Multidomain Lifestyle Interventions for the Prevention of Cognitive Decline After Ischemic Stroke
Background and Purpose—Cognitive impairment occurs in ≤30% of all stroke survivors. However, effective therapies aimed at preventing poststroke cognitive decline are lacking. We assessed the efficacy of a multidomain intervention on preventing cognitive decline after stroke.
Methods—In this randomized, observer-blind trial patients were recruited within 3 months after an acute stroke in 5 Austrian neurological centers. Patients were assigned to a 24-month lifestyle-based multidomain intervention or standard stroke care. Primary outcomes were the cognitive subscale of the Alzheimer Disease Assessment Scale (ADAS-cog) and occurrence of cognitive decline in the composite scores of at least 2 of 5 cognitive domains at 24 months.
Results—A total of 101 patients were randomized into multi-intervention and 101 into standard care during June 2010 and November 2012. Of them, 76 patients in the intervention group and 83 in the control group were included in the final intention-to-treat analysis. At 24 months, 8 of 76 (10.5%) patients in the intervention group and 10 of 83 (12.0%) patients in the control group showed cognitive decline corresponding to a relative risk reduction of 0.874 (95% confidence interval, 0.364–2.098). The change in ADAS-cog from baseline to 24 months was not different either (median 0 [IQR, −1 to 2] in both groups; P=0.808).
Conclusions—This trial found no benefit of 24-month multidomain intervention with focus on improvement in lifestyle and vascular risk factors on the incidence of poststroke cognitive decline in comparison with standard stroke care. Studies with a larger sample size are needed.
With an estimated total annual cost of €105 billion in Europe in 2010 and an expected increase due to the increasing life expectancy, dementia represents one of the biggest challenges of the century for the economic, social and healthcare system in Europe.1 There is a strong relationship between stroke and dementia. The prevalence data show that approximately ten percent of patients with stroke already have dementia when stroke occurs, another 10% will develop dementia after a first-ever stroke and additional 30% will develop dementia after a recurrent stroke.2 Depending on the criteria used for cognitive impairment—up to 76% of patients with stroke have mild cognitive impairment at three months after an acute stroke.3 Although up to 50% of patients improve cognitively, 30% deteriorate in a delayed fashion between 3 and 15 months poststroke.3 The mechanisms for the delayed onset remain unclear but vascular as well as neurodegenerative mechanisms are involved and stroke seems to accelerate ongoing gradual processes of cognitive decline.
The preservation of cognitive abilities after stroke is increasingly recognized as a crucial target, but no therapeutic strategy has shown convincing clinical evidence of restoring cognitive function or preventing its further decline.4–6
Modifiable risk factors for stroke such as hypertension, dyslipidaemia, diabetes mellitus, smoking, physical inactivity, and poor diet have been associated with an increased risk of cognitive impairment.7–10 Thus, it is plausible that effective secondary prevention strategies in combination with lifestyle-oriented interventions adjusted to individual risk factors can reduce the risk of cognitive decline in stroke survivors. One recent randomized controlled trial showed that multidomain lifestyle intervention was effective in preventing or delaying cognitive decline in people at high dementia risk.11
A meta-analysis of 3 small randomized controlled trials showed that multifactorial lifestyle interventions in patients with stroke can be effective to change lifestyle-related behaviors and physiological outcomes, such as blood pressure.12 However, there is insufficient evidence from randomized controlled trials about the effectiveness of lifestyle intervention on preservation of cognition.4–6,13
The primary objective of this study was to assess the efficacy of an intensive multifactorial intervention focused on adapting lifestyle and improving medication adherence in the prevention of cognitive decline after stroke.
The Austrian Polyintervention Study to Prevent Cognitive Decline after Ischemic Stroke (ASPIS) is a randomized, multicenter, observer-blind trial designed to test whether a 24-month intensive multidomain intervention can prevent poststroke cognitive decline compared with a control group receiving standard care. The setup of the study and baseline data have previously been reported.14
In brief, patients aged 40 to 80 years with a clinical diagnosis of ischemic stroke within the previous 3 months, with a National Institutes of Health Stroke Scale (NIHSS) score of 1 to 14 on admission, a modified Rankin Scale (mRS) score <3 on admission, and sufficient communication ability were recruited at 5 (3 regional and 2 tertiary academic) of the 6 hospitals with stroke units situated in the Austrian federal district Lower-Austria (1.62 million inhabitants). Patients with Mini Mental State Examination <24, pre-existing dementia, Parkinson disease, persistently disturbed level of consciousness, persistent aphasia, significant psychiatric disease, severe sensory impairment, severe comorbidities, and those who were likely to be noncompliant during the follow-up were excluded.
The study was approved by the Ethics committee of Lower-Austria and all patients gave written informed consent.
Randomization and Masking
Patient were randomly assigned (1:1) to the multidomain intervention or to standard care according to a computer-generated randomization list in blocks of 10, stratified by site and by level of education. The person who generated and stored the randomization list centrally was not involved in recruitment or any other activities at local study sites. Information on treatment allocation was given to the recruiting physician in each center in opaque, sealed, and numbered envelopes.
Given the type of intervention patients, neurologists and therapists involved in the intervention, and project coordinators could not be masked to treatment assignment. The primary outcome measure was, however, assessed by neuropsychologists blinded to the study group; patients were instructed not to reveal their study group allocation to the neuropsychologist.
The intervention consisted of intensive management and motivation for compliance with clinical therapy, adequate blood pressure, lipid and glycemic control, healthy diet, regular physical activity, and cognitive training during a period of 24 months. Interventions took place at local study sites and were performed by local nutritionists, physiotherapists, occupational therapists, and neurologists experienced with stroke patients and trained in a standardized way by the study coordinators according to a study manual. Interventions were previously described in detail.14
Study goals were defined as following: perform moderate or vigorous physical activity at least 3× to 5×/week, a body mass index <25 kg/m2 or weight loss and maintenance of at least 5% weight loss during the first year in obese individuals, individually defined dietary goals according to the composition of energy intake and the composition of foods, smoking cessation in patients who smoked, and a blood pressure <140/90 mm Hg (if diabetes mellitus, <130/85 mm Hg) of >75% of self-recorded measurements. Compliance with pharmacological treatment including lipid-lowering drugs, antithrombotic/oral anticoagulation therapy, and glucose-lowering drugs was targeted according to the treatment goals of stroke prevention European Stroke Organization (ESO) guideline.15
Adherence to the study goals were checked by analyzing patient’s diaries, monthly calls from the coordinating center, and regular calls from the local study physicians.
Cognition was assessed at baseline, 12 and 24 months by trained clinical neuropsychologists blinded to group allocation using a neuropsychological test battery assessing 5 cognitive domains (executive functions, working memory, general memory, speed of cognitive processing, and visual-spatial ability; Table I in the online-only Data Supplement).
The first primary outcome was cognitive decline at 24 months after randomization. For each of the 5 cognitive domains, standardized composite scores were calculated from the differences between baseline and 24 months in individual neuropsychological test results using the SD of a control population. Cognitive decline at 24 months after randomization was defined as a statistically significant decrease of function in at least 2 of 5 cognitive domains. The critical level for decision for each domain was chosen by using an overall α level for decision of 0.05 and by calculating the probability that a patient has a cognitive decline in at least 2 of 5 cognitive domains using the binomial distribution. Accordingly, the critical level for a single domain is the 0.076 quantile of the standard normal distribution, that is, a cutoff of −1.43. This criterion was chosen to ensure that cognitive decline was not an artifact of measurements.
To allow comparison with previous stroke trials, change in Alzheimer Disease Assessment Scale-cognitive subscale (ADAS-cog) from baseline to 24 months after randomization was chosen as second primary outcome.
Prespecified secondary outcomes included cognitive decline at 12 months, change in ADAS-cog at 12 months, mean composite scores of each cognitive domain at 12 and 24 months, stroke severity (NIHSS), functional outcome (mRS), activity of daily living (Barthel index), vascular events, all-cause mortality, quality of life (EQ-5D), and depression (Center for Epidemiological Studies Depression Scale [CES-D]) at 12 and 24 months.
Adverse events were recorded by study neurologists during regular phone calls and during study visits (12 and 24 months).
Efficacy analyses were based on the modified intention-to-treat (ITT) principle, using data of all patients who had a valid baseline assessment and at least 1 valid postbaseline assessment for the primary outcome variables (ie, completed the neuropsychological test battery). Missing values were imputed using the last observation carried forward imputation. As last observation carried forward may introduce a bias, we also present primary efficacy analyses for the per protocol (PP) analysis, which included all patients who had a valid baseline assessment and a valid 24-month assessment for the primary outcome variables. Patients with recurrent stroke during the trial were excluded from efficacy analyses.
We estimated a sample size of 158, assuming an incidence of cognitive decline at 24 months of 50% in the control group and a relative risk (RR) reduction of 40%, that is, 30% cognitive decline in the intervention group with a 1-sided significance level of 5% and a power of 80%. The final sample size was estimated at 200, which included adjustment for dropouts and recurrent strokes.
Results for the first primary outcome (cognitive decline) are presented as RR with the corresponding 95% confidence interval, with values of RR of <1 indicating a treatment effect in favor of the intervention group. For the second primary outcome group, differences in changes of ADAS-cog from baseline to 24 months were compared using the Mann–Whitney U test. In addition, a random effects model was performed with differences in ADAS-cog between baseline and 24 months as dependent variable and center as random factor. ANOVAs were calculated to compare different models including combinations of the following independent variables, such as group, sex, age, education, time from stroke to baseline testing, lifestyle factors, Mini Mental State Examination, depressive symptoms, NIHSS, and ADAS-cog at baseline. Secondary outcome variables were analyzed using the Mann–Whitney U test.
Analyses were performed with R version 22.214.171.124
Between June 2010 and November 2012, 101 patients were randomized into the multi-intervention and 101 into the standard care group. Three patients did not meet the inclusion criteria and 9 did not complete the baseline neuropsychological tests. Of the remaining 190 participants, 166 (87%) completed the 12-month cognitive test battery and 154 (81%) completed the 24-month cognitive test battery. Overall, 17 of 94 in the intervention group compared with 13 of 96 in the control group dropped out for nonmedical reasons (Figure). Seven patients who suffered recurrent stroke and 1 patient who was unable to complete the entire test battery at baseline were excluded from the data analysis. Thus, data of 159 (76 intervention and 83 control) persons entered the ITT analyses, whereas 146 (72 intervention and 74 control) entered the PP analyses (Figure).
The baseline characteristics have been presented previously14 and were similar for the control and intervention group in the ITT population (Table II in the online-only Data Supplement), except for body mass index, which was higher in the intervention group.
At 24 months, 8 of 76 (10.5%) in the intervention group and 10 of 83 (12.0%) in the control group had cognitive decline corresponding to a RR reduction of 0.874 (95% confidence interval, 0.364–2.098; Table 1) for the intervention group compared with the control group according to ITT. In the PP population, the corresponding numbers were 7 of 72 (9.7%) for the intervention group and 7 of 74 (9.5%) for the control group (RR, 1.028; 95% confidence interval, 0.380–2.783; Table 1). Of the 7 participants excluded from analysis because of recurrent stroke, 3 had a cognitive decline at 24 months.
During the 24 months, participants in the intervention group and in the control group showed no improvement on the ADAS-cog (median, 0 [−1 to 2] points for both groups; ITT analysis; group difference, P=0.808, Mann–Whitney U test; Table 2). Similarly, for the PP analysis the groups did not differ significantly with a median improvement of 0 (−1 to 2) in the intervention group and of 1 (−1 to 2) in the control group (group difference, P=0.834). In the random effects model, also no group difference was found (Table III in the online-only Data Supplement).
Cognitive decline was mainly found in the cognitive domains speed of mental processing and general memory (Table 1). Of those 18 persons with cognitive decline (≥2 domains affected), 13 (72%) had a significant decline in speed of mental processing, 12 (67%) in executive functions and 11 (61%) in general memory, whereas the domains visual-spatial and working memory were affected in 4 (22%) persons each.
None of the secondary outcome variables (cognitive decline at one year, cognitive domain summary scores, depressive symptoms and quality of life) differed significantly between groups (Table 2). Two participants, both in the control group, died. Five participants in the intervention group and 4 in the control group suffered a vascular event (7 recurrent strokes, 1 transient ischemic attack, and 1 myocardial infarction).
Adverse events were reported by 37 of 94 (40%) patients in the intervention group and by 29 of 96 (30%) in the control group. Of these, 16 of 94 (17%) patients in the intervention group and 17 of 96 (18%) in the control group reported serious adverse events (Table IV in the online-only Data Supplement).
The 72 patients randomized into the intervention group and compliant at the 24-month visit attended in median 5 (IQR, 3–6) of the 7 recommended individual dietary counselings, 7 (3–9) of the 7 recommended dietary group meetings, 8 (4–11) of the 9 recommended physical activity group meetings, and 12 (5.5–17) of the optional monthly cognitive group meetings. For the ITT population, the proportion of patients meeting the lifestyle goals after 24 months had increased by 12.0% in the intervention group in respect to physical activity (≥90 minutes/wk) compared with 11.0% in the control group, by 10.5% versus 1.2% for body mass index (<30 kg/m2), and by 2.6% versus 7.2% for currently not smoking. The proportion of patients meeting the blood pressure goals (<140 mm Hg systolic and <90 mm Hg diastolic) had increased by 9.2% in the intervention group compared with a decrease by 2.4% in the control group. These changes were only significant for physical activity in the intervention group (P=0.035, McNemar test) but not in the control group (P=0.093); changes in body mass index were marginally significant in the intervention group (P=0.057).
In this trial, no beneficial effect of an intensive -based multidomain intervention on cognitive decline could be detected in patients with acute stroke compared with standard stroke care. The trial was also neutral for all secondary outcomes. A similar neutral finding was shown in a previous trial, where interventions to control vascular risk factors did not lead to better cognitive performance in neuropsychological tests in comparison with the control group.17 But compared with our trial, fewer interventions were used and follow-up was restricted to 12 months. To date, no other trial was designed to test the effect of lifestyle interventions on poststroke cognition.4,13 In the Finnish Geriatric Intervention Study to Prevent Cognitive Impairment and Disability (FINGER), comparable interventions during the same period have been shown effective in preventing cognitive decline in a cohort of people at increased risk of dementia.11 It is unclear whether this is because of the larger power of the FINGER study, or whether prevention measures might be less effective once stroke becomes manifest. However, our study is the first to address the question of multiple interventions to prevent poststroke cognitive decline in a comprehensive way.
The study was randomized and a loss to follow-up of 19% is not uncommon for lifestyle intervention studies. It is unlikely that dropouts may have biased our results because the baseline characteristics of both the groups were similar and results of the ITT analysis were comparable with the PP analysis.
Several reasons can be accounted for this neutral result. It is possible that our study was too underpowered to detect a small difference between the randomized groups because the overall incidence of cognitive decline of 10% in our study was lower than expected in our sample size estimation. The low rate of cognitive decline in our population may be because of our relative strict inclusion criteria and the rigorous definition of cognitive decline (a statistically significant cognitive decline in ≥2 cognitive domains). However, even at the level of single cognitive domain no group differences were detected. It is also possible, that a longer follow-up period than 24 months would have led to a higher incidence of cognitive decline and to the detection of a significant group difference. Furthermore, in our sample size estimation, we assumed a large effect of interventions on cognition. In the absence of intervention trials investigating cognition in patients with stroke at the time of study planning, the estimation was based on studies, such as the Systolic Hypertension in Europe (Syst-Eur),18 which found that after a median follow-up period of 2 years antihypertensive therapy reduced the risk of dementia by 55% compared with the controls. The choice of a minimum clinically relevant RR reduction might have been better suited. Nevertheless, the ASPIS provided useful information for planning new studies and their power calculations and future studies will have to use larger samples and perhaps longer periods of observation.
Our data are representative for an Austrian population having mild strokes. Cognitive performance was highly variable among participants. As cognition at baseline was tested at a median of 19 days after the onset of incident stroke, it is possible that early spontaneous improvement of cognitive dysfunctions may have masked any decline in cognitive function during the later time course, thus reducing the probability to detect cognitive decline. However, the timing of baseline neuropsychological examinations did not differ between the treatment groups. Patients included in the study had mild strokes, which might have led to a ceiling effect or to a delay in cognitive decline compared with patients with more severe strokes. Further studies might also include patients with more severe strokes who might be at higher risk of cognitive decline and assess the benefit from such interventions according to initial assessment of the severity of the index stroke event as well as the stroke subtype. Unfortunately, there is insufficient evidence for an early identification of those being at risk of developing poststroke cognitive impairment and profiting most from interventions.
It is noteworthy that the decline in executive function domain was found less often in the intervention group, whereas its frequency doubled in the controls. Although not significant, this difference might be seen as a result of the multifactorial interventions as problems in executive functions are the main consequences of poststroke cognitive deterioration. In the FINGER study, significant group difference in cognitive scores were found for executive functions and processing speed but not for memory.11
We excluded people with recurrent strokes from our analysis according to prespecified criteria to eliminate the confounding effects of further strokes on cognition in the group comparison. Recurrent stroke is known to double the risk of poststroke dementia, and the occurrence of stroke itself seems to accelerate cognitive decline.2 In our study, patients with recurrent stroke had a 4-fold higher incidence of cognitive decline (3/7; 43%) than patients without recurrent stroke (18/159; 11%).
Overall, the interventions were well accepted by patients with stroke and adherence to interventional meetings was good. Dropouts for nonmedical reasons were more frequent in the intervention group in the first year. But once participants became accustomed to the interventions commitment to the study was high, and dropouts were more frequent in the control group in the second year. The lack of blinding of patients may have decreased the adherence to follow-up in the control group because of the fact that only standard treatment was offered to this group.
Analysis of the lifestyle data of our study showed that favorable lifestyle changes were more prominent in the intervention group than in the control group (data not presented in detail here). However, these improvements in the intervention group were possibly not pronounced enough to protect cognition in poststroke patients. In general, patients recruited in secondary prevention trials may be more ready to adhere to lifestyle interventions and improvements in standard stroke care may have further reduced group differences and diluted the effects of our interventions.
Twenty-four months of intensive multifactorial interventions with focus on improvement in lifestyle and vascular risk factors did not reduce the incidence of poststroke cognitive decline in comparison with standard treatment. Further studies with larger sample size are needed.
Participants in the ASPIS Study Group in addition to the authors named: Peter Schnider(Department of Neurology, Landesklinikum Wr. Neustadt, Austria); Christian Bancher, Michaela Pinter (Department of Neurology Landesklinikum Horn/Allentsteig, Austria); Berthold Kepplinger, Susanne Asenbaum-Nan (Landesklinikum Amstetten/Mauer, Austria); Stefan Oberndorfer (University Hospital St. Pölten, Austria); Wolf-Dieter Heiss (Department for Clinical Neuroscience and Preventive Medicine, Danube University Krems, Austria).
We thank the motivated and motivating neurologists, neuropsychologists, occupational therapists, physiotherapists, nutritionists in the participating centers as well as the likewise motivated patients.
Sources of Funding
The Austrian Polyintervention Study to Prevent Cognitive Decline after Ischemic Stroke trial is supported by a Grant from the NÖ Forschungs- und Bildungsges.m.b.H. (former Life Science Krems GmbH), Austria with the Grant Agreement Number LS09-002.
*The contributors of the ASPIS Study Group are listed in the Appendix.
The online-only Data Supplement is available with this article at http://stroke.ahajournals.org/lookup/suppl/doi:10.1161/STROKEAHA.115.009992/-/DC1.
- Received May 12, 2015.
- Revision received August 12, 2015.
- Accepted August 13, 2015.
- © 2015 American Heart Association, Inc.
- Ballard C,
- Rowan E,
- Stephens S,
- Kalaria R,
- Kenny RA
- Dichgans M,
- Zietemann V
- Sahathevan R,
- Brodtmann A,
- Donnan GA
- Ngandu T,
- Lehtisalo J,
- Solomon A,
- Levälahti E,
- Ahtiluoto S,
- Antikainen R,
- et al
- Lawrence M,
- Kerr S,
- McVey C,
- Godwin J
- Brainin M,
- Matz K,
- Nemec M,
- Teuschl Y,
- Dachenhausen A,
- Asenbaum-Nan S,
- et al
- 16.↵R Development Core Team. A Language and Environment for Statistical Computing. Vienna, Austria: R Foundation for Statistical Computing; 2012. http://cran.r-project.org/ Accessed May 6, 2015.
- Ihle-Hansen H,
- Thommessen B,
- Fagerland MW,
- Øksengård AR,
- Wyller TB,
- Engedal K,
- et al
- Forette F,
- Seux ML,
- Staessen JA,
- Thijs L,
- Babarskiene MR,
- Babeanu S,
- et al