predicting success of endoscopic third ventriculostomy ... · true positives (benefits) and false...

J Neurosurg Volume 123 • December 2015

cliNical articleJ Neurosurg 123:1447–1455, 2015

Endoscopic third ventriculostomy (ETV) represents an alternative to shunts in most cases of obstructive hydrocephalus. This procedure allows the CSF to be

internally diverted to the basal cisterns and eventually be reabsorbed by arachnoid granulations, avoiding the need

to implant exogenous material. The success rate of ETV is variable among series, with reported rates between 50% and 94%. In an exclusively adult population, the success rate seems to be somewhat higher, with 55% to 83% of the patients having satisfactory outcomes. However, relative-

abbreviatioNs ETV = endoscopic third ventriculostomy; ETVSS = ETV Success Score; EVD = external ventricular drain; ROC = receiver operating characteristic.submitted June 6, 2014. accepted December 18, 2014.iNclude wheN citiNg Published online July 24, 2015; DOI: 10.3171/2014.12.JNS141240.disclosure This study received funds from the Fondation de l’Hôpital de l’Enfant-Jésus and an educational grant from Medtronic of Canada.

Predicting success of endoscopic third ventriculostomy: validation of the ETV Success Score in a mixed population of adult and pediatric patientsmoujahed labidi, md,1 pascale lavoie, md, msc, Frcsc,1 geneviève lapointe, md, Frcsc,1 sami obaid, md,2 alexander g. weil, md, Frcsc,2 michel w. bojanowski, md, Frcsc,2 and andré turmel, md, msc, cspQ1

1Neurological Sciences Department, Division of Neurosurgery, CHU de Québec—Hôpital de l’Enfant-Jésus, Québec City; and 2Surgery Department, Division of Neurosurgery, CHUM—Hôpital Notre-Dame, Montréal, Québec, Canada

obJect Endoscopic third ventriculostomy (ETV) has become the first line of treatment in obstructive hydrocephalus. The Toronto group (Kulkarni et al.) developed the ETV Success Score (ETVSS) to predict the clinical response following ETV based on age, previous shunt, and cause of hydrocephalus in a pediatric population. However, the use of the ETVSS has not been validated for a population comprising adults. The objective of this study was to validate the ETVSS in a “closed-skull” population, including patients 2 years of age and older.methods In this retrospective observational study, medical charts of all consecutive cases of ETV performed in two university hospitals were reviewed. The primary outcome, the success of ETV, was defined as the absence of reop-eration or death attributable to hydrocephalus at 6 months. The ETVSS was calculated for all patients. Discriminative properties along with calibration of the ETVSS were established for the study population. The secondary outcome is the reoperation-free survival.results This study included 168 primary ETVs. The mean age was 40 years (range 3–85 years). ETV was success-ful at 6 months in 126 patients (75%) compared with a mean ETVSS of 82.4%. The area under the receiver operating characteristic curve was 0.61, revealing insufficient discrimination from the ETVSS in this population. In contrast, calibra-tion of the ETVSS was excellent (calibration slope = 1.01), although the expected low numbers were obtained for scores < 70. Decision curve analyses demonstrate that ETVSS is marginally beneficial in clinical decision-making, a reduction of 4 and 2 avoidable ETVs per 100 cases if the threshold used on the ETVSS is set at 70 and 60, respectively. However, the use of the ETVSS showed inferior net benefit when compared with the strategy of not recommending ETV at all as a surgical option for thresholds set at 80 and 90.In this cohort, neither age nor previous shunt were significantly associated with unsuccessful ETV. However, better outcomes were achieved in patients with aqueductal stenosis, tectal compressions, and other tumor-associated hydro-cephalus than in cases secondary to myelomeningocele, infection, or hemorrhage (p = 0.03).coNclusioNs The ETVSS did not show adequate discrimination but demonstrated excellent calibration in this popu-lation of patients 2 years and older. According to decision-curve analyses, the ETVSS is marginally useful in clinical scenarios in which 60% or 70% success rates are the thresholds for preferring ETV to CSF shunt. Previous history of CSF shunt and age were not associated with worse outcomes, whereas posthemorrhagic and postinfectious causes of the hydrocephalus were significantly associated with reduced success rates following ETV.http://thejns.org/doi/abs/10.3171/2014.12.JNS141240Key words third ventriculostomy; neuroendoscopy; minimally invasive neurosurgery; hydrocephalus

1447©AANS, 2015

Unauthenticated | Downloaded 08/28/20 07:01 AM UTC

m. labidi et al.

ly few studies report on the adult population,1,8,18,29,30,41 and there is significant variation, both in patient characteristics and outcome definition.

In adults, previous studies have established an associa-tion between success of ETV with age at surgery, cause of hydrocephalus, and previous history of shunt.18,29,41 In the pediatric population, Kulkarni et al. recently proposed a clinical prediction rule allowing for the approximation of operative success (ETV Success Score [ETVSS], Table 1) based on these same 3 clinical factors.20 This model has been validated in several pediatric series.7,21,24,39 However, no adult data have been used in the derivation or valida-tion of the ETVSS, and consequently its use and predictive ability in the adult population is unknown. Moreover, the population from which this clinical prediction rule was derived was composed of a high percentage of very young children. Still, age is the most important predictive factor, accounting for 50% of the model variability. It therefore seems reasonable to hypothesize that calibration of this prediction rule will be slightly inadequate for a mixed population.

The primary objective of the present study is thus to validate the ETVSS in a mixed population, including adults and children. Discriminative properties and score calibration for our population will be assessed. Clinical parameters susceptible of being associated with positive outcome following ETV in this population will also be studied.

methodsA retrospective observational study of all consecutive

cases of ETVs was conducted at two tertiary care universi-ty hospitals, the CHU de Québec—Hôpital de l’Enfant-Jé-sus and the CHUM—Hôpital Notre-Dame between 1998 and 2011 through medical chart review. Data collection was done in an anonymized fashion and approval was ob-tained from the CHU de Québec’s and CHUM—Hôpital Notre-Dame’s research ethics committee. Two methods of patient identification were used: medical archives were queried for all ventricular endoscopy cases, and personal surgical logs of the senior authors (A.T. and M.W.B.) were reviewed to identify all ETV cases. Previous authors have described the operative technique for ETV used in both our centers.15,17 All the operative reports were reviewed to verify that a stoma was created in the floor of the third ventricle, which was the minimal criterion for inclusion.

inclusion and exclusion criteriaAll patients 2 years and older who underwent an ETV

in our institutions during the study period were included. After 2 years of age fontanelles are closed in most chil-dren, allowing for a more homogeneous group in terms of CSF physiology. ETVs performed for conditions of a temporary nature (e.g., hydrocephalus preceding surgical treatment of a tumor) or palliation and redo ETVs were excluded from the analyses. Patients with a clinical follow-up of less than 6 months, unless they were treated sur-gically for a problem related to hydrocephalus, were also excluded.

data collection and outcome measuresDemographic, preoperative, and postoperative vari-

ables were collected through medical chart review. Pa-tients lost to follow-up were contacted by telephone to arrange a clinic visit and MRI when possible, and if not, their outcome was assessed at the time of the phone inter-view. Two independent reviewers collected data (M.L. and S.O.); one reviewer in each center. Data were compiled in a single database, and both reviewers used the same definitions for independent variables and outcomes. The cause of hydrocephalus was recorded as per the surgeon’s preoperative assessment. The operative note was reviewed for surgical details, including opacity of third ventricular floor, occurrence of significant bleeding, opening of the Liliequist membrane, concurrent biopsy, and the decision to leave an external ventricular drain (EVD) at the end of the surgery. Postoperative complications that occurred in the first 30 days following the ETV and were documented by the treating physician were also recorded. They were divided into the following categories: “new neurologic deficit,” “seizure,” “fistula,” “infection,” “hyponatremia,” “diabetes insipidus,” and “systemic.” Postoperative Day 1 imaging reports, either CT scan or MRI, were reviewed for any new hemorrhage, excluding only small hemor-rhagic sediments in the occipital horns.

The primary outcome, the success of ETV, was defined as the absence of reoperation or death attributable to hy-drocephalus at 6 months, in accordance with the ETVSS derivation study.20 The secondary outcome is the reopera-tion-free survival.

statistical analysisTo validate the ETVSS in our population, the score was

calculated for each patient included in our study. The pre-dicted success rate from the ETVSS was then compared with the actual success rate of ETV at 6 months (primary outcome) in our population. To that end, included pa-tients were categorized in 9 subgroups, according to their ETVSS results. Actual success rates were determined for each of these strata. The receiver operating characteristic (ROC) curve (equivalent to the c-statistic in dichotomized outcomes) was used to assess the discriminative proper-ties of the score.32 A result superior to 0.7 is usually satis-factory for a clinical prediction rule.26

A Hosmer-Lemeshow test (formal goodness-of-fit test) was conducted to study the ETVSS calibration for the different strata. The calibration slope was also obtained through a linear regression, weighted according to size (number) of each ETVSS stratum, testing the correlation between observed and predicted success for each score category. A calibration slope close to 1.0 usually indicates a good calibration of the score, whereas a score less than 0.8 is generally unsatisfactory.33

Decision-analytical measures were conducted to as-sess the clinical utility of the ETVSS in this population. These measures are based on the relative impact of the true positives (benefits) and false positives (harms) rela-tive to the threshold on the ETVSS.37 This threshold may be defined as the score at which the expected benefit of ETV is equal to the expected benefit of avoiding ETV. Different thresholds were tested because the benefit may



validation of the etv success score in a mixed population

vary according to clinical setting and surgeon and patient preferences. Metrics reported are the net benefit, the rate of reduction of avoidable ETVs per 100 patients, and the decision curves, which compare the use of ETVSS with two alternative clinical strategies (treat all patients with ETV, and treat none with ETV). Details of the statistical method used have been published, and public access to the software used is available.37

We conducted a multivariate analysis using binomial regression, including all variables with p < 0.20 in the uni-variate analysis. Associations were considered significant at p < 0.05. A Kaplan-Meier analysis was used to estimate the total reoperation-free survival. For those patients still alive and who underwent no further hydrocephalus sur-gery, the length of follow-up after surgery was determined with the date of the last clinic visit. To ascertain if exclu-sions due to missing data or loss to follow-up affected the results, best-case and worst-case analyses were done dur-ing univariate and multivariate analyses, ETVSS valida-tion, and Kaplan-Meier analysis.

Because approximately 5–10 predicted events per inde-pendent predictor are usually required to have an adequate sample for statistical inference,22,40 a sample size allowing detection of a total of 30 failures of ETV was required. A pilot study on 41 primary ETV cases performed in the last 2 years in a single institution (CHU de Québec) revealed a failure rate of 17%. Assuming that the failure rate of approximately 17% will be maintained throughout the se-ries, a sample size of 200 procedures is needed to obtain the minimum absolute number of 30 failures required to achieve the ratio events/predictor.

resultsDuring the study period, 205 ETVs were performed in

both institutions. Following application of the exclusion criteria, 168 patients were included in the final analysis (Fig. 1). Eight patients were lost to follow-up at 6 months postsurgery. There were 86 female patients (51.2%) and the mean age was 40 years. Only 12 patients (7.1%) were younger than 10 years, and 137 (81.5%) were 18 years or older. The most common indications were aqueductal ste-nosis with 59 patients (35.1%), and tectal tumor with 39 patients (23.2%). Patients’ characteristics are detailed in Table 2.

At 6 months, 126 patients (75%) achieved the primary

outcome, while the mean ETVSS was 82.4%. Table 3 de-picts actual success rates and predicted success accord-ing to the ETVSS. A higher ETVSS was correlated with higher success rate (p = 0.045). However, the area under the ROC curve (c-statistic) for the score was 0.61 (95% CI 0.52–0.71), which was under the 0.7 mark for adequate discrimination (Fig. 2). A logistic regression was used to assess calibration of the ETVSS model in this cohort (Fig. 3). The calibration slope was 1.01, implying excellent cali-bration for this population. The Hosmer-Lemeshow test was composed of 3 strata (23, 72, and 73 patients). The test results were not significant (p = 0.89), which indicates good calibration of the model (Table 4).

To estimate the impact of age in our cohort, we exclud-ed patients younger than 10 years, which did not result in different measures of discrimination (c-statistic = 0.61) or calibration (calibration slope = 1.01, Hosmer-Lemes-how test p value = 0.77). Excluding age as a factor of the ETVSS (all patients were assigned 50 years for the vari-able age in the ETVSS) did not affect the results either, with metrics of both discrimination (c-statistic = 0.63) and calibration (calibration slope = 1.00, Hosmer-Lemeshow test p value = 0.67) remaining practically unchanged.

Because the cutoff used on the ETVSS may vary from one case to another, the net benefit is reported for a range of different thresholds (Fig. 4). The reductions of avoid-able ETVs were 2 per 100 patients and 4 per 100 patients if the thresholds were set at 60 and 70, respectively. For scores 80 and 90, the “treat none” strategy proved to be superior to both the “treat all” strategy and use of ETVSS to predict success.

In multivariate analysis, the only preoperative factors independently associated with ETV success were cause of the hydrocephalus and sex. When analyzed together, posthemorrhagic and postinfectious hydrocephalus had the lowest success rate (RR 0.54, 95% CI 0.34–0.84), and patients with myelomeningocele also showed a tendency toward lower success rates (RR 0.67, 95% CI 0.41–1.09). Female patients presented a tendency toward better out-comes than male patients, with a RR of 0.85 (95% CI 0.73–1.00). In univariate analysis of surgical variables (Ta-ble 5), the only factor significantly associated with ETV failure was the decision to leave an EVD at the completion of the surgery (p = 0.0092). Best-case/worst-case analysis revealed no significant impact on Kaplan-Meier, univari-ate, multivariate, and ETV validation analyses.

table 1. etv success score*

VariableETVSS

0 10 20 30 40 50

Age <1 mo 1 mo to <6 mos 6 mos to <1 yr 1 yr to <10 yrs ≥10 yrsEtiology Postinfectious Myelomeningocele, IVH,

nontectal brain tumorAqueductal stenosis, tectal tumor, other etiology

Previous shunt Yes No

IVH = intraventricular hemorrhage.* Based on the scale published by Kulkarni et al. in 2009. Reprinted from J Pediatr, 155(2) Aug: Kulkarni AV, Drake JM, Mallucci CL, et al. Endoscopic third ventriculostomy in the treatment of childhood hydrocephalus, 254–259, 259e1, copyright 2009, with permission from Elsevier.

J Neurosurg Volume 123 • December 2015 1449


m. labidi et al.

The majority (77.8%) but not all of the ETV failures oc-curred in the first 6 months after the surgery. Of the 52 pa-tients with ETV failures, 13 had a redo ETV at the time of failure, with the remainder treated by CSF shunt implan-tation. Figure 5 represents a Kaplan-Meier analysis of the reoperation-free survival, which corroborates the finding that most failures occurred early in the follow-up period.

In this series the overall complication rate at 30 days was 11.9% (Table 6), and major complications occurred in 6 patients (3.6%). Two fatalities following ETV occurred in the study population (1.2%). One patient had an ETV following a high-grade subarachnoid hemorrhage with in-traventricular hemorrhage, but later succumbed to pneu-

monia and sepsis. The second patient had hydrocephalus associated with cerebellar lymphoma and was treated by ETV and EVD placement, but experienced upward her-niation in the immediate postoperative period.

discussionIn our cohort of 168 patients 2 years and older, discrim-

ination of the ETV success score was low (c-statistic = 0.61), whereas calibration appeared excellent (calibration slope = 1.01). In addition, results of the Hosmer-Lemeshow test were not significant, also suggesting good calibration. Interpretation of this test and of the calibration slope result are limited, however, by the small number of patients who have low scores on the ETVSS.

Although this is the first appraisal of the ETVSS in a mixed adult and pediatric population, external validation of this scoring system has been assessed by a few studies in pediatric populations. In contrast to our results, Naftel et al. reported, in their cohort of 136 pediatric patients, a c-statistic at 0.74 (95% CI 0.65–0.83) with the mean ETVSS score (76.5% ± 12.5%) predicting very closely the actual success rate (68.4%).24 It seemed, however, that the ETVSS slightly overestimated success in scores ≤ 70.

table 2. characteristics in 168 patients who underwent etv

Variable Total

No. of patients 168Age ≥10 yrs 156 (92.9%)Sex Female 86 (51.2%)Etiology Aqueductal stenosis 59 (35.1%) Nontectal tumor 23 (13.7%) Posthemorrhagic 20 (11.9%) Tectal tumor 39 (23.2%) Myelomeningocele 12 (7.1%) Postinfectious 5 (3.0%) Other 10 (6.0%)Previous shunt 48 (28.6%)Follow-up (mean ± SD) 799 ± 848 days

table 3. predicted and actual success of etv

ETV Success

ETVSS40 50 60 70 80 90

No. predicted

2 2 2 17 72 73

Actual success (%)

1 (50.0) 1 (50.0) 0 (0.0) 11 (64.7) 53 (73.6) 60 (82.2)

Fig. 1. Flowchart. Flow diagram detailing patient inclusions and exclusions.




Likewise, Breimer et al. reported a c-statistic of 0.82 for the ETVSS in their group of 104 children, with a mean ETVSS of 66.2 ± 17.0, accurately approximating a suc-cess rate of 70.2% at 6 months.5 However, in a popula-tion of pediatric patients with hydrocephalus undergoing ETV in Uganda, Warf et al. have shown inadequate cali-bration (Hosmer-Lemeshow p < 0.0001) and insufficient discrimination (c-statistic = 0.57). In their large series of 979 patients, choroid plexus coagulation replaced previous history of shunt placement in the prediction rule, whereas age and cause of hydrocephalus remained significant pre-dictive factors.39 Others have examined the utility of the ETVSS in predicting longer-term (> 6 months) success11,12 or reported good prediction of ETV success, but have not reported on the score’s discriminative properties in their population.7,38

From a statistical standpoint, the low discrimination of the ETVSS in our study, in contrast to the pediatric literature, may be explained by the facts that 50% of the model variability depends on age and that > 30% of pa-tients in the initial training set were < 1 year old. Hence, a group like ours, comprising mostly adult patients (92.9% ≥ 10 years, 81.5% ≥ 18 years), has a considerably reduced distribution of patients in the different categories of the ETVSS. We conducted analyses exploring the impact of age in our cohort, and these showed that exclusion of patients younger than 10 years of age as a factor in the ETVSS did not result in different discrimination or cali-bration metrics. These analyses further emphasize that the variability conferred by age to the ETVSS is practically

negated in our cohort. However, this may have been offset by the better outcomes obtained following ETV among adults, although there is conflicting evidence to this effect in the literature.1,8,18,29,30,41 A higher success rate may also explain in part the excellent calibration of the model in our population.

The discrepancy between the low discrimination and excellent calibration obtained in the present study leaves unanswered the question of how to apply the ETVSS in our population. In fact, even in validation studies in which both calibration and discrimination measures are in accord, they do not translate easily into the clinical realm.16,34,36 Decision curve analysis conducted for the ETVSS in our population allowed us to better define the role of this clinical prediction rule in our population. When comparing the strategy of treating all eligible patients who have undergone ETV (those with obstructive hydrocepha-lus) with a more selective approach using the ETVSS as a tool to identify candidates with a better chance of clinical success, the results of these analyses indicate that the ben-efit conferred by the ETVSS is marginal. If the minimally acceptable success rate is 70% for ETV, the prediction model allowed for the detection of only 4 avoidable ETVs per 100 cases, and if the threshold is 60%, that figure was

table 4. partition for the hosmer-lemeshow test*

Group TotalETV Success ETV Failure

Observed Expected Observed Expected

1 23 13 13.25 10 9.752 72 53 53.05 19 18.953 73 60 59.70 13 13.30

* Details of the partitions and sample sizes used for the Hosmer-Lemeshow test. The result of the test was not significant (p = 0.89), indicating good calibration of the model.

Fig. 2. The ROC curve. Patients were categorized in 9 subgroups, ac-cording to their ETVSS. Actual success rates were determined for each of these strata. The ROC curve plots the sensitivity (true-positive rate) against 1 − specificity (false-positive rate) for the consecutive cutoffs of the ETVSS. The concordance (c) statistic is equivalent to the area under the ROC curve and is 0.61 (95% CI 0.52–0.71). Figure is available in color online only.

Fig. 3. Calibration logistic regression. The calibration slope was also obtained through a linear regression, weighted according to size (N) of each ETVSS stratum, testing correlation between observed and predict-ed success for each score category. The calibration slope obtained is 1.01, with slight overestimation of the actual success with ETVSS ≥ 70 and underestimation of success in ETVSSs 40 or 50. The low number of patients with ETVSSs 60 and ≤ 30 limits statistical interpretation.



m. labidi et al.

reduced to 2 avoidable ETVs per 100 cases. In other words, in clinical scenarios in which 30% or 40% failure rates are acceptable, treating all communicating hydrocephalus with ETV may yield comparable clinical results as would decision-making based on the use of the ETVSS. If the clinician or the patient considers that an acceptable failure rate should be lower, however; say 10% or 20% (scores of 80 or 90 on the ETVSS), the use of the ETVSS showed an inferior net benefit when compared with the strategy of not recommending ETV at all as a surgical option.

In addition to the loss of variability due to age, the low generalizability of the ETVSS to our cohort may also be

due to factors related to study design or outcome definition and, possibly more importantly, to etiology definition. For example, an ETV done in the immediate posthemorrhagic setting may have a lower chance of succeeding than one undertaken for a shunt malfunction in a patient with post-hemorrhagic hydrocephalus. This introduces confusion bias in the ETVSS when applied to adult patients. It may also reflect the different pathophysiological mechanisms of hydrocephalus in this population. One hypothesis that might explain this discrepancy is the underdevelopment of CSF absorption capacity in patients with long-term shunt treatment,35 although previous CSF shunt was not associ-ated with worse outcomes in our series or those of others (see below). A second mechanism by which one can ex-plain the different outcomes, especially when comparing the very young children (< 1 year) with older ETV candi-dates, is altered CSF hydrodynamics in a “closed-skull” environment.14 In fact, closure of the fontanelles probably plays a role in allowing the establishment of a pressure gra-dient between the ventricles and the subarachnoid space, thus favoring stoma patency and CSF absorption through the arachnoid villi. However, CSF physiology continues to evolve after closure of the fontanelles, and this alone cannot explain the different results obtained after ETV, or CSF shunting for that matter.

In the univariate and multivariate analyses, only etiol-ogy was significantly associated with outcome at 6 months postsurgery. In fact, the worst outcomes were achieved in posthemorrhagic and postinfectious hydrocephalus. Both age and previous history of shunt placement did not ad-versely influence the success of ETV in our series. This is in keeping with the majority of previous series reporting on prognostic factors following ETV in mixed or exclusively adult populations.8,29,41 The literature is divided, however, on the prognostic significance of the cause of hydrocepha-lus; there are some studies indicating lower success rates in postinfectious and posthemorrhagic hydrocephalus,9,10,29 and others reporting no such association.8,14,41 The prog-nostic significance of the factors used in the ETVSS in this

Fig. 5. Kaplan-Meier analysis. Kaplan-Meier analysis of ETV “survival” limited to the first 5 years following the procedure. The median follow-up time is 463 days (1st quartile–3rd quartile; 109–1273 days). Figure is available in color online only.

Fig. 4. Chart showing the net benefit of the use of ETVSS compared with “treat all” and “treat none” strategies. Decision curve analysis for ETVSS compared with two simpler clinical strategies: doing ETV for all patients and avoiding ETV in all patients. The y axis shows the net benefit, calculated by summing the true positives and subtracting the false positives, in which the latter are weighted by the relative harm of an unsuccessful ETV compared with the harm of denying a potentially successful ETV.

table 5. univariate analysis of surgical variables

Variable Total Success p Value

Surgical experience 0.5733 6/1999 to 5/2005 38 (22.6%) 29 (76.3%) 6/2005 to 8/2007 46 (27.4%) 35 (76.1%) 8/2007 to 3/2009 41 (24.4%) 33 (80.5%) 4/2009 to 2/2012 43 (25.6%) 29 (67.4%)Liliequist opening* 115 (98.3%) 91 (79.1%) Not done†Opaque 3rd ventricular floor‡ 33 (36.3%)§ 26 (78.8%) 0.1295Significant bleeding‡ 22 (13.1%) 16 (72.7%) 0.7934Biopsy 11 (6.5%) 8 (72.7%) 1EVD left in 36 (21.4%) 21 (58.3%) 0.0092

* Opening of the Liliequist membrane when specified in the operative report (117 cases total).† The p value was not calculated because of the very low number being compared (the Liliequist membrane was opened in all but 2 patients).‡ When mentioned in the operative report. § The appearance of the third ventricular floor was discussed specifically in the operative reports in only 91 cases.




population was thus found to be different from that in the patient cohort used for score derivation.

In the present series, 72.9% of patients with previously implanted shunts (or “secondary ETV”) had a satisfactory outcome at 6 months after ETV, compared with 75.8% in primary ETV cases (p = 0.63). This adds data to the claim that one can achieve good outcomes in the setting of shunt malfunction.3,4,6,23,25 Furthermore, 7 of 12 patients with previously shunted myelomeningocele (58.3%) were reoperation free at 6 months following ETV. Teo and Jones have reported an even better 80% success rate in their group of previously shunt-treated patients with spina bifida when they were > 6 months old at the time of endoscopy.35 Spennato et al. have hypothesized that changes in ventric-ular anatomy induced by chronic shunting may actually increase the probability of ETV success by, among other mechanisms, reexpansion of the subarachnoid compart-ment and production of a pressure gradient between the supra- and infratentorial compartments due to secondary aqueductal stenosis.31

To adapt the ETVSS to the adult population, or to devise a new prediction model altogether, the first step is to iden-tify predictive factors that might be specific to this popula-tion. Previous studies have submitted different models of predicting ETV success, including variables not taken into account by the ETVSS. Some of these models relied on preoperative clinical factors or neuroradiological parame-ters,2,19 whereas others were built on operative findings13,27 or early postoperative variables.28 In our series, none of the studied surgical variables (third ventricular floor opacity, significant bleeding during endoscopy, concurrent biopsy, surgeon experience, and Liliequist membrane opening) were predictive of ETV success, except for the decision by the neurosurgeon to leave an EVD at the end of the case. This association is probably due to the fact that EVDs were, in most cases, selectively inserted in those patients who were deemed poorer candidates for ETV.

In contrast to our results, Greenfield et al., in a mixed adult and pediatric population, devised a grading sys-tem that combined intraoperative findings (thickened or scarred membranes in the subarachnoid space and ab-sence of third ventricular floor pulsatility) and preopera-tive variables, and these authors were able to demonstrate an increasing failure rate with increasing grade.13 Romero et al. showed a similar association between ETV failure

and absence of pulsatility of the floor of the third ventricle at the end of ETV, scarring or thickening in the subarach-noid membranes, and the need for a second attempt be-fore opening of the Liliequist membrane.27 Kehler et al. published a grading system that combined two different neuroradiological parameters (identification of an obstruc-tive cause and downward bulging of the floor of the third ventricle) with progression in symptomatology at presenta-tion. Grades ranged from 0 to 5, with higher grades cor-relating with better outcomes (40% in Grade 3 compared with 95% in Grade 5, p < 0.001).19 However, these findings were not replicated in a subsequent study8 and we were not able to analyze neuroradiological parameters in the pres-ent series, owing to an unacceptably high rate of missing data.

Among the study limitations, of note is the fact that 8 patients (4.5%) were lost to follow-up, even if this number was minimized by telephone interviews. Moreover, best-case and worst-case analyses have demonstrated their lack of impact on study results. Although the projected total number of patients was not achieved, a higher than expect-ed failure rate allowed for adequate statistical inference. The lack of blinded reviewers in the assessment of predic-tive variables and outcome may have introduced an obser-vational bias. In accordance with the outcome definition used in the derivation of the ETVSS, the outcome relied on the surgeon’s decision to reoperate, based on his as-sessment of the clinical and radiological evolution during the follow-up period. In most patients ETV failure is quite clear, with renewed manifestations of intracranial hyper-tension and ventricular dilation. Yet, in a subset of patients symptomatology is more insidious and, in others, ventricu-lar dilation improves only partially after the ETV even if the patient is doing well clinically. Thus, in some cases of clinical successes, one may ask if a superior clinical out-come would have been achieved with shunt implantation.

conclusionsIn this population of patients comprising mostly adults

and children 2 years and older, the ETVSS did not show adequate discriminative ability. However, even in a mixed adult and pediatric population, the ETVSS showed an ad-equate calibration. The ETVSS is only marginally useful in clinical scenarios when 60% and 70% success rates are used as the thresholds for preferring ETV to CSF shunt. In such cases, the ETVSS can result in a reduction of 2 and 4 avoidable ETVs per 100 patients, respectively.

In our cohort, previous history of CSF shunt and age were not associated with worse outcomes, whereas post-hemorrhagic and postinfectious causes of the hydrocepha-lus were significantly associated with reduced success rates following ETV. Surgical variables, including third ventricular floor opacity, significant bleeding during en-doscopy, concurrent biopsy, surgeon experience, and Lili-equist membrane opening were not associated with suc-cess of the ETV at 6 months. Further studies should focus on the predictive factors that are specific to the older popu-lation. A combination of clinical, radiological, and intra-operative findings may be necessary to devise a clinical prediction rule better suited to this group of patients.

table 6. complications following etv

Complication No. %

Morbidity 20 11.9 Pneumonia 2 1.2 Thromboembolic event 1 0.6 CSF fistula 4 2.4 Epilepsy 4 2.4 Motor deficit 3 1.8 Significant IPH/IVH 3 1.8Mortality 2 1.2

IPH = intraparenchymal hemorrhage.



m. labidi et al.

references 1. Amini A, Schmidt RH: Endoscopic third ventriculostomy

in a series of 36 adult patients. Neurosurg Focus 19(6):E9, 2005

2. Anik I, Etus V, Anik Y, Ceylan S: Role of interpeduncular and prepontine cistern cerebrospinal fluid flow measurements in prediction of endoscopic third ventriculostomy success in pediatric triventricular hydrocephalus. Pediatr Neurosurg 46:344–350, 2010

3. Bilginer B, Oguz KK, Akalan N: Endoscopic third ven-triculostomy for malfunction in previously shunted infants. Childs Nerv Syst 25:683–688, 2009

4. Boschert J, Hellwig D, Krauss JK: Endoscopic third ventricu-lostomy for shunt dysfunction in occlusive hydrocephalus: long-term follow up and review. J Neurosurg 98:1032–1039, 2003

5. Breimer GE, Sival DA, Brusse-Keizer MG, Hoving EW: An external validation of the ETVSS for both short-term and long-term predictive adequacy in 104 pediatric patients. Childs Nerv Syst 29:1305–1311, 2013

6. Brichtova E, Chlachula M, Hrbac T, Lipina R: Endoscopic third ventriculostomy in previously shunted children. Minim Invasive Surg 2013:584567, 2013

7. Durnford AJ, Kirkham FJ, Mathad N, Sparrow OC: Endo-scopic third ventriculostomy in the treatment of childhood hydrocephalus: validation of a success score that predicts long-term outcome. J Neurosurg Pediatr 8:489–493, 2011

8. Dusick JR, McArthur DL, Bergsneider M: Success and complication rates of endoscopic third ventriculostomy for adult hydrocephalus: a series of 108 patients. Surg Neurol 69:5–15, 2008

9. Feng H, Huang G, Liao X, Fu K, Tan H, Pu H, et al: Endo-scopic third ventriculostomy in the management of obstruc-tive hydrocephalus: an outcome analysis. J Neurosurg 100:626–633, 2004

10. Fukuhara T, Vorster SJ, Luciano MG: Risk factors for failure of endoscopic third ventriculostomy for obstructive hydro-cephalus. Neurosurgery 46:1100–1111, 2000

11. Furlanetti LL, Santos MV, de Oliveira RS: The success of endoscopic third ventriculostomy in children: analysis of prognostic factors. Pediatr Neurosurg 48:352–359, 2012

12. García LG, López BR, Botella GI, Páez MD, da Rosa SP, Rius F, et al: Endoscopic Third Ventriculostomy Success Score (ETVSS) predicting success in a series of 50 pediat-ric patients. Are the outcomes of our patients predictable? Childs Nerv Syst 28:1157–1162, 2012

13. Greenfield JP, Hoffman C, Kuo E, Christos PJ, Souweidane MM: Intraoperative assessment of endoscopic third ventricu-lostomy success. J Neurosurg Pediatr 2:298–303, 2008

14. Grunert P, Charalampaki P, Hopf N, Filippi R: The role of third ventriculostomy in the management of obstructive hy-drocephalus. Minim Invasive Neurosurg 46:16–21, 2003

15. Hellwig D, Grotenhuis JA, Tirakotai W, Riegel T, Schulte DM, Bauer BL, et al: Endoscopic third ventriculostomy for obstructive hydrocephalus. Neurosurg Rev 28:1–38, 2005

16. Holmberg L, Vickers A: Evaluation of prediction models for decision-making: beyond calibration and discrimination. PLoS Med 10:e1001491, 2013

17. Hopf NJ, Grunert P, Fries G, Resch KD, Perneczky A: Endo-scopic third ventriculostomy: outcome analysis of 100 con-secutive procedures. Neurosurgery 44:795–806, 1999

18. Jenkinson MD, Hayhurst C, Al-Jumaily M, Kandasamy J, Clark S, Mallucci CL: The role of endoscopic third ventricu-lostomy in adult patients with hydrocephalus. J Neurosurg 110:861–866, 2009

19. Kehler U, Regelsberger J, Gliemroth J, Westphal M: Outcome prediction of third ventriculostomy: a proposed hydrocepha-lus grading system. Minim Invasive Neurosurg 49:238–243, 2006

20. Kulkarni AV, Drake JM, Mallucci CL, Sgouros S, Roth J, Constantini S: Endoscopic third ventriculostomy in the treat-ment of childhood hydrocephalus. J Pediatr 155:254–259, 259.e1, 2009

21. Kulkarni AV, Riva-Cambrin J, Browd SR: Use of the ETV Success Score to explain the variation in reported endoscopic third ventriculostomy success rates among published case series of childhood hydrocephalus. J Neurosurg Pediatr 7:143–146, 2011

22. Laupacis A, Sekar N, Stiell IG: Clinical prediction rules. A review and suggested modifications of methodological stan-dards. JAMA 277:488–494, 1997

23. Melikian A, Korshunov A: Endoscopic third ventriculostomy in patients with malfunctioning CSF-shunt. World Neuro-surg 74:532–537, 2010

24. Naftel RP, Reed GT, Kulkarni AV, Wellons JC: Evaluating the Children’s Hospital of Alabama endoscopic third ven-triculostomy experience using the Endoscopic Third Ven-triculostomy Success Score: an external validation study. J Neurosurg Pediatr 8:494–501, 2011

25. Neils DM, Wang H, Lin J: Endoscopic third ventriculostomy for shunt malfunction: What to do with the shunt? Surg Neu-rol Int 4:3, 2013

26. Randolph AG, Guyatt GH, Calvin JE, Doig G, Richardson WS: Understanding articles describing clinical prediction tools. Crit Care Med 26:1603–1612, 1998

27. Romero L, Ros B, Ibanez G, Rius F, Gonzalez L, Arraez M: Endoscopic third ventriculostomy: can we predict success during surgery? Neurosurg Rev 37:89–97, 2013

28. Roytowski D, Semple P, Padayachy L, Carara H: Intracranial pressure monitoring as an early predictor of third ventricu-lostomy outcome. World Neurosurg 80:605–611, 2013

29. Sacko O, Boetto S, Lauwers-Cances V, Dupuy M, Roux FE: Endoscopic third ventriculostomy: outcome analysis in 368 procedures. J Neurosurg Pediatr 5:68–74, 2010

30. Santamarta D, Díaz Alvarez A, Gonçalves JM, Hernández J: Outcome of endoscopic third ventriculostomy. Results from an unselected series with noncommunicating hydrocephalus. Acta Neurochir (Wien) 147:377–382, 2005

31. Spennato P, Ruggiero C, Aliberti F, Nastro A, Mirone G, Cinalli G: Third ventriculostomy in shunt malfunction. World Neurosurg 79:S22 e21–e26, 2013

32. Steyerberg EW: Evaluation of performance, in Clinical Pre-diction Models: A Practical Approach to Development, Validation, and Updating. New York: Springer, 2009, pp 255–280

33. Steyerberg EW: Patterns of external validity, in Clinical Pre-diction Models: A Practical Approach to Development, Validation, and Updating. New York: Springer, 2009, pp 335–360

34. Steyerberg EW, Vickers AJ: Decision curve analysis: a dis-cussion. Med Decis Making 28:146–149, 2008

35. Teo C, Jones R: Management of hydrocephalus by endoscop-ic third ventriculostomy in patients with myelomeningocele. Pediatr Neurosurg 25:57–63, 1996

36. Vickers AJ: Prediction models: revolutionary in principle, but do they do more good than harm? J Clin Oncol 29:2951–2952, 2011

37. Vickers AJ, Elkin EB: Decision curve analysis: a novel method for evaluating prediction models. Med Decis Mak-ing 26:565–574, 2006

38. Vogel TW, Bahuleyan B, Robinson S, Cohen AR: The role of endoscopic third ventriculostomy in the treatment of hydro-cephalus. J Neurosurg Pediatr 12:54–61, 2013

39. Warf BC, Mugamba J, Kulkarni AV: Endoscopic third ven-triculostomy in the treatment of childhood hydrocephalus in Uganda: report of a scoring system that predicts success. J Neurosurg Pediatr 5:143–148, 2010

40. Wasson JH, Sox HC, Neff RK, Goldman L: Clinical predic-




tion rules. Applications and methodological standards. N Engl J Med 313:793–799, 1985

41. Woodworth GF, See A, Bettegowda C, Batra S, Jallo GI, Rigamonti D: Predictors of surgery-free outcome in adult endoscopic third ventriculostomy. World Neurosurg 78:312–317, 2012

author contributionsConception and design: Labidi, Bojanowski, Turmel. Acquisition of data: Labidi, Obaid, Weil, Bojanowski, Turmel. Analysis and interpretation of data: Labidi, Lavoie, Lapointe, Bojanowski, Turmel. Drafting the article: Labidi. Critically revising the article: all authors. Reviewed submitted version of manuscript: all authors. Approved the final version of the manuscript on behalf of all authors: Labidi. Statistical analysis: Labidi, Lavoie.

Administrative/technical/material support: Labidi, Bojanowski, Turmel. Study supervision: Labidi, Turmel.

supplemental informationPrevious PresentationPart of this work was presented, along with accompanying pro-ceedings and abstracts, during the 2013 meeting of the Société de Neurochirurgie de Langue Française (SNCLF), which was held May 27–30, 2013, in Quebéc City, QC, Canada, and during the 2013 meeting of the Canadian Neurological Sciences Federation (CNSF), which was held June 12–14, 2013, in Montréal, QC, Canada.

correspondenceMoujahed Labidi, Neurological Sciences Department, Division of Neurosurgery, CHU de Quebéc, Hôpital de l’Enfant-Jésus, 1401, 18e rue, Quebéc City, QC G1J 1Z4, Canada. email: [email protected].



predicting success of endoscopic third ventriculostomy ... · true positives (benefits) and false...

Documents