Results

Patient demographics

A total of 426 NPH patients were screened, of which 114 had preoperative spinal MRI data available. 100 of these had cervical MRI, 86 lumbar MRI, and 60 of these patients had both, cervical and lumbar MRI. Mean age was 73 ± 10 years, and 53% were males (Table 1). In all patients planned for shunt surgery, previous spinal tap test or lumbar drainage had been proven to be therapeutically effective. Preoperatively, mean NPH-grading score was 6.38 ± 3.25, with significant improvement after shunt surgery (with a median follow-up of 926 days) to 5.75 ± 3.64 (p = .03).

 | Spinal MRI (n = 114) | CS (n = 90) | LS (n = 61)
Gender (m, f) | 60, 54 | 51, 39 | 34, 27
Age (years) | 73 ± 10 | 73 ± 10 | 74 ± 7
Evans index | 0.36 ± 0.07 | 0.36 ± 0.08 | 0.36 ± 0.07
Radscale | 8.04 ± 2.63 | 8.11 ± 2.55 | 8.40 ± 2.64
Callosal angle (°) | 67 ± 16 | 67 ± 17 | 67 ± 16
NPH grading score (pre-OP)* | 6.24 ± 3.44 | 6.24 ± 3.33 | 6.67 ± 3.26
NPH grading score (post-OP)* | 5.73 ± 3.58 | 5.40 ± 3.69 | 6.40 ± 3.2
Stenosis grade | - | Muhle 1:22 | Muhle 2:33 | Muhle 3:35 | Schizas A: 18 | Schizas B: 11 | Schizas C: 22 | Schizas D: 10

Descriptive statistics of spinal imaging data

Cervical stenosis

Out of 100 patients with cervical MRI, 90 patients (90%) had cervical stenosis (CS), with the maximally stenotic segment corresponding in 22 patients to Muhle grade 1 (24%), in 33 patients to Muhle grade 2 (37%), and in 35 patients to Muhle grade 3 (39%). Mean AP of the maximally stenotic cervical segment was 6.4 ± 1.5 mm, and mean DSCSA was 89.1 ± 24.6 mm2. The cervical segments most often affected were C5/6 and C4/5 (for frequencies of segment levels affected see Fig. 3)

Fig. 3: Frequencies of stenoses per segment level affected. On the left for cervical stenoses (CS), on the right for lumbar stenoses (LS).

Lumbar stenosis

Out of 86 patients with lumbar MRI, 61 patients (71%) had spinal lumbar stenoses (LS), with the maximally stenotic segment in 18 patients corresponding to Schizas grade A (29%), in 11 patients to Schizas grade B (18%), in 22 patients to Schizas grade C (36%), and in 10 patients to Schizas grade D (16%). Mean AP of maximal lumbar stenosis was 6.6 ± 2.6 mm, and mean DSCSA was 80.5 ± 38.2 mm2. Segment levels most often affected in the lumbar spine were L3/4 and L4/5 (see Fig. 3).

Plurisegmental stenoses

Plurisegmental stenoses, defined as the involvement of more than one spinal segment, appeared to be more frequent in the cervical spine as compared to the lumbar spine. 70 patients with CS (78%) had more than one cervical stenotic segment. 32 patients with LS (53%) had more than one lumbar stenotic segment. Of those patients with both, cervical and lumbar imaging (n = 60, 53% of the whole cohort), 3% had no stenosis, 25% had only CS, 8% had only LS, while the majority (63%) showed combined cervical and lumbar (dual-level) stenoses. For frequency of occurrence of mono- and plurisegmental CS and LS please see Fig. 4.

Fig. 4: Number of patients with mono- or plurisegmental stenoses regarding the cervical spine (on the left) and the lumbar spine (on the right).

Descriptive statistics of cranial imaging data

For analysis of the cranial hydrocephalus-related imaging parameters, 85 preoperatively acquired cranial CTs were included in the analysis, showing pathological enlargement of the lateral ventricles with a mean Evans-Index of 0.36 ± 0.07, corresponding well to the diagnosis of NPH. The mean Radscale-Score was 8.04 ± 2.63, and the mean callosal angle was 67° ± 0.07, and thus well within the typically expected range (<90°) for NPH patients (For subgroup specifics please see Table 1.).

Subgroup comparisons

Group comparisons of imaging parameters between subgroups stratified by DSCSA and AP diameter, including significances for one- and two-sided testing, Bonferroni-Holm-corrected significances, effect sizes and 95%-confidence intervals (CI) can be found in Table 2.

Comparison | U | N (na vs nb) | p1 (one-sided) | p2 (two-sided) | p (Holm)c | r | 95% CI | Effect sized
Cervical
DSCSA small vs. large (Evans index) | 636 | 39 vs 33 | .468 | .932 | .854 | .012 | [−.221, .243] | very small
DSCSA small vs. large (Radscale) | 573 | 39 vs 33 | .210 | .426 | .630 | .110 | [−.125, .333] | small
DSCSA small vs. large (Callosal angle) | 611 | 38 vs 33 | .427 | .854 | .854 | .026 | [−.209, .257] | very small
AP small vs. large (Evans index) | 644 | 38 vs 34 | .511 | .982 | .511 | .003 | [−.229, .235] | very small
AP small vs. large (Radscale) | 509 | 38 vs 34 | .059 | .122 | .135 | .212 | [−.021, .423] | medium
AP small vs. large (Callosal angle) | 481 | 37 vs 34 | .045* | .088 | .135 | .235 | [.002, .444] | medium
Lumbar
DSCSA small vs. large (Evans index) | 237 | 27 vs 24 | .050 | .101 | .100 | .269 | [−.008, .507] | medium
DSCSA small vs. large (Radscale) | 294 | 27 vs 24 | .287 | .571 | .287 | .093 | [−.188, .359] | very small
DSCSA small vs. large (Callosal angle) | 217 | 27 vs 24 | .022* | .043 | .066 | .330 | [.060, .555] | large
AP small vs. large (Evans index) | 274 | 27 vs 24 | .172 | .345 | .516 | .154 | [−.127, .412] | small
AP small vs. large (Radscale) | 314 | 27 vs 24 | .424 | .850 | .516 | .031 | [−.247, .304] | very small
AP small vs. large (Callosal angle) | 287 | 27 vs 24 | .242 | .485 | .516 | .114 | [−.167, .378] | small

Comparing CS patients with small and large DSCSA

In patients with cervical stenosis, DSCSA-subgroup comparisons did not reveal significant differences with regard to Evans-Index, Radscale Score, or callosal angle (all p > .05, _p_Holm > .630) (please see Table 2).

Comparing CS patients with high and low AP

Comparing CS subgroups based on AP-diameter did not show significant differences with regard to Evans-Index or Radscale Score (all p > .05, _p_Holm > .135). However, subgroups differed regarding callosal angle, indicating higher mean callosal angle in patients with small as compared to large AP (callosal angle: medianAPsmall = 70.7°, medianAPlarge = 64.9°, p = .045), although this difference did not remain significant after correction for multiple comparisons (_p_Holm = .135). Despite the lack of statistical significance after correction, the effect size was medium, r = .235, 95% CI [.002, .444] (please see Table 2).

Comparing LS patients with small and large DSCSA

In LS patients, small/large DSCSA subgroups differed neither in Evans-Index, nor Radscale-Score (all p > .05, _p_Holm > .066). Results revealed however subgroup differences in callosal angle, indicating higher mean callosal angle in the small as compared to the large DSCSA subgroup (callosal angle: medianDSCSAsmall = 71.0°, medianDSCSAlarge = 64.3°, p = .022), although this difference did not remain significant after correction for multiple comparisons (_p_Holm = .066, please see Fig. 5), effect size was large, r = .330, 95% CI [.060, .555] (please see Table 2).

Fig. 5: A) Comparison of the callosal angle between CS patients with small and large anterior-posterior (AP) diameter of the maximally stenotic cervical segment, and B) between LS with small and large dural sac cross sectional area (DSCSA) of the maximally stenotic lumbar segment (for further details please see Table 2).

Comparing LS patients with small and large AP

Group analyses comparing LS subgroups with high/low AP revealed no significant differences in callosal angle, Evans-Index or Radscale Score (all p > .05, _p_Holm = .516).

Partial correlations

Partial correlations, controlling for age, were computed between Evans Index, Radscale-Score, and Callosal Angle, and cervical and lumbar DSCSA as well as AP diameter. A significant negative correlation was observed between lumbar AP diameter and Radscale (p = .041, _r_partial = −.31, 95%CI [−.561, −.014]) (please see Fig. 6). However, this association did not remain statistically significant after Holm correction (_p_Holm = .492). All remaining partial correlations were non-significant both before and after correction (all p > .05, _p_Holm = 1.000). An overview including partial correlations between spinal and imaging parameters, significances, effect sizes and 95% confidence intervals are given in Table 3.

Fig. 6: NPH-patients showed a trend for higher Radscale-Scores to be associated with smaller minimal lumbar AP diameter. The p-value is uncorrected, 95%CI [−.561, −.014]. Bonferroni-Holmes corrected pHolm = .492 (for details please see Table 3).

Imaging parameter | rpartial | 95% CI | p | p (Holm) | Effect sizea
DSCSA, cervical
Evans Index | −.071 | [−.364, .234] | .651 | 1.000 | very small
Radscale | .135 | [−.172, .418] | .387 | 1.000 | small
Callosal angle [°] | −.091 | [−.381, .215] | .561 | 1.000 | very small
AP, cervical
Evans Index | .024 | [−.278, .322] | .879 | .879 | very small
Radscale | .204 | [−.103, .475] | .189 | .567 | medium
Callosal angle [°] | −.130 | [−.414, .177] | .407 | .814 | small
DSCSA, lumbar
Evans Index | .073 | [−.232, .365] | .642 | 1.000 | very small
Radscale | −.139 | [−.422, .168] | .373 | 1.000 | small
Callosal angle [°] | .034 | [−.269, .331] | .826 | 1.000 | very small
AP, lumbar
Evans Index | −.024 | [−.322, .278] | .877 | .877 | very small
Radscale | −.313 | [−.561, −.014] | .041* | .123 | large
Callosal angle [°] | .142 | [−.165, .424] | .362 | .724 | small