Methods
Setting
The current study was embedded within NILVAD, a randomised controlled trial of the antihypertensive
Nilvadipine in mild-to-moderate Alzheimer Disease (AD). The trial’s primary outcome of a therapeutic effect of
Nilvadipine in mild/ moderate AD was not achieved. Briefly, participants were recruited from 23 centres in 9 European countries (
Clincaltrials.gov NCT02017340; EudraCT number 2012–002764-27). The full trial protocol, the sub-study protocol in addition to the main trial results have been published elsewhere [
28‐
30]. Ethical approval for the trial was granted from the appropriate National Competent Authorities, Independent Ethics Committees and Institutional Review Boards for all study sites.
Participants once enrolled were randomized to either Nilvadapine 8 mg or placebo for the 18 month duration of the study.
Inclusion/exclusion criteria
The inclusion/exclusion criteria for the NILVAD trial have been previously published [
28]. Of note, included participants were those aged > 50 years who met the National Institute of Neurological and Communicative Disorders and Stroke/Alzheimer’s Disease Criteria: NINCDS-ADRDA for AD. Criteria for inclusion was a diagnosis of mild-to-moderate AD, defined as a Mini-Mental State (MMSE) Score between 12 and 26 at trial enrolment and a diagnosis of AD as per the NINCDS-ADRDA criteria. Further, those included had a Mini-Mental State (MMSE) Score between 12 and 26 at trial enrolment, meeting criteria for mild to moderate AD.
Relevant exclusion criteria to the current analysis include a comorbid neurological condition such as Parkinson’s Disease or Huntington’s Disease. Similarly, participants with a history of significant head trauma, known structural brain abnormalities or any other condition known to interfere with cognitive function were excluded. Patients meeting criteria for a substance use disorder were also excluded as were participants with significant cardiovascular disease. For full exclusion criteria, readers are directed to the initial trial protocol.
Gait measurement
Gait speed was measured as part of the NILVAD frailty sub-study, the protocol of which has been previously published [
28]. Briefly, gait speed was measured both at baseline and 18 months. Four markings were placed on an even floor with adhesive tape over a distance of 6 m total. The markings were spaced as follows (i) first mark at start, (ii) second mark at 1 m, (iii) third mark 4 m from mark 2 and (iv) fourth mark 1 m from mark 3. Participants were asked to stand at mark 1 and then instructed to walk from mark 1 to mark 4 at normal walking speed. The stopwatch was started once the participant’s first foot touched/crossed mark two and was stopped once the participant’s last foot crossed mark 3. A stopwatch accurate to two decimal places was used. Thus the total speed represented the time for the participant to walk 4 m. If the total time was > 6 s to complete this 4 m walk (< 0.67 ms
− 1), gait speed was recorded as slow gait speed. This pre-specified binary cut-off was applied as per previous literature and included in the initial sub-study protocol [
23,
28].
Baseline slow gait speed was considered as those below this cut-off at trial initiation and incident slow gait speed defined as those below this cut off at 18 months who did not meet the slow great criteria at baseline.
Cognitive function and dementia severity assessment
The Alzheimer’s Disease Assessment Scale, Cognitive Subsection (ADAS-Cog) was used to assess cognition and was the primary cognitive endpoint in the NILVAD trial. For the current study, we considered ADAS-Cog at initial visit to be baseline ADAS-Cog and analysed the change in ADAS-Cog at 18 months as the main cognitive outcome. ADAS-Cog sub-scores analysed included the following: word recall task, naming, commands, constructional praxis, delated word recall, ideational praxis, orientation, word recognition task, spoken language ability, comprehension, word finding difficult in spontaneous speech, remembering test instructions.
Dementia severity at baseline and at 18-months was assessed using the Clinical Dementia Rating scale sum of boxes (CDR-Sb) score.
Medical history and regular medications
A comprehensive medical history was taken from participants/carers at time of study enrolment and updated on follow up visits. A list of concomitant medications was also obtained at time of study enrolment and reviewed at follow up visits in order to identify any changes in concomitant medication usage. Medications were classified according to Anatomic Therapeutic Classification (ATC) codes to ensure consistency. For the current analysis, only medications taken for the entire 18-month study duration were considered. Short term and historical medication use were excluded.
Statistical analysis
STATA V15.0 (Stata Corp, College Station, Texas, USA) was used for all analysis in the current study, with the significance level considered as p < 0.05. Baseline descriptive statistics are reported as means with standard deviations where parametric and medians with Interquartile Ranges (IQR) where appropriate. Univariate analysis of between group differences employed T-tests, Wilcoxon rank-sum tests and Chi-square tests. Logistic regression was used to assess predictors of slow gait speed at baseline and adjusted Odds Ratios (OR), 95% Confidence Intervals (CI) and p-values presented as appropriate. Change in ADAS-Cog Scores was considered the difference between the baseline ADAS-Cog score and ADAS-Cog score at 18 months, with an increasing score indicating greater cognitive decline.
In the current analysis, we sought to assess the longitudinal relationship between slow gait speed and cognitive function in older adults with mild-to-moderate Alzheimer Disease. Our research questions were as follows: (i) is there an association at baseline between slow gait speed and baseline cognitive function?, (ii) does slow gait speed at baseline predict greater cognitive decline at 18 months in mild-to-moderate AD?, (iii) does cognitive function at baseline predict the development of slow gait speed at 18 months (incident slow gait speed)?, and finally, (iv) does slow gait speed predict falls in those with AD over an 18-month period.
To assess the cross-sectional relationship at baseline between cognitive function (independent variable, linear) and slow gait speed (dependent variable; binary), we used multivariate logistic regression adjusting for important confounders known to influence gait speed and/or cognition. We adjusted for age, gender, Body Mass Index, study group (Nilvadipine vs.placebo), diagnosis duration, symptom duration, years of formal education, total number of medications, antidepressant and benzodiazepine use, total medical comorbidities, diabetes, hypertension, arthritis, baseline ADAS-Cog and baseline CDR-Sb.
In order to assess whether baseline slow gait speed (independent variable; categorical) was associated with cognitive outcomes at 18 months (dependent variable; linear), mixed effects linear regression was used with country as a random effect. The association was tested unadjusted in the first instance (model 1). The association was then tested controlling for important demographic and AD-related covariates including baseline cognitive function (ADAS-Cog Score), age, gender, study arm, body mass index, years of formal education and diagnosis duration (model 2). Finally we incorporated total number of medical comorbidities and total number of medications (model 3).
For assessment of whether baseline cognition (independent variable; linear) was associated with incident slow gait speed (dependent variable; categorical), logistic regression was used Again the association was tested unadjusted followed by adjustment for baseline ADAS-Cog, age, gender, study arm, body mass index, years of formal education, diagnosis duration (model 2) followed by robust adjustment (model 3) for total number of medical conditions, diabetes mellitus, arthritis, total number of medications, antidepressant use and benzodiazepine use based on known impact on gait speed.
Finally, in order to assess the relationship between slow gait speed (independent variable; categorical) and falls (dependent variable; count), a Poisson regression was used, again unadjusted in the first instance, followed by adjustment for baseline ADAS-Cog, age, gender, study arm, body mass index, years of formal education, diagnosis duration (model 2) in addition to total medical comorbidities and total number of medications (model 3).
Analyses were then repeated by ADAS-Cog sub-scores for each task, with the number of errors on each task analysed by Poisson regression using the same covariates the above models. Finally, we repeated the above models using the CDR-Sb as the dependent variable in order to assess the relationship between slow gait speed and dementia severity at baseline at 18 months.
In all above analyses, the following data were categorical: gender, group (Nilvadipine vs. placebo), slow gait speed (present vs absent), antidepressant use (user vs non-user), benzodiazepine use (user vs non-user), diabetes (present vs absent), arthritis (present vs absent), hypertension (present vs absent). All other variables were continuous as detailed above.
Discussion
In this study of older adults mild-to-moderate AD, greater cognitive impairment at baseline was significantly associated with both baseline slow gait speed and incident slow gait speed at 18 months. There was no association between baseline slow gait speed and longitudinal cognitive performance at 18 months. This is despite a clinically meaningful cognitive decline experienced by the cohort overall (with the mean ADAS-Cog score increasing by a mean of 8.87 points) [32]. Further, slow gait speed at baseline was associated with a more than threefold increased risk of falls over an 18 month period in those with mild to moderate AD. This is one of the first studies in the literature to assess the longitudinal relationship between slow gait speed and cognitive function in those with a diagnosis of AD
At baseline, the only significant predictors of slow gait speed were greater age and poorer cognition in multivariate analysis. This finding is noteworthy and is largely consistent with previous studies in the literature which have demonstrated a relationship between cognitive function and gait speed in a cross-sectional manner in those with AD [
17‐
19].
The lack of association between baseline slow gait speed and cognitive outcomes is of interest. The only previous study in mild to moderate dementia to assess gait speed and cognition in a longitudinal fashion similarly found no association in longitudinal analysis for general cognition, but found a positive association for particular cognitive sub-domains, namely executive function [
21]. One of the reasons for the lack of association in the current study may be that gait was considered as a dichotomous variable based on a pre-specified cut-off. It may be the case that more subtle abnormalities in gait speed may correlate with cognitive function and that such a stringent cut off may mask such an association. Nevertheless, it is possible that whilst slow gait speed may be a predictor of a later diagnosis of dementia, it may lose its predictive value once a diagnosis of dementia has been established.
One of the most interesting findings from the current analysis is the relationship between baseline cognitive function and incident slow gait speed over the duration of the study. Even in a fully adjusted model, baseline cognition was the strongest (apart from age) predictor of incident slow gait speed. This is particularly interesting given that 14% of the participants with normal gait speed at baseline converted to slow gait speed (incident slow gait speed). Thus, this may represent a subpopulation within those with mild-to-moderate AD who may be particularly vulnerable to adverse consequences.
The association between baseline slow gait speed and incident falls is particularly stark. Baseline slow gait speed was associated with a greater than threefold increased risk in incident falls over the 18 month study period. Indeed, baseline slow gait speed was the strongest predictor of incident falls under a fully adjusted model. This is particularly pertinent given that more than 15% of the participants with mild to moderate AD experienced at least one fall over the study period. Targeted interventions for falls prevention focussed on those with AD and slow gait speed are warranted and slow gait speed may be a way to select those with mild to moderate AD who are particularly vulnerable to experiencing falls, given the adverse consequences of falls in such a cohort [
23].
An important consideration in the present study is the nature of the included cohort. Notably, patients with other neurological disorders such as Parkinson’s Disease were excluded. Further, patients with a significant cardiac history, previous stroke etc. were all excluded. Thus, the cohort of included patients had fewer comorbidities which may impair cognition than in other naturalistic studies. This enhances the current study in assessing the relationship between gait and cognition specifically in mild-to-moderate AD.
Other strengths of this study include its international nature as well as the high fidelity of follow up in included patients and the large amount of clinical and medical history data available. This enabled us to control for other causes of gait disorders such as arthritis and diabetes in our analysis. Further, it enabled us to include the effects of certain medications on incident slow gait speed and falls, such as benzodiazepines, which have been previously linked with falls in this vulnerable population [
32].
Our study has several limitations. Principal amongst these is the fact that gait was assessed as a binary variable. Applying such a stringent cut-off in a dichotomous fashion may not produce the same results as analysis repeated with gait speed as a continuous variable in meters per second for instance. However, in the current study we were limited by the fact that gait speed was recorded as an individual binary variable and not a specific speed in metres per second. This may introduce less accuracy into our results than more detailed recording of gait speeds. Future longitudinal studies should examine the intricacies of the gait-cognition relationship in those with a diagnosis of dementia. A further limitation of the current study is that gait and cognition was only considered over an 18 month period. Such a period may be too short to demonstrate the relationship between gait speed and cognition. Previous studies in the literature have demonstrated a relationship over a longer period of time and we cannot rule out that following this cohort for a longer period of time may demonstrate a significant relationship between baseline gait speed and cognition at follow-up.
In conclusion, we demonstrate a significant baseline association between cognition and slow gait speed in a large cohort of participants with mild to moderate Alzheimer Disease. Poorer baseline cognition was associated with incident slow gait speed over 18 months. However, there was no association between baseline slow gait speed and cognition at follow-up. We also report a significant increase in falls risk over an 18 month period in those with baseline slow gait speed. Further studies should continue to explore the nature of the gait-cognition relationship specifically in those with Alzheimer Disease in helping to tease out which patients may be particularly vulnerable to further cognitive decline and adverse health outcomes.
Acknowledgements
NILVAD Study Group.
Brian Lawlor, Mercer’s Institute for Research on Ageing, St. James’s Hospital and Department of Medical Gerontology, Trinity College, Dublin, Ireland; Ricardo Segurado, CSTAR and School of Public Health, Physiotherapy and Sport Science, University College Dublin (UCD), Dublin, Ireland; Sean Kennelly, Department of Age Related Healthcare, Tallaght Hospital, Dublin 24 and Department of Medical Gerontology, Trinity College Dublin; Marcel G. M. Olde Rikkert, Department of Geriatric Medicine, Radboud umc Alzheimer Center, Donders Institute of Medical Neurosciences, Radboudumc, Nijmegen, the Netherlands; Robert Howard, Division of Psychiatry, University College London and King’s College London; Florence Pasquier, CHU Lille, Univ. Lille, DISTALZ Laboratory of Excellence, F-59000 Lille, France; Anne Bo¨rjesson-Hanson, Department of Psychiatry and Neurochemistry, Institute of Neuroscience and Physiology, Sahlgrenska Academy, University of Gothenburg; Magda Tsolaki, Papanikolaou General Hospital of Thessaloniki, Greece; Ugo Lucca, Laboratory of Geriatric Neuropsychiatry, IRCCS Istituto di Ricerche Farmacologiche Mario Negri, Milano, Italy; D. William Molloy, University College Cork Centre for Gerontology and Rehabilitation, Cork, Ireland; Robert Coen, Mercer’s Institute for Research on Ageing, St. James’s Hospital, Dublin, Ireland; Matthias W. Riepe, Department of Geriatrics and Old Age Psychiatry, Psychiatry II, Ulm University at BKH Gu¨nzburg, Germany; Ja’nos Ka’lma’n, Department of Psychiatry, University of Szeged, Hungary; Rose Anne Kenny, Department of Medical Gerontology, Trinity College Dublin (TCD), Dublin, Ireland; Fiona Cregg, Department of Medical Gerontology, Trinity College Dublin (TCD), Dublin, Ireland; Sarah O’Dwyer, Mercer’s Institute for Research on Ageing, St. James’s Hospital, Dublin, Ireland; Cathal Walsh, Health Research Institute and MACSI, Department of Mathematics and Statistics, University of Limerick, Ireland; Jessica Adams, Department of Old Age Psychiatry, King’s College London; Rita Banzi, Laboratory of Geriatric Neuropsychiatry, IRCCS Istituto di Ricerche Farmacologiche Mario Negri, Milano, Italy; Laetitia Breuilh, CHU Lille, Univ. Lille, DISTALZ Laboratory of Excellence, F-59000 Lille, France; Leslie Daly, CSTAR and School of Public Health, Physiotherapy and Sport Science, University College Dublin (UCD), Dublin, Ireland; Suzanne Hendrix, Pentara Corporation, 2180 Claybourne Avenue, Salt Lake City, Utah; Paul Aisen, Department of Neurology, University of Southern California; Siobhan Gaynor, Molecular Medicine Ireland (MMI), Dublin, Ireland; Ali Sheikhi, Health Research Institute and MACSI, Department of Mathematics and Statistics, University of Limerick, Ireland; Diana G. Taekema, Department of Geriatric Medicine, Rijnstate Hospital, Arnhem, the Netherlands; Frans R. Verhey, Department of Psychiatry and Neuropsychology, School of Mental Health and Neuroscience, Alzheimer Center Limburg, Maastricht University, Maastricht, the Netherlands; Raffaello Nemni, IRCCS Don Gnocchi Foundation-University of Milan, Italy; Flavio Nobili, Dept. of Neuroscience (DINOGMI), University of Genoa, and IRCCS AOU Polyclinic, Hospital San Martino, Genoa, Italy; Massimo Franceschi, Neurology Department, Multimedica, Castellanza, Italy; Giovanni Frisoni, Centro San Giovanni di Dio—IRCCS Fatebenefratelli, Brescia, Italy; Orazio Zanetti, Centro San Giovanni di Dio—IRCCS Fatebenefratelli, Brescia, Italy; Anastasia Konsta, Aristotle University of Thessaloniki (AUTH), First Psychiatric Department, Papageorgiou General Hospital, Greece; Orologas Anastasios, Ahepa University General Hospital of Thessaloniki, Greece; Styliani Nenopoulou, Papanikolaou General Hospital of Thessaloniki, Greece; Fani Tsolaki-Tagaraki, Papanikolaou General Hospital of Thessaloniki, Greece; Magdolna Pakaski, Department of Psychiatry, University of Szeged, Hungary; Olivier Dereeper, Centre Hospitalier de Calais, France; Vincent de la Sayette, Centre Hospitalier Universitaire de Caen, France; Olivier Se′ne’chal, Centre Hospitalier de Lens, France; Isabelle Lavenu, Centre Hospitalier de Be′thune, France; Agnès Devendeville, Centre Hospitalier Universitaire d’Amiens, France; Gauthier Calais, Groupement des Hoˆpitaux de l’Institut Catholique de Lille (GHICL), France; Fiona Crawford, Archer Pharmaceuticals, Sarasota, Florida, and Roskamp Institute, Sarasota, Florida; Michael Mullan, Archer Pharmaceuticals, Sarasota, Florida, and Roskamp Institute, Sarasota, Florida, Pauline Aalten, PhD, Department of Psychiatry and Neuropsychology, School of Mental Health and Neurosciences, Alzheimer Center Limburg, Maastricht University, Maastricht, the Netherlands; Maria A. Berglund, RN, Sahlgrenska University Hospital, Gothenburg, Sweden; Jurgen A. Claassen MD, PhD, Department of Geriatric Medicine, Radboudumc Alzheimer Center, Donders Institute of Medical Neurosciences, Radboudumc, Nijmegen, the Netherlands; Rianne A. De Heus, MSc, Department of Geriatric Medicine, Radboudumc Alzheimer Center, Donders Institute of Medical Neurosciences, Radboudumc, Nijmegen, the Netherlands; Daan L. K. De Jong, MSc, Department of Geriatric Medicine, Radboudumc Alzheimer Center, Donders Institute of Medical Neurosciences, Radboudumc, Nijmegen, the Netherlands; Olivier Godefroy, MD, PhD, Centre Hospitalier Universitaire d’Amiens, France; Siobhan Hutchinson, MD, St. James’s Hospital, Dublin, Ireland; Aikaterini Ioannou, MD, 1st Department of Neurology, Ahepa University General Hospital, Aristotle University of Thessaloniki, Greece; Michael Jonsson, MD, PhD, Department of Psychiatry and Neurochemistry, Institute of Neuroscience and Physiology, Sahlgrenska Academy, University of Gothenburg, Sweden; Annette Kent, PhD, Trinity College Dublin (TCD), Dublin, Ireland; Ju¨rgen Kern MD, PhD, Department of Psychiatry and Neurochemistry, Institute of Neuroscience and Physiology, Sahlgrenska Academy, University of Gothenburg, Sweden; Petros Nemtsas MD, PhD, 1st Department of Neurology, Ahepa University General Hospital, Aristotle University of Thessaloniki, Greece; Minoa-Kalliopi Panidou, BSc, MA, 1st Department of Neurology, Ahepa University General Hospital, Aristotle University of Thessaloniki, Greece; Laila Abdullah, PhD, Roskamp Institute, Sarasota, Florida; Daniel Paris, PhD, Roskamp Institute, Sarasota, Florida; Angelina M. Santoso, MD, MSc, Department of Geriatric Medicine, Radboudumc Alzheimer Center, Donders Institute of Medical Neurosciences, Radboudumc, Nijmegen, the Netherlands; Gerrita J. van Spijker, MSc, Department of Geriatric Medicine, Radboudumc Alzheimer Center, Donders Institute of Medical Neurosciences, Radboudumc, Nijmegen, the Netherlands; Martha Spiliotou MD, PhD, 1st Department of Neurology, Ahepa University General Hospital, Aristotle University of Thessaloniki, Greece; Georgia Thomoglou, BSc, 1st Department of Neurology, Ahepa University General Hospital, Aristotle University of Thessaloniki, Greece; and Anders Wallin, MD, Department of Psychiatry and Neurochemistry, Institute of Neuroscience and Physiology, Sahlgrenska Academy, University of Gothenburg, Sweden.