Value of MRI-based semi-quantitative structural neuroimaging in predicting the prognosis of patients with idiopathic normal pressure hydrocephalus after shunt surgery

Idiopathic normal-pressure hydrocephalus (iNPH) is a senile syndrome of unknown etiology characterized by gait disorders, cognitive impairment, and urinary incontinence [1, 2]. Its incidence and disability rate increase significantly with age. On brain imaging, the main manifestation is the enlargement of the lateral ventricle [3], with normal cerebrospinal fluid (CSF) pressure. Early shunt surgery can significantly improve the clinical symptoms and patient prognosis [4]. At present, CSF tap test is still the main diagnostic method and for judging prognosis after iNPH surgery. However, it is an invasive test with complications such as low back pain, low intracranial pressure, headache, and low negative predictive value [5]. Patients with a negative CSF tap test may also benefit from surgery. Therefore, the study of non-invasive prognostic indicators for iNPH surgery has clinical significance.

Several scholars have conducted extensive research on the structural imaging features of iNPH. For example, the Japanese guidelines emphasize that the presence of disproportionately enlarged subarachnoid space hydrocephalus (DESH) in neuroimaging is an important indicator for the diagnosis of iNPH [6] and that the DESH sign reportedly plays a positive role in predicting the prognosis of shunt surgery [7]. Based on this, Shinoda et al developed the DESH score based on iNPH’s MRI features and explored its value in prognostic predictions for iNPH [8]. In their study, the DESH score included five items: Evans’ index (EI), Sylvian fissures, tight high convexity, callosal angle (CA), and focal sulcal dilatation. In addition, Kockum et al developed the iNPH Radscale to explore the relationship between imaging features and clinical symptoms of iNPH. Based on CT image features, it included seven items: the EI, Sylvian fissures, tight high convexity, CA, focal sulcal dilatation, temporal horns, and periventricular hyperintensities (PVH) [9]. However, the accuracy and practicability of traditional structural imaging in the diagnosis and prognosis of iNPH are still debated [10‐13].

We hypothesized that there were differences in structural imaging between iNPH patients who improved and those who did not improve after shunt surgery. By comparing the difference of preoperative DESH score and iNPH Radscale score between the improvement and non-improvement groups, the value of structural imaging features in the prognostic evaluation of iNPH patients undergoing shunt surgery was clarified.

The initial screening involved a clinical sample of 111 consecutive patients suspected of iNPH who underwent brain MRI examination at our hospital between January 2018 and December 2020. Figure 1 shows the flowchart for this study, from initial screening to the final analysis. This study initially included 111 patients who visited the hospital due to ≥ 1 of progressive gait disorder, cognitive impairment, and urinary incontinence, and who were evaluated by neurologists and underwent relevant head imaging examinations. Among them, nine patients were diagnosed with obstructive hydrocephalus after a detailed MRI examination. iNPH-related symptoms were confirmed in 111, and the imaging manifestations were signs of ventricular dilatation (EI ≥ 0.3). These patients were suspected of iNPH and admitted to the neurosurgery department for further examination and treatment; 12 refused admission due to family reasons. The hospitalized patients underwent more detailed examinations, including CSF biochemical and stress examinations and behavioral and cognitive examinations. Eight patients were diagnosed with Parkinson’s disease (PD) and were excluded, 25 patients were excluded because of failure to fulfill the CSF inclusion criteria, and 2 patients refused surgery; therefore, 55 patients underwent shunt surgery. During the 12 months of follow-up, one patient died after an accidental fall and seven were lost to follow-up because they moved to other hospitals in their hometown. Finally, the remaining 47 cases were evaluated. The protocol was approved by our hospital’s bioethics committee (approval no. KS20190114001).

For patient selection, we used the following criteria based on international standards [2] and actual conditions: (1) age ≥ 60 years; (2) ≥ 1 clinical manifestations of the triad of gait disorder, cognitive impairment, and urinary incontinence; (3) manifestations of ventricular enlargement (EI ≥ 0.3) on imaging, with the exclusion of other diseases that may cause it, such as traumatic brain injury, various types of cerebral hemorrhage, brain tumors, encephalitis, meningitis, and large-area cerebral infarction, among others; (4) CSF pressure ≤ 200 mmH₂O on lumbar puncture, with normal CSF biochemical examination results; (5) positive response on the CSF tap test [14]; and (6) no severe cardiovascular and cerebrovascular diseases and other diseases that are contraindications for surgery. In addition, all patients underwent head MRI examination within 1 month before surgery, and a ventricular-abdominal shunt was selected for surgery.

The demographic characteristics and preoperative imaging parameters of the 47 patients included in the study are shown in Table 2. Of the participants, 49% were men. The age at the time of the shunt surgery was 69.2 ± 5.9 years. The median (interquartile range, IQR) duration of the symptoms (having at least two or more of the triad) before imaging was 12.0 (6.0–36.0) months. Based on the iNPHGS, the typical preoperative symptoms were distributed as follows: 47 patients had gait disturbance, 45 had cognitive impairment, 37 had urinary symptoms, and 37 had the classic triad.

Table 3 shows the interrater reliability. For continuous variables, the reliability ranged between 0.95 and 0.98 (ICC), and for variables on an ordinal scale, between 0.56 and 0.73 (κ). Except for the focal sulcal dilatation, all evaluations showed substantial consistency. For all patients, the preoperative EI was > 0.3. The median DESH score and iNPH Radscale were 6.0 (4.0–6.0) and 8.0 (6.5–10.0), respectively. Severe PVHs (white matter hyperintensities extending from the paraventricular to the deep white matter) were observed in 48% of the patients. The callosal angle was < 90° in 64% of patients. A temporal horns’ diameter ≥ 6 mm was observed in 80% of the patients.

An mRS improvement ≥ 1 within the first year after shunt placement was the primary index. iNPH patients were divided into an improvement group and a non-improvement group; 28 patients (59.6%) showed improvement, and the median (IQR) mRS improved from 2.0 (2.0–3.0) to 2.0 (1.0–3.0) (p = 0.039). The preoperative imaging parameters and clinical results showed that the median DESH scores of the improvement (Md = 6.0, IQR 4.5–6.0) and non-improvement (Md = 4.5, IQR 3.25–6.0) groups were not significantly different (p = 0.230; Table 4). Similarly, the improvement (Md = 8.0, IQR 7.0–9.5) and non-improvement (Md = 8.0, IQR 6.25–10.0) groups showed no significant differences in iNPH Radscale scores (p = 0.657), EI (p = 0.397), and CA (p = 0.43; Fig. 3).

The secondary outcome was an iNPHGS improvement ≥ 1 point. We found that 29 (61.7%) patients had improvement in the iNPHGS; the total iNPHGS improved from 6.0 (4.0–7.0) to 4.0 (2.5–7.0) (p = 0.011). Specifically, 24 patients showed improvement in gait, 20 showed improvement in cognitive ability, and 14 showed improvement in urination. Similarly, we divided the iNPH patients into improvement and non-improvement groups based on the iNPHGS scores (iNPHGS total, iNPHGS gait, iNPHGS cognitive, iNPHGS urinary) before and after shunt surgery, and the preoperative imaging parameters and clinical outcomes were compared. Similarly, there were no significant differences between patients with and without improvements in preoperative imaging parameters and clinical outcomes (Table 5).

Using different outcome indicators, the ROC analysis of the two imaging scores showed no significant difference between groups (p > 0.05). Under the primary result, the AUCs of the two scores were 0.65 (95% CI: 0.49−0.81, p = 0.079) and 0.59 (95% CI: 0.43−0.76, p = 0.28), respectively (Fig. 4a). Similarly, under the secondary result, the AUCs of the two scores were 0.66 (95% CI: 0.50−0.82, p = 0.06) and 0.62 (95% CI: 0.46−0.79, p = 0.15), respectively (Fig. 4b). In conclusion, the diagnostic performance for the treatment response of the two scores was poor and not significant, further corroborating our previous results.

Binary logistic regression analysis was performed on all imaging indexes included in the two scores with the primary result as classification standard. The predicted values are expressed after adjustment for sex and age (Fig. 5). No MRI marker was significantly associated with postoperative improvement.

Furthermore, we analyzed the relationship between the preoperative imaging parameters and the severity of clinical symptoms in patients with iNPH (Supplementary Materials). The total DESH and iNPH Radscale scores were not associated with clinical symptoms. However, there were significant, but not strong, correlations between EI (p = 0.017) and the temporal horns diameter (p = 0.016) and clinical symptoms of gait disorders.

In this study of 47 patients with iNPH, we used quantitative imaging features (DESH score, iNPH Radscale, EI, and CA) to predict the prognosis of shunt surgery. We found that neither the DESH score nor iNPH Radscale was predictive of the postoperative outcome; high EI value, low corpus callosum angle, or high DESH score and iNPH Radscale score were not associated with a favorable outcome.

Obviously, both the DESH score and iNPH Radscale were based on the DESH sign, which is characterized by tight high-convexity and medial subarachnoid spaces and enlarged Sylvian fissures with ventriculomegaly [6]. For both scoring criteria, higher scores were associated with more unfavorable imaging findings. The iNPH Radscale has two more sub-items than the DESH score, which are the temporal angle and PVH, and the same item has different scoring standards. This is related to the differences in inclusion criteria and research purposes.

In a study by Shinoda et al, there was a significant difference between the DESH score based on MRI in patients with and without postoperative improvement, and a strong correlation between the DESH score and the degree of postoperative improvement [8]. This contradicts the results of the present study. The explanation for this difference is as follows: the DESH score is the result of quantifying some imaging features of iNPH, but some parameters are based on the subjective impressions of the scorer, such as the Sylvian fissure enlargement, the high-convex tightness, and the local expansion of the sulcus [12]. This may lead to differences in scores by raters, accounting for the difference in the research results. Furthermore, recent studies have questioned the prognostic prediction for surgery based on structural imaging [10, 11], mostly due to the uncertainty of the measurement standard for imaging markers. Ryska et al reported that the measurements of EI, CA, and other imaging markers will deviate with different angles or scanning sequence parameters, and this deviation often occurs in different centers. Therefore, a more standard and unified definition of the imaging plane or development of a computer-assisted system is warranted in the future [24]. In addition, a more important reason may be the sample bias caused by the different etiology of NPH in different regions of the world.

Kockum et al proposed the iNPH Radscale and confirmed its association with the severity of clinical symptoms in patients with iNPH [9]. However, our results showed no significant correlation between the total score of the iNPH Radscale and the clinical symptoms of iNPH patients; only gait disorder was associated with EI and temporal horns diameter, but the correlation was low. This is consistent with the report by Agerskov et al [11]. A possible explanation is that the changes in nerve function and structure are not synchronous in iNPH patients; Kockum et al only screened suspected iNPH patients using a questionnaire survey and the evaluation of clinical symptoms, but this was not confirmed by surgery. There may be other patients with similar symptoms in the sample, such as those with Alzheimer’s disease and PD. Our cases were confirmed after shunting. Moreover, the results of a recent study showed similar total scores of the iNPH Radscale for CSF tap test responders and non-responders. Therefore, iNPH Radscale does not predict clinical improvement after the CSF tap test [25]. This is consistent with the present results.

In the study of Hong et al, the DESH sign was considered a relevant factor for good results, while white matter hyperintensities, CA, and EI were not [26]. However, the iNPH sub-population with typical imaging features is only one subgroup. The rate of detection of DESH is only approximately 30–50%, and its ability to diagnose treatment response is poor [12, 16, 27]. This can lead to selection bias and exclude iNPH patients with atypical imaging findings from the operation schedule. In this case, it is unfair to discuss the prognosis of the surgery.

Recent studies have shown that the improvement of symptoms after shunt surgery is related to mild symptoms and short disease course [28], while delayed shunt surgery is related to poor improvement of symptoms [29]. Meanwhile, Wu et al compared systematic volumetric analysis with traditional structural imaging and pointed out that systematic volumetric analysis has a high classification performance for predicting the results of shunt surgery [30]. Furthermore, traditional structural imaging has been reported to have poor diagnostic performance for treatment response [10, 11, 28, 31], which is consistent with our results.

This study has some limitations. It was conducted at a single center using a single MRI scanner. However, this prevented any discrepancies related to the use of different scanners. Due to the strict differential diagnosis and robust inclusion criteria, the cohort was relatively small. Further multicenter large-sample studies are needed to verify the neuroimaging features predictive of the outcome of iNPH surgery. Furthermore, after 1-year follow-up, many patients could not return to the hospital due to the long distance, physical condition, or COVID pandemic. Thus, telephone interviews were conducted to ask patients and their families about the patient’s status to try and objectively evaluate the clinical symptoms.

In our study, there were no significant differences in DESH and iNPH scale scores between iNPH patients with and without improved clinical symptoms a year after shunt surgery. Therefore, some patients with lower DESH and iNPH Radscale scores may also improve after surgery. In summary, structural imaging appears to have limited value and should, therefore, not be used to exclude patients from shunt surgery at this time.

Relationship between preoperative imaging parameters and clinical outcomes of iNPH patients