Statistics in Clinical Trials28-10-2021
Statistics helps to effectively manage and analyze the data gathered from Clinical Trials (CT). It is mainly aimed at:
- Ensuring efficient investigational designs and certifying consistency of results by controlling and reducing some types of bias.
- Analyzing and presenting the results obtained in a clear manner.
Three phases can be identified when conducting clinical trials:
- Preparation of the protocol.
- Development of the research.
- Analysis, documentation, and presentation of results.
Statistics is involved in each of these phases.
Preparation of the Protocol
Calculating the sample size is an essential step, which is carried out in the protocol design phase. The risk of false-negative results increases if sample sizes are determined solely based on the patients available. However, the probability of false-negative results in the clinical trial is reduced if the sample size is adjusted based on type I and type II risks assumed, the minimum clinically significant differences expected, and the response variability.
Data Cleaning and Internal Information Validation
Based on study case report forms, databases are created. Files are created in the computer, with information that must be free of errors to be properly analyzed later.
Analysis and Documentation of Results
Statistical analyses from clinical trials are aimed at showing that differences are not random. To that end, the “p” value is calculated. The p-value helps differentiate results that are derived from the random nature of the sampling from non-random or “statistically significant” results. The p-value is based on the null hypothesis or starting point, which means that it is assumed that there is no difference between treated patients and untreated patients. Its value ranges from 0 to 1, and a cut-off value must be established below which the null hypothesis may be rejected. In general, this value is usually 0.05 or 0.01.
It should be noted that neither the p-value nor the statistical significance measures the size of an effect or the importance of a result. Therefore, it is advisable to use a confidence interval indicating the limits of the effect, with the same units as the study variable. The confidence interval describes the variability between the measurement obtained in a trial and the actual measurement in the population (the true value). It corresponds to a range of values whose distribution must be normal and where there is a high probability of finding the true value of a given variable. This “high probability” has been consensually established at 95%. A 95% confidence interval indicates that within the given range lies the true value of a parameter with a 95% certainty. The level of confidence and the width of the interval vary together, which means that a wider interval will have a higher probability of accuracy (higher confidence level), while a smaller interval, offering a more accurate estimate, will have a higher probability of error.
Presentation of Results
The final data of a trial must be presented clear and concisely. There are official guidelines, such as those of the U.S. Food & Drug Administration (FDA) or the European Medicines Agency (EMA), which help organize the statistical report.