Analysis of missing data in experimental medicine: an example of postresuscitation disorders in mononucleotide metabolism

Abstract

Background: In experimental medicine, some values may be missing from the data due to death of a part of experimental animals from the study pathology. These missing values are not accidental and lead to errors in results and conclusions of the studies, since conventional statistical methods are not suitable for them. It is necessary to develop new methods for analyzing missing data and testing their applicability in medical studies.
Aim: Testing a previously developed analytical method for samples with non-random, missing values on an example of a study of postresuscitation disorders in mononucleotide metabolism.
Methods: Modeling 6.5-min asphyxia and postresuscitation disease in rats; measurement of mononucleotides, nucleosides and nucleobases in the rat brain; and analysis of their arrays. Processing of results using a CensMed statistical software previously created by the authors.
Results: In the first 30 minutes after starting the resuscitation, mononucleotide catabolism increased in the rat brain, which resulted in accumulation of nucleosides and nucleobases. In later post-resuscitation periods, due to death of a part of animals from the study pathology, comparison of resuscitated rats with the control group became ambiguous. Using the method previously proposed by the authors it appeared possible to reduce this ambiguity and to establish upper and lower limits for possible influence of the missing values. The authors defined limitations of the proposed method.
Conclusion: The results of studies including non-random, missing values should be processed with special rather than regular statistical methods.