Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Comparison of multiple interventions for older adults with Alzheimer disease or mild cognitive impairment

Comparison of multiple interventions for older adults with Alzheimer disease or mild cognitive... Background: The increasing prevalence of Alzheimer disease (AD) emphasizes the need for effective treatments. Both pharmacological therapies such as nutrition therapy (NT) and nonpharmacologic therapies including traditional treatment or personalized treatment (e.g., physical exercise, music therapy, computerized cognitive training) have been approved for the treatment of AD or mild cognitive impairment (MCI) in numerous areas. Methods: The aim of this study was to compare 4 types of interventions, physical exercise (PE), music therapy (MT), computerized cognitive training (CCT), and NT, in older adults with mild to moderate AD or MCI and identify the most effective intervention for their cognitive function. We used a system of search strategies to identify relevant studies and include randomized controlled trials (RCTs), placebo-controlled trials evaluating the efficacy and safety of 4 interventions in patients with AD or MCI. We updated the relevant studies which were published before March 2017 as a full-text article. Using Bayesian network meta-analysis (NMA), we ranked cognitive ability based objectively on Mini-Mental State Examination (MMSE), and assessed neuropsychiatric symptoms based on Neuropsychiatric Inventory (NPI). Pairwise and network meta-analyses were sequentially performed for efficacy and safety of intervention compared to control group through RCTs included. Results: We included 17 RCTs. Fifteen trials (n=1747) were pooled for cognition and no obvious heterogeneity was found (I = 21.7%, P=.212) in NMA, the mean difference (MD) of PE (MD=2.1, confidence interval [CI]: 0.44–3.8) revealed that PE was significantly efficacious in the treatment group in terms of MMSE. Five trials (n=660) assessed neuropsychiatric symptoms with an obvious heterogeneity (I =61.6%, P=.034), the MD of CCT (MD=7.7, CI: 14 to 2.4), revealing that CCT was significantly efficacious in NPI. Conclusions: As the first NMA comparing different interventions for AD and MCI, our study suggests that PE and CCT might have a significant improvement in cognition and neuropsychiatric symptoms respectively. Moreover, nonpharmacological therapies might be better than pharmacological therapies. Abbreviations: AD = Alzheimer disease, ADAS-Cog = Alzheimer disease Assessment Scale, cognitive subscale, ADL = activities of daily living, AEs = adverse events, CCT = computerized cognitive training, CDR = Clinical Dementia Rating, CG = control group, MCI = mild cognition impairment, MD = mean difference, MMSE = Mini-Mental State Examination, MT = music therapy, NMA = network meta-analysis, NPI = Neuropsychiatric Inventory, NT = nutrition therapy, PE = physical exercise, RCT = randomized controlled trial, SD = standard deviation, SUCRA = surface under the cumulative ranking. Keywords: Alzheimer disease, cognitive interventions, mild cognitive impairment, network meta-analysis 1. Introduction Editor: Helen Gharaei. Alzheimer disease (AD) is a neurological degenerative disease J-hL and LL contributed equally to the work presented here and therefore should be considered equivalent authors. that would obtain progressive development but concealed in the The authors have no conflicts of interest to disclose. early days. Clinically characterized by memory impairment, aphasia, disability, visual impairment, executive dysfunction, and Department of Child Health, Jiangsu Key Laboratory of Preventive and [1,2] Translational Medicine for Geriatric Diseases, School of Public Health, Soochow personality and behavioral changes, patients living with AD University, School of Nursing, Medical College of Soochow University, Suzhou, have a poor self-living ability and impose a series of burden on PR China. their family, caregivers, health-care system even society. As a Correspondence: Yong Xu, School of Public Health, Medical College of significant potential risk factor for AD, MCI is generally Soochow University, No. 199 Ren Ai Road, Suzhou 215123, China [3] considered a precursor to AD. Although there was a large (e-mail: childhealth@suda.edu.cn). amount of objective evidence that MCI patients have experienced Copyright © 2018 the Author(s). Published by Wolters Kluwer Health, Inc. a decline in cognitive function, their abilities in activities of daily This is an open access article distributed under the Creative Commons [4] Attribution License 4.0 (CCBY), which permits unrestricted use, distribution, and living (ADL) are still functional. reproduction in any medium, provided the original work is properly cited. The world’s older population currently comprises nearly 900 Medicine (2018) 97:20(e10744) million people, most of which come from relatively poor countries. Nowadays, more than 46 million people around the Received: 6 September 2017 / Accepted: 22 April 2018 world suffer from AD, and by 2050 the number is estimated to http://dx.doi.org/10.1097/MD.0000000000010744 1 Liang et al. Medicine (2018) 97:20 Medicine reach 131.5 million. The total estimated worldwide cost of 2. Methods [5] dementia is US $818 billion. Recently, a study showed that 2.1. Literature search the risk of AD is increasing in men and women as their age increases, but more prevalent in women (rate per 100 person- PUBMED, EMBASE, and the Cochrane Central Register of years=2.50 (1.85–3.41)) than in men (rate per 100 person- Controlled Trials were used for preliminary literature search [6] years=1.89 (1.22–2.94)). Several organizations, such as the before March 2017. With a highly sensitive strategy, we identified National Institute for Health and Care Excellence, suggest that relevant randomized controlled trials (RCTs). We used the management of patients with AD should be tailored to their MESH terms “Alzheimer’s Disease, Cognitive Therapy, Physical needs. The organizations of AD focus on retarding the Exercise, Music Therapy, Computer-Assisted, Nutrition therapy, progressive cognitive dysfunctions, maintaining functional randomized controlled trials” and keywords “Disease, Alz- status, improving quality of life, minimizing adverse events heimer” or “Alzheimer Dementia” or“ Alzheimer Type Demen- (AEs), modulating caregiver stress, and relieving the economic tia” or “ Alzheimer Type Senile Dementia” or “Cognitive burden of the family. Methods” or “Computer-Assisted Therapy” or “Computer Pharmacological therapies consist of multifarious cognitive Assisted Protocol Directed Therapy” or “Physical Activity” or enhancers and it is still not clear whether they are the optimal “Aerobic Exercise” or “Exercise Trainings” or “Nutritional treatment for AD. Moreover, evidence established by several Support” to search for related literature. studies has strongly shown that the use of some specialist drugs Moreover, we additionally scanned the bibliography of the like cholinesterase inhibitors increases the risk of AEs in patients included studies, such as studies in reports and reference lists of with AD. For example, cardiac medications like b-blockers may identified studies from published meta-analyses. The search increase risk of bradycardia, and antiinflammatories may covered the full-text of the reports published before March 1, [7] increase risk for gastrointestinal bleeding. 2017. This NMA was prepared according to the preferred Nonpharmacological therapies have attracted considerable reporting items for systematic reviews and meta-analyses [40] attention as a safe, relatively inexpensive and scalable interven- (PRISMA) guidelines. All analyses were based on previous tion that aims to maintain cognition in patients with AD and mild published studies, thus no ethical approval and patient consent cognitive impairment (MCI), which include social support, daily are required. activities, personalized cognitive treatment, advanced technical assistance, and support from the caregivers. Nonpharmacolog- 2.2. Eligibility criteria and data abstraction ical cognitive interventions for AD and MCI include physical We used population, interventions, comparisons, outcomes, exercise (PE) and music therapy (MT), as well as computerized study designs (PICOS) criteria. Eligible studies are RCTs that cognitive training (CCT), an efficacious and intelligently included older adults with AD or MCI and conducted a cognitive cognitive intervention. Several randomized, controlled trials intervention compared with each other, or control group. The (RCTs) have assessed the efficacy and safety of nonpharmaco- [8–27] particular PICOS criteria are: logical therapies (PE, MT, CCT) or pharmacological [28–33] therapies (NT) compared with control group (CG). Population: Older adults with AD or MCI diagnosed using However, the sample size of the previous studies was too small. various criteria. (Individual research has different measurement In addition, no direct comparisons between cognitive interven- method, but it does not affect the normal assessment of AD and tions have been made. Generally speaking, almost no study have MCI.) tried to answer such a sharp and debatable question—how to Interventions: Cognitive interventions including physical exercise, choose an optimal therapy from these interventions to treat older music therapy, computerized cognitive training (nonpharmaco- adults with AD or MCI. In the absence of direct evidence and logical therapies), and nutrition therapy (pharmacological large sample size, recently, a promising but much controversial therapy). extension of meta-analysis, network meta-analysis (NMA), has Comparisons: Cognitive interventions, control group alone or in been increasingly used. any combination. As the extension of traditional meta-analyses, NMA can Outcomes: The principal outcome was evaluated by validated simultaneously compare at least 2 interventions and pool data assessment of MMSE, as the efficacy of cognitive interventions. from different trials. It also enhances the relative effectiveness of As the second outcome, NPI was used to assess the neuropsychi- inference for each intervention through direct and indirect atric symptoms. The above outcomes were employed by an [34,35] information. Transitivity assumption is the pivotal as- adequate number of the included trials and thus our NMA can be sumption in NMA, which requires the balance of the distribution conducted. of potential effect modifiers across the treatment compari- Study design: We confined to RCTs for they are the optimal [36–38] sons. standard for examining interventions. And we did not include NMA is helpful when investigators are interested in summa- other types of trials in our NMA. rizing 2 or more of the treatment results and the hierarchy of these Two authors (J-hL, H-bZ) independently identified and treatments. Although there are doubts about these methodologi- evaluated articles during the initial literature search according cal issues for sample size, relevant outcomes, and heterogeneity to the above criteria and extracted information into an electronic sources, but for more comparisons, NMA may obtain more [39] database. Appropriateness of group allocation, blinding, accurate and reliable results than traditional meta-analysis. In intended indication, population characteristics, specific inter- this study, we employed this novel differential meta-analysis ventions, and the completeness of outcome report. Titles and method to estimate the comparative efficacy and safety associated abstracts were screened firstly, and if the article was potentially with cognitive interventions versus CG for AD or MCI. Our aim relevant, full-text article was retrieved. Once any discrepancies was to provide relatively effective and safe comparative evidence emerged, the authors would discuss with each other, and the third when identifying the optimal intervention for AD or MCI author (LH) was asked to resolve the divergence if necessary. A patients. 2 Liang et al. Medicine (2018) 97:20 www.md-journal.com unanimous agreement must be reached for these eligibility criteria The probability that which intervention was the most by all authors. efficacious intervention was derived from the proportion of the [45] Each study we extracted included demographic characteristics best ranking in all simulation operations. The Bayesian (e.g., gender, age mean and standard deviation, type of AD), approach has a superiority of being able to provide the most study aims, treatment time, outcomes (e.g., ADAS-Cog, MMSE, effective cognitive intervention probability even if the standard NPI), and study areas. If reports were of the same trial at different method may determine that there is no significant difference follow-up periods, data of the last report were used for analysis. between them. We used the network rank option to estimate the We used the mean, sample size and their standard deviation (SD) ranking probabilities. Probability values were summarized and [42] from each trial to analyze the group-specific of participants for reported as surface under the cumulative ranking (SUCRA). If continuous outcome. the corresponding cognitive intervention of the SUCRA is always 1, it is ranked first and 0 if it always ranks last. We also analyzed relative rankings among each cognitive intervention (second, 2.3. Outcome measures third, best, etc.), for some additional cases, the best cognitive intervention might be unavailable, more expensive, or contra- Different from traditional meta-analysis, our NMA did not indicated in some patients. Since our study only made the indirect extract the relevant outcome for each output, and only analyzed comparison between each cognitive intervention, we could not intersected outcomes reported in the original RCTs. The primary calculate the difference of standardized mean differences (MDs) outcome was MMSE which evaluated the cognitive domain. between direct and indirect comparisons to estimate the Scores and cognitive ability are proportional. The secondary [46] consistency of direct and indirect evidence. The above outcome was NPI, which assessed the neuropsychiatric symp- analyses were performed using Gemtc package (3.32 version) toms. Scores and neuropsychiatric symptoms are inversely in R. At last, we used a slightly adapted version of the risk of bias proportional. The means and SDs of the change from baseline approach of the Cochrane Collaboration to assess the quality of were extracted. To ensure data precision, 2 authors (J-hL, H-bZ) [47] each included study, which performed in Review Manager independently extracted all of the data and discrepancies were (5.3 version). settled by discussion or the involvement of a third author (LH). All authors were completely unanimous in selecting the outcome. 3. Results Figure 1 summarizes the selection process. We identified relevant 2.4. Statistical analysis studies for review of title and abstract at an initial screening. We We first analyzed the summary data and demographic character- used an extensive search strategy to retrieve the full texts of istics of each study. We also quantitatively estimated heteroge- potentially eligible RCTs. It therefore seems unlikely that we 2 [41] [48] neity across studies with the help of I statistic (ranges from missed some relevant trial. Three thousand five hundred 0% to 100%, the higher the I , the greater the heterogeneity), and twenty-six RCTs evaluating 4 different cognitive treatments were looked at the funnel plots to evaluate obvious publication biases identified from initial screening, and 20 studies met our inclusion based on visual inspection, after which the NMA was conducted. criteria, of which 17 studies were designed as RCTs and 3 studies The above random effects models in traditional meta-analysis were ruled out for having not identified a control group. All was used to estimate variance between studies by using STATA, participating authors agreed on the methodology for selection version 12 (Stata Corp, College Station, TX). The reason why we and assessment. Fifteen studies assessed MMSE and 5 assessed used the random effects model rather than the fixed effects model NPI. is that this might be the most appropriate and most conservative Table 1 presents the baseline data of demographic character- analysis of the variance between the studies. istics from 17 trials included. A total of 1931 AD patients As a natural extension of traditional meta-analysis for underwent PIO (Population, Intervention, Outcomes) strategies. [34] summarizing comparisons between treatment pairs, the The trials were published between 2004 and 2016 and the random effect Bayesian statistical model was implemented to majority of them were from US and Europe (N=1748, 91%). compare the indirect evidence for 4 cognitive interventions with Trials recruited participants mostly from their home. The mean placebo (cognitive interventions comparison: physical exercise vs age of all samples ranged from 69.8 to 86.1 years (one study [17] musical therapy vs computerized cognitive training vs nutrition lacked the data of baseline age). Fifty-five percent of the therapy vs placebo) combining all the descriptive data from participants were women, and the average scores of MMSE for [34,35,38,42] various studies. In the Bayesian framework, all all samples ranged from 7.9 to 27.9 at baseline. At last, the parameters are treated as random variables. For each incorpo- average scores of NPI at baseline ranged from 5.0 to 18.7. rated parameter, its posterior distribution is estimated by placing Figures 2 and 3 present the degree of risk of bias for all studies [43] the appropriate prior distribution using the Markov chain. included. The vast majority of studies had a low risk of outcome The number of tuning iterations was set at 5000 and the number data integrity. By contrast, the blinding of patient and of simulation iterations at 20,000. The degree of convergence of investigator were unclear generally. The overall quality of the the model was evaluated by visually inspecting the trace plot studies included in our study was modest. combined with density plot and the potential scale reduction [44] factors. We extracted the mean and SD of the MMSE at the 3.1. Efficacy and ranking of treatment arm last observation of the studies, and computed the standardized 3.1.1. Primary outcome. Among 17 studies included, 15 mean change (Hedges’ adjusted g) from baseline as the gist of recorded relevant data about cognition, 4 for PE, 3 for MT, 4 difference between the treatment groups. We also use the uniform for CCT, and 4 for NT. The absence of obvious heterogeneity method to evaluated the NPI scale as the measure of the (I =21.7%, P=.212) was shown by preliminary meta-analysis neuropsychiatric symptoms of patients. For each summary (Fig. 4A). The funnel plot showed a symmetric distribution statistic, a 95% credible interval (95% CI) was computed. (Fig. 5A), indicating no hint of publication bias. 3 Liang et al. Medicine (2018) 97:20 Medicine Figure 1. Literature review flowchart. AD=Alzheimer disease, ADAS-Cog=Alzheimer Disease Assessment Scale, cognitive subscale, CCT=computerized cognitive training, MCI=mild cognition impairment, MMSE=Mini-Mental State Examination, MT=music therapy, NPI=Neuropsychiatric Inventory, NT=nutrition therapy, PE=physical exercise, RCT=randomized controlled trial. At last, 15 eligible studies were finally included and reflected in by preliminary meta-analysis (I =61.6%, P=.034) (Fig. 4B). the network relationship plot (Fig. 6A). Our analysis revealed Since only 5 studies contributed data, the meta-regression could that only PE had a significantly greater improvement than CG not be conducted. The number of small studies we analyzed could (Fig. 7A). The accumulate histogram (Fig. 8A) presents the further explain the potential source of the heterogeneity. And the probability of rank for each cognitive intervention, which funnel plots (Fig. 5B) showed a fairly symmetric distribution, indicated that PE was the highest in probability among all the indicating no hint of publication bias. 4 cognitive interventions (SUCRA=0.45), followed by CCT Figure 6B shows a network of 5 eligible studies. Our analysis (SUCRA=0.30), and MT (SUCRA=0.17). In contrast, NT revealed that only CCT had a significantly greater improvement seemed to have the lowest probability. than CG (Fig. 7B). The accumulate histogram (Fig. 8B) presents the probability of rank for each cognitive intervention, which 3.1.2. Secondary outcome. NPI: Five studies reported relevant indicated that CCT was the highest in probability among all the 4 data about neuropsychiatric symptoms, 3 for PE, 1 for CCT, and cognitive interventions (SUCRA=0.87), followed by PE (SUCRA 1 for NT. The presence of an obvious heterogeneity was shown =0.54), and NT (SUCRA=0.29). 4 Liang et al. Medicine (2018) 97:20 www.md-journal.com Table 1 Baseline chart. Age, mean (SD) Gender (female, %) Baseline MMSE, mean (SD) Trial Outcome Intervention Study Type of intervention Diagnosis Exp Con Exp Con Exp Con duration, wk assessment tool Country [11] Physical Vreugdenhil et al Exercise program AD 73.5 (—) 74.7 (—) 45 75 22.9 (—) 21.0 (—) 16 MMSE Australia exercise [22] Hoffmann et al Aerobic exercise MAD 69.8 (7.4) 71.3 (7.3) 48 39 23.8 (3.4) 24.1 (3.8) 16 MMSE, NPI, France ADAS-Cog [12] Gustavo et al Physiotherapeutic intervention MD 72.9 (2.3) 79.4 (2.0) 71 70 12.7 (2.1) 14.6 (1.2) 24 MMSE Brazil [10] Van de Winckel et al Daily physical exercises AD 81.3 (4.2) 81.9 (4.2) 100 100 12.9 (5.0) 10.8 (5.0) 12 MMSE Belgium supported by music [15] Yang et al Aerobic exercise MAD 72.0 (6.7) 71.9 (7.3) 60 72 21.3 (2.2) 20.0 (3.5) 12 MMSE, NPI, China ADAS-Cog [20] Rolland et al Exercise program Mild to 82.8 (7.8) 83.1 (7.0) 72 79 9.7 (6.8) 7.9 (6.4) 54 NPI France severe AD [21] Music therapy Arroyo-Anllo et al Music intervention AD 74.4 (3.6) 75.2 (4.2) 90 95 19.3 (3.7) 19.9 (2.9) 12 MMSE Salamanca, Spain [49] Satoh et al Singing training AD 78.1 (7.0) 77.0 (6.1) 60 80 19.1 (3.9) 20.9 (3.5) 24 MMSE, NPI, Japan ADAS-Cog [9] Suzuki et al Music therapy SD 82.0 (8.4) 85.2 (4.2) 60 69 11.6 (6.6) 9.1 (7.0) 8 MMSE Japan [24] Computerized Barban et al Process-based cognitive training MCI 74.4 (5.7) 72.9 (6.0) 46 48 27.3 (2.1) 28.1 (1.4) 24 MMSE Italy, Greece, cognitive computerized pb-CT Norway, Spain training [19] Lee et al Computerized errorless learning- AD —— 86 67 15.3 (2.7) 17.6 (4.7) 6 MMSE Hongkong based memory training program (CELP) [17] Rozzini et al Multidimensional software MCI —— — — 26.0 (1.6) 26.4 (1.9) 54 MMSE, Italy NPI [25] Tarraga et al Interactive multimedia Internet- AD 75.8 (5.9) 77.4 (4.7) 87 88 20.6 (2.1) 22.5 (2.9) 24 MMSE, Spain based system (IMIS) + ADAS-Cog integrated psychostimulation program (IPP) [28] Nutrition Rondanelli et al Oily emulsion of docosahexaenoic MCI 85.3 (5.3) 86.1 (6.5) 82 79 26.2 (2.6) 26.7 (2.8) 12 MMSE Italy therapy acid (DHA)-phospholipids containing melatonin and tryptophan [30] Aisen et al Folate, vitamin b6, vitamin B12 Mild to 75.7 (8.0) 77.3 (7.9) 58 54 20.9 (3.4) 20.9 (3.7) 76 MMSE, USA moderate AD ADAS-Cog, NPI [29] Scheltens et al Medical food AD 74.1 (7.2) 73.3 (7.8) 49 51 23.8 (2.7) 24.0 (2.5) 12 MMSE Germany, Belgium, United Kingdom, United States [32] Petersen et al Vitamin E MCI 72.8 (7.3) 72.9 (7.6) 46 47 27.2 (1.9) 27.4 (1.8) 156 MMSE. US, Canada ADAS-Cog AD= Alzheimer disease, ADAS-Cog= Alzheimer Disease Assessment Scale, cognitive subscale, CCT= computerized cognitive training, CON= control group, EXP= experiment group, MAD= mild Alzheimer disease, MCI= mild cognition impairment, MMSE= Mini-Mental State Examination, MT= music therapy, NPI= Neuropsychiatric Inventory, NT= nutrition therapy, PE= physical exercise, RCT= randomized controlled trial, SD= senile dementia, —= lose data. Liang et al. Medicine (2018) 97:20 Medicine above findings will be reinforced by our analysis of previous meta-analyses. We applied a trial sequential analysis to detect the robustness and reliability of evidence for relative effectiveness of each [50–59] cognitive intervention. The trials in previous meta-analyses (PUBMED search March 1, 2017) only investigated the efficacy and safety of PE, MT, CCT, and NT, respectively, and lacked a synthesized analysis among them. By contrast, our NMA assessed PE, MT, CCT and incorporated NT using 4 pairwise MDs. Integrating indirect comparisons in our NMA resulted in higher statistical precision in scientific comparisons of cognitive interventions against a control group. This integration makes the comparison of different interventions more explicit and facilitates interpretation. The potential correlations between these 4 MDs were accounted for in our NMA, and linking to modeling of indirect comparisons provided greater statistical [60–62] power and more precise estimates. The totality of the evidence we extracted, largely based on trials in PE, MT, CCT, NT showed that further trials of cognitive interventions versus no cognitive interventions or control group are likely to have [30,32] positive effect, except for some specific trials. Based on relative effect estimates and SUCAR, nowadays, PE seems to be the most effective cognitive intervention when we consider a cognitive therapy and CCT is the most effective cognitive intervention for neuropsychiatric symptoms. The cumulative probability ranking obtained through the Bayesian NMA cannot be considered as decisive conclusion because it was probably compromised by the lack of a significant difference among the cognitive interventions. For example, PE ranked the first in cognition but did not have superiority over any of the other cognitive interventions, which might be due to the fact that PE-relevant studies contributed a relatively greater deal of evidence in the network (6 out of 15 studies), and thus significant differences between these cognitive interventions were not found. Previous studies have consistently demonstrated that almost all of these 4 cognitive interventions have beneficial effects on older adults with AD or MCI, PE in particular. Various kinds of [10–12,14–16,20,22,26,27] moderate PE including “Walking pro- gram”, “Whole-body vibration”, “Treadmill training” had demonstrated that it was useful for AD and MCI through improvement in cognitive function or other areas. Some previous studies demonstrated low intensity or multiple exercise were able to improve neuropsychiatric symptoms in older adults with AD [63,64] or MCI. moderate-to-high intensity PE can also improve [22] cognitive ability. It seemed that PE as a relatively common intervention can effectively improve the core domains in patients of AD. However, NT is a double-edged sword, because it can Figure 2. Risk of bias assessment. improve the cognitive ability but at the same time may cause some [29,30,32] AEs such as vomiting and diarrhea, which might explain why NT did not make an obvious improvement in cognitive ability of older adults with AD or MCI. By contrast, CCT as a relatively safe and inexpensive cognitive intervention has been 4. Discussion increasingly applied. Some trials using a variety of computer- [17,18,25] To the best of our knowledge, no previous study has solved the related advanced technologies to achieve CCT. It is worth problem that which cognitive intervention is the relatively best mentioning that there are 2 meta-analyses of which the subjects intervention for AD or MCI. Therefore, as the first NMA of were healthy older adults and which concluded that CCT were cognitive interventions for patients with AD or MCI in which moderately effective in long-term improvement of cogni- [65,66] indirect evidence was used appraising the relative effectiveness tion. Moreover, the majority of studies suggest that and safety of cognitive interventions across trials simultaneous- cognitive intervention is a long-term not temporary pro- [17,50,51,55] ly, our study attempted to summarize available data to suggest cess. However, network meta-analyses synthesizes that the highest probability of being the best intervention for various cognitive interventions, and the measures of discrepancy cognitive decline and neuropsychiatric symptoms lies in PE between them are fairly obvious. This is probably the main (SUCAR=0.45) and CCT (SUCAR=0.87) respectively. The reason for heterogeneity. NT is the only cognitive intervention we 6 Liang et al. Medicine (2018) 97:20 www.md-journal.com Figure 3. Risk of bias assessment (summary graph). Figure 4. (A) The forest plot of primary outcome (summary graph). (B) The forest plot of secondary outcome (summary graph). CCT=computerized cognitive training, CG=control group, MT=music therapy, NT=nutrition therapy, PE=physical exercise, SMD=standard mean difference. 7 Liang et al. Medicine (2018) 97:20 Medicine Figure 5. (A) Funnel plot of cognition. (B) Funnel plot of neuropsychiatric symptoms. included which may cause AEs, which is why the probability rank of this intervention was low. Although previous meta-analy- [50–59] ses provide high-quality evidence that PE, MT, CCT, NT Figure 6. (A) Network of cognitive interventions comparison of cognition for can improve cognitive ability and quality of life in people with the network meta-analysis. (B) Network of cognitive interventions comparison AD or MCI, but the trials they included only compared single of neuropsychiatric symptoms for the network meta-analysis. CCT= cognitive intervention with only a control group. By contrast, our computerized cognitive training, CG=control group, MT=music therapy, NT=nutrition therapy, PE=physical exercise. Bayesian network meta-analyses actualized the integration of different interventions. Since our results were based on indirectly randomized evidence, we were convinced that our study probably provides the best evidence of the efficacy and safety of these 4 AD and MCI are progressive neurodegenerative disorders, and cognitive interventions. are still incurable. Any cognitive intervention that could possibly In summary, PE had the best effective improvement in slow down the progressive of AD and MCI patients, it worth cognitive ability and the second best in neuropsychiatric disseminating. We may create an assumption that PE and MT as symptoms. CCT had the best result in improving neuropsychiat- a potent, convenient, selective cognitive interventions were play a ric symptoms and was relatively inexpensive. MT has a relatively positive role in helping improve the cognitive function for older low probability of being the best intervention for cognitive ability adults with AD or MCI. and neuropsychiatric symptoms. However, its safety factor and cost is relatively the best compared with other cognitive 4.1. Strengths and limitations interventions. It should be noted the efficacy of a series of nondrug interventions to improve cognitive ability of AD and Rather than only grouping various interventions into CCT or MCI patients have all been proven by research, such as estrogen human intervention, as the biggest strength, our NMA assessed [67,68] [69,70] replacement therapy, psychotherapy. But quite a each intervention individually and compared all major inter- ventions simultaneously. Then, potential bias was reduced in the number of trails were restricted by appropriate endpoints, which conduction of our review by having 2 independent authors (J-hL, resulted in this individually cognitive interventions lack of the H-bZ) scan through the search output, extract the relevant data, relevant endpoints whose efficacy we must adopt to analysis. classify each intervention, and assess the methodological quality Therefore, from our conclusion above, PE, MT, CCT all have of each trial. We performed an extensive search strategy across beneficial effects on older adults with AD and MCI, especially [58,71] [72,73] several databases and sources to obtain an adequate number of PE and MT as relatively obtained easily interventions. 8 Liang et al. Medicine (2018) 97:20 www.md-journal.com Figure 7. (A) Forest plot of cognition. (B) Forest plot of neuropsychiatric symptoms. CCT=computerized cognitive training, CG=control group, MT=music therapy, NT=nutrition therapy, PE=physical exercise. eligible studies, with no language restriction. We also extensively when they are making choices among different alternatives than searched the bibliographies of published studies. In addition, the results from multiple separate traditionally meta-analyses, cognitive intervention of CCT is complex and multifaceted and because several relevant outcomes have been assessed simulta- the number of relevant trials is very small, which proves the neously. It uses common random parameters to compare particular significance of our NMA. different interventions, which combines experimental evidence From the methodological point of view, our NMA demon- from indirectly randomized comparisons with observational strates a series of preponderance of Bayesian NMA for evidence from adjusted indirect comparisons derived from [74] comparing various cognitive interventions and for evaluating trials. the relative effectiveness and safety of multifarious interventions. The limitations of our study also need to be acknowledged. In this context, the results of our NMA are likely to be more Firstly, as the biggest limitation, the number of studies and the useful for decision makers, service commissioners and caregivers number of patients included in the study were relatively small. Figure 8. (A) Accumulate histogram of cognition. (B) Accumulate histogram of neuropsychiatric symptoms. CCT=computerized cognitive training, CG=control group, MT=music therapy, NT=nutrition therapy, PE=physical exercise. 9 Liang et al. Medicine (2018) 97:20 Medicine The studies included in our NMA used the same scale as the basis, Author contributions the outcomes of which were presented as a continuous variable. J-hL conducted the database search, screened and extracted data In the analysis section, we extracted the mean, SD, and sample for the meta-analysis, prepared extracted data for the procedures, size values at baseline and at last observation for analysis. and had primary responsibility in writing this article. LL However, a few studies lost their data, which made the number of performed statistical analysis and interpretation of data. H-bZ available studies even less. There are significant differences and LH contributed to the discussion and editing. YX critically among cognitive interventions such as in the method section. The revised the draft manuscript. All authors read and approved the above-mentioned reasons explain why the number of our final manuscript. included studies was limited. In particular, the informative Data curation: Lu Lin, Hong-bo Zhang, Lei Hang. evidence of the direct comparisons between cognitive interven- Formal analysis: Jing-hong Liang. tions was limited by the absence of relevant studies. Because no Investigation: Lu Lin, Hong-bo Zhang, Lei Hang. direct comparative trial was found through our search strategy, Methodology: Jing-hong Liang, Yong Xu. our study lacked direct evidence. No direct evidence was Project administration: Yong Xu. available when we performed NMA, and thus the evaluation Resources: Jing-hong Liang, Hong-bo Zhang, Lei Hang. of consistency could not be achieved. Secondly, only 5 of 17 Software: Jing-hong Liang, Lu Lin. studies included in the analyses were double-blind, and details of Supervision: Yong Xu. allocation were noted in 15 of 17 studies, indicating that Validation: Yong Xu, Lu Lin. publication bias and selective reporting biases could not be ruled Writing – original draft: Jing-hong Liang. out. Specific intervention regimens and patient populations Writing – review & editing: Yong Xu, Rui-xia Jia. varied across studies, which might cause heterogeneity. In addition, our study data were limited by the outcome of the intersection, a number of studies used their specific scales to References present outcomes. And quite a number of trials were restricted by [1] Zhang S, Zhang M, Cai F, et al. Biological function of Presenilin and its appropriate endpoints, which resulted in individual cognitive role in AD pathogenesis. Transl Neurodegener 2013;2:15. intervention lacking the relevant endpoints we must adopt, for [2] Di Iulio F, Palmer K, Blundo C, et al. Occurrence of neuropsychiatric example MT lack NPI data that only 5 studies included. That is symptoms and psychiatric disorders in mild Alzheimer’s disease and mild cognitive impairment subtypes. Int Psychogeriatr 2010;22:629–40. the reason why we could not evaluate this intervention [3] Knopman DS, Beiser A, Machulda MM, et al. Spectrum of cognition objectively in the end. Moreover, similar to previous traditional short of dementia: Framingham Heart Study and Mayo Clinic Study of meta-analyses, our study yielded heterogeneity due to the small Aging. Neurology 2015;85:1712–21. number of studies, although funnel plots did not suggest presence [4] Albert MS, DeKosky ST, Dickson D, et al. The diagnosis of mild of heterogeneity and an extensive search strategy was used to cognitive impairment due to Alzheimer’s disease: recommendations from the National Institute on Aging-Alzheimer’s Association workgroups on identify relevant trials. The NMA is complex, and it is difficult for diagnostic guidelines for Alzheimer’s disease. Alzheimers Dement decision makers to explain the results. For example, under the 2011;7:270–9. accumulate histogram analysis of MMSE, PE was most likely to [5] Prince M, Wimo A, Guerchet M, et al. World Alzheimer Report 2015. be the best cognitive intervention. This might be because the PE [6] Katz MJ, Lipton RB, Hall CB, et al. Age-specific and sex-specific intervention had a relative larger sample size used for NMA. At prevalence and incidence of mild cognitive impairment, dementia, and last, we have not extracted the number of patients who have been Alzheimer dementia in blacks and whites: a report from the Einstein observed in trials of numerous AEs or other reasons, which is Aging Study. Alzheimer Dis Assoc Disord 2012;26:335–43. [29,30,32] because only three studies mentioned the AEs and we [7] Veroniki AA, Straus SE, Ashoor HM, et al. Comparative safety and effectiveness of cognitive enhancers for Alzheimer’s dementia: protocol could not evaluate them in this analysis. for a systematic review and individual patient data network meta- analysis. BMJ Open 2016;6:e010251. 5. Conclusion [8] Li CH, Liu CK, Yang YH, et al. Adjunct effect of music therapy on cognition in Alzheimer’s disease in Taiwan: a pilot study. Neuropsychiatr In conclusion, our NMA suggested that PE is the optimum Dis Treat 2015;11:291–6. cognitive intervention for patients with AD or MCI while CCT is [9] Suzuki M, Kanamori M, Watanabe M, et al. Behavioral and endocrinological evaluation of music therapy for elderly patients with the optimum one for neuropsychiatric symptoms. Relatively dementia. Nurs Health Sci 2004;6:11–8. speaking, MT is the most safe intervention but its efficacy is [10] Van de Winckel A, Feys H, De Weerdt W, et al. Cognitive and moderate. And NT is the last choice to manage AD or MCI behavioural effects of music-based exercises in patients with dementia. because of its 2-sidedness. The results of our NMA suggest that Clin Rehabil 2004;18:253–60. [11] Vreugdenhil A, Cannell J, Davies A, et al. A community-based exercise nonpharmacological therapies are better than pharmacological programme to improve functional ability in people with Alzheimer’s therapies. In the future, there is a need to include more studies of disease: a randomized controlled trial. Scand J Caring Sci 2012;26:12–9. high methodological quality related to comprehensive cognitive [12] Christofoletti G, Oliani MM, Gobbi S, et al. A controlled clinical trial on interventions to help establish a more extensive literature the effects of motor intervention on balance and cognition in foundation. Researchers should shift their research interest to institutionalized elderly patients with dementia. Clin Rehabil 2008;22: 618–26. outcomes other than MMSE, such as ADL and Clinical Dementia [13] Kim HJ, Yang Y, Oh JG, et al. Effectiveness of a community-based Rating (CDR) so that more network meta-analyses of cognitive multidomain cognitive intervention program in patients with Alzheimer’s interventions for AD or MCI patients can be performed. disease. Geriatr Gerontol Int 2016;16:191–9. [14] Lam FM, Liao LR, Kwok TC, et al. Effects of adding whole-body vibration to routine day activity program on physical functioning in Acknowledgments elderly with mild or moderate dementia: a randomized controlled trial. Int J Geriatr Psychiatry 2017;33:21–30. We would like to gratefully acknowledge the help by the [15] Yang SY, Shan CL, Qing H, et al. The effects of aerobic exercise on following authors: Chun-Hua Zhao, Deng-Juan Qian. All cognitive function of Alzheimer’s disease patients. CNS Neurol Disord Drug Targets 2015;14:1292–7. authors agreed to contribute to this study. 10 Liang et al. Medicine (2018) 97:20 www.md-journal.com [16] Holthoff VA, Marschner K, Scharf M, et al. Effects of physical activity [41] Higgins JP, Thompson SG. Quantifying heterogeneity in a meta-analysis. training in patients with Alzheimer’s dementia: results of a pilot RCT Stat Med 2002;21:1539–58. study. PLoS ONE 2015;10:e0121478. [42] Salanti G, Ades AE, Ioannidis JP. Graphical methods and numerical [17] Rozzini L, Costardi D, Chilovi BV, et al. Efficacy of cognitive summaries for presenting results from multiple-treatment meta-analysis: rehabilitation in patients with mild cognitive impairment treated with an overview and tutorial. J Clin Epidemiol 2011;64:163–71. cholinesterase inhibitors. Int J Geriatr Psychiatry 2007;22:356–60. [43] Mavridis D, Salanti G. A practical introduction to multivariate meta- [18] Fernandez-Calvo B, Rodriguez-Perez R, Contador I, et al. [Efficacy of analysis. Stat Methods Med Res 2013;22:133–58. cognitive training programs based on new software technologies in [44] Brooks SP, Gelman A. General methods for monitoring convergence of patients with Alzheimer-type dementia]. Psicothema 2011;23:44–50. iterative simulations. J Comput Graph Stat 1997;7:434–55. [19] Lee GY, Yip CC, Yu EC, et al. Evaluation of a computer-assisted [45] Dias S, Welton NJ, Sutton AJ, et al. NICE Decision Support Unit errorless learning-based memory training program for patients with early Technical Support Documents. A Generalised Linear Modelling Alzheimer’s disease in Hong Kong: a pilot study. Clin Interv Aging Framework for Pairwise and Network Meta-Analysis of Randomised 2013;8:623–33. Controlled Trials. London: National Institute for Health and Care [20] Rolland Y, Pillard F, Klapouszczak A, et al. Exercise program for nursing Excellence (NICE) unless otherwise stated. All rights reserved, home residents with Alzheimer’s disease: a 1-year randomized, 2014. controlled trial. J Am Geriatr Soc 2007;55:158–65. [46] Salanti G, Marinho V, Higgins JP. A case study of multiple-treatments [21] Arroyo-Anllo EM, Diaz JP, Gil R. Familiar music as an enhancer of self- meta-analysis demonstrates that covariates should be considered. J Clin consciousness in patients with Alzheimer’s disease. Biomed Res Int Epidemiol 2009;62:857–64. 2013;2013:752965. [47] Higgins JPT, Green S (Eds.). Cochrane Handbook for Systematic [22] Hoffmann K, Sobol NA, Frederiksen KS, et al. Moderate-to-high Reviews of Interventions Version 5.1.0. March 2011. intensity physical exercise in patients with Alzheimer’s disease: a [48] Egger M, Juni P, Bartlett C, et al. How important are comprehensive randomized controlled trial. J Alzheimers Dis 2016;50:443–53. literature searches and the assessment of trial quality in systematic [23] Gomez Gallego M, Gomez Garcia J. Music therapy and Alzheimer’s reviews? Empirical study. Health Technol Assess 2003;7:1–76. disease: cognitive, psychological, and behavioural effects. Neurologia [49] Satoh M, Ogawa JI, Tokita T, et al. The effects of physical exercise with (Barcelona, Spain) 2016;32:300–8. music on cognitive function of elderly people: Mihama-Kiho project. [24] Barban F, Annicchiarico R, Pantelopoulos S, et al. Protecting cognition PloS one 2014. from aging and Alzheimer’s disease: a computerized cognitive training [50] Hill NT, Mowszowski L, Naismith SL, et al. Computerized cognitive combined with reminiscence therapy. Int J Geriatr Psychiatry training in older adults with mild cognitive impairment or dementia: a 2016;31:340–8. systematic review and meta-analysis. Am J Psychiatry 2017;174: [25] Tarraga L, Boada M, Modinos G, et al. A randomised pilot study to 329–40. assess the efficacy of an interactive, multimedia tool of cognitive [51] Garcia-Casal JA, Loizeau A, Csipke E, et al. Computer-based cognitive stimulation in Alzheimer’s disease. J Neurol Neurosurg Psychiatry interventions for people living with dementia: a systematic literature 2006;77:1116–21. review and meta-analysis. Aging Mental Health 2017;21:454–67. [26] Venturelli M, Scarsini R, Schena F. Six-month walking program changes [52] Fang R, Ye S, Huangfu J, et al. Music therapy is a potential intervention cognitive and ADL performance in patients with Alzheimer. Am J for cognition of Alzheimer’s disease: a mini-review. Transl Neurodegener Alzheimers Dis Other Demen 2011;26:381–8. 2017;6:2. [27] Arcoverde C, Deslandes A, Moraes H, et al. Treadmill training as an [53] Groot C, Hooghiemstra AM, Raijmakers PG, et al. The effect of physical augmentation treatment for Alzheimer’s disease: a pilot randomized activity on cognitive function in patients with dementia: a meta-analysis controlled study. Arq Neuropsiquiatr 2014;72:190–6. of randomized control trials. Ageing Res Rev 2016;25:13–23. [28] Rondanelli M, Opizzi A, Faliva M, et al. Effects of a diet integration with [54] Mohajeri MH, Troesch B, Weber P. Inadequate supply of vitamins and an oily emulsion of DHA-phospholipids containing melatonin and DHA in the elderly: implications for brain aging and Alzheimer-type tryptophan in elderly patients suffering from mild cognitive impairment. dementia. Nutrition 2015;31:261–75. Nutr Neurosci 2012;15:46–54. [55] Coyle H, Traynor V, Solowij N. Computerized and virtual reality [29] Scheltens P, Kamphuis PJ, Verhey FR, et al. Efficacy of a medical food in cognitive training for individuals at high risk of cognitive decline: mild Alzheimer’s disease: a randomized, controlled trial. Alzheimers systematic review of the literature. Am J Geriatr Psychiatry 2015;23: Dement 2010;6:1.e1–0.e1. 335–59. [30] Aisen PS, Schneider LS, Sano M, et al. High-dose B vitamin [56] Ueda T, Suzukamo Y, Sato M, et al. Effects of music therapy on supplementation and cognitive decline in Alzheimer disease: a random- behavioral and psychological symptoms of dementia: a systematic review ized controlled trial. JAMA 2008;300:1774–83. and meta-analysis. Ageing Res Rev 2013;12:628–41. [31] Gangwar AK, Rawat A, Tiwari S, et al. Role of vitamin-D in the [57] Shah R. The role of nutrition and diet in Alzheimer disease: a systematic prevention and treatment of Alzheimer’s disease. Indian J Physiol review. J Am Med Dir Assoc 2013;14:398–402. Pharmacol 2015;59:94–9. [58] Littbrand H, Stenvall M, Rosendahl E. Applicability and effects of [32] Petersen RC, Thomas RG, Grundman M, et al. Vitamin E and donepezil physical exercise on physical and cognitive functions and activities of for the treatment of mild cognitive impairment. N Engl J Med daily living among people with dementia: a systematic review. Am J Phys 2005;352:2379–88. Med Rehabil 2011;90:495–518. [33] Ford AH, Flicker L, Thomas J, et al. Vitamins B12, B6, and folic acid for [59] Heyn P, Abreu BC, Ottenbacher KJ. The effects of exercise training on onset of depressive symptoms in older men: results from a 2-year elderly persons with cognitive impairment and dementia: a meta- placebo-controlled randomized trial. J Clin Psychiatry 2008;69:1203–9. analysis. Arch Phys Med Rehabil 2004;85:1694–704. [34] Lu G, Ades AE. Combination of direct and indirect evidence in mixed [60] Ades AE, Sculpher M, Sutton A, et al. Bayesian methods for evidence treatment comparisons. Stat Med 2004;23:3105–24. synthesis in cost-effectiveness analysis. Pharmacoeconomics 2006; [35] Salanti G, Higgins JP, Ades AE, et al. Evaluation of networks of 24:1–9. randomized trials. Stat Methods Med Res 2008;17:279–301. [61] Sutton A, Ades AE, Cooper N, et al. Use of indirect and mixed treatment [36] Salanti G. Indirect and mixed-treatment comparison, network, or comparisons for technology assessment. Pharmacoeconomics 2008;26: multiple-treatments meta-analysis: many names, many benefits, many 753–67. concerns for the next generation evidence synthesis tool. Res Synth [62] Song F, Altman DG, Glenny AM, et al. Validity of indirect comparison Methods 2012;3:80–97. for estimating efficacy of competing interventions: empirical evidence [37] Jansen JP, Naci H. Is network meta-analysis as valid as standard pairwise from published meta-analyses. BMJ 2003;326:472. meta-analysis? It all depends on the distribution of effect modifiers. BMC [63] Nascimento CM, Teixeira CV, Gobbi LT, et al. A controlled clinical trial Med 2013;11:159. on the effects of exercise on neuropsychiatric disorders and instrumental [38] Caldwell DM, Ades AE, Higgins JP. Simultaneous comparison of activities in women with Alzheimer’s disease. Rev Bras Fisioter 2012;16: multiple treatments: combining direct and indirect evidence. BMJ 197–204. 2005;331:897–900. [64] Stevens J, Killeen M. A randomised controlled trial testing the impact of [39] Mills EJ, Thorlund K, Ioannidis JP. Demystifying trial networks and exercise on cognitive symptoms and disability of residents with dementia. network meta-analysis. BMJ 2013;346:f2914. Contemp Nurse 2006;21:32–40. [40] Moher D, Liberati A, Tetzlaff J, et al. Preferred reporting items for [65] Lampit A, Hallock H, Valenzuela M. Computerized cognitive training in systematic reviews and meta-analyses: the PRISMA statement. BMJ cognitively healthy older adults: a systematic review and meta-analysis of 2009;339:b2535. effect modifiers. PLoS Med 2014;11:e1001756. 11 Liang et al. Medicine (2018) 97:20 Medicine [66] Shah TM, Weinborn M, Verdile G, et al. Enhancing cognitive [70] Hirsch RD. [Socio- and psychotherapy in patients with Alzheimer functioning in healthly older adults: a systematic review of the clinical disease]. Z Gerontol Geriatr 2001;34:92–100. significance of commercially available computerized cognitive training in [71] Jensen CS, Portelius E, Hogh P, et al. Effect of physical exercise on preventing cognitive decline. Neuropsychol Rev 2017;27:62–80. markers of neuronal dysfunction in cerebrospinal fluid in patients with [67] Mulnard RA, Cotman CW, Kawas C, et al. Estrogen replacement Alzheimer’s disease. Alzheimers Dement (NY) 2017;3:284–90. therapy for treatment of mild to moderate Alzheimer disease: a [72] Särkämö T, Laitinen S, Numminen A, et al. Clinical and demographic randomized controlled trial. Alzheimer’s Disease Cooperative Study. factors associated with the cognitive and emotional efficacy of regular JAMA 2000;283:1007–15. musical activities in dementia. J Alzheimers Dis 2015;49:767–81. [68] Levine AJ, Battista M. Estrogen replacement therapy: effects on the [73] Liesk J, Hartogh T, Kalbe E. [Cognitive stimulation and music cognitive functioning and clinical course of women with Alzheimer’s intervention for people with dementia in nursing homes: a pilot study, disease. Arch Clin Neuropsychol 2004;19:769–78. problems and perspectives]. Z Gerontol Geriatr 2015;48:275–81. [69] Burns A, Guthrie E, Marino-Francis F, et al. Brief psychotherapy in [74] Cipriani A, Higgins JP, Geddes JR, et al. Conceptual and technical Alzheimer’s disease: randomised controlled trial. Br J Psychiatry challenges in network meta-analysis. Ann Intern Med 2013;159: 2005;187:143–7. 130–7. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Medicine Wolters Kluwer Health

Comparison of multiple interventions for older adults with Alzheimer disease or mild cognitive impairment

Download PDF
12 pages

Loading next page...
 
/lp/wolters-kluwer-health/comparison-of-multiple-interventions-for-older-adults-with-alzheimer-7lkaDL063h

References (82)

Publisher
Wolters Kluwer Health
Copyright
Copyright © 2018 the Author(s). Published by Wolters Kluwer Health, Inc.
ISSN
0025-7974
eISSN
1536-5964
DOI
10.1097/MD.0000000000010744
pmid
29768349
Publisher site
See Article on Publisher Site

Abstract

Background: The increasing prevalence of Alzheimer disease (AD) emphasizes the need for effective treatments. Both pharmacological therapies such as nutrition therapy (NT) and nonpharmacologic therapies including traditional treatment or personalized treatment (e.g., physical exercise, music therapy, computerized cognitive training) have been approved for the treatment of AD or mild cognitive impairment (MCI) in numerous areas. Methods: The aim of this study was to compare 4 types of interventions, physical exercise (PE), music therapy (MT), computerized cognitive training (CCT), and NT, in older adults with mild to moderate AD or MCI and identify the most effective intervention for their cognitive function. We used a system of search strategies to identify relevant studies and include randomized controlled trials (RCTs), placebo-controlled trials evaluating the efficacy and safety of 4 interventions in patients with AD or MCI. We updated the relevant studies which were published before March 2017 as a full-text article. Using Bayesian network meta-analysis (NMA), we ranked cognitive ability based objectively on Mini-Mental State Examination (MMSE), and assessed neuropsychiatric symptoms based on Neuropsychiatric Inventory (NPI). Pairwise and network meta-analyses were sequentially performed for efficacy and safety of intervention compared to control group through RCTs included. Results: We included 17 RCTs. Fifteen trials (n=1747) were pooled for cognition and no obvious heterogeneity was found (I = 21.7%, P=.212) in NMA, the mean difference (MD) of PE (MD=2.1, confidence interval [CI]: 0.44–3.8) revealed that PE was significantly efficacious in the treatment group in terms of MMSE. Five trials (n=660) assessed neuropsychiatric symptoms with an obvious heterogeneity (I =61.6%, P=.034), the MD of CCT (MD=7.7, CI: 14 to 2.4), revealing that CCT was significantly efficacious in NPI. Conclusions: As the first NMA comparing different interventions for AD and MCI, our study suggests that PE and CCT might have a significant improvement in cognition and neuropsychiatric symptoms respectively. Moreover, nonpharmacological therapies might be better than pharmacological therapies. Abbreviations: AD = Alzheimer disease, ADAS-Cog = Alzheimer disease Assessment Scale, cognitive subscale, ADL = activities of daily living, AEs = adverse events, CCT = computerized cognitive training, CDR = Clinical Dementia Rating, CG = control group, MCI = mild cognition impairment, MD = mean difference, MMSE = Mini-Mental State Examination, MT = music therapy, NMA = network meta-analysis, NPI = Neuropsychiatric Inventory, NT = nutrition therapy, PE = physical exercise, RCT = randomized controlled trial, SD = standard deviation, SUCRA = surface under the cumulative ranking. Keywords: Alzheimer disease, cognitive interventions, mild cognitive impairment, network meta-analysis 1. Introduction Editor: Helen Gharaei. Alzheimer disease (AD) is a neurological degenerative disease J-hL and LL contributed equally to the work presented here and therefore should be considered equivalent authors. that would obtain progressive development but concealed in the The authors have no conflicts of interest to disclose. early days. Clinically characterized by memory impairment, aphasia, disability, visual impairment, executive dysfunction, and Department of Child Health, Jiangsu Key Laboratory of Preventive and [1,2] Translational Medicine for Geriatric Diseases, School of Public Health, Soochow personality and behavioral changes, patients living with AD University, School of Nursing, Medical College of Soochow University, Suzhou, have a poor self-living ability and impose a series of burden on PR China. their family, caregivers, health-care system even society. As a Correspondence: Yong Xu, School of Public Health, Medical College of significant potential risk factor for AD, MCI is generally Soochow University, No. 199 Ren Ai Road, Suzhou 215123, China [3] considered a precursor to AD. Although there was a large (e-mail: childhealth@suda.edu.cn). amount of objective evidence that MCI patients have experienced Copyright © 2018 the Author(s). Published by Wolters Kluwer Health, Inc. a decline in cognitive function, their abilities in activities of daily This is an open access article distributed under the Creative Commons [4] Attribution License 4.0 (CCBY), which permits unrestricted use, distribution, and living (ADL) are still functional. reproduction in any medium, provided the original work is properly cited. The world’s older population currently comprises nearly 900 Medicine (2018) 97:20(e10744) million people, most of which come from relatively poor countries. Nowadays, more than 46 million people around the Received: 6 September 2017 / Accepted: 22 April 2018 world suffer from AD, and by 2050 the number is estimated to http://dx.doi.org/10.1097/MD.0000000000010744 1 Liang et al. Medicine (2018) 97:20 Medicine reach 131.5 million. The total estimated worldwide cost of 2. Methods [5] dementia is US $818 billion. Recently, a study showed that 2.1. Literature search the risk of AD is increasing in men and women as their age increases, but more prevalent in women (rate per 100 person- PUBMED, EMBASE, and the Cochrane Central Register of years=2.50 (1.85–3.41)) than in men (rate per 100 person- Controlled Trials were used for preliminary literature search [6] years=1.89 (1.22–2.94)). Several organizations, such as the before March 2017. With a highly sensitive strategy, we identified National Institute for Health and Care Excellence, suggest that relevant randomized controlled trials (RCTs). We used the management of patients with AD should be tailored to their MESH terms “Alzheimer’s Disease, Cognitive Therapy, Physical needs. The organizations of AD focus on retarding the Exercise, Music Therapy, Computer-Assisted, Nutrition therapy, progressive cognitive dysfunctions, maintaining functional randomized controlled trials” and keywords “Disease, Alz- status, improving quality of life, minimizing adverse events heimer” or “Alzheimer Dementia” or“ Alzheimer Type Demen- (AEs), modulating caregiver stress, and relieving the economic tia” or “ Alzheimer Type Senile Dementia” or “Cognitive burden of the family. Methods” or “Computer-Assisted Therapy” or “Computer Pharmacological therapies consist of multifarious cognitive Assisted Protocol Directed Therapy” or “Physical Activity” or enhancers and it is still not clear whether they are the optimal “Aerobic Exercise” or “Exercise Trainings” or “Nutritional treatment for AD. Moreover, evidence established by several Support” to search for related literature. studies has strongly shown that the use of some specialist drugs Moreover, we additionally scanned the bibliography of the like cholinesterase inhibitors increases the risk of AEs in patients included studies, such as studies in reports and reference lists of with AD. For example, cardiac medications like b-blockers may identified studies from published meta-analyses. The search increase risk of bradycardia, and antiinflammatories may covered the full-text of the reports published before March 1, [7] increase risk for gastrointestinal bleeding. 2017. This NMA was prepared according to the preferred Nonpharmacological therapies have attracted considerable reporting items for systematic reviews and meta-analyses [40] attention as a safe, relatively inexpensive and scalable interven- (PRISMA) guidelines. All analyses were based on previous tion that aims to maintain cognition in patients with AD and mild published studies, thus no ethical approval and patient consent cognitive impairment (MCI), which include social support, daily are required. activities, personalized cognitive treatment, advanced technical assistance, and support from the caregivers. Nonpharmacolog- 2.2. Eligibility criteria and data abstraction ical cognitive interventions for AD and MCI include physical We used population, interventions, comparisons, outcomes, exercise (PE) and music therapy (MT), as well as computerized study designs (PICOS) criteria. Eligible studies are RCTs that cognitive training (CCT), an efficacious and intelligently included older adults with AD or MCI and conducted a cognitive cognitive intervention. Several randomized, controlled trials intervention compared with each other, or control group. The (RCTs) have assessed the efficacy and safety of nonpharmaco- [8–27] particular PICOS criteria are: logical therapies (PE, MT, CCT) or pharmacological [28–33] therapies (NT) compared with control group (CG). Population: Older adults with AD or MCI diagnosed using However, the sample size of the previous studies was too small. various criteria. (Individual research has different measurement In addition, no direct comparisons between cognitive interven- method, but it does not affect the normal assessment of AD and tions have been made. Generally speaking, almost no study have MCI.) tried to answer such a sharp and debatable question—how to Interventions: Cognitive interventions including physical exercise, choose an optimal therapy from these interventions to treat older music therapy, computerized cognitive training (nonpharmaco- adults with AD or MCI. In the absence of direct evidence and logical therapies), and nutrition therapy (pharmacological large sample size, recently, a promising but much controversial therapy). extension of meta-analysis, network meta-analysis (NMA), has Comparisons: Cognitive interventions, control group alone or in been increasingly used. any combination. As the extension of traditional meta-analyses, NMA can Outcomes: The principal outcome was evaluated by validated simultaneously compare at least 2 interventions and pool data assessment of MMSE, as the efficacy of cognitive interventions. from different trials. It also enhances the relative effectiveness of As the second outcome, NPI was used to assess the neuropsychi- inference for each intervention through direct and indirect atric symptoms. The above outcomes were employed by an [34,35] information. Transitivity assumption is the pivotal as- adequate number of the included trials and thus our NMA can be sumption in NMA, which requires the balance of the distribution conducted. of potential effect modifiers across the treatment compari- Study design: We confined to RCTs for they are the optimal [36–38] sons. standard for examining interventions. And we did not include NMA is helpful when investigators are interested in summa- other types of trials in our NMA. rizing 2 or more of the treatment results and the hierarchy of these Two authors (J-hL, H-bZ) independently identified and treatments. Although there are doubts about these methodologi- evaluated articles during the initial literature search according cal issues for sample size, relevant outcomes, and heterogeneity to the above criteria and extracted information into an electronic sources, but for more comparisons, NMA may obtain more [39] database. Appropriateness of group allocation, blinding, accurate and reliable results than traditional meta-analysis. In intended indication, population characteristics, specific inter- this study, we employed this novel differential meta-analysis ventions, and the completeness of outcome report. Titles and method to estimate the comparative efficacy and safety associated abstracts were screened firstly, and if the article was potentially with cognitive interventions versus CG for AD or MCI. Our aim relevant, full-text article was retrieved. Once any discrepancies was to provide relatively effective and safe comparative evidence emerged, the authors would discuss with each other, and the third when identifying the optimal intervention for AD or MCI author (LH) was asked to resolve the divergence if necessary. A patients. 2 Liang et al. Medicine (2018) 97:20 www.md-journal.com unanimous agreement must be reached for these eligibility criteria The probability that which intervention was the most by all authors. efficacious intervention was derived from the proportion of the [45] Each study we extracted included demographic characteristics best ranking in all simulation operations. The Bayesian (e.g., gender, age mean and standard deviation, type of AD), approach has a superiority of being able to provide the most study aims, treatment time, outcomes (e.g., ADAS-Cog, MMSE, effective cognitive intervention probability even if the standard NPI), and study areas. If reports were of the same trial at different method may determine that there is no significant difference follow-up periods, data of the last report were used for analysis. between them. We used the network rank option to estimate the We used the mean, sample size and their standard deviation (SD) ranking probabilities. Probability values were summarized and [42] from each trial to analyze the group-specific of participants for reported as surface under the cumulative ranking (SUCRA). If continuous outcome. the corresponding cognitive intervention of the SUCRA is always 1, it is ranked first and 0 if it always ranks last. We also analyzed relative rankings among each cognitive intervention (second, 2.3. Outcome measures third, best, etc.), for some additional cases, the best cognitive intervention might be unavailable, more expensive, or contra- Different from traditional meta-analysis, our NMA did not indicated in some patients. Since our study only made the indirect extract the relevant outcome for each output, and only analyzed comparison between each cognitive intervention, we could not intersected outcomes reported in the original RCTs. The primary calculate the difference of standardized mean differences (MDs) outcome was MMSE which evaluated the cognitive domain. between direct and indirect comparisons to estimate the Scores and cognitive ability are proportional. The secondary [46] consistency of direct and indirect evidence. The above outcome was NPI, which assessed the neuropsychiatric symp- analyses were performed using Gemtc package (3.32 version) toms. Scores and neuropsychiatric symptoms are inversely in R. At last, we used a slightly adapted version of the risk of bias proportional. The means and SDs of the change from baseline approach of the Cochrane Collaboration to assess the quality of were extracted. To ensure data precision, 2 authors (J-hL, H-bZ) [47] each included study, which performed in Review Manager independently extracted all of the data and discrepancies were (5.3 version). settled by discussion or the involvement of a third author (LH). All authors were completely unanimous in selecting the outcome. 3. Results Figure 1 summarizes the selection process. We identified relevant 2.4. Statistical analysis studies for review of title and abstract at an initial screening. We We first analyzed the summary data and demographic character- used an extensive search strategy to retrieve the full texts of istics of each study. We also quantitatively estimated heteroge- potentially eligible RCTs. It therefore seems unlikely that we 2 [41] [48] neity across studies with the help of I statistic (ranges from missed some relevant trial. Three thousand five hundred 0% to 100%, the higher the I , the greater the heterogeneity), and twenty-six RCTs evaluating 4 different cognitive treatments were looked at the funnel plots to evaluate obvious publication biases identified from initial screening, and 20 studies met our inclusion based on visual inspection, after which the NMA was conducted. criteria, of which 17 studies were designed as RCTs and 3 studies The above random effects models in traditional meta-analysis were ruled out for having not identified a control group. All was used to estimate variance between studies by using STATA, participating authors agreed on the methodology for selection version 12 (Stata Corp, College Station, TX). The reason why we and assessment. Fifteen studies assessed MMSE and 5 assessed used the random effects model rather than the fixed effects model NPI. is that this might be the most appropriate and most conservative Table 1 presents the baseline data of demographic character- analysis of the variance between the studies. istics from 17 trials included. A total of 1931 AD patients As a natural extension of traditional meta-analysis for underwent PIO (Population, Intervention, Outcomes) strategies. [34] summarizing comparisons between treatment pairs, the The trials were published between 2004 and 2016 and the random effect Bayesian statistical model was implemented to majority of them were from US and Europe (N=1748, 91%). compare the indirect evidence for 4 cognitive interventions with Trials recruited participants mostly from their home. The mean placebo (cognitive interventions comparison: physical exercise vs age of all samples ranged from 69.8 to 86.1 years (one study [17] musical therapy vs computerized cognitive training vs nutrition lacked the data of baseline age). Fifty-five percent of the therapy vs placebo) combining all the descriptive data from participants were women, and the average scores of MMSE for [34,35,38,42] various studies. In the Bayesian framework, all all samples ranged from 7.9 to 27.9 at baseline. At last, the parameters are treated as random variables. For each incorpo- average scores of NPI at baseline ranged from 5.0 to 18.7. rated parameter, its posterior distribution is estimated by placing Figures 2 and 3 present the degree of risk of bias for all studies [43] the appropriate prior distribution using the Markov chain. included. The vast majority of studies had a low risk of outcome The number of tuning iterations was set at 5000 and the number data integrity. By contrast, the blinding of patient and of simulation iterations at 20,000. The degree of convergence of investigator were unclear generally. The overall quality of the the model was evaluated by visually inspecting the trace plot studies included in our study was modest. combined with density plot and the potential scale reduction [44] factors. We extracted the mean and SD of the MMSE at the 3.1. Efficacy and ranking of treatment arm last observation of the studies, and computed the standardized 3.1.1. Primary outcome. Among 17 studies included, 15 mean change (Hedges’ adjusted g) from baseline as the gist of recorded relevant data about cognition, 4 for PE, 3 for MT, 4 difference between the treatment groups. We also use the uniform for CCT, and 4 for NT. The absence of obvious heterogeneity method to evaluated the NPI scale as the measure of the (I =21.7%, P=.212) was shown by preliminary meta-analysis neuropsychiatric symptoms of patients. For each summary (Fig. 4A). The funnel plot showed a symmetric distribution statistic, a 95% credible interval (95% CI) was computed. (Fig. 5A), indicating no hint of publication bias. 3 Liang et al. Medicine (2018) 97:20 Medicine Figure 1. Literature review flowchart. AD=Alzheimer disease, ADAS-Cog=Alzheimer Disease Assessment Scale, cognitive subscale, CCT=computerized cognitive training, MCI=mild cognition impairment, MMSE=Mini-Mental State Examination, MT=music therapy, NPI=Neuropsychiatric Inventory, NT=nutrition therapy, PE=physical exercise, RCT=randomized controlled trial. At last, 15 eligible studies were finally included and reflected in by preliminary meta-analysis (I =61.6%, P=.034) (Fig. 4B). the network relationship plot (Fig. 6A). Our analysis revealed Since only 5 studies contributed data, the meta-regression could that only PE had a significantly greater improvement than CG not be conducted. The number of small studies we analyzed could (Fig. 7A). The accumulate histogram (Fig. 8A) presents the further explain the potential source of the heterogeneity. And the probability of rank for each cognitive intervention, which funnel plots (Fig. 5B) showed a fairly symmetric distribution, indicated that PE was the highest in probability among all the indicating no hint of publication bias. 4 cognitive interventions (SUCRA=0.45), followed by CCT Figure 6B shows a network of 5 eligible studies. Our analysis (SUCRA=0.30), and MT (SUCRA=0.17). In contrast, NT revealed that only CCT had a significantly greater improvement seemed to have the lowest probability. than CG (Fig. 7B). The accumulate histogram (Fig. 8B) presents the probability of rank for each cognitive intervention, which 3.1.2. Secondary outcome. NPI: Five studies reported relevant indicated that CCT was the highest in probability among all the 4 data about neuropsychiatric symptoms, 3 for PE, 1 for CCT, and cognitive interventions (SUCRA=0.87), followed by PE (SUCRA 1 for NT. The presence of an obvious heterogeneity was shown =0.54), and NT (SUCRA=0.29). 4 Liang et al. Medicine (2018) 97:20 www.md-journal.com Table 1 Baseline chart. Age, mean (SD) Gender (female, %) Baseline MMSE, mean (SD) Trial Outcome Intervention Study Type of intervention Diagnosis Exp Con Exp Con Exp Con duration, wk assessment tool Country [11] Physical Vreugdenhil et al Exercise program AD 73.5 (—) 74.7 (—) 45 75 22.9 (—) 21.0 (—) 16 MMSE Australia exercise [22] Hoffmann et al Aerobic exercise MAD 69.8 (7.4) 71.3 (7.3) 48 39 23.8 (3.4) 24.1 (3.8) 16 MMSE, NPI, France ADAS-Cog [12] Gustavo et al Physiotherapeutic intervention MD 72.9 (2.3) 79.4 (2.0) 71 70 12.7 (2.1) 14.6 (1.2) 24 MMSE Brazil [10] Van de Winckel et al Daily physical exercises AD 81.3 (4.2) 81.9 (4.2) 100 100 12.9 (5.0) 10.8 (5.0) 12 MMSE Belgium supported by music [15] Yang et al Aerobic exercise MAD 72.0 (6.7) 71.9 (7.3) 60 72 21.3 (2.2) 20.0 (3.5) 12 MMSE, NPI, China ADAS-Cog [20] Rolland et al Exercise program Mild to 82.8 (7.8) 83.1 (7.0) 72 79 9.7 (6.8) 7.9 (6.4) 54 NPI France severe AD [21] Music therapy Arroyo-Anllo et al Music intervention AD 74.4 (3.6) 75.2 (4.2) 90 95 19.3 (3.7) 19.9 (2.9) 12 MMSE Salamanca, Spain [49] Satoh et al Singing training AD 78.1 (7.0) 77.0 (6.1) 60 80 19.1 (3.9) 20.9 (3.5) 24 MMSE, NPI, Japan ADAS-Cog [9] Suzuki et al Music therapy SD 82.0 (8.4) 85.2 (4.2) 60 69 11.6 (6.6) 9.1 (7.0) 8 MMSE Japan [24] Computerized Barban et al Process-based cognitive training MCI 74.4 (5.7) 72.9 (6.0) 46 48 27.3 (2.1) 28.1 (1.4) 24 MMSE Italy, Greece, cognitive computerized pb-CT Norway, Spain training [19] Lee et al Computerized errorless learning- AD —— 86 67 15.3 (2.7) 17.6 (4.7) 6 MMSE Hongkong based memory training program (CELP) [17] Rozzini et al Multidimensional software MCI —— — — 26.0 (1.6) 26.4 (1.9) 54 MMSE, Italy NPI [25] Tarraga et al Interactive multimedia Internet- AD 75.8 (5.9) 77.4 (4.7) 87 88 20.6 (2.1) 22.5 (2.9) 24 MMSE, Spain based system (IMIS) + ADAS-Cog integrated psychostimulation program (IPP) [28] Nutrition Rondanelli et al Oily emulsion of docosahexaenoic MCI 85.3 (5.3) 86.1 (6.5) 82 79 26.2 (2.6) 26.7 (2.8) 12 MMSE Italy therapy acid (DHA)-phospholipids containing melatonin and tryptophan [30] Aisen et al Folate, vitamin b6, vitamin B12 Mild to 75.7 (8.0) 77.3 (7.9) 58 54 20.9 (3.4) 20.9 (3.7) 76 MMSE, USA moderate AD ADAS-Cog, NPI [29] Scheltens et al Medical food AD 74.1 (7.2) 73.3 (7.8) 49 51 23.8 (2.7) 24.0 (2.5) 12 MMSE Germany, Belgium, United Kingdom, United States [32] Petersen et al Vitamin E MCI 72.8 (7.3) 72.9 (7.6) 46 47 27.2 (1.9) 27.4 (1.8) 156 MMSE. US, Canada ADAS-Cog AD= Alzheimer disease, ADAS-Cog= Alzheimer Disease Assessment Scale, cognitive subscale, CCT= computerized cognitive training, CON= control group, EXP= experiment group, MAD= mild Alzheimer disease, MCI= mild cognition impairment, MMSE= Mini-Mental State Examination, MT= music therapy, NPI= Neuropsychiatric Inventory, NT= nutrition therapy, PE= physical exercise, RCT= randomized controlled trial, SD= senile dementia, —= lose data. Liang et al. Medicine (2018) 97:20 Medicine above findings will be reinforced by our analysis of previous meta-analyses. We applied a trial sequential analysis to detect the robustness and reliability of evidence for relative effectiveness of each [50–59] cognitive intervention. The trials in previous meta-analyses (PUBMED search March 1, 2017) only investigated the efficacy and safety of PE, MT, CCT, and NT, respectively, and lacked a synthesized analysis among them. By contrast, our NMA assessed PE, MT, CCT and incorporated NT using 4 pairwise MDs. Integrating indirect comparisons in our NMA resulted in higher statistical precision in scientific comparisons of cognitive interventions against a control group. This integration makes the comparison of different interventions more explicit and facilitates interpretation. The potential correlations between these 4 MDs were accounted for in our NMA, and linking to modeling of indirect comparisons provided greater statistical [60–62] power and more precise estimates. The totality of the evidence we extracted, largely based on trials in PE, MT, CCT, NT showed that further trials of cognitive interventions versus no cognitive interventions or control group are likely to have [30,32] positive effect, except for some specific trials. Based on relative effect estimates and SUCAR, nowadays, PE seems to be the most effective cognitive intervention when we consider a cognitive therapy and CCT is the most effective cognitive intervention for neuropsychiatric symptoms. The cumulative probability ranking obtained through the Bayesian NMA cannot be considered as decisive conclusion because it was probably compromised by the lack of a significant difference among the cognitive interventions. For example, PE ranked the first in cognition but did not have superiority over any of the other cognitive interventions, which might be due to the fact that PE-relevant studies contributed a relatively greater deal of evidence in the network (6 out of 15 studies), and thus significant differences between these cognitive interventions were not found. Previous studies have consistently demonstrated that almost all of these 4 cognitive interventions have beneficial effects on older adults with AD or MCI, PE in particular. Various kinds of [10–12,14–16,20,22,26,27] moderate PE including “Walking pro- gram”, “Whole-body vibration”, “Treadmill training” had demonstrated that it was useful for AD and MCI through improvement in cognitive function or other areas. Some previous studies demonstrated low intensity or multiple exercise were able to improve neuropsychiatric symptoms in older adults with AD [63,64] or MCI. moderate-to-high intensity PE can also improve [22] cognitive ability. It seemed that PE as a relatively common intervention can effectively improve the core domains in patients of AD. However, NT is a double-edged sword, because it can Figure 2. Risk of bias assessment. improve the cognitive ability but at the same time may cause some [29,30,32] AEs such as vomiting and diarrhea, which might explain why NT did not make an obvious improvement in cognitive ability of older adults with AD or MCI. By contrast, CCT as a relatively safe and inexpensive cognitive intervention has been 4. Discussion increasingly applied. Some trials using a variety of computer- [17,18,25] To the best of our knowledge, no previous study has solved the related advanced technologies to achieve CCT. It is worth problem that which cognitive intervention is the relatively best mentioning that there are 2 meta-analyses of which the subjects intervention for AD or MCI. Therefore, as the first NMA of were healthy older adults and which concluded that CCT were cognitive interventions for patients with AD or MCI in which moderately effective in long-term improvement of cogni- [65,66] indirect evidence was used appraising the relative effectiveness tion. Moreover, the majority of studies suggest that and safety of cognitive interventions across trials simultaneous- cognitive intervention is a long-term not temporary pro- [17,50,51,55] ly, our study attempted to summarize available data to suggest cess. However, network meta-analyses synthesizes that the highest probability of being the best intervention for various cognitive interventions, and the measures of discrepancy cognitive decline and neuropsychiatric symptoms lies in PE between them are fairly obvious. This is probably the main (SUCAR=0.45) and CCT (SUCAR=0.87) respectively. The reason for heterogeneity. NT is the only cognitive intervention we 6 Liang et al. Medicine (2018) 97:20 www.md-journal.com Figure 3. Risk of bias assessment (summary graph). Figure 4. (A) The forest plot of primary outcome (summary graph). (B) The forest plot of secondary outcome (summary graph). CCT=computerized cognitive training, CG=control group, MT=music therapy, NT=nutrition therapy, PE=physical exercise, SMD=standard mean difference. 7 Liang et al. Medicine (2018) 97:20 Medicine Figure 5. (A) Funnel plot of cognition. (B) Funnel plot of neuropsychiatric symptoms. included which may cause AEs, which is why the probability rank of this intervention was low. Although previous meta-analy- [50–59] ses provide high-quality evidence that PE, MT, CCT, NT Figure 6. (A) Network of cognitive interventions comparison of cognition for can improve cognitive ability and quality of life in people with the network meta-analysis. (B) Network of cognitive interventions comparison AD or MCI, but the trials they included only compared single of neuropsychiatric symptoms for the network meta-analysis. CCT= cognitive intervention with only a control group. By contrast, our computerized cognitive training, CG=control group, MT=music therapy, NT=nutrition therapy, PE=physical exercise. Bayesian network meta-analyses actualized the integration of different interventions. Since our results were based on indirectly randomized evidence, we were convinced that our study probably provides the best evidence of the efficacy and safety of these 4 AD and MCI are progressive neurodegenerative disorders, and cognitive interventions. are still incurable. Any cognitive intervention that could possibly In summary, PE had the best effective improvement in slow down the progressive of AD and MCI patients, it worth cognitive ability and the second best in neuropsychiatric disseminating. We may create an assumption that PE and MT as symptoms. CCT had the best result in improving neuropsychiat- a potent, convenient, selective cognitive interventions were play a ric symptoms and was relatively inexpensive. MT has a relatively positive role in helping improve the cognitive function for older low probability of being the best intervention for cognitive ability adults with AD or MCI. and neuropsychiatric symptoms. However, its safety factor and cost is relatively the best compared with other cognitive 4.1. Strengths and limitations interventions. It should be noted the efficacy of a series of nondrug interventions to improve cognitive ability of AD and Rather than only grouping various interventions into CCT or MCI patients have all been proven by research, such as estrogen human intervention, as the biggest strength, our NMA assessed [67,68] [69,70] replacement therapy, psychotherapy. But quite a each intervention individually and compared all major inter- ventions simultaneously. Then, potential bias was reduced in the number of trails were restricted by appropriate endpoints, which conduction of our review by having 2 independent authors (J-hL, resulted in this individually cognitive interventions lack of the H-bZ) scan through the search output, extract the relevant data, relevant endpoints whose efficacy we must adopt to analysis. classify each intervention, and assess the methodological quality Therefore, from our conclusion above, PE, MT, CCT all have of each trial. We performed an extensive search strategy across beneficial effects on older adults with AD and MCI, especially [58,71] [72,73] several databases and sources to obtain an adequate number of PE and MT as relatively obtained easily interventions. 8 Liang et al. Medicine (2018) 97:20 www.md-journal.com Figure 7. (A) Forest plot of cognition. (B) Forest plot of neuropsychiatric symptoms. CCT=computerized cognitive training, CG=control group, MT=music therapy, NT=nutrition therapy, PE=physical exercise. eligible studies, with no language restriction. We also extensively when they are making choices among different alternatives than searched the bibliographies of published studies. In addition, the results from multiple separate traditionally meta-analyses, cognitive intervention of CCT is complex and multifaceted and because several relevant outcomes have been assessed simulta- the number of relevant trials is very small, which proves the neously. It uses common random parameters to compare particular significance of our NMA. different interventions, which combines experimental evidence From the methodological point of view, our NMA demon- from indirectly randomized comparisons with observational strates a series of preponderance of Bayesian NMA for evidence from adjusted indirect comparisons derived from [74] comparing various cognitive interventions and for evaluating trials. the relative effectiveness and safety of multifarious interventions. The limitations of our study also need to be acknowledged. In this context, the results of our NMA are likely to be more Firstly, as the biggest limitation, the number of studies and the useful for decision makers, service commissioners and caregivers number of patients included in the study were relatively small. Figure 8. (A) Accumulate histogram of cognition. (B) Accumulate histogram of neuropsychiatric symptoms. CCT=computerized cognitive training, CG=control group, MT=music therapy, NT=nutrition therapy, PE=physical exercise. 9 Liang et al. Medicine (2018) 97:20 Medicine The studies included in our NMA used the same scale as the basis, Author contributions the outcomes of which were presented as a continuous variable. J-hL conducted the database search, screened and extracted data In the analysis section, we extracted the mean, SD, and sample for the meta-analysis, prepared extracted data for the procedures, size values at baseline and at last observation for analysis. and had primary responsibility in writing this article. LL However, a few studies lost their data, which made the number of performed statistical analysis and interpretation of data. H-bZ available studies even less. There are significant differences and LH contributed to the discussion and editing. YX critically among cognitive interventions such as in the method section. The revised the draft manuscript. All authors read and approved the above-mentioned reasons explain why the number of our final manuscript. included studies was limited. In particular, the informative Data curation: Lu Lin, Hong-bo Zhang, Lei Hang. evidence of the direct comparisons between cognitive interven- Formal analysis: Jing-hong Liang. tions was limited by the absence of relevant studies. Because no Investigation: Lu Lin, Hong-bo Zhang, Lei Hang. direct comparative trial was found through our search strategy, Methodology: Jing-hong Liang, Yong Xu. our study lacked direct evidence. No direct evidence was Project administration: Yong Xu. available when we performed NMA, and thus the evaluation Resources: Jing-hong Liang, Hong-bo Zhang, Lei Hang. of consistency could not be achieved. Secondly, only 5 of 17 Software: Jing-hong Liang, Lu Lin. studies included in the analyses were double-blind, and details of Supervision: Yong Xu. allocation were noted in 15 of 17 studies, indicating that Validation: Yong Xu, Lu Lin. publication bias and selective reporting biases could not be ruled Writing – original draft: Jing-hong Liang. out. Specific intervention regimens and patient populations Writing – review & editing: Yong Xu, Rui-xia Jia. varied across studies, which might cause heterogeneity. In addition, our study data were limited by the outcome of the intersection, a number of studies used their specific scales to References present outcomes. And quite a number of trials were restricted by [1] Zhang S, Zhang M, Cai F, et al. Biological function of Presenilin and its appropriate endpoints, which resulted in individual cognitive role in AD pathogenesis. Transl Neurodegener 2013;2:15. intervention lacking the relevant endpoints we must adopt, for [2] Di Iulio F, Palmer K, Blundo C, et al. Occurrence of neuropsychiatric example MT lack NPI data that only 5 studies included. That is symptoms and psychiatric disorders in mild Alzheimer’s disease and mild cognitive impairment subtypes. Int Psychogeriatr 2010;22:629–40. the reason why we could not evaluate this intervention [3] Knopman DS, Beiser A, Machulda MM, et al. Spectrum of cognition objectively in the end. Moreover, similar to previous traditional short of dementia: Framingham Heart Study and Mayo Clinic Study of meta-analyses, our study yielded heterogeneity due to the small Aging. Neurology 2015;85:1712–21. number of studies, although funnel plots did not suggest presence [4] Albert MS, DeKosky ST, Dickson D, et al. The diagnosis of mild of heterogeneity and an extensive search strategy was used to cognitive impairment due to Alzheimer’s disease: recommendations from the National Institute on Aging-Alzheimer’s Association workgroups on identify relevant trials. The NMA is complex, and it is difficult for diagnostic guidelines for Alzheimer’s disease. Alzheimers Dement decision makers to explain the results. For example, under the 2011;7:270–9. accumulate histogram analysis of MMSE, PE was most likely to [5] Prince M, Wimo A, Guerchet M, et al. World Alzheimer Report 2015. be the best cognitive intervention. This might be because the PE [6] Katz MJ, Lipton RB, Hall CB, et al. Age-specific and sex-specific intervention had a relative larger sample size used for NMA. At prevalence and incidence of mild cognitive impairment, dementia, and last, we have not extracted the number of patients who have been Alzheimer dementia in blacks and whites: a report from the Einstein observed in trials of numerous AEs or other reasons, which is Aging Study. Alzheimer Dis Assoc Disord 2012;26:335–43. [29,30,32] because only three studies mentioned the AEs and we [7] Veroniki AA, Straus SE, Ashoor HM, et al. Comparative safety and effectiveness of cognitive enhancers for Alzheimer’s dementia: protocol could not evaluate them in this analysis. for a systematic review and individual patient data network meta- analysis. BMJ Open 2016;6:e010251. 5. Conclusion [8] Li CH, Liu CK, Yang YH, et al. Adjunct effect of music therapy on cognition in Alzheimer’s disease in Taiwan: a pilot study. Neuropsychiatr In conclusion, our NMA suggested that PE is the optimum Dis Treat 2015;11:291–6. cognitive intervention for patients with AD or MCI while CCT is [9] Suzuki M, Kanamori M, Watanabe M, et al. Behavioral and endocrinological evaluation of music therapy for elderly patients with the optimum one for neuropsychiatric symptoms. Relatively dementia. Nurs Health Sci 2004;6:11–8. speaking, MT is the most safe intervention but its efficacy is [10] Van de Winckel A, Feys H, De Weerdt W, et al. Cognitive and moderate. And NT is the last choice to manage AD or MCI behavioural effects of music-based exercises in patients with dementia. because of its 2-sidedness. The results of our NMA suggest that Clin Rehabil 2004;18:253–60. [11] Vreugdenhil A, Cannell J, Davies A, et al. A community-based exercise nonpharmacological therapies are better than pharmacological programme to improve functional ability in people with Alzheimer’s therapies. In the future, there is a need to include more studies of disease: a randomized controlled trial. Scand J Caring Sci 2012;26:12–9. high methodological quality related to comprehensive cognitive [12] Christofoletti G, Oliani MM, Gobbi S, et al. A controlled clinical trial on interventions to help establish a more extensive literature the effects of motor intervention on balance and cognition in foundation. Researchers should shift their research interest to institutionalized elderly patients with dementia. Clin Rehabil 2008;22: 618–26. outcomes other than MMSE, such as ADL and Clinical Dementia [13] Kim HJ, Yang Y, Oh JG, et al. Effectiveness of a community-based Rating (CDR) so that more network meta-analyses of cognitive multidomain cognitive intervention program in patients with Alzheimer’s interventions for AD or MCI patients can be performed. disease. Geriatr Gerontol Int 2016;16:191–9. [14] Lam FM, Liao LR, Kwok TC, et al. Effects of adding whole-body vibration to routine day activity program on physical functioning in Acknowledgments elderly with mild or moderate dementia: a randomized controlled trial. Int J Geriatr Psychiatry 2017;33:21–30. We would like to gratefully acknowledge the help by the [15] Yang SY, Shan CL, Qing H, et al. The effects of aerobic exercise on following authors: Chun-Hua Zhao, Deng-Juan Qian. All cognitive function of Alzheimer’s disease patients. CNS Neurol Disord Drug Targets 2015;14:1292–7. authors agreed to contribute to this study. 10 Liang et al. Medicine (2018) 97:20 www.md-journal.com [16] Holthoff VA, Marschner K, Scharf M, et al. Effects of physical activity [41] Higgins JP, Thompson SG. Quantifying heterogeneity in a meta-analysis. training in patients with Alzheimer’s dementia: results of a pilot RCT Stat Med 2002;21:1539–58. study. PLoS ONE 2015;10:e0121478. [42] Salanti G, Ades AE, Ioannidis JP. Graphical methods and numerical [17] Rozzini L, Costardi D, Chilovi BV, et al. Efficacy of cognitive summaries for presenting results from multiple-treatment meta-analysis: rehabilitation in patients with mild cognitive impairment treated with an overview and tutorial. J Clin Epidemiol 2011;64:163–71. cholinesterase inhibitors. Int J Geriatr Psychiatry 2007;22:356–60. [43] Mavridis D, Salanti G. A practical introduction to multivariate meta- [18] Fernandez-Calvo B, Rodriguez-Perez R, Contador I, et al. [Efficacy of analysis. Stat Methods Med Res 2013;22:133–58. cognitive training programs based on new software technologies in [44] Brooks SP, Gelman A. General methods for monitoring convergence of patients with Alzheimer-type dementia]. Psicothema 2011;23:44–50. iterative simulations. J Comput Graph Stat 1997;7:434–55. [19] Lee GY, Yip CC, Yu EC, et al. Evaluation of a computer-assisted [45] Dias S, Welton NJ, Sutton AJ, et al. NICE Decision Support Unit errorless learning-based memory training program for patients with early Technical Support Documents. A Generalised Linear Modelling Alzheimer’s disease in Hong Kong: a pilot study. Clin Interv Aging Framework for Pairwise and Network Meta-Analysis of Randomised 2013;8:623–33. Controlled Trials. London: National Institute for Health and Care [20] Rolland Y, Pillard F, Klapouszczak A, et al. Exercise program for nursing Excellence (NICE) unless otherwise stated. All rights reserved, home residents with Alzheimer’s disease: a 1-year randomized, 2014. controlled trial. J Am Geriatr Soc 2007;55:158–65. [46] Salanti G, Marinho V, Higgins JP. A case study of multiple-treatments [21] Arroyo-Anllo EM, Diaz JP, Gil R. Familiar music as an enhancer of self- meta-analysis demonstrates that covariates should be considered. J Clin consciousness in patients with Alzheimer’s disease. Biomed Res Int Epidemiol 2009;62:857–64. 2013;2013:752965. [47] Higgins JPT, Green S (Eds.). Cochrane Handbook for Systematic [22] Hoffmann K, Sobol NA, Frederiksen KS, et al. Moderate-to-high Reviews of Interventions Version 5.1.0. March 2011. intensity physical exercise in patients with Alzheimer’s disease: a [48] Egger M, Juni P, Bartlett C, et al. How important are comprehensive randomized controlled trial. J Alzheimers Dis 2016;50:443–53. literature searches and the assessment of trial quality in systematic [23] Gomez Gallego M, Gomez Garcia J. Music therapy and Alzheimer’s reviews? Empirical study. Health Technol Assess 2003;7:1–76. disease: cognitive, psychological, and behavioural effects. Neurologia [49] Satoh M, Ogawa JI, Tokita T, et al. The effects of physical exercise with (Barcelona, Spain) 2016;32:300–8. music on cognitive function of elderly people: Mihama-Kiho project. [24] Barban F, Annicchiarico R, Pantelopoulos S, et al. Protecting cognition PloS one 2014. from aging and Alzheimer’s disease: a computerized cognitive training [50] Hill NT, Mowszowski L, Naismith SL, et al. Computerized cognitive combined with reminiscence therapy. Int J Geriatr Psychiatry training in older adults with mild cognitive impairment or dementia: a 2016;31:340–8. systematic review and meta-analysis. Am J Psychiatry 2017;174: [25] Tarraga L, Boada M, Modinos G, et al. A randomised pilot study to 329–40. assess the efficacy of an interactive, multimedia tool of cognitive [51] Garcia-Casal JA, Loizeau A, Csipke E, et al. Computer-based cognitive stimulation in Alzheimer’s disease. J Neurol Neurosurg Psychiatry interventions for people living with dementia: a systematic literature 2006;77:1116–21. review and meta-analysis. Aging Mental Health 2017;21:454–67. [26] Venturelli M, Scarsini R, Schena F. Six-month walking program changes [52] Fang R, Ye S, Huangfu J, et al. Music therapy is a potential intervention cognitive and ADL performance in patients with Alzheimer. Am J for cognition of Alzheimer’s disease: a mini-review. Transl Neurodegener Alzheimers Dis Other Demen 2011;26:381–8. 2017;6:2. [27] Arcoverde C, Deslandes A, Moraes H, et al. Treadmill training as an [53] Groot C, Hooghiemstra AM, Raijmakers PG, et al. The effect of physical augmentation treatment for Alzheimer’s disease: a pilot randomized activity on cognitive function in patients with dementia: a meta-analysis controlled study. Arq Neuropsiquiatr 2014;72:190–6. of randomized control trials. Ageing Res Rev 2016;25:13–23. [28] Rondanelli M, Opizzi A, Faliva M, et al. Effects of a diet integration with [54] Mohajeri MH, Troesch B, Weber P. Inadequate supply of vitamins and an oily emulsion of DHA-phospholipids containing melatonin and DHA in the elderly: implications for brain aging and Alzheimer-type tryptophan in elderly patients suffering from mild cognitive impairment. dementia. Nutrition 2015;31:261–75. Nutr Neurosci 2012;15:46–54. [55] Coyle H, Traynor V, Solowij N. Computerized and virtual reality [29] Scheltens P, Kamphuis PJ, Verhey FR, et al. Efficacy of a medical food in cognitive training for individuals at high risk of cognitive decline: mild Alzheimer’s disease: a randomized, controlled trial. Alzheimers systematic review of the literature. Am J Geriatr Psychiatry 2015;23: Dement 2010;6:1.e1–0.e1. 335–59. [30] Aisen PS, Schneider LS, Sano M, et al. High-dose B vitamin [56] Ueda T, Suzukamo Y, Sato M, et al. Effects of music therapy on supplementation and cognitive decline in Alzheimer disease: a random- behavioral and psychological symptoms of dementia: a systematic review ized controlled trial. JAMA 2008;300:1774–83. and meta-analysis. Ageing Res Rev 2013;12:628–41. [31] Gangwar AK, Rawat A, Tiwari S, et al. Role of vitamin-D in the [57] Shah R. The role of nutrition and diet in Alzheimer disease: a systematic prevention and treatment of Alzheimer’s disease. Indian J Physiol review. J Am Med Dir Assoc 2013;14:398–402. Pharmacol 2015;59:94–9. [58] Littbrand H, Stenvall M, Rosendahl E. Applicability and effects of [32] Petersen RC, Thomas RG, Grundman M, et al. Vitamin E and donepezil physical exercise on physical and cognitive functions and activities of for the treatment of mild cognitive impairment. N Engl J Med daily living among people with dementia: a systematic review. Am J Phys 2005;352:2379–88. Med Rehabil 2011;90:495–518. [33] Ford AH, Flicker L, Thomas J, et al. Vitamins B12, B6, and folic acid for [59] Heyn P, Abreu BC, Ottenbacher KJ. The effects of exercise training on onset of depressive symptoms in older men: results from a 2-year elderly persons with cognitive impairment and dementia: a meta- placebo-controlled randomized trial. J Clin Psychiatry 2008;69:1203–9. analysis. Arch Phys Med Rehabil 2004;85:1694–704. [34] Lu G, Ades AE. Combination of direct and indirect evidence in mixed [60] Ades AE, Sculpher M, Sutton A, et al. Bayesian methods for evidence treatment comparisons. Stat Med 2004;23:3105–24. synthesis in cost-effectiveness analysis. Pharmacoeconomics 2006; [35] Salanti G, Higgins JP, Ades AE, et al. Evaluation of networks of 24:1–9. randomized trials. Stat Methods Med Res 2008;17:279–301. [61] Sutton A, Ades AE, Cooper N, et al. Use of indirect and mixed treatment [36] Salanti G. Indirect and mixed-treatment comparison, network, or comparisons for technology assessment. Pharmacoeconomics 2008;26: multiple-treatments meta-analysis: many names, many benefits, many 753–67. concerns for the next generation evidence synthesis tool. Res Synth [62] Song F, Altman DG, Glenny AM, et al. Validity of indirect comparison Methods 2012;3:80–97. for estimating efficacy of competing interventions: empirical evidence [37] Jansen JP, Naci H. Is network meta-analysis as valid as standard pairwise from published meta-analyses. BMJ 2003;326:472. meta-analysis? It all depends on the distribution of effect modifiers. BMC [63] Nascimento CM, Teixeira CV, Gobbi LT, et al. A controlled clinical trial Med 2013;11:159. on the effects of exercise on neuropsychiatric disorders and instrumental [38] Caldwell DM, Ades AE, Higgins JP. Simultaneous comparison of activities in women with Alzheimer’s disease. Rev Bras Fisioter 2012;16: multiple treatments: combining direct and indirect evidence. BMJ 197–204. 2005;331:897–900. [64] Stevens J, Killeen M. A randomised controlled trial testing the impact of [39] Mills EJ, Thorlund K, Ioannidis JP. Demystifying trial networks and exercise on cognitive symptoms and disability of residents with dementia. network meta-analysis. BMJ 2013;346:f2914. Contemp Nurse 2006;21:32–40. [40] Moher D, Liberati A, Tetzlaff J, et al. Preferred reporting items for [65] Lampit A, Hallock H, Valenzuela M. Computerized cognitive training in systematic reviews and meta-analyses: the PRISMA statement. BMJ cognitively healthy older adults: a systematic review and meta-analysis of 2009;339:b2535. effect modifiers. PLoS Med 2014;11:e1001756. 11 Liang et al. Medicine (2018) 97:20 Medicine [66] Shah TM, Weinborn M, Verdile G, et al. Enhancing cognitive [70] Hirsch RD. [Socio- and psychotherapy in patients with Alzheimer functioning in healthly older adults: a systematic review of the clinical disease]. Z Gerontol Geriatr 2001;34:92–100. significance of commercially available computerized cognitive training in [71] Jensen CS, Portelius E, Hogh P, et al. Effect of physical exercise on preventing cognitive decline. Neuropsychol Rev 2017;27:62–80. markers of neuronal dysfunction in cerebrospinal fluid in patients with [67] Mulnard RA, Cotman CW, Kawas C, et al. Estrogen replacement Alzheimer’s disease. Alzheimers Dement (NY) 2017;3:284–90. therapy for treatment of mild to moderate Alzheimer disease: a [72] Särkämö T, Laitinen S, Numminen A, et al. Clinical and demographic randomized controlled trial. Alzheimer’s Disease Cooperative Study. factors associated with the cognitive and emotional efficacy of regular JAMA 2000;283:1007–15. musical activities in dementia. J Alzheimers Dis 2015;49:767–81. [68] Levine AJ, Battista M. Estrogen replacement therapy: effects on the [73] Liesk J, Hartogh T, Kalbe E. [Cognitive stimulation and music cognitive functioning and clinical course of women with Alzheimer’s intervention for people with dementia in nursing homes: a pilot study, disease. Arch Clin Neuropsychol 2004;19:769–78. problems and perspectives]. Z Gerontol Geriatr 2015;48:275–81. [69] Burns A, Guthrie E, Marino-Francis F, et al. Brief psychotherapy in [74] Cipriani A, Higgins JP, Geddes JR, et al. Conceptual and technical Alzheimer’s disease: randomised controlled trial. Br J Psychiatry challenges in network meta-analysis. Ann Intern Med 2013;159: 2005;187:143–7. 130–7.

Journal

Medicine – Wolters Kluwer Health

Published: May 1, 2018

There are no references for this article.