Software to compute and conduct sequential Bayesian phase I or II dose-ranging clinical trials with stopping rules

doi:10.1016/S0169-2607(02)00120-7

Computer Methods and Programs in Biomedicine

Volume 72, Issue 2, October 2003, Pages 117-125

https://doi.org/10.1016/S0169-2607(02)00120-7 Get rights and content

Abstract

The aim of dose-ranging phase I (resp. phase II) clinical trials is to rapidly identify the maximum tolerated dose (MTD) (resp., minimal effective dose (MED)) of a new drug or combination. For the conduct and analysis of such trials, Bayesian approaches such as the Continual Reassessment Method (CRM) have been proposed, based on a sequential design and analysis up to a completed fixed sample size. To optimize sample sizes, Zohar and Chevret have proposed stopping rules (Stat. Med. 20 (2001) 2827), the computation of which is not provided by available softwares. We present in this paper a user-friendly software for the design and analysis of these Bayesian Phase I (resp. phase II) dose-ranging Clinical Trials (BPCT). It allows to carry out the CRM with stopping rules or not, from the planning of the trial, with choice of model parameterization based on its operating characteristics, up to the sequential conduct and analysis of the trial, with estimation at stopping of the MTD (resp. MED) of the new drug or combination.

Introduction

The purpose of dose-ranging phase I clinical trials is to rapidly identify the maximal tolerated dose (MTD) of a new drug or association. For the conduct and analysis of such trials in cancer patients, the Continual Reassessment Method (CRM) has been proposed [1]. Based on a sequential design, it provides a Bayesian estimation of the MTD from a fixed set of dose levels, once the fixed sample size has been reached. It has been widely studied and slightly modified [1], [2], [3], [4], [5], [6], [7], [8], [9], either by treating patients in cohort of c (>1) patients, by administering the lowest dose level to the first cohort, or by extending the approach to phase II dose-ranging trials, i.e., when estimating the minimal effective dose (MED) of a new drug or combination. By contrast, little attention has been devoted to address ethical concerns such as to stop the trial early if it is likely that no dose is safe (resp. efficacious), or statistical concerns such as to treat a sufficiently large number of patients to estimate the probability of toxicity (resp. efficacy) at the selected dose with a given level of reliability. Recently, stopping criteria were developed for the CRM to address these issues [10]. However, even if a software has already been proposed to implement the CRM [11], it does not provide computation of stopping rules. Therefore, we developed a new software for the CRM that allows to compute these stopping criteria [10], together with a Graphic User Interface (GUI) to simplify the use of the software for non initiated users.

In the next section, we briefly present the CRM with the stopping criteria. The software itself is described in Section 3. Two examples are implemented in 4 Example: stopping for wrong dose scale, 5 Example: stopping for suitable estimation of the MTD. Finally, some concluding remarks are given in Section 6.

Section snippets

Model

In the further sections, we will place ourselves in the setting of a cancer phase I dose-ranging clinical trial, although it can be easily extended to other settings or phase II clinical trials.

Program description

This program has been written in C language, and compiled on PC compatible hardware for Microsoft Windows (either NT 4.0 or Windows 98) platforms. The program is to be executed with Tcl/Tk interface (version 8.2).

It allows to implement, through the use of three options, the planning, conduct and analysis of Bayesian Phase I (phase II) dose-ranging Clinical Trials (BPCT), from the initial choice of model parameterization (options 1 and 2), to the sequential computation of the dose level to be

Example: stopping for wrong dose scale

We illustrated the features of BPCT on a previously published dose-ranging phase II clinical trial [15]. This trial aimed at assessing the MED of midazolam, defined as the dose level associated with 0.10 failure probability. Six dose levels were initially chosen, and associated with the following guesses of failure probabilities: p₁=0.01, p₂=0.10, p₃=0.20, p₄=0.40, p₅=0.60 and p₆=0.70. The cohort size was one and the total number of enrolled patients was 16.

We first performed an operating

Example: stopping for suitable estimation of the MTD

To illustrate now what could happen in a favorable case where the MTD is one of the tested dose levels, we simulated a phase I dose-ranging as follows. We considered six dose levels, with actual toxicity probabilities of 0.05, 0.10, 0.30, 0.50, 0.60 and 0.65, respectively. The aim of the trial was to estimate the MTD, defined as the 30th percentile of dose–toxicity relationship (i.e., dose level 3). The sample size was 25, the size cohort was 1 and initial guesses of toxicity probabilities were

Concluding remarks

The Bayesian stopping rules associated with the CRM give clinicians and biostatisticians an accurate tool for decision making during a dose-ranging trial. These stopping rules should be used in different ways during the trial. The first two rules should be used at the beginning of the trial and prior to the other rules to point out a wrong dose scale. The third one allows the user to predict if the administered dose level is expected to not change in future patients. Finally, the last four

References (15)

M. Bensadon et al.
Integral evaluation for continual reassessment method
Comput. Methods Programs Biomed.
(1994)
J. O'Quigley et al.
Continual reassessment method: a practical design for phase I clinical trials in cancer
Biometrics
(1990)
J. O'Quigley et al.
Methods for dose finding studies in cancer clinical trials: a review and results of a Monte Carlo study
Stat. Med.
(1991)
S. Chevret
The continual reassessment method in cancer phase I clinical trials: a simulation study
Stat. Med.
(1993)
D. Faries
Practical modifications of the continual reassessment method for phase I cancer clinical trials
J. Biopharm. Stat.
(1994)
E.L. Korn et al.
A comparison of two phase I trial designs
Stat. Med.
(1994)
S.N. Goodman et al.
Some practical improvements in the continual reassessment method for phase I studies
Stat. Med.
(1995)

There are more references available in the full text version of this article.

Cited by (30)

Effective concentration of levobupivacaine and ropivacaine in 80% of patients receiving epidural analgesia (EC80) in the first stage of labour: A study using the Continual Reassessment Method
2018, Anaesthesia Critical Care and Pain Medicine
Citation Excerpt :
We aimed to determine the EC80 of ropivacaine or levobupivacaine i.e., the concentration leading to satisfactory analgesia in 80% of the patients. EC80 was first modelled using the Continual Reassessment Method (CRM) with the software BPCT [11,17,18]. In a second step, the whole probability-dose curve (logistic equation) was fitted to the data using non-linear regression with NONMEM VI, level 12 (Appendix).
A comparison of the effective dose in 50% of patients (ED50) has suggested that the potency of levobupivacaine lies between that of bupivacaine and ropivacaine. However, for clinical purposes, knowledge and use of doses close to the ED95 are more relevant. This study was designed to determine the EC80 (effective concentration) for both epidural levobupivacaine and ropivacaine using the Continual Reassessment Method (CRM) during obstetric analgesia.
In this double-blind randomised study, term parturients were included by cohorts of 6 if cervical dilatation was ≤ 5 cm and visual analogue pain score (VAPS) > 30 mm. Efficacy was defined by a decrease of VAPS to a value ≤ 10, thirty minutes after epidural injection of 20 mL of levobupivacaine or ropivacaine. The first cohort received the lowest dose. Every next cohort received a dose according to the response's probability calculated using a Bayesian method, incorporating data from all consecutive previous patients. In addition, a logistic equation was fitted a posteriori to the whole data set to determine the whole dose-probability curve.
Fifty-four patients were enrolled. Levobupivacaine 0.17% and ropivacaine 0.2% gave probabilities of success of 82% and 72% respectively. By fitting the logistic model to the data, the concentration leading to a probability of 0.8 (EC80) was 0.14% for levobupivacaine and 0.24% for ropivacaine while the EC50 were 0.09% for levobupivacaine and 0.17% for ropivacaine, respectively.
This study suggests that epidural levobupivacaine used as the sole drug for labour analgesia has an EC80 lower than that of ropivacaine.
Pomalidomide, cyclophosphamide, and prednisone for relapsed/refractory multiple myeloma: A multicenter phase 1/2 open-label study
2013, Blood
Citation Excerpt :
A recent survey on 286 relapsed myeloma patients who were refractory to bortezomib and relapsed/refractory to an immunomodulatory drug reported at least MR in 44% of patients, including 32% PR, and median event-free survival and OS of 5 and 9 months, respectively.3 Initial reports demonstrated encouraging activity with pomalidomide alone or with low-dose dexamethasone in relapsed/refractory myeloma.6-12,24,25 In a phase 3 study comparing pomalidomide (4 mg/d for 21 days in a 28-day cycle) plus low-dose dexamethasone (160 mg monthly) with high-dose dexamethasone (480 mg monthly) in multirelapsed/refractory myeloma, median PFS was 3.9 vs 2 months (P < .001), and median OS was not reached vs 8.5 months (P < .001), in the pomalidomide and the high-dose dexamethasone groups, respectively.26
We performed a phase 1/2 trial to determine the maximum tolerated dose (MTD) of pomalidomide and to explore its efficacy when combined with cyclophosphamide-prednisone in relapsed/refractory myeloma patients. Pomalidomide was given at 1 to 2.5 mg/d, cyclophosphamide at 50 mg every other day, prednisone at 50 mg every other day, for 6 28-day cycles, followed by pomalidomide-prednisone maintenance therapy. Thromboprophylaxis was recommended. Sixty-nine patients were enrolled, 55 received the MTD (2.5 mg/d) and were evaluated. Best responses included complete response in 3 patients (5%), very good partial response in 10 (18%), partial response in 15 (27%), minimal response in 11 (20%), stable disease in 15 (27%), and progressive disease in 1 (3%), for an overall response rate of 51%. The median time-to-response was 1.83 months. After a median follow-up of 14.8 months, median progression-free survival was 10.4 months and 1-year overall survival was 69%. At the MTD, grade 3 to 4 toxicities included anemia (9%), thrombocytopenia (11%), neutropenia (42%), neurologic events (7%), dermatologic events (7%), and thromboembolism (2%). Grade 3 to 5 infections occurred in 5 patients (9%). Five patients (9%) discontinued treatment for toxicity. New grade 3 to 4 adverse events were low during maintenance. Pomalidomide-cyclophosphamide-prednisone is safe and effective in relapsed/refractory myeloma patients. This trial was registered at www.clinicaltrials.gov as #NCT01166113.
17 Early Phase Clinical Trials: Phases I and II
2007, Handbook of Statistics
Citation Excerpt :
TITE–CRM has also been modified to allow inclusion of late-onset toxicities, a useful feature when the agents involve are known to have lingering effects (Cheung and Chappell, 2000). Software for CRM trials and other practical considerations are presented by Zohar et al. (2003) and Piantadosi et al. (1998). An extension of the EWOC method by Tighiouart et al. (2005) has shown that efficient estimation of the MTD using the EWOC method can improve the safety profile of the trial as a whole by assuming a prior negative correlation structure for the two primary parameters of the dose–toxicity model, the probability of DLT at the initial dose and the MTD.
A clinical trial is a planned experiment on human subjects to assess one or more potentially beneficial therapies. A central problem in early phase clinical trials is the limited knowledge on the new treatment of interest. As a consequence, extreme caution needs to be taken in study designs to minimize the risk of participants while maximizing the benefit. This chapter provides an overview of the recent advances in statistical designs of early phase clinical trials. Since formal statistical methods for phase I and II trials have been mostly developed for cancer drugs, a considerable portion of this chapter addresses statistical issues in this particular setting.
The limited efficacy of tramadol in postoperative patients: A study of ED80 using the continual reassessment method
2008, Anesthesia and Analgesia
The aim of this study was to reevaluate the efficacy of tramadol for postoperative analgesia and to determine its ED₈₀ (the clinical dose for which 80% of the patients had their pain adequately relieved) using the Continual Reassessment Method. Because the preliminary results of the first 24 patients were contradictory to the literature, we performed a second trial to verify and validate our data.
The study was double-blind and prospective. Participants were allocated to a dose of tramadol by 3-patient cohorts, in order of inclusion. The dose of tramadol received in each cohort was determined by the reaction of all previous patients. Five doses were chosen before beginning, with a probability of a positive reaction associated with each: 60 (0.4), 100 (0.55), 140 (0.7), 190 (0.8), and 260 mg (0.9). Tramadol was considered effective if the numeric pain scale was ⩾3/10 at T30.
The effective dose in 80% of patients was 260 mg for both trials. The probability of success of the 260 mg dose was 0.699 (95% credibility interval, 0.471–0.874) and 0.657 (95% credibility interval, 0.437–0.853) for trial 1 and trial 2, respectively.
Tramadol used as a sole drug cannot be considered the drug of choice after moderately painful surgery. The doses needed to relieve pain in 80% of patients are much larger than the usual dose of 100 mg. The Continual Reassessment Method allowed us to determine the ED₈₀ of tramadol with a limited number of patients.
Romidepsin-CHOEP followed by high-dose chemotherapy and stem-cell transplantation in untreated Peripheral T-Cell Lymphoma: results of the PTCL13 phase Ib/II study
2023, Leukemia
Early stopping in seamless phase I/II clinical trials
2021, Pharmaceutical Statistics

View all citing articles on Scopus

View full text

Software to compute and conduct sequential Bayesian phase I or II dose-ranging clinical trials with stopping rules

Abstract

Introduction

Section snippets

Model

Program description

Example: stopping for wrong dose scale

Example: stopping for suitable estimation of the MTD

Concluding remarks

Comput. Methods Programs Biomed.

Continual reassessment method: a practical design for phase I clinical trials in cancer

Biometrics

Methods for dose finding studies in cancer clinical trials: a review and results of a Monte Carlo study

Stat. Med.

The continual reassessment method in cancer phase I clinical trials: a simulation study

Stat. Med.

Practical modifications of the continual reassessment method for phase I cancer clinical trials

J. Biopharm. Stat.

A comparison of two phase I trial designs

Stat. Med.

Some practical improvements in the continual reassessment method for phase I studies

Stat. Med.