Improvements in the Technique of Spinal Subarachnoid Recirculatory Perfusion

Abstract

With 14C-labeled dextran as the tracer, studies of the original configuration of spinal recirculatory perfusion and the original model for data analysis demonstrated that this approach does not yield acceptable accuracy in determining cerebrospinal fluid (CSF) formation (Fcsf) and absorption (Acsf) rates. A significant component of this error was due to the fact that the method of data analysis used originally is not based on a realistic mathematical model of the system. A more realistic mathematical model resulted in two simultaneous differential equations that did not have simple analytical solutions and, therefore, could not be used easily for data analysis. By computer simulation, a comparison of the more realistic model with the original model demonstrated that, under ideal conditions, there was a 10% error inherent in the original data analysis method. In the experimental setting, the magnitude of this inherent error is probably 20%. There were three other major problems with the original system: (a) one could not tell when enough data had been collected to ensure convergence of the data analysis algorithm; (b) calibration of the syringe pump in the external circuit was not accurate for short infusion periods; and (c) the presence of the syringe in the external circuit unnecessarily lengthened the period of nonhomogeneous mixing. A new system configuration and new data analysis methods have been developed. In the new system, the syringe is removed from the external circuit and intracranial pressure is controlled by infusion from a separate reservoir where the pressure head is maintained at any desired level by feedback control. Spectrophotometry is used to measure tracer concentration in the external circuit. Data collection and analysis are fully automated under computer control so that, during an experimental run, the investigators are updated at 1- to 2-second intervals as to the convergence of the data analysis routine. Data analysis methods for the new system are superior to previous methods because the models are realistic and no extrapolation is required. In addition, all of the data during the initial period of nonhomogeneous mixing are used to calculate Fcsf and Acsf. With the new system and a simple phantom of the CSF system, the mean error in finding Acsf was 1.7 +/- 1.2% for 27 determinations covering a wide range of absorption rates. Fcsf could be determined to within 0.001 ml/minute. In up to six sequential pressure plateaus, the magnitude of error did not increase with each subsequent run.(ABSTRACT TRUNCATED AT 400 WORDS)