Phase I CAR-T Clinical Trials Review

Abstract

Background/Aim:

Chimeric antigen receptor (CAR) T cells with tumor specificity are being increasingly investigated. Phase I trials are the first step of testing for safety of novel CAR-T therapy to determine the maximum tolerated dose (MTD). Several dose escalation methods have been developed over time including rule-based, model-based and model-assisted designs. The goal of this project is to overview the phase I designs used in current CAR-T trials.

Materials and Methods:

We searched PubMed for peer-reviewed literature published between January 1, 2015 and December 31, 2021. The search was limited to human studies in the English language using the keywords “CAR-T phase I”, “clinical trials”, and “full text”.

Results:

One hundred nine papers with at least partial phase I components were included for analysis. 31.2% of the trials used the traditional 3+3 or a variation of said design, and 60.6% did not mention the dose escalation design. The majority of the manuscripts (59.6%) did not report cohort size while 19.3% did not specify the timing of evaluation. Although most of the studies were registered with CT.gov, only 33.9% had any results submitted or posted to CT.gov. These trends persisted even in manuscripts published in journals with high impact factors.

Conclusion:

Standardizing the publication criteria and providing basic elements of phase I clinical trials are critical to ensure high quality of manuscripts. With the quick development and high costs of CAR-T cell therapy, adoption of advanced designs such as model-based and model-assisted should increase to improve efficiency of clinical trials.

Chimeric antigen receptor (CAR) T cells with a tumor specificity have been increasingly investigated as revolutionary cancer immunotherapy. Over the past decade, the use of CAR T cell therapy has rapidly emerged as a novel therapeutic option for cancer treatment. Since the safety and efficacy of CD19 CAR-T cells used in lymphoma was initially reported in 2010, development of new targets used for generating the CAR-T cells and their clinical applications, along with the number of early phase clinical trials of CAR-T cell therapies, has increased dramatically. Different CAR-T therapeutic approaches have been investigated, and the number of clinical trials has increased to demonstrate remarkable response outcomes. The Food and Drug Administration (FDA) agency has approved several CAR-T cell therapies for treating certain patients with multiple myeloma, acute lymphocytic leukemia (ALL), and advanced lymphomas (1–5). Many other CAR-T therapies are also in development, with CD19 remaining the most popular cell therapy target. These trials aim to improve safety and efficacy of CAR-T cell therapies and help in ensuring their continued approval.

Phase I trials, also known as dose finding studies, are usually the first step with the goal of testing for safety of novel CAR-T cell therapies to evaluate the safety at accelerating dose levels and determine the maximum tolerated dose (MTD). The principle of phase I trials is to treat fewer patients at sub-optimal doses while maintaining accelerated dose escalations without compromising patient safety. Several dose escalation methods have been developed over time including rule-based designs, model-based designs, and the relatively new class of model-assisted designs (6–13). Although these novel designs showed favorable operating characteristics to improve efficiency and accuracy of finding the MTD, the adoption of these designs has not increased in clinical trials. Despite poor performance and lack of flexibility(13–17), the 3+3 design continues to be used in the majority of phase I trials (18).

Given the high costs of CAR-T cell therapy development and the high failure rate of early phase trials (19), it is essential to consider alternative designs to optimize the clinical development process and ensure efficiency and patient safety. The goal of this project is to review and assess the phase I designs and quality of reporting of current CAR-T clinical trials.

Materials and Methods

We searched PubMed for all the peer-reviewed literatures published between January 1, 2015, and December 31, 2021. The search was limited to human studies in the English language using the keywords “CAR-T phase I”, “clinical trials”, and “full text”. One hundred fifty-six papers were retrieved from 47 journals, and forty-seven that did not meet the selection criteria (1 animal study, 1 cohort study, 1 treatment guidelines paper, 1 transcriptome profiling paper, 1 paper unrelated to CAR-T therapy, 1 device study, 1 pilot study, 3 design papers, 4 letters to the editor, 4 secondary analyses, 6 case reports, and 23 phase II trials) were excluded. Publication characteristics, study design aspects, and objectives/reporting information were summarized using descriptive statistics. The phase I/II studies were only summarized based on their phase I components.

Results

Characteristics of the reviewed studies are presented in Table I. One hundred fifty-six papers were retrieved, and 109 (69.9%) of these papers with at least partial phase I components were included for analysis (20–128). Out of 109, 30 (27.5%) had partial phase II components (27, 28, 47, 49, 52, 53, 57, 59, 61, 63, 64, 69, 73, 75, 78, 81, 87–89, 93, 94, 97, 98, 101, 103, 112, 115, 124, 127, 128), and 79 (72.5%) were solely phase I studies (20–26, 29–46, 48, 50, 51, 54–56, 58, 60, 62, 65–68, 70–72, 74, 76, 77, 79, 80, 82–86, 90–92, 95, 96, 99, 100, 102, 104–111, 113, 114, 116–123, 125, 126). The number of CAR-T publications gradually increased over the specified years. Nearly 50% of the trials were conducted in the United States and the majority of the trials (88.1%) were registered with ClinicalTrials.gov. Over 75% of these trials were published in journals with high impact factor (IF, high as ≥10).

Table 1:

Characteristics of reviewed phase I studies.

Characteristics	(N=109)	Characteristics	(N=109)
	n (%)		n (%)
Publication characteristics		Dose level and infusions
Publication Year		Number of Dose Levels
2015	4 (3.7%)	1	23 (21.1%)
2016	13 (11.9%)	2	14 (12.8%)
2017	14 (12.8%)	3	30 (27.5%)
2018	16 (14.7%)	4	9 (8.3%)
2019	20 (18.4%)	5	2 (1.8%)
2020	20 (18.4%)	6	2 (1.8%)
2021	22 (20.2%)	8	1 (0.9%)
Publication Location		Range	17 (15.6%)
US	50 (45.9%)	NA/Unreported/Unclear	11 (10.1%)
China	26 (23.9%)	Number of Infusions
Others	31 (28.4%)	Single	80 (73.4%)
Multi-country	2 (1.8%)	Multiple	25 (22.9%)
Registered with CT.gov		Various	3 (2.7%)
Yes	96 (88.1%)	NA/Unreported	1 (0.9%)
No	13 (11.9%)	Study design
Journals
Blood	14 (12.8%)	Study Phase
Clin Cancer Res	8 (7.3%)	Phase I	79 (72.5%)
J Clin Oncol	8 (7.3%)	Phase I/II	30 (27.5%)
Nat Med	8 (7.3%)	Sample Size
J Clin Invest	7 (6.4%)	≤30	83 (76.2%)
J Hematol Oncol	6 (5.5%)	>30	26 (23.9%)
Mol Ther	5 (4.6%)	Phase I Design
N Engl J Med	4 (3.7%)	Rule-based 3+3	34 (31.2%)
Others	49 (45.0%)	Rule-based other	3 (2.8%)
Impact Factor		Model-based/model-assisted	6 (5.5%)
High (≥10)	82 (75.2%)	NA/Unreported	66 (60.6%)
Low (<10)	27 (24.8%)	Cohort Size
Disease and study population		2	3 (2.8%)
		3	41 (37.6%)
Disease		NA/Unreported	65 (59.6%)
Leukemia	35 (32.1%)	Safety Evaluation Period
Lymphoma	21 (19.3%)	≤W4	27 (24.8%)
Myeloma	10 (9.2%)	>W4-W6	42 (38.5%)
Solid	26 (23.9%)	>W6	19 (17.4%)
Leukemia/Lymphoma	14 (12.8%)	NA/Unreported	21 (19.3%)
Leukemia/Lymphoma/Solid	2 (1.8%)	Results
Leukemia/Myeloma	1 (0.9%)
Study Population		Dose Limiting Toxicities (DLTs)
Adult	75 (68.8%)	0	35 (32.1%)
Pediatric	3 (2.8%)	1 or more	25 (22.9%)
Adult/Pediatric	30 (27.5%)	NA/Unreported	49 (45.0%)
NA/Unreported	1 (0.9%)	Cytokine Release Syndrome (CRS) ≥ grade 3
T-Cell Source		0	36 (33.0%)
Autologous	102 (93.6%)	1 or more	57 (52.3%)
Allogeneic	7 (6.4%)	NA/Unreported	16 (14.7%)
CAR-T Type		Neurotoxicity ≥ grade 3
CD19	67 (61.5%)	0	37 (33.9%)
Others	42 (38.5%)	1 or more	50 (45.9%)
CAR-T Generation		NA/Unreported	22 (20.2%)
I	6 (5.5%)	Objectives and reporting
I/II	1 (0.9%)
II	82 (75.2%)	Primary Endpoint
III	19 (17.4%)	Safety	101 (92.7%)
IV	1 (0.9%)	Response/Efficacy	8 (7.3%)
Lymphodepletion		Secondary Endpoint
Yes	83 (76.2%)	Response	62 (56.9%)
No	14 (12.8%)	Safety	29 (26.6%)
Various	5 (4.6%)	Survival	12 (11.0%)
NA/Unreported	7 (6.4%)	NA/Unreported	6 (5.5%)
		Results
		Reported	106 (97.3%)
		NA/Unreported	3 (2.8%)
		CT.gov Results
		Posted	37 (33.9%)
		NA/Unreported/Unposted	72 (66.1%)

About 64.2% of the phase I manuscripts centered on either leukemia, lymphoma, or a combination of the two as the disease of interest. The vast majority of the trials (96.3%) studied adults or a combination of adults and pediatric patients, with 93.6% using autologous T cells, 61.5% targeting CD19, and 94.5% using at least 2nd generation CARs. Lymphodepletion chemotherapy was given in 76.2% of the trials prior to the T cell infusions with either single (73.4%) or multiple (22.9%) infusions at 1–3 dose levels (61.4%).

Concerning the trials with phase I components, 60.6% did not mention the dose escalation design, and 31.2% used the traditional 3+3 or a variation of said design. Only 6 (5.5%) used model-based or model-assisted designs, with 3 being model-based [Bayesian adaptive dose-finding design based on Efficacy-Toxicity Trade-Offs (EffTox) and modified continual reassessment method (mCRM)] (44, 88, 103) and 3 being model-assisted [modified toxicity probability interval (mTPI) and Bayesian Optimal Interval (BOIN)] (45, 78, 106). Almost 76.2% of the phase I trials had sample sizes of 30 or less. More than half of the phase I trials (59.6%) did not report cohort size, and those that did report had sizes of 2 (2.8%) or 3 (37.6%). Approximately, 63.3% of the trials had a safety evaluation period within 6 weeks, while 19.3% did not specify the timing of evaluation. We also examined a subset of 75 trials with multiple dose levels, and among these, 46.7% did not mention the dose escalation design, 45.3% did not report cohort size, and 20.0% did not specify the safety evaluation period (data not shown). About 22.9% of the trials reported at least one DLT, 52.3% had ≥ grade 3 CRS and 45.9% had ≥ grade 3 neurotoxicity. The vast majority of the trials (92.7%) had safety as the primary endpoint, and 56.9% had response as the secondary endpoint. Although a majority of the phase I studies were registered with CT.gov, only 33.9% had any results submitted or posted to ClinicalTrials.gov. This might be due to the fact that result submission is not required under FDAAA 801 for phase I trials that are not applicable clinical trials.

The key items of reporting are described in Table II. About 60.6% of trials did not explicitly describe study design with 10.1% not reporting numbers of dose levels, 59.6% having no information about cohort size, and 19.3% not specifying time frame for safety evaluation. In addition, 45.0% did not report dose limiting toxicity (DLT) with no clear statement on cytokine release syndrome (CRS) (14.7%) and neurotoxicity (20.2%). We observed that study design was poorly reported even in the high impact factor (IF) journals with 58.5% not reporting study design.

Table II.

Reporting quality of key items.

Study aspects	Unreported or Unclear N (%)
	All (N=109)	High IF (IF≥10) (N=82)	Low IF (IF<10) (N=27)
Design essentials
Study design	66 (60.6)	48 (58.5)	18 (66.7)
Number of dose levels	11 (10.1)	8 (9.8)	3 (11.1)
Cohort size	65 (59.6)	48 (58.5)	17 (63.0)
Safety evaluation period	21 (19.3)	12 (14.6)	9 (33.3)
Safety outcomes
DLTs	49 (45.0)	30 (36.6)	19 (70.4)
CRS events	16 (14.7)	8 (9.8)	8 (29.6)
Neurotoxicity	22 (20.2)	11 (13.4)	11 (40.7)

IF: Impact factor, DLTs: Dose-limiting toxicities, CRS: Cytokine release syndrome

Discussion

Our results showed that the basic elements and key points of publications were illustrated in only a fraction of papers (Table II). More than 50% of the trials did not report their study designs, and a number of trials did not clearly describe or report the relevant components for the design of these phase I trials such as number of dose levels, cohort size, and time frame of safety evaluation.

Table III lists criteria and essential items concerning phase I clinical trials in general. Specifically, the description of any DLTs as well as the MTD chosen are listed as essential items to include. Additionally, explicit definitions of the primary and secondary objectives, as well as information on safety and efficacy outcomes, are listed as items that should be reported. Lastly, more statistics focused items, such as cohort size and the number of patients on each dose level, as well as the number of dose levels, are also stated to be essential items. Many of the items listed, like these, are the exact items we’ve found to be lacking in overall reporting.

Table III.

Proposed detailed criteria and essential items for reporting Phase I trials.

General aspects

Rational for study

Current status of study

Study population and disease type

Description of CAR-T (type, generation)

Primary and secondary objectives (endpoints)

Description of treatment (including number of infusions w/o lymphodepletion, dose, treatment cycles/schedule)

Statistical aspects

Study dose escalation design and safety monitoring method

Cohort size

Definition of DLT, MTD

Toxicity evaluation period

Number of patients treated at each dose level and overall

Outcome aspects

Safety outcomes

DLT encountered on study

Description of grade 3 toxicity and above at least possibly related to study drug

Description of cytokine release syndrome (>=grade 3)

Description of neurotoxicity (>=grade 3)

Conclusion explicitly states MTD, RP2D or reason for early trial closure

Efficacy outcome

CAR-T: Chimeric antigen receptor T-cells, DLT: Dose-limiting toxicity, MTD: Maximum tolerated dose, RP2D: Recommended phase II dose

Our findings revealed that 31.2% of trials used rule-based designs, such as the 3+3 and its variations. These designs continued to be commonly used for phase I trials and are broadly recognized by clinicians primarily because of their simplicity and transparency. However, these designs are deemed less accurate in estimating the toxicity rates and allocate more patients to sub-optimal dose levels, which can potentially result in high costs and long trial durations. Selecting an optimal dose for subsequent trials to provide a higher chance of sufficient effectiveness in treating patients is an important objective for phase I trials. Many new dose escalation methods have been proposed in recent years to improve operating characteristics of phase I trials, and they were proved to outperform the 3+3 design (13, 129–132) in terms of accuracy to identify the MTD, under or overdose control, and trial duration. Unlike the rule-based designs that are easy to understand and implement for non-statisticians, these innovative model-based designs remain unfamiliar to, and are used less by, large parts of the clinical community due to the fact that they are complicated and require computations and advanced statistics. Although software packages are available for these model-based designs, they don’t fit all of the trial characteristics and that makes implementation challenging. To promote the use of model-based designs, publications need to more clearly enumerate the statistical properties and aspects of the trials (133). The new model-assisted designs such as BOIN design have comparable performance to the model-based designs with the characteristics of simplicity and flexibility (7, 15). In recent years, model-based and model–assisted designs extended to monitor efficacy in addition to safety in seamless phase I/II trials (134–136). Nevertheless, they may not be commonly known to the peers and applied in the clinical trials.

Conclusion

We performed a review of phase I clinical trials for CAR-T immunotherapy from a range of journals. The quality of reporting phase I CAR-T trials was poor even in high IF journals. Standardizing the criteria and basic elements of publications are critical to ensure high-quality reports of phase I clinical trials. As previously stated, many of the reviewed trials still use rule-based designs, such as the traditional 3+3 design. With the quick development and high costs of CAR-T cell therapy, adoption of advanced designs such as model-based and model-assisted should increase to improve efficiency of clinical trials and expedite the drug development process.