LAUSR

Sir, We read with great interest the article “Stability of complete blood count in different storage conditions using the ABX PENTRA 60 analyzer” by Oliveira, Simionetto, and Cruz et al published in your journal.1 The authors take an interesting approach to estimate stability of hematological parameters on storage. We feel that the analysis may have been more effective with a different approach to the analysis, which leads us to share our experience in storage stability studies. Sample stability on storage is important because there is always some time gap between blood sample collection and analysis. It is important to know confirm and document that the parameters to be measured in the samples do not undergo major shift in their measured values at least during the typical time from collection to analysis, the maximum permissible time from storage to analysis, and the time of sample retention in the laboratory. As the storage temperature and transport temperature are often different, a sample stability study should be carried out at the ideal transport temperature, the ideal storage temperature, and outlying temperatures often encountered during transport with potential effect on stability. For the statistical inference, the primary emphasis should be on statistical estimation of bias (with information about the uncertainty associated with the estimation of bias, ie, the confidence intervals) rather than just hypothesis testing. Therefore, the percentage shift with its confidence interval at each and every time point should have been calculated and compared to the maximum acceptable bias, and the conclusion confirmed with proper hypothesis testing. Just a test for statistical significance using ANOVA or its nonparametric equivalent is not enough as it gives no information about the estimates of interest, that is, presence or absence of a clinically significant bias at the time points measured. A sample with unacceptable and statistically significant bias should be declared unstable at that time point; laboratory protocols should be designed such that for the important parameters, the maximum time from collection to analysis does not exceed the time at which significant bias appears. At this point, importance has to be given to how much bias is acceptable. The desired maximum permissible systematic error in a laboratory as calculated from biological variation experiments comes out to be 0.25∗ √