We investigated relationships between the ptau/Aβ1-42 ratio in ventricular CSF and cortical AD pathology, initial clinical features, shunt outcome, and lumbar CSF ptau/Aβ1-42 ratios in patients being shunted for iNPH. We found that the ventricular CSF ptau/Aβ1-42 ratio differentiated patients with cortical AD pathology from those without pathology with a sensitivity of 68% and a specificity of 67% at the optimal cutoff point. Patients with the highest levels of ventricular ptau/Aβ1-42 (more suggestive of AD) performed more poorly at baseline on a variety of psychometric measures but performed better on gait measures. Higher ptau/Aβ1-42 ratios were associated with poorer improvement in both cognition and gait after shunting. Finally, we demonstrated a significant correlation between ptau/Aβ1-42 ratios in the ventricular and lumbar CSF in a subset of our cohort, suggesting that measures of lumbar ptau/Aβ1-42 may also correlate with patterns of baseline performance and shunt outcomes in patients with suspected iNPH.
The impact of AD pathology on the specific cognitive measures used in this study is informative. Improvement in the entire cohort of subjects after shunting was most significant on certain tests of executive function, attention, and speeded psychomotor action (category fluency, digit symbol, and trails A time), and not memory. Deficits in these cognitive domains (classically considered part of a constellation of frontostriatal or so-called "subcortical" deficits) have long been associated with iNPH [16
], such that improvement in these areas in response to shunting was expected. However, it is notable that significant baseline differences were identified between subjects with high and low ptau/Aβ1-42 ratios in ventricular CSF on measures that also tested these aspects of cognition (Trails B, digit span and digit symbol). The finding that comorbid AD pathology worsens performance in these specific areas reinforces the notion that it is difficult to disambiguate patients with the clinical syndrome of iNPH from those with co-morbid AD based on patterns of cognitive impairment. On the other hand, the finding that patients with high ventricular CSF ptau/Aβ1-42 ratios have better baseline gait performance suggests that cognitive symptoms tend to predominate the overall clinical picture of those patients who are affected by comorbid AD and iNPH. The finding that patients with high ptau/Aβ1-42 ratios show less recovery of gait function after shunting is accounted for at least partly by their higher baseline gait performance compared to patients lacking comorbid AD pathology.
Patients with high ptau/Aβ1-42 ratios showed poorer post-shunt improvement mainly on tests of memory (immediate and delayed word list recall) and visuospatial function (clock copy)--cognitive domains that are typically impaired in patients with AD. In other words, while the baseline cognitive profile of these patients may not readily distinguish patients with comorbid AD pathology from those without, their response to shunting may suggest selective failure of improvement in cognitive domains typically associated with AD. Examining individual psychometric measures more closely, there also emerged an interesting dissociation between performance on different memory tests in our battery. While the ptau/Aβ1-42 ratio made a difference with respect to outcomes for verbal list learning, it did not impact logical memory performance. One possible explanation is that the two tasks differ with respect to specific cognitive demands. Evidence suggests that performance on story memory tasks relies, to some extent, on intrinsic semantic organization of the material, while word-list memory tasks require subjects to self-generate organizational strategies between unrelated items [17
]. Word list memory is thus thought to rely more on executive function, a domain thought to be preferentially affected in iNPH [16
]. The finding that patients with low ptau/Aβ1-42 ratios show selective improvement in word list recall is therefore potentially consistent with the idea that iNPH may influence performance on word list recall tasks to a greater extent than performance on story memory tasks. However, the finding that baseline logical memory performance did not differ significantly between high and low ptau/Aβ1-42 ratio groups was somewhat unexpected, since more severe baseline impairments in this area might have been predicted in patients with AD pathology.
Our finding that total tau does not predict the presence of cortical AD pathology in patients with iNPH is consistent with a previous study which found elevated total tau levels in the lumbar CSF of patients with iNPH and those with AD [18
]. In addition, Tisell and colleagues [19
] have investigated the relationship between ventricular total tau levels and shunt outcome in iNPH, reporting no significant relationship between total tau and shunt response as assessed using measures of gait, balance, alertness, and cognition. One likely explanation for the difference between these findings and our own is our use of a ratio of biomarkers that may be more specifically associated with AD pathological processes than total tau. This is further supported by a study conducted by Tarnaris and colleagues [20
] that found that total tau achieved a sensitivity of only 17% and a specificity of only 20% for predicting favorable shunt outcome, as assessed by determining the extent to which patients demonstrated non-transient improvement and resumption of pre-illness activity at 6 months. Interestingly, the combination of total tau and Aβ1-42 predicted favorable outcome with a sensitivity of 80% and a specificity of 82.4%.
Ventricular CSF biomarker information, although not available as part of a typical clinical outpatient evaluation for iNPH, may potentially be used to guide postoperative management for those patients who receive magnetically programmable shunts. CSF AD biomarker data obtained intraoperatively may identify a subset of patients for whom repeated shunt adjustments are unlikely to yield clinical benefit, thus sparing those patients from additional risk of overdrainage and other complications. Bauman and colleagues [21
] recently observed that cortical pathology data may similarly inform shunt management strategy. In addition, ventricular biomarker levels consistent with AD may potentially suggest that a patient previously presumed to have iNPH should be re-evaluated and possibly treated for AD.
Our finding that ptau/Aβ1-42 ratios in the ventricular and lumbar CSF are correlated suggests that lumbar ptau/Aβ1-42 may also predict baseline performance and shunt outcome in patients with suspected iNPH. The regression equation suggests that the optimal cutoff ratio for the lumbar CSF is much smaller than the ventricular CSF ratio. This is consistent with the findings of Tarnaris and colleagues [22
] that the levels of various biomarkers, including total tau and Aβ1-42, differ between the ventricular and lumbar CSF. Although we have provided a regression equation that predicts the lumbar CSF ptau/Aβ1-42 ratio from the ventricular ratio, the optimal lumbar ratio cutoff should ideally be determined by replicating our analysis in a larger sample using lumbar CSF biomarkers directly.
This study has several limitations. One is its relatively small sample size. In addition, a number of patients who had received shunts to treat presumed iNPH did not participate in the study, either because they were excluded by the study criteria or because they failed to follow up. Restriction of the patients participating in the study in these ways may limit the generalizability of the findings and may also have led to a degree of selection bias. A comparison of baseline performance between patients who completed the study versus those who enrolled but did not complete the study reveals a statistically significant difference in the cognitive subscale measure of the NPH scale. Notably, this result was not supported by performance on any individual psychometric instrument. Nonetheless, this finding suggests that patients who completed the study may have had slightly worse baseline cognition than those who dropped out. While we believe that the results of the study are still valid and informative, we also acknowledge that the data may be somewhat more informative for patients who have cognitive impairments as part of their presenting clinical picture. Finally, we assessed outcome at 4 months following surgery. A longer timeframe for assessment of outcome in these patients may be more useful for gauging long-term outcome and we acknowledge the possibility that clinical and cognitive characteristics of these patients may evolve long after 4 months following surgery. However, given that the temporal dynamics of post-surgical symptom improvement in iNPH patients are not well characterized, these findings do contribute meaningfully to what is currently known about recovery of function after shunting.