Individualized Molecular Profiling for Allocation to Clinical Trials Singapore Study—An Asian Tertiary Cancer Center Experience

Abstract

PURPOSE

Precision oncology has transformed the management of advanced cancers through implementation of advanced molecular profiling technologies to identify increasingly defined subsets of patients and match them to appropriate therapy. We report outcomes of a prospective molecular profiling study in a high-volume Asian tertiary cancer center.

PATIENTS AND METHODS

Patients with advanced cancer were enrolled onto a prospective protocol for genomic profiling, the Individualized Molecular Profiling for Allocation to Clinical Trials Singapore study, at the National Cancer Center Singapore. Primary objective was to identify molecular biomarkers in patient's tumors for allocation to clinical trials. The study commenced in February 2012 and is ongoing, with the results of all patients who underwent multiplex next-generation sequencing (NGS) testing until December 2018 presented here. The results were discussed at a molecular tumor board where recommendations for allocation to biomarker-directed trials or targeted therapies were made.

RESULTS

One thousand fifteen patients were enrolled with a median age of 58 years (range 20-83 years). Most common tumor types were lung adenocarcinoma (26%), colorectal cancer (15%), and breast cancer (12%). A total of 1,064 NGS assays were performed, on fresh tumor tissue for 369 (35%) and archival tumor tissue for 687 (65%) assays. TP53 (39%) alterations were most common, followed by EGFR (21%), KRAS (14%), and PIK3CA (10%). Of 405 NGS assays with potentially actionable alterations, 111 (27%) were allocated to a clinical trial after molecular tumor board and 20 (4.9%) were enrolled on a molecularly matched clinical trial. Gene fusions were detected in 23 of 311 (7%) patients tested, including rare fusions in new tumor types and known fusions in rare tumors.

CONCLUSION

Individualized Molecular Profiling for Allocation to Clinical Trials Singapore demonstrates the feasibility of a prospective broad molecular profiling program in an Asian tertiary cancer center, with the ability to develop and adapt to a dynamic landscape of precision oncology.

INTRODUCTION

The identification of targetable and actionable genomic alterations for therapeutic decision making has transformed the management of advanced cancers in recent decades. Molecular profiling is an important cornerstone of standard clinical practice for many tumor subtypes, such as non–small-cell lung cancer (NSCLC),¹ colorectal cancer (CRC),² and breast cancer.³ Technological advancements with next-generation sequencing (NGS) have heralded a new age resulting in a rapidly expanding list of potential targets—in some instances, tumor agnostic^4,5 and increasing accessibility to such high-throughput sequencing platforms.⁶ In parallel, genomic biomarker–directed clinical trials of targeted and immunotherapeutics have seen significant growth.⁷

CONTEXT

• Key Objective

• Molecular profiling is often regarded as the cornerstone to enabling precision oncology. Here, we present a single-center experience of profiling > 1,000 solid tumors within the setting of an academic phase I unit where 80 early-phase trials were recruiting during the study period.

• Knowledge Generated

• Our study illustrates the substantial interdisciplinary effort required to enable a precision oncology program, where a range of fit-for-purpose laboratory assays including next-generation sequencing panels, immunohistochemistry, and fluorescence in situ hybridization are required. About a third of molecular studies were performed on contemporaneous tissue obtained from an active re-biopsy program. Twenty-seven percent of patients with actionable molecular results were enrolled into trials, although only 4.9% of patients were enrolled into a genomically matched trial.

• Relevance

• Our study highlights the complexities in delivering precision oncology, where key priorities include appropriate timing of testing, choice of assays, integration of molecular tumor boards into routine clinical care, and capture of patient outcomes.

With greater complexity in the mutational landscape both within and across tumor subtypes, the application and implementation of appropriate assays become paramount, so as to delineate molecularly defined patient subsets and match them to the appropriate therapy. Broad molecular screening protocols to facilitate clinical trial enrollment have been previously demonstrated to result in improved therapeutic outcomes.⁸ With growing demand for access to newer and more effective therapies, access to comprehensive profiling and relevant trials remains significant barriers, especially in the context of a rapidly evolving and dynamic landscape.

Here, we report the findings from a prospective molecular profiling study in a high-volume Asian tertiary cancer center, demonstrating the evolution in the breadth and depth of sequencing and testing approaches to facilitate matched therapies. We further highlight the challenges and lessons learnt through implementing such a program for precision oncology.

PATIENTS AND METHODS

Patient Selection and Enrollment

Patients were enrolled onto a prospective protocol for genomic profiling, the Individualized Molecular Profiling for Allocation to Clinical Trials (IMPACT) study (ClinicalTrials.gov identifier: NCT02806388), at the National Cancer Centre Singapore. The primary objective of the study was to identify molecular targets from tumors in patients eligible for early-phase trials. The study commenced in February 2012 and is ongoing, with the results of all patients who underwent NGS testing from study commencement until December 2018 presented here. Patients were recruited, and NGS testing was ordered by their primary treating oncologist or by phase I trial oncologists after referral and assessment for trial enrollment. Patients were recruited at any time during their standard care or after referral to the phase I unit, and molecular profiling was conducted at no additional cost to the patient. Patients with all solid tumors were eligible, with no restrictions on the number of lines of prior therapy, performance status, or organ function. Clinical information was collected from the electronic medical record and a prospective database that was included as part of the study. Data acquisition was locked on December 31, 2018. The study was approved by the SingHealth Centralised Institutional Review Board, and all patients provided written informed consent.

Genomic Analysis

Molecular profiling was conducted at the Division of Pathology, Singapore General Hospital, a laboratory accredited by the College of American Pathologists. Early versions of the study protocol consisted of singleplex molecular testing, either used in combination or alone, immunohistochemistry (IHC), fluorescence in situ hybridization (FISH), and single-gene sequencing (Fig 1). Singleplex assays were conducted at the discretion of the ordering clinician. Multiplex NGS consisted of various targeted panels with each subsequent panel having broader genomic coverage (Appendix Table A1). Successive panels from the Lung Colon Panel onward (representing 86% of assays conducted) contained the same genes within the panel with additional genes in each subsequent panel. Tumor tissue was obtained from archival formalin-fixed paraffin-embedded samples collected during routine clinical care or from fresh frozen tissue collected prospectively from tumor biopsies or surgical resections. Genomic DNA and RNA from tumor were extracted from formalin-fixed paraffin-embedded or fresh frozen tissue. Library was constructed using different panels, and the resulting amplicons were treated to partially digest, phosphorylate, and ligate to ion adapters with barcoding and purified. Quality and concentration of the libraries were determined using the Qubit 2.0 Fluorometer. Emulsion polymerase chain reaction and enrichment of template-positive Ion Sphere Particles, which contained clonally amplified DNA, was conducted using the Ion PGM OneTouch 2 system. Sequencing of this amplified DNA was subsequently performed on the Ion Torrent PGM sequencer. Data were analyzed primarily using Torrent Suite Variant Caller plugin, Ion Reporter software, and in-house analysis pipeline v1.0.0 using Oncomine Reporter with reference genome hg19. All genomic analysis in this study from the use of tumor tissue specimen, analysis pipeline, and database references was developed to detect for somatic changes and was not designed or validated to interrogate for germline changes. Tumor mutation burden (TMB) was calculated based on the size of the panel and number of nonsynonymous mutations and mathematically approximated for 1 Mbp. The results of all molecular profiling, including singleplex assays and NGS, were returned to the ordering clinician.

FIG 1.

Assay utilization over time in the Individualized Molecular Profiling for Allocation to Clinical Trials Singapore study. FISH, fluorescence in situ hybridization; IHC, immunohistochemistry; MMR, mismatch repair; PD-L1, programmed death-ligand 1.

Molecular Tumor Board

All patients and NGS results were discussed at a monthly molecular tumor board (MTB), attended by medical oncologists (treating primary and phase I), radiation oncologists, pathologists (including molecular pathologists), scientists (including bioinformaticians and laboratory scientists), and clinical trial coordinators. Patients were discussed when the molecular profiling results were available—meaning that patients might have been continuing on standard therapies. Following scientific discussion and interpretation of the NGS molecular profiling results, recommendations were made regarding allocation to biomarker-directed clinical trials or alternative targeted or immunotherapies. These recommendations were communicated to the treating primary oncologist, and patients were followed up for outcomes in terms of recruitment to trials. To determine the rate of enrollment to genomically matched phase I clinical trials, we cross-referenced all patients enrolled on phase I trials for the duration of the IMPACT study with all patients with sequencing results from IMPACT before enrollment on the clinical trial. A patient was considered to have matched to a clinical trial if the mechanism of action of the therapy was targeted to a detected alteration from the sequencing assay.

Statistical Analysis

Alterations deemed actionable, for therapeutic targeting with established or investigational drugs, are annotated in Appendix Table A1. Mutation rates were calculated based on their prevalence in the tested samples. Descriptive statistics, including median and range for continuous variables and percentages for categorical variables, were used. Statistical analysis was conducted using GraphPad Prism 7 (GraphPad Software, San Diego, CA).

RESULTS

Description of Sequencing Cohort

One thousand fifteen patients with advanced malignancies were enrolled onto the study and underwent molecular profiling with multiplex NGS testing from February 2012 to December 2018 (Table 1). The median age was 58 years (range 20-83 years), and 569 (56%) patients were female. Most had an Eastern Cooperative Oncology Group performance status of ≤ 2 (77%, with 23% unknown) and were never smokers (63%). The most common tumor types are shown in Figure 2, with the cohort consisting predominantly of lung adenocarcinoma (26%), CRC (15%), and breast cancer (12%).

TABLE 1.

Patient and Assay Characteristics

FIG 2.

Overview of the distribution of tumor types sequenced in the Individualized Molecular Profiling for Allocation to Clinical Trials Singapore cohort.

There were a total of 1,064 NGS assays including patients who had more than one assays performed at serial time points, with molecular profiling performed on fresh tumor tissue in 369 (35%) assays and archival tumor tissue in 687 (65%) assays. Testing was ordered by the primary treating oncologist in 741 (70%) cases, and the phase I team in 343 (30%) cases. The specific multiplex assay and the performance of the assay are shown in Table 1. Of 687 assays performed on archival tissue, the archival tissue was from within the last six months in 265 (39%), between 6 and 12 months prior in 122 (18%), between 1 and 2 years prior in 133 (19%), from > 2 years prior in 130 (19%), and unknown or not recorded in 37 (5%).

Landscape of Tumor Mutational Profiles

The most frequently altered genes and the spectrum of kinase gene fusions detected by multiplex NGS testing are illustrated in Figure 3. In the overall cohort, TP53 (38.5%) alterations were most common, followed by EGFR (20.5%), KRAS (13.8%), and PIK3CA (10.0%) alterations. There were high proportions of TP53 alterations (100%) in head and neck squamous cell carcinoma (n = 8), KRAS alterations (69.4%) in pancreatic adenocarcinoma (n = 36), and EGFR alterations (64.4%) in lung adenocarcinoma (n = 264). The presence of concurrent alterations in selected cohorts is shown in Appendix Figure A1. In EGFR-mutated NSCLC patients, TP53 (54.1%) and PIK3CA (10.6%) co-mutations were most common. In RAS-mutated CRC patients, TP53 (54.8%), APC (22.6%), and PIK3CA (11.3%) mutations were most common. TMB was reported in 299 assays (Oncomine Comprehensive Panel version 1 and version 3), with a median TMB of eight mutations/Mbp (range 0-36 mutations/Mbp). Gene fusions were detected in 23 of 311 (7%) patients tested (Fig 3B).

FIG 3.

Next-generation sequencing findings in the Individualized Molecular Profiling for Allocation to Clinical Trials Singapore cohort (N = 1,064). (A) Frequency of gene alterations across common tumor types. (B) Spectrum of kinase gene fusions in patients with fusion testing (n = 311).

Correlation of Sequencing with Singleplex Molecular Testing

Of patients who had multiplex NGS assay performed, the results of concurrent singleplex assay testing are illustrated in Figure 4A. Programmed death-ligand 1 (PD-L1) IHC (n = 450) was conducted most commonly, followed by PTEN IHC (n = 357), MET IHC (n = 265), and MET FISH (n = 189). Figure 4B shows the concordance between different modalities of testing. For the majority of alterations tested, the results were negative on both NGS and singleplex assay (IHC or FISH)—indicated by negative concordance. There were 231 patients with the results for TMB and PD-L1 IHC, with no correlation between both assays (r² < 0.001, P = .99).

FIG 4.

Singleplex assay testing. (A) Results of IHC and FISH testing. (B) Concordance with NGS results. FISH, fluorescence in situ hybridization; IHC, immunohistochemistry; MMR, mismatch repair; NGS, next-generation sequencing; PD-L1, programmed death-ligand 1.

Clinical Actionability and Allocation to Clinical Trials

Figure 5 illustrates the detection of molecular alterations on NGS and the subsequent outcomes of patients in terms of allocation to clinical trials and subsequent enrollment on genomically matched clinical trials. Of 405 assays with potentially actionable alterations, 111 (27.4%) were allocated to a therapeutic clinical trial of targeted or immune-based therapy after discussion at MTB and 20 (4.9%) patients were then enrolled on a genomically matched clinical trial (Appendix Table A2). Altogether, there were 53 (4.9%) patients who were enrolled on a genomically matched phase I trial after molecular profiling.

FIG 5.

CONSORT diagram of the outcomes of the Individualized Molecular Profiling for Allocation to Clinical Trials Singapore study. Patients with potentially actionable alteration(s) categorized as not discussed or not applicable included patients who had died before the MTB, were still on standard-of-care therapies, or were treated at external institutions. MTB, molecular tumor board; NGS, next-generation sequencing.

Of these 53 patients enrolled on genomically matched phase I clinical trials, 9 (17%) achieved stable disease, 15 (28%) partial response, and 26 (49%) progressive disease as best response. Three (6%) patients had withdrawn consent or did not reach their first assessment of response in view of toxicities.

There were 80 early-phase trials open within the unit during the study period from 2012 to 2018. Of these, 45 (56%) were for multiple or all tumor types, 34 (43%) were molecularly or biomarker selected, 41 (51%) consisted of combination therapies, and 12 (15%) were trials evaluating immunotherapy treatments. The biomarker-selected trials included trials for patients with alterations in genes such as ALK, BRAF, EGFR, ERBB2, FGFR, IDH1, KRAS, MET, NRAS, NTRK, PIK3CA, PTEN, RET, and wtTP53.

Common targetable mutations were also detected in uncommon tumor types (Appendix Table A3), such as EGFR L858R (one patient with tongue squamous cell carcinoma), EGFR exon 19 deletion (two patients with invasive breast carcinoma and pseudomyxoma peritonei), BRAF V600E (two patients with invasive breast carcinoma), and KRAS G12C (two patients with gastric cancer and pancreatic adenocarcinoma).

Identification of New Tumor Types Harboring Known Fusions

In our cohort, numerous cases of known fusions were detected in new tumor types, not previously reported to our knowledge (Table 2). Notably, this includes both rare fusions in common tumor types such as TBL1XR1-PIK3CA fusions in two patients with NSCLC and known fusions in uncommon tumor types such as an SLC45A3-ERG fusion in a patient with paraganglioma and an EIF3E-RSPO2 fusion in a patient with cervical round cell sarcoma. None of these patients, however, went on to receive targeted therapies for their fusions.

TABLE 2.

Rare or Uncommon Fusions Detected in Individualized Molecular Profiling for Allocation to Clinical Trials

DISCUSSION

Our study demonstrates the feasibility of implementing broad multiplex NGS testing with subsequent MTB discussion in a high-volume Asian tertiary cancer center, and the profile of our cohort reveals interesting insights into our phase I referral base. Overall, in our patient cohort, the proportions of tumor types profiled were consistent with cancer prevalence in Singapore.¹¹ Patients were generally younger and of good performance status, as might be expected with referrals for phase I trials. Although only 30% of assays were ordered in the phase I clinic, many patients might have been consented during standard care by primary oncologists who were also members of the phase I team. Furthermore, molecular profiling was not restricted to the phase I clinic to allow for responsive testing and aid in the timely allocation onto early-phase trials. Despite the potential for selection bias in the patients selected for molecular profiling with NGS, the overall molecular profile was also reflective of the known molecular characteristics of common tumor types in Singapore. In particular, this includes a high rate of EGFR mutations in patients with lung adenocarcinoma in an Asian compared with Caucasian population.¹² Frequencies of KRAS mutations in CRC and pancreatic cancer¹³ and PIK3CA mutations in breast cancer¹⁴ were consistent with the literature for an Asian population. Concordance of singleplex assay testing with NGS generally displayed negative concordance. There were low rates of positive concordance; however, singleplex assays were conducted predominantly in tumor types for which singleplex assay testing is not routinely performed or validated, limiting interpretation of these findings. Additionally, the TMB findings should be interpreted with caution. Key parameters in TMB estimation using the Oncomine Comprehensive Panel v3, such as the relatively small size of the panel, may influence its accuracy significantly.¹⁵

Notably, 35% of patients were agreeable to a fresh biopsy sample for the purpose of molecular profiling, providing the most contemporaneous and clinically relevant sample. This highlights the challenges that arise with the implementation of broad molecular profiling, particularly the timing of testing whether before progression on current therapy to guide future therapeutic options or at the time of progression to decide on the next line of therapy. The former option allows for more focused therapy planning especially in regard to a patient’s fitness for clinical trials but may overlook important resistance alterations that develop as tumors evolve. Similarly, the use of archival versus fresh tumor tissue is an important parameter with wide ranging implications on potential therapeutic options and efficacy.

About a quarter of patients with potentially actionable alterations were allocated to a clinical trial, of which only 5% were genomically matched, despite running up to 80 trials during this period of time. A major limitation of our study was the lack of a prospectively applied formal framework to categorize actionable alterations and clearly define therapeutic tractability stratified by the strength of the scientific literature. There are numerous guidelines such as ESMO Scale for Clinical Actionability of Molecular Targets¹⁶ and knowledge bases such as OncoKB¹⁷ that have been developed in an effort to standardize reporting and interpretation of genomic variants. These are currently being incorporated, as patients continue to be enrolled in the IMPACT Singapore study. Nevertheless, despite the seemingly low proportion of patients enrolled to genomically matched trials, this was comparable with other international centers,¹⁸ which have reported that molecular screening programs result in only 4%-11% of patients matching to biomarker-enriched clinical trials (Appendix Table A4).^20,21,24 This reflects a number of important factors such as the availability of specific profiling assays, emergence of immuno-oncology therapeutic options vs targeted therapies over time, and access to clinical trials at the time when testing was performed. As we transition from single-gene assays to NGS, there is a crucial need for careful consideration of the relative performance (sensitivity and specificity) of each biomarker technique. Importantly, in our cohort, we demonstrate how incorporation of fusion detection can yield both common and rare fusions in new tumor types not previously reported. Continual technological advancements further emphasize the need for flexibility to adapt to a dynamic early-phase trials landscape. As the recently reported ProfiLER trial from France forewarns, large-scale implementation with broad inclusion criteria can yield disappointing outcomes.²⁵

Our study further not only underscores how delivering on the promise of precision oncology cannot be solely dependent on molecular profiling alone but also encompasses two additional crucial aspects: (1) the ability to screen a large, trial-eligible patient population and (2) having access to relevant novel therapies. Stringent eligibility criteria for phase I trials, dynamic windows for slots, and declining physical fitness of our cohort of heavily pretreated patients with advanced-stage cancer all likely contribute to only a minority of patients enrolled on trials. There was significant heterogeneity in the NGS panels used; however, common cancer driver genes such as EGFR and KRAS were covered in all patients. Our study demonstrates the feasibility of a prospective broad molecular profiling program in an Asian tertiary cancer center, with the ability to develop and adapt to a dynamic therapeutic and biomarker landscape. With the increasing enthusiasm for molecular profiling and rapid expansion in early-phase trials in regions such as the Asia Pacific, careful integration of precision oncology programs into routine clinical practice is crucial.²⁶ Particularly, in resource-limited settings, careful consideration of patient selection and timing and breadth of testing should be undertaken in conjunction with available therapeutics, which may be in the form of locally available clinical trial or access programs, with prospective capture of outcomes.

Appendix

FIG A1.

Concurrent alterations in selected cohorts within the Individualized Molecular Profiling for Allocation to Clinical Trials Singapore study. (A) EGFR-mutated non–small-cell lung cancer. (B) RAS-mutant colorectal cancer.

TABLE A1.

Genes Tested by NGS-Targeted Panel Assays

TABLE A2.

Patients Enrolled on a Genomically Matched Early-Phase Trial

TABLE A3.

Targetable Mutations Detected in Uncommon Tumor Types

TABLE A4.

Enrollment of Patients Onto Genomically Matched Treatments Across Selected Previously Published Studies

AUTHOR CONTRIBUTIONS

Conception and design: Aaron C. Tan, Tira J. Tan, Matthew C. H. Ng, Tony K. H. Lim, Daniel S. W. Tan

Provision of study materials or patients: Amanda O. L. Seet, Aaron C. Tan, Tira J. Tan, David W. M. Tai, Bien Soo Tan, Tse Hui Lim, Rebecca Alexandra Dent, Yoon-Sim Yap, Iain B. H. Tan, Su Pin Choo, Elaine H. Lim, N. Gopalakrishna Iyer, Eng Huat Tan, Tony K. H. Lim, Daniel S. W. Tan

Collection and assembly of data: Amanda O. L. Seet, Aaron C. Tan, Tira J. Tan, David W. M. Tai, Justina Y. C. Lam, Gek San Tan, Apoorva Gogna, Bien Soo Tan, Angela Takano, Alvin Lim, Tse Hui Lim, Soon Thye Lim, Rebecca Alexandra Dent, Mei Kim Ang, Su Pin Choo, Chee Keong Toh, Elaine H. Lim, Wan Teck Lim, Daniel S. W. Tan

Data analysis and interpretation: Amanda O. L. Seet, Aaron C. Tan, Tira J. Tan, Matthew C. H. Ng, Gek San Tan, Chow Wei Too, Alvin Lim, Soon Thye Lim, Yoon-Sim Yap, Iain B. H. Tan, Mohamad Farid, Anders Jacobsen Skanderup, N. Gopalakrishna Iyer, Wan Teck Lim, Tony K. H. Lim, Daniel S. W. Tan

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 the authors of this manuscript. All relationships are considered compensated unless otherwise noted. 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 ascopubs.org/po/author-center.

Open Payments is a public database containing information reported by companies about payments made to US-licensed physicians (Open Payments).