Sampling errors and the precision associated with counting very low numbers of white cells in blood components

Attention to the accurate and precise measurement of the white cell (WBC) content of transfused products has risen in response to awareness of the potential benefits of WBC-depleted components and the development of technical capabilities to produce these components. The techniques thus far reported have focused on the reliability of detecting a WBC, provided it is present in the test system. The likelihood of selecting a WBC from the product of interest for counting in the analytical system--that is, the sampling error--must also be considered. The occurrence of a WBC in a WBC-depleted component is a rare event and may be modeled with the binomial or the Poisson distribution. Several assay techniques were analyzed by using these distribution models to determine the confidence intervals of the WBC content. The 95-percent confidence intervals spanned more than 2 logs10 for some methods at 3 x 10(5) WBCs per product. It is concluded that the reporting of WBC content for research provides not only the estimate of the mean but also a confidence interval for this estimate. Quality control procedures should be designed to verify that the WBC content is less than the targeted amount and should provide an associated statement of confidence.