Letter: Antibiotic-Impregnated Ventriculoperitoneal Shunts Decrease Bacterial Shunt Infection: A Systematic Review and Meta-Analysis

To the Editor: Kovács et al1 recently published their systematic review and meta-analysis on a very important topic for both adult and pediatric neurosurgeons that antibiotic-impregnated shunts can improve certain shunt outcomes. They are to be congratulated on continuing the important dialog on how to continue to improve shunt performance to minimize both dreaded failure and infection. It is challenging to incorporate such a multifaceted topic into a singular PICOS (population, intervention, comparator, outcome, study type) question format, and we believe that further clarity is required to better define their important message. First, this study included all etiologies in which a shunt could be used. This is clinically difficult then to generalize across all possible indications for antibiotic-impregnated shunts given the vast array of possible causes leading to a shunting procedure, further amplified by the inclusion of studies of all ages as adult and pediatric shunt indications can be very different. For example, postinfectious hydrocephalus in an infant requiring a shunt is clinically very different vs that of normal pressure hydrocephalus in an elderly adult also requiring a shunt. It is known that age and etiology within themselves can both affect shunt infection risk, irrespective of antibiotic-impregnation status.2-4 The greater the heterogeneity in the selection criteria of this study, the greater likelihood the overall summary value has trended toward the null hypothesis meaning that an absence of statistical significance may not be entirely applicable to all clinical situations as much as the authors imply. Second, the authors to a degree have recognized some of this clinical heterogeneity by including subanalyses comparing different groups, such as adult vs pediatric studies, indeed finding that there were differences in statistical significance when considering age group alone. However, we respectfully disagree with the authors' statement that there was relatively low heterogeneity in the case of the primary outcome. Indeed, the P-heterogeneity values for both the overall primary outcome and the one based on age were statistically significant (both P < .050). As such, we would caution in assuming homogeneous data distribution and would assume this would lower the certainty in the summarized outcomes reported even more. This concern is further exacerbated by a report that 22/27 studies included were deemed to have high risk of bias. Next, it would be an interesting statistical exercise for the authors to perform a leave-one-out analysis for the study by Parker et al5 or demonstrate where this specific study lies on within the bias funnel plots. This is because the overall contributing control cohort size of this single study was more than 50% of the entire metadata for the overall primary outcomes, 25% of the pediatric subgroup primary outcome, and 98% of the adult subgroup primary outcome. This possible "large-study" bias should be addressed to further enhance our confidence in the reported results. Finally, we would like to clarify the implication of the authors stating that there was a tendency in favor of antibiotic-impregnated shunt use regarding shunt failure. Statistically, summary value for this outcome in the study was odds ratio 0.73, 95% CI 0.51-1.06, with P-overall 0.086 and P-heterogeneity <0.001. Although this may be semantics, we do not believe that any relation should be implied if P-overall is not significant, particularly when the included outcomes are statistically heterogeneous. Indeed, of the 9 studies included in this outcome, only 1 study actually reported statistical significance, and the remaining 8 studies did not. The observation is that 8/9 (89%) studies indicated no statistical relationship between antibiotic-impregnated shunt use and shunt failure. Without meta-regression analysis, it is even more difficult to infer any type of trend or tendency. Assuming P-overall values close to 0.050 indicates a tendency is a statistical fallacy without any post hoc justification, and one that we hope readers do not equate to a clinically justifiable finding. In summary Kovács et al1 have embarked and completed an ambitious task of summarizing the performance of antibiotic-impregnated shunts across all etiologies and age groups. Particular clinical and statistical clarifications will greatly enhance the translation of these findings into appropriately actionable items to improve shunt outcomes for all our patients.