software to determinate the sample size in clinical trials

Software to Determinate the Sample Size in Clinical Trials
Lic. Evian Fernández García
National Center of Clinical Trials
Calle 200 esq. 21 Rpto. Atabey
Ciudad de La Havana, Cuba.
[email protected]
Carmen Viada Gonzalez
National Center of Clinical Trials
Calle 200 esq. 21 Rpto. Atabey
Ciudad de La Havana, Cuba.
[email protected]
ABSTRACT
In clinical trials, the number of subjects to include is one of he most important aspects to
define because of its ethics and economic consequences. The lack of a powerful software for such
functions made us to program it in WINDOWS using Borland Delphi 2.0. With the software the
user can determine the sample size of twenty research designs. Each design give the possibility of
make the inverse calculations, information about the theoretical fundaments and bibliographic
references. In addition the software have mathematical tools to make easier the calculations.
1. Introduction
In a clinical research the number of patients to include (sample size) is one of most important
issues to define in the design in the research because of its ethics and economics implications.
When we recruit a insufficient sample size in the research its result can not be use for answer the
objective of the research because they are not scientifically supported. In the other hand, when we
recruit a excessive sample size we expend resources to obtain something that could have be
obtained with a less investment. In both cases this is not ethically justified because we are dealing
with human beings and the efficacy of the issue under research is not proved yet. In the first case we
submit a number of patients for obtain nothing and in the second case we do a unnecessary waste.
In our institutions and in the country we found a lack of a software with the most important
study designs. That is why our department created a project to implement such software.
2. Results
We reviewed all the clinical trials made in our institution and we defined the most frequents
study designs used. In each of them we search the objectives, theoretical fundament and the
mathematics formulas to be implemented. In additions we elaborated a specialized system of help.
The implementation was made in a WINDOWS interface using the programming language Borland
Delphi 2.0.
Each design give the possibility of make the inverse calculations, information about the
theoretical fundaments and bibliographic references. In addition the software have mathematical
tools to make easier the calculations. With the software we can determine the sample size of twenty
different designs with the following structure:
v Comparative design
A- Qualitative variables
I- 2×2 contingence tables
a- Independent samples
1- Comparison
2- Estimation
3- Equivalence
b- Dependent samples
c- Agreement inter raters
d- Log-lineal
II- k×m contingence tables
III- Stratified analysis
B- Quantitative variables
I- Two groups comparisons
a- Independent samples
1- Hypothesis tests
À Parametric
À Non parametric
2- Estimation
b- Dependent samples
II- ANOVA
a- One-Way
b- Two-Way
c- Repeated measures
v Non comparative designs
A- Qualitative variables
I- Parameter estimation
II- Hypothesis tests
a- Normal
b- Exact
B- Quantitative variables
I- Parameter estimation
II- Hypothesis tests
C- Correlation
v Another ones
A- Testing Procedure hypothesis
I- Fleming: 1 step
II- Gehan
B- Survival analysis
I- Log-rank
II- Exponential
C- Cross over design
3. Conclusions
We have freely distribute copies of the software to research medical institutions and have
proved to be a useful and efficient tool in the research sample size calculations. In addition the help
system implemented contribute to elevate the scientific knowledge and the mathematical
background of the users. The next step in the development of the software is its distribution in the
Cuban research institutions.
REFERENCES
Machin D, Campbell MJ. (1987). Statistical Tables for the design Clinical Trials. Oxford: Blackwell
Scientific Publications.
Rochon J. (1989). The Applications of the Method GSK to the Determination of Minimum Sample
Sizes. Biometrics 45, 193-205.
Fleming, T R. (1989). One-Sample Multiple Testing Procedure for Phase II Clinical Trials.
Biometrics 1982; 38: 143-151.
FRENCH RÉSUMÉ
Dans le dessin d’un essai clinique, le nombre de sujets à inclure c’est l’un des aspects les plus
importants à définir pour les conséquences étiques et économiques qui entraîne. C’est la genèse de
nous proposer dessiner et programmer une système automatisé pour calculer le nombre des sujets
nécessaire dans le dessin des essais cliniques. Le système est programmé avec le Borland Delphi
2.0 sur une ambiance interactive en WINDOWS 95. Avec le système on peut calculer nombre des
sujets des dessins les plus utilisés (vingt dessins). Chacun des dessins offre la possibilité de faire le
calcule á l’inverse, des informations sur le fondement théorique de et des références
bibliographiques. D’ailleurs on trouve des outils mathématiques qui aident le calcule.