Nicholas J. Brandmeir, MD, Brian Anderson, MD, Einar Bogason, MD, Ephraim W. Church, MD, Jonathan Cooke, MD*, Gareth M Davies, MD, Koijan Kainth, MD, John Kelleher, MD, Akshal S. Patel, MD, Russell Payne, MD, Pratik Rohatgi, MD, Emily Sieg, MD, Omar Zalatimo, MD
Department of Neurosurgery, Penn State Hershey Medical Center, Hershey, Pennsylvania, USA
Journal Club Article: Pinto, F, Saad, F, Oliveira, M, et al. Role of Endoscopic Third Ventriculostomy and Ventriculoperitoneal Shunt in Idiopathic Normal Pressure Hydrocephalus: Preliminary Results of a Randomized Clinical Trial. Neurosurgery. 2013; 72(5):845-854.
Significance and importance of study
The prevalence of Idiopathic Normal Pressure Hydrocephalus (INPH) is 22 cases per 100,000 people, with an incidence of 5.5 cases per 100,000 people per year 2. NPH is one of the few treatable causes of dementia and has been shown to respond well to cerebrospinal fluid (CSF) shunting 1. More recently, a group showed that endoscopic third ventriculostomy (ETV) could be an effective treatment for INPH 3. ETV offers some theoretical advantages over CSF shunt placement including avoidance of permanent hardware and less risk of mechanical failures and/or infections. For these reasons, it is important to determine if ETV represents a viable alternative for INPH treatment. This is the first trial to compare ETV and CSF shunting for the treatment of NPH.
Originality of Work
The originality of this study is in its direct comparison of ETV and CSF shunting for the treatment of INPH. Previous studies have examined the value of CSF shunting for INPH in various configurations as well as looking at the complication rates expected from this procedure 1. Others have investigated the success of ETV for the treatment of INPH 3. Further, current guidelines on the surgical management of INPH were unable to site any Class I evidence 1. Since their publication, there have been no published randomized, controlled clinical trials to evaluate the treatment of INPH. This study is the first randomized clinical trial to evaluate the treatment of INPH.
Appropriateness of Study Design
The study was designed as a randomized trial, where patients who met strict diagnostic criteria for INPH were included and randomized at the time of surgery to receive either a CSF ventriculo-peritoneal shunt with a fixed pressure valve or an ETV.
The diagnosis of INPH and inclusion into the trial required improvement by two points or more on the NPH Japanese Scale (NPH Scale) after a tap-test. The exclusion criteria omitted many patients from the study who may have had some element of either comorbid disease or long-standing symptoms. This had the effect of improving the likelihood that the study population would respond to CSF shunting, but may decrease the generalizability of the study as many patients with probable and possible INPH could benefit from CSF shunting and would have been excluded from the study 2,1. A recent study that validated the NPH Scale with more accepted measures of dementia and gait instability used only a 1 point improvement as a treatment success for both CSF shunting and a tap test 4. This would also contribute to decreased patient enrollment and decreased generalizability of the study.
The authors powered their study to detect a difference in the means of the Berg Balance Scale (BERG) of 85%. This required them to have a total sample of 50 patients, with 25 in each arm. The actual accrual of patients was limited to 21 in each arm, and after cross-over, 16 in the ETV treated group and 28 in the CSF shunting group. Further the primary end-point was a decrease in 2 points or more on the NPH Scale. This relationship between the results on the BERG and the NPH Scale is never given in mathematical terms, only that the BERG was dependent on changes in the NPH Scale. Using the difference in percentages the authors report in Table 3, the actual power of their study with their sample size was only 44.8%, short of the beta of 80% that the authors had determined a priori as necessary to draw conclusions from their experimental design. To achieve a power of 80%, the study would have needed 50 patients assigned to each treatment arm. In the study's initial interpretation, their data reached statistical significance making a power analysis of little concern. However, our analysis showed that the difference in treatments did not reach significance, raising concern that the study was underpowered to answer the clinical question of interest.
The randomization was carried out by means of selecting a white envelope at random at the time of surgery; this is a reasonable and effective design. Five patients randomized to receive ETV had their assignment switched to CSF shunting after their anatomy was found to be inappropriate for ETV. They were analyzed as part of the CSF shunting arm of the trial and no intention-to-treat analysis was done. This decision has merit as the ETV is limited by anatomic constraints and analyzing their results in an intention-to-treat fashion may have been the less appropriate option for answering the authors’ main inquiry. Without an intention-to-treat analysis the benefits of randomization are abolished and the study should not be regarded as a truly randomized trial. A better study design may have been to exclude patients with anatomy that precludes ETV and include only patients that could undergo either intervention under investigation. Alternatively, the data could have been analyzed in both an intention-to-treat and as-treated fashion to address concerns of the lost randomization and confounding between treatment arms.
The masking of evaluation was not mentioned, but because of the obvious appearance of ventriculo-peritoneal shunt implants, it seems obvious that evaluations were not blinded. The NPH Scale was the primary endpoint. This scale is largely objective and has high inter-rater reliability, and is valid when compared to more established measurements of the cardinal features of INPH4. Since the strengths of randomized trials is there is their ability to address bias via blinded evaluation, the lack of blinding in the current study limits the confidence in the results.
Adequacy of Experimental Techniques
The authors placed fixed valves in all patients in the CSF shunting arm of the trial. The authors propose in their discussion that placement of adjustable valves could have prevented subdural collections that required reoperation in that treatment arm. The largest study to date on adjustable shunt valves had a subdural collection rate of ~7% while the rate in this study was 19%, so this contention seems reasonable5. The inability of the authors to use the favored technique for CSF shunting hurts the external validity of the study.
The authors analyzed their results by means of a Student’s t-test. While a Student's t-test provides an estimate of the risk of chance, it does not provide an estimate of effect size. Moreover, the primary endpoint was an improvement on the NPH scale, which is a dichotomous variable, not optimally tested with a t-test, which requires a paired, continuous data set. Perhaps, a more appropriate analysis would have been calculation of the relative risk (RR) as well as the number needed to treat (NNT) as this would provide both an estimate of the effect size of the interventions when compared to each other as well as an estimate of the impact of chance on the outcome. With the data provided in Table 3, those values can be easily calculated. The RR of improvement in the NPH Scale by 2 points or more for CSF shunting compared to ETV is 1.54 (0.99-2.39) with a p=.07 and NNT=3.71. With this calculation, it is plain that the intervention difference between the two interventions is not statistically significant and comments on clinical significance must be withheld. Further a calculation of the number needed to harm for CSF shunting based on the complications reported by the authors in Table 2 shows a NNH of 5.2 with p=.06. From these calculations, though neither reached the level of statistical significance, we can be slightly more confident of the harms posed by CSF shunting compared to ETV than the benefits.
The authors’ conclusion that “VPS shows the best gait outcome at 12 months after surgery” is not supported by the data. The study was underpowered to detect the reported level of difference in the NPH Scale; the comparisons did not reach statistical significance based on our statistical analysis and there is no discussion of harms in a cost-benefit type analysis. Further, the study lost the benefit of its randomization when no intention-to-treat analysis was done and evaluation was unblinded. Both limit the level of confidence behind any conclusions drawn from the data. To add to these concerns, the NPH Scale improvement cutoff chosen by the authors is different than that used to validate the score, which again limits its generalizability.
Relevance of Discussion
The discussion was focused and addressed many of the limitations of the study. The authors' claim that the use of programmable valves would have avoided subdural collections and over-drainage is supported by available data5, but the effect on outcomes in this study is unknown.
The authors’ discussion of the “flag signal” during ETV as an intraoperative check on the functioning of the ventriculostomy is interesting. Checking for the “flag signal” after ETV is described in most technical descriptions of the procedure, but the suggestion of immediate placement of a ventriculo-peritoneal shunt if it is not observed is intriguing and bears investigation.
In their discussion, the authors did not address any of the several statistical limitations of the study or the problem of treatment crossover and intention-to-treat vs. as-treated analysis. This would have made interpretation and generalization of the results easier.
Clarity of Writing, Strength, and Organization of Paper
The writing was clear and easy to follow. The paper was well organized. The primary end point, assessment tools, interventions and statistical analyses were all clearly presented and transparent.
Economy of Words
The writing was to the point and appropriate.
Relevance, Accuracy and Completeness of Bibliography
The bibliography was in general broad and complete in covering the surgical management of INPH. There were some notable exceptions. Most importantly, the authors did not cite the study validating their primary end-point4. The study they did site for the use of the NPH Scale also used a 1 point difference as a clinically significant outcome, begging the question why the authors of the current study chose a 2 point difference. This was not addressed in the discussion or methods and a justification would have been helpful in interpretation of the study.
Another noted absence is the latest guideline on the surgical management of INPH including pre-surgical evaluation and diagnosis. This includes a systematic review of the available trials on CSF shunting for INPH as well as an excellent discussion on the pre-operative evaluation of INPH and current treatment recommendations. No study on the surgical management of INPH can be complete without the inclusion or mention of these guidelines.
Number and Quality of Figures, Tables, Drawings
The figures in this study were clear and clearly labeled. Figure 1 was exceptionally clear and showed the allocation of all patients screened for enrollment in the study in an easily read manner.
Table 1 showed the average values in each treatment arm for the assessment metrics. This showed clearly that both arms were the same in regards to these factors. Table 2 was a clear reporting of the harms in each arm of the trial. Table 3 showed the number and percentage of patients in each arm who improved or didn’t improve on the NPH Scale at 12 months. The legend for this table is unclear, and exactly what is displayed is found much earlier in the paper. The legend should have included the name of the metric and the difference considered an improvement so that information could have been gathered at a glance.
A truly randomized, blinded controlled trial comparing ETV and CSF shunting for treatment of INPH remains to be done, and should be the next step. Alternatively, a well-designed, audited, large prospectively collected registry would be helpful and potentially less costly in answering this question. Both approaches would require a multi-center approach. This paper provides an excellent framework for inclusion/exclusion criteria, assessments, and shows some common pitfalls in this investigation.
Another next step could be to evaluate the authors conjecture that immediate placement of ventriculo-peritoneal shunt for patients undergoing ETV and failing to develop a “flag sign” could prevent failure of CSF diversion in this group. This could be done with a randomized trial or cohort study.
1. Bergsneider M, Black PM, Klinge P, Marmarou A, Relkin N: Surgical Management of Idiopathic Normal-pressure Hydrocephalus. Neurosurgery 57:S2–29–S2–39, 2005
2. Brean a, Eide PK: Prevalence of probable idiopathic normal pressure hydrocephalus in a Norwegian population. Acta neurologica Scandinavica 118:48–53, 2008
3. Gangemi M, Maiuri F, Naddeo M: Endoscopic third ventriculostomy in idiopathic normal pressure hydrocephalus: an Italian multicenter study. … 63:62–69, 2008,
4. Kubo Y, Kazui H, Yoshida T, Kito Y, Kimura N, Tokunaga H, et al.: Validation of grading scale for evaluating symptoms of idiopathic normal-pressure hydrocephalus. Dementia and geriatric cognitive disorders 25:37–45, 2008,
5. Zemack G, Romner B: Adjustable valves in normal-pressure hydrocephalus: a retrospective study of 218 patients. Neurosurgery 51:34–36, 2002,