Introduction

Endoscopic third ventriculostomy (ETV) has been widely applied in pediatric patients over the last several decades [1–6] with many centers reporting successful outcomes. Complications have also been reported, some quite serious [7–9]. Determining the best candidates for ETV has been difficult, with conflicting reports on who are the best candidates, particularly with regards to the effect of age and etiology. Reports have indicated that outcome is a function of age [10, 11], independent of age [12–14], a function of etiology [12, 13, 15–17], or a function of both age and etiology [18]. More recent evidence from larger, and in one case, multicentered series, has supported the finding that age is the main determinant of outcome with younger children, particularly neonates, faring worse [6, 19, 20]. The other standard technique of cerebrospinal fluid (CSF) diversion, a CSF shunt, has rarely been compared to ETV, with results suggesting no difference [21] or a slight advantage in terms of cost-effectiveness for ETV [22]. CSF shunt outcomes are also known to be influenced by age, with younger children also faring more poorly [23, 24]. We compared the quality of life (QOL) outcomes of pediatric patients with hydrocephalus having CSF diversion by either ETV or a CSF shunt using the technique of decision analysis to try and determine if there were any specific age groups where one procedure was favored over another.

Decision analysis compares overall outcome using probabilities of various treatment events, including complications, using a decision tree where the outcome events are scored with a quality-adjusted life year (QALY) value. For each set of values, the decision tree calculates a winner between two options; in this case, ETV or shunt. Accurate estimates of the outcome events probabilities, often extracted from the literature, improves the validity of the analysis. For this comparison, we extracted data from what appeared to be the most accurate and up to date information on these two procedures, a multicentered ETV outcome study [19] and two prospective randomized trials of CSF shunts [25, 26].

Materials and methods

A decision tree with the most common and salient outcomes in either ETV or CSF shunt surgery was constructed using Treeage Software (Williamstown, MA, USA) (Fig. 1). Data on the probability of the various outcome events from ETV was extracted from the raw data from a Canadian cooperative ETV study involving 368 patients [19]. Data for the outcomes for CSF shunt insertion were extracted from two prospective randomized CSF shunt trials; one, a comparison of three shunt designs in 344 patients from nine centers, seven North American and two European [25], and the other, the use of endoscopic ventricular catheter placement in 393 patients from 16 centers, 14 North American and two European [26]. Summary data from the studies is shown in Table 1.

Fig. 1
figure 1

Decision analysis tree created with Treeage Software (Williamstown, MA, USA), used for the analysis, delineating the paths for the various outcomes with the 1-year QALY outcomes for each path

Full size image
Table 1 Comparison of salient patient demographics and outcomes in ETV and shunt trials used to obtain event rates for decision analysis
Full size table

Health utility scores (a measure of QOL) from 0 (death) to perfect health (1.0) for the various outcomes in the decision tree were estimated from the literature [22, 27–31] and are shown in Table 2. QALY outcomes for 1 year following the initial procedure were summated on a month-by-month basis according to the path in the decision tree followed (Fig. 1). For example, if a patient had an uncomplicated ETV (health utility score of 0.7 for that month) and then was healthy with a functioning ETV for the rest of the year (health utility score of 0.95 for each month), the final average health utility score for the year would be 0.93, calculated as \({{\left( {{\text{0}}{\text{.7 + }}\left( {{\text{0}}{\text{.95}} \times {\text{11}}} \right)} \right)} \mathord{\left/ {\vphantom {{\left( {{\text{0}}{\text{.7 + }}\left( {{\text{0}}{\text{.95}} \times {\text{11}}} \right)} \right)} {{\text{12}}}}} \right. \kern-\nulldelimiterspace} {{\text{12}}}}\). The decision analysis outcome was determined for predetermined age groups that seemed to follow reasonable physiologic categories: (0–1 month, 1–6 months, 6 months–1 year, 1–5 years, 5–10 years, and >10 years). Sensitivity analysis for each variable and each outcome’s assigned health utility score was also performed. In sensitivity analysis, each variable or outcome is varied over a large range, typically with ranges that far exceed anything reported in the literature, or would be expected clinically, and the decision analysis repeated. Threshold values of the variable or outcome where the decision changes from favoring one outcome to the other are flagged, and the decision analysis is said to be sensitive to that variable in the region of the threshold. Variables or values so identified, particularly with thresholds in the region of reported incidences or outcome values, are important in decision-making for the outcomes being studied.

Table 2 Health utility values for various outcomes from CSF diversion, used in the calculation of QALY estimates over the year following initial surgery
Full size table

Results

Table 1 shows the salient features of the patients from the ETV and shunt studies. The ETV patients had a higher average age and slightly higher male preponderance. As previously reported, ETV had a significant worse outcome as a function of age, with younger patients faring worse (Fig. 2). Similarly, pooled data from the two prospective randomized shunt insertion trials indicated that the youngest children fared worse (Fig. 3). This was partly accounted for in the shunt insertion trials by an increased infection rate in younger patients (Fig. 4). A comparison of 1-year complication-free survival rates, at each of the specified age groups, is shown in Fig. 5. Both show an incremental improvement with age, with shunts faring slightly better at each age group.

Fig. 2
figure 2

Complication-free survival for patients following ETV according to age group. Outcome improves with increasing age (with permission [19])

Full size image
Fig. 3
figure 3

Complication-free survival for patients following CSF shunt insertion according to age group. Pooled data from two prospective randomized trials [25, 26]

Full size image
Fig. 4
figure 4

Infection rate according to age group for ETV and CSF shunt. Overall average infection rate is shown for ETV for comparison, as the numbers of cases in some age groups were too small to provide an accurate estimate. Infection rate increases with younger age following CSF shunt insertion

Full size image
Fig. 5
figure 5

One-year complication-free survival for ETV and CSF shunt as a function of age group

Full size image

Table 2 shows the health utility estimates for either ETV or shunt and complications used in the decision analysis calculation. The results of the decision analysis are shown in Fig. 6. ETV had a higher net QALY at 1 year compared to CSF shunt for each age category (and pooled over all age groups), but the differences were small. Sensitivity analysis revealed that the outcome was dependent on two factors only:

  1. (a)

    The probability of a severe outcome post-ETV. This had a threshold value of 7%, indicating that once the probability of a severe outcome rose above this level, ETV lost any net QALY benefit and shunt became superior.

  2. (b)

    The assigned health utility score for a patient with either a functioning ETV or CSF shunt, with very similar threshold values of 0.923 and 0.926. If the assigned utility score fell below these levels, then the other procedure would have a net benefit in QALY.

Fig. 6
figure 6

Comparison of QALY outcomes at 1 year for ETV and CSF shunt as a function of age group. ETV has a slightly higher QALY than CSF shunt, but the difference is small

Full size image

None of the other probability values or health utility scores had any significant effect on the decision analysis model when varied over very wide ranges.

Discussion

Determining the best form of CSF diversion in pediatric patients is difficult, particularly in very young patients, especially given the recent information which supports the notion that outcome for ETV is a function of age, with younger patients having a low probability of success [19, 20]. As shown in our pooled data, and perhaps less well-appreciated, is that success rates are also significantly worse in younger children [23, 24] treated with shunts. This outcome is partially accounted for by an increase in infection rate in younger children. Failure in shunt and ETV studies has been defined as any subsequent surgical procedure or death; outcomes which are fairly easy to measure [25, 26]. However, each surgical procedure carries with it the risk of adverse outcomes such as infection or neurological injury, which can have an adverse effect on QOL. QOL is increasingly recognized as an extremely important measurement of outcome of surgical procedures. A validated QOL outcome measure for hydrocephalus, the Hydrocephalus Outcome Questionnaire, has been developed in the form of a parental questionnaire for children over the age of 5 years by some of the authors of this report [28, 29]. Results indicate that QOL is adversely by affected by length of stay in hospital or number of shunt catheters in the head [32], for example. Unfortunately, this measure would not be applicable to many of the young children in the analysis reported in this study because many of them were less than 5 years of age. For this reason, estimates of QALY were extracted from the literature [22, 27–31]. Hopefully, in the future, more accurate estimates will be available, particularly for young children.

The decision analysis in this study allowed QALY estimates for two different CSF diversion procedures to be compared, using event rates extracted from large studies. The two prospective randomized shunt trials data represent the most accurate event rates to date available. The data from the multicenter Canadian endoscopy study, while not all prospective and not randomized, does also represent the largest North American experience to date and was contemporaneous with the shunt trials. They, therefore, represent the best estimates of event rates available.

That ETV is not a better procedure than shunt, by QALY decision analysis, particularly in older age groups, is somewhat counterintuitive. Many of these patients, after all, are leading a virtually normal life devoid of any shunt hardware. However, as the data in the figures remind us, younger children fare worse, and older children fare better, with either operation, with similar event rates. Sensitivity analysis revealed that the outcomes were quite sensitive to health utility score estimates in those with a functioning ETV or shunt with no previous serious complications. We estimated a slightly higher health utility value for successful ETV patients than successful shunt patients, to which the analysis was quite sensitive. This is a debatable point for which virtually no good data currently exists. Accurate and validated estimates for these groups would be quite important to confirm the results of this decision analysis, and we intend to pursue this in future work.

A very important finding in this study is the effect of age on outcome. While this has generally been acknowledged, the data in this study really emphasizes this effect of poorer outcomes in the youngest patients. While several plausible mechanisms for this young age effect have been put forth, i.e., thin skin, propensity for CSF leak, immature immune system, underdeveloped CSF absorption pathways, etc., these are very speculative concepts with no clear pathophysiology identified. Identifying these pathophysiological mechanisms should also be an effort of future research as modifiable risk factors might be determined.

Limitations

There are several limitations to this analysis. The health utility score values were estimated from the literature which is sparse, particularly so for young children. As stated, we hope to pursue, with future work, validated estimates of QOL in patients following ETV and shunts, and these results will help confirm our current findings. It will be particularly important to get accurate health utility estimates for children who are functioning well following these procedures, since the decision analysis appeared to be sensitive to these values. The ETV data was not all prospective in nature and might be subject to bias. As well, although the data we used represent the largest and most accurate assessments of outcome following ETV and shunt, we recognize that direct comparison, even within prespecified age groups, can be difficult due to differences between patients selected for one procedure over the other. There may be other important outcomes which, although rare, might be important in decision analysis.