An easily customized, random allocation system using the minimization method for multi-Institutional clinical trials

Abstract

In a randomized clinical trial, random allocation of patients to treatment groups should be done to balance in the distribution of prognostic factors. Random allocation in a multi-institutional randomized clinical trial is conducted by a coordinating center, independent of the medical institution the attending doctor uses for his/her practice. This study provides a sophisticated system for doing an exact random allocation of patients to treatment groups. The minimization method proposed by Pocock was applied to this system to balance the distribution of prognostic factors between two treatment groups, even when the number of registered patients is relatively small (S.J. Pocock, Allocation of patients to treatment in clinical trial, Biometrics 35 (1979) 183–197). Furthermore, Zelen’s method is used to balance the number of patients allocated to the two groups within each institution (M. Zelen, The randomization and stratification of patients to clinical trials, J. Chron. Dis. 27 (1974) 365–375.). This system was created by the ‘perl’ language for writing common gateway interface (CGI) script, and can therefore, be easily extended to include data entry function by attending doctors as well as the random allocation function. This system is being used effectively in thirteen multi-institutional randomized clinical trials for stomach, colon–rectum and breast cancers in Japan.

Introduction

Random allocations are performed to balance the distribution of known and unknown prognostic factors between two treatment groups. An unbalanced distribution yields power loss and increases the size of statistical testing to compare the effects of two treatments, which may cause erroneous decisions [1], [2]. Stratified blocked randomization has also been employed as a procedure for randomization of patients to treatments, however, it has the disadvantage that as the number of strata increases, random allocation becomes rather irrelevant [3]. Pocock and Simon’s minimization method and Zelen’s method for institution balancing enable balancing of treatment numbers for each level of several prognostic factors over the entire trial and at the same time, also balance the allocation of treatments within an institution [4], [5].

In many multi-institutional randomized clinical trials, central telephone registration is probably the most frequently employed procedure for patient registration and treatment allocation. The attending doctor of each institution calls or faxes a centralized coordinating center to register the patient meeting and necessary data on the trial to get information on the assigned treatment for the patient from a registration personnel. The randomization procedures with the minimization method and Zelen’s method for institution balancing require a certain amount of calculation, as each patient is randomized. These calculations may be time consuming for the coordinating centers in charge, especially in the case of central telephone randomization, the amount of time for the call should be kept to a minimum. Thus, a collaborative study group for randomized clinical trials needs to set up a computer system enabling easy and reliable random allocation at the coordinating center [6].

We developed a random allocation system by which registration personnel can perform random allocation precisely and quickly. The system is available, on-line, to world wide web (WWW) users. We created a graphical user interface by using common gateway interface (CGI), and developed the program over a shorter period by using perl language rather than C language, pascal or Macro language of Microsoft excel. The collaborative study group applied this program to a new randomized clinical trial with slight modification, and saved the cost of developing new programs. At present, this system is being used conveniently in 13 randomized clinical trials for gastric, colon–rectum and breast cancers in Japan.