A Statistical Approach to Drug Sampling: A Case Study

In many countries it is left to the discretion of the court to accept or reject conclusions based on sampling procedures as applied to the total drug exhibit. As an alternative to this subjective approach, a statistical basis is presented using binomial and hypergeometric distributions to determine a lower limit for the proportion of units in a population which contains a drug, at a given confidence level. A method for calculating the total weight of a drug present in a population within a given confidence interval is also presented. In the event of no failures (all units sampled contain a drug), a sample size of six or seven units is generally sufficient to state that a proportion of at least 0.70 of the population contains a drug at a confidence level of at least 90%. When failures do occur in the sample, point estimation is used as the basis for selecting the appropriate sample size.