Initiative for Molecular Profiling and Advanced Cancer Therapy (IMPACT): An MD Anderson Precision Medicine Study

Abstract

Purpose

Genomic profiling is increasingly used in the management of cancer. We have previously reported preliminary results of our precision medicine program. Here, we present response and survival outcomes for 637 additional patients who were referred for phase I trials and were treated with matched targeted therapy (MTT) when available.

Patients and Methods

Patients with advanced cancer who underwent tumor genomic analyses were treated with MTT when available.

Results

Overall, 1,179 (82.1%) of 1,436 patients had one or more alterations (median age, 59.7 years; men, 41.2%); 637 had one or more actionable aberrations and were treated with MTT (n = 390) or non-MTT (n = 247). Patients who were treated with MTT had higher rates of complete and partial response (11% v 5%; P = .0099), longer failure-free survival (FFS; 3.4 v 2.9 months; P = .0015), and longer overall survival (OS; 8.4 v 7.3 months; P = .041) than did unmatched patients. Two-month landmark analyses showed that, for MTT patients, FFS for responders versus nonresponders was 7.6 versus 4.3 months (P < .001) and OS was 23.4 versus 8.5 months (P < .001), whereas for non-MTT patients (responders v nonresponders), FFS was 6.6 versus 4.1 months (P = .001) and OS was 15.2 versus 7.5 months (P = .43). Patients with phosphatidylinositol 3-kinase (PI3K) and mitogen-activated protein kinase pathway alterations matched to PI3K/Akt/mammalian target of rapamycin axis inhibitors alone demonstrated outcomes comparable to unmatched patients.

Conclusion

Our results support the use of genomic matching. Subset analyses indicate that matching patients who harbor a PI3K and mitogen-activated protein kinase pathway alteration to only a PI3K pathway inhibitor does not improve outcome. We have initiated IMPACT2, a randomized trial to compare treatment with and without genomic selection.

INTRODUCTION

Genomic profiling is increasingly used in the management of patients with cancer. In 2007, we initiated the IMPACT (Initiative for Molecular Profiling and Advanced Cancer Therapy) trial, a personalized medicine program for patients who were referred to the phase I clinical trials program at MD Anderson Cancer Center. Personalized, or precision, medicine uses traditional and emerging concepts with regard to the genetic and environmental basis of disease to individualize prevention, diagnosis, and treatment,^1,2 and integrates tumor genetics of individual patients into medical practice.³ The aim of the personalized medicine program is to use tumor molecular profiling to optimize the selection of targeted therapies for patients who are considered for phase I clinical trial participation. The rationale for designing the precision medicine program was based on three factors: the emergence of technologies that enable the identification of genomic alterations in patient tumors; the dramatic improvement in survival that has resulted from the use of imatinib to treat newly diagnosed chronic myeloid leukemia⁴; and the BATTLE (Biomarker Integrated Approaches of Targeted Therapy for Lung Cancer Elimination) trial concept from 2007.⁵

We have previously reported the preliminary results of our first IMPACT personalized medicine trial, as well as validation and landmark analyses.^6,7 Here, we present results for 1,436 additional patients who underwent clinical-grade molecular profiling before treatment in the phase I clinical trials program. We demonstrate that matched targeted therapy (MTT) is associated with improved outcomes. Prior preclinical data suggest that the presence of alterations in the mitogen-activated protein kinase (MAPK) pathway confers resistance to targeted agents against alterations in the phosphatidylinositol 3-kinase (PI3K)/AKT/mammalian target of rapamycin (mTOR) pathway.⁸ We demonstrate that matching of the PI3k/AKT/mTOR pathway alone in the presence of MAPK pathway alterations is associated with outcomes that are similar to nonmatched therapy.

PATIENTS AND METHODS

Patients

By using the IMPACT1 database, we identified consecutive patients who were referred to the phase I clinical trials program from February 2012 to October 2013 and who underwent molecular analysis, per previously described methodology.⁶ All patients provided written informed consent in accordance with the guidelines of the MD Anderson Cancer Center institutional review board (Appendix, online only).

Analysis of Molecular Aberrations

Molecular and immunohistochemical profiling was performed in the Clinical Laboratory Improvement Amendments (CLIA)–certified Molecular Diagnostics Laboratory at MD Anderson Cancer Center as previously described.⁶

The Appendix provides detailed data on molecular profiling used.

Treatment

Assignment to a clinical trial was determined after clinical, laboratory, and/or pathologic data from patient records were reviewed. Patients whose tumors had a molecular aberration were preferably treated in a clinical trial with a matched targeted agent when available.

If two or more molecular aberrations were present, patients were preferably treated in a trial of therapies that targeted both aberrations. If such a trial was unavailable, a physician’s choice of a trial that targeted one aberration was permitted (Appendix).

Definition of Matched Therapy

A patient’s tumor was considered matched with a targeted therapy if a drug was known to inhibit the aberration at low nanomolar concentrations or if an antibody targeted the alteration product, per literature data ^9-12 (Appendix). Subset analyses are listed in Table 1.

Table 1.

Response by RECIST (N = 619 evaluable patients)

End Points and Statistical Methods

Statistical analyses were performed by a biostatistician (K.H.) using S-Plus software for Windows (version 8.2; TIBCO Software, Palo Alto, CA). By using Response Evaluation Criteria in Solid Tumors (RECIST) guidelines,^14,15 we assessed objective response rates (complete response [CR] and partial response [PR]), clinical benefit rates (stable disease [SD] ≥ 6 months plus PR plus CR), and rates of failure-free survival (FFS) and overall survival (OS). Details are in the Appendix.

RESULTS

Patient Characteristics

Tumor molecular profiling was ordered in 1,436 consecutive patients. Overall, 1,179 (82%) of 1,436 patients had one or more aberration. The remaining 257 patients had no alterations and were excluded. Of 1,179 patients with one or more aberration, 914 (77.5%) had one or more targetable aberration (Appendix Fig A1, online only). Of these 914 patients, 277 patients with targetable molecular aberrations did not receive evaluable treatment by the end of the analysis period for the following reasons: treatment was received in other centers (n = 118; 42.6%); worsening of performance status (n = 39; 14.1%); regional therapy (n = 31; 11.2%); ineligibility for targeted therapy (n = 27; 9.7%); lost to follow-up (n = 23; 8.3%); declined (n = 15; 5.4%); insurance issues (n = 12; 4.3%); and other (n = 12; 4.3%).

Overall, 637 patients had at least one targetable molecular alteration and received treatment (specific aberrations are shown in Appendix Fig A2 [online only]; patient characteristics in Table 2). The most frequently identified alterations were estrogen receptor overexpression, TP53 mutation, KRAS mutation, PTEN loss or mutation, PIK3CA mutation, and BRAF mutations.

Table 2.

Baseline Characteristics of 637 Patients With Mutations by Type of Therapy

Group 1 Analysis

In a group 1 analysis, we used the traditional definition of matched therapy, as previously published^6,7,13 (Appendix and Table 1 footnote). Overall, 390 patients (27% of 1,436 total patients; or 61.2% of 637 patients) received matched therapy and 247 (17.2%) received nonmatched therapy. Tumor types treated are listed in Table 3. The respective clinical outcomes of patients treated with MTT versus non-MTT were as follows: CR/PR rate, 11% versus 5% (P = .0099); SD ≥ 6 months/CR/PR rate, 29% versus 24% (P = .13); FFS (hazard ratio [HR], 0.81; 95% CI, 0.69 to 0.96; P = .015); and OS (HR, 0.84; 95% CI, 0.71 to .099; P = .041; Table 1 and Figs 1A and 1B). These results show that matched therapy compared with nonmatched therapy was associated with significantly higher rates of objective response, FFS, and OS (nonsignificant for SD ≥ 6 months/CR/PR).

Table 3.

Patient Characteristics: Tumor Type by Type of Therapy

Fig 1.

Failure-free survival and overall survival in patient subgroups. Group 1: Matched targeted therapy (MTT) versus non-MTT. (A) Failure-free survival according to matched targeted therapy. (B) Overall survival according to matched targeted therapy. Group 2: MTT with TP53 (includes matching as previously published plus inclusion of anti–vascular endothelial growth factor [VEGF] therapy as targeted therapy against TP53) versus non-MTT. (C) Failure-free survival according to matched targeted therapy. (D) Overall survival according to matched targeted therapy. Group 3: MTT minus negative matches (includes patients who had matched targeted therapy, without considering TP53 mutations matched to VEGF/VEGF receptor [VEGFR] inhibitors and excluding the negative matches) versus non-MTT versus negative matches, with the latter defined as matched targeted therapy against an alteration in the PI3K axis in the presence of KRAS or BRAF or another MEK pathway mutation (the latter being a known resistance pathway) and no MEK/RAF inhibitors. (E) Failure-free survival. (F) Overall survival. Group 4: MTT with TP53 match minus negative matches (includes TP53 mutation matches to VEGF/VEGFR inhibitors but excludes negative matches) versus negative matches (see definition for group 3; Table 1) versus non-MTT. (G) Failure-free survival. (H) Overall survival. Group 5: Non-MTT (includes nontargeted therapy or targeted therapy against an alteration in the PI3K axis in the presence of KRAS or BRAF or other MEK pathway mutation and no MEK/RAF inhibitor) versus MTT (includes all other groups, including TP53 matches). (I) Group 5: failure-free survival. (J) Overall survival. HR, hazard ratio.

Two-month landmark analysis demonstrated that in patients who received MTT, FFS duration in responders versus nonresponders was 7.6 versus 4.3 months, respectively (P < .001; Figs 2A to 2D), whereas in patients who received non-MTT, FFS duration was 6.6 versus 4.1 months, respectively (P = .001). Among patients who received MTT, median OS duration in responders versus nonresponders was 23.4 versus 8.5 months, respectively (P < .001), whereas in patients who were treated with non-MTT, respective median OS duration was 15.2 versus 7.5 months (P = .43). Therefore, the 2-month landmark analysis demonstrated that FFS duration was significantly longer in responders than in nonresponders. For OS, the matched group of responders versus nonresponders had significantly longer OS duration (reaching a median of 23.4 months for responders). No difference was observed in the unmatched group.

Fig 2.

Two-month landmark analyses. Group 1: Matched targeted therapy (MTT), as previously published. (A) Failure-free survival according to objective response (complete response [CR] + partial response [PR], yes) versus no response (No) in patients who received matched therapy (hazard ratio [HR], 0.49; 95% CI, 0.34 to 0.70; P < .001). (B) Failure-free survival according to objective response (CR + PR, yes) versus no response (No) in patients who received nonmatched therapy (HR, 0.36; 95% CI, 0.18 to 0.70; P = .001). (C) Overall survival according to objective response (CR + PR, yes) versus no response (No) in patients who received matched therapy (HR, 0.49; 95% CI, 0.34 to 0.71; P < .001). (D) Overall survival according to objective response (CR + PR, yes) versus no response (No) in patients who received nonmatched therapy (HR, 0.66; 95% CI, 0.35 to 1.25; P = .17). Group 2: MTT plus inclusion of anti–vascular endothelial growth factor (VEGF) therapy as targeted therapy against TP53. (E) Failure-free survival according to objective response (CR + PR, yes) versus no response (No) in patients who received matched therapy (HR, 0.50; 95% CI, 0.36 to 0.71; P < .001). (F) Failure-free survival according to objective response (CR + PR, yes) versus no response (No) in patients who received nonmatched therapy (HR, 0.33; 95% CI, 0.16 to 0.71; P = .001). (G) Overall survival according to objective response (CR + PR, yes) versus no response (No) in patients who received matched therapy (HR, 0.49; 95% CI, 0.34 to 0.70; P < .001). (H) Overall survival according to objective response (CR + PR, yes) versus no response (No) in patients who received nonmatched therapy (HR, 0.73; 95% CI, 0.36 to 1.49; P = .36).

Group 2 Analysis

The group 2 analysis was similar to that of group 1, except that TP53 mutations were considered matched to vascular endothelial growth factor (VEGF)/VEGF receptor (VEGFR) inhibitors, as has been previously reported.^11,12 Clinical outcomes for matched versus nonmatched patients were as follows: CR/PR rate, 11% versus 5% (P = .02); SD ≥ 6 months/CR/PR rate, 29% versus 22% (P = .055); FFS (HR, 0.80; 95% CI, 0.67 to 0.95; P = .012); and OS (HR, 0.85; 95% CI, 0.71 to 1.01; P = .07; Table 1 and Figs 1C and 1D). These results demonstrated that, compared with unmatched therapy, matched therapy, including TP53 mutations matched to VEGF/VEGFR inhibitors, was associated with significantly higher rates of objective response and FFS and a trend toward higher rates of SD ≥ 6 months/CR/PR and OS.

Two-month landmark analyses demonstrated that median FFS duration for responders (CR/PR) versus nonresponders in patients who received matched therapy was 7.6 versus 4.4 months (P ≤ .001; Figs 2E to 2H), whereas for nonmatched patients, median FFS duration was 6.6 versus 4.1 months (P = .001).

For OS, the 2-month landmark analyses for responders (CR/PR) versus nonresponders in patients who received matched therapy demonstrated a median OS of 23.6 versus 8.4 months (P < .001), whereas nonmatched patients had a median OS of 9.8 versus 7.7 months, respectively (P = .29). As with group 1, the group 2 landmark analysis showed that FFS was significantly longer in responders than in nonresponders. For OS, the matched group of responders versus nonresponders had significantly longer OS, reaching a median of 23.6 months for responders. No difference was observed in the unmatched group.

Group 3 Analysis

The group 3 analysis was similar to that of group 1, except that we isolated a group of 36 patients who had negative matches, a term applied when patients were matched by virtue of a PI3K pathway alteration to a PI3K/Akt/mTOR axis inhibitor, but a RAS/RAF/MEK alteration was also present and not matched to a drug. Because the latter alterations are known to cause resistance to targeted agents, we sought to determine the outcomes of patients in this group separately.

For unmatched versus negatively matched patients, clinical outcomes were as follows: CR/PR rate, 5% versus 6% (P = 1.0); SD ≥ 6 months/CR/PR rate, 24% versus 19% (P = .6); FFS (HR, 0.9; 95% CI, 0.6 to 1.3; P = .51); and OS (HR, 0.9; 95% CI, 0.6 to 1.3; P = .48; Table 1 and Figs 1E and 1F). In contrast, when comparing matched patients—excluding the negative matches—with negatively matched patients, outcomes were as follows: CR/PR rate, 12% versus 6% (P = .25); SD ≥ 6 months/CR/PR rate, 30% versus 19% (P = .12); FFS (HR, 0.7; 95% CI, 0.5 to 1.0; P = .045); and OS (HR, 0.7; 95% CI, 0.5 to 1.0; P = .06). P value (log-rank test) for multiple comparisons of FFS was .0072 (Fig 1E) and for OS, .024 (Fig 1F). These results showed that negative matches attained rates of CR/PR and SD ≥ 6 months/CR/PR that were similar to those of unmatched patients, and that FFS and OS were equivalent. Conversely, depending on the outcome parameter, negatively matched patients either had significantly worse outcomes or demonstrated a trend toward worse outcomes compared with positively matched patients. These results are consistent with the hypothesis that matching patients who harbor a PI3K and a MAPK pathway alteration to a PI3K pathway inhibitor alone does not improve outcomes.^8,16

Group 4 Analysis

The group 4 analysis was similar to that of group 2, which included TP53 mutations matched to VEGF/VEGFR inhibitors, except that we again isolated the group of 36 patients who had negative matches. For unmatched versus negatively matched patients, results were as follows: CR/PR rate, 5% versus 6% (P = .9); SD ≥ 6 months/CR/PR rate, 22% versus 19% (P = .5); FFS (HR, 0.91; 95% CI, 0.63 to 1.3; P = .6); and OS (HR, 0.88; 95% CI, 0.61 to 1.27; P = .5; Table 1 and Figs 1G and 1H). In contrast, when comparing matched patients—excluding the negative matches—with negatively matched patients, outcomes were as follows: CR/PR rate, 11% versus 6% (P = .3); SD ≥ 6 months/CR/PR rate, 30% versus 19% (P = .18); FFS (HR, 0.71; 95% CI, 0.5 to 1.0; P = .05); and OS (HR, 0.73; 95% CI, 0.51 to 1.04; P = .079; Table 1 and Figs 1G and 1H). P value (log-rank test) for multiple comparisons of FFS was .0063 (Fig 1G) and for OS, .043 (Fig 1H). These results demonstrate that negative matches had rates of CR/PR and SD ≥ 6 months/CR/PR that were similar to those of unmatched patients, and that FFS and OS were similar. Conversely, depending on the outcome parameter, negatively matched patients either had significantly worse outcomes or trended toward worse outcomes compared with positively matched patients.

Group 5 Analysis

The group 5 analysis compared positively matched patients, which included TP53 mutations matched to VEGF/VEGFR inhibitors, with negatively matched patients—as defined for group 4—plus unmatched patients. Results are listed in Table 1 and Figs 1I and 1J: CR/PR rate, 11% versus 5% (P = .0068); SD ≥ 6 months/CR/PR rate, 30% versus 22% (P = .017); FFS (HR, 0.77; 95% CI, 0.65 to 0.91; P = .002); and OS (HR, 0.81; 95% CI, 0.68 to 0.96; P = .017). These results demonstrate that, for all outcome parameters, matched patients, which included TP53 mutations matched to VEGF/VEGFR inhibitors, did significantly better than the combined group of unmatched and negatively matched patients.

DISCUSSION

The application of precision medicine using targeted agents to inhibit the function of tumor molecular alterations is a complex process. Our results support the use of genomic matching in patients with advanced cancer. Patients who were treated with MTT had higher rates of CR and PR (11% v 5%; P = .0099), longer FFS (3.4 v 2.9 months; P = .0015), and longer OS (8.4 v 7.3 months; P = .041) than unmatched patients. Of interest, the 2-month landmark analysis showed that median FFS was significantly longer in responders than in nonresponders, but that difference was more pronounced in the matched group. For OS, the matched group of responders versus nonresponders—but not the unmatched group of responders versus nonresponders—had significantly longer OS, reaching a median of 23.4 months for matched responders.

In this retrospective analysis, we report 1,436 patients who underwent prospective molecular profiling with a CLIA-certified test. Overall, 1,179 patients (82%) had molecular alterations (Appendix Fig A1). Of 637 treated patients with at least one alteration, 61% (n = 390 patients; 27% of the total 1,436 patients) received MTT and 39% received non-MTT. This number is similar to that of other studies; tumors that were interrogated with larger panels of genes are more likely to have alterations.^13,17,18 Indeed, in some reports, approximately 95% of patients from whom adequate tissue is obtained are found to have molecular alterations.^13,16-20 The complexity of using precision medicine is explained, in part, by the coexistence of multiple molecular alterations, as seen in our population, the lack of efficient targeted agents for all alterations, and the fact that some alterations are more difficult to target than others.¹⁹ This observation is consistent with previously published results that have demonstrated that a matching score—number of matches divided by number of alterations—correlates well with all outcome parameters.¹³ A limitation of the current analysis relates to the fact that gene panels of different sizes were used for molecular profiling, and, hence, the impact of the number of alterations in each patient could be confounded by the panel size.

Several previous studies have been performed with variable results. In a multicenter study in metastatic lung adenocarcinoma, an oncogenic driver was found in 64% of 1,007 patients.²⁰ Patients with an oncogenic driver who received genotype-directed therapy experienced longer survival than did those with an oncogenic driver who did not receive genotype-directed therapy (median, 3.5 v 2.4 years; adjusted HR, 0.69; P = .006).²⁰ Our previous studies—MD Anderson personalized therapy initiative and University of California San Diego PREDICT studies—have also demonstrated improved outcomes with the selection of therapy that is matched to patients’ tumor molecular profiles in patients with diverse cancers.^6,7,13,18 Other trials, however, have reported negative results. For instance, the SHIVA trial—a randomized molecular navigation trial—did not achieve its endpoints¹⁶; however, approximately 80% of patients in SHIVA were matched to single-agent hormone modulators or mTOR inhibitors.^21,22 Our current results confirm that single-agent PI3K pathway inhibitors are generally not effective, especially in patients with concomitant MAPK pathway alterations. Indeed, for these patients, targeting the PI3K pathway without targeting the MAPK pathway was associated with outcomes that were indistinguishable from, or numerically worse than, those in unmatched patients (Table 1 and Figs 1E to 1H). This observation confirms previously published data^8,16,21,22 and should be taken into consideration in the design and reporting of future clinical trials. In contrast, combination therapy that includes PI3K/Alt/mTOR axis–matched therapy seems to be associated with high rates of response and disease stabilization.⁸

Several recent studies from our group and others have suggested that patients with TP53 mutations may have better outcomes when treated with antiangiogenesis agents, perhaps because TP53 alterations result in increased VEGF-A levels.^9-12 We have also reported that patients with TP53 alterations who were treated with anti-VEGF/VEGR agents—with no other matches—had higher rates of response, PFS, and OS compared with patients with wild-type TP53.¹² We therefore performed subanalyses to determine outcomes if groups of matched patients included TP53 matched to VEGF/VEGFR inhibitors (groups 2, 4, and 5; Table 1 and Fig 1). Outcome parameters were favorable in groups that included patients who were matched in this way, and the strongest P values were observed in the group of patients that both included TP53 alterations matched to VEGF/VEGFR inhibitors and excluded patients who harbored both PI3K and MAPK pathway alterations matched to only PI3K/Akt/mTOR axis inhibitors (group 5; Table 1 and Fig 1). These observations are supportive of previously published findings that TP53 may be a targetable alteration, and suggest that, in patients with PI3K pathway anomalies, the presence of MAPK resistance alterations must be addressed.

Our study had several limitations. First, it was a retrospective, nonrandomized study. Second, although statistical significance was obtained for almost all subset analyses for matching, differences between matched and unmatched groups were sometimes relatively small; however, when the 2-month landmark analysis was applied, more substantial differences were observed—for instance, OS by response was 23.4 versus 8.5 months for the matched group (P < .001). Third, molecular analysis was performed by using archival tissue in most patients, tissue was sometimes inadequate for testing all available molecular aberrations, and the availability of CLIA-certified tests varied over the years and, thus, some molecular profiles became suboptimal by later standards. Fourth, patients were heavily pretreated and were treated in the phase I setting; thus, treatment was experimental and doses and schedules were generally imperfect. To address the limitations of this study, we have initiated the IMPACT 2 trial, a phase II randomized study evaluating molecular profiling and targeted therapy in metastatic cancer (NCT02152254).^6,7 Multiple other nonrandomized trials of molecular profiling have been completed or are ongoing (NCT01771458, NCT01827384, NCT02154490, NCT01042379, and NCT01856296).^{6,7,13,16,17,23,24}

In conclusion, our analysis demonstrates that precision medicine is feasible in an academic setting and that it is associated with superior outcomes compared with non-MTT in patients with advanced cancer. The rate of matching was approximately 27% in our group of patients, which is consistent with some previous studies but higher than that observed in other trials.^6,7,13,17,18 Higher rates of matching may be associated with larger gene panels, which produce more potentially actionable alterations,¹⁸ with more robust clinical trial portfolios and with molecular studies with flexible algorithms and eligibility criteria that are not overly restrictive. Trials are beginning to address the critical issue of combination therapies.¹³ Other crucial issues include beginning treatment earlier in the course of disease when tumor genomics are less complex, as well as the emerging role of immunotherapy.

SUPPLEMENTAL METHODS

Patients

In brief, patients with advanced or metastatic cancer for whom standard-of-care therapy had been exhausted or for whom no Food and Drug Administration–approved therapy was available for their indication were referred for participation in phase I clinical trials. All protocols required that participants have evidence of evaluable or measurable disease according to RECIST guidelines^14,15 and an Eastern Cooperative Oncology Group performance status of 0 to 2. Additional eligibility criteria varied depending on which specific phase 1 clinical trial in which the patient enrolled.

Methods

Archival formalin-fixed, paraffin-embedded tissue blocks or tissue from fine-needle aspiration or surgical biopsies was used for profiling. Molecular and/or immunohistochemistry assays were performed when tissue was available by using standard operating procedures and PCR-based sequencing technology.

Tumor molecular profiling for 637 patients was performed in the following laboratories: MD Anderson Cancer Center (hotspot testing of 11 genes, n = 282; 46 genes, n = 145; and 50 genes, n = 20; patients, n = 447), Foundation Medicine (next-generation sequencing of 182 genes; patients, n = 151), Knight Diagnostics (48 genes; patients, n = 20), and other CLIA-certified laboratories (patients, n = 19).

Analysis of Molecular Aberrations

DNA was extracted from microdissected, paraffin-embedded tumor samples, and the coding regions for specific exons, depending on the test ordered, were analyzed for the following genes: PIK3CA (exon 9: codons 532 to 554; exon 20: codons 1,011 to 1,062); BRAF (exon 15: codons 595 to 600); KRAS and NRAS (exon 2: codons 12, 13, and 61); EGFR (exons 18 to 21 of the kinase domain); KIT (exons 9, 11, 13, and 17); GNAQ (exon 5); TP53 (exons 4 to 9); MET (exon 2: codon 375; exon 11: codon 848; exon 14: codons 988 and 1010; exon 16: codons 1112 and 1124; exon 19: codons 1,248, 1,253, and 1,268); and RET (exon 10: codons 609, 611, 618, and 620; exon 11: codon 634; exon 16: codon 918). The sensitivity of mutation assays for the detection of PIK3CA, BRAF, KRAS, NRAS, GNAQ, and MET mutations was approximately one in 10 mutation-bearing cells in the microdissected area. For EGFR, KIT, and TP53 mutations, the lower limit of detection was approximately one in five mutation-bearing cells. Loss of expression of the tumor suppressor nuclear protein, PTEN, was determined by using immunohistochemical staining with the monoclonal mouse anti-human PTEN clone 6H2 (code M3627; Dako, Carpinteria, CA). Anaplastic lymphoma kinase (ALK) translocation was assessed using fluorescent in situ hybridization (commercial probe; Abbott Molecular, Des Plaines, IL). With the availability of next-generation sequencing, some patient tumors were assessed by more extensive molecular profiling of 48 or more genes at MD Anderson or in other CLIA-certified laboratories. Selected patients also had tumor molecular profiling using next-generation sequencing performed by Foundation Medicine (Cambridge, MA). Genomic libraries were captured for 3,230 exons in 182 cancer-related genes plus 37 introns from 14 genes that are often rearranged in cancer and sequenced to an average median depth of 734 × with 99% of bases covered > 100 ×.

HER2/neu and estrogen and progesterone receptor status were also determined by immunohistochemistry. Estrogen and progesterone receptors were assessed by using the 6F11 antibody (Novocastra Laboratories, Newcastle-upon-Tyne, United Kingdom). Alternatively, fluorescent in situ hybridization analysis was used to measure the copy number of HER2/neu.

Before the use of next-generation sequencing, the CLIA pathology laboratory prioritized the testing of molecular aberrations on the basis of their known frequency in cancer and/or whether they were perceived as actionable or as having other clinical relevance to patients. Treating physicians requested all available molecular tests that were CLIA-certified at MD Anderson Cancer Center at the time a patient who was interested in receiving treatment in the phase I clinical trials program presented to the phase I clinic. If tissue available for analysis was limited, the treating physician prioritized molecular testing on the basis of tumor type and the availability of clinical trials that could impact specific targets.

Treatment

Treatment regimens included one or more drugs. Allocation of patients to investigational treatment varied over time according to protocol availability, eligibility criteria, histologic diagnosis, the patient's prior response to therapy, potential toxicity, insurance coverage, and patient preference or physician choice. In addition, molecular profiling was added as a screening procedure. Physicians prioritized matched therapy—versus nonmatched therapy—on the basis of the following criteria: patients had an actionable molecular aberration; matched targeted therapy (MTT) was available; patients met the eligibility criteria; insurance coverage was obtained; and patients agreed to comply with study requirements. Because of the 3 + 3 design in most phase I clinical trials (requiring monitoring of three patients for 1 month before dose escalation, and, therefore, sometimes resulting in a lack of immediate protocol availability), the multi-institutional study design in several sponsored studies (further limiting the number of patients enrolled per institution), and restrictions associated with eligibility criteria, not all patients with an actionable aberration could be treated on a protocol with matched therapy.

Definition of Matched Therapy

PIK3CA mutations and PTEN loss could be targeted by inhibitors of AKT and mammalian target of rapamycin, as well as phosphatidylinositol 3-kinase (PI3K), as AKT and mammalian target of rapamycin are downstream of activated PIK3CA and because both PIK3CA mutations and PTEN loss, which usually reflects PTEN mutation, activate PI3K. GNAQ, RAS, and BRAF mutations could be targeted by inhibitors of MEK. BRAF mutations could also be targeted by BRAF inhibitors. Other aberrations, such as RET, EGFR, KIT, and MET mutations, were targeted by drugs that inhibit the respective activated kinase with 50% inhibitory concentration (IC₅₀) in the low nanomolar range. EGF receptor could be targeted by anti–EGF receptor antibodies. TP53 mutations were not initially considered actionable by drugs available in our trials; however, with more recent emerging data,8-11 TP53 mutations were considered actionable by vascular endothelial growth factor (VEGF)/VEGF receptor (VEGFR) inhibitors, and an appropriate secondary analysis on this basis was included. We performed subset analyses for clinical outcome comparisons using the following (Table 1, footnote): Group 1: MTT, as previously published6,7; non-MTT (all others). Group 2: MTT with TP53, which includes MTT and VEGF/VEGFR therapy as targeted therapy against TP53; non-MTT (all others). Group 3: Negative matches, which includes only MTT against an alteration in the PI3K axis in the presence of KRAS or BRAF or another MEK pathway mutation (the latter being a known resistance pathway) and no MEK/RAF inhibitor.⁸ Negative matches do not include patients with additional positive matches; positive MTT minus negative matches, which includes patients who had MTT (without considering TP53 mutations matched to VEGF/VEGFR inhibitors) minus the negative matches; non-MTT refers to patients who had no MTT minus the negative matches. Group 4: Negative matches; MTT with TP53 match minus negative matches; non-MTT. Group 5: MTT with TP53 match minus negative matches refers to all matched groups, including TP53 matches to VEGF/VEGFR inhibitors, but excluding negative matches (group 4 negative matches excluded); non-MTT plus negative matches includes patients without a match as well as MTT against an alteration in the PI3K axis in the presence of KRAS or BRAF or other MEK pathway mutation (the latter being a known resistance pathway) and no MEK/RAF inhibitor (group 4 negative matches included).

The rationale for subset analyses is as follows: Group 1 is straightforward and consistent with our previously published analyses.^6,7 Group 2 included anti-VEGF/VEGFR agents matched to TP53 alterations. Recent studies demonstrated that altered TP53 upregulates angiogenesis and VEGF-A.¹¹ In addition, published data from our group and other investigators showed that patients with TP53 alterations have improved outcomes if treated with anti-VEGF/VEGFR agents.^9-12 Therefore, we did a subset analysis (group 2) to examine matched versus unmatched therapy, including TP53 alterations matched to anti-VEGF/VEGFR agents. We also performed another subset analysis, which we designated as negative matches (group 3). Negative matches pertained to individuals who had both a PI3K axis alteration and a KRAS or BRAF or other MEK pathway mutation—the latter being a known resistance pathway—and were matched only with a PI3K axis inhibitor. Several investigators suggest that these patients are resistant to a PI3K axis inhibitor, likely because the MAPK pathway is not targeted.^8,17,18 In groups 4 and 5, we further explored analyses that included negative matches together with unmatched patients as well as TP53 alterations matched to VEGF/VEGFR inhibitors.

End Points and Statistical Methods

Matched versus unmatched and grouping was verified by one of the coauthors (R.K.) who was blinded to outcome data at the time that designation was assigned.

Overall survival (OS) was measured from the initiation of participation in the phase I trial until death or last follow-up. Failure-free survival (FFS) was measured from the first day of treatment on a clinical trial until the date of discontinuation of treatment of any reason, including disease progression, treatment toxicity, or death, whichever came first. The decision to discontinue treatment on protocol was made by the treating physician and was based on the patient’s history, clinical presentation, and imaging studies (response assessment using RECIST criteria). Patient characteristics were analyzed by using descriptive statistics. Survival functions were estimated by using the Kaplan-Meier method. All P values presented are two sided, and the statistical significance level was P ≤ .05.

We also performed landmark analyses for FFS and OS at 2 months, as we previously published.⁷ We omitted patients with event times or censoring times of < 2 months and stratified outcome by overall response status.

We also compared matched responders with nonmatched responders (Supplemental file).

Results

Objective response in patients who were treated with matched and nonmatched therapy by tumor type and molecular alterations is shown in Appendix Tables A1 and A2. The small numbers of patients in the subgroups preclude robust analyses.

We also performed a 2-month landmark analysis that compared FFS and OS between matched responders and nonmatched responders. Median FFS was 7.6 months in the matched therapy group and 6.6 months in the nonmatched group (hazard ratio, 0.9; 95% CI, 0.4 to 1.9; P = .78; Appendix Fig A3A). Median OS was 23.4 months in the matched therapy group and 15.2 months in the nonmatched group (hazard ratio, 0.6; 95% CI, 0.3 to 1.2; P = .15; Appendix Fig A3B). The small numbers of patients in the subgroups preclude robust analyses for most comparisons.

We then performed an analysis, excluding patients with estrogen receptor/progesterone receptor overexpression from the matched therapy group. Results were as follows: Median FFS was 3.4 months in patients who were treated with matched therapy and 2.9 months in patients who were treated with nonmatched therapy (P = .034). Median OS was 7.3 months in patients who were treated with matched therapy and 7.5 months in patients who were treated with nonmatched therapy (P = .16; Appendix Figs A3C and A3D).

Fig A1.

CONSORT diagram. The 257 patients who had no alterations were excluded from the analysis. Overall, 277 patients with molecular aberrations did not receive treatment by the time of this analysis for the following reasons: treated elsewhere (n = 118; 42.6%); worsening performance status (n = 39; 14.1%); received regional therapy (n = 31; 11.2%); ineligibility (n = 27; 9.7%); lost to follow-up (n = 23; 8.3%); declined investigational therapy (n = 15; 5.4%), insurance issues (n = 12; 4.3%); treated after the cutoff date of the data (n = 12; 4.3%).

Fig A2.

Numbers of patients with specific alterations. The following alterations were found in fewer than 10 patients: CCNE1 and CTNNB1 each in eight patients; CDH1, FGF3, MYST3, NFKBIA, and STK11 each in seven patients; ARID1A, FGF4, FGF19, HRAS, MLL2, and SOX2 each in six patients; AKT1, AKT3, CCND3, GNAS, IDH1, PIK3R1, RET, VHL, and ZNF703 each in five patients; BRCA2, CCND2, EMSY, FGFR2, FGFR3, KDM6A, MYCL1, and PDGFRA each in four patients; ATRX, FGF10, IRS2, LRP1B, MDM4, NF2, PTPRD, RICTOR, and TSC2 each in three patients; AR, ATR, AURKA, BAP1, BARD1, BCL2L2, CSF1R, EP300, ERBB2, ESR1, FANCA, FGF23, FLT4, GNAQ, GRIN2A, HGF, IKBKE, JAK2, JUN, MAP2K4, MLH1, MPL, MSH6, NKX2-1, NOTCH1, PALB2, PBRM1, PTCH1, SMAD2, and TSC1 each in two patients; and ABL1, AKT, AKT2, ALK, ARIDA, ARID2, ARFRP1, AURKB, BCORL1, BRIP1, CCDC6-RET, CDK6, CMARCB1, COX2, CREBBP, CRKL, DAXX, DNMT3A, DOT1L, EPHB1, ERBB3, ERBB4, ESR6, ESR13, ETV1, FAM123B, FANCC, FGF6, FLT1, FLT3, FOXL2, GATA3, IGF1R, IKZF1, JAK3, JAK2, MAP2K2, MAP3K1, MEN1, MSH2, YD88, PAX5, PDGFRB, PPP2R1A, PRKDC, PTPN11, RPTOR, STK11 loss, SMARCCB1, SMARCA4, SRC, TET2, TOP1, USP9X, WT1, XPO1, and ZNF217 each in one patient.

Fig A3.

(A) Landmark analysis of responders: Failure-free survival (hazard ratio [HR], 0.9; 95% CI, 0.4 to 1.9; P = .78. (B) Landmark analysis of responders: Overall survival (HR, 0.6; 95% CI, 0.3 to 1.2; P = .15. (C) Analysis excluding patients with estrogen receptor (ER)/progesterone receptor (PR) overexpression-based matching from the matched therapy group 1: Failure-free survival. (D) Analysis excluding patients with ER/PR overexpression-based matching from the matched therapy group 1: Overall survival. MTT, matched targeted therapy.

Table A1.

Objective Response by Tumor Type

Table A2.

Objective Response by Molecular Alteration

AUTHOR CONTRIBUTIONS

Conception and design: Apostolia-Maria Tsimberidou, Razelle Kurzrock

Financial support: Apostolia-Maria Tsimberidou

Administrative support: Apostolia-Maria Tsimberidou

Provision of study materials or patients: Apostolia-Maria Tsimberidou, David S. Hong, Jennifer J. Wheler, Gerald S. Falchook, Aung Naing, Siqing Fu, Sarina Piha-Paul, Filip Janku, Funda Meric-Bernstam, Patrick Hwu, Bryan Kee, Merrill S. Kies, Russell Broaddus, John Mendelsohn, Razelle Kurzrock

Collection and assembly of data: Apostolia-Maria Tsimberidou, Yang Ye, Carrie Cartwright

Data analysis and interpretation: Apostolia-Maria Tsimberidou, Kenneth R. Hess, Razelle Kurzrock

Manuscript writing: All authors

Final approval of manuscript: All authors

Accountable for all aspects of the work: All authors

AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST

The following represents disclosure information provided by authors of this manuscript. All relationships are considered compensated. Relationships are self-held unless noted. I = Immediate Family Member, Inst = My Institution. Relationships may not relate to the subject matter of this manuscript. For more information about ASCO's conflict of interest policy, please refer to www.asco.org/rwc or po.ascopubs.org/site/ifc.

Apostolia-Maria Tsimberidou

Research Funding: EMD Serono (Inst), Baxter (Inst), Foundation Medicine (Inst), Onyx Pharmaceuticals (Inst), Bayer (Inst), Boston Biomedical (Inst), Placon Therapeutics (Inst)

David S. Hong

Stock and Other Ownership Interests: MolecularMatch, Oncorena

Honoraria: Adaptimmune, Baxter, Merrimack Pharmaceuticals, Bayer

Consulting or Advisory Role: Baxter, Bayer

Research Funding: Novartis, Genentech, Eisai, AstraZeneca, Pfizer, miRNA Therapeutics, Amgen, Daiichi Sankyo, Merck, Mirati Therapeutics, Eli Lilly

Travel, Accommodations, Expenses: Loxo, miRNA Therapeutics

Other Relationship: Oncorena

Yang Ye

No relationship to disclose

Carrie Cartwright

No relationship to disclose

Jennifer J. Wheler

No relationship to disclose

Gerald S. Falchook

Employment: Sarah Cannon Research Institute, HealthONE

Research Funding: GlaxoSmithKline, Millennium Pharmaceuticals, EMD Serono, Celgene, MedImmune, Genmab, Vegenics, Novartis, AstraZeneca, Incyte, ARMO BioSciences, Kolltan Pharmaceuticals, 3-V Biosciences, AbbVie, Aileron Therapeutics, DelMar Pharmaceuticals, eFFECTOR Therapeutics, Strategia Therapeutics, FujiFilm, Hutchison MediPharma, Regeneron, Biothera, Curegenics, Curis, Eli Lilly, Jounce Therapeutics, OncoMed, Precision Oncology, Syndax, Taiho Pharmaceutical, Tesaro

Patents, Royalties, Other Intellectual Property: Handbook of Targeted Cancer Therapy: Millennium

Travel, Accommodations, Expenses: Millennium Pharmaceuticals, Sarah Cannon Research Institute, EMD Serono, Bristol-Myers Squibb

Aung Naing

Consulting or Advisory Role: Novartis, CytomX Therapeutics

Research Funding: National Cancer Institute, EMD Serono, MedImmune, Healios, Atterocor, Amplimmune, ARMO BioSciences, Karyopharm Therapeutics, Incyte, Novartis, Regeneron, Baxter

Travel, Accommodations, Expenses: ARMO BioSciences

Siqing Fu

No relationship to disclose

Sarina Piha-Paul

Consulting or Advisory Role: Genentech

Research Funding: GlaxoSmithKline, XuanZhu, Puma Biotechnology, Novartis, Merck Sharp & Dohme, Curis, Principa Biopharma, Biomarin, Helix BioPharma, Bayer, AbbVie, Incyte, Five Prime Therapeutics, Cerulean Pharma, MedImmune, Medivation

Filip Janku

Stock and Other Ownership Interests: Trovagene

Consulting or Advisory Role: Deciphera, Trovagene, Novartis, Sequenom, Foundation Medicine, Guardant Health

Research Funding: Biocartis, Trovagene, Novartis, BioMed Valley Discoveries, Foundation Medicine, Roche, Agios, Astellas Pharma, Deciphera, Symphogen, Plexxikon, Piqur, FujiFilm

Other Relationship: Bio-Rad

Funda Meric-Bernstam

Honoraria: Dialecta

Consulting or Advisory Role: Genentech, Inflection Biosciences, Pieris Pharmaceuticals, Clearlight Diagnostics, Darwin Health

Research Funding: Novartis, AstraZeneca, Taiho Pharmaceutical, Genentech, Calithera Biosciences, Debiopharm Group, Bayer, Aileron Therapeutics, Puma Biotechnology, Verastem, CytomX Therapeutics, Jounce Therapeutics, Zymeworks, Effective Pharmaceuticals, Curis, Patrick Hwu

Stock and Other Ownership Interests: Lion Biotechnologies, Immatics

Consulting or Advisory Role: Lion Biotechnologies

Research Funding: Genentech

Patrick Hwu

Stock and Other Ownership Interests: Lion Biotechnolgies, Immatics

Consulting or Advisory Role: Lion Biotechnolgies

Research Funding: Genentech (Inst)

Bryan Kee

Stock and Other Ownership Interests: Medtronic

Merrill S. Kies

No relationship to disclose

Russell Broaddus

No relationship to disclose

John Mendelsohn

Leadership: Merrimack Pharmaceuticals

Stock and Other Ownership Interests: Merrimack Pharmaceuticals

Patents, Royalties, Other Intellectual Property: Royalty payments from University of California San Diego

Travel, Accommodations, Expenses: Merck

Kenneth R. Hess

Travel, Accommodations, Expenses: Angiochem

Razelle Kurzrock

Leadership: CureMatch

Stock and Other Ownership Interests: Actuate Therapeutics, CureMatch

Honoraria: XBiotech, Mayo Clinic Cancer Center, Kaiser Permanente, Health Advances, Wiley, Scripps Translational Research Institute, Defined Health, Roche

Consulting or Advisory Role: Sequenom, Actuate Therapeutics, XBiotech

Research Funding: EMD Serono (Inst), Genentech (Inst), Foundation Medicine (Inst), Pfizer (Inst), Guardant Health (Inst), Sequenom (Inst)

Patents, Royalties, Other Intellectual Property: No relationship to disclose

Travel, Accommodations, Expenses: EMD Serono, Gate-way Research Advisory Committee, Guardant Health, Global Source Ventures, Meyers Consulting, Genentech, Sylvester Cancer Center, CureMatch, Lynx Group, Mayo Clinic Cancer Center, Kaiser Permanente, Cedars-Sinai, MedImmune