Predictors of Response to Cholinesterase Inhibitors Treatment of Alzheimer’s Disease: Date Mining from the TREDEM Registry

Abstract

Background:

The pharmacological treatment of Alzheimer’s disease (AD) is based largely on cholinesterase inhibitors (ChEI).

Objective:

To investigate whether or not some non-pharmacological and contextual factors measured prior to starting treatment such as past occupation, lifestyles, marital status, degree of autonomy and cognitive impairment, living alone or with others, and the degree of brain atrophy are associated with a better response to ChEI treatment.

Methods:

Eighty-four AD and six AD with cerebrovascular disease (AD + CVD) outpatients of Treviso Dementia (TREDEM) Registry, with an average cholinesterase inhibitors treatment length of four years, were considered. The outpatients had undergone a complete evaluation and some non-pharmacological and contextual factors were collected. We defined responder a patient with a delta score T0 – T1 equal or inferior to 2.0 points per year of MMSE and a non-responder a patient with a delta score T0 – T1 superior to 2.0 points per year. In order to identify hidden relationships between variables related to response and non-response, we use a special kind of artificial neural network called Auto-CM, able to create a semantic connectivity map of the variables considered in the study.

Results:

A higher cognitive profile, a previous intellectual occupation, healthier lifestyles, being married and not living alone, a higher degree of autonomy, and lower degree of brain atrophy at baseline resulted in affecting the response to long-term ChEI therapy.

Conclusion:

Non-pharmacological and contextual factors appear to influence the effectiveness of treatment with ChEI in the long term.

Keywords

Alzheimer’s disease Auto-CM system brain and cognitive reserve brain atrophy cholinesterase inhibitors environmental enrichment TREDEM

Introduction

Dementia is one of the most common diseases that affect older people, with prevalence rates ranging between 5.9–9.4% for subjects aged over 65 years in the European Union [1]. Alzheimer’s disease (AD) is the most common type of dementia that affects 50–60% of all demented patients.

Pharmacological treatment of AD is based on cholinesterase inhibitors (ChEIs) and NMDA receptor antagonists (memantine). These drugs induce modest improvements on cognitive functions, activities of daily living, and behavioral and psychological symptoms of dementia, in patients whose level of disease severity ranges from mild to severe [2 –6]. ChEIs have proved effective in slowing down clinical progression of disease in 12, 24, 30, and 52 weeks placebo controlled trials [7 –14]. Many published studies are double blind randomized controlled trials (RCT) investigating the efficacy of pharmacological treatment, in which donepezil, rivastigmine, or galantamine are compared with placebo. Cholinesterase inhibition appears to reduce decline in cognitive performances and the risk of nursing home admission [15, 16]. Treatment with galantamine, at 6 months, significantly improved cognition and global function and these results were also maintained for 12 months with high dosage [17].

There are some studies that tested the efficacy of ChEIs in long-term treatment. Donepezil is an effective and safe drug for the long-term symptomatic treatment of mild to moderate AD up to 2.8 years [18], and in a “US multicenter open-label study” it was even shown to be efficacious over a period up to 4.9 years [19]. The authors detected major evidence of clinical improvement from baseline during the first 6–9 months of treatment. Although after this period the cognitive decline continued, it was less evident compared to patients who had not been treated [19].

Studies on ChEIs and memantine showed that they are beneficial for both cognitive and non-cognitive symptoms associated with AD [20], but that it was difficult to identify possible predictors of positive response to therapy.

Many studies have focused on the relationship between the degree of disease severity or rate of progression, and response to drug treatment. Rivastigmine is effective and safe for patients with mild and moderate degrees of illness [21, 22]. In others studies, authors claimed that the rate of progression might be more important than disease severity in determining prognosis [23, 24].

Most of the information about ChEIs treatment results from RCTs, while there is little evidence of their efficacy in the real world. Attention has mostly been focused on dosage and duration of treatments and less on research of contextual factors that may facilitate or reduce the response to therapy. In this study, we investigated whether or not some non-pharmacological and contextual factors measured prior to starting treatment (i.e., past occupation, lifestyles, age, marital status, degree of autonomy and cognitive impairment, living alone or with others, the degree of brain atrophy, etc.) are associated with a better response to ChEIs treatment.

In order to identify hidden relationships between variables related to response and non-response, i.e., perform data mining to provide a better understanding of the condition complexity, we use a special kind of artificial neural network (ANN) called an Auto-Contractive Map (Auto-CM), able to create a semantic connectivity map of the variables considered in the study.

Particularly, Auto-CM is a peculiar ANN able to define the strength of the associations of each variable with all the others and to visually show the map of the main connections of the variables and the basic semantic of their ensemble.

PATIENTS AND METHODS

Population

Eighty-four AD and six AD with cerebrovascular disease (AD + CVD) outpatients of Treviso Dementia (TREDEM) Registry with an average treatment length with ChEI of four years and minimum of two years constituted the study group. The TREDEM Registry is an observational prospective data collection on dementia conducted at the Cognitive Impairment Center of the Local Health Authority n. 9 of Treviso (LHA9) and consists of data of 1,364 subjects of the TREDEM study [25 –28] collected from 2000 up to 2010 plus those collected later by the same methods; in total 2,315 subjects up to December 2014, 431 AD patients of whom have started treatment with ChEI. The TREDEM Registry was approved by the Treviso Province Hospitals Ethics Committee.

Inclusion criteria for entry in the study

The criteria of the National Institute of Neurological and Communicative Disorders and Stroke and the Alzheimer’s Disease and Related Disorders Association (NINCDS-ADRDA) [29] and the diagnostic criteria of the National Institute of Neurological Disorders and Stroke and the Association Internationale pour la Recherche et l’Enseignement en Neurosciences –NINDS–AIREN [30] were used to identify the diagnoses of AD and AD + CVD, respectively. All of the eligible patients examined in the outpatient clinic during the study period were invited to participate in the research. The ninety outpatients attended the Treviso Cognitive Impairment Center at different times between 2000 and 2014.

Exclusion criteria

Patients were excluded from the study if they presented with mild cognitive impairment; other types of dementia, such as pure vascular dementia, frontotemporal dementia, dementia with Lewy bodies, etc.; severe dementia (MMSE <14/30 or CDR 3).

We considered only the subjects that had undergone a complete biological, clinical, neuroradiological (brain CT), and neuropsychological evaluation. We did not consider patients with psychosis and major depression diagnoses. The patient were reassessed after 1 month, 3 months, 9 months, and then twice a year, every six months.

We have not considered patients with a follow up of less than two years, to exclude the variability of the MMSE score in the short term and preferring a consolidated data in the longer term [31]. Therefore the number of patients has decreased from 431 to 90.

Drug treatment

Donepezil, galantamine, or rivastigmine was prescribed to the patients. For each patient it had reached maintenance dose, gradually increasing the dose, from 5 mg to 10 mg for donepezil, from 8 mg to 16 mg for galantamine, and from 4.6 to 9.5 mg or 13.3 mg for rivastigmine patch. None of the 90 patients included in the analysis was being treated with memantine.

Definition of response to the treatment

The definition of response to the treatment has been established according to the delta between baseline (T0) and last available follow-up point (T1) in MMSE score. The length of follow-up (years) had an average of 4.07 ± 1.85 years. Assuming that the natural history of untreated AD is characterized by an annual average decrease of 2.7–3.3 MMSE points [32, 33], we operatively and conservatively defined responder a patient with a delta score T0 – T1 equal or inferior to 2.0 points per year and a non-responder a patient with a delta score T0 – T1 superior to 2.0 points per year.

Data collection

The data used in this work are shown in Table 1. Age, gender, lifestyles such as alcohol intake and smoking, marital status, living alone or with others, and urinary and fecal incontinence were obtained at the baseline clinic visit. Past occupations such as farmers, artisans, workmen, and tradesmen were classified as blue collar, while office workers, teachers, and professionals were considered white collar. The characteristics of responders and non-responders are shown in Table 2.

Clinical and disability assessment

Clinical assessment included a comprehensive medical neurological and geriatric assessment. Functional disability was measured using the activities of daily living (ADL) [34] and the instrumental ADL (IADL) scales [35, 36].

Neuropsychological assessment

A trained psychologist performed a complete neuropsychological assessment of participants by administrating several tests [25]. In the current study, we used Mini-Mental State Examination (MMSE) [37] and Clinical Dementia Rating (CDR) scale [38, 39].

Structural neuroimaging

CT scans were acquired with the volumetric scanner EMOTION 6 Siemens. Section orientation was parallel to the orbitomeatal plane. Sections on the same plane (time of 2 s, 120 kV, 130 mA, section thickness of 5 mm, no intersection gap) covered the brain from the inferior aspect of the cerebellum to the vertex of the cranium.

All data were analyzed from 8 to 10 images in the temporal lobe region, and from 16 to 18 images rostral to the temporal region. No contrast medium was used. All the measurements were taken and recorded by two independent trained raters blinded to clinical information. We considered the reliability of repeated CT scans evaluations of two raters as described in Statistical analysis: A physician neuro-radiologist and another one, of the Cognitive Impairment Center, experienced in reading the images.

Brain atrophy

Severity (low, moderate, or severe) of atrophy in (frontal, parietal, temporal, and occipital) cortical and (periinsular, basal, vault) subcortical regions as well as lateral ventricular enlargement were recorded. The severity of atrophy was detected by the widening of sulci and narrowing of gyri and by the reduction in amplitude of the respective regions [25].

Statistical analysis

Interrater reliability refers to the accuracy of repeated CT scans evaluations of any given subject by the two different raters. We used the Kappa statistic as the index of agreement [40]. Values greater than 0.80 are considered to indicate good agreement. Statistical analyses were conducted using the software package JMP7 (SAS Institute Inc., Cary, NC, USA). The agreement statistic was high (0.86).

We constructed a semantic connectivity map through Auto-CM system (Semeion), a fourth generation ANN, to offer some insight regarding the complex biological connections between the studied variables and the degree of brain atrophy. The system highlights the natural links among variables with a graph based on minimum spanning tree theory, where distances among variables reflect the weights of the ANN after successful training phase.

The Auto-CM was born as a new ANN. It was designed by Massimo Buscema at the Semeion Research Center [41]. The Auto-CM system finds, by a specific learning algorithm, a square matrix of weighted connections among the variables of any dataset. This matrix of connections presents many suitable features: a) non linear associations among variables are preserved; b) connections schemes among clusters of variables are captured, and c) complex similarities among variables become evident. Once an Auto-CM weights matrix is obtained, it is then filtered by a minimum spanning tree algorithm (MST) generating a graph whose biological evidence has already been tested in the medical field [42 –44]. The MST algorithm, described originally by the Czech scientist Otakar Borůvka in 1926, was later refined by Kruskal with a specific deterministic algorithm [45]; however, is still rarely used in the medical field. Auto-CM has been developed to explore the concomitant associations of different variables, and the potential relationships among variables in a multi-factor network relevant for the disease. The ultimate goal of this data mining model is to discover hidden trends and associations among variables, since this algorithm is able to create a semantic connectivity map in which non linear associations are preserved and explicit connection schemes are described. This approach shows the map of relevant connections between and among variables and the principal hubs of the system. Hubs can be defined as variables with the maximum amount of connections in the map. From a mathematical point of view, the specificity of Auto-CM algorithm is to minimize a complex cost function with respect to the traditional ones.

Traditional minimization cost function:

$E = Min {\sum_{i}^{N} \sum_{j}^{N} \sum_{q}^{M} u_{i}^{q} \cdot u_{j}^{q} \cdot σ_{i, j}}$

Auto-CM minimization cost function:

$\begin{matrix} E = Min {\sum_{i}^{N} \sum_{j}^{N} \sum_{k}^{N} \sum_{q}^{M} u_{i}^{q} \cdot \\ u_{j}^{q} \cdot u_{k}^{q} \cdot A_{i, j} \cdot A_{i, k}}; \\ A = (1.0 - \frac{w}{C}); \\ N = Number of Variables (Columns); \\ M = Number of Patterns (Rows) . \end{matrix}$

Comparing the two cost functions, it is evident how the traditional minimization includes only second order effects, while the Auto-CM considers also a third order effect. Practically, this means that the Auto-CM algorithm is able to discover variable similarities completely embedded in the dataset and invisible to the other classical tools. This approach describes a context that is typical of living systems where a continuous time dependent complex change in the variable value is present. Auto-CM can also learn under difficult circumstances such as when the connections of the main diagonal of the second connections matrix are removed. When the learning process is organized in this way, Auto-CM identifies specific relationships between each variable and all others. Consequently, from an experimental point of view, it appears that the ranking of its connections matrix is equal to the ranking of the joint probability between each variable and the others. Auto-CM requires a training phase necessary to learn how variables are interconnected. The learning algorithm of CM may be summarized in four orderly steps: a) Signal Transfer from the Input into the Hidden layer; b) Adaptation of the connections value between the Input layer and the Hidden layer; c) Signal Transfer from the Hidden layer into the Output layer; d) Adaptation of the connections value between the Hidden layer and the Output layer. All 90 subjects available have been included in the analysis. We transformed ADL scoring, for which a paradigmatic cut-off value is not available, into two variables by constructing a complementary variable scaled from 0 to 1. The transformation technique is explained elsewhere [27]. In the map, we named these two different forms as high and low. This pre-processing scaling is necessary to make possible a proportional comparison among ADL score and the other variables and to understand the existing links of the variable when the values tend to be high or low.

RESULTS

The sample had an average age of 77.8 ± 6.7 and was composed predominantly of women (81.1%). Nearly half of the subjects were married (48.8%), while 44.4% were widowers. 75% of the sample lived at home with one or more family members or with private assistance; 21% lived alone and only 3% in nursing homes. The majority of the subjects were housewives (52.2%); the blue collar workers (farmers, workmen, artisans and tradesmen) were 27.7% , while the white collar subjects (office workers, teachers and professionals) were 20% . In our sample, 53.3% of patients showed moderate or severe cortical atrophy and 56.6% moderate or severe subcortical atrophy. The patients had a mild degree of cognitive impairment (MMSE 20.4 ± 3.4) and a low-moderate degree of disability (ADL 4.9 ± 1.2) at baseline (T0). The intake of alcohol (mainly wine) included 37% of patients; former smokers and active smokers were 16.6% and 3% , respectively. Comorbidity was present but not high (CIRS Severity Index 1.4 ± 0.2) [46] in a similar way to that recorded in the TREDEM study previously (CIRS Severity Index 1.5 ± 0.3) [26] with in particular high blood pressure (64%), cardiac (30%), musculoskeletal (30%), peripheral vascular (26%), endocrine and metabolic diseases (22%). The number of drugs taken daily was equal to 4.3 ± 2.8, exactly what was reported in a previous TREDEM analysis (4.3 ± 2.7) [26].

According to the criteria adopted, 58 (64.4%) of the patients are considered as responders and 32 (35.6%) as non-responders. The map (Fig. 1) represents the natural connection scheme among variables.

We can observe the different characteristics of the variables most closely related with the condition of non-responders in the right side of the map and of those most closely associated with the condition of responders in central low part of the map. Better cognitive profile at baseline (MMSE >20; CDR 0.5 or 1) appears associated with the status of responder; the opposite (MMSE <20; CDR 2) appears with regard to the non-responder. “Blue collar” and “housewife” are linked to the status of non-responders while “white collar” is closely connected with responders. “Use of alcohol” and “former smoking” appear associated with “non-responder” status and not with the condition of responders. “Live alone” and “widower” are linked to the status of non-responders while “live with assistance” and “married” are closely connected with responders. A low degree of autonomy in basic activities of daily living (ADL low) is associated with non-responders while a higher degree of autonomy (ADL high) is correlated with responders.

“Moderate degree of cortical and subcortical atrophy” is related with the situation of non-responders while “low degree of cortical and subcortical atrophy” are connected with responders. Male gender and age less than 75 years appear to be connected with the responders, while the female gender, a more advanced age (between 75 and 85 years), and urinary incontinence to non-responders.

DISCUSSION

One of the main factors preventing a more efficient use of new pharmacological treatments for chronic diseases like, for example, hypertension, cancer, AD, or obesity, is represented by the difficulty in predicting a priori the chance of response of the single patient to a specific drug. While RCTs lack for external generalizability, a major methodological setback in drawing inferences and making predictions from data collected in the real world setting, such as observational studies, is that variability in the underlying biological substrates of the studied population and the quality and content of medical intervention influence outcomes. Because there is no reason to believe that these, like other health factors, work together in a linear manner, the traditional statistical methods, based on the generalized linear model, have limited value in predicting outcomes such as responsiveness to a particular drug. Therefore, in order to identify hidden relationships between variables related to response and non-response, we used an Auto-CM ANN. The Auto-CM matrix of connections preserves non linear associations among variables, while at the same time capturing elusive connection schemes among clusters that are often overlooked by traditional cluster analyses, and highlighting complex similarities among variables on various dimensions, such as role, connectivity, essentiality, and so on. The consistency of this system in picking-up real connection links with a causal relationship also in cross-sectional contexts is sustained by some publications. In fact,Auto-CM has been previously and successfully applied to several medical datasets. Among others, some examples include its application to detect the main connections of folate metabolism and chromosome damage [47], AD and brain atrophy [48], and to reveal the genetic risk factors for sporadic amyotrophic lateral sclerosis [49], and also the connections between symptoms in gastro-esophageal reflux disease [50] and fetal growth retardation [51].

In the case of AD, data from the RCTs of ChEIs, in addition to poorly reflecting real world situation, provide little information about their long-term effects. In fact most of the studies were of short (3 or 6 months) duration. Only some open-label extensions [52] and observational studies followed treated patients for more than a year and have been used to look for evidence of long-term efficacy and safety. In AD, placebo-controlled long-term studies conducted with ChEIs are not permitted for ethical reasons. Therefore, in these studies, patients’ outcomes on cognitive and functional assessment scales were compared with mathematical models or historical data from untreated cohorts [53, 54].

Our study is a prospective observational survey conducted in the real world and covers patients treated continuously for an average period of four years. The evaluation of the effectiveness was measured by identifying responders with a restrictive criterion such as a delta MMSE score T0 - T1 equal or inferior to 2.0 points per year and non-responders with a delta score superior to 2.0 points per year. Another group of researchers used similar criteria for the definition of treatment response and also obtained similar percentages [55]. In fact, in the Miranda et al. study, 97 patients completed the study after a 12-month treatment. Good clinical responders were defined as those who scored≥2 on the MMSE after 12 months of treatment, the neutrals had MMSE scores between –1 and +1 in the same period, and the bad responders scored ≤–2 on the MMSE at the end of 12 months. The definition and the rate of bad responders (37.1%) were similar to those of our non-responders (35.6%). Similarly, the sum of good responders (27.8%) and of the neutral responders (35.1%) totaling 62.9% of patients, was close to our responders (64.4%).

The response to treatment with ChEIs looks good, and this data is consolidated by a long period of observation of four years on average. Also we believe that a likely explanation for the high rate of clinical response found in both studies could be the fact that the patients were kept under careful and strict control and an appropriate orientation to the family caregivers was always provided. This global clinical approach may have contributed to an improved response.

As we know, the studies on the efficacy of treatment with ChEIs have considered slightly the influence of non-pharmacological and contextual factors such as type of previous work done, cognitive profile at baseline, lifestyles, marital status, possible social isolation, the degree of brain atrophy, and the degree of autonomy of the patients at the beginning of therapy.

With ordinary statistics, it is extremely difficult to differentiate responders from non-responders, as exemplified by Table 2 in which no statistically significant difference has been found in any of the variables; this explains why we used a more advanced statistical analysis tool like Auto-CM. Thanks to this approach, our study suggests that a higher cognitive profile, a previous intellectual occupation (white collar), healthier lifestyles such as abstinence from drinking alcohol and smoking, being married, not being widower and not living alone, a higher degree of autonomy, and lower degree of cortical and subcortical brain atrophy at baseline positively affect the response to ChEIs therapy in the long term.

The effects of these non-pharmacological and contextual factors could take place through the maintenance and enhancement of brain and cognitive reserve protecting the substrate on which the drug acts. Brain reserve refers to neuroprotective brain capacity, which can be induced by chronic enhancement of mental and physical activity [56].

Cognitive reserve is a more specific term and encompasses increased cognitive function, and enhanced complex mental activity, as protective factors against dementia and other brain disorders [57].

Synthetically, other authors define brain and cognitive reserve (BCR) as changes occurring in the brain, in response to chronic life experiences, which positively modulate susceptibility to brain disorders and age-dependent dysfunction, via neuroprotective and/or compensatory mechanisms [57]. Molecular mechanisms may underlie the changes observed in animal models of BCR. The main initial effects of increased mental and physical activity in the brain will be to enhance synaptic transmission. In order to mediate long-term changes in brain function, it is expected that signal transduction pathways and epigenetics will induce chronic changes in gene expression and other molecular processes. This in turn may induce cellular changes such as experience-dependent synaptogenesis, which may also mediate BCR [56, 57].

Work complexity, social network, healthy lifestyles, better cognitive profile, low degree of brain atrophy, greater autonomy in activities of daily living and leisure activities may positively contribute to BCR allowing cognitive function to be maintained in the old age [56 –59]. Adult-life occupational work complexity, as well as a mentally and socially integrated lifestyle in late life may realize an environmental enrichment and could postpone the onset of clinical dementia and AD [56 –58].

These non-pharmacological and contextual factors, considered in our work, appear to positively affect the effectiveness of treatment with ChEI in the long term. Considered from the beginning, these factors could allow the identification of those subjects who will better respond to therapy with ChEI, making the best use of available resources and to reduce costs.

Finally, male gender appears to be connected with the responders, while the female gender is associated with non-responders according to what already reported [53]. It should be noted that the imbalances in the distribution of some variables, for example, in gender, do not affect in any way the analysis with this ANN.

An age less than 75 years appear to be connected with the responders, while a more advanced age (between 75 and 85 years) to non-responders; younger age could express better anatomical and functional conditions of the brain, thus promoting better BCR.

We have not evaluated separately the three ChEIs drugs (donepezil, galantamine, and rivastigmine) because no significant difference between the cognitive response according to the ChEI used was previously observed [55].

We have not considered the presence or absence of concomitant diseases either because it is already proven that the absence of concomitant disease is significantly correlated with a good clinical response to ChEI [60] or because we wanted to focus on the contextual factors that we have described.

In the upper left region of Fig. 1, some connections are depicted that are not always easily explainable and referring particularly to over 85 year old patients, who are so frail from having to live in nursing home and present with sphincter incontinence. The biological and pathophysiological mechanisms in this population are poorly understood because often it is not included in clinical trials and therefore has been little studied. Understanding how diseases behave in the very old patients is one of the challenges for the future. The connection of the variable “live with one relative” with the variable “non-responders” is an apparent inconsistency. In fact the relative that resides with the patient often is a worker who spends most of the day working outside the home.

This study has some limitations. Firstly, the cross sectional design does not allow us to assess causality, but only association between variables. Secondly, the sample size is relatively small and this does not exclude that with larger sample size some associations could be different. In the future, we would like to extend the analysis to a larger sample. However, there are also several strengths. Firstly, as we know, there are few other studies that have focused on the effect of non-pharmacological and contextual factors with regard to response to ChEIs treatment. Secondly, the time interval of the survey is especially large and confers robustness to the results. Thirdly, the sample observed is representative of the real world residing at home. Conversely, many past studies were conducted on selected samples, such as in the clinical trials. Fourthly, the study sample is part of TREDEM Registry, which is well characterized by a large amount of data that was in part set out in previous published works [25 –28]. Finally, through intelligent data mining, the Auto-CM ANN identifies specific relationships between each variable and all others, that ordinary statistics has not caught.

CONCLUSION

A growing body of epidemiological evidence supports the idea that environmental enrichment and BCR may delay the onset of a range of brain disorders as well as possibly slow normal brain aging. This study shows that non-pharmacological and contextual factors linkable to environmental enrichment and BCR might affect also the long-term response to ChEI treatment.

These associations are weak from a traditional point of view based on classical statistics. In complex diseases, most variables have weak and loose relationships from mathematical point of view and for this reason depicting the overall underlying schema is very difficult. However, the use of advanced data mining techniques like Auto-CM allows us to understand better how these factors should deserve further attention in epidemiological and pharmacological studies.

Our findings, if confirmed in larger observational studies, will improve the present standard of care of AD allowing a better identification of future responders.

Footnotes

ACKNOWLEDGMENTS

We are grateful to all people who kindly agreed to participate in this study and to Cristina Bastianello for manuscript editing. This study has received support from the Local Health Authority n. 9 (LHA 9) of Treviso.

Authors’ disclosures available online ().

References

Berr

, Wancata

, Ritchie

(2005) Prevalence of dementia in the elderly in Europe. Eur Neuropsychopharmacol 15, 463–471.

Black

, Doody

, Li

, McRae

, Jambor

, Xu

, Sun

, Perdomo

, Richardson

(2007) Donepezil preserves cognition and global function in patients with severe Alzheimer disease. Neurology 69, 459–469.

Erkinjuntti

, Kurz

, Gauthier

, Bullock

, Lilienfeld

, Damaraju

(2002) Efficacy of galantamine in probable vascular dementia and Alzheimer’s disease combined with cerebrovascular disease: A randomised trial. Lancet 359, 1283–1290.

Bullock

, Erkinjuntti

, Lilienfeld

, GAL-INT-6 Study Group (2004) Management of patients with Alzheimer’s disease plus cerebrovascular disease: 12-month treatment with galantamine. Dement Geriatr Cogn Disord 17, 29–34.

Birks

, Grimley

, Iakovidou

(2002) Rivastigmine for Alzheimer’s disease. Cochrane Database Syst Rev, CD001191.

Reisberg

, Doody

, Stöffler

, Schmitt

, Ferris

, Möbius

(2003) Memantine in moderate-to-severe Alzheimer’s disease. N Engl J Med 348, 1333–1341.

Farlow

, Gracon

, Hershey

, Lewis

, Sadowsky

, Dolan-Ureno

, Asher

, Beaver

, Hamilton

, Bergman

, Roger

, Black

, Carr

, Winchester

, Layne

, Clark

, Dexter

, DuBoff

, Hendrie

, Caress

, Shatz

, Hanna

, Brashear

, Damgaard

, Hershey

, Donnelly

, Burch

, Homan

, McSweeney

, Barczak

, Mattes

, Hermann

, Mohr

, Mendis

, Roberts

, Begin

, Sampson

, Ott

, Lannon

, Prendergast

, Madan

, Hanning

, Martinez

, Stone

, Winner

, Zuniga

, Mate

, Ehlert

, Lessard

, Seltzer

, Taylor

, Calabrese

, Harkins

, Weis

, Slade

, Sommer

, Wichter

, Schwartz

, Eastman

, Thein

Jr SG

, Williams

, Dewar

, Foster

, Bluemlein

, Gelb

, Berent

, Giordani

, Baron

, Myers

, Tylor

(1992) A controlled trial of tacrine in Alzheimer’s disease. JAMA 268, 2523–2529.

Rösler

, Bayer

, Anand

, Cicin-Sain

, Gauthier

, Agid

, Dal Bianco

, Stähelin

, Hartman

, Gharabawi

(1999) Efficacy and safety of rivastigmine in patients with Alzheimer’s disease: International randomised controlled trialCommentary: Another piece of the Alzheimer’s jigsaw. BMJ 318, 633–640.

Tariot

, Solomon

, Morris

, Kershaw

, Lilienfeld

, Ding

, Galantamine USA-Study Group (2000) A 5-month, randomized, placebo-controlled trial of galantamine in AD. Neurology 54, 2269–2276.

10.

Rogers

, Doody

, Mohs

, Friedhoff

(1998) Donepezil improves cognition and global function in Alzheimer disease: A 15-week, double-blind, placebo-controlled study. Arch Intern Med 158, 1021–1031.

11.

Thal

, Ferguson

, Mintzer

, Raskin

, Targum

(1999) A 24-week randomized trial of controlled-release physostigmine in patients with Alzheimer’s disease. Neurology 52, 1146–1152.

12.

Rogers

, Farlow

, Doody

, Mohs

, Friedhoff

(1998) A 24-week, double-blind, placebo-controlled trial of donepezil in patients with Alzheimer’s disease. Neurology 50, 136–145.

13.

Qizilbash

, Whitehead

, Higgins

, Wilcock

, Schneider

, Farlow

, Dementia Trialists’ Collaboration (1998) Cholinesterase inhibition for Alzheimer disease: A meta-analysis of the tacrine trials. JAMA 280, 1777–1782.

14.

Winblad

, Engedal

, Soininen

, Verhey

, Waldemar

, Wimo

, Wetterholm

, Zhang

, Haglund

, Subbiah

(2001) Donepezil Nordic Study Group. A 1-year, randomized, placebo-controlled study of donepezil in patients with mild to moderate AD. Neurology 57, 489–495.

15.

Lopez

, Becker

, Wisniewski

, Saxton

, Kaufer

, DeKosky

(2002) Cholinesterase inhibitor treatment alters the natural history of Alzheimer’s disease. J Neurol Neurosurg Psychiatry 72, 310–314.

16.

Knopman

, Schneider

, Davis

, Talwalker

, Smith

, Hoover

, Gracon

(1996) Long-term tacrine (Cognex) treatment effects on nursing home placement and mortality. Neurology 47, 166–177.

17.

Raskind

, Peskind

, Wessel

, Yuan

, Galantamine USA-Study Group (2000) Galantamine in AD A 6-month randomized, placebo-controlled trial with a 6-month extension. Neurology 54, 2261–2268.

18.

Doody

, Geldmacher

, Gordon

, Perdomo

, Pratt

(2001) Open-label, multicenter, phase 3 extension study of the safety and efficacy of donepezil in patients with Alzheimer disease. Arch Neurol 58, 427–433.

19.

Rogers

, Doody

, Pratt

, Ieni

(2000) Long-term efficacy and safety of donepezil in the treatment of Alzheimer’s disease: Final analysis of a US multicentre open-label study. Eur Neuropsychopharmacol 10, 195–203.

20.

Tan

, Yu

, Wang

, Tan

, Meng

, Wang

, Jiang

, Zhu

, Tan

(2014) Efficacy and safety of donepezil, galantamine, rivastigmine, and memantine for the treatment of Alzheimer’s disease: A systematic review and meta-analysis. J Alzheimers Dis 41, 615–631.

21.

Rösler

, Anand

, Cicin-Sain

, Gauthier

, Agid

, Dal-Bianco

, Stähelin

, Hartman

, Gharabawi

(1999) Efficacy and safety of rivastigmine in patients with Alzheimer’s disease: International randomised controlled trial. Commentary: Another piece of the Alzheimer’s jigsaw. BMJ 318, 633–640.

22.

Burns

, Spiegel

, Quarg

(2004) Efficacy of rivastigmine in subjects with moderately severe Alzheimer’s disease. Int J Geriatr Psychiatry 19, 243–249.

23.

Farlow

, Hake

, Messina

, Hartman

, Veach

, Anand

(2001) Response of patients with Alzheimer disease to rivastigmine treatment is predicted by the rate of disease progression. Arch Neurol 58, 417–422.

24.

Farlow

, Small

, Quarg

, Krause

(2005) Efficacy of rivastigmine in Alzheimer’s disease patients with rapid disease progression: Results of a meta-analysis. Dement Geriatr Cogn Disord 20, 192–197.

25.

Gallucci

, Mariotti

, Saraggi

, Stecca

, Oddo

, Bergamelli

, Boldrini

, Mazzucco

, Ongaro

, Mecocci

, Di Paola

, Bendini

, Forloni

, Albani

, Antuono

, Caberlotto

, Zanardo

, Siculi

, Gajo

, Durante

, Buscato

(2012) The Treviso Dementia (TREDEM) Study: A biomedical, neuroradiological, neuropsychological and social investigation of dementia in North-Eastern Italy. J Frailty Aging 1, 24–31.

26.

Gallucci

, Battistella

, Bergamelli

, Spagnolo

, Mazzuco

, Carlini

, Di Giorgi

, Boldrini

, Pilotto

(2014) Multidimensional Prognostic Index (MPI) in an outpatient setting: Mortality and hospitalizations. The Treviso Dementia (TREDEM) Study. J Alzheimers Dis 42, 1461–1468.

27.

Gallucci

, Zanardo

, Bendini

, Di Paola

, Boldrini

, Grossi

(2014) Serum folate, homocysteine, brain Atrophy, and auto-CM system: The Treviso dementia (TREDEM) Study. J Alzheimers Dis 38, 581–587.

28.

Spagnolo

, Aricò

, Bergamelli

, Mazzucco

, Boldrini

, Di Giorgi

, Gallucci

(2015) A pilot study on behavioral and cognitive effects of Cognitive Stimulation Training based on 3R Mental Stimulation. The Treviso Dementia (TREDEM) Study. Neurorehabilitation 36, 151–156.

29.

McKhann

, Drachman

, Folstein

, Katzman

, Price

, Stadlan

(1984) Clinical diagnosis of Alzheimer’s disease Report of the NINCDS-ADRDA Work Group* under the auspices of Department of Health and Human Services Task Force on Alzheimer’s Disease. Neurology 34, 939.

30.

Roman

, Tatemichi

, Erkinjuntti

, Cummings

, Masdeu

, Garcia

, Amaducci

, Orgogozo

, Brun

, Hofman

(1993) Vascular dementia: Diagnostic criteria for research studies. Report of the NINDS-AIREN International Workshop. Neurology 43, 250–260.

31.

Clark

, Sheppard

, Fillenbaum

, Galasko

, Morris

, Koss

, Mohs

, Heyman

, the CERAD Investigators (1999) Variability in annual Mini-Mental State Examination score in patients with probable Alzheimer disease: A clinical perspective of data from the Consortium to Establish a Registry for Alzheimer’s Disease. Arch Neurol 56, 857–862.

32.

Teri

, Hughes

, Larson

(1990) Cognitive deterioration in Alzheimer’s disease: Behavioral and health factors. J Gerontol 45, 58–63 .

33.

Han

, Cole

, Bellavance

, McCusker

, Primeau

(2000) Tracking cognitive decline in Alzheimer’s disease using the mini-mental state examination: A meta-analysis. Int Psychogeriatr 12, 231–247.

34.

Katz

, Ford

, Moskowitz

, Jackson

, Jaffe

(1963) Studies of illness in the aged. The index of ADL: A standardized measure of biological and psychosocial function. J Am Med Assoc 185, 914–919.

35.

Lawton

, Brody

(1969) Assessment of older people: Self maintaining and instrumental activities of daily living. Gerontologist 9, 176–186.

36.

Spector

, Katz

, Murphy

, Fulton

(1987) The hierarchical relationship between activities of daily living and instrumental activities of daily living. J Chron Dis 40, 481–489 (1987).

37.

Folstein

, Folstein

, McHugh

(1975) “Mini-Mental State” A practical method for grading the cognitive state of patients for the clinician. J Psychiatric Res 12, 189–198.

38.

Hughes

, Berg

, Danziger

, Coben

, Martin

(1982) A new clinical scale for the staging of dementia. Br J Psychiatry 140, 566–572.

39.

Morris

(1993) The Clinical Dementia Rating (CDR): Current version and scoring rules. Neurology 43, 2412–2414.

40.

Agresti

(1987) An agreement model with Kappa as parameter. Stat Probab Lett 7, 227–271.

41.

Buscema

, Grossi

(2008) The Semantic Connectivity Map: An adapting self-organizing knowledge discovery method in data bases. Experience in Gastro-oesophageal reflux disease. Int J Data Min Bioinform 2, 362–404 .

42.

Buscema

, Grossi

, Snowdon

, Antuono

(2008) Auto-Contractive Maps: An artificial adaptive system for data mining. An Application to Alzheimer disease. Curr Alzheimer Res 5, 481–498.

43.

Buscema

, Helgason

, Grossi

(2008) Auto contractive maps, H function and maximally regular graph: Theory and applications. Special Session on “Artificial Adaptive Systems in Medicine: Applications in the real world, NAFIPS 2008 (IEEE), New York, May 19–22.

44.

Licastro

, Porcellini

, Chiappelli

, Forti

, Buscema

, Ravaglia

, Grossi

(2010) Multivariable network associated with cognitive decline and dementia. Neurobiol Aging 31, 257–269.

45.

Kruskal

(1956) On the shortest spanning subtree of a graph and the traveling salesman problem. J Am Math Soc 7, 48–50.

46.

Linn

, Linn

, Gurel

(1968) Cumulative Illness Rating Scale. J Am Geriatr Soc 16, 622–626.

47.

Coppedè

, Grossi

, Migheli

, Migliore

(2010) Polymorphisms in folate-metabolizing genes, chromosome damage, and risk of Down syndrome in Italian women: Identification of key factors using artificial neural networks. BMC Med Genomics 3, 42.

48.

Licastro

, Porcellini

, Chiappelli

, Forti

, Buscema

, Ravaglia

, Grossi

(2010) Multivariable network associated with cognitive decline and dementia. Neurobiol Aging 31, 257–269.

49.

Buscema

, Penco

, Grossi

(2012) A novel mathematical approach to define the genes/SNPs conferring risk or protection in sporadic amyotrophic lateral sclerosis based on Auto Contractive Map neural networks and graph theory. Neurol Res Int 2012, 478560.

50.

Buscema

, Grossi

(2008) The semantic connectivity map: An adapting self-organising knowledge discovery method in data bases. Experience in gastro-oesophageal reflux disease. Int J Data Min Bioinform 2, 362–404 .

51.

Street

, Grossi

, Volta

, Faleschini

, Bernasconi

(2008) Placental determinants of fetal growth: Identification of key factors in the insulin-like growth factor and cytokine systems using artificial neural networks. BMC Pediatr 17, 8–24.

52.

Winblad

, Jelic

(2004) Long-term treatment of Alzheimer disease; efficacy and safety of acetylcholinesterase inhibitors. Alzheimer Dis Assoc Disord 18, S2–S8.

53.

Wattmo

(2013) Prediction models for assessing long-term outcome in Alzheimer’s disease: A review. Am J Alzheimers Dis Other Demen 28, 440–449.

54.

Hogan

(2014) Long-term efficacy and toxicity of cholinesterase inhibitors in the treatment of Alzheimer disease. Can J Psychiatry 59(12), 618–23.

55.

Miranda

, Gomes

, Silveira

, Pianetti

, Byrro

, Peles

, Pereira

, Santos

, Assini

, Ribeiro

, Tito

, Matoso

, Lima

, Moraes

, Caramelli

(2015) Predictive factors of clinical response to cholinesterase inhibitors in mild and moderate Alzheimer’s disease and mixed dementia: A one-year naturalistic study. J Alzheimers Dis 45, 609–620.

56.

Valenzuela

(2008) Brain reserve and the prevention of dementia. Curr Opin Psychiatry 21(3), 296–302.

57.

Nithianantharajah

, Hannan

(2009) The neurobiology of brain and cognitive reserve: Mental and physical activity as modulators of brain disorders. Prog Neurobiol 89, 369–382.

58.

Stern

, Habeck

, Moeller

, Scarmeas

, Anderson

, Hilton

, Flynn

, Sackeim

, van Heertum

(2005) Brain networks associated with cognitive reserve in healthy young and old adults. Cereb Cortex 15, 394–402.

59.

Fratiglioni

, Wang

(2007) Brain reserve hypothesis in dementia. J Alzheimers Dis 12, 11–22.

60.

Raschetti

, Maggini

, Sorrentino

, Martini

, Caffari

, Vanacore

(2005) A cohort study of effectiveness of acetylcholinesterase inhibitors in Alzheimer’s disease. Eur J Clin Pharmacol 61, 361–368.