use of CGP in clinical practice was associated with minimal benefit to patients in terms of disease control. Furthermore, we were able to identify various barriers to implementation of GDT that should be taken into consideration before requesting CGP on tumor specimens. Until the results of prospective trials are reported, it will be difficult to curb the runaway train that is the routine clinical use of CGP. However, given the available data we recommend against routine CGP outside the context of a prospective precision medicine program or well-designed clinical trials.