Understanding on Basic Statistics in Clinical Trials

by | May 24, 2019 |

Statistics is a fundamental aspect of clinical trials. Statistical concepts are used at every stage of a clinical trial, i.e., in protocol development and design, conduct and data monitoring, data analyses, data management and reporting of results. Statistical analysis enables the researchers to make accurate interpretations from the collected data. Some of the key concepts of statistics that help in the demonstration of accurate results include:

  • randomization
  • blinding
  • hypothesis testing
  • missing data handling
  • p-values
  • multiplicity adjustment

Statistics for Randomization and Blinding Techniques

Randomization is the process of assigning the patients to treatment arms at random. The most common techniques in statistics used to randomize patients are simple random sampling, permuted block randomization, and stratified random sampling. In randomized trials, mostly blinding methods are implemented in which one (single-blinded) or more (double- or triple-blinded) parties are unaware of the treatment given to avoid the bias. Concealed allocation is also used in some studies to avoid the bias at the time of allocating the patients to a particular treatment arm.

Hypothesis in Statistics

A hypothesis is an assumption that states something about an assessment to guide scientific investigation and assumes highly probable things. Hypotheses in statistics are of two types: Null hypothesis and Alternative hypothesis. The Null hypothesis assumes that an observation result from a chance factor and Alternative hypothesis assumes observation results from a real factor. A hypothesis can result in two kinds of errors- Type I error (false positives) and Type II error (false negatives). In Type I error, a result indicates the presence of a condition (such as disease), but actually, it is not present whereas in Type II error the test results indicate that the condition is not present, but actually it does.

p-Value and Statistics Solution for Missing Data Handling

A probability value (p-value) in statistics indicates the significance of the difference between groups. The between-group difference is considered significant when the p-value is less than or equal to 0.05. But the power of significant decreases when there are some missing data. Missing data is one of the greatest dangers to the integrity of a clinical trial as it creates challenges in the interpretation of results. Before handling the missing data, it is necessary to understand what type of data are missing and then find out the methods to impute missing data. Most common methods to impute missing data are mean/median/mode imputation and last observation carried forward (LOCF). In mean/median/mode imputation, missing values in a column are substituted with the mean/median/mode of the values available in that column. In the LOCF method, missing value in a row is substitute with the value in the preceding row.

Statistics Approach for Adjusting Multiplicity

Uncertainties in the conclusions of trials may also be caused by multiplicity, i.e., testing multiple hypotheses or testing a hypothesis at multiple time points or in multiple ways. To conduct multiple hypotheses, a chance of making false positive conclusions increases. There are a number of approaches in statistics to adjust multiplicity and one can choose the best-suited method per type of study design. In a clinical study with a single analysis, a p-value below 0.05 is considered significant, and the rate of false positive error is set at 5%. But in the case of multiple hypotheses, the chance of making false positive conclusion becomes more than 5%. These errors can be controlled by determining the significance level, e.g., if a researcher wants to conduct four hypothesis analyses, then the study-wise error rate needs to be controlled at 5% by performing an individual test at 0.0125 rather than 0.05.

Errors can take occur at any stage in a clinical trial, but the use of ideal methods of statistics can help to minimize these errors. A statistician must have an understanding of all the basic concepts of statistics apart from the formulas. The deep knowledge of statistical methods will help to guarantee the correct interpretation from clinical trials.