gms | German Medical Science

50. Jahrestagung der Deutschen Gesellschaft für Medizinische Informatik, Biometrie und Epidemiologie (gmds)
12. Jahrestagung der Deutschen Arbeitsgemeinschaft für Epidemiologie (dae)

Deutsche Gesellschaft für Medizinische Informatik, Biometrie und Epidemiologie
Deutsche Arbeitsgemeinschaft für Epidemiologie

12. bis 15.09.2005, Freiburg im Breisgau

RANDI- Online Randomisation for small-scale Clinical Trials

Meeting Abstract

Suche in Medline nach

  • Dirk Hoffmann - Deutsches Krebsforschungszentrum, Heidelberg
  • Lothar Pilz - Deutsches Krebsforschungszentrum, Heidelberg

Deutsche Gesellschaft für Medizinische Informatik, Biometrie und Epidemiologie. Deutsche Arbeitsgemeinschaft für Epidemiologie. 50. Jahrestagung der Deutschen Gesellschaft für Medizinische Informatik, Biometrie und Epidemiologie (gmds), 12. Jahrestagung der Deutschen Arbeitsgemeinschaft für Epidemiologie. Freiburg im Breisgau, 12.-15.09.2005. Düsseldorf, Köln: German Medical Science; 2005. Doc05gmds321

Die elektronische Version dieses Artikels ist vollständig und ist verfügbar unter: http://www.egms.de/de/meetings/gmds2005/05gmds361.shtml

Veröffentlicht: 8. September 2005

© 2005 Hoffmann et al.
Dieser Artikel ist ein Open Access-Artikel und steht unter den Creative Commons Lizenzbedingungen (http://creativecommons.org/licenses/by-nc-nd/3.0/deed.de). Er darf vervielfältigt, verbreitet und öffentlich zugänglich gemacht werden, vorausgesetzt dass Autor und Quelle genannt werden.


Gliederung

Text

Introduction

A prototype of an online randomisation system for small-scale clinical trials called RANDI is presented in this abstract. RANDI is a tool which can shorten the process of randomisation. Compared to manual randomisation, it is more accurate and available 24 hours a day.

Clinical trials are planned experiments to prove the effectiveness of a therapy for patients having a defined disease. The intention is to define the effects on potency, dosage, and safety of clinical treatments by using statistical methods, and to find a better treatment in comparison with the standard treatment available for other patients.

The ideal clinical trial is a randomised clinical trial, where patients are assigned to a group (treatment, stratum, etc.) by a random mechanism [1]. Considering a prior defined distribution, each patient must have the same probability to get one of the available therapies (restricted random). It is fundamental that neither the patient nor the physician can bias this choice. The aim of this procedure is to guarantee equal conditions for all treatments.

In order to avoid a conflict of interest and the possible impact on ideal case planning, implementation and control of a clinical trial is best done by an independent biometric centre. For these purposes, many organisers of clinical trials cooperate with an independent institution like the Central Department of Biostatistics at the German Cancer Research Center [2], [3].

Manual Randomisation

Currently at the beginning of a randomisation for a small-scale trial, the required randomisation lists must be manually created, e.g., by common statistical software which has a built-in stable generator for pseudo random numbers. The medical and statistical context is analysed, and the accordant parameters such as stratification and block formation are chosen.

Having created the randomisation lists, patients can be admitted to the trial. Often this job is done by a telephone- or facsimile-randomisation during the daily tasks. If the data of a patient fulfils the required criteria, the responsible physician submits those to the biometric centre by telephone or facsimile. A therapy is assigned to the patient according to the randomisation lists, and this assignment is normally documented manually by simple lists. If more study centres take part in a trial, the randomisation is more complex. Each centre has its own randomisation list controlling the assignments of the patients to the different therapy branches. Even with a small number of participating centres, it still requires a folder full of different paper lists. If these centres conduct the randomisation on their own, misapplication can be avoided by the use of enveloped written assignments. With this method, alternative therapies are masked and each envelope is signed with a running number that will be executed sequentially.

Conclusion of the Manual Randomisation

The described manual workflow proves to be ineffective because the operative effort is too high. Creating an attentive handling of the randomisation lists can be presently supported easily by electronic systems. There already exist several systems for the electronic management of clinical trials. These systems show, in most cases, functional diversity, and are too large and expensive [4], [5]. Especially taking into account that often only a small number of patients are involved, the use of an extensive and non-adjusted system is neither required nor desired. Due to these facts, we developed an online randomisation system that is specially adapted for small-scale clinical trials.

Methods and Processes

System Environment

RANDI has been implemented as an internet application in PHP (Hypertext Preprocessor). PHP is a widely-used scripting language that is especially suited for web development and can be embedded into HTML (Hyper Text Markup Language). To store the information, RANDI can communicate with common database systems. On the client’s side, the user can access the system by any common web browser or any mobile phone that supports WAP (Wireless Markup Language) [6], [7].

Randomisation in RANDI

In RANDI we integrate three user groups: Physician, Statistician and Administrator (study coordination). These groups are administrated in a user management. In order to gain access to the system, a new Physician has to fill out a registration form. RANDI submits this request to the Administrator, and as soon as the Administrator has checked and activated the account, the new Physician will have access to the system. Furthermore, the Administrator can create accounts for Statisticians and Administrators.

To set up the randomisation, the Administrator first describes and defines the alternative therapies of the study within RANDI. In addition, he chooses the randomisation type. He can choose between “global list”, “blocked list” and “random generator”. If he selects “blocked list”, he must define the exact block length. The length must be 6, 8 or 9. In case of stratification, RANDI can stratify for institute, gender and one other optional binary attribute. To guarantee the applicability, a block factor had been implemented. In case of multilevel stratification, there are probably not enough suitable randomisation table-entries available. In order to avoid this problem, the Administrator should apply the block factor, generating n permutations of the randomisation block. Afterwards he only has to define the number of entries. This means that a maximum number of patients will be randomised. Finally, the Administrator starts the generation of the randomisation table. This generation occurs by creating permutations which depend on the randomisation settings including therapy alternatives. In the next step, the randomisation table is stored in the data base and can be checked by the Administrator.

After a successful generation of the randomisation table, the Physician can start with the randomisation of his patients. Therefore the Physician enters in RANDI all relevant patient data (e.g., initials, gender, date of birth, data of accordance, body surface area, WHO performance status). After the data transmission, patient randomisation takes place in the following three steps: validity check, randomisation control and finally the randomisation. The randomisation control assures that no patient has already been randomised with the same initials, birth date and gender from the same institute. The randomisation is conducted as follows: search all unused entries that comply with the randomisation settings, pull at random one entry and assign the actual patient to this entry. By now a therapy is assigned to the current patient. A confirmation via e-mail is sent to the Physician and also for documentation purposes to the Administrator.

For statistical purposes, RANDI can automatically create Excel-Files of the actual randomisation status. These files can be used by the Statistician for further analysis.

Results

The developed system was tested by the staff of the Biometric Unit of the German Cancer Research Center and is now ready to be implemented in a phase II medical study as a pilot project.

In an internal study, RANDI has effected a significant time saving during the whole randomisation process. Sources of errors could be minimized by validity checks in a predefined process. In addition, this high quality was assisted by a statistical evaluation of the randomisation process in runtime.

Further on RANDI will be upgraded to support more and more different tasks in a clinical trial. Currently we are working on a generator of Case Report Forms (CRF) which will enable the simple creation of CRFs in a clinical trial.


References

1.
Rosenberger WF, Lachin JM. Randomization in Clinical Trials: Theory and Practice. New York: Wiley; 2002
2.
Edler L, Pilz LR. Klinische Studien. In: Therapieoptionen beim nicht-kleinzelligen Bronchialkarzinom, Chap. 14. Manegold C, Hrsg. Bremen: UNI-MED, 2002: 148-163.
3.
Schumacher M, Schulgen G. Methodik klinischer Studien - Methodische Grundlagen der Planung, Durchführung und Auswertung. Berlin: Springer; 2002
4.
Chow S-C, Liu J-P. Design and Analysis of Clinical Trials. Concepts and Methodologies. 2nd Edition. New York: Wiley, 2004
5.
Girling DJ, Parmar MKB et al. Clinical Trials in Cancer: Principles and Practice. Oxford: University Press; 2003
6.
Lerdorf R, Tatroe K. Programming PHP. Cambridge, MA: O'Reilly, 2002
7.
Williams HE, Lane D. Web Database Applications with PHP and MySQL, 2nd Edition. Cambridge, MA: O'Reilly, 2002