Can intraoperative assessment of endoscopic third ventriculostomy predict success?

In the article that follows, Greenfield et al. have identified and investigated an intuitive observation that many surgeons doing endoscopic third ventriculostomy (ETV) have made. There seem to be patients in whom one can predict a successful outcome; the ventricles are pristine, the floor of the third ventricle is thin, it balloons up when perforated, and one has a clear view of the basilar artery in a widely patent subarachnoid space. By contrast there are other patients in whom there is scarring, a thickened floor of the third ventricle that is relatively immobile after perforation, scarred membranes below the floor, and so on, where failure seems to follow. These authors have attempted to quantify these observations, in a retrospective fashion, to develop a prediction model of these intraoperative findings, perhaps so that surgeons could immediately resort to an alternative treatment (such as ventriculoperitoneal shunt placement) if it looked like the chances of success were small. The authors’ findings match their intuition: there do appear to be clinical and surgical features that predict success. Although they plan to follow this study with a prospective study, how convinced can we be with their current model, and what might they include in their prospective study? There is a specific method for this type of model generation and testing to which the authors have followed to some extent. One first needs a group of patients with the clinical problem at hand who undergo the treatment of interest and experience an outcome. One is immediately aware that this report is from a “quaternary” referral center, particularly devoted to management of intracranial neoplasms, and this patient population might be different from other centers. Their determination of success or failure was the requirement for a subsequent cerebrospinal fluid (CSF) diversion. This is a convenient, often used outcome, but it is potentially problematic in this type of analysis because it depends on the decision making of the surgeon, usually the one performing the first procedure. If the treating surgeon had preconceived notions of what intraoperative factors may lead to failure, it could affect his or her decision-making process on who requires an additional procedure (and thus in whom it would fail), fulfilling the prophesy and acting as a form of bias. There are methods for defining CSF diversion failure at least for shunts, which can be blindly adjudicated, reducing this bias, and probably something the authors should do for ETV for their next study. Second, one needs factors that might have been previously reported or suspected to lead to success or failure. This is perhaps the most vital process in developing clinical prediction rules. The authors effectively used previous literature and clinical expert opinion to generate the potential predictors for study; however, the opinions were limited to a single institution. There is an emerging field in clinical epidemiology of quantitative methods, such as Delphic panels, which would significantly expand the scope of opinions and evaluate them in a more scientific manner. These factors should also be stringently defined, particularly if they are to be used across centers. The authors analyze 2 preoperative factors: 1) patient age < 2 years and 2) documented intraventricular hemorrhage (IVH), a history of meningitis, or a shunt infection. This second factor is really a group of factors that might all be envisioned to create some arachnoid or ventricular scarring. These factors are, however, not strictly speaking separate etiologies; IVH is different from meningitis, not all shunt infections have meningitis, and so on. Similarly for intraoperative findings under “abnormal anatomy” they group intraventricular scarring and intraventricular synechiae (probably the same process), but also leptomeningeal tumor dissemination, which is intuitively quite different. Normally the effect of each individual factor is analyzed separately, and they are only grouped for statistically valid reasons. The authors may have grouped these factors to stay within the confines of their statistical power. For this analysis they would require ~ 10 failures per factor to be analyzed. With 35 failures, the number of factors they J Neurosurg Pediatrics 2 295 297

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.