OBJECTIVEThe multicenter, randomized pediatric cerebrospinal fluid shunt valve design trial found no difference in the rate of shunt failure between a standard valve, a siphon-reducing valve (Delta; Medtronic PS Medical, Goleta, CA), and a flow-limiting valve (Orbis Sigma; Cordis, Miami, FL); however, the valves were expected to have different effects on ultimate ventricular size. Also, the catheter position or local environment of the ventricular catheter tip might have affected shunt failure. Therefore, we performed a post hoc analysis to understand what factors, other than valve design, affected shunt failure and to identify strategies that might be developed to reduce shunt failure.METHODSVentricular size was measured at as many as six different intervals, using a modified Evans’ ratio (with incorporation of the frontal and occipital dimensions), in 344 patients. Ventricular catheter location was defined as being in the frontal horn, occipital horn, body of the lateral ventricle, third ventricle, embedded in brain, or unknown. The ventricular catheter tip was described as surrounded by cerebrospinal fluid, touching brain, or surrounded by brain parenchyma within the ventricle (slit ventricle). Repeated measures analysis of variance for unbalanced data was used to analyze ventricular size. A Cox model (with incorporation of time-dependent covariates) was used to evaluate the contribution of age, etiology, shunt design, ventricular size, ventricular catheter location, and environment among the cases.RESULTSVentricular volume decreased in an exponential fashion, forming a plateau at 14 months, and was similar for the three valves (P= 0.4). Frontal and occipital ventricular catheter tip locations were associated with a reduced risk of shunt failure (hazard ratios, 0.60 [P= 0.02] and 0.45 [P= 0.001], respectively). Ventricular catheter tips surrounded by cerebrospinal fluid or touching the brain were associated with a reduced risk of failure (hazard ratios, 0.21 and 0.33, respectively;P= 0.0001). Patients with myelomeningocele or large ventricles had increased risk of malfunction (hazard ratios, 1.78 [P= 0.006] and 2.33 [P= 0.03], respectively).CONCLUSIONDecline of ventricular size over time is not affected by these different shunt valve designs. This suggests that the mechanical models of hydrocephalus on which the designs were based are inadequate. Ventricular catheter tip location and ventricular catheter environment are important. Techniques to accurately place ventricular catheters and new valve designs that effectively control ventricular size might reduce shunt malfunction.