Identification of risk factors for shunt infection and predictors of infectious pathogens may improve current methods to prevent and treat shunt infections. We reviewed data on 820 consecutive ventriculoperitoneal (VP) shunt placement procedures in 442 pediatric patients at our institution during 1992-1998. Ninety-two shunts (11%) developed infection a median of 19 days (interquartile range, 11-35 days) after insertion. Premature birth (relative risk [RR], 4.81; 95% confidence interval [CI], 2.19-10.87), previous shunt infection (RR, 3.83; 95% CI, 2.40-6.13), and intraoperative use of the neuroendoscope (RR, 1.58; 95% CI, 1.01-2.50) were independent risk factors for shunt infection. The bacterial organisms early after shunt surgery (<14 days) were the same as those late after shunt surgery (>14 days). As determined by an analysis of the 92 infected shunts, hospital stay of >3 days at the time of shunt insertion (odds ratio [OR], 5.27; 95% CI, 1.15-25.3) and prior Staphylococcus aureus shunt infection (OR, 5.91; 95% CI, 1.35-25.9) independently increased the odds that S. aureus was the causal pathogen.

Background: Innovations in shunt technology and neuroendoscopy have been increasingly applied to shunt management. However, the relative life span of shunts and the etiology of shunt failure have not been characterized recently. Methods: We reviewed the records of all shunting procedures at our institution between January 1992 and December 1998. Independent predictors of shunt failure were analyzed via multivariate Cox regression analysis in 836 shunting procedures. Independent predictors of the etiology of failure (infection, proximal obstruction, distal malfunction) were analyzed via multivariate logistic regression analysis in the 383 shunts which failed. Results: A total of 353 pediatric patients underwent 308 shunt placements and 528 revisions. The risk (hazard ratio; HR) of shunt failure decreased as a function of time in both primary placements and revised shunts. In failed shunts, the odds of infection decreased 4-fold per year of shunt function, while the odds of distal malfunction increased 1.45-fold per year. Increasing number of shunt revisions (HR 1.31, p < 0.05), decreasing patient age in years (HR 1.04, p < 0.001), gestational age <40 weeks (HR 2.15, p < 0.001) but not the etiology of hydrocephalus were associated with an increased risk of shunt failure. Revisions versus primary placements, Dandy-Walker cysts and gestational age <40 weeks were independently associated with proximal, distal and infectious causes of failure, respectively. Conclusions: The long-term shunt revision rates observed here are similar to those reported over the past 2 decades. Shunt life span remains poorer in shunt revisions and in younger patients. Patient characteristics may suggest a specific risk and mechanism of failure, aiding in the long-term management of shunted hydrocephalus.

Vaccination with cytokine-producing tumor cells generates potent immune responses against tumors outside the central nervous system (CNS). The CNS, however, is a barrier to allograft and xenograft rejection, and established tumors within the CNS have failed to respond to other forms of systemic immunotherapy. To determine what barriers the "immunologically privileged" CNS would pose to cytokine-assisted tumor vaccines and what cytokines would be most efficacious against tumors within the CNS, we irradiated B16 murine melanoma cells producing murine interleukin 2 (IL-2), IL-3, IL-4, IL-6, y-interferon, or granulocytemacrophage colony stimulating factor (GM-CSF) and used these cells as subcutaneous vaccines against tumors within the brain. Under conditions where untransfected B16 cells had no effect, cells producing IL-3, IL-6, or GM-CSF increased the survival of mice challenged with viable B16 cells in the brain.Vaccination with B16 cells producing IL-4 or r-interferon had no effect, and vaccination with B16 cells producing IL-2 decreased survival time. GM-CSF-producing vaccines were also able to increase survival in mice with pre-established tumors. The response elicited by GM-CSF-producing vaccines was found to be specific to tumor type and to be abrogated by depletion of CD8+ cells. Unlike the immunity generated against subcutaneous tumors by GM-CSF, however, the effector responses generated against tumors in the CNS were not dependent on CD4+ cells. These data suggest that cytokineproducing tumor cells are very potent stimulators ofimmunity against tumors within the CNS, but effector responses in the CNS may be different from those obtained against subcutaneous tumors.