Normal pressure hydrocephalus (NPH) is characterized by pathologically enlarged ventricles without elevated cerebrospinal fluid (CSF) pressure along with a triad of clinical symptoms including gait disturbances, urinary incontinence, and cognitive impairment. NPH is evaluated with lumbar drain trials (LDTs) where CSF is removed over several days to determine if patients would benefit from ventricular shunting. Candidate selection and success for these surgeries remains challenging because other diseases such as Alzheimer’s disease (AD) share common features with NPH in cognitive impairment and enlarged ventricles. Prior research has found that 20%-40% of presumed NPH cases have AD pathology as determined by brain biopsy or autopsy. CSF biomarkers of AD can be altered in NPH and are not always conclusive, complicating the interpretation of results when formulating diagnoses and prognoses. Studies to refine the analyses of AD CSF biomarkers in NPH are needed. We aimed to examine the frequency of CSF biomarker results among patients presenting for NPH evaluations with LDTs.

62 patients presented for LDTs upon physician recommendations. CSF specimens were sent to Mayo Clinic Laboratories for Alzheimer Disease Evaluation (ADEVL) that utilizes Elecsys (Lenexa, KS) CSF electrochemiluminescence immunoassays (Roche Diagnostics, Basel, Switzerland) to measure levels of amyloid-beta 42 (Aβ42), total tau (t-tau), and phosphorylatedtau (p-tau), and p-tau:Aβ42 ratio. Results were classified based on interpretation through the Amyloid/Tau/Neurodegeneration (ATN) framework1: 1) AD - biomarker profile consistent with AD pathologic change, 2) non-AD profile - biomarker levels normal or inconsistent with AD pathologic change, or 3) indeterminate - biomarkers were incongruous with only one or two abnormal levels of Aβ42, t-tau, p-tau, or ptau: Aβ42. Indeterminate cases may represent altered protein levels due to CSF dynamics or AD-related pathologic change. In reviewing recent research on CSF dynamics and AD biomarkers in NPH2 a p-tau threshold of 15 pg/mL was derived and implemented such that cases with Aß42 <=1026 pg/mL and p-tau <15 pg/mL were designated as suspected non-AD, and those with Aß42 <=1026 pg/mL and p-tau >15 pg/mL were designated suspected AD.

Of the 62 LDT cases, 12 (19.35%) were classified as AD, 31 (50%) were indeterminate and 22 (35.48%) were non-AD. Of the 31 indeterminate cases, 21 (33.87% of the overall sample) were suspected non-AD and 7 (11.29% of the full sample) were categorized as suspected AD.

Our findings show that 20%-30% of patients presenting for LDT showed evidence for AD-type pathologic change, consistent with prior reports of AD pathology in cases of possible NPH. Half of all LDT cases had indeterminate AD CSF biomarker results, the interpretations of which were confounded by the potential alterations of CSF biomarkers levels due to NPH independent of AD. Our findings emphasize the need to establish better approaches to interpreting CSF AD biomarkers in evaluating NPH. Future research should examine the discriminative utility of CSF AD biomarkers and the selected p-tau threshold in indeterminate cases for predicting response to LDT and shunting.

Patients and their families often ask clinicians to estimate when full-time care (FTC) will be needed after Alzheimer's Disease (AD) is diagnosed. Although a few algorithms predictive algorithms for duration to FTC have been created, these have not been widely adopted for clinical use due to questions regarding precision from limited sample sizes and lack of an easy, user friendly prediction model. Our objective was to develop a clinically relevant, data-driven predictive model using machine learning to estimate time to FTC in AD based on information gathered from a) clinical interview alone, and b) clinical interview plus neuropsychological data.

The National Alzheimer's Coordinating Center dataset was used to examine 3,809 participants (M age at AD diagnosis = 76.05, SD = 9.76; 47.10% male; 87.20% Caucasian) with AD dementia who were aged >50 years, had no history of stroke, and not dependent on others for basic activities of daily living at time of diagnosis based on qualitative self or informant report. To develop a predictive model for time until FTC, supervised machine learning algorithms (e.g., gradient descent, gradient boosting) were implemented. In Model 1, 29 variables captured at the time of AD diagnosis and often gathered in a clinical interview, including sociodemographic factors, psychiatric conditions, medical history, and MMSE, were included. In Model 2, additional neuropsychological variables assessing episodic memory, language, attention, executive function, and processing speed were added. To train and test the algorithm(s), data were split into a 70:30 ratio. Prediction optimization was examined via cross validation using 1000 bootstrapped samples. Model evaluation included assessment of confusion matrices and calculation of accuracy and precision.

The average time to requiring FTC after AD diagnosis was 3.32 years (Range = 0.53-14.57 years). For the clinical interview only model (Model 1), younger age of onset, use of cholinesterase inhibitor medication, incontinence, and apathy were among the clinical variables that significantly predicted duration to FTC, with the largest effects shown for living alone, a positive family history of dementia, and lower MMSE score. In Model 2, the clinical predictors remained significant, and lower Boston Naming Test and Digit-Symbol Coding scores showed the largest effects in predicting duration to FTC among the neuropsychological measures. Final prediction models were further tested using five randomly selected cases. The average estimated time to FTC using the clinical interview model was within an average of 5.2 months of the recorded event and within an average of 5.8 months for the model with neuropsychological data.

Predicting when individuals diagnosed with AD will need FTC is important as the transition often carries significant financial costs related to caregiving. Duration to FTC was predicted by clinical and neuropsychological variables that are easily obtained during standard dementia evaluations. Implementation of the model for prediction of FTC in cases showed encouraging prognostic accuracy. The two models show promise as a first step towards creation of a user friendly prediction calculator that could help clinicians better counsel patients on when FTC after AD diagnosis may occur, though the development of separate models for use in more diverse populations will be essential.

Definitive diagnosis of Alzheimer’s disease (AD) is often unavailable, so clinical diagnoses with some degree of inaccuracy are often used in research instead. When researchers test methods that may improve clinical accuracy, the error in initial diagnosis can penalize predictions that are more accurate to true diagnoses but differ from clinical diagnoses. To address this challenge, the current study investigated the use of a simple bias adjustment for use in logistic regression that accounts for known inaccuracy in initial diagnoses.

A Bayesian logistic regression model was developed to predict unobserved/true diagnostic status given the sensitivity and specificity of an imperfect reference. This model considers cases as a mixture of true (with rate = sensitivity) and false positives (rate = 1 - specificity) while controls are mixtures of true (rate = specificity) and false negatives (rate = 1 - sensitivity). This bias adjustment was tested using Monte Carlo simulations over four conditions that varied the accuracy of clinical diagnoses. Conditions utilized 1000 iterations each generating a random dataset of n = 1000 based on a true logistic model with an intercept and three arbitrary predictors. Coefficients for parameters were randomly selected in each iteration and used to produce a set of two diagnoses: true diagnoses and observed diagnoses with imperfect accuracy. Sensitivity and specificity of the simulated clinical diagnosis varied with each of the four conditions (C): C1 = (0.77, 0.60), C2 = (0.87, 0.44), C3 = (0.71, 0.71), and C4 = (0.83, 0.55), which are derived from published values for clinical AD diagnoses against autopsy-confirmed pathology. Unadjusted and bias-adjusted logistic regressions were then fit to the simulated data to determine the models’ accuracy in estimating regression parameters and prediction of true diagnosis.

Under all conditions, the bias-adjusted logistic regression model outperformed its unadjusted counterpart. Root mean square error (the variability of estimated coefficients around their true parameter values) ranged from 0.23 to 0.79 for the unadjusted model versus 0.24 to 0.29 for the bias-adjusted model. The empirical coverage rate (the proportion of 95% credible intervals that include their true parameter) ranged from 0.00 to 0.47 for the unadjusted model versus 0.95 to 0.96 for the bias-adjusted model. Finally, the bias-adjusted model produced the best overall diagnostic accuracy with correct classification of true diagnostic values about 78% of the time versus 62-72% without adjustment.

Results of this simulation study, which used published AD sensitivity and specificity statistics, provide evidence that bias-adjustments to logistic regression models are needed when research involves diagnoses from an imperfect standard. Results showed that unadjusted methods rarely identified true effects with credible intervals for coefficients including the true value anywhere from never to less than half of the time. Additional simulations are needed to examine the bias-adjusted model’s performance under additional conditions. Future research is needed to extend the bias adjustment to multinomial logistic regressions and to scenarios where the rate of misdiagnosis is unknown. Such methods may be valuable for improving detection of other neurological disorders with greater diagnostic error as well.

Learning curve patterns on list-learning tasks can help clinicians determine the nature of memory difficulties, as an “impaired” score may actually reflect attention and/or executive difficulties rather than a true memory impairment. Though such pattern analysis is often qualitative, there are quantitative methods to assess these concepts that have been generally underutilized. This study aimed to develop a model that decomposes learning over repeated trials into separate cognitive processes and then include other testing data to predict performance at each trial as a function of general cognitive functioning.

Data for CVLT-II learning trials were obtained from an outpatient neuropsychology service within an academic medical center referred for clinical reasons. Participants with a cognitive diagnosis of non-demented (ND) or probable Alzheimer’s disease (AD) were included. The final sample consisted of 323 ND [Mage = 58.6 (14.8); Medu = 15.4 (2.7); 55.7% female] and 915 AD [Mage = 72.6 (9.0); Medu = 14.2 (3.1); 60.1% female cases. A Bayesian non-linear beta-binomial multilevel model was used, which uses three parameters to predict CVLT-II recall-by-trial: verbal attention span (VAS), maximal learning potential (MLP), and learning rate (LR). Briefly, VAS predicts expected first trial performance while MLP, conversely, predicts the expected best performance as trials are repeated, and LR weights the influence of VAS versus MLR over repeated trials. Predictors of these parameters included age, education, sex, race, and clinical diagnosis, in addition to raw scores on Trail Making Test Parts A and B, phonemic (FAS) fluency, animal fluency, Boston Naming Test, Wisconsin Card Sorting Test (WCST) Categories Completed, and then age-adjusted scaled scores from WAIS-IV Digit Span, Block Design, Vocabulary, and Coding. Random intercepts were included for each parameter and extracted for comparison of residual differences by diagnosis.

The model explained 84% of the variance in CVLT-II raw scores. VAS reduced with age and time-to-complete Trails B but improved with both verbal fluencies and confrontation naming. MLP increased as a function of WAIS Digit Span, animal fluency, confrontation naming, and WCST categories completed. Finally, LR was greater for females and WAIS-IV Coding and Vocabulary performances but reduced with age. Participants with AD had lower estimates of all three parameters: Cohen’s d = 2.49 (VAS) - 3.48 (LR), though including demographic and neuropsychological tests attenuated differences, Cohen’s d = 0.34 (LR) - 0.95 (MLP).

The resulting model highlights how non-memory neuropsychological deficits affect list-learning test performance. At the same time, the model demonstrated that memory patterns on the CVLT-II can still be identified beyond other confounding deficits since having AD affected all parameters independent of other cognitive impairments. The modeling approach can generate conditional learning curves for individual patient data, and when multiple diagnoses are included in the model, a person-fit statistic can be computed to return the mostly likely diagnosis for an individual. The model can also be used in research to quantify or adjust for the effect of other patient data (e.g., neuroimaging, biomarkers, medications).

Higher baseline dispersion (intra-individual variability) across neuropsychological test scores at a single time-point has been associated with more rapid cognitive decline, onset of Mild Cognitive Impairment (MCI) and Alzheimer’s disease (AD), faster rates of hippocampal and entorhinal atrophy, and increased AD neuropathology. Comparison between predictions made from test score dispersion within a cognitive domain versus global, cross-domain dispersion is understudied. Global dispersion may be influenced by ability-and test-specific characteristics. This study examined the performance of global versus domain-specific dispersion metrics to identify which is most predictive of cognitive decline over time.

Data for baseline and five follow-up visits of 308 participants with normal cognition (Mage=73.90, SD=8.12) were selected from the National Alzheimer’s Coordinating Center (NACC) Dataset. Participants were required to have no focal neurological deficits, or history of depression, stroke, or heart attack. Diagnoses and progression to MCI and/or dementia were determined at each visit through consensus conferences. Raw neuropsychological scores were standardized using NACC norms. Global baseline dispersion was defined as the intraindividual standard deviation (ISD) across the 10 scores in the NACC battery. Domain-specific dispersions were calculated by constructing composites and ISD was computed across tests sampling their respective domains (executive functioning/attention/processing speed [EFAS], language, and memory; see Table 1 for details on these tests). Higher values on each of these metrics reflect greater dispersion. Multinomial logistic regression model fit statistics and parameter estimates were compared across four different models (global, EFAS, Language, and Memory dispersion) covarying for age, years of education, sex, race, ethnicity, and ApoE4 status. Models were compared using the Likelihood Ratio Test (LRT) and the Akaike Information Criteria (AIC) of Models statistics.

Of the 308 participants, 70 (22.7%) progressed to MCI, and 82 (26.6%) progressed to dementia. Tables 1 and 2 show the results of the logistic regressions for the four models. All models fit the data well, with statistically significant predictions of conversion. Model 1 (global dispersion) showed a better fit than domain-specific models of dispersion per LRT and AIC values. Consistent with the results from mean differences between groups, parameter estimates showed that only global dispersion and EFAS dispersion significantly predicted conversion to dementia (when included with other covariates in models), with higher dispersion reflecting a greater risk of conversion.

In this sample, baseline global and EFAS dispersion measures significantly predicted conversion to dementia. Although global dispersion was a stronger predictor of dementia progression, findings suggest that executive functioning performance may be driving this relationship. A single index of global variability, from the calculation of standard deviation across test scores, may be supplementary for clinicians when distinguishing individuals at risk for dementia progression. None of the models were predictive of conversion to MCI. Further research is required to examine cognitive variability differences among patients who progress to MCI and patient-specific factors that may relate to test score dispersion and its utility in predicting the progression of symptoms.

Evidence-based diagnostic methods have clinical and research applications in neuropsychology. A flexible Bayesian model was developed to yield diagnostic posttest probabilities from a single person’s neuropsychological score profile by utilizing sample descriptive statistics of the test battery across diagnostic populations of interest.

Three studies examined the model’s performance. One simulation examined estimation accuracy of true z-scores. A diagnostic accuracy simulation utilized descriptive statistics from two popular neuropsychological tests, the Wechsler Adult Intelligence Scale–IV (WAIS-IV) and Repeatable Battery for the Assessment of Neuropsychological Status (RBANS). The final simulation examined posterior predictive accuracy of scores to those reported in the WAIS manual.

The model produced minimally biased z-score estimates (root mean square errors: .02–.18) with appropriate credible intervals (95% credible interval empirical coverage rates: .94–1.00). The model correctly classified 80.87% of simulated normal, mild cognitive impairment, and Alzheimer’s disease cases using a four subtest WAIS-IV and the RBANS compared to accuracies of 60.67–65.60% from alternative methods. The posterior predictions of raw scores closely aligned to percentile estimates published in the WAIS-IV manual.

This model permits estimation of posttest probabilities for various combinations of neuropsychological tests across any number of clinical populations with the principal limitation being the accessibility of applicable reference samples. The model produced minimally biased estimates of true z-scores, high diagnostic classification rates, and accurate predictions of multiple reported percentiles while using only simple descriptive statistics from reference samples. Future nonsimulation research on clinical data is needed to fully explore the utility of such diagnostic prediction models.