Patient-specific analysis of sequential haematological data by multiple linear regression and mixture distribution modelling

Automated storage and analysis of the results of serial haematologic studie s are now technically feasible with present-day laboratory instruments and devices for data storage and processing. In current practice, physicians me ntally compare a laboratory result with previous values and use their clini cal judgement to determine the significance of any change. To provide a sta tistical basis for this process, we describe a new approach for the detecti on of changes in patient-specific sequential measurements of standard haema tologic laboratory tests. These methods include hierarchical multiple regre ssion modelling, with a weighted minimum risk criteria for model selection, to choose models indicating changes in mean values over time. This study i s the first to analyse sequential patient-specific distributions of laborat ory measurements, utilizing mixture distribution modelling with systematic selection of starting values for the EM algorithm. To evaluate these statis tical methods under controlled conditions, we studied 11 healthy human volu nteers who were depleted of iron by serial phlebotomy to iron-deficiency an aemia, then treated with oral iron supplements to replete iron stores and c orrect the anaemia. Application of sequential patient-specific analyses of haemoglobin, haematocrit, and mean cell volume showed that significant depa rtures from past values could be identified, in many cases, even when value s were still within the population reference ranges. Additionally, for all subjects sequential alterations in red blood cell volume distributions duri ng development of iron-deficiency anaemia could be characterized and quanti fied. These methods promise to provide more sensitive techniques for improv ed diagnostic evaluation of developing anaemia and serial monitoring of res ponse to therapy. Copyright (C) 2000 John Wiley & Sons, Ltd.