Trends in Grade 5 Toxicity and Response in Phase I Trials in Hematologic Malignancy: 20-Year Experience From the Cancer Therapy Evaluation Program at the National Cancer Institute

Abstract

PURPOSE

Cancer drug development has largely shifted from cytotoxic chemotherapy to targeted treatment in the past two decades. Although previous studies have highlighted improvement in response rates in recent phase I trials, disease-focused reporting is limited.

METHODS

We integrated patient-level data for patients with hematologic malignancies who participated in phase I trials sponsored by the National Cancer Institute Cancer Therapy Evaluation Program between January 2000 and May 2019 and estimated the trend of grade 5 toxicity and response by disease subtype over time.

RESULTS

We analyzed 161 trials involving 3,308 patients, all of whom were assessed for toxicity and 2,404 of whom were evaluable for response to therapy. The overall rate of grade 5 toxicities was 1.81% (95% CI, 1.36 to 2.27), with no significant change in the rate over time. Baseline characteristics associated with higher risk of grade 5 toxicity were age and performance status ≥ 2 at enrollment. Overall response rate (ORR) and complete response (CR) rate for all trials during the study period were 25.1% and 14.7%, respectively. A significant increase in both ORR and CR rate was observed over time (ORR, 18.5% in 2000-2005, 25.9% in 2006-2012, and 50.6% in 2013-2019, P < .001). ORR in phase I trials varied across disease subtypes: 20.2% in acute myeloid leukemia, 9.1% in myelodysplastic syndrome, 43.2% in lymphoma, 42.9% in chronic lymphocytic leukemia, 15.1% in acute lymphoblastic leukemia, and 16.5% in myeloma.

CONCLUSION

Over time, the ORR and CR rates in phase I trials for hematologic malignancy have improved meaningfully, whereas the rate of toxicity-related death remains stable. This study provides broad experience that physicians can use when discussing the potential outcomes for patients with hematologic malignancy considering participation in phase I trials.

BACKGROUND

Cancer drug development is conducted in a stepwise fashion. Phase I trials are a first step in which the primary objective is to evaluate the safety and tolerability of a new agent or regimen and to determine the recommended phase II dose (RP2D) for subsequent clinical trials. The treatments investigated in phase I trials include first-in-human agents, novel agents exploring different dosing schedules, or new combinations of drugs proven to be effective as single agents in prior studies. On the basis of the results of the phase I trial, clinical investigation proceeds to phase II and phase III studies focusing on acquiring further safety data, assessing drug efficacy, and determining survival benefit.

CONTEXT

• Key Objective

• We conducted the study to evaluate the toxicity and response in phase I trials for hematologic malignancies and determine how these changed over time.

• Knowledge Generated

• The response rate in phase I trials has increased significantly in the last 20 years, and was 50.6% in 2013-2019. Grade 5 toxicity remained stable over this period with 1.8% overall. Trials that evaluated combination treatment focused on hematologic malignancy or specific disease showed higher response rates.

• Relevance

• These study results aid physicians in discussing the potential outcomes for patients with hematologic malignancy considering participation in phase I trials.

Expectations of direct clinical benefit to patients participating in phase I clinical trials have historically been low because the primary objectives are typically to evaluate toxicity and safety as a primary end point along with defining the optimal drug dosing. Often, these oncology treatments are expected to have a dose-dependent response. Although the maximum tolerated dose may not be the most effective dose, most patients are exposed to relatively lower doses than the maximum tolerated dose, thus the anticipated efficacy across the trial can also be lower. As a result, phase I trials are commonly considered a last resort for patients. Some researchers have raised ethical concerns regarding performing phase I trials in patients with cancer who are desperately seeking treatment,¹ since the perception of benefit versus risk from participating in a phase I trial is low.^2-5 Nevertheless, ASCO strongly encourages participation in phase I trials, for possible clinical benefit, societal benefit, and to support patients, since clinical trials include rigorous institutional review board assessment, informed consent, and protocol-specified monitoring and reporting procedures.^6,7 Recent meta-analyses of phase I trials have shown improved responses.^8,9 However, disease-specific results in hematologic malignancies have been lacking.

The National Cancer Institute's (NCI) Cancer Therapy Evaluation Program (CTEP) sponsors clinical trials by funding an extensive national network of cancer research that evaluates new anticancer agents and combinations. Early-phase clinical trials supported by CTEP are conducted through the Experimental Therapeutics Clinical Trials Network (ETCTN) consisting of NCI-designated academic cancer centers in the United States and Canada. CTEP particularly emphasizes translational research to understand molecular targets and mechanisms of drug effects and attempts to fill critical gaps between government, academic, and industry cancer research efforts. Through funded investigator-initiated clinical trials, CTEP maintains an extensive database dating back to its inauguration, including patient demographics, and agent toxicity and efficacy data. On the basis of previously published analysis of CTEP-sponsored early-phase clinical trials conducted between 1974 and 1982, the overall response rate (ORR) in phase I trials was only 4.2%.⁵ The follow-up CTEP analysis showed that the ORR in phase I trials was slightly improved to 10.6% between 1991 and 2002, although no significant trend was observed for improvement within this period.² Notably, these studies analyzed response and toxicity for trials that predominantly involved cytotoxic chemotherapy and included all patients with solid tumors and hematologic malignancies.

Since the late 1990s, the treatment of cancer has dramatically changed from cytotoxic chemotherapy to targeted agents such as monoclonal antibodies, small molecule inhibitors, and immunotherapy including cellular therapy. For hematologic malignancies, targeted agents such as imatinib¹⁰ and rituximab¹¹ brought a paradigm shift in treatment in the early 2000s, building on advances in the understanding of disease biology and identification of therapeutic targets. As a result, phase I trials became more disease-focused rather than evaluating the safety of a new agent across patients of all cancer subtypes. In recognition of the 50th anniversary of the National Cancer Act,¹² we performed a retrospective evaluation of a large cohort of patients with hematologic malignancies who received treatment on CTEP supported phase I clinical trials to assess trends in toxicity, response, and survival outcomes over the past 20 years.

METHODS

Study Population

The data consisted of all patients with hematologic malignancies receiving treatment on CTEP-sponsored investigator-initiated phase I oncology trials conducted between January 2000 and May 2019. CTEP receives comprehensive trial data at regular intervals from investigators and actively monitors all trials through routine data submission and periodic audits. The collected data are managed by two different monitoring systems, the Clinical Trials Monitoring System (CTMS) and the Clinical Data Update System (CDUS). The CTMS includes electronically submitted case-report forms for trials of novel agents first evaluated in humans and combinations of investigational new drugs and at least one US Food and Drug Administration–approved drug. The CDUS receives electronic data according to course of therapy and for the individual patient. It is updated every 3 months and is generally used for phase I trials of agents with a toxicity profile established in earlier studies. Grade 5 toxicity was reported within 24 hours to CTEP when it occurred any time over the duration of the study. All phase I clinical trials during the defined period were identified using the Integrated Platform for Agents and Diseases, an enterprise search tool for the CTEP Enterprise System database. Although patients who received treatment on the dose-expansion phase during a phase I trial were included, we excluded patients who received treatment on phase I/II trials.

Statistical Analysis

We computed the rates of grade 5 toxicity (treatment-related death), response to treatment, and survival following enrollment to the trials. Agents used in the trials analyzed were grouped by investigators (D.C., L.M.C., and N.T.) according to the mechanism of action (Data Supplement, online only). Toxicity grade was based on the Common Terminology Criteria for Adverse Events (CTCAE), and attribution of all grade 5 adverse events to the intervention (unrelated, unlikely, possible, probable, and definite) was included as assessed by the phase I study investigators when reported to CDUS.¹³ Response to the treatment (best response during the trial) was reported for each patient if available according to the standard response criteria for each disease.

The aim of this study was to assess grade 5 toxicity rate, ORR, and complete response (CR) rate over time. Univariate associations of these outcomes with time, treatment, and patient clinical variables were assessed using generalized linear mixed models. Inferences on odds ratios (ORs) were performed. A likelihood ratio test was used to compare a model containing only study cohort as a random effect to one containing the study cohort and the variables as a fixed effect. Survival rate was measured using the Kaplan-Meier method, and the differences between groups were assessed using the log-rank test. Univariate associations of overall survival (OS) with these factors were assessed through Cox mixed models in a similar manner. Benjamini-Hochberg procedures were applied to correct for multiple testing.¹⁴ Analyses were performed in R. Functions for fitting the generalized linear mixed models and Cox mixed models can be found in the lme4 and coxme libraries.

RESULTS

Trial Characteristics

Overall, data from 161 trials involving 106 agents (Data Supplement) and 3,308 patients during the study period were analyzed (Table 1). The median age was 61 years (range, 0-96 years) and 61% of patients were male; 126 trials (78%) treated only patients with hematologic malignancies and 35 trials (22%) treated patients with hematologic malignancies and solid tumors. Forty-nine trials (30%) focused on a single histology or specified patient population on the basis of target of investigational agent (disease-specific trial). The median number of patients treated per trial was 25 (range, 1-122). Acute myeloid leukemia (AML) was the most common disease (n = 1,502), followed by lymphoma (n = 689), myelodysplastic syndrome (MDS; n = 266), chronic lymphocytic leukemia (CLL; n = 232), acute lymphoblastic leukemia (ALL; n = 193), and myeloma (n = 153). Sixty-six trials (41%) used agents as monotherapy and 95 trials (59%) tested combination treatment. The most commonly used nonchemotherapy investigational agents were monoclonal antibodies and histone deacetylase (HDAC) inhibitors (28 trials each), followed by DNA methyltransferase inhibitors (18 trials) and proteasome inhibitors (17 trials).

TABLE 1.

Characteristics of the Phase I Trials

Grade 5 Toxicity in Phase I Trial for Hematologic Malignancies

Overall, 468 patients died while on the study. Sixty of these deaths were attributed to the trial treatment (grade 5 toxicity rate, 1.8%, 95% CI, 1.4 to 2.3). Grade 5 toxicities occurred in 2.2% of patients with AML, 3.8% in MDS, 0.7% in lymphoma, 1.7% in CLL, 1.6% with ALL, and 0.7% with myeloma (P = .091, Table 2). Patients with lymphoma had significantly lower risk of grade 5 toxicity compared with patients with AML (OR, 0.31, 95% CI, 0.11 to 0.87). The overall death rate on the trial, regardless of the cause, was 19.6% for patients with AML, 15.8% with MDS, 8.4% with lymphoma, 6.0% with CLL, 11.4% with ALL, and 3.3% with myeloma (P < .001).

TABLE 2.

Risk Factors Associated With Death During Trial

Baseline characteristics associated with a higher risk of grade 5 toxicity were age (OR, 1.02 for each 1-year increase in age; 95% CI, 1.01 to 1.04; P = .007) and performance status (PS) ≥ 2 at enrollment (OR, 2.76; 95% CI, 1.42 to 5.37; P = .006). No specific class of agents were associated with a higher rates of grade 5 toxicities. The grade 5 toxicity rate for combination trials, namely those using two or more agents, was not higher compared with monotherapy trials (1.7% v 2.1%); however, patients who received treatment on disease-specific trials experienced higher risk of grade 5 toxicity (2.7% v 1.4%; OR, 1.87; 95% CI, 1.01 to 3.47). Older patients, those with PS ≥ 2, and those with albumin < 3.5 g/dL had significantly higher risk of death (all causes of death during trial) during the trial (Table 2). Compared with patients with AML, patients with lymphoma, CLL, or myeloma had lower risk of death during a trial. Patients who received an investigational agent on a disease-specific trial had significantly lower risk of death during the trial.

Rates of grade 5 toxicities did not change significantly over time (2000-2005 v 2006-2012 v 2013-2019; Fig 1 and Table 3). The rate of grade 5 toxicity significantly increased in lymphoma in the 2013-2019 time period since the rate of grade 5 toxicity was 0% in the 2000-2012 time period. The overall rate of death during phase I protocol treatment has been stable over the period (15.2% in 2000-2005, 13.6% in 2006-2012, and 12.2% in 2013-2019; Table 3); however, the rate significantly differed among diseases.

FIG 1.

Trend of grade 5 toxicity and response by time between 2000 and 2019 with 95% CI.

TABLE 3.

Trend of Grade 5 Toxicity and Response Over Time

Response Rates in Phase I Trials for Hematologic Malignancies

Response assessment was available for 2,404 patients (72.7%). The ORR and CR rate for all trials during the study period was 25.1% (95% CI, 23.3 to 26.8) and 14.7% (95% CI, 13.3 to 16.2), respectively. ORR was significantly higher in lymphoma (43.2%) and CLL (42.9%) compared with that in AML (19.8%), MDS (9.0%), ALL (13.3%), or myeloma (16.5%). There was a significant increase in the ORR and the CR rate over time (Fig 1). The ORR increased significantly from 18.5% in the 2000-2005 time period to 50.6% in the 2013-2019 time period (Table 3). Meanwhile, the CR rate increased from 10.5% in the 2000-2005 time period to 26.0% in the 2013-2019 time period. Both ORR and CR rate were significantly higher in hematologic malignancy–focused trials than in trials that enrolled all cancers (ORR, 25.9% v 8.4%, CR rate 15.4% v 2.5%, both P < .001), and in disease-specific trials than in nonspecific trials (ORR, 38.5% v 18.7%, CR rate 22.7% v 10.9%, both P < .001).

The trend over time differed by disease. The ORR and CR rate improved over time in lymphoma (38.2% in the 2000-2005 time period to 61.0% in the 2013-2019 time period) and CLL (21.2% in the 2000-2005 time period to 78.0% in the 2013-2019 time period). Lower ORR was observed in patients with AML and MDS enrolled in trials conducted in the 2013-2019 time period. However, the number of patients with AML or MDS evaluable for response during this period was small.

Disease-specific ORRs and CR rates by class of investigational agent are shown in Data Supplement. Although the numbers are limited in this analysis, agents such as immunomodulatory drugs (iMIDs) in AML and lymphoma, monoclonal antibodies in lymphoma, CLL, and ALL, and BTK inhibitors in lymphoma were associated with higher ORRs and CR rates, whereas HDAC inhibitor was associated with lower ORR in lymphoma. The proportion of combination trials increased over time (2000-2005: 42%, 2006-2012: 69%, and 2013-2019: 79%, P < .001). Combination trials had significantly higher ORRs and CR rates compared with trials that used monotherapy across all diseases: AML (ORR, 26.1% v 6.7%, CR rate, 22.2% v 4.1%), MDS (ORR, 13.3% v 6.7%, CR rate, 6.7% v 3.7%), lymphoma (ORR, 47.6% v 18.4%, CR rate, 26.4% v 4.0%), CLL (ORR, 54.3% v 6.8%, CR rate, 10.9% v 0%), ALL (ORR, 26.0% v 5.4%, CR rate, 22.0% v 0%), and myeloma (ORR, 18.8% v 10.0%, CR rate, not assessed).

Survival in Patients With Hematologic Malignancies Who Participated in Phase I Trials

Overall, 776 deaths were recorded, 448 of which occurred within 3 months after trial enrollment. The 6-month OS was 52% in AML, 77% in MDS, 84% in lymphoma, 94% in CLL, 30% in ALL, and 92% in myeloma. The median OS among all patients after enrollment in phase I studies was 384 days (95% CI, 337 to 491 days). Patients with ALL had significantly shorter median OS (100 days) compared with patients with other diseases (Fig 2). OS by time period and disease subtype is shown in Appendix Fig A1 (online only). There was significant improvement in OS by time period in AML (P < .001) and ALL (P = .0024). PS ≥ 2 (hazard ratio [HR], 2.35; 95% CI, 1.91 to 2.91) and albumin < 3.5 g/dL (HR, 1.99; 95% CI, 1.63 to 2.42) were associated with significantly shorter survival after trial enrollment (Data Supplement). Patients who were on combination trials (HR, 0.53; 95% CI, 0.31 to 0.89) and on disease-specific trial (HR, 0.41; 95% CI, 0.25 to 0.68) experienced longer survival.

FIG 2.

Kaplan-Meier survival curve from trial enrollment by disease. ALL, acute lymphoblastic leukemia; AML, acute myeloid leukemia; CLL, chronic lymphocytic leukemia; MDS, myelodysplastic syndrome.

DISCUSSION

A significant improvement in ORRs and CR rates has occurred over the past 20 years in CTEP-sponsored phase I trials for patients with hematologic malignancies. Approximately half of all patients achieved response, and one in four achieved CR in the 2013-2019 time period, whereas grade 5 toxicity rates have been stable at approximately 2%. In addition, this study observed favorable 6-month survival following phase I trial enrollment, particularly in patients with lymphoma, CLL, and myeloma. These findings contradict the common belief of short expected survival for patients with hematologic malignancies who have received multiple lines of standard therapy and are enrolled on a phase I trial. Schwaederle et al⁹ performed meta-analysis of 346 phase I trials published between January 2011 and December 2013 and found that response rate was significantly higher in hematologic malignancies compared with solid tumors, and higher in trials with biomarker-based eligibility criteria compared with those without. Subsequently, Chakiba et al⁸ conducted meta-analyses of 224 phase I trials published between January 2014 and June 2015, confirming that the response rate is higher in biomarker-based trial and reported that response rate was higher in single disease-focused phase I trials. These studies highlighted recent improvement in response rates and provide encouragement for phase I trials.¹⁵ However, publication bias and lack of disease-specific analyses remain significant limitations for these literature search–based retrospective meta-analyses.¹⁶ A study by Zeidner et al¹⁷ of CTEP-sponsored phase I trials for AML conducted from 1986 to 2009 demonstrated significant improvement in ORR over time (8.9% in the 1986-1990 time period to 22.6% in the 2006-2009 time period). However, the study did not analyze the class of agents used in the trials. Since the Zeidner study only focused on AML, they reported early mortality rates instead of grade 5 toxicity. The 30-day and 60-day mortality rates were 11.1% and 22.6%, respectively. Otherwise, few studies have focused on phase I trials for hematologic malignancies that reported outcomes by disease subgroups. In this study, we confirmed that patients who received treatment on hematologic malignancy or on disease-specific trials experienced higher response rates and longer survival. Our study is the first comprehensive evaluation of phase I trials using patient-level data that avoids publication bias, and the toxicity and response rate presented can serve as a reference frame for future phase I trials for hematologic malignancies.

During the study period, we observed dramatic changes in treatment, transitioning from cytotoxic chemotherapy to disease-specific targeted treatment. In the 2000-2005 time period, nine of 67 trials (13%) used only cytotoxic agents. This ratio decreased to one in 33 trials (3%) in the 2013-2019 time period. Among these targeted agents, monoclonal antibodies, iMIDs, and BTK inhibitors were associated with higher response rate in various diseases. Notable agents that were used in these CTEP trials included blinatumomab, lenalidomide, and ibrutinib. The use of combination therapies also increased significantly over time. Increasing toxicity is a major concern for combination treatment; however, combination treatment did not increase the rate of grade 5 toxicity in this large review. Targeted agents have different toxicity profiles than cytotoxic treatment and allowed investigators to explore more effective combination strategies over time that likely increased response rates during the study period without increases in deaths from treatment-related toxicity.

This analysis had limitations related to study selection and generalizability of the results. This study summarized CTEP-sponsored phase I trials, which often involved combination trials structured through CTEP agreements with industry partners,¹⁸ and these trials were funded on the basis of the strength of preclinical studies of the combinations. The general perception of a phase I trial as equivalent to a first-in-human single-agent trial is mistaken and such trials are under-represented in the current study. Many recent phase I trials have an expansion cohort after determining the RP2D. Our study could not assess how the responses were affected by the dose level in the trial. Higher response rates may be due to the expansion cohort in a disease-specific cohort, not by the classic phase I dose escalation cohort. Because of the lack of data, we could not estimate duration of response and progression-free survival or examine whether toxicity and response were influenced by inclusion/exclusion criteria or number of prior regimens.

In conclusion, this study of more than 3,000 patients with hematologic cancers treated on 161 trials over two decades demonstrates that response rates in phase I trials have improved without significant increases in treatment-related deaths. Phase I trials vary from first-in-human agent trials to novel combination therapy of approved agents, and therefore, discussion with patients is necessary to assure that they are well informed in the informed consent process for participation. These results suggest that physicians and patients should be encouraged to consider phase I studies, as meaningful therapeutic outcome can be clearly observed in hematologic malignancies.

APPENDIX

FIG A1.

Overall survival by year period in each disease: (A) acute myeloid leukemia, (B) myelodysplastic syndrome, (C) lymphoma, (D) chronic lymphocytic leukemia, (E) acute lymphoblastic leukemia, and (F) multiple myeloma.

AUTHOR CONTRIBUTIONS

Conception and design: Dai Chihara, Naoko Takebe

Administrative support: James H. Doroshow, Naoko Takebe

Provision of study materials or patients: S. Percy Ivy, Naoko Takebe

Collection and assembly of data: Dai Chihara, Shanda R. Finnigan, Lisa M. Cordes, Nebojsa Skorupan, Yoko Fukuda, Naoko Takebe

Data analysis and interpretation: Dai Chihara, Erich P. Huang, Loretta J. Nastoupil, Christopher R. Flowers, Naoko Takebe

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

Trends in Grade 5 Toxicity and Response in Phase I Trials in Hematologic Malignancy: 20-Year Experience From the Cancer Therapy Evaluation Program at the National Cancer Institute

The following represents disclosure information provided by 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/jco/authors/author-center.

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