Referrals to a Phase I Clinic and Trial Enrollment in the Molecular Screening Era

Molecular profiling of patients with advanced cancers can identify specific genomic alterations to inform decisions about investigational treatments. This article describes the clinic experience and clinical impact of a molecular profiling program at the Princess Margaret Cancer Center, in Canada, and assesses factors associated with subsequent early phase clinical trial enrollment.

Abstract

Background.

Enrichment of patients based on molecular biomarkers is increasingly used in early phase clinical trials. Molecular profiling of patients with advanced cancers can identify specific genomic alterations to inform decisions about investigational treatment(s). Our aim was to evaluate the outcomes of new patient referrals to a large academic solid tumor phase I clinical trial program after the implementation of molecular profiling.

Materials and Methods.

Retrospective chart review of all new referrals to the Princess Margaret Cancer Centre (PM) phase I clinic from May 2012 to December 2014. Molecular profiling using either MALDI‐TOF hotspot mutation genotyping or targeted panel DNA sequencing was performed for patients at PM or community hospitals through the institutional IMPACT/COMPACT trials.

Results.

A total of 971 new patient referrals were included for this analysis. Twenty‐seven percent of referrals assessed in clinic were subsequently enrolled in phase I trials. Of all new referrals, 41% had prior molecular profiling, of whom 11% (n = 42) were enrolled in genotype‐matched trials. Patients with prior molecular profiling were younger, more heavily pretreated, and had more favorable Princess Margaret Hospital Index (PMHI) scores. Eastern Cooperative Oncology Group (ECOG) performance status 0–1 (p = .002), internal referrals within PM (p = .002), and PMHI (p ≤ .001) were independently associated with successful trial enrollment in multivariable analysis.

Conclusion.

Although nearly half of new patients referred to a phase I clinic had prior molecular profiling, the proportion subsequently enrolled into clinical trials was low. Prior molecular profiling was not an independent predictor of clinical trial enrollment.

Implications for Practice.

The landscape of oncology drug development is evolving alongside technological advancements. Recently, large academic medical centers have implemented clinical sequencing protocols to identify patients with actionable genomic alterations to enroll in therapeutic clinical trials. This study evaluates patient referral and enrollment patterns in a large academic phase I clinical trials program following the implementation of a molecular profiling program. Performance status and referral from a physician within the institution were associated with successful trial enrollment, whereas prior molecular profiling was not an independent predictor.

Introduction

There has been a shift in oncology drug development over the last couple of decades, from the development of cytotoxic chemotherapy to molecular targeted agents, and more recently immunotherapy. Traditional first‐in‐human phase I clinical trials with novel cancer drug therapies enroll patients with advanced cancers, agnostic of tumor type, following progression on all standard treatments, who demonstrate adequate organ function and good performance status [1]. The primary objectives of these early phase studies was to assess the safety of novel agents and identify the maximum tolerated dose, with the assumption that the risk for toxicity and clinical benefit increases with escalating doses. Low objective responses rates between 4%–10% in unselected patients have been reported [1], [2].

In the last decade, large‐scale DNA sequencing projects have improved our understanding of cancer biology [3]. With widespread availability of bench‐top sequencing platforms in clinical laboratories, tumor tissues from patients with advanced cancers increasingly undergo molecular profiling to identify “actionable” driver mutations that may benefit from genotype‐matched targeted treatment. Enrichment trial designs restrict enrollment during initial dose finding and/or subsequent dose expansion at the recommended phase II dose to patients whose cancers harbor molecular specific alterations that are predicted to be most likely to benefit from novel targeted drug therapies. There are several examples, such as the BRAF inhibitor vemurafenib for BRAF V600E/K mutant melanoma or the ALK inhibitor crizotinib for ALK‐translocated non‐small cell lung cancer, where robust early efficacy results in phase I testing restricted to biomarker‐selected patients contributed to accelerated clinical development and regulatory agency approval [4], [5]. The introduction of other innovative early phase clinical trial designs, such as “basket trials” that include patients with specific molecular alterations irrespective of tumor histology or “umbrella trials” that include multiple drug treatments for patients with a particular tumor type based upon specific molecular alterations, has further increased the demand for patients with advanced cancers to undergo molecular profiling of their tumor specimens to inform decisions about investigational treatment. Recent reviews of contemporary phase I oncology trials suggest a higher overall response rate approximating 20% and a greater probability of clinical benefit when phase I studies are based on enrichment strategies either through identification of specific histological characteristics, biomarker, or both [6], [7], [8].

The impact of this change in early phase clinical trial landscape on patterns of new patient referral and enrollment at large academic medical centers with specialized early phase clinical trial programs is largely unknown. The phase I Drug Development Program (DDP) at the Princess Margaret Cancer Center (PM) is the largest early phase clinical trial program in Canada. We previously reported our phase I referral and enrollment process in the years 2000–2005, providing insight into outcomes of patients referred to a phase I clinic and identifying barriers to trial enrollment in an era of “traditional” drug development [9]. Our goal in this analysis was to review new patients referred to the PM phase I DDP following the introduction of an institutional molecular profiling program and to assess factors associated with subsequent early phase clinical trial enrollment.

Materials and Methods

New Patient Referrals

Prior to accepting a new patient referral for a scheduled consultation appointment at a weekly ambulatory clinic, medical records, including clinical notes, radiology reports, and recent laboratory investigations, were reviewed by a PM phase I DDP principal investigator. Referrals were declined and patients were not seen in phase I clinic because of poor performance status, inadequate organ function, ongoing systemic treatment, or lack of available clinical trials. All consecutive, unique new patients assessed at the PM phase I clinic from May 2012, following the introduction of a prospective institutional molecular profiling program, until December 2014 were retrospectively reviewed. Patients who had previously been assessed in the phase I clinic before May 2012 were excluded.

Molecular Profiling

The Integrated Molecular Profiling in Advanced Cancers Trial (IMPACT) and Community Molecular Profiling in Advanced Cancers Trial (COMPACT) trials provided molecular profiling for patients treated at PM or local community hospitals with advanced solid tumors, who were ≥18 years, who had Eastern Cooperative Oncology Group (ECOG) performance status ≤1, and who had available formalin‐fixed paraffin‐embedded archival tumor tissue [NCT01505400]. Patients were consented for testing by their treating PM medical oncologist or at a dedicated weekly clinic for patients treated at other hospitals. Molecular profiling included either a matrix‐assisted laser desorption/ionization time‐of‐flight mass spectrometry (MALDI‐TOF) hotspot panel that included 23 genes and 279 hotspots or a commercial, small targeted next‐generation sequencing panel covering regions of 48–50 genes (testing details provided in [10]). Molecular profiling results were included in the patient's electronic medical record and returned to the treating oncologist, along with regular summary tables of testing results and mutation‐specific clinical trial listings available at PM [10].

Retrospective Data Collection

Medical records of new patients assessed in the PM phase I DDP ambulatory clinic were reviewed retrospectively. The following data were captured: demographics (age, gender, postal code), past medical history (tumor site, date of diagnosis, patient comorbidities), molecular profiling results, whether the patient was offered participation in one or more clinical trials at their phase I clinic appointment, and reasons for subsequent trial enrollment or nonenrollment. Delay to trial enrollment was defined as greater than 1 month from the date of phase I consultation appointment. Clinical parameters such as ECOG performance status (PS), Princess Margaret Hospital Index (PMHI; 1 point for each of the following: ECOG ≥1, albumin <lower limit of normal [LLN], and >2 metastatic sites) [11], and Royal Marsden Hospital Prognostic Index (RMHPI; 1 point for each of the following: albumin <35 g/L, lactase dehydrogenase [LDH] >upper limit of normal [ULN] and >2 metastatic sites) [12] were extracted from review of clinical notes and/or laboratory tests performed at the date of the initial consultation.

Distance to PM was calculated using the first three digits of the patient's postal code. The postal codes were translated to longitudinal and latitudinal measures using an online geocoder (https://www.doogal.co.uk/BatchGeocoding.php), which was then used to calculate the distance in kilometers to PM using an excel formula: = ACOS(COS[RADIANS(90 − Lat1)] * COS[RADIANS(90 − Lat2)] + SIN[RADIANS(90 − Lat1)] * SIN[RADIANS(90 − Lat2)] * COS[RADIANS(Long1 − LONG2)]) * 6,371; where Lat1 and Long1 are the latitude and longitude of the patient's postal code and Lat2 and Long2 are the latitude and longitude of PM's postal code. (http://bluemm.blogspot.com/2007/01/excel‐formula‐to‐calculate‐distance.html). Reasons for trial nonentry were categorized into one of the following subclassifications outlined in the consort diagram (Fig. 1). Because the reasons for trial nonentry were often multifactorial, the patient was assigned to the single best category based on the most limiting feature toward trial participation stated in the phase I clinic consult notes. If the reason for nonentry was not explicitly stated, judgement was made by the first and last authors (M.R. and P.B.) based on the clinical flow of the consultation as well as prioritization of patient‐centered factors, such as refusal, stabilized disease, and consideration of other treatment.

Figure 1.

Consort diagram of flow of patient from clinic visit to trial enrollment.

Inclusion criteria of patients into the statistical analysis included all new referrals to the phase I clinic from May 2012 to December 2014. Patients who were previously assessed by the PM phase I program were excluded from this analysis. This study was approved by the University Health Network Research Ethics Board (no. 15‐9073).

Statistical Analysis

Descriptive statistics (mean, median) were used for continuously scaled measurements, whereas qualitative data pertaining to clinical trial enrollment was converted into binary data in order to facilitate statistical analysis. Chi‐square analysis was performed to examine the association between two categorical variables. A t test was carried out for comparison of continuous variables. Univariate and multivariate logistic analysis were performed to identify predictors of successful trial enrollment. The following covariates were included in the analysis: gender, age, prior molecular profiling (yes vs. no), PMHI (0, 1, 2, and 3), RMHPI (0, 1, 2 and 3), number of prior lines of treatment, distance to PM (<50 km vs. >50 km), ECOG PS (0–1 vs. 2–3), disease site (gastrointestinal [GI] vs. non‐GI), and referring physician (internal vs. external). Patients with missing values for the covariates were excluded from the analysis. All statistical tests were two‐sided, and p < .05 was considered statistically significant.

Results

Study Population

In total, 971 unique new patient referrals assessed in the PM phase I DDP clinic between May 2012 and December 2014 were identified (Table 1). Of these referrals, approximately 55% were referred from internal physicians at PM, and 45% were referred from an external institution. The median age of referred patients was 60 years (range, 18–84), 56% were female, and the median distance travelled to PM was 27 km (range, 0–3,363 km). The most common tumor types were GI (31%), pancreatobiliary (14%), and gynecological (12%). The median number of lines of prior systemic therapy was two (range, 0–6), and 5% of patients had received no prior systemic therapy. Patients were generally fit, and 85% were ECOG PS 1 or better and had favorable phase I prognostic scores (PMHI 0/1 52% and RMHPI 0/1 54%).

Table 1. Patient demographics.

^aOther: chordoma (2), hemangioendothelioma (1), lymphoepithelioma (2), Merkel cell (1), mesothelioma (11), neuroendocrine tumor (8), pheochromocytoma (1), Sertoli‐Leydig sex cord (1), thyroid (9), urachal (1), pseudomyxoma peritonei (2), adrenal (3), lymphoma (2).

^bPrincess Margaret Hospital Index (0–3); 1 point for each of the following: >2 metastatic site, albumin >35 g/L, Eastern Cooperative Oncology Group >1.

^cRoyal Marsden Prognostic Index (0–3); 1 point for each of the following: >2 metastatic site, albumin >35 g/L, lactase dehydrogenase >upper limit of normal.

Molecularly Profiled Subgroup

Of all new referrals to the phase I clinic, 396 patients (41%) had prior molecular profiling through IMPACT or COMPACT. 132 patients who had molecular profiling, were enrolled on a trial but only 42 of these patients were enrolled in a genotype‐matched trial. This represents 11% (42 of 396) of all the new patient referrals with prior molecular profiling or 4% (42 of 971) of all new referrals to the phase I clinic. Patients with prior molecular profiling were younger (p = .013), more heavily pretreated with systemic chemotherapy (p = .014), lived closer to PM (p = .025), and had more favorable phase I prognostic scores as compared with patients without prior molecular profiling (p ≤ .001; Table 2). Patients with prior molecular profiling were also more likely to be offered clinical trials and subsequently enrolled when compared with patients without prior molecular profiling (p = .001; Table 2).

Table 2. Differences between patients with prior molecular profiling versus no prior molecular profiling.

Abbreviations: PM, Princess Margaret; PMHI, Princess Margaret Hospital Index.

Phase I Assessment Outcomes

Figure 1 demonstrates the sequential flow of patients from the point of initial consult and informed consent to clinical trial enrollment. Of 971 patients, there were 411 (42%) patients that were not offered a trial. The most common reasons were poor performance status (n = 111, 27% of those not offered trial), comorbid medical condition(s) (n = 69, 17%), and ongoing anticancer treatment (n = 33, 8%). A total of 560 (58%) were offered a trial at the initial consult. Of the 560 patients who were offered a trial, 160 (29%) patients did not enroll, most often because of patient refusal after reviewing the informed consent (n = 66, 41%). Reasons for patient refusal included desire to preserve quality of life, pursuit of other treatment, travel distance to PM, uncertainty of treatment benefit, and concerns about treatment‐related toxicity.

Of the 400 patients who completed the informed consent process for a clinical trial, 119 (30%) were subsequently determined to be ineligible and were considered “screen failures.” Common reasons for screen failure included negative result in a trial‐mandated biomarker prescreening (n = 45, 38% of screen failures), unacceptable laboratory parameter(s) (n = 21, 18%), and patient deterioration (n = 18, 15%). Sixteen (2%) screened‐eligible patients did not start trial because of preference for other treatment(s) (n = 6), patient refusal (n = 5), rapid deterioration (n = 2), trial unavailability (n = 2), or lost to follow‐up (n = 1).

A total of 265 (27%) patients were subsequently enrolled in a clinical trial. Of patients enrolled in clinical trials, 73 (28% of those enrolled in trials) patients were delayed more than 1 month from their initial consultation appointment to their trial start date for various reasons, including scheduling delays (n = 26, 10%), patients considering other treatment options (n = 14, 5%), or patients were still receiving other treatment(s) prior to clinical trial enrollment (n = 8, 3%).

Predictors of Trial Enrollment

Factors associated with successful trial enrollment on univariate analysis include age (p = .010), prior IMPACT/COMPACT molecular profiling (p < .001), PMHI score (p < .001), ECOG 0–1 (p < .001), and internal referrals (p < .001). On multivariate analysis, only PMHI (p ≤ 0.001), ECOG status (≤1 vs ≥2; p < .002), and internal referrals (vs. external; p = .002) were independently associated with successful trial enrollment (Table 3).

Table 3. Univariate and multivariable analysis of factors associated with trial enrollment.

^aNote: an OR of >1 indicates a higher likelihood of trial enrolment in the presence of listed factors vs. reference as indicated in brackets.

Abbreviations: CI, confidence interval; ECOG, Eastern Cooperative Oncology Group; GI, gastrointestinal; OR, odds ratio; PM, Princess Margaret; PMHI, Princess Margaret Hospital Index; PS, performance status; RMHPI; Royal Marsden Hospital prognostic index.

Discussion

In our single institutional cohort in which molecular profiling was available on 41% of patients with advanced solid tumors who were referred to a large academic phase I program, only 27% of new referrals assessed were successfully enrolled in a therapeutic clinical trial. Gustav Roussy enrolled 24%–39% of new patients assessed from 2008–2016. [13] Both Guy's and St Thomas’ hospital in the U.K. and the unit at MD Anderson, U.S., report higher rates of 40% and 52% respectively [14], [15]. Previously, our group reported a 29% enrollment rate in a retrospective review of 667 new referrals assessed over a 5 year period (2000–2005). Although the proportion of patients enrolled are similar, the absolute numbers (current analysis of 265 over 2.5 years compared with previous, 197 over 5.5 years) has increased [9], [16]. This finding is not surprising, given that the number of clinical trials available through PM phase I program has expanded over the last decade and the introduction of an institutional molecular profiling program led to increased visibility of the trials offered through the phase I program.

In spite of triage of all referrals by a phase I principal investigator to evaluate eligibility prior to a consultation appointment, 42% (n = 411) of new patients who were phase I referrals were not offered a clinical trial because of poor performance status, comorbid medical condition(s) or inadequate organ function(s), and ongoing anticancer treatment. Although phase I investigators are best positioned to assess potential trial eligibility, limited clinical resources could be more efficiently allocated on patients most likely to participate in a clinical trial. Referrals of patients with poor performance status, multiple comorbid conditions, rapidly progressive disease, and significant symptoms can and should be avoided. In our cohort, only 2% of new referrals were not offered a clinical trial because of lack of interest. Similar to our prior analysis, new patient referrals from internal physicians at PM were more likely to be offered participation in a trial than referrals from outside physicians that may be related to improved knowledge of trial eligibility considerations and available clinical trials within the institution. Deterioration of patient condition between the last assessment of the referring physician and the phase I new patient consultation appointment does not explain the high proportion of patients deemed unsuitable for clinical trial participation, as the vast majority of referrals were assessed within two weeks of receiving a consultation request. Referring physicians may view a negative assessment by a phase I program as an opportunity to reaffirm to a patient with poor performance status that there are no further treatment options available following progression on standard therapy.

It is noteworthy that 71% of patients offered a clinical trial completed the consent process and initiated screening, compared with 39% in our prior analysis [9]. This likely reflects the increased availability of clinical trials and greater efficiencies in the consent and screening process with a larger phase I trials program infrastructure. The most frequent reason why patients offered a trial did not subsequently screen was because of patient refusal. The main reasons for declining trial participation despite being assessed as eligible are consistent with previous literature and include a lack of interest in trial participation, preference for quality of life, and pursuit of other treatment [17].

Despite the availability of molecular profiling for patients receiving standard treatment at PM and local community hospital, only 41% of new patient referrals had prior molecular profiling before their phase I assessment. Reasons why patients with prior profiling were not offered a trial included limited targeted hotspot panel testing, which failed to identify “actionable” mutations for clinical trial matching, lack of available spots at the time of assessment for a genotype‐matched trial, and trial‐specific exclusion criteria that precluded trial participation. Interpretative challenges and lack of guidance to support clinical action has also been shown to contribute to low clinical utility of molecular profiling [18], [19]. This may be addressed, in part, through decision support teams, such as The Precision Oncology Decision Support system as described by The University of Texas MD Anderson Cancer Center [20]. At PM, our experience with the IMPACT/COMPACT study of 1,893 patients was previously reported [10]. Of the 1,640 patients with molecular profiling results, 15% were enrolled onto therapeutic clinical trials and 5% were treated on genotype‐matched trials. We facilitated interpretation of returned molecular profiling reports through multidisciplinary tumor board discussions, physician‐directed e‐mail alerts with genotype‐matched trial listings, and individual physician summaries of profiling results. In spite of this, the rate of genotype‐matched clinical trial enrolment was low [10]. Several large academic cancer centers have similarly implemented clinical sequencing protocols to guide management [10], [21], [22], [23]. Memorial Sloan Kettering Cancer Center reported their experience of sequencing tumors from more than 10,000 patients [21]. In a subset analysis of the clinical utility of sequencing on the MSK‐IMPACT assay, of the 5,009 patients tested 1 year before analysis, 11% enrolled on a genotype‐matched trial [21]. In the current analysis, although prior molecular profiling was associated with a higher likelihood of being offered and subsequent enrollment onto a clinical trial, this was not an independent predictor in multivariate analysis.

Over the last 5 years, there has been another paradigm shift, with the focus of drug development turned to immunotherapy. Immune checkpoint inhibitors targeting the cytotoxic T‐lymphocyte‐associated protein 4, programmed cell death protein 1, or its ligand (programmed death ligand 1) have entered the clinic and are now standard of care treatment for a variety of tumor types such as melanoma, non‐small cell lung cancer, and renal cell carcinoma, among others. The ability of immunotherapies to induce durable clinical responses with favorable toxicity profiles has led to a wave of next‐generation combinatory immune‐oncology trials. Despite the challenges in biomarker development, genomic factors indicative of tumor mutational burden such as nonsynonymous somatic mutational load, microsatellite instability, and defects in DNA polymerase epsilon have been shown to play a role in predicting response to immunotherapy, making molecular profiling of continued relevance [24]. As research in immunotherapy continues to advance, we anticipate a shift in focus away from molecularly targeted drugs and patient enrollment onto genotype‐matched clinical trials to immunotherapy trials within which novel biomarkers such as gene expression profiling, T‐cell clonal diversity, peripheral blood immunophenotyping, and systemic cytokines continue to be developed.

There are several limitations of our study. This is a single institution, retrospective review that is subject to information bias and residual confounding due to incomplete or unknown covariate information. There were few immunotherapy phase I trials available during the study period, which may influence referral patterns and the impact of prior molecular testing on subsequent phase I trial enrollment. Finally, there were differences in the molecular profiling platforms used during the study period that may affect detection of “druggable” genetic aberrations.

Conclusion

We report here our phase I clinic experience and the clinical impact of a molecular profiling program. Although nearly half of referred patients had prior molecular profiling, this did not predict for successful enrollment onto a phase I trial, and the overall rate of genotype‐matched trial enrollment was low. Through this analysis, we have identified and described factors and barriers contributing to non‐trial participation. With rapid advances in sequencing methods and an expanding repertoire of novel agents including that of immune‐oncology drugs, longitudinal follow‐up is necessary to continually evaluate the clinical utility of genomic profiling and phase I patient outcomes.

Author Contributions

Conception/design: Tira Tan, Michael Rheaume, Lillian L. Siu, Philippe L. Bedard

Provision of study material or patients: Anna Spreafico, Aaron R. Hansen, Albiruni R. A. Razak, Lillian L. Siu, Philippe L. Bedard

Collection and/or assembly of data: Tira Tan, Michael Rheaume, Lisa Wang, Helen Chow, Philippe L. Bedard

Data analysis and interpretation: Tira Tan, Michael Rheaume, Lisa Wang, Lillian L. Siu, Philippe L. Bedard

Manuscript writing: Tira Tan, Michael Rheaume, Lisa Wang, Helen Chow, Anna Spreafico, Aaron R. Hansen, Albiruni R.A. Razak, Lillian L. Siu, Philippe L. Bedard

Final approval of manuscript: Tira Tan, Michael Rheaume, Lisa Wang, Helen Chow, Anna Spreafico, Aaron R. Hansen, Albiruni R.A. Razak, Lillian L. Siu, Philippe L. Bedard

Disclosures

Anna Spreafico: Merck, Bristol‐Myers Squibb, Novartis, Oncorus (C/A), Novartis, Bristol‐Myers Squibb, Symphogen AstraZeneca/Medimmune, Merck, Bayer, Surface Oncology, Northern Biologics, Janssen Oncology/Johnson & Johnson (RF); Aaron R. Hansen: Genentech/Roche, Merck, GlaxoSmithKline, Bristol‐Myers Squibb, Novartis, Boston Biomedical, Boehringer Ingelheim, AstraZeneca, Medimmune (C/A, RF); Albiruni R.A. Razak: Boehringer Ingelheim Consulting (H), Eli Lilly & Co., Merck, Boehringer Ingelheim (C/A), CASI Pharmaceuticals, Boehringer Ingelheim, Eli Lilly & Co., Novartis, Decipher‐a, Karyopharm Therapeutics, Pfizer, Roche/Genentech, Boston Biomedical, Bristol‐Myers Squibb, MedImmune, Amgen, GlaxoSmithKline, Blueprint Medicines, Merck, Abbvie, Adaptimmune (RF); Lillian L. Siu: Merck, Pfizer, Celgene, AstraZeneca/Medimmune, Morphosys, Roche, GeneSeeq, Loxo, Oncorus, Symphogen (C/A), Novartis, Bristol‐Myers Squibb, Pfizer, Boerhinger‐Ingelheim, Regeneron, GlaxoSmithKline, Roche/Genentech, Karyopharm, AstraZeneca/Medimmune, Merck, Celgene, Astellas, Bayer, Abbvie, Amgen, Symphogen, Intensity Therapeutics (RF‐Clinical Trials), Agios (OI‐spouse); Philippe L. Bedard: Bristol‐Myers Squibb, Pfizer, Sanofi (C/A), Bristol‐Myers Squibb, Sanofi, Novartis, GlaxoSmithKline, AstraZeneca, Merck, Seattle Genetics, Nektar, Immunomedics, Mersana, Servier (RF). The other authors indicated no financial relationships.

(C/A) Consulting/advisory relationship; (RF) Research funding; (E) Employment; (ET) Expert testimony; (H) Honoraria received; (OI) Ownership interests; (IP) Intellectual property rights/inventor/patent holder; (SAB) Scientific advisory board