Lack of Availability and Efficacy of Phase I and Basket Trials for Patients With Gastrointestinal Cancers

Abstract

In the modern era of targeted and immune-based therapies, investigator and patient expectations of availability and efficacy in phase I trials have increased. We assessed availability of, and benefit from, early drug development trials, specifically in patients with gastrointestinal cancers. We reviewed computerized referral records of the Early Drug Development Service at our institution to identify patients internally referred from our Gastrointestinal Oncology Service in calendar year 2018. End points were treatment on a trial, 3- and 6-month progression-free survival (PFS), and any tumor shrinkage. Of 394 gastrointestinal cancer patients referred in 2018, 54 enrolled on a trial and 53 (13.5%) were treated (1 withdrew before treatment): 34 on immune-based and 19 on targeted (3 to phase II basket) studies. None of the 52 patients who had exhausted standard therapy achieved 6-month PFS, two (3.8%) met 3-month PFS with tumor growth below Response Evaluation Criteria in Solid Tumors progression at 3 months, and both came off study for progression at 4 months. One patient who was to receive an irinotecan-based regimen as standard therapy instead received irinotecan plus an investigational targeted agent and remained stable for 8 months. No patients achieved any degree of tumor shrinkage. The most common reasons for nonaccrual were lack of available protocol treatment openings and failure to meet eligibility criteria for specific trials. Thus, availability and benefit from investigational treatment in this treatment-refractory gastrointestinal cancer patient population was extremely modest. Expectations regarding both availability and efficacy of phase I investigational therapy in gastrointestinal cancer patients likely exceed what our experience suggests.

Patients with cancer who have exhausted standard treatments often seek phase I drug development or phase II basket studies, with the hope that this will result in therapeutic benefit (1,2). Historically, the purpose of a phase I trial has been to establish the maximum tolerable dose, the recommended phase II dose, and the dose-limiting and nondose- limiting toxicities of the drug or regimen under study. Even the expansion cohorts of traditional phase I trials, in which an additional number of patients were treated with what had been identified as the recommended phase II dose, were typically stated to be for better characterization of the tolerability and toxicities of that dose. Efficacy has not been a primary purpose or end point of a phase I trial, although efficacy and evidence of clinical benefit have historically occurred, albeit infrequently, in some but not all phase I trials (3,4). The lack of focus on efficacy was, and still is, often emphasized in both verbal discussion of and the written consent forms for such trials. Although altruism may be a contributing factor in some patients’ decisions to seek such trials, patient misconception in terms of the unrealistically high expectation that receipt of trial drugs will result in direct benefit has long been recognized (5,6).

Approximately a decade ago, two important avenues of investigation into the management of metastatic cancer—targeted therapy and immune checkpoint inhibitors—began to show clinically significant progress. Phase I studies of therapies targeting a particular mutation or fusion began to be structured to screen patients and make the presence of the specified target a prerequisite for trial eligibility, thereby enriching the study population with patients likely to respond if the mechanistic hypothesis of the drug was correct (7,8). This was greatly facilitated by the concurrent advent of the ability to accurately identify these mutations from archived, paraffin-embedded tissue and the expanded use of next- generation sequencing (NGS), first in investigational settings and then in more routine practice (9,10). Immune checkpoint inhibitors represented a second transformative approach to cancer treatment. Many tumors previously thought to be refractory to oncologic management, such as melanoma, renal cell cancer, and non–small cell lung cancer, were suddenly transformed into immune-responsive diseases, with durable benefit in some patients (11,12).

Unfortunately, neither of these approaches has thus far demonstrated widespread transformative impact on common tumors of gastrointestinal (GI) origin, such as colorectal or pancreas cancers (with the noted exception of the subset with microsatellite instability). Yet, the excitement engendered by progress in other cancers has led to apparent increased expectations in GI oncology patients, fueling enthusiasm for patients and their doctors to seek out early drug development clinical trials when standard therapeutic options have failed. We recognized, however, that even in the setting of a large academic center with a robust, cutting- edge drug development program, a substantial number of patients who sought investigational agents were unable, for one reason or another, to obtain them, and with GI patients distributed across many different trials, overall efficacy specifically within the GI population was often difficult to discern. We therefore explored the recent levels of availability and efficacy of early drug development trials at our institution for patients with tumors of the GI tract.

To do this, we obtained institutional review board permission to review the electronic referral records of our Early Drug Development Service, inclusive of the developmental therapy clinic and the investigational immunotherapy clinic, to identify those patients who had been referred for a phase I or phase II basket trial from our GI Oncology Service in calendar year 2018 and to review their electronic medical records. Referral information was generated from data visualization software (Tableau, Seattle, WA) specifically designed to track internal patient referrals to the Early Drug Development Service with the capability to populate cohorts of patients based on disease type. Patients who for any reason were not formally referred by their treating Gl oncologist were not captured and therefore were not considered in this analysis.

The end points we chose to investigate were, first, what percentage of referred patients achieved enrollment on a trial, and of those who did enroll, we sought to establish evidence of clinical benefit. For this, we reviewed the individual electronic medical records for evidence of clinical complete or partial Response Evaluation Criteria in Solid Tumors (RECIST) response or report of any degree of tumor shrinkage. We also looked for 3- and 6-month progression-free survival (PFS) as surrogates for degree of clinical benefit. We chose not to examine either overall survival or any protocol-specific surrogate end points, given the substantial heterogeneity of the patient population and of the study portfolio.

Our review of the electronic referral records of the Early Drug Development Service of Memorial Sloan Kettering Cancer Center identified 394 GI oncology patients referred in calendar year 2018. Of these 394 patients referred, 54 were enrolled and 53 (13.5%) initiated treatment (one patient withdrew consent prior to initiating treatment). Thirty-four were accrued to immunotherapy-based trials and 19 were accrued to targeted therapy trials (n = 16 to a phase I trial and n = 3 to a phase II basket trial). Of the 53 patients treated, the median age was 57 (range = 23–80) years; 26 were female. Gastrointestinal cancer types included colorectal (n = 26), pancreas (n = 10), cholangiocarcinoma (n = 6), appendiceal (n = 3), small bowel adenocarcinoma (n = 3), ampullary (n = 1), esophageal (n = 1), gastric (n = 1), peritoneal mesothelioma (n = 1), and cancer of unknown primary (n = 1). The most common reasons for nonaccrual were lack of available trials for specific molecular cohorts, lack of availability of treatment spots on a protocol, and failure at the time of evaluation to meet either general performance status, organ function, or specific molecular eligibility criteria for specific trials.

Of 53 patients who initiated treatment on a trial, 52 received treatment with either an investigational agent alone or an investigational agent plus an approved drug known not to have single- agent activity in that patient’s tumor (eg, an investigational agent plus a PD-1 inhibitor in a microsatellite stable colorectal cancer). None (0%) of these 52 patients achieved 6-month PFS. Two of these 52 patients (3.8%) achieved 3-month PFS by RECIST criteria. One of these patients had tumor growth just below RECIST criteria for progression at 3 months and came off study for progression at 4 months. The other was assessed as stable on first scan at 2 months and came off study for progression on the second scan at 4 months. None (0.0%) of the 52 patients achieved any degree of documented tumor shrinkage while receiving trial drugs.

One patient with locally advanced, unresectable pancreatic cancer and no definitive metastases was initially treated with standard FOLFIRINOX (oxaliplatin, irinotecan, fluorouracil, leucovorin) for 4 months with stable disease. Radiation therapy was then given to the pancreatic mass. Two months later, with the tumor still stable on CT scan, the patient was recommended to start FOLFIRI (irinotecan, fluorouracil, leucovorin) as a standard treatment option. The patient instead chose to enroll on a clinical trial of irinotecan plus an investigational oral targeted agent. She remained on this treatment for 8 months with continuation of stable disease. She subsequently developed new peritoneal metastases, without change to the pancreatic mass.

One of the supreme challenges faced by clinicians, hospitals, pharmaceutical companies, and other stakeholders in medical oncology is that of providing accurate and, at the same time, compassionate communication to patients with treatment-refractory cancers regarding their prognoses and clinical options. In striving to meet this challenge, we work to maintain optimism tempered by realism and realism tempered by optimism. This well-intentioned, compassionate approach to communication of what is too often tragic and devastating information can lead to a failure to accurately convey the true limitations of prognoses and the often-extreme limitations in both the quality and quantity of viable treatment options. This may negatively impact patient judgment in the decisions that they must make for the remaining time of life that they have. Clinical trials of new agents are a highly appropriate consideration for many patients who have exhausted standard treatments. However, the degree to which GI oncology patients, as well as referring clinicians, perceive that these trials are accessible and have established potential for benefit may be overstated.

Our experience as noted here demonstrates two important issues that are likely underappreciated in the cancer community: the relative difficulty for even medically fit GI oncology patients to access early drug development trials because of the limited availability of study openings and the specifics of individual protocol eligibility, and the extreme paucity of benefit seen from these trials in the GI oncology population. This discordance between availability and demand is likely even greater than our data suggest, in that the nearly 400 GI oncology patients referred internally to our Early Drug Development Service in the year 2018 represent a subset of what is likely a far larger group. Because it is well understood that early drug development trials routinely require adequate performance status and renal, hepatic, and bone marrow function, patients who fail to meet such generic eligibility criteria are not routinely referred. As such, the patients identified in this study represent an appropriately selected subset of those who have exhausted standard therapy, largely excluding patients with poor or marginal performance status, jaundice or renal insufficiency, or substantially suppressed neutrophil counts, platelet counts, or hemoglobin. In addition, patients determined by informal “curbside” consultation to be ineligible or to not have availability of a trial would not have been captured in our data.

It is noteworthy that our institution has among the largest and most robust early drug development programs in the country, so that the experience of our patients is likely to represent a “best case” scenario of what GI oncology patients might expect in terms of access to and benefit from trials. That said, in contradistinction to what was standard in the past several decades, most phase I clinical trials are now open at a large number of sites simultaneously, such that each site may have only a very limited number of openings for patients. As such, a trial may be open at a center for several years but only be able to offer entry to a small number of patients, and only at limited times. Thus, having many open phase I trials may not translate into many available openings.

Simply gaining access to a clinical trial, however, is far from an indication of therapeutic benefit, because by definition the treatments on an early drug development clinical trial are of unknown efficacy. It is reasonable to presume, however, that patients seeking these trials are hoping for a substantial probability of clinical benefit in the form of effective tumor control. Our data suggest that for GI oncology patients such hopes may be overstated. Of the nearly 400 patients in our study cohort who sought clinical benefit from a phase I trial, only 53 were enrolled, and none showed either tumor regression or prolonged control to an investigational agent.

Some prior phase I efficacy reviews noted more activity than we encountered (3,4). These reported across all tumor types, whereas we report exclusively on GI tumors. We also report the results of a single year. Results in prior or future years may differ. Our study focused on drugs in the earliest phases of clinical development and included only phase I and basket trials run by our Early Drug Development Service. It did not include disease-specific protocols within our GI Oncology Service. The results of those trials have been, or will be, reported separately. For purposes of expediency, we chose to study the results of patients referred over a single calendar year and chose the most contemporary year for which we have a minimum of 6-month follow-up on all patients.

To be clear, we feel strongly that phase I clinical trials remain an essential component of the drug development process. The main purposes of a phase I trial have always been to define a sufficiently safe dose and schedule of a new drug and to identify that drug’s dose-limiting and nondose-limiting toxicities. A lways hoping for clinical benefit, patients were informed that these were the main objectives of the phase I trial and that clinical benefit, which is always hoped for, was not a specific goal or expectation of the study. However, we believe that expectations of benefit on early drug development clinical trials have undergone a change over the past decade. Much of the increase in expectations is likely related to two important paradigm shifts. The first of these was the advent of clinically effective immunotherapy in the form of, first, CTLA4 inhibitors and then PD-1 or PD-L1 inhibitors. The phase I trials of these agents demonstrated considerable activity (9,10). Furthermore, because the therapeutic window of these agents was found to be larger than what is seen with traditional cytotoxic agents, activity was often seen at dose levels far below what was determined to be a recommended phase II dose, and specific reproducible dose-limiting toxicities were not necessarily identified. This led to greater frequency of clinical benefit in the dose-escalation phase of the phase I trials than had been traditionally seen in many tumors that were previously poorly responsive to cytotoxic therapies such as melanoma, kidney, and lung cancers. In addition, many patients were treated on the “disease-specific cohorts” of those phase I trials, with a number of favorable results. The use of these active agents has now moved rapidly into standard practice, and immunotherapy trials in early drug development are now exploring other targets beyond CTLA4 and PD-1, in the hopes of expanding immunotherapy efficacy for those whom CTLA4 and or PD-1 or PD-L1 inhibitors are not efficacious, such as many GI oncology patients. Unfortunately, thus far, additional immune targets have been far less successful. The quantum leaps forward that CTLA4 and PD-1/PD-L1 represented in many areas of oncology are not as yet being reproduced by inhibitors of other immune targets. As such, the activity seen in drug development clinics with newer investigational immunotherapies is far more modest across the board than had been experienced a decade ago, and a widely targetable immune checkpoint for colorectal, pancreatic, and many other GI malignancies has yet to be identified.

A second paradigm shift that likely influenced expectations for phase I trials was the development of targeted small molecule agents along with the concurrent development of NGS technology that allowed for identification of those patients whose tumors actually had the target that the targeted agent was designed to inhibit (7,8). Thus, phase I studies were now able to “enrich” the study population by enrolling only those patients whose tumors demonstrated the particular molecular mutation or amplification that the small molecule was designed to inhibit (5,6). Over time, the presence of the targeted mutation became an eligibility criterion for these phase I trials, resulting in a much more restrictive accrual but a much higher likelihood of benefit for patients enrolling on those trials than had been expected from phase I trials of cytotoxic agents or from phase I trials of targeted agents not enriched for patients whose tumors harbored the appropriate target. Similarly, a number of promising targets, including ALK, ROS, RET, and BRAF, as well as fusions involving NTRK and FGF, have shown striking clinical vulnerability when treated with an agent specifically designed to inhibit that particular mutated tyrosine kinase. The identification of these prominent driver mutations, coinciding with the broad availability of technology allowing mutational profiling of tumors using achieved, paraffin-embedded tissue, led to dramatic examples of patients on phase I trials, with cancers that progressed on all standard therapies, having spectacular and durable responses of the sort that was previously inconceivable.

These high-value molecular targets, however, are thus far rare in GI cancers and have shown lower degrees of activity and clinical benefit in GI tumors as compared with many other solid tumors. For example, the targeted BRAF inhibitor vemurafenib was found to be essentially inactive as a single agent in BRAF-mutated colorectal cancer (13,14), and of four colorectal patients with NTRK fusions, the NTRK inhibitor larotrectinib showed only one response durable for 5.6 months (15,16). In addition, it is often underappreciated that the presence of a dominant driver mutation such as RAS or BRAF typically overrides the clinical relevance of other potentially targetable mutations and renders the concurrent presence of such targets clinically moot.

In those few areas within GI oncology where we have seen efficacy of these approaches, these strategies have already been rapidly incorporated into standard practice and so are likely no longer the subject of phase I investigations. For example, screening for mismatch repair deficiency or microsatellite instability and treatment of mismatch repair deficient or microsatellite instable patients with a PD-1 inhibitor has become standard practice (17). BRAF V600E mutated colorectal cancer, although not responsive to single agent BRAF inhibition, is now routinely treated with a combination of BRAF, MEK, and EGFR inhibition (18), although the results are more modest than in melanoma and other BRAF-mutated solid tumors.

An additional factor that may contribute to inflated expectations of targeted therapy trials in GI malignancies is the presumption that the presence of a targetable mutation and the availability of an agent designed to target that mutation will necessarily produce a favorable outcome. This is certainly the primary motivation behind phase II basket trials, which focus on such a match between targeted agent and tumor expression of the target. Although the potential for such favorable outcomes certainly exists, reports from numerous trials refute the generality of this presumption of efficacy. Several reports from the National Cancer Institute MATCH trial (19,20), as well as other large, high-profile trials (21,22), have failed to show substantial efficacy for targeted agents matched to patients with tumors harboring those targets. Furthermore, as previously noted, the activity of several targeted approaches, such as small molecule inhibitors of BRAF mutations or NTRK fusions, have demonstrated substantially inferior results specifically in GI tumors harboring these mutations vs non-GI tumors with the same molecular aberrations (11–14). Thus, the probability or extent of benefit of an investigational targeted approach may be less than what might be achieved for patients with other tumor types.

There are, of course, limitations to our analyses. Our data are retrospective, although they are obtained from a prospectively populated database. Although our perception, based on clinical experience, is that most patients who did not enter a trial did not do so because we were not able to offer them a treatment slot because of unavailability of a study or a study treatment slot for which they met eligibility criteria. We do not know what percentage of patients were offered a study slot and then declined it, although our sense is that this is not a common scenario in our phase I clinic and, in such cases, would have to be acknowledged as a study that failed to meet the patients’ expectations of what such a trial would offer in terms of expected risks and benefits. We do not have data on what the subjective experiences of participation, or nonparticipation, in a phase I trial were for those who did and did not receive treatment. Nevertheless, as we assume that therapeutic benefit was the predominate goal for most patients seeking investigational treatment after failure of standard options, we believe it is a valid assumption that the failure to obtain investigational treatment, or the failure to gain benefit in terms of substantial disease control from such treatments, would constitute a major disappointment to most who sought out these trials.

In conclusion, the hope that patients, families, and clinicians have regarding both availability and efficacy of modern-day phase I and phase II basket trials needs to be balanced with the harsh reality of the very modest potential for efficacy that such trials have shown for patients with treatment-refractory GI malignancies. Alongside ongoing research efforts to identify clinically impactful molecular and immune targets, accurate and compassionate communication about the current state of investigational therapy in this vulnerable population remains paramount for patients to make informed decisions for their cancer treatment and to plan, to the best of their ability, for their future.

 

Funding

Supported in part by National Cancer Institute P30-17 CA008748, Craig Thompson, Principal Investigator.

Notes

The funder had no role in the writing of this commentary or the decision to submit it for publication.

The authors declare the following potential conflicts of interest: Rajiv Agarwal: none; Nicholas Cangemi: none; Andrew Epstein: none; James Harding: consulting fees unrelated to the current work from Bristol-Myers Squibb, Eiasi, Eli Lilly, and CytomX and research support from Bristol-Myers Squibb; Diane Reidy-Lagunes: advisory board for Novartis, Lexicon; Research support from Merck, Novartis, Ibsen; Leonard Saltz: research support from Taiho.