SAE: An R package for early stopping rules in clinical trials
Introduction
The monitoring of Serious Adverse Events (SAE) during a clinical trial is becoming a major preoccupation for sponsors of clinical trials. SAEs are life-threatening events of a patient such as grade 4 clinical (for example, diarrhea …) or biological toxicity (for example, neutropenia …), or which cause a clinically significant invalidity or incapacity. Toxic deaths of course are also considered as SAE.
The investigator is required to declare each SAE within 24–48 h to the sponsor who transmits the information to the appropriate health authorities (for example AFSSAPS in France). The use of this information is essential in the decision to prematurely stop a trial or one arm of the trial if the treatment is considered too toxic.
Multi-stage clinical trial designs have been developed for defining stopping rules in terms of efficacy and safety in phase II trials [1], but they rely on pre-defined fixed sample sizes at each stage. In phase III designs, group sequential interim analyses for efficacy are planned after a fixed number of patients or events have occurred. Stopping rules for excess toxicity are not commonly predefined in the protocol, except for ad hoc statements such as, for example, not more than 5% or 3%. Sequential stopping rules for safety are not usually incorporated into the protocol design of clinical trials. However, in the case of an unexpected high frequency of serious adverse events (SAE), statistical methods are needed to help in the decision making process as to continuation of accrual to the trial.
Such methods have been developed by defining stopping rules after each observed SAE [2], [3]. Our approach extends the ideas of sequential surveillance of SAEs by comparing the total number of patients included to the maximal number of patients that satisfies maximal acceptable SAE criteria while preserving the nominal type I error [4]. The proposed methods are more adapted to phase III clinical trials, since in this situation only ad hoc meetings are organized for prompt decision making as to the continuation of the trial.
This paper describes the implementation in R language of both approaches directly during the planning phase of the trial or after the observation of each SAE. Then, decision rules are written into the protocol as “multi-stage early stopping rules”. The following section briefly describes the method implemented. In the third section, we present a description of the SAE package. Different functions of the package are detailed as well as input parameters, output and results. Data from a clinical trial are presented as an example in Section 4. The paper is concluded by a discussion.
Section snippets
Methods
The aim of this paper is to present a computer package applied after the occurrence of each SAE, rather than having to rely on a posteriori ad hoc decision once the events have occurred. This sequential method needs to define events which incorporate unacceptable toxicity types such as toxic death, a maximal acceptable rate of toxicity (τ), and three a priori parameters: a one-sided type I error α, an expected total number of patients to include in the trial (Nmax) and the parameter γ defined
Program description
The package SAE has been developed using R language, a popular language and environment for statistical computing (http://www.r-project.org). SAE implements the method presented in the previous section and contains three main functions: SAE() to use after each observed event during the trial and DESIGN() or DESIGN_plus() which provide statistical rules set up before the beginning of accrual in the trial. A flow chart describing this package is presented in Fig. 2.
The package SAE has
Discussion and conclusion
This paper describes an R package aiming at monitoring the SAEs during a trial according to a pre-specified rule. The SAE package provides access to statistical methodology related to the different stages of sequential monitoring. First, we use dynamic calculations to determine the information fraction, the values of sequential boundaries and the value of confidence intervals associated with the sequential boundaries. Then, we find the value of for which the lower bound of the confidence
Acknowledgement
We would like to thank ROCHE for their financial support.
References (11)
- et al.
Incorporating toxicity considerations into the design of two-stage phase II trials
Biometrics
(1995) - et al.
A Bayesian stopping rule for a single arm study: with a case study of stem cell transplantation
Stat. Med.
(2006) - et al.
Optimal continuous sequential boundaries for monitoring toxicity in clinical trials: a restricted search algorithm
Stat. Med.
(2001) - et al.
Early stopping rules in clinical trials based on sequential monitoring of serious adverse events
Med. Decis. Making
(2009) Group sequential methods in the design and analysis of clinical trials
Biometrika
(1977)
Cited by (2)
Quantitative Methods for Blinded Safety Monitoring
2021, Quantitative Drug Safety and Benefit Risk Evaluation: Practical and Cross-Disciplinary Approaches