Gait apraxia evaluation in normal pressure hydrocephalus using inertial sensors. Clinical correlates, ventriculoperitoneal shunt outcomes, and tap-test predictive capacity

Idiopathic normal pressure hydrocephalus (iNPH) is a syndrome characterised by normal cerebrospinal fluid (CSF) pressure associated with chronic ventricular dilation. Progressive gait impairments, urinary incontinence, and cognitive deficits define the clinical triad that highlights its onset [1]. Usually, gait alterations represent an early manifestation of the syndrome, accompanied by motor disabilities and increased fall risk, whereas cognitive impairment and urinary incontinence might appear in a later stage [2]. For this reason, the correct definition of gait alterations and their clinical and quantitative description, together with a differential diagnosis from similar motor conditions (iNPH mimics), represent critical challenges to address the syndrome in its early phase [3]. Motor alterations often appear as a complex syndrome, characterised by large base of support, reduced gait speed, stride length, foot clearance, left-to-right coordination, and difficulty in postural transitions, not depending on motor, sensory or cerebellar deficits [4]. For the first time in 1977 Miller-Fisher thoroughly described iNPH gait alterations using the term "gait apraxia" [5]. Considering that the classical definition of apraxia of gait is “loss of ability to properly use lower limbs in the act of walking which cannot be accounted for by demonstrable sensory impairment or motor weakness” [6], also today this term represents a broadly accepted definition of iNPH gait alterations in literature.

Once iNPH is adequately diagnosed, treating it with the insertion of a ventriculoperitoneal shunt (VPS) allows draining the CSF, which in turn can lead up to a partial or complete recovery [7]. On the other hand, delayed intervention can cause disease progression to a point where treatment may be no longer effective. Patient selection for VPS treatment is crucial to date, but clinical predictors for a successful outcome are poorly investigated and known [8].

As reported by a recent review and meta-analysis by Scully et al. [9], there is still a strong need for an accurate tool to predict shunt outcomes to facilitate standardization of care and optimal management of patients. The CSF-Tap Test (CSF-TT) is an assessment test used to predict iNPH patients with positive shunt-response [10]. It consists in removing 30–50 ml of CSF via lumbar puncture and clinically assessing the patient before and after the CSF removal to verify symptoms modifications [10]. The simplicity of this procedure allows it to be performed in an outpatient environment and an improvement in gait quality is considered to be predictive of a positive response to VPS surgery. However, evidence supporting its validity is still inadequate [11]. Moreover, the degree of change in motor symptoms to be considered a positive response has not yet been established [8, 12]. In a recent review, Mihalj et al. quantified the accuracy of CSF-TT in screening patients for shunting as 62%, but with a negative predictive value as low as 37% [13].

Visual observation is still the main procedure to assess motor performance. However, clinical tests and scales such as the timed up and go test (TUG), the Tinetti Performance Oriented Mobility Assessment (POMA), and the self-selected gait speed are considered beneficial ancillary evaluations when associated with CSF-TT [11]. In particular, Scully et al. [9] found a moderate prognostic value of TUG and 18-m walking test (18mW) gait speed in predicting shunt outcomes among iNPH patients. Still, these tests suffer particularly in terms of specificity, estimated at about 0.63 and 0.67 for TUG and 18mW respectively. To avoid underestimating VPS responders, Scully et al. also highlighted the importance of not using these tests in isolation [9].

In clinical practice, POMA is one of the most used scales and is considered a helpful tool when associated with a CSF-TT [14]. However, as reported in a systematic review by Gor-Garcia-Fogeda et al. [14], this type of assessment fails to capture objectively the possible diversities of motor manifestations and can result in subjective, inconsistent interpretations of the motor response. Convenient and accessible, yet highly accurate objective tools are therefore needed for patients’ assessment and management.

Instrumented gait analysis represents a promising paradigm for addressing this need by objectively assessing gait and balance improvements and unveiling details that are not easily accessible by simply using clinical observations or clinical scales [15‐18]. The inertial measurement units (IMUs) are increasingly employed as motion sensors in instrumenting motor tests in the clinical routine [19]. Nowadays, IMUs are wireless, lightweight, cost-effective, and operate on commercially available mobile devices, thus allowing the plug-and-play execution of gait analysis tests without the need for specialized personnel or dedicated laboratories [20, 21].

Only a few studies employed an instrumented gait analysis for studying iNPH gait apraxia and its improvement after CSF-TT [17, 21‐27]. Stolze et al. reported gait speed and stride length as the most responsive parameters in a pre vs. 24 h post-CSF-TT comparison [22]. A few years later, Williams et al. confirmed an improvement in gait velocity and described a reduction of double support time and cadence [26]. Recently, Allali et al. found a significant improvement of gait speed in single and dual tasking during the 10 m walk test in a pre vs. post-CSF-TT comparison [23]. Colella et al. [24] confirmed the stride length to be one of the most reliable parameters in evaluating iNPH’s gait performance. In a pre vs. 24 h post-CSF-TT study, Lim et al. reported significant improvements of gait velocity, stride length, and step width and decreases in stride time and stride length variability [27]. Other studies focused on TUG parameters. Bovonsunthonchai et al. compared performances on TUG pre vs. post-CSF-TT, revealing a significant improvement in the sit-to-stand transition, walking time, and the number of steps employed to turn [25]. Ferrari et al. [17] reported a reduction of double support duration and an increase of cadence from baseline to 72 h post-CSF-TT in a 18mW and TUG, and an increase of stride length in the 18mW.

On iNPH patients instrumented with inertial sensors, He et al. found a decreased cadence, reduced gait speed, a higher duration of double support, decreased elevation at mid-swing, reduced foot strike angle, shorter stride length, difficulty in turning, and impaired balance functions [21]. They also showed that most previously listed gait manifestations were significantly improved after external lumbar drainage in responders to external lumbar drainage.

Eighty-six patients included in the Bologna PRO-Hydro study between May 2015 and November 2019 were diagnosed with iNPH and underwent the VPS. Among these, 42 patients who were able to move independently and had completed the instrumental TUG acquisition at preTT, postTT, and postVPS, were considered for the analyses of this study (Fig. 2). Mean (± standard deviation, SD) age was 75.2 ± 4.0 years. Most patients were women (n = 27, 64%), with a body mass index of 26.9 ± 4.0 kg/m². At baseline, median iNPH-GS and Tinetti POMA were respectively 6 (interquartile range, IQR 4.75–7) and 20 (IQR 18–23). The median Rankin Scale was 2 (IQR 1–3) (Table 2). The mean number of days between preTT and postTT was 4.4, between preTT and VPS was 121, and bewteeen VPS and postVPS was 166.8.

Most instrumental gait parameters were significantly correlated with clinical scores (Fig. 3, [see Additional file 1: Table S1]). Correlation coefficients were in the expected direction of the association. Only FIM, INPH-GS and nSteps2turn did not show high correlation coefficients with any instrumental variables, with values ranging from 0.33 to 0.58. Tinetti Balance, Tinetti Gait, Tinetti total, GSS and Rankin scale obtained statistically significant coefficients in the range 0.32–0.77, with at least one instrumental parameter showing a coefficient with an absolute value above 0.6. Five TUG variables (testDuration, walkTime, sitTime, doubleSupport, gaitSpeed) and three 18mW variables (testDuration, doubleSupport, gaitSpeed) correlated with all clinical parameters.

Most instrumental variables showed excellent reliability (ICC above 0.90, [see Additional file 1: Table S2]). Besides, MDC testDuration, totalSteps, strideLength, doubleSupport, gaitSpeed and minTC were smaller than the mean group-level change after VPS. All clinical evaluations except FIM significantly improved between preTT and postVPS. In particular, Tinetti POMA score improved of 4.5 points (95% confidence interval (C.I.) [3.5, 5.5], median preTT: 20, postVPS: 24), while GSS improved of 2.5 points (95% C.I. [− 3.5, − 1.5], median preTT: 5, postVPS: 2). All instrumental variables significantly changed between preTT and postVPS, except trunkInclination and pitchAtTC2 measured in TUG and 18mWT. TUG testDuration decrease was − 7.66 s (95% C.I. [− 9.47, − 6.09] s, mean preTT: 24.07 s, postVPS: 16.41 s), while 18mW testDuration decrease was -8.1 s (95% C.I. [− 10.62, − 5.73] s, mean preTT: 33.85 s, postVPS: 25.76 s; [see Additional file 1: Table S2 ]). Figure 4 reports the relative changes of instrumental variables after VPS, measured as the difference between preTT and postVPS divided by the inter-subject standard deviation (Cohen’s effect size). Such effect sizes range, in module, between 0.21 and 1.39.

Table 3 and Fig. 5 show the results on the added prognostic value of the CSF-TT for predicting the clinical and instrumental outcomes postVPS. Among clinical variables, the Tinetti POMA Total score change after the CSF-TT had significant prognostic capacity for its postVPS value (p = 0.013). The Tinetti POMA score postVPS was predicted from its preTT value and its change after CSF-TT with an adjusted R² of 0.41. All other clinical variables after CSF-TT did not show prognostic capacity for their values postVPS, with adjusted R² of the prediction models ranging from 0.14 to 0.30.

Among TUG instrumental variables, CSF-TT prognostic capacity was significant for testDuration, walkTime, standTime, sitTime, cadence, strideLength, gaitSpeed, trunkInclination, maxTC2, minTC, pitchAtTC2, pci, psdF, and psdW. The highest predictive accuracy was obtained for strideLength (adjusted R² = 0.53). The strongest evidence for the added prognostic value of CSF-TT was obtained for walkTime (χ² = 18.15, p = 2∙10^–5).

Among 18mW variables, CSF-TT prognostic capacity was significant for cadence, trunkInclination, pci, psdF, and psdW. The highest predictive accuracy and evidence for the added prognostic value of CSF-TT were obtained for trunkInclination (adjusted R² = 0.70, χ² = 16.055, p = 10^–4).

Idiopathic normal pressure hydrocephalus is a complex neurological syndrome with gait apraxia manifestations. To the best of our knowledge, this is the first observational cohort study able to combine objective instrumental measures from wearable motion sensors with the major clinical iNPH scales. We aimed to test the hypothesis that instrumental measures provide valid tool for iNPH management, including the evaluation of gait apraxia symptoms, their improvements after VPS, and prediction of VPS outcomes by means of CSF-TT.

We employed clinical and instrumental measurements for gait assessment on 42 patients at baseline (preTT), 72 h after the CSF-TT (postTT), and six months after VPS (postVPS).

At baseline, we found significant correlations among some clinical and instrumental variables, that indicate concurrent validity. Tinetti Balance, Tinetti Gait, Tinetti total, GSS, and Rankin Scale showed strong correlation with at least one instrumental variable. In particular, TUG variables (except StandTime) strongly correlated with Tinetti Gait and Total scores. Variables related to gait coordination (pci), variability (strideSD), and frequency (psdF) showed higher correlations with clinical variables when computed on 18mW than TUG, possibly due to the shorter path in TUG.

Both in TUG and 18mW cadence showed weak (0.32) or non-significant correlations with clinical variables. Thus, cadence is an abnormal feature of iNPH gait that improves post CSF-TT [17, 21, 26] but is not captured by clinical variables. As expected, most instrumental variables (in particular gait speed, both in TUG and 18mW) were correlated with the number of steps to make a turn (nSteps2turn). An increased nSteps2Turn is usually associated with a reduced ability to dissociate the motor program among head, trunk, and lower limbs (“enbloc” movement). This clinical feature typically represents a pathological sign of turning instability [40].

FIM and INPH-GS were the two clinical variables showing the weakest correlations with instrumental variables, probably due to their multidimensional nature with different non-motor domains.

Both instrumental and clinical variables (except trunkInclination, pitchAtTC2, and FIM) showed an improvement six months after VPS, suggesting a reduction of motor impairment. In particular, the decreased double support (doubleSupport), the increased foot clearance (maxTC1 and maxTC2), and the improved gait bilateral coordination (pci), are consistent with the idea of a reduction of shuffling gait. Interestingly, some TUG variables (standTime, sitTime) can instead reflect a favorable modification in apraxia of postural transitions [4]. The MDC for most instrumental variables were smaller than the mean group-level change after VPS. This demonstrates how instrumental variables can possibly be used in clinical practice to sensibly measure VPS improvements (see Additional file 1: Table S2).

Among clinical variables, FIM did not show a statistically significant improvement after VPS. Tinetti POMA significantly improved, with a median value of 4.5 points [95% C.I. 3.5–5.5]. This value is slightly inferior to the MDC at the individual level that is of at least 5 points [41]. GSS, Rankin scale, iNPH-GS and the number of steps to turn also improved six months after VPS.

This study also revealed a predictive capacity of the CSF-TT for VPS outcomes. Using motion sensors for CSF-TT evaluation enables to infer VPS outcomes on many different aspects of gait apraxia [15]. The modifications of many instrumental variables after the CSF-TT proved to be informative for predicting gait apraxia improvements six months after surgery. Trunk flexion reduction after CSF-TT demonstrated good predictive accuracy for VPS outcome although its improvement after CSF-TT was not statistically significant. On the other hand, among clinical variables, only the Tinetti POMA demonstrated prognostic capacity to infer VPS response.

Our study has some limitations. First, the exclusion from the studied population of 44 patients characterized by very severe gait disturbances, abasia or bed-ridden which could not complete motor tests (Fig. 1) limit the external validity of the study. Secondly, for feasibility reasons the Bologna PRO-Hydro study is affected by a variability in the time that elapses between the inpatient program and the VPS that goes from one week to two months [16]. Finally, despite its good predictive accuracy for VPS outcome, trunk flexion reduction after CSF-TT did not demonstrate a statistically significant improvement. Further research must be carried out to confirm this result.

The results of this study support the hypothesis that the instrumental analysis of multiple spatiotemporal parameters of gait, derived from wearable inertial sensors applied to TUG and 18mW, can capture important clinical correlates and describe motor apraxia modification of iNPH patients six months after VPS. Interestingly, instrumental variables resulted sensible enough for measuring improvements of motor apraxia after VPS on individual patients thus can possibly be introduced in clinical practice.

Besides an overall correlation between instrumental variables and motor clinical scales, instrumented TUG and 18mW showed to be able to broaden the spectrum of analysis of iNPH gait by allowing a quantitative measure of those features contributing to overall motor performance. In particular, instead of considering only the total time and/or the total number of steps (possibly affected directly by cognitive impairments), the application of inertial sensors to the TUG and 18mW could expand the range of clinical assessment to parameters potentially less influenced by individual variability (for instance, TUG total time can be). Among clinical scales measured before and after CSF-TT, only the Tinetti POMA demonstrated a predictive capacity for VPS outcomes, whereas instrumented TUG and 18mW showed a predictive capacity for multiple spatiotemporal parameters. The findings of this study suggest that an evaluation protocol implementing the analysis of gait instrumental parameters together with clinical examinations and scales (TUG, 18mWT, Tinetti POMA, Rankin, iNPH-GS, GSS), represents a powerful tool in the management of iNPH patients affected by gait apraxia.