MONITORING SMALL-CELL LUNG-CANCER (SCLC) BY SERUM NEURON-SPECIFIC ENOLASE (S-NSE) ANALYSES USING DYNAMIC LINEAR-MODELS

In clinical monitoring based on quantitative data the identification a nd interpretation of changes in pattern of time series is important. A nalysis of such series is difficult due to influence from random biolo gical and analytical variation. Simple statistical procedures are insu fficient to describe the series and limited to static situations. The multiprocess dynamic linear time series model offers an on-line comput ation of the probability of changes of a patient's health status. In t he present study of the usefulness of the tumour marker serum neuron s pecific enolase (S-NSE) in predicting relapse of small cell lung cance r (SCLC) this model was combined with a statistical procedure, applica tion of the Kalman filter. This was used to be able to discriminate be tween changes of clinical significance (i.e. relapse of disease) and v ariations in the tumour marker baseline level due to random analytical and biological variations. The alarm criteria set for suspicion of re lapse were a posterior probability for a transition to the recurrence phase at present time bigger than 0.5 or a posterior probability of al ready being in the recurrence phase bigger than 0.5. Time series from 64 patients were divided at random into two groups each comprising 32 cases. The population parameters of the statistical model were estimat ed in one group and applied in the other and vice versa. The populatio n parameters did not differ significantly between the two groups (p < 0.05). In cases with clinical relapse the monitoring procedure identif ied 79 %. The lead time varied from one observation period prior to cl inically overt recurrence to three periods after. Six clinically unide ntified relapses with fatal outcome were predicted using the monitorin g procedure. Besides 7 false negative and 5 false positive cases were identified. By modulating the criteria set for alarm the method may be suitable to predict various clinical events such as response to thera py as well in the introductory phase as after therapy change into seco ndline treatment, provided they are reflected in the marker level.