Can the radiological scale “iNPH Radscale” predict tap test response in idiopathic normal pressure hydrocephalus?

BACKGROUND
Idiopathic normal pressure hydrocephalus (iNPH) presents typical radiological signs that have been summarised in a semi-quantitative scale named the iNPH Radscale. However, the iNPH Radscale's predictive value for response to cerebrospinal fluid (CSF) tap test has never been studied. This study aims to investigate if the iNPH Radscale can predict locomotion improvement after CSF tap test.


METHODS
A total of 100 patients with iNPH (age: 76.3 ± 7.9, gender: 36% female) were included in this retrospective study. Two raters, blinded to the response of the CSF tap test, evaluated the iNPH Radscale and its seven subitems (Evan's index, callosal angle, size of temporal horns, narrow high-convexity sulci, dilated Sylvian fissures, focally dilated sulci, and periventricular hypodensities). Locomotion improvement was assessed by the Timed Up and Go (TUG) performed before, and 24 h after, the CSF tap test.


RESULTS
The iNPH Radscale (total score) was similar between the CSF tap test responders and non-responders (responders: 8.31 ± 1.96, non-responders: 9.18 ± 2.51, p = 0.128). However, the temporal horns score was smaller in the responders group (1.66 ± 0.57 versus 1.94 ± 0.24, p = 0.045), even after adjusting for age, gender, education level, white matter changes, and global cognition (β: -0.250, C.I. 95%: [-3.185; -0.161], p = 0.031).


CONCLUSION
The iNPH Radscale (total score) does not predict locomotion improvement after CSF tap test, while a smaller temporal horns score at baseline is associated with a positive tap test responder status.


Introduction
Idiopathic normal pressure hydrocephalus (iNPH) is characterised by locomotion disturbance, cognitive impairment, and urinary incontinence, with enlarged ventricles at brain imaging [1,2]. Its treatment relies on an invasive neurosurgical shunt procedure [3]. iNPH is often associated with comorbid neurological conditions, such as Alzheimer's disease or other neurodegenerative conditions, which interfere with the reversibility of symptoms [4]. Various neuroimaging biomarkers, such as the hippocampal volume or microstructural changes in specific white matter tracts [5,6], have been studied to identify iNPH patients with good prognosis after shunt surgery. However, these neuroimaging features rely on complex image processing approaches, which are difficult to apply in clinical practice. Furthermore, these neuroimaging biomarkers mainly apply to MRI and not to CT-Scans that are largely performed in iNPH patients.
Recently, the iNPH Radscalea semi-quantitative radiological scale that includes seven radiological parameters (Evan's index, narrow sulci, Sylvian fissures, focally enlarged sulci, temporal horns, callosal angle, and periventricular hypodensities) -has been validated to identify patients with iNPH [7]. The severity of the iNPH Radscale correlates with the severity of clinical symptoms in iNPH patients [7], and demonstrates high specificity and sensitivity for distinguishing iNPH patients from asymptomatic individuals in the general population [8]. Furthermore, the iNPH Radscale demonstrates a similar validity when assessing brain imaging with CT or MRI [9]. However, the iNPH Radscale and its subitems have never been studied for their ability to identify iNPH responders from non-responders.
This study aims to investigate if the iNPH Radscale and its subitems can predict if iNPH patients improve their locomotion after cerebrospinal fluid (CSF) tap test (i.e. responders), by using the Timed Up and Go (TUG) -a largely used clinical test that has been also validated in iNPH [10,11]. Establishing this would contribute to the identification of appropriate iNPH candidates for shunt surgery.

Sample
A total of 179 consecutive patients enrolled in the Geneva Protocol [12] from September 2009 to February 2019 were selected for this study. The study procedures were previously described [12]. Briefly, inclusion criteria for this analysis were: patients with (i) a diagnosis of possible or probable iNPH that fulfilled the iNPH consensus guidelines criteria [13], (ii) a brain MRI or CT, and (iii) a Timed Up and Go (TUG) test performed before and 24 h after CSF tap testing. Seventy patients were excluded because they were diagnosed with iNPH mimics, and nine due to missing pre or post CSF tap test locomotion assessment. The final sample consisted of 100 patients with a possible or probable iNPH diagnosis. A total of 41 of the 100 patients performed a shunt surgery; 37 of which (90.2%) improved their gait according to the subjective postoperative evaluation from a neurosurgeon at one-year post-shunting. The study protocol was approved by the ethical committee of Geneva University Hospitals. Patients gave informed consent for their participation.

The iNPH Radscale
The iNPH Radscale is a semi-quantitative imaging scale validated to screen for iNPH [7]. Scores range from 0 to 12, with higher values indicating more severe radiological signs. The iNPH Radscale includes seven subitems: Evan's index, narrow sulci, Sylvian fissures, focally enlarged sulci, temporal horns, callosal angle, and periventricular hypodensities. Details about the method used to evaluate these subitems can be found in the original iNPH Radscale article [7]. In our cohort, iNPH Radscale scores were evaluated by two independent raters (TL and GA) blinded to the CSF tap test responder status.

Tap test responder status
Tap test responder status was defined with the Timed Up and Go (TUG), as the TUG has been validated for evaluating improvement in locomotion in iNPH patients [11,14]. The TUG was measured before, and 24 h after, the CSF tap test. Participants were asked to perform the TUG at their self-selected, normal speed. We used a very sensitive approach for defining the responder status: TUG improvement was defined by any improvement in TUG time 24 h after the CSF tap test, when compared to the pre-CSF tap test TUG. Furthermore, a sensitivity analysis was carried out at TUG improvement cut-offs between 0% and 10%. The formula used to measure TUG improvement (in %) was: (post tap TUG) − (pre tap TUG) 0.5*[(post tap TUG) + (pre tap TUG) ] *100

Covariates
The NPH grading scale [15] was used to quantify the severity of iNPH symptoms. Global cognition was evaluated using the Mini-Mental State Examination (MMSE). Behavioural symptoms were evaluated using the Starkstein for apathy [16] and the Hospital Anxiety and Depression (HAD) scale for depression and anxiety [17]. The Global Health Status (GHS: range 0-10) was defined based on the presence of diabetes, chronic heart failure, arthritis, hypertension, depression, stroke, Parkinson's disease, chronic obstructive pulmonary disease, angina, and myocardial infarction [12]. White matter lesions were rated with the age-related white matter changesa validated semiquantitative scale [18]. Medial temporal lobe atrophy was evaluated using the Scheltens scale (average score) [19][20][21].

Statistics
Descriptive statistics were compared between responders and nonresponders with a Mann-Whitney U test. Multivariate linear regression adjusted for age, gender, education level, white matter changes, and global cognition was performed to assess the association between responder status (independent variable) and temporal horns (dependent variable). All analyses were conducted using SPSS version 25 (IBM Corp., Armonk, NY., USA).  [7]. Both groups presented a similar total composite iNPH Radscale score (p = 0.128) ( Table 2). However, for individual subitems of the iNPH Radscale (Table 2), the tap test responders exhibited a smaller temporal horns score compared to the non-responders (p = 0.045). This difference in temporal horns scores (using the semi-quantitative iNPH Radscale), was confirmed by the quantitative measurement of temporal horns width (Fig. 1): 7.71 ± 2.47 mm for the tap test responders, versus 9.12 ± 2.29 mm for the non-responders (p = 0.031). This association between smaller temporal horns width and positive tap test responder status remained significant even after adjusting for age, gender, education level, white matter changes, and global cognition (β: -0.250, C.I. 95%: [− 3.185; − 0.161], p = 0.031). Scheltens' scale for medial temporal lobe atrophy yielded a score of 1.63 ± 0.86 for tap test responders vs 2.27 ± 0.99 for non-responders (p = 0.010), and 1.74 ± 0.91 for the entire sample of iNPH patients.

Results
The sensitivity analysis using cut-offs of TUG improvement between 0% and 10% yielded similar results to Table 2. The association between smaller temporal scores and positive tap test responder status was not significant anymore at a cut-off of TUG improvement ≥6%. However, when using a TUG improvement cut-off ≥10% for responders and < 0% for non-responders, the results closely resemble those shown in Table 2, including for the temporal horns observation (temporal horns scores 1.64 ± 0.61 n = 61 vs 1.94 ± 0.24 n = 17, respectively, p = 0.044). Furthermore, all of our findings were similar when defining TUG improvement in % as (post tap TUG)− (pre tap TUG) (pre tap TUG) *100.

Main findings
While the iNPH Radscale (total score) does not predict locomotion improvement after CSF tap test, a smaller temporal horns scorea subitem of the iNPH Radscale -is associated with improvement in locomotion.

Temporal horns
Smaller temporal horns (iNPH Radscale score and quantitative measurement) are associated with a positive tap test responder status in iNPH patients. This result is in contradiction with previous reports showing that increased temporal horns predict positive outcomes after shunt surgery [22,23], as well as with studies showing that temporal horns cannot predict shunt outcome [24]. Virhammar's study [23] reported 13% smaller average temporal horns widths than our study, using a similar measuring method. Few explanations may contribute to these discrepancies. Patients in our sample are older than those reported in previous studies [22,23], suggesting a longer disease duration with more brain atrophy. Furthermore, the older age and wider temporal horns in our sample may reflect more comorbid Alzheimer's pathology [25], which could be linked with a poorer responder status. To test this hypothesis, a post-hoc evaluation of medial temporal lobe atrophy was carried out using Scheltens' scale. This analysis shows that the tap test responder group is below the threshold for Alzheimer's Disease (≥2 for individuals aged 75-84) [21], while tap test non-responders are above. It should, however, be noted that Scheltens' scale has not been validated to distinguish patients with Alzheimer's disease from iNPH. Future studies should investigate the ability of the temporal horns subitem of the iNPH Radscale to distinguish temporal lobe atrophy from ventricular enlargement. Another explanation may be associated with the definition of responder status. Most prognostic studies define response to treatment as any improvement in locomotion, cognition or urinary symptoms. Here, we restricted the definition to any improvement in locomotion, as locomotion represents the key symptom of iNPH [26,27].

The iNPH Radscale
The iNPH Radscale total score does not predict the CSF tap test responder status. This finding is supported by previous studies demonstrating that a combination of radiological parameters is unable to predict response to shunt surgery in iNPH [22,24,28,29]. The iNPH Radscale linearly correlates with iNPH symptom severity [7], indicating that more severe symptoms are associated with more severe radiological signs. However, it is well known that disease severity increases with disease duration [30], and a longer disease duration has a negative effect on shunting outcomes [27,[31][32][33]. Consequently, more severe iNPH Radscale scores would be associated with worse treatment outcomes,  thus reducing the ability of the iNPH Radscale to predict the responder status. Furthermore, focusing on specific radiological signs rather than on composite scores, could uncover underlying comorbidities such as Alzheimer's pathology.

TUG and CSF tap test
The CSF tap test is not a perfect proxy of a shunt, but TUG change after a tap test is a good predictor of shunt outcomes [10,11]. Previous studies have used a TUG improvement cut-off of ≥10% [10,11]. Meanwhile, false-negatives are a concern when using the tap test for predicting a positive shunt outcome [10,34,35]. We deliberately used a cut-off of 0% to maximise the sensitivity of the approach. Furthermore, a sensitivity analysis using different cut-offs for TUG improvement suggests a ceiling effect, as the association between smaller temporal scores and positive tap test responder status was not significant anymore at a cut-off of TUG improvement ≥6%. In addition, when separating the sample in clear responders (TUG improvement ≥10%) and clear nonresponders (TUG improvement <0%), our results remained unchanged.

Post-shunt outcome
Our post-shunt evaluations could not be used in this analysis because of the small size of the shunt non-responder sample (n = 4), and because the outcome of the post-shunt evaluation was not standardised (i.e. TUG), but based on the subjective evaluation of the neurosurgeon during the post-shunt clinical evaluation.

Strengths and weaknesses
The main strength of this study is the inclusion of a high number of iNPH patients with a quantitative locomotion measurement before and after the CSF tap test. Another strength relies on the use of a validated measure of locomotion (i.e. TUG) for defining tap test responder status in iNPH patients. However, the main limitation of the study is the absence of neuropathological confirmation for iNPH (and other comorbid neurological conditions) and the absence of standardised quantitative outcomes after shunt surgery.

Conclusion
The iNPH Radscale (total score) does not predict CSF tap test responder status, while a subitem of the iNPH Radscalesmaller temporal hornsis associated with locomotion improvement after CSF tap test. Future studies should verify these findings by including standardised post-shunt outcomes, and investigate the ability of the iNPH Radscale temporal horns width subitem to differentiate atrophy from ventricular enlargement without atrophy.

Data availability statement
Anonymised data that are not published in this article will be made available on request from any qualified investigator, after the approval by the Institutional Review Board of the Geneva University Hospitals.

Funding
This work was supported by the Geneva University Hospitals (grant numbers PRD 11-I-3 and PRD 12-2013-I) and the Swiss National Science Foundation (grant number 320030_173153). The funding bodies had no involvement in the preparation of this article.

Disclosure of conflict of interest
The authors have no conflicts of interest.