On the Road to Precision Cancer Medicine: Analysis of Genomic Biomarker Actionability in 439 Patients

The strategy of matching targeted drugs to biologically relevant targets using molecular profiling techniques is becoming better established, though many challenges remain (1–3). Indeed, the presence of genomic aberrations in tumors may be critical to achieving response, especially when using agents with specific molecular targets. For instance, KIT kinase inhibitors are effective in patients with KIT mutations (4, 5). Similarly, preclinical and emerging clinical data suggest that PI3K inhibitors may be most effective in patients with PI3K or PTEN aberrations (6, 7); RAF inhibitors, in patients with RAF mutations (8); MEK inhibitors, in patients with RAS or RAF mutations (9, 10), and so on. In some cases, matching patients with targeted therapies has resulted in transformative changes. For instance, treatment of chronic myelogenous leukemia, a disease driven by an aberrant BCR-ABL kinase, with imatinib (a BCR-ABL kinase inhibitor) has dramatically increased median survival (11–13). In the phase I setting, molecular matching was associated with improved outcomes in multivariate analysis (14). Further, a systematic review of phase II clinical trials in advanced/metastatic non–small cell lung cancer showed that molecular matching of patients' tumors to drugs was independently associated with better outcomes, including higher median response rate (48.8% vs. 9.7%; P = 0.005), longer median progression-free survival (6 vs. 2.8 months; P = 0.005), and overall survival (11.3 vs. 7.5 months; P = 0.05), as compared with those of unselected patients (15).

The percentage of patients with cancer who have potentially “actionable” aberrations remains a matter of debate. In order to determine the proportion of individuals who have druggable alterations, we analyzed the molecular profile results of patients who had had a targeted next-generation sequencing panel performed on their tumor.

We retrospectively reviewed the medical charts of 439 patients with diverse cancers, for whom molecular testing had been performed, and who were seen at the UCSD Moores Cancer Center (La Jolla, CA) from October 2012 until July 2014. This study was performed and consents were obtained in accordance with UCSD Institutional Review Board guidelines.

Next-generation sequencing was performed by Foundation Medicine (FoundationOne, http://www.foundationone.com), which is a clinical-grade CLIA-approved next-generation sequencing test that sequences the entire coding sequence of 236 cancer-related genes and 47 introns from 19 genes often rearranged in cancer [9 patients (2.3%) were tested with prior version of the panel comprising 182 cancer–related genes]. Amplification was noted when there was ≥8-fold change in copy number.

An actionable alteration was defined as an alteration that was either the direct target or a pathway component that could be targeted by at least one approved or investigational drug. For consistency purposes, the interpretations provided on the next-generation sequencing reports were utilized as a basis for actionability determination.

Demographic information such as gender, age at diagnosis, race, and clinical information such as cancer histology, presence of metastasis at diagnosis, and presence of metastatic disease at the time of the biopsy used for molecular testing were extracted from patients' electronic medical charts. The biopsy site used for molecular testing was also recorded. From the molecular testing reports, the following information were extracted: number of total alterations, number of actionable alterations, and, more specifically, the number of alterations with an approved drug available in the disease (on-label use), the number of alterations with an approved drug in another disease (off-label use), and the number of alterations with experimental drug(s) available (clinical trials).

Most of the analysis was descriptive in nature. When appropriate, linear or binary logistic regression analyses were performed; coefficients and 95% confidence intervals (95% CI) were reported. Spearman's rho coefficients were computed to assess the correlation between two continuous variables. The sample size was determined by the available patients with genetic testing information. The medians and 95% CI were reported for continuous variables. Demographic and genetic characteristics were compared between groups. All statistical analyses were performed by MS with SPSS version 22.0.

Our population comprised 439 patients who were seen at the cancer center and had molecular testing performed. Patient characteristics are listed in Table 1. The median age at diagnosis was 54 years (95% CI, 52.5–55.8 years); 57% were women. The majority of our patient population was Caucasian (73%), followed by other (13%) and Asian (6%). The most common primary tumor sites were gastrointestinal (n = 110, 25%), followed by breast (n = 83, 19%) and brain tumors (n = 62, 14%). Seventy patients (n = 70, 16%) had metastatic disease at the time of diagnosis, and 257 patients (58.5%) had metastatic disease at the time of the biopsy used for molecular testing. Biopsy used for molecular testing mostly originated from the primary tumor (n = 255, 58%) versus a metastatic site.

Overall, 1,813 alterations (found in 207 distinct genes) were identified (Supplementary Table S1). The three most commonly altered genes were TP53 (44%), followed by KRAS (16%) and PIK3CA (12%; Fig. 1A). Most of the alterations identified were mutations (62%) or amplifications (29%; Fig. 1B).

Four hundred and twenty patients had at least one molecular alteration (96%). The median number of alterations per patient was 3 (range, 0–16). Nineteen patients (4%) had no alterations and 48 patients (11%) had only one alteration. Interestingly, the majority of our patients had different molecular profiles. Indeed, only 7 patients (1.6%) had precisely the same molecular profile, when looking at both the gene and the location/type of alteration. If only the gene was taken into consideration (and not the location/type of alteration), 49 patients (11.2%) had the same profile.

In total, 393 patients of 439 (90%) had at least one potentially actionable alteration identified, for which at least one or several experimental drugs in clinical trial were usable (Fig. 2). Among the 10% of patients who had no actionable alterations, 4% had no reportable genetic alteration found, and 6% had one or more alterations, but none actionable. The median number of potentially actionable alteration per patient was 2 (range, 0–8; Fig. 3A). In considering the drug options for actionable aberrations, the total number of patients who had an aberration targetable by an approved drug was 307 [70%; 296 patients (67%) had at least one drug that was approved for another disease (off-label use), and 89 patients (20%) had at least one or more approved agents in their disease available (on-label use); Table 2 and Fig. 2]. There was a positive correlation between the number of alterations found and the number of actionable alterations (Spearman's rho coefficient = 0.808, P < 0.0001; Fig. 3B). A multiple linear regression model, including breast cancer versus not (breast cancer being the only histology type that was statistically significantly correlated with a higher number of actionable alterations in univariable analysis), the number of alterations, the origin of biopsy used for the testing (primary vs. metastatic), whether the disease was metastatic or not at the time of biopsy that was used for testing, confirmed that only the number of alterations was an independent predictor of a higher number of actionable alterations (P < 0.0001).

A subanalysis revealed that across all different malignancies tested, there consistently was a majority of patients that had potentially actionable alteration (Table 2 and Fig. 3C). Of note, breast cancer cases had a significantly higher median of molecular alterations (median of 5 alterations vs. 3; P = 0.0002). A multivariable analysis, including both the breast cancer histology and whether or not the disease was already metastatic at the time of biopsy, confirmed that breast cancers were associated with a higher number of alterations (P = 0.003); the presence of metastatic disease at the time of biopsy also correlated with a higher number of alterations (P = 0.024).

A binary logistic regression model showed that there were significantly more patients with breast cancers (P < 0.0001) and melanoma (P = 0.0002) who had approved matched targeted drug options in their disease (on-label use). Patients with breast cancer had more often mTOR (P < 0.0001) and ERBB2 (P = 0.005) approved inhibitor treatments identified, whereas patients with melanoma frequently had RAS/RAF/MEK inhibitors (P < 0.0001) as treatment options.

Table 3 provides the potentially actionable aberrations and examples of the FDA-approved drugs for them. The mTOR inhibitors everolimus and temsirolimus, targeting the PI3K/Akt/mTOR pathway alterations, were the most commonly potentially matched drugs (38%; 151/393 patients with ≥1 actionable alteration), followed by trametinib (26%, 101/393 patients with ≥1 actionable alteration; Table 3).

For malignancies comprising the most patients, we were able to identify the more common doublets of actionable alterations. For patients with gastrointestinal disease, KRAS alterations were found with either APC or PIK3CA alterations in 14 and 6 patients, respectively. For patients with breast cancer, the most frequent actionable co-alterations were PIK3CA with either ERBB2 or PTEN, and CCND1 with ERBB2 (n = 5 patients each doublet). Patients with brain cancers frequently harbored CDKN2A/B loss with either PTEN (n = 7) or EGFR alterations (n = 12).

Herein, we studied the molecular alterations identified in 439 patients with diverse cancers (58.5% of whom had metastatic disease at the time of biopsy) using next-generation sequencing. We found that 96% of our patients demonstrated at least one molecular alteration, and 85% had two or more alterations. The most frequent alteration was TP53 mutation, found in 44% of our patients, similar to previous reports (16).

Ninety percent of patients had an actionable aberration. Similarly, a recent report on a smaller number of patients suggested that 83% (of 103 tested individuals with cancer) had actionable abnormalities (17). The median number of actionable aberrations in our patients was 2 (range, 0–8). In regard to the types of agents that could be used, 296 patients (67%) had abnormalities that could be prosecuted by at least one drug that was approved for another disease (off-label use); 89 patients (20%) had abnormalities that could be prosecuted by at least one approved agent in their disease (on-label use). The total number of patients who had an aberration targetable by an approved drug was 307 (70%). All patients who had a least one potentially actionable alteration also had one or more experimental drugs that targeted the anomaly as possible treatment options. However, previous experience suggests that patient eligibility for these clinical trials or their conduct at a limited number of distant enrolling sites might severely limit patients' access (18).

Actionability by drugs that were approved in the same disease was more common in breast cancer and melanoma. Indeed, ERBB2 and mTOR inhibitors are approved in breast cancer, and BRAF/MEK inhibitors are approved for melanoma. For drugs approved in other disease types (off-label use), logistical problems such as insurance coverage may limit access (18). Beyond logistics, a fundamental question relates to the amount and quality of data needed to qualify a patient for a drug, or inversely to determine that withholding that drug might be detrimental. Because many molecular abnormalities do not appear to segregate well by histology, and because many mutations and amplifications can be found in a subset of patients with almost any cancer (14), it appears unlikely that classical methods of approval that require high-level scientific evidence (usually randomized phase III trials) will be feasible for the subset of patients with a particular molecular anomaly in each histology. On the other hand, new classifications based on molecular diagnosis or the use of genomically driven bucket trials that cross canonical disease boundaries may be able to provide adequate scientific evidence to determinate actionability. As an example, a bucket trial using imatinib (19; reported in 2008) assessed a variety of uncommon disorders that harbored an imatinib target (KIT, PDGFR, Bcr-Abl); the study led to approval for rare disorders bearing these targets, including myeloproliferative/myelodysplastic disorders, dermatofibrosarcoma protuberans, aggressive systemic mastocytosis, and hypereosinophilic syndrome, with only 5 to 14 patients in each subgroup (20).

There were several limitations to our study. First, it included a limited number of patients and comprised different malignancies. However, the latter may suggest that the results are generalizable across malignancies. Second, the definition of “actionable” and the level of evidence needed for such a determination are a matter of debate and in constant evolution. Finally, whether or not the patients would have responded to these drugs was not ascertained in the study. Also, we were able to observe frequent actionable co-alterations, although the number of patients with each disease subgroup limited this analysis; still, this type of data may inform the development of combination therapy protocols.

In conclusion, our observations suggest that the vast majority of patients (90%) have theoretically actionable aberrations. Furthermore, 70% of the patients had an aberration that can be targeted by an approved drug. However, only a minority of individuals (approximately 20%) had aberrations that could be targeted by drugs approved for their type of cancer (on-label). Only 6% of patients had no aberration that was actionable by either an approved or experimental drug in clinical trials. Interestingly, the molecular portfolio of almost all patients was unique, consistent with previously reported results (21). Indeed, only 1.6% of patients harbored identical molecular aberrations. These observations suggest that individualization of therapy is likely to become increasingly important. Challenges such as the expense and/or restriction in use of off-label drugs, as well as the strict eligibility criteria and the distance of sites for clinical trials, however, limit access to matched targeted medications. One possible solution is for pharmaceutical sponsors or government bodies to provide an online registry where patients with a given aberration would be provided free drug as long as baseline key information is documented. The registry could then track outcome parameters or time on drug (as a surrogate for clinical benefit). These challenges, in addition to the important question regarding the level of evidence needed to define actionability, are crucial issues that must be addressed in order to fully deploy precision medicine in patients with cancer. New paradigms for clinical research and drug access are needed at the sponsor and national levels.

R. Kurzrock is founder of RScueRX, has ownership interest (including patents) in RScueRX, and is a consultant/advisory board member for Sequenom. No potential conflicts of interest were disclosed by the other authors.

Conception and design: M. Schwaederle

Development of methodology: M. Schwaederle

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M. Schwaederle, G.A. Daniels, D.E. Piccioni, P.T. Fanta, R.B. Schwab, K.A. Shimabukuro, B.A. Parker

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M. Schwaederle, D.E. Piccioni, P.T. Fanta, B.A. Parker, R. Kurzrock

Writing, review, and/or revision of the manuscript: M. Schwaederle, G.A. Daniels, D.E. Piccioni, P.T. Fanta, R.B. Schwab, B.A. Parker, R. Kurzrock

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M. Schwaederle

Other (final approval): R. Kurzrock

R. Kurzrock received funding by the Joan and Irwin Jacobs Fund and My Answer To Cancer philanthropic fund.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.