Endoscopic Third Ventriculostomy for Adults: A Prognostic Model for Success

Endoscopic Third Ventriculostomy for Adults with Hydrocephalus: Creating a Prognostic Model for Success - a retrospective multicenter study

Background:

The most common treatment for hydrocephalus is a ventriculoperitoneal shunt (VPS), to divert excess CSF from the ventricles to be absorbed in the peritoneum. The treatment can be applied to different aetiologies of hydrocephalus, but there is a high complication risk both short and long term. A retrospective review of 17,035 adult patients who had undergone their first VPS surgery for hydrocephalus, report that one third (33.4%) of the patients experienced at least one complication, during the follow up (3.9 years), and 22% required revision. 21.4% of the complications occurred within the first year. Another revew of 683 adult patients, found that 32% experienced shunt failure, with 74% occurring within 6 months.

Endoscopic third ventriculostomy (ETV) is an alternative treatment option, creating a passage between the ventricles and the subarachnoid space, by perforating the floor of the third ventricle. It is minimally invasive and leaves no mechanical foreign body behind, thereby avoiding many of the implant complications associated with VPS.

The overwhelming majority of research on endoscopic third ventriculostomy (ETV) is conducted in paediatric or mixed paediatric/adult populations. ETV Success Score (ETVSS) was created in 2009 to predict the outcome of ETV treatment in children. The ETVSS consists of three factors: age, aetiology and shunt history. Based on these factors a score from 0-90 is given, representing the predicted probability of successful ETV outcome 6 months postoperatively. The ETVSS was tested in a mixed population of 168 patients with a mean age of 40 years (range 3-85 years) and found to have inadequate discrimination with an AUC of 0.61 but good calibration.

The ETVSS is based on paediatric populations, and the age differentiation stops after the patient has reached 10 years. 50 out of a possible 90 points are given if the patient is more than 10 years old, making this parameter in the ETVSS redundant when used in adults. Furthermore, the ETVSS does not include several common aetiologies for adult hydrocephalus such as idiopathic normal pressure hydrocephalus (iNPH), subarachnoid haemorrhage (SAH) and long-standing overt ventriculomegaly in adults (LOVA). Previous shunt treatment seems to play an important role in adults as well. A review of 163 adult patients reported an overall success rate of 80%, with patients treated with ETV as the primary treatment faring significantly better than those with a previous shunt, at 87% and 65% respectively.

Radiological findings are not included in the ETVSS, and although most radiological signs of obstruction are subjective evaluations based on the observer's experience, some quantifiable signs have been identified. Downward bowing of the third ventricular floor has been identified as a strong predictor of ETV success. The bowing was measured by placing a line through the chiasma to the top of the mesencephalon or the mamillary bodies. Downward bowing was defined as inferior displacement of the third ventricle floor below this line.

Although there are a few studies analysing long-term ETV survival in adults, most are in paediatric or mixed populations. The existing long-term series on adult patients show most failures occurring shortly after the procedure although late failures are reported. Kaplan Meier curves for ETV survival have an initial steep decline, followed by a gradual fall-off before it seems to stabilise with few failures after a certain point. By determining the composition of patient characteristics these three different parts of the curve. The initial fall off is hypothesised to represent patients without benefit of the procedure and significant symptoms requiring early re-operation. The second group are also ineffective from the beginning, but present with more chronic symptomatology giving more time to evaluate the effect before re-operation. The failures occurring in the stable part of the curve represents initial success with a late closure of the stoma.

With ETV becoming an increasingly widespread treatment for adult hydrocephalus there is a need for a new prognostic model specific for this patient population. The ability to predict who will benefit from an ETV will allow better primary patient selection both for ETV and shunting, reduce additional second procedures because of primary treatment failure, and possibly prevent further unnecessary procedures.

Objective

The purpose of this study is to create a prognostic model to predict the short-term success of ETV for adult patients with hydrocephalus.

Specific aims for this research project is to:

1. Identify factors associated with both success and failure of ETV in adults, to establish a prognostic model.
2. Report on ETV success rates, complications and survival in adult patients at the partici-pating centres.

A success score specific for adults will improve patient selection and can also be used as a communication tool to provide better information and guidance to patients. The prognostic model will be tested in a future prospective study conducted at the participating centres. As well as follow up on the study population in the future to report long-term outcomes.

Hypotheses

Main hypothesis:

An adult ETVSS can be created based on patient demographics, symptomatology, aetiology, shunt history and radiologic findings.

Secondary hypotheses:

• Age is still a relevant factor but has the inverse effect in adults with less successful out-comes with older age.
• Aetiology and shunt history have prognostic value but must be recalibrated to reflect the spectrum of hydrocephalus conditions in adults.
• It is possible to develop a radiological hydrocephalus classification and scoring system providing additional prognostic value.
• There are different characteristics in the failures occurring during the different phases seen on the Kaplan Meier survival curve.

Methodology

Study design:

The study will be conducted as a retrospective review of electronic patient charts and will adhere to the TRIPOD guidelines in the development of the prediction model. A multivariate logistic regression model will be used to identify prognostic factors for success of ETV treatment. This model is expected to be simplified to include only 3 to 4 strong predictors to make it useful in daily clinical practice.

Data collection and monitoring:

Each of the participating centres will be responsible for the data collection in a standardised database, that will include demographic information (date of birth, sex) and details on aetiology (haemorrhage, infection, tumour or cyst, trauma, congenital or idiopathic), symptomatology (acute and chronic symptoms), previous shunt treatment (number revisions, cause of malfunction, duration of shunt treatment), radiological investigations (visible obstruc-tion or radiological signs of obstruction), surgical details (date, equipment, technique, concur-rent procedures, following procedures) complications (intra- and postoperative, length of stay) and follow-up (at 3-12 months and most recent) All ETVs performed at the participating centres will be entered in the database and then includ-ed or excluded based on the below delineated criteria. Reason for exclusion will be registered, and the in-built tools in REDCap used to uncover missing data.

Oversight:

Approval from the Danish Patient Safety Authority (Styrelsen for Patientsikkerhed) and The Danish Data Protection Agency (Datatilsynet) has been granted.

Definition of ETV-success Success is defined as clinical improvement at first follow up, with no further CSF-diversion procedures within the first year of follow up. Clinical improvement will be registered based on the records from the first available follow up, but also include patients where the ETV was deemed ineffective before the patient was discharged or where a second procedure was per-formed during the same admission. If the patient's chart leaves any doubt when registering if the patient's symptoms improved postoperatively, it should be registered as "not improved". If the patient receives further CSF-diversion procedures during the first year, the ETV is considered a failure. Implantation of ICP-monitoring equipment does not render the ETV unsuccessful, un-less it is followed by CSF-diversion. Patients undergoing repeat ETV are counted as failures when calculating the success rate, but results are registered in order to determine the efficacy of re-ETVs. Additional CSF-diversion procedures will be registered for the entire observation pe-riod (beyond the first year) to determine long-term ETV-survival.

Sample size:

There is no clear consensus on the required sample size for prognostic models based on retrospective data. However, the paediatric ETVSS was based on 618 paediatric patients, using 70% (455) of the population for the creation of the model and the remaining 30% (163) for validation. Based on this the minimum number of patients required, is approximately 500. There is no upper limit, as more patients would give a better foundation for the prediction model, especially in the rarer patient categories such as iNPH or hydrocephalus caused by infection or SAH. approximately 220 adult ETV patients have been identified in Copenhagen 2010-2018. With cooperation between several centres the sample should easily meet the minimum required sample size and provide the necessary power to create a robust prognostic model. Approximately 250 ETVs are expected from each of the participating countries, resulting in 1000 patients. The model will be tested in a separate prospective study.

Statistical Analysis:

Patient demographics, aetiology and shunt history, as well as complications, will be summarised using descriptive statistics. The patient's symptoms are categorised as 'improved' or 'not improved' following treatment. If the patient requires subsequent CSF diversion procedures or fails to show clinical improvement, the ETV is considered a failure. Each of the proposed predictors are analysed in a univariate statistical analysis and are subsequently included in a multivariate logistic regression model to construct a unified prediction model. Statistical significance is defined at p<0.05. The model will be validated using K-fold cross validation. Discriminative ability will be assessed using AUROC and calibration using calibration belt plots. Significant missing data will be handled using multiple imputation. Time to ETV failure will be analysed using Kaplan-Meier curves.

Financial plan:

As this is a retrospective chart review, limited funds are required. ST has been partially funded by the Lundbeck Foundation and the Hjerne- og Nervekirurgisk Forskningspulje, Rigshospitalet. No other specific funding has been received for this study.