Background

Idiopathic normal pressure hydrocephalus (iNPH) is characterized by impairment of balance, gait, and cognition as well as urinary incontinence [1, 2]. The cerebral ventricles are enlarged and deformed [3]. INPH is more common than previously thought with a prevalence of 3.7% in the general population above 65 years [4]. The only effective treatment is shunt surgery to divert cerebrospinal fluid (CSF), with 70–80% of patients improving afterwards [5]. The CSF composition differ between patients with iNPH, Alzheimer’s Disease (AD) and healthy controls (HC). The CSF levels of amyloid β 1–42 (Aβ42) have been reported to be lower in iNPH than in HC. Accordingly, iNPH patients had decreased or similar levels of tau and phosphorylated tau (p-tau) compared with HC [6,7,8,26,27,28,29,30]. Females with AD or mild cognitive impairment have a higher atrophy rate [31] and suffer a faster disease progression rate [32]. In a cognitively healthy population, no relationship was seen between gender and tau/p-tau while a complex relationship between age, gender and APO-E carriership was seen for Aβ42 [33]. In iNPH, less is known about potential sex differences and although a few iNPH studies include sex-matched controls, subgroup analysis and comparisons are rare. This leaves the question of sex differences in the pathophysiology of iNPH largely unanswered, a question of special interest when investigating Aβ42, tau and p-tau due to the sex differences in AD and possible overlap between the diseases.

The volume of intracranial CSF can be measured by magnetic resonance imaging (MRI), with different methods available from manual to automatic segmentation of the images [34,35,36]. A reduction in VV have been reported after shunt surgery and clinical symptoms have been seen to correlate with the reduction in volume, implicating the volume change in the pathophysiology of the disease [37, 38]. Given the hypothesis of dilution of the biomarkers in a larger amount of CSF [19, 20], a negative correlation is expected between VV and the concentration Aβ42, tau and p-tau. The objective of this study was to investigate whether such a correlation exists in an attempt to shed further light on the pathophysiology of iNPH.

Method

Participants

One hundred and thirty-six patients (47 females and 89 males with a median age of 76 and 75 years, respectively) who underwent surgery due to possible or probable iNPH [39] were consecutively recruited from Sahlgrenska University Hospital (n = 108) during 2013–2016 and Östersund Regional Hospital (n = 28) during 2017–2020. The mean time between imaging and CSF collection was 38.3 days (SD = 72.8). Two patients were excluded because of > 1 year between imaging and lumbar puncture.

A positive shunt response was defined as at least one of the following (a – c) [40, 41]:

a. A 20% reduction in the time or the number of steps in at least one of the two motor function tests (Timed Up and Go, and 10 m walk).

b. Four points increase in the Mini-Mental State Examination.

c. One level increase in the continence scale and two points increase in the Mini-Mental State Examination.

Ventricular volumetry

MRI images were acquired before operation with the resolution of 1mm3/voxel. The MRI used at Sahlgrenska University hospital was either a 1.5 T Intera (Philips Medical Systems, Best, The Netherlands) or a 1.5 T Achieva dStream (Philips Medical Systems, Best, The Netherlands) whereas the MRI used at Östersund Regional Hospital was either a 1.5 T GE Optima MR450w (GE Healthcare, Wausheka, WI, USA) or a 1.5 T GE Signa Voyager (GE Healthcare, Wausheka, WI, USA). The images were exported using DICOM format. The lateral and third ventricle volumes were semi-automatically segmented using the ITK-SNAP software [34], a method used in a recent study of volumetric change after shunt surgery [42] and exemplified in fig. 1.

Analysis of CSF

Analyses of Aβ42, tau and p-tau were performed as part of clinical routine workup and the results were retrieved from the participants’ patient charts. Like most Swedish hospitals, both the hospitals in Gothenburg and Östersund purchase this service from the clinical Neurochemistry Laboratory at Sahlgrenska University Hospital, Mölndal, Sweden. CSF was collected in polypropylene tubes during the diagnostic tap test after discarding the initial 3-5 mL to avoid contamination of the sample by small initial bleeding, then centrifuged at 2000 g for 10 min before being frozen at − 70 °C awaiting transportation. Biochemical analysis was performed by the standard clinical procedures at the laboratory using commercially available kits for quantitative solid-phase enzyme immunoassay: INNOTEST hTAU Ag, INNOTEST PHOSPHO-TAU(181P), INNOTEST β-AMYLOID (Fujirebio Europe N.V, Gent, Belgium).

Statistical analysis

Spearman's rank order correlation was used to analyze associations due to non-normal distribution of the data and the presence of outliers. Confidence intervals for spearman’s rho were acquired by bootstrap** with 3000 samples using the bias correction accelerated method [43]. Between-group comparisons were made using the Mann–Whitney U Test. One participant had a tau value of < 75 pg/mL which was included in the statistical analysis as the value 75 pg/mL. Because of the non-parametric tests used and it being the single lowest value in the data set this should not have impacted the reliability of results. The level of statistical significance was set to p < 0.05. Statistical analysis was performed using R version 4.0.3 (R Foundation for Statistical Computing, Vienna, Austria). Power calculation was made using G*Power 3.1.9.4. [44]. The sample size was large enough to have discovered correlations of effect size 0.3 or larger with 95.4% power, per Cohen’s convention a medium effect [45].

Results

For the total sample (n = 136) VV had a negative correlation with Aβ42 (rs = − 0.17, p = 0.042) which is visualized in Fig. 2, but not with tau or p-tau. However, the correlation between Aβ42 and VV was slightly stronger in the male subgroup (rs = − 0.22, p = 0.039), visualized in Fig. 3. For males there was also a positive correlation between Aβ42 and tau (rs = 0.302 p = 0.004) as well as p-tau (rs = 0.264 p = 0.001). In females, no significant correlation between VV and the CSF biomarkers was found (Table 1). The relationship between Aβ42 and VV for females is shown in Fig. 4 and the interrelationship of the biomarkers for the subgroups in Fig. 5 and 6.

Fig. 1
figure 1

Segmentation example. Example image of the segmentation process using ITK-SNAP. A Axial view of the segmentation. B 3D reconstruction of the segmented volume

Full size image
Fig. 2
figure 2

Ventricular Volume and Aβ42 concentration. Scatterplot of ventricular volume Aβ42 concentration for the full sample, n = 136

Full size image
Table 1 Correlation between ventricular volume, Aβ42, tau and p-tau
Full size table
Fig. 3
figure 3

Ventricular Volume and Aβ42 concentration in the male subgroup. Scatterplot of ventricular volume Aβ42 concentration for the male subgroup, n = 89

Full size image
Fig. 4
figure 4

Ventricular Volume and Aβ42 concentration in the female subgroup. Scatterplot of ventricular volume Aβ42 concentration for the female subgroup, n = 47

Full size image
Fig. 5
figure 5

Tau and Aβ42 in the male subgroup. Scatterplot of tau and Aβ42 concentration for the male subgroup, n = 89

Full size image
Fig. 6
figure 6

Tau and Aβ42 in the female subgroup. Caption: Scatterplot of tau and Aβ42 concentration for the female subgroup, n = 47

Full size image

Median VV and CSF biomarker concentrations are shown in Table 2. Males had significantly larger VV than females, but no significant differences were found in CSF biomarker concentration, age nor in improvement rate. Neither did the two clinics differ significantly in these variables nor in sex distribution. Detailed clinical information was available for a subset of the participants consisting of 83 individuals. At the first follow-up, 63 (76%) had improved after surgery.

Table 2 Participant characteristics
Full size table

Discussion

Considering previous reports of generally low CSF concentrations of Aβ42, tau and p-tau in iNPH [6,7,8,27,28,29,30] and a high comorbidity between AD pathology and iNPH has been found in perioperative brain biopsies [52, 53]. With this in mind, one possible explanation for our discrepant findings between the sexes could be a higher prevalence of preclinical AD pathology in the female group. However, this explanation is contradicted by the lack of difference between the sexes regarding absolute concentration of Aβ42, t-tau and p-tau. Another explanation is that there are sex differences in the complex pathophysiology of iNPH affecting the pattern of CSF-biomarkers. In a recent study, Graff-Radford et al. [54] underlined the need to take CSF disorders like iNPH into account in the interpretation of Aβ42, tau and p-tau levels. Further studies on a possible dilution effect would gain from complementary information on AD pathology, for example APO-E genotype and AD pathology in perioperative brain biopsies, to explore interaction between the effects of sex, iNPH and AD on CSF biomarker patterns.

In a previous study, an increase in Aβ42 has been observed in patients with iNPH one year after shunt surgery, though only in a low tau-subgroup [55]. In the high tau-subgroup, Aβ42 remained unchanged. Due to a cognitive decline in the high tau group the authors considered the possibility of AD pathology in the group, this could in theory cause a progressive lowering of Aβ42 canceling out the effects of shunting at follow-up. A recent longitudinal study on CSF biomarkers before and after shunt surgery found a postsurgical decrease in Aβ42 and simultaneous increase in tau, p-tau [52]. The authors argued that decrease in Aβ42 was mainly driven by APOE epsilon 4 carriers, indicating a possible role of AD pathology in the evolution of biomarker concentrations. In other studies, an increase in the levels of different lumbar CSF biomarkers was seen after shunt insertion, except for Aβ42 that was unchanged [8, 56]. The latter findings seem to contradict the dilution theory, given that the expected outcome of a decrease in VV following shunting would be an overall increase in the concentrations of CSF biomarkers, including Aβ42. However, lumbar CSF composition of brain-derived proteins is affected by absorption and metabolism along the spinal canal [57] and as such it susceptible to the changes in fluid dynamics introduced by a shunt and effects of surgery on CSF content should be interpreted with caution.

In summary, our results do not support dilution to be the major mechanism behind the lower lumbar CSF concentrations of Aβ42, t-tau and p-tau found in iNPH. Due to the unexplained sex differences, however, the possibility of a minor dilution effect cannot be dismissed. As our study focused on whether dilution could explain the lower concentrations of Aβ42, tau and p-tau in iNPH, our results provide no further insight regarding other possible pathophysiological explanations.

Strengths and limitations

That our material consists of patients undergoing routine clinical care for iNPH with a shunt response rate and levels of Aβ42, tau and p-tau in accordance with previous reports supports generalizability. Lumbar CSF was used which is a methodological limitation shared with most previous studies on CSF composition in iNPH. Levels of biomarkers of ventricular CSF would probably have been more relevant to relate with VV and better reflect the pathophysiology of iNPH. While there is a known difference between ventricular and lumbar CSF composition [48, 57], how predictable the difference is is largely unknown. One study showed a strong correlation between ventricular and lumbar CSF composition, although based on only five participants [58]. Further research in this area would be valuable, however ventricular CSF from healthy individuals is seldom available.

In our study, neither the spinal nor other extraventricular CSF was covered by the volumetric measurements. As the largest increase in CSF volume in iNPH is intraventricular this is unlikely to have masked a dilution effect.

The material provided sufficient power to detect a medium or larger effect supporting its reliability. For the subgroup analyses the power was weaker, though considering the small overlap between the confidence intervals of rs for the correlation between Aβ42 and tau in particular, a true difference between the subgroups is likely to exist.

Conclusion

A major dilution effect was not found when analyzing the correlation between VV and lumbar CSF levels of Aβ42, tau and p-tau biomarker in the total sample of 136 iNPH patients. Subgroup analyses revealed sex-based differences which highlights a need for stratification when investigating iNPH pathophysiology.