Introduction

Idiopathic normal pressure hydrocephalus (iNPH) is considered a curable form of dementia characterized by the association of gait disturbances, cognitive impairment, and urinary incontinence and by radiographically evident ventriculomegaly. Ventricular shunting using programmable valves is the current standard of care for iNPH patients, with peritoneum or right heart atrium the most common shunting sites [30]. However, despite several non-invasive and invasive diagnostic tools are commonly used in preoperative workup, the selection of patients that will benefit from shunting is still the main clinical challenge in this syndrome [1, 13, 15, 32, 37].

Among invasive tests to predict shunt response, Katzman’s infusion test evaluates CSF hydrodynamics [16, 22]: the measurement of CSF outflow resistance (R-out) is generally regarded as the most significant parameter to predict improvement after shunt placement [3, 4, 20]; however, a prospective European multicenter study concluded that R-out should not be used as a parameter to exclude patients from treatment [37].

In 2010 our group summarized almost 30 years of experience (1977–2005) in the treatment of iNPH showing that an Intracranial Elastance Index (IEI) above 0.3 was a robust predictor of positive response after shunting [2]. This index was automatically computed by a dedicated software developed at our institution by measuring the slope of the linear regression between the diastolic intracranial pressure (ICP) values and the corresponding amplitude of each CSF pulse pressure wave during an intraventricular infusion test. The test was considered as reliable if the coefficient of determination (R2) was >0.8. All the patients who were selected for shunting using a threshold of IEI≥0.3 had a clinical improvement at 6- and 12-month follow-up indeed. On the other hand, patients with IEI<0.3 did not improve at the same time point follow-up. In the same series, R-out values did not correlate with clinical outcome, and reduction in ventricular size was not associated to clinical improvement, as it was found in only about 40% of improved cases.

On the basis of the above findings, starting from 2006, we systematically used intraventricular infusion test with IEI≥0.3 to select for shunting patients who presented with clinical and radiological stigmata of iNPH.

The aim of this paper is to verify the accuracy of IEI in predicting response to shunt at both short- and long-term follow-up in patients with suspected iNPH and at least 3 years of follow-up.

Methods

We reviewed clinical data of 64 consecutive patients who underwent ventricular shunting for iNPH between January 1, 2006, and December 31, 2015, based on a positive ventricular infusion test, namely, a test that showed an IEI≥0.3 and an R2>0.8. The method used for the ventricular infusion test has been previously described [2]. All patients gave written informed consent for the analysis of clinical data.

Patients with secondary NPH (e.g., post-traumatic, post-hemorrhagic, post-infective, etc.) or evidence of long-standing overt ventriculomegaly (LOVA) were excluded as, due to different onset mechanisms [26], different parameters or thresholds could be appropriate for this subset of patients.

All iNPH patients were selected for the ventricular infusion test according to clinical evidence of partial or complete gait/urinary/cognitive clinical triad and evidence of ventriculomegaly with increased Evan’s ratio, disproportionally effaced superior frontal sulci with respect to other sulci and Sylvian fissures, and reduced callosal angle at brain high-field MRI (≥1.5 T). All the patients also underwent pre-admission neuropsychological testing by dedicated and experienced neurologists and neuropsychologists. Patients who did not fulfill the above clinical, radiological, and neuropsychological selection criteria were excluded from the intraventricular infusion test. Patients who showed an IEI<0.3 were excluded from shunting and followed-up clinically and radiologically.

All the patients selected for surgery had a ventriculo-peritoneal shunt using a Codman® Hakim® programmable valve. The proximal catheter was placed in the frontal horn of the right lateral ventricle in all cases. Valve initial opening pressure was set mainly according to infusion test opening pressure and adjusted by patients’ anthropometric characteristics as height and abdominal circumference.

Data collection

All the ventricular infusion tests were reviewed, and the following parameters were collected: preoperative duration of symptoms, opening pressure, closing pressure, IEI, R-out, and ICP wave morphology before infusion and at the end of infusion according to the 4 class classifications based on changes in the relations of the three ICP peaks (percussion, tidal, and dicrotic peak) previously reported by our group [31].

Valve opening pressure and pressure modifications, shunt revisions/shunting site modifications, and surgical complications were recorded.

Clinical data were collected at 3 time points: preoperative period, at 12-month follow-up and at last follow-up, which occurred either in our outpatient clinic or via telephonic interview with patient/caregiver in January 2019. Clinical data were classified using the Krauss scale [24] and the modified Rankin scale (mRS): patients with mRS≤2 are generally considered to have a good level of function [10].

Patients were classified at 12 months and at last follow-up in three categories according to each of the above evaluation scales: improved, if there was an improvement of at least one point at total Krauss score or at mRS scale; unchanged, namely, no changes at total Krauss or mRS score; and worsened, decrease of at least one point at total Krauss or mRS score.

Results at last follow-up were compared both with preoperative status and with patients’ status at 1 year. Improvement was considered as a good outcome, while stable or worsened patients were grouped as poor outcome (not improved).

Functional status was regarded as good if mRS≤2 and poor if mRS>2.

Statistical analysis

Mean Krauss score and mRS at 1-year and last follow-up were compared with preoperative values using a paired samples Student’s t test.

Receiver operating characteristic (ROC) curves were used to determine IEI and R-out diagnostic accuracy in predicting shunt response at both 1-year follow-up and last follow-up.

Sensitivity, specificity, positive likelihood ratio (LR+), and negative likelihood ratio (LR−) of initial waveform characteristics to predict clinical outcomes response were calculated.

Independent samples Student’s t test and Chi-squared test were used to analyze the differences of age, gender, length of symptoms (in months and dichotomized as ≤ 12/>12 months), IEI, R-out, and preoperative/1-year clinical status according to Krauss scale and mRS (the latter both as continuous value and dichotomized as ≤2/>2), between patients who were improved and not improved at Krauss and mRS at 1 year, last follow-up, and between last follow-up and 1 year. If the equal variance assumption was violated, a Welch test instead of a Student’s test was used. Independent samples Student’s t test was also used to assess possible differences in IEI, R-out, and preoperative clinical status between patients with different length of symptoms (≤12/>12 months).

Multivariate binomial logistic regression was also used to analyze the association between outcomes at both Krauss and mRS (improved/not improved) and length of symptoms, age, IEI, R-out, preoperative Krauss and mRS, and gender at 1 year, last follow-up, and between last follow-up and 1 year. Multivariate linear regression was used to determine the association between the same covariates and the total score at Krauss and mRS at 1-year and last follow-up.

Significance level was set at 0.05. Statistical analysis was performed using IBM SPSS Statistics V22.0.

Results

Study population

Sixty-four patients were analyzed. Demographics are reported in Table 1. All the patients had a minimum follow-up of 3 years (mean 6.6±2.5). One patient (1.5%) had a bleeding related to the intraventricular infusion test without any neurological sequelae. Twelve patients (18.7%) underwent revision surgery: 8 patients for abdominal catheter repositioning; 3 patients with previous history of abdominal surgeries had a conversion to ventriculo-atrial shunt since a poor peritoneal absorption was suspected; 2 patients (3%) developed a chronic subdural hematoma, one treated by transitory increasing of valve opening pressure up to the maximum setting level (20 cmH2O) and the other requiring surgical evacuation and temporary shunt ligation. Twenty patients (31.3%) had a reduction of the valve opening pressure during follow-up in order to try to improve clinical benefit.

Table 1 Demographics
Full size table

Clinical outcomes at 1 year

All the patients were alive at 1 year: 40 (62.5%) and 31 (48.4%) had a good outcome (improved) at Krauss scale and at mRS, respectively. Among patients with poor outcome (not improved) at Krauss score, 23 (36%) had an unchanged score, and one (1.5%) complained worsening of symptoms. All the 33 patients (51.6%) who had a poor outcome at mRS had an unchanged level of function, namely, none had a worse score at mRS.

Clinical outcomes at last follow-up

Overall, mean scores of Kraus scale both as total score and as score per domain, and of mRS, were all significantly reduced at both 1-year and last follow-up with the exception of cognition at Krauss at long term (Table 2).

Table 2 Descriptive results
Full size table

Comparing the last follow-up with the follow-up at 1 year, 17 (30.4%) and 12 (21.4%) patients had an improvement of their outcome according to Krauss scale and mRS, respectively.

Outcome predictors analysis

Neither R-out nor IEI were shown to be accurate in predicting the outcome as the area under the curves (AUC) were all under 0.70 with the exception of elastance with respect to the change in mRS from first year to last follow-up (AUC 0.784; 95% CI 0.624–0.943). Also the waveform morphology was not proved accurate as LR+ and LR− were below 5 and above 0.2. Based on these results, waveform was not included in univariate and multivariate analysis.

A significant difference in both neurological and functional preoperative status was seen between patients who improved after shunting compared to patients who did not improve, at both 1-year follow-up and last follow-up. In particular, improved patients had worse preoperative Krauss and mRS than unimproved ones. However, after treatment, improved patients showed significant worse scores at both scales than not improved patients. Also R-out and IEI showed some significant differences between the two outcome groups (Tables 3 and 4).

Table 3 Independent samples t test
Full size table
Table 4 Chi-squared test
Full size table

Patients with length of symptoms >12 months showed significant lower values of R-out compared with patients treated at shorter onset of symptoms (mean 13.3 vs. 17.8, p=0.016). No differences of mean IEI, Krauss, and mRS were seen between patients with length of symptoms ≤/>12 months.

Regression analysis showed an association between preoperative clinical status and post-operative outcomes, with a significant association also between IEI and last follow-up outcomes at both Krauss and mRS (Tables 5 and 6). At last follow-up, an association between IEI and mRS and between R-out and Krauss score was also found (Table 6).

Table 5 Binomial logistic regression: outcomes (improved or not improved) according to Krauss scale total score and mRS
Full size table
Table 6 Linear regression: outcomes according to Krauss scale total score and mRS total value
Full size table

Results are summarized in Tables 2, 3, 4, 5, and 6 and Figs. 1 and 2.

Fig. 1
figure 1

Intracranial Elastance Index (IEI) and outcome at Krauss scale at 1-year (a) and last follow-up (b) and at mRS at 1-year (c) and last follow-up (d). White bars, improved; gray bars, unchanged; black bars, worsened. Outcomes are compared to preoperative status

Full size image
Fig. 2
figure 2

CSF outflow resistance (R-out) and outcome at Krauss scale at 1-year (a) and last follow-up (b) and at mRS at 1-year (c) and last follow-up (d). White bars, improved; gray bars, unchanged; black bars, worsened. Outcomes are compared to preoperative status

Full size image

Discussion

iNPH is currently the only form of dementia that can be treated by surgery, with high chance of neurological improvement after a correct diagnosis, which is clinical and radiological. However, since the literature reported percentage of shunt non-responders ranges between 20 and 40% of patients [11], some ancillary, invasive tests have been developed to help clinicians to select patients that are more likely to improve after surgical treatment [19, 22, 32].

Among ancillary tests, two different categories of tests are described: subtraction tests, namely, tap test or prolonged lumbar drainage, and infusion tests, either lumbar or ventricular. Several studies have previously addressed the question of the predictive role of these invasive tests as standalone procedure or in combination: while specificity and positive predictive value are elevated, low sensitivity and negative predictive value are generally reported [4, 7,8,9, 20, 32, 37]. Briefly, in 2015 the American Academy of Neurology (AAN) guidelines summarized the literature evidences of the percentage of iNPH patients improving with shunting according to the results of ancillary tests [13]: 96% of patients with a positive tap test improved after treatment, as well as 94% of patients with a positive external lumbar drain test and 52% of patients with increased Rout at lumbar infusion test. However, a clinical improvement after shunting was seen in 71% of patients with a negative tap test, in 33% of patients with a negative external lumbar drain test, and in 69% of patients with no increased Rout at lumbar infusion test.

Historically our group preferred ventricular to lumbar infusion test as in the past decades a prolonged overnight ICP monitoring was performed through the same access. Intraventricular infusion test was also deemed as more reliable than lumbar infusion test [6] and allowed a deeper insight of pathophysiology of iNPH [2]. Summarizing a clinical experience of 3 decades, an IEI≥0.3 at ventricular infusion test was detected as highly accurate in prediction of shunt response, with a 100% positive response at 1 year. Moreover, patients with IEI<0.3 had no improvement at 1 year. This diagnostic accuracy outstood the predictive values reported for other ancillary tests.

Therefore, despite intraventricular infusion test is a more invasive diagnostic tool than spinal infusion or subtraction diagnostic tests, our results encouraged its use in our clinical practice.

However, present results at 1 year are less satisfactory than our previously published ones. A number of reasons could explain these differences. One could be an increased indication to shunt surgery based on the positive influence of our previous findings about IEI. Indeed, several conditions may mimic iNPH, with only about 30% of patients presenting with enlarged ventricles and suggestive clinical picture actually harboring iNPH [25]. Moreover, a beta-amyloid deposit and hyperphosphorylated tau have been reported in up to 60% of patients undergoing shunting for iNPH [19]. Worse post-operative outcomes, particularly in the cognitive domain, are reported in these patients [18]. The possible interweaving of iNPH, Alzheimer’s disease, Parkinson’s disease, Parkinsonisms, and Lewy body disease stresses the need of a multidisciplinary team to select iNPH patients to submit to neurosurgical treatment [15, 24, 25, 27, 33, 35]. However, all our patients underwent accurate neuroradiologic assessment and neuropsychological testing before neurosurgical management, and none of them received a diagnosis of other neurological conditions during post-operative follow-up. Indeed, while some studies on the association between tap test and amyloid deposits are available [27], a dedicated study to compare infusion test metrics between patients affected by standalone iNPH and by concomitant iNPH and neurodegenerative diseases is currently lacking.

Moreover, in the present study, shunt efficacy may have been reduced by an elevated initial valve opening pressure setting or by a lack of progressive adjustment of the optimal valve setting for individual patient in the context of an evolving disease. Some subtle shunt malfunctions may have also been missed. These have been recognized as frequent causes of long-term loss of efficacy of the treatment [12].

Nonetheless, this series shows that at long-term follow-up, outcome according to Krauss scale and mRS was good in 64.3% and 55.4% of cases, respectively. Maybe more relevant, there was an overall shift towards better levels of function as 57% of patients were in mRS≤2 at 1 year, compared to 25% in the preoperative period, and this percentage was maintained at last follow-up. Namely, the majority of patients were independent more than 6 years after treatment. This trend towards a long-term improvement is different to what has been recently reported by other groups, whose early results were better than long-term ones [19, 21, 28, 34].

The analysis of outcome predictors showed that IEI is associated with outcome, as it was significantly lower in patients who improved at mRS at last follow-up (Table 3), and showed a strong positive association (odds ratio>30) with outcomes at last follow-up at both Krauss and mRS, meaning that increasing IEI significantly increases the odds of poor outcome (Table 5, Fig. 1). Increasing IEI has also significant association with higher mRS at last follow-up (Table 6). On the other hand, we previously showed a lack of improvement in patients with IEI<0.3 [2]. This seems to suggest that both a low and an elevated IEI can predict a poor response to shunting. Further studies are needed to confirm these data and to try to identify also an upper threshold of IEI values for non-responders.

Moreover, patients who were in worse clinical status before shunting had a higher chance to improve their outcome at both 1-year and last follow-up (Tables 3, 4, 5, and 6). Indeed, 21 (44%) of the 48 patients who were operated upon with mRS>2 had an improvement to mRS≤ 2 at 1 year (p<.001; Table 4).

Lastly, our results confirm that R-out is not an accurate parameter to predict the shunt response. While R-out was initially regarded as the most sensitive parameter to predict surgical outcome [3,4,5, 20], more recently, it has been shown that this value could be influenced by length of pathology [7] and should not be used to preclude patients from treatment [37]. In agreement with data reported by Czosnyka et al. [7], duration of symptoms longer than 1 year was associated to lower mean R-out in our series (p=0.01).

Despite the above considerations, our current results suggest that no single one CSF dynamics parameter tested by means of the intraventricular infusion test can be used as absolute predictor of shunt response.

Indeed, differently from our previous more encouraging experience, results from this series are in line with those reported by other groups using other one or more diagnostic tests such as tap test, extended lumbar drainage, or spinal infusion test and even to some studies where only clinical-radiological parameters were used to select patients for shunting [14, 17, 19, 23, 29, 32].

This may raise the legitimate doubt about the opportunity to perform a diagnostic test that is as invasive as the actual treatment in its cranial part. Regarding safety, in the same time period considered in this manuscript, a strict pre-selection of patients on the basis of clinical, neuropsychological, and neuroradiological allowed us to reduce to about 7% the number of patients with suspected iNPH that had a negative intraventricular infusion test (IEI<0.3) and were therefore not elected for shunting. None of these patients had complications from the ventriculostomy. Among the 64 patients with a positive test that underwent definitive shunting, only 1 (1.5%) had a complication (Table 1), namely, a bleeding with no neurological sequalae, related to ventriculostomy/ventricular infusion test, while no infections were recorded. This complication rate is in line with the intrinsic risk of the ventricular puncture needed for a shunting procedure. However, despite our experience confirms that the intraventricular infusion test is safe for patients and easy to be performed by a trained group of neurosurgeons using a dedicated computer software, its supposed superior predictive value over other ancillary tests is not confirmed by our results. Therefore, this invasive procedure should be subject to special consideration. It could indeed represent the last tier of a diagnostic algorithm that privileges alternative diagnostic methods, including the lumbar subtraction and/or infusion tests, and reserves the ventricular infusion test to those patients with a high suspicion of iNPH that could not be fully confirmed by previous less invasive tests.

Furthermore, lumbo-peritoneal shunting is emerging as a possible alternative in iNPH treatment, with similar outcomes but still higher failure rate compared to ventriculo-peritoneal shunting [23, 36]. Should lumbo-peritoneal shunting be recognized as the treatment of choice in iNPH in the next future, the role of intraventricular infusion test could be further diminished.

Study limitations

Main limitations of the present series are its retrospective nature, the limited number of cases, and the pre-selection of patients according to IEI. The latter prevented to calculate the negative predictive value of IEI. However, we previously showed that patients below 0.3 of IEI had no benefit from shunting at 6- and 12-month evaluation [2].

A prospective, multicentric, randomized trial could further support the role of IEI in the prediction of both short- and long-term outcomes after shunting and compare it to the diagnostic accuracy of other ancillary tests.

Conclusions

iNPH is a benign pathology that can severely limit the quality of life of patients, with up to 75% of patients living in a dependent functional status (mRS>2). Accurate selection of patients according to clinical, neuroradiological, and neuropsychological data is essential. Intraventricular infusion test is an invasive adjuvant method to select patients who will benefit from surgery. Preoperative status and IEI are significantly associated with last follow-up outcomes. Our past and present data show that a threshold of 0.3 of IEI can be helpful to predict the response to shunt at both short- and long-term follow-up. However, no single one CSF dynamics parameter analyzed in this series seems to outstand the predictive values of other ancillary tests. Further studies are needed to investigate whether also an upper threshold of IEI exists to predict poor response to shunting.