Machine learning predicts risk of cerebrospinal fluid shunt failure in children: a study from the hydrocephalus clinical research network

PURPOSE

While conventional statistical approaches have been used to identify risk factors for cerebrospinal fluid (CSF) shunt failure, these methods may not fully capture the complex contribution of clinical, radiologic, surgical, and shunt-specific variables influencing this outcome. Using prospectively collected data from the Hydrocephalus Clinical Research Network (HCRN) patient registry, we applied machine learning (ML) approaches to create a predictive model of CSF shunt failure.

METHODS

Pediatric patients (age < 19 years) undergoing first-time CSF shunt placement at six HCRN centers were included. CSF shunt failure was defined as a composite outcome including requirement for shunt revision, endoscopic third ventriculostomy, or shunt infection within 5 years of initial surgery. Performance of conventional statistical and 4 ML models were compared.

RESULTS

Our cohort consisted of 1036 children undergoing CSF shunt placement, of whom 344 (33.2%) experienced shunt failure. Thirty-eight clinical, radiologic, surgical, and shunt-design variables were included in the ML analyses. Of all ML algorithms tested, the artificial neural network (ANN) had the strongest performance with an area under the receiver operator curve (AUC) of 0.71. The ANN had a specificity of 90% and a sensitivity of 68%, meaning that the ANN can effectively rule-in patients most likely to experience CSF shunt failure (i.e., high specificity) and moderately effective as a tool to rule-out patients at high risk of CSF shunt failure (i.e., moderately sensitive). The ANN was independently validated in 155 patients (prospectively collected, retrospectively analyzed).

CONCLUSION

These data suggest that the ANN, or future iterations thereof, can provide an evidence-based tool to assist in prognostication and patient-counseling immediately after CSF shunt placement.

Surgical resource utilization after initial treatment of infant hydrocephalus: comparing ETV, early experience of ETV with choroid plexus cauterization, and shunt insertion in the Hydrocephalus Clinical Research Network

OBJECTIVE

Few studies have addressed surgical resource utilization-surgical revisions and associated hospital admission days-following shunt insertion or endoscopic third ventriculostomy (ETV) with or without choroid plexus cauterization (CPC) for CSF diversion in hydrocephalus. Study members of the Hydrocephalus Clinical Research Network (HCRN) investigated differences in surgical resource utilization between CSF diversion strategies in hydrocephalus in infants.

METHODS

Patients up to corrected age 24 months undergoing initial definitive treatment of hydrocephalus were reviewed from the prospectively maintained HCRN Core Data Project (Hydrocephalus Registry). Postoperative courses (at 1, 3, and 5 years) were studied for hydrocephalus-related surgeries (primary outcome) and hospital admission days related to surgical revision (secondary outcome). Data were summarized using descriptive statistics and compared using negative binomial regression, controlling for age, hydrocephalus etiology, and HCRN center. The study population was organized into 3 groups (ETV alone, ETV with CPC, and CSF shunt insertion) during the 1st postoperative year and 2 groups (ETV alone and CSF shunt insertion) during subsequent years due to limited long-term follow-up data.

RESULTS

Among 1090 patients, the majority underwent CSF shunt insertion (CSF shunt, 83.5%; ETV with CPC, 10.0%; and ETV alone, 6.5%). Patients undergoing ETV with CPC had a higher mean number of revision surgeries (1.2 ± 1.6) than those undergoing ETV alone (0.6 ± 0.8) or CSF shunt insertion (0.7 ± 1.3) over the 1st year after surgery (p = 0.005). At long-term follow-up, patients undergoing ETV alone experienced a nonsignificant lower mean number of revision surgeries (0.7 ± 0.9 at 3 years and 0.8 ± 1.3 at 5 years) than those undergoing CSF shunt insertion (1.1 ± 1.9 at 3 years and 1.4 ± 2.6 at 5 years) and exhibited a lower mean number of hospital admission days related to revision surgery (3.8 ± 10.3 vs 9.9 ± 27.0, p = 0.042).

CONCLUSIONS

Among initial treatment strategies for hydrocephalus, ETV with CPC yielded a higher surgical revision rate within 1 year after surgery. Patients undergoing ETV alone exhibited a nonsignificant lower mean number of surgical revisions than CSF shunt insertion at 3 and 5 years postoperatively. Additionally, the ETV-alone cohort demonstrated significantly fewer hospital admission days related to surgical management of hydrocephalus within 3 years after surgery. These findings suggest a time-dependent benefit of ETV over CSF shunt insertion regarding surgical resource utilization.

Predictors of success for combined endoscopic third ventriculostomy and choroid plexus cauterization in a North American setting: a Hydrocephalus Clinical Research Network study

OBJECTIVE

Endoscopic third ventriculostomy combined with choroid plexus cauterization (ETV+CPC) has been adopted by many pediatric neurosurgeons as an alternative to placing shunts in infants with hydrocephalus. However, reported success rates have been highly variable, which may be secondary to patient selection, operative technique, and/or surgeon training. The objective of this prospective multicenter cohort study was to identify independent patient selection, operative technique, or surgical training predictors of ETV+CPC success in infants.

METHODS

This was a prospective cohort study nested within the Hydrocephalus Clinical Research Network's (HCRN) Core Data Project (registry). All infants under the age of 2 years who underwent a first ETV+CPC between June 2006 and March 2015 from 8 HCRN centers were included. Each patient had a minimum of 6 months of follow-up unless censored by an ETV+CPC failure. Patient and operative risk factors of failure were examined, as well as formal ETV+CPC training, which was defined as traveling to and working with the experienced surgeons at CURE Children's Hospital of Uganda. ETV+CPC failure was defined as the need for repeat ETV, shunting, or death.

RESULTS

The study contained 191 patients with a primary ETV+CPC conducted by 17 pediatric neurosurgeons within the HCRN. Infants under 6 months corrected age at the time of ETV+CPC represented 79% of the cohort. Myelomeningocele (26%), intraventricular hemorrhage associated with prematurity (24%), and aqueductal stenosis (17%) were the most common etiologies. A total of 115 (60%) of the ETV+CPCs were conducted by surgeons after formal training. Overall, ETV+CPC was successful in 48%, 46%, and 45% of infants at 6 months, 1 year, and 18 months, respectively. Young age (< 1 month) (adjusted hazard ratio [aHR] 1.9, 95% CI 1.0-3.6) and an etiology of post-intraventricular hemorrhage secondary to prematurity (aHR 2.0, 95% CI 1.1-3.6) were the only two independent predictors of ETV+CPC failure. Specific subgroups of ages within etiology categories were identified as having higher ETV+CPC success rates. Although training led to more frequent use of the flexible scope (p < 0.001) and higher rates of complete (> 90%) CPC (p < 0.001), training itself was not independently associated (aHR 1.1, 95% CI 0.7-1.8; p = 0.63) with ETV+CPC success.

CONCLUSIONS

This is the largest prospective multicenter North American study to date examining ETV+CPC. Formal ETV+CPC training was not found to be associated with improved procedure outcomes. Specific subgroups of ages within specific hydrocephalus etiologies were identified that may preferentially benefit from ETV+CPC.