Dose-finding design using mixed-effect proportional odds model for longitudinal graded toxicity data in phase I oncology clinical trials

An Overview of Regression Models for Adverse Events Analysis

Over the last few years, several review articles described the adverse events analysis as sub-optimal in clinical trials. Indeed, the context surrounding adverse events analyses often imply an overwhelming number of events, a lack of power to find associations, but also a lack of specific training regarding those complex data. In randomized controlled trials or in observational studies, comparing the occurrence of adverse events according to a covariable of interest (e.g., treatment) is a recurrent question in the analysis of drug safety data, and adjusting other important factors is often relevant. This article is an overview of the existing regression models that may be considered to compare adverse events and to discuss model choice regarding the characteristics of the adverse events of interest. Many dimensions may be relevant to compare the adverse events between patients, (e.g., timing, recurrence, and severity). Recent efforts have been made to cover all of them. For chronic treatments, the occurrence of intercurrent events during the patient follow-up usually needs the modeling approach to be adapted (at least with regard to their interpretation). Moreover, analysis based on regression models should not be limited to the estimation of relative effects. Indeed, absolute risks stemming from the model should be presented systematically to help the interpretation, to validate the model, and to encourage comparison of studies.

Dose finding studies for therapies with late-onset toxicities: A comparison study of designs

An objective of phase I dose-finding trials is to find the maximum tolerated dose; the dose with a particular risk of toxicity. Frequently, this risk is assessed across the first cycle of therapy. However, in oncology, a course of treatment frequently consists of multiple cycles of therapy. In many cases, the overall risk of toxicity for a given treatment is not fully encapsulated by observations from the first cycle, and hence it is advantageous to include toxicity outcomes from later cycles in phase I trials. Extending the follow up period in a trial naturally extends the total length of the trial which is undesirable. We present a comparison of eight methods that incorporate late onset toxicities while not extensively extending the trial length. We conduct simulation studies over a number of scenarios and in two settings; the first setting with minimal stopping rules and the second setting with a full set of standard stopping rules expected in such a dose finding study. We find that the model-based approaches in general outperform the model-assisted approaches, with an interval censored approach and a modified version of the time-to-event continual reassessment method giving the most promising overall performance in terms of correct selections and trial length. Further recommendations are made for the implementation of such methods.

A Road Map for Designing Phase I Clinical Trials of Radiotherapy-Novel Agent Combinations

Radiotherapy has proven efficacy in a wide range of cancers. There is growing interest in evaluating radiotherapy-novel agent combinations and a drive to initiate this earlier in the clinical development of the novel agent, where the scientific rationale and preclinical evidence for a radiotherapy combination approach are high. Optimal design, delivery, and interpretation of studies are essential. In particular, the design of phase I studies to determine safety and dosing is critical to an efficient development strategy. There is significant interest in early-phase research among scientific and clinical communities over recent years, at a time when the scrutiny of the trial methodology has significantly increased. To enhance trial design, optimize safety, and promote efficient trial conduct, this position paper reviews the current phase I trial design landscape. Key design characteristics extracted from 37 methodology papers were used to define a road map and a design selection process for phase I radiotherapy-novel agent trials. Design selection is based on single- or dual-therapy dose escalation, dose-limiting toxicity categorization, maximum tolerated dose determination, subgroup evaluation, software availability, and design performance. Fifteen of the 37 designs were identified as being immediately accessible and relevant to radiotherapy-novel agent phase I trials. Applied examples of using the road map are presented. Developing these studies is intensive, highlighting the need for funding and statistical input early in the trial development to ensure appropriate design and implementation from the outset. The application of this road map will improve the design of phase I radiotherapy-novel agent combination trials, enabling a more efficient development pathway.

Bayesian interval-based oncology dose-finding design with repeated quasi-continuous toxicity model

In oncology dose-finding clinical trials, the key to accurately estimating the maximum tolerated dose (MTD) is to use all data efficiently given small sample sizes. Currently, popular designs dichotomize adverse events of various types and grades that occur within the first treatment cycle into binary toxicity outcomes of dose-limiting toxicity (DLT) events. Such compression of toxicity data from multiple treatment cycles causes huge loss of information, often resulting in MTD estimation with large bias and variance. To improve this, a continuous endpoint (the total toxicity profile, TTP) was proposed to incorporate adverse event types and grades. The Bayesian Repeated Measures Design (RMD) was further developed by Yin et al. (2017) to account for the cumulative toxicity information from multiple treatment cycles. However, the existing RMD method selects the dose that minimizes the loss function based on point estimates, which may generate inconsistent results due to small sample sizes in phase I trials. To reduce the variability in dose escalation decision-making, we propose an improved repeated measures design with an interval-based decision rule that selects the dose with the highest posterior probability of falling in a pre-specified target toxicity interval. Through comprehensive simulations, we compared this proposed design with the existing RMD design, along with well-established DLT-based designs such as Continual Reassessment Method (CRM) and Bayesian Logistic Regression Model (BLRM). The results demonstrated that our proposed design outperforms all other designs in terms of accurately identifying the MTD and assigning fewer patients to sub-therapeutic or overly toxic doses.

Proportional odds assumption for modeling longitudinal ordinal multiple toxicity outcomes in dose finding studies of targeted agents: A pooled analysis of 54 studies

BACKGROUND

Data generated by phase I trials is richer than the classical binary DLT measured at the first cycle used as primary endpoints. Several works developed designs for more informative endpoints, e.g. ordinal toxicity grades and/or longitudinal data which relied however on strong assumptions, in particular the proportional odds (PO) assumption.

METHODS

We evaluated this PO assumption for the dose and cycle on a large database of individual patient data from 54 phase I clinical trials of molecularly targeted agents. The PO model is a specific case of the continuation ratio logit model (CRLM) with null parameters. We compared the PO and CRLM models using the widely applicable information criterion (WAIC). We considered a longitudinal multivariate ordinal toxicity outcome (cutaneous, digestive, hematological, general disorders, and other toxicities).

RESULTS

WAIC suggested that the CRLM model (WAIC = 30911.58) outperformed the PO model (WAIC = 31432.10). Deviance from PO assumption for dose was observed for digestive and general disorder toxicities. There was moderate cycle effect with slight deviance from PO assumption for the other type of toxicity.

CONCLUSIONS

Designs based on PO for dose should be a useful tool for drug with low expected digestive or general disorder toxicity dose-related incidence.

Dose-finding designs for cumulative toxicities using multiple constraints

This article addresses the concern regarding late-onset dose-limiting toxicities (DLT), moderate toxicities below the threshold of a DLT and cumulative toxicities that may lead to a DLT, which are mostly disregarded or handled in an ad hoc manner when determining the maximum tolerated dose (MTD) in dose-finding cancer clinical trials. An extension of the Time-to-Event Continual Reassessment Method (TITE-CRM) which allows for the specification of toxicity constraints on both DLT and moderate toxicities, and can account for partial information is proposed. The method is illustrated in the context of an Erlotinib dose-finding trial with low DLT rates, but a significant number of moderate toxicities leading to treatment discontinuation in later cycles. Based on simulations, our method performs well at selecting the dose level that satisfies both the DLT and moderate-toxicity constraints. Moreover, it has similar probability of correct selection compared to the TITE-CRM when the true MTD based on DLT only and the true MTD based on grade 2 or higher toxicities alone coincide, but reduces the probability of recommending a dose above the MTD.

Bayesian Adaptive Design for Finding the Maximum Tolerated Sequence of Doses in Multicycle Dose-Finding Clinical Trials

Purpose

Statistical designs for traditional phase I dose-finding trials consider dose-limiting toxicity in the first cycle of treatment. In reality, patients often go through multiple cycles of treatment and may experience toxicity events in more than one cycle. Therefore, it is desirable to identify the maximum tolerated sequence of three doses across three cycles of treatment.

Methods

Motivated by a three-cycle dose-finding clinical trial for a rare cancer with a JAK inhibitor, we proposed and implemented a simple Bayesian adaptive dose-cycle finding (BaSyc) design that allows intercycle and intrapatient dose modification. Because of the patient-specific dosing strategy over cycles, the BaSyc design is suited as a method in precision oncology.

Results

BaSyc is simple and transparent because its algorithm can be summarized as two tabulated decision rules before the trial starts, allowing physicians to visually examine these rules. In addition, BaSyc employs a time-saving enrollment scheme that speeds up the trial. Extensive simulation studies show that BaSyc has desirable operating characteristics in identifying the maximum tolerated sequence.

Conclusion

The BaSyc design provides a first-of-kind multicycle approach for dose finding and will likely lead to better and safer patient care and drug development.

Early Phase Clinical Trial Designs - State of Play and Adapting for the Future

The process of anti-cancer drug development is complex, with high attrition rates. Factors that may optimise this process include well-constructed and relevant pre-clinical testing and use of biomarkers for patient selection. However, the design of early phase clinical trials will probably play a vital role in both the robust clinical investigation of new targeted therapies and in streamlining drug development. In this overview, we assess current concepts in phase I clinical trials, highlighting issues and opportunities to improve their meaningfulness. The particular challenge of how to design combination trials is addressed, with focus on the potential of new adaptive and model-based designs.

Adaptive phase I/II clinical trials for drug combination assessment in oncology using the outcomes of each cycle

Many new anticancer agents can be combined with existing drugs, as combining a number of drugs may be expected to have a better therapeutic effect than monotherapy owing to synergistic effects. Furthermore, to drive drug development and to reduce the associated cost, there has been a growing tendency to combine these as phase I/II trials. With respect to phase I/II oncology trials for the assessment of dose combinations, in the existing methodologies in which efficacy based on tumor response and safety based on toxicity are modeled as binary outcomes, it is not possible to enroll and treat the next cohort of patients unless the best overall response has been determined in the current cohort. Thus, the trial duration might be potentially extended to an unacceptable degree. In this study, we proposed a method that randomizes the next cohort of patients in the phase II part to the dose combination based on the estimated response rate using all the available observed data upon determination of the overall response in the current cohort. We compared the proposed method to the existing method using simulation studies. These demonstrated that the percentage of optimal dose combinations selected in the proposed method is not less than that in the existing method and that the trial duration in the proposed method is shortened compared to that in the existing method. The proposed method meets both ethical and financial requirements, and we believe it has the potential to contribute to expedite drug development.

Challenges and Innovations in Phase I Dose-Finding Designs for Molecularly Targeted Agents and Cancer Immunotherapies

Phase I oncology trials are designed to identify a safe dose with an acceptable toxicity profile. In traditional phase I dose-finding design, the dose is typically determined based on the probability of severe toxicity observed during the first treatment cycle. The recent development of molecularly targeted agents and cancer immunotherapies call for new innovations in phase I designs, because of prolonged treatment cycles often involved. Various phase I designs using toxicity and efficacy endpoints from multiple treatment cycles have been developed for these new treatment agents. Here, we will review the novel endpoints and designs for the phase I oncology clinical trials.

Pharmacogenetics-Guided Phase I Study of Capecitabine on an Intermittent Schedule in Patients with Advanced or Metastatic Solid Tumours

The FDA-approved starting dosage of capecitabine is 1,250 mg/m², and market research indicates that U.S. physicians routinely prescribe 1,000 mg/m². Retrospective analyses however report reduced toxicity and efficacy in a subset of patients with the 3R/3R genotype of the thymidylate synthase gene enhancer region (TSER). This study sought to develop TSER genotype-specific guidelines for capecitabine dosing. Capecitabine was dose-escalated in advanced and/or metastatic cancer patients with TSER 3R/3R (Group A; N = 18) or 2R/2R + 2R/3R (Group B; N = 5) from 1,250 to 1,625 mg/m² b.i.d., every 2 weeks on/1 week off for up to 8 cycles. Parent and metabolites pharmacokinetics, adverse events, and tumour response were assessed. The maximum tolerated and recommended doses in 3R/3R patients are 1,625 mg/m² and 1,500 mg/m². At 1,500 mg/m², one in nine 3R/3R patients experienced a dose-limiting toxicity. Dosing guidelines for 2R/2R + 2R/3R remain undetermined due to poor accrual. The results indicate that 3R/3R patients may be amenable to 1,500 mg/m² b.i.d. on an intermittent schedule, and is the first to prospectively validate the utility of TSER pharmacogenetic-testing before capecitabine treatment.

Innovations for phase I dose-finding designs in pediatric oncology clinical trials

Phase I oncology clinical trials are designed to identify the optimal dose that will be recommended for phase II trials. In pediatric oncology, the conduct of those trials raises specific challenges, as the disease is rare with limited therapeutic options. In addition, the tolerance profile is known from adult trials. This paper provides a review of the major recent developments in the design of these trials, inspired by the need to cope with the specific challenges of dose finding in cancer pediatric oncology. We reviewed simulation studies comparing designs dedicated to address these challenges. We also reviewed the design used in published dose-finding trials in pediatric oncology over the period 2009-2014. Three main fields of innovation were identified. First, designs that were developed in order to relax the rules for more flexible inclusions. Second, methods to incorporate data emerging from adult studies. Third, designs accounting for toxicity evaluation at repeated cycles in pediatric oncology. In addition to this overview, we propose some further directions for designing pediatric dose-finding trials.

Towards using a full spectrum of early clinical trial data: a retrospective analysis to compare potential longitudinal categorical models for molecular targeted therapies in oncology

Following the pattern of phase I clinical trials for cytotoxic drugs, dose-finding clinical trials in oncology of molecularly targeted agents (MTA) aim at determining the maximum tolerated dose (MTD). In classical phase I clinical trials, MTD is generally defined by the number of patients with short-term major treatment toxicities (usually called dose-limiting toxicities, DLT), occurring during the first cycle of study treatment (e.g. within the first 3weeks of treatment). However, S. Postel-Vinay (2011) highlighted that half of grade 3 to 4 toxicities, usually considered as DLT, occur after the first cycle of MTA treatment. In addition, MTAs could induce other moderate (e.g. grade 2) toxicities which could be taken into account depending on their clinical importance, chronic nature and duration. Ignoring these late toxicities may lead to an underestimation of the drug toxicity and to wrong dose recommendations for phase II and III clinical trials. Some methods have been proposed, such as the time-to-event continuous reassessment method (Cheung 2000 and Mauguen 2011), to take into account the late toxicities. We suggest approaches based on longitudinal models (Doussau 2013). We compare several models for longitudinal data, such as transitional or marginal models, to take into account all relevant toxicities occurring during the entire length of the patient treatment (and not just the events within a predefined short-term time-window). These models allow the statistician to benefit from a larger amount of safety data which could potentially improve that accuracy in MTD assessment.

Statistical controversies in clinical research: requiem for the 3 + 3 design for phase I trials

BACKGROUND

More than 95% of published phase I trials have used the 3 + 3 design to identify the dose to be recommended for phase II trials. However, the statistical community agrees on the limitations of the 3 + 3 design compared with model-based approaches. Moreover, the mechanisms of action of targeted agents strongly challenge the hypothesis that the maximum tolerated dose constitutes the optimal dose, and more outcomes including clinical and biological activity increasingly need to be taken into account to identify the optimal dose.

PATIENTS AND METHODS

We review key elements from clinical publications and from the statistical literature to show that the 3 + 3 design lacks the necessary flexibility to address the challenges of targeted agents.

RESULTS

The design issues raised by expansion cohorts, new definitions of dose-limiting toxicity and trials of combinations are not easily addressed by the 3 + 3 design or its extensions.

CONCLUSIONS

Alternative statistical proposals have been developed to make a better use of the complex data generated by phase I trials. Their applications require a close collaboration between all actors of early phase clinical trials.

Multivariate Markov models for the conditional probability of toxicity in phase II trials

In addition to getting a preliminary assessment of efficacy, phase II trials can also help to determine dose(s) that have an acceptable toxicity profile over repeated cycles as well as identify subgroups with particularly poor toxicity profiles. Correct modeling of the dose-toxicity relationship in patients receiving multiple cycles of the same dose in oncology trials is crucial. A major challenge lies in taking advantage of the conditional nature of data collection, that is each cycle is observed conditional on having no previous toxicities on earlier cycles. We develop a novel and parsimonious model for the probability of toxicity during a kth cycle of therapy, conditional on not seeing toxicity in any of the k-1 previous cycles using a Markov model, hereafter we refer to these probabilities as conditional probabilities of toxicity. Our model allows the conditional probability of toxicity to depend on randomized dose group, cumulative dose from prior cycles, a measure of how consistently a patient responds to the same dose exposure and individual risk factors influencing the ability to tolerate the treatment regimen. Simulations studying finite sample properties of the model are given. Finally, the approach is demonstrated in a phase II trial studying two dose levels of ifosfamide plus doxorubicin and granulocyte colony-stimulating factor in soft tissue sarcoma patients over four cycles. The Markov model provides correct estimates of the probabilities of toxicity in finite sample simulations. It also correctly models the data from the phase II clinical trial, and identifies particularly high cumulative toxicity in females.

Dose finding with longitudinal data: simpler models, richer outcomes

Phase I oncology clinical trials are designed to identify the optimal dose that will be recommended for phase II trials. This dose is typically defined as the dose associated with a certain probability of severe toxicity at cycle 1, although toxicity is repeatedly measured over cycles on an ordinal scale. Recently, a proportional odds mixed-effect model for ordinal outcomes has been proposed to (i) identify the optimal dose accounting for repeated events and (ii) to provide some framework to explore time trend. We compare this approach to a method based on repeated binary variables and to a method based on an under-parameterized model of the dose-time toxicity relationship. We show that repeated binary and ordinal outcomes both improve the accuracy of dose-finding trials in the same proportion; ordinal outcomes are, however, superior to detect time trend even in the presence of nonproportional odds models. Moreover, less parameterized models led to the best operating characteristics. These approaches are illustrated on two dose-finding phase I trials. Integration of repeated measurements is appealing in phase I dose-finding trials.

Adaptive Phase I clinical trial design using Markov models for conditional probability of toxicity

Many Phase I trials in oncology involve multiple-dose administrations on the same patient over multiple cycles, with a typical cycle lasting 3 weeks and having about six cycles per patient with a goal to find the maximum tolerated dose (MTD) and study the dose-toxicity relationship. A patient's dose is unchanged over the cycles and the data are reduced to a binary endpoint and the occurrence of a toxicity and analyzed by considering the toxicity either from the first dose or from any cycle on the study. In this article, an alternative approach allowing an assessment of toxicity from each cycle and dose variations for patient over cycles is presented. A Markov model for the conditional probability of toxicity on any cycle given no toxicity in previous cycles is formulated as a function of the current and previous doses. The extra information from each cycle provides more precise estimation of the dose-toxicity relationship. Simulation results demonstrating gains in using the Markov model as compared to analyses of a single binary outcome are presented. Methods for utilizing the Markov model to conduct a Phase I study, including choices for selecting doses for the next cycle for each patient, are developed and presented via simulation.

Early phase clinical trials to identify optimal dosing and safety

The purpose of early stage clinical trials is to determine the recommended dose and toxicity profile of an investigational agent or multi-drug combination. Molecularly targeted agents (MTAs) and immunotherapies have distinct toxicities from chemotherapies that are often not dose dependent and can lead to chronic and sometimes unpredictable side effects. Therefore utilizing a dose escalation method that has toxicity based endpoints may not be as appropriate for determination of recommended dose, and alternative parameters such as pharmacokinetic or pharmacodynamic outcomes are potentially appealing options. Approaches to enhance safety and optimize dosing include improved preclinical models and assessment, innovative model based design and dose escalation strategies, patient selection, the use of expansion cohorts and extended toxicity assessments. Tailoring the design of phase I trials by adopting new strategies to address the different properties of MTAs is required to enhance the development of these agents. This review will focus on the limitations to safety and dose determination that have occurred in the development of MTAs and immunotherapies. In addition, strategies are proposed to overcome these challenges to develop phase I trials that can more accurately define the recommended dose and identify adverse events.

Phase 1 trial design: is 3 + 3 the best?

BACKGROUND

Concerns have been recognized about the operating characteristics of the standard 3 + 3 dose-escalation design. Various innovative phase 1 trial designs have been proposed to address the issues and new challenges posed by molecularly targeted agents. However, in spite of these proposals, the conventional design is still the most widely utilized.

METHODS

A review of the literature of phase 1 trials and relevant statistical studies was performed.

RESULTS

Beyond statistical simulations, sparse clinical data exist to support or refute many of the shortcomings ascribed to the 3 + 3 rule method. Data from phase 1 trials demonstrate that traditional designs identified the correct dose and relevant toxicities with an acceptable level of precision in some instances; however, no single escalation method was proven superior in all circumstances.

CONCLUSIONS

Design selection should be guided by the principle of slow escalation in the face of toxicity and rapid dose increases in the setting of minimal or no adverse events. When the toxicity of a drug is uncertain or a narrow therapeutic window is suggested from preclinical testing, then a conservative 3 + 3 method is generally appropriate. However, if the therapeutic window is wide and the expected toxicity is low, then rapid escalation with a novel rule- or model-based design should be employed.

A new approach to integrate toxicity grade and repeated treatment cycles in the analysis and reporting of phase I dose-finding trials

BACKGROUND

Safety assessment beyond the dose-limiting toxicity evaluation period provides relevant information to define the recommended phase II dose (RP2D) of a new treatment. We retrospectively analyzed three phase I trials to illustrate two indicators: per-cycle probability of graded toxicity and cumulative probability of severe toxicity over the treatment period.

PATIENTS AND METHODS

Data were collected from two continual reassessment method (CRM) trials (T1: aviscumine in solid tumors with short time on treatment; T2: erlotinib + radiotherapy in brainstem gliomas with longer time on treatment) and one 3 + 3 design (T3: liposomal doxorubicin + cyclophosphamide combination in ovarian carcinoma). The probability of severe and moderate or severe toxicity per cycle was estimated at each dose level with mixed proportional odds model. The cumulative probability of severe toxicity was also estimated with the time-to-event CRM.

RESULTS

Eighty-three patients were included in the three trials; 94, 96 and 72 treatment cycles were administered, in T1, T2 and T3, respectively. Moderate toxicities were at least twice as frequent as severe toxicities. An increased probability of toxicity over time was detected in T3 [P = 0.04; per-cycle probability of severe toxicity: 27% (cycle 1) to 59% (cycle 6) at the RP2D]. At the RP2D, 37% of patients experienced at least one severe toxicity over the first six cycles in T2, and 78% in T3.

CONCLUSIONS

Dedicated methods can be used to analyze toxicities from all cycles of treatment. They do not delay accrual and should be integrated in the analysis and reporting of phase I dose-finding trials.