Quantification of pathological gait parameter thresholds of idiopathic normal pressure hydrocephalus patients in clinical gait analysis
Descriptive statistics
NPH patient characteristics are depicted in Table 1. The anthropometric measures between groups were comparable without statistically significant differences between the groups. 37 out of 55 (67.3%) patients suffered from all three symptoms of the Hakim’s triad. During gait assessment, “stops walking while talking” episodes were observed in 28 patients (50.9%), two patients (3.6%) showed a freezing of gait. No such phenomena were present in healthy subjects. NPH patients showed a reduced step length and walking speed with elongated double support phase, broadened base of support, and an increased variability of stride length (see Supplementary Table 1). In the course of disease, 51 patients (92.7%) received a CSF drainage procedure. Thereafter, gait improved in 43 (78.2%) patients, cognitive improvement was observed in 37 (67.3%) patients. 31 patients (56.4%) underwent VP shunt surgery.
Gait parameter thresholds during preferred walking speed
The statistical approach is visualized in Fig. 1. AUC values of the classification models are presented in Table 2. Gait parameter thresholds, sensitivity as well as specificity values are illustrated for all gait conditions in Fig. 2.
Overview of gait parameter thresholds, sensitivity and specificity values of patients with idiopathic normal pressure hydrocephalus (NPH) compared to age-matched healthy subjects. Gait parameters are arranged in five distinct global gait domains (x-axis). Different gait conditions are summarized in speed, sensory, and dual task conditions (y-axis). Each tile consists of the gait parameter threshold in its unit on the top, the small numbers on the bottom show the corresponding sensitivity (left) and specificity (right) in %. Tiles are color-coded according to the area under the curve (AUC) of the receiver operating characteristic (ROC) curve (see legend). Blue tiles illustrate that NPH patients fall below the threshold value, red tiles illustrate that NPH patients fall above the threshold value. Grey tiles indicate a non-significant regression fit. Abbreviations: Gait parameters: SLen: stride length; STime: stride time; Swing: percentage of swing phase; Dsupp: percentage of double support phase; CV: coefficient of variation; ASYM: asymmetry; SWidth: stride width. Gait conditions: PS: preferred walking speed; SS: slow walking speed; MS: maximal walking speed; HR: head reclination; EC: eyes closed; DTC: walking and serial 7 dual task; DTS: walking and verbal fluency dual task; DTM: walking and carrying a tray dual task.
During preferred walking speed, the most distinct gait thresholds were variables in the Pace domain, namely stride length ≤ 1.02 m (sensitivity 0.93, specificity 0.91, AUC 0.96) and gait velocity ≤ 0.83 m/s (0.80, 0,91, AUC 0.93). Other significant variables include Cycle variables such as swing phase ≤ 36.5% (0.93, 0.78, AUC 0.91) and double support phase ≥ 27.0% (0.96, 0.76, AUC 0.91) as well as the Support variable stride width ≥ 0.13 m (0.80, 0.91, AUC 0.89). The most distinct variability parameter was stride length CV ≥ 3.4% (0.93, 0.72, AUC 0.90). Asymmetry parameters showed non-significant to only moderate discriminative power.
Gait parameter thresholds during other gait conditions
The most noticeable thresholds were observed during DTS: Stride length ≤ 0.97 m (0.97, 0.93, AUC 0.98), swing phase ≤ 35.1% (0.91, 0.90, AUC 0.96), double support phase ≥ 30.2% (0.88, 0.90, AUC 0.95), and stride width ≥ 0.13 m (0.91, 0.86, AUC 0.89). During EC, the gait velocity threshold was ≤ 0.69 m/s (0.90, 0.91, AUC 0.96). The mean AUC values of all gait conditions for each gait parameter ranged from 0.52 (stride time asymmetry) to 0.94 (stride length). Further high overall AUC values were found for velocity (0.92) and for double support phase (0.91; see Table 2). Overall, the most distinct gait parameters thresholds were observed for the Pace, Cycle, and Support gait domains.
AUC values and the influence of gait examination conditions
The mean AUC values of all examined gait parameters for each examination condition ranged from 0.76 (SS) to 0.84 (DTM). AUC values for PS and HR were 0.82. Highest single AUC values were found in DTS (stride length AUC: 0.98, swing phase AUC: 0.96, double support phase AUC: 0.95), DTM (stride length AUC: 0.96, velocity AUC: 0.95), EC (velocity AUC: 0.96, double support phase AUC: 0.95), and PS (stride length AUC: 0.96, velocity AUC: 0.93).
The most distinct dual task condition throughout all gait parameters was DTM with following noticeable threshold parameters: gait velocity ≤ 0.87 m/s (0.91, 0.87, AUC 0.95), stride length ≤ 1.03 m (0.94, 0.89, AUC 0.96), swing phase ≤ 35.5% (0.83, 0.98, AUC 0.91), double support phase ≥ 28.0% (0.89, 0.87, AUC 0.93), stride length CV ≥ 3.7% (0.91, 0.78, AUC 0.91), and stride width ≥ 0.12 m (0.91, 0.83, AUC 0.91).