1295
Journal of Alzheimer’s Disease 42 (2014) 1295–1310 DOI 10.3233/JAD-140954 IOS Press
Midlife Vascular Risk Factors and the Risk of Alzheimer’s Disease: A Systematic Review and Meta-Analysis Xiang-Fei Menga , Jin-Tai Yua,b,c,∗ , Hui-Fu Wangb , Meng-Shan Tanc , Chong Wanga , Chen-Chen Tana and Lan Tana,b,c,∗ a Department
of Neurology, Qingdao Municipal Hospital, School of Medicine, Qingdao University, China of Neurology, Qingdao Municipal Hospital, Nanjing Medical University, China c Department of Neurology, Qingdao Municipal Hospital, College of Medicine and Pharmaceutics, Ocean University of China, China b Department
Accepted 8 May 2014
Abstract. Background/Objective: We examine whether midlife vascular risk factors (VRFs) are associated with increased risk of incident Alzheimer’s disease (AD) in a systematic review and meta-analysis of published cohort studies. Methods: Original cohort studies were included if they reported adjusted combined odds ratio (COR) and corresponding 95% confidence intervals (CIs) or enough information to quantify the association between risk for AD in late-life and baseline VRFs of midlife. Results: There were positive and significant associations between high blood pressure (COR 1.31; 95% CI: 1.01–1.70), hypercholesterolemia (COR 1.72; 95% CI: 1.32–2.24), obesity (COR 1.88; 95% CI: 1.32–2.69), and diabetes mellitus in midlife (COR 1.4; 95% CI: 1.25–1.57). Smoking and hyperhomocysteinemia (although only one high-quality paper) were also associated with an increased risk of AD generally. Conclusions: These results strengthen the epidemiological evidence that VRFs of midlife significantly increase risk for AD. Keywords: Alzheimer’s disease, diabetes mellitus, high blood pressure, hypercholesterolemia, hyperhomocysteinemia, metaanalysis, midlife vascular risk factors, obesity, smoking, systematic review
INTRODUCTION Alzheimer’s disease (AD) is the most common cause of dementia and has been identified as a research priority. A cure for AD appears unlikely when significant ∗ Correspondence to: Lan Tan and Jin-Tai Yu, Department of Neurology, Qingdao Municipal Hospital, School of Medicine, Qingdao University, No. 5 Donghai Middle Road, Qingdao, Shandong Province 266071, PR China. Tel.: +86 532 8890 5659; Fax: +86 532 85968434; E-mails: [email protected] (Lan Tan); [email protected] (Jin-Tai Yu).
cognitive loss has occurred because the neuronal networks that controlled the perturbed cognitive abilities are either dead or irreversibly damaged and replacing them, even if it were technically possible, would not reconstruct the intellectual identity of the host. Prevention of risk factors to sporadic AD appears as a more realistic stratagem, and treatment, when indicated, ideally should begin in cognitively intact individuals. The vascular hypothesis of AD, first proposed in 1993 [1] and refined over the years as more data became available [2–5], provides substantial evidence that suggests vascular risk factors (VRFs) play
ISSN 1387-2877/14/$27.50 © 2014 – IOS Press and the authors. All rights reserved
1296
X.-F. Meng et al. / Midlife Vascular Risk Factors and the Risk of Alzheimer’s Disease
a critical role in the onset and development of AD during aging. This represents an opportunity for interventions that could prevent or postpone the onset of AD with a possible major effect on public health. A number of primary studies have assessed the associations between VRFs and incidence of AD, suggesting that VRFs could be independently associated with AD; VRFs may therefore be modifiable risk factors for AD. The mass screening and management of traditional VRFs in the midlife group are general public health strategies against the onset of mild to severe cognitive impairment in advanced age. Better insights into the strength and the nature of the associations between these midlife risk factors or diseases and AD are of great importance. We therefore conducted a systematic review and meta-analysis of cohort studies describing the associations between midlife modifiable risk factors (obesity, smoking, and hyperhomocysteinemia) or midlife treatable medical conditions (high blood pressure, diabetes mellitus (DM), hypercholesterolemia) and the incidence of AD in late-life. We aimed to quantify the extent to which VRFs or vascular diseases might be associated with or might be risk factors for AD in late life, and to discuss underlying mechanisms and relevant neuropathological studies. We also discuss the limitations and gaps in current studies. METHODS AND MATERIALS Literature search We conducted a systematic literature search of PubMed, EMBASE for studies using key words including Alzheimer, dementia, weight, overweight, adiposity, obesity, diabetes mellitus, insulin resistance, metabolic syndrome, body mass index (BMI), hypertension, blood pressure, homocysteine, hyperhomocysteinemia, tobacco smoking, cigarette smoking, smoking, cholesterol, hypercholesterolemia, and dyslipidemia. We also searched the reference lists of relevant studies and previous meta-analysis as well as systematic review for additional studies. We required that the articles were published in English (1966 to January 2014). Inclusion criteria and data extraction We systematically searched for studies that investigated the associations between mid-aged VRFs and AD in late life. For inclusion into the meta-analysis, papers therefore had to meet the following criteria: studies had to be original cohort studies (prospec-
tive cohort or historical cohort) with the population level (i.e., not hospital based) which were recruited at midlife (defined here as 40 to 65 years of age and with follow-up duration of at least 10 years) and exposed to aforementioned VRFs with an appropriate comparison group (a comparison group that was not exposed to VRFs per se); the study reported separate risk statistics (relative risks (RRs), odds ratios (ORs), or hazard ratios (HRs) with 95% confidence intervals (95% CIs)) for AD (excluding familial AD) by contrasted in smoking habits as well as BMI, systolic blood pressure (SBP), diastolic blood pressure (DBP), fasting blood glucose value, serum cholesterol consistent with World Health Organization (WHO) criteria for obesity, i.e., BMI ≥30 kg/m2 or for hypertension, i.e., SBP/ DBP ≥160/95 mm Hg, or for DM, i.e., >126 mg/dl (6.99 mmol/l), or for hypercholesterolemia, i.e., >251 mg/dl (6.49 mmol/l). We searched all studies evaluating the association between midlife homocysteine level and incident AD but identified only one cohort study which reported adjusted results for hyperhomocysteinemia using threshold of >12.6 mol/l. AD was mostly diagnosed according to the National Institute of Neurological and Communicative Diseases and Stroke and the Alzheimer’s Disease and Related Disorders Association (NINCDSADRDA) criteria [6], including both cases of probable and possible AD. Studies on people with cognitive impairments or other types of dementia but not AD were excluded. The titles and abstracts of all articles identified by the search were screened and potentially relevant articles were retrieved and assessed according to the criteria. Papers that addressed more than one VRF were included in multiple risk factor sections in this analysis [7–9]. When more than one paper reported on the same population (e.g., FINMONICA Study, CAIDE study) and they provided the same information [10, 11], only the paper with the largest sample size and/or the most detailed information on that risk factor and/or AD was included (the approach which was used to identify eligible studies of meta-analysis is detailed in Fig. 1). The search yielded 4,142 hits for high BP, 3,012 hits for hypercholesterolemia, 1,461 hits for obesity, 4,683 hits for diabetes, 1,080 hits for smoking, and 942 hits for hyperhomocysteinemia. 16 fit the inclusion criteria and were included for meta-analysis. A formal meta-analysis of the AD-risk estimates from the different studies investigating smoking or hyperhomocysteinemia in the middle-aged was not done because of limited number of studies (only one study of hyperhomocysteinemia and five studies of smoking) and the
X.-F. Meng et al. / Midlife Vascular Risk Factors and the Risk of Alzheimer’s Disease
1297
Fig. 1. Flowchart describing the approach used to identify all eligible studies of meta-analysis. VRFs, vascular risk factors; BP, blood pressure.
differences between studies in smoker types (e.g., former versus never smokers; heavy versus light). Some results which measured changes in these VRFs are also shown in Tables 1 and 2 but could not be used for meta-analysis. First author’s name, year of publication, participants’ age, gender rate, country of origin, follow-up time in years, numbers of participants and events, adjusted ORs, RRs, and HRs, and corresponding 95% CIs, as well as covariates adjusted in the statistical analysis were extracted from the included studies and are presented in Tables 1 and 2. Assessment of study quality and statistical analysis The qualities of reporting of all included studies were appraised with checklists developed for longitudinal observational study designs [12, 13] (Table 3). A maximum score of two points were given for each of the following five parameters: sample representatives; participation at follow-up; VRFs measurement; outcome measurement; confounding factors. We conducted meta-analyses combining RRs, HRs, and ORs for AD. We reported separate meta-analytic results for high BP, for hypercholesterolemia, for obesity, and for DM using the most fully adjusted risk statistics for AD available. In some studies, authors listed results for more than one subgroup
for the given risk factor, which were included separately in the meta-analysis [8, 14–16]. Moreover, we estimated prevalence of each risk factor in meta-analysis using published prevalence rates in United States (National Center for Health Statistics, 2012, http://www.cdc.gov/nchs/data/hus/hus12.pdf). We calculated the population attributable risk using the formula [(RR–1)PF] / [1 + (RR–1)PF] in which RR is the adjusted relative risk for a particular risk factor and PF the population fraction with that risk factor [17]. Meta-analysis was performed using Stata version 12.0 (StataCorp, College Station, TX). We used the “metan” command in Stata with the DerSimonian and Laird method [18]. Forest plots were used to visually assess the individual study odds ratio and corresponding 95% confidence intervals (CIs) across studies. A random-effects model was used to calculate the combined odds ratio (COR). To assess for heterogeneity of COR across studies, the Cochrane Q statistic (significance level of p ≤ 0.10) and the I2 statistic (25%, 50%, and 75% were considered to represent low, medium, and high heterogeneity, respectively) were calculated [19, 20]. We assessed publication bias using the method of Begg’s test and visual inspection of a funnel plot [21, 22]. Subgroup analyses were performed to estimate the relationships between midlife high BP and risk of AD because of varied definitions of high BP.
Country
FU (Y)a N Total Baseline/FU
ND/870 ND/ND
VRFs Baseline/FU
74 men 446 women 371
nb AD
477
ND/1027
Obesity Studies that compare the risk of AD in subjects above a cut-off for obesity to subjects with normal weight Whitmer et al., [15]f USA 36 ND/10136 ND/891
30
31
Japanese American
ND/ND
Stewart et al., 2007 [79]
Systolic blood pressure Studies that compare the risk of AD in subjects above a cut-off for high SBP to subjects with normal SBP Launer et al., [24] Japanese American 27 8006/3734 ND/228 118 Morris et al., [25] USA 13.6 ND/378 ND/ND 41 Kivipelto et al., [7] Finland 21 2293/1449 ND/ND 48 Ronnemaa et al., [9] Sweden 29 2268/551 ND/ND 127 Joas et al., [26] Sweden 37 1462/707 ND/ND 103 Studies that express the results per 10 mm Hg increase or per SD increase Morris et al., [25] USA 13.6 ND/378 ND/ND 41 Yamada et al., [77] Japan 30 ND/1774 ND/ND 51 Diastolic blood pressure Studies that compare the risk of AD in subjects above a cut-off for high DBP to subjects with normal DBP Launer et al., [24] Japanese American 27 8006/3734 ND/393 118 Morris et al., [25] USA 13.6 ND/378 ND/ND 41 Kivipelto et al., [7] Finland 21 2293/1449 ND/ ND 48 Studies that express the results per 10 mm Hg increase or per SD increase Morris et al., [25] USA 13.6 ND/378 ND/ND 41 Hypercholesterolemia Studies that compare the risk of AD in subjects above a cut-off for high cholesterol to subjects with normal cholesterol levels Notkola et al., [29]d Finland 23 1711/444 ND/260 27 Kivipelto et al., [7] Finland 21 2293/1449 ND/ND 48 Solomon et al., [27]e USA 30 ND/9844 ND/3147 469 Mielke et al., [28] Sweden 32 1462/648 ND/ND 46 Ronnemaa et al., [9] Sweden 29 2268/551 ND/ND 127 Studies that express the results per mmol/l increase or per SD increase Tan et al., [78] USA 30 5209/853 ND/ND 60
Hypertension Studies that compare the risk of AD in subjects with hypertension to subjects without hypertension Stewart et al., [23] Japanese American 32 ND/1890 Kimm et al., [8] c Korean 14 ND/ND
Study
Table 1 Population characteristics in studies of AD and midlife vascular risk factors
100 ND 62 ND
100 62 46 0 100 45
53 (45–68) 60 50 60
40–59 50 40–45 40–60 50 40 (Mean of first 15 measurements (1948–1978) Mean age 50 (Recruited between 1965–1968)
42.56 (40–45)
ND 27
60 43
45
100
100 ND 62 100 0
100 ND
Gender (% men)
53 (45–68) 60 estimated 50 50 45
46–68 men 52 women 54
Age Baseline
8
8
8
6 9 8 8 7
8
6 8 9
8 7
6 8 9 7 8
9 8
Quality Rating
1298 X.-F. Meng et al. / Midlife Vascular Risk Factors and the Risk of Alzheimer’s Disease
9 9 9 7 8 7 8 8
5 7 6 8 7 9
8 7 7
8
8
9
37 40.9 31 100 39 27 100 39
100 27 58.1 ND 43 50
100 40 100
43
ND
0
n, number of persons; ND, not determined; FU, follow-up; AD, Alzheimer’s disease; SD, standard deviation; HP, hypertension; SBP, systolic blood pressure; DBP, diastolic blood pressure; DM, diabetes mellitus; BMI, body-mass index; VRFs, vascular risk factors; DSM-III, Diagnostic and Statistical Manual of Mental Disorders-III; DSM- IV, Diagnostic and Statistical Manual of Mental Disorders- IV; ICD-9, International Statistical Classification of Diseases and Related Health Problems, 9th Revision; ICD-10, International Statistical Classification of Diseases and Related Health Problems, 10th Revision; NINCDS-ADRDA, National Institute of Neurological and Communicative Disease and Stroke-Alzheimer’s Disease and Related Disorders Association. a Follow-up duration of all studies of meta-analysis was at least 10 years except Wang et al. [16]. b AD diagnostic criteria according to NINCDS-ADRDA and dementia diagnostic criteria according to DSM-III or DSM-IV unless otherwise indicated. c,h AD diagnostic criteria according to ICD 10. d Dementia was diagnosed according to DSM-III-R; AD was diagnosed, if there was a gradual onset and progressive cognitive decline without other conditions known to cause irreversible dementia. e,f,g AD diagnostic criteria according to ICD 9.
Beydoun et al., [14] USA 23.4 ND/2322 ND/32 187 30 at enrollment Fitzpatrick et al., [32] USA 30.1 ND/2798 ND/261 236 50 Hassing et al., [30] Sweden 30 1465/1152 ND/ND 181 45–65 Ronnemaa et al., [9] Sweden 29 2268/551 ND/ND 127 50 Tolppanen et al., [31] Finland 26 3559/1304 ND/199 77 50.4 Studies that express the results in terms of weight change or per BMI point or SD increase Yamada et al., [77] Japan 30 ND/1774 ND/ND 51 43 Stewart et al., [80] Japanese American 32 ND/1890 ND/ND 74 45–68 Tolppanen et al., [30] Finland 26 3559/1304 ND/199 77 50.4 Diabetes Mellitus (The studies did generally not distinguish between type 1 and type 2 diabetes mellitus, but given the age of populations involved, the vast majority of the participants is likely to have type 2 DM) Curb et al., [34] Japanese American 25 8006/3734 ND/259 ND 53 (45–68) Yamada et al., [77] Japan 30 ND/1774 ND/ND 51 43 Xu et al., [33] Swedish 14 ND/ND ND/ND 16 90 mmHg)
Men HR 1.3 – (1.0–1.5) Women HR 1.0 (0.8–1.3) Studies that compare the risk of AD in subjects above a cut-off for high SBP to subjects with normal SBP. Launer et al., [24] SBP – OR 1.2 (0.4–4.0) Morris et al., [25] SBP OR 1.13 – (0.24–5.37) Kivipelto et al., [7] SBP – OR 2.6 (1.1–6.6) Ronnemaa et al., [9] SBP HR 1.0 (0.8–1.2) – Joas et al., [26]d SBP RR 4.92 – (1.35–8.48) Studies that express the results per 10 mm Hg increase or per SD increase of high SBP. Morris et al., [25] SBP OR 1.03 – (0.80–1.32) Yamada et al., [77] SBP OR 1.03 – (p < 0.001) Studies that compare the risk of AD in subjects above a cut-off for high DBP to subjects with normal DBP. Launer et al., [24]e DBP – OR 4.5 (1.5–13.1) Morris et al., [25] DBP OR 1.56 – (0.46–5.32) Kivipelto et al., [7] DBP – OR 2.0 (0.9–4.6) Studies that express the results per 10 mm Hg increase or per SD increase of high DBP. Morris et al., [25] DBP OR 1.16 – (0.75–1.81) Studies that compare the risk of AD in subjects above a cut-off for high cholesterol to subjects with normal cholesterol levels Notkola et al., [29] Hypercholesterolemia OR 2.6 (1.0–6.6) OR 3.1 (1.2–8.5) Kivipelto et al., [7] Hypercholesterolemia – OR 2.8 (1.2–6.7) Solomon et al., [27] Hypercholesterolemia – HR 1.57 (1.23–2.01) Mielke et al., [28] Hypercholesterolemia HR 1.5 HR 1.48 (0.76–2.99) (0.73–2.96) Ronnemaa et al., [9] Hypercholesterolemiaf HR 1.0 (0.9–1.2) –
Kimm et al., [8]
Studies that compare the risk of AD in subjects with hypertension to subjects without hypertension (High SBP or High DBP). Stewart et al., [23] HP (SBP >160 mmHg or – RR 0.89 DBP >95 mmHg) (0.57–1.4)
Study
Table 2 Risk of AD in patients with vascular risk factors—longitudinal studies with late-life assessment
–
BMI, DBP
APOE APOE, smoking, alcohol race/ethnic group, BMI, DM, HP, VaD
–
APOE, smoking, alcohol
VaD, APOE, smoking, alcohol –
–
–
APOE, smoking, alcohol – –
VaD, APOE, smoking, alcohol –
His stroke, his hyp, DM, smoking, BMI, impaired physical function, depressive symptoms –
Additional adjustmentsc
1300 X.-F. Meng et al. / Midlife Vascular Risk Factors and the Risk of Alzheimer’s Disease
Obesity Obesityk Obesity
BMI (decrease)
Tolppanen et al., [31]
OR = 0.989 p = 0.809 −0.01 (−0.13 to +0.10) kg/year –
– HR 1.1 (0.9–1.3) –
Studies that express the risk of AD associated with type 2 diabetes mellitus. Curb et al., [34] DM RR 0.98 (0.48–1.99) Yamada et al., [77] DM OR 4.4 (p = 0.07) Xu et al., DM – Ronnemaa et al., [8] DM HR 1.0 (0.8–1.2) Kimm et al., [9] DM Men HR1.6 (1.2–2.1) Women HR 1.2 (0.8–1.7)
BMI (decrease)
Stewart et al., [80]
Studies that express the results of obesity in terms of weight change. Yamada et al., [77] BMI (increase)
Hassing et al., [30] Ronnemaa et al., [9] Tolppanen et al., [31]
– – VaD, HP, BMI. – –
– OR 2.25 (1.29–3.92) – –
Vascular factors, impaired physical function, depression HP, DM, VaD, APOE, total cholesterol level, smoking.
−0.01 (−0.13 to +0.10) kg/year HR1.20 (1.09–1.33)
–
–
Race, C-reactive protein level, interleukin 6 level, HP, DM, VaD, APOE, smoking, kilocalories expended per week HP, VaD, DM, smoking, alcohol habits – SBP, DM, VaD, APOE, total cholesterol level, smoking.
ethnicity, smoking
HP, DM, VaD, smoking, hyperlipidemia, race
vascular risk factors, weight change, alcohol intake, use of lipid-lowing agents
APOE, VaD, therapy to lower lipid levels, BMI
–
OR 1.68 (1.21–2.33) – HR 1.57 (0.75–3.29)
HR each 10 mg/dL increase = 0.95 (0.87–1.04)g Stewart et al., [79] Cholesterol –  = −0.33; p = 0.03 (change in cholesterol over timeh , fully adjusted for covariates, as a predictor of incident AD)i Studies that compare the risk of AD in subjects above a cut-off for obesity to subjects with normal weight Whitmer et al., [15] Obesity HR 1.39 Total HR 3.10 (2.19, 4.38) (1.00–1.94) Men HR 2.60(1.44–4.69) Women HR3.38(2.20–5.19) Beydoun et al., [14] Obesityj – Men HR 0.52 (0.13–2.17) Women HR 2.43 (0.84–7.02) Fitzpatrick et al., [32] Obesity – HR 1.25 (0.74–2.11)
Studies that express the results per mmol/l increase or per SD increase of high cholesterol. Tan et al., [78] Cholesterol –
X.-F. Meng et al. / Midlife Vascular Risk Factors and the Risk of Alzheimer’s Disease 1301
DM
Wang et al., [16]
Results (95% CI)a
Additionally adjusted resultsb
BMI, SBP, DBP, cholesterol, triglycerides, smoking, creatinine and vitamin B12
HR 2.13 (1.22–3.73)
Studies are listed according to year of publication, data are presented as odds ratios (OR), relative risks (RR), and hazard ratios (HR). ND, not determined; FU, follow-up; AD, Alzheimer’s disease; CI, confidence interval. a All studies adjusted for age (or plus gender, education). b Analyses additionally adjusted for other vascular risk factors are listed in this column, the variables that were adjusted for are listed in the last column. Moreover, most studies did not have information on treatment of those midlife treatable medical conditions (When data on untreated participants were available (i.e., Launer et al., [24]) these were included in the tables.). c Additional adjustments: HP, hypertension; SBP, systolic blood pressure; DBP, diastolic blood pressure; his hyp, history of hypertension; BMI, body-mass index; VaD, vascular disease (including cerebrovascular disease, stroke, cardiac disease, myocardial infarction, coronary artery disease, congestive heart failure, angina pectoris, ankle-brachial index); DM, diabetes mellitus; BMI, body-mass index; APOE, apolipoprotein E status. d Definition of high SBP was not shown in this paper. e Investigators (Launer et al., [24]) also found significantly elevated odds ratios for AD in those with untreated borderline high DBP between 90 and 94 mm Hg (OR = 3.49; 95% CI, 1.28 - 9.52), but not untreated mixed DBP (OR = 1.33; 95% CI,.54 –3.26). f Serum cholesterol > 7.0 mmol/l. g Cox proportional hazards regression. h There was a greater decline in cholesterol level from baseline in subjects with AD at follow-up. i Random effects model. j Obesity at age 50 years; Beydoun et al. [14] also reported HRs for AD for obesity at age 30 years, 35 years, 40 years, and 45 years. k Obesity determined by BMI ≥ 28 kg/m2 . l The diagnosis of diabetes was based on medical history or medication use only. m No association was found between very heavy smoking and AD which probably explained by the hardy survivor bias effect. n Data of AD were not shown in this paper.
–
– – HP, DM, BMI, VaD, high cholesterol, alcohol drinking –
OR 1.37 (0.53–3.44)
OR 1.08 (0.43–2.63)
APOE, VaD, SBP, DBP, use of antihypertensive medication, respiratory (forced expiratory volume in 1 s adjusted for height) variables
VaD, HP, DM, hyperlipidemia (except for age and gender)
Additional adjustmentsc
– – HR 2.57 (1.63–4.03)
OR 2.93 (1.37–6.53)
OR 2.40 (1.16–5.17)
– HR 1.0 (0.7–1.4) HR 2.36 (1.54–3.61)
OR 2.55 (1.22–5.58)
OR 1.17 (0.69–1.98)
Men HR1.40 (1.16–1.68) Women HR 1.34, (1.15–1.56)
OR 2.18 (1.07–4.69)
OR 1.00 (0.61–1.63)
–
AD
Men HR 1.3 (1.0–1.7) Women HR 1.4 (0.9–2.1) Studies that compare the risk of AD in subjects above a cut-off for the highest quartile to the lowest quartile. Zylberstein et al., [42] Hyperhomocysteinemia –
Kimm et al., [8]
Kivipelto et al., [36]n Ronnemaa et al., [9] Rusanen et al., [37]
medium versus light (>26.7–40.5 pack-years versus ≤26.7 pack-years) heavy versus light (>40.5–55.5 pack-years versus ≤26.7 pack-years) very heavy versus light (>55.5–156 pack-years versus ≤26.7 pack-years)n current versus never current versus never and former heavy versus nonsmokers (≥2 packs daily versus not smoking) current versus never
Studies that express the risk of AD associated with smoking Tyas et al., [35]m current versus never
Risk factors
Study
Table 2 (Continued)
1302 X.-F. Meng et al. / Midlife Vascular Risk Factors and the Risk of Alzheimer’s Disease
X.-F. Meng et al. / Midlife Vascular Risk Factors and the Risk of Alzheimer’s Disease
1303
Table 3 Quality checklist Quality parameters
Score 2
Sample representatives
Participation at follow-up
70% response or more
Vascular risk factors measurement
Recognized international criteria (the most standard method, e.g., fasting glucose/oral glucose tolerance test for diabetes mellitus) Recognized international criteria by a central consensus committee Age (or plus gender and education) and pertinent independent risk factors (vascular diseases or other closely related vascular risk factors)
Outcome measurement
Confounding factors
RESULTS
1
0
Defined geographic area or baseline response rate 50% or more Participation rate at follow-up 60–69.9% Self-report, examination, blood tests, medical records (combination of those, e.g., random glucose measurement for diabetes mellitus) Active screening with ad-hoc criteria
Only self-report or one off measurement of blood pressure or computed from insufficient data Based on medical records
Only sociodemographic variables (age or plus gender and education)
random-model COR is likely attributable mostly to the studies of hypertension as previously noted (Table 4).
Results of meta-analysis Factors considered under this heading include high BP, hypercholesterolemia, obesity, and DM of midlife. Midlife high BP We included five studies in meta-analysis that examined the association between midlife high BP and AD [7, 8, 23–25]. We performed subgroup analysis of BP to explore the association between marked differences of the method of analysis in studies and their results because of varied definitions of high BP (high SBP, or high DBP). Meta-analysis yielded a randommodel COR for AD of 1.1 (95% CI. 88–1.37; z = 0.83; p = 0.405) with hypertension (high SBP or high DBP) (Fig. 2A), of 1.77 (95% CI. 93–3.37; z = 1.74; p = 0.082) with high SBP (Fig. 2B), and a pooled COR for AD of 2.38 (95% CI 1.34–4.23; z = 2.95; p = 0.003) with high DBP (Fig. 2C). Furthermore, the COR was 1.31 (95% CI, 1.01–1.7; z = 2.06, p = 0.039; Fig. 2) for the combined results of these three subgroups on the association between high BP and AD risk. The studies of high SBP (Q = 1.47; p = .48; I2 = .0%), and of high DBP (Q = 1.96; p = .375; I2 = .0%) were homogeneous, but the studies were moderately heterogeneous of hypertension (Q = 3.89; p = .143; I2 = 48.6%). Medium heterogeneity (Q = 14.72; p = .065; I2 = 45.7%) in the
Midlife hypercholesterolemia We identified four studies on the relationship of hypercholesterolemia measured for the midle-aged to AD [7, 27–29]. After pooling these four studies, the random-model COR was 1.72 (95% CI 1.32–2.24; z = 4.05; p = .000). The low I2 statistic showed low heterogeneity for the pooled studies (Q = 3.28; p = .351; I2 = 8.5%) (Fig. 3A; Table 4). Midlife obesity Five studies of midlife obesity [14, 15, 30–32] were included in our meta-analysis. The meta-analysis also indicated that midlife obesity was associated with higher risk of incident AD (COR, 1.88; 95% CI 1.32–2.69; z = 3.49; p = .000). Again, there was evidence of medium heterogeneity (Q = 14.67; p = 0.023; I2 = 59.1%) (Fig. 3B; Table 4). Midlife diabetes mellitus We only included four studies in our analysis for the association between midlife DM and AD [8, 16, 33, 34]. The meta-analysis also demonstrated a positive and clear effect of midlife DM on the rates of AD (COR, 1.4; 95% CI 1.25–1.57; z = 5.79; p = .000). Heterogeneity was low for analysis of midlife DM. (Q = 5.59; p = .348; I2 = 10.6%) (Fig. 3C; Table 4).
1304
X.-F. Meng et al. / Midlife Vascular Risk Factors and the Risk of Alzheimer’s Disease
Fig. 2. Association of high blood pressure (HP (2A); High SBP (2B); High DBP (2C); Combined results (2) of midlife and AD risk in cohort studies. Squares indicate study-specific odds ratio (size of the square reflects the study-specific statistical weight); horizontal lines indicate 95% CIs; diamond indicates the combined odds ratio with its 95% CI. The summary estimates were obtained using a random-effects model. CI, confidence interval; COR, combined odds ratio; HP, hypertension; SBP, systolic blood pressure; DBP, diastolic blood pressure; AD, Alzheimer’s disease.
Publication bias Begg’s test did not find significant evidence for publication bias in the meta-analysis for hypertension (z = .00, p = 1.000), for hypercholesterolemia (z = 1.7, p = .089), for obesity (z = .6, p = .548), or for DM (z = 0.00, p = 1.000). Although Begg’s test did not find significant evidence for publication bias in the final COR of the four VRFs respectively, the possibility of publication bias is not fully excluded by this method. Population attributable risk of AD Population attributable risks of AD for these four VRFs are presented in Table 5. The population attributable risk was highest for midlife obesity, up to 25% of cases of late life AD. These estimated population attributable risks should be interpreted with some caution since the results are highly dependent on the data that are entered in the equation. The population attributable risks presented in Table 5 cannot simply be added up across risk factors to obtain an estimate
of the combined effects of the risk factors, because it is unlikely that these risks are independent. Nevertheless, these data demonstrate that a substantial number of AD cases can be attributed to these VRFs, indicating an evident target for therapeutic intervention. Other midlife VRFs (Smoking and Hyperhomocysteinemia) Five publications reported the risk of midlife smoking for AD, which are described in Table 1 and Table 2 [8, 9, 35–37]. The association between the amount of smoking in midlife and the risk of AD is deserved to note. Two studies [35, 37], which specifically examined this association, were consistent indicating that heavier smoking is associated with a greater AD risk. Upon autopsy in this study, it was found that the number of amyloid plaques increased with amount smoked. Two studies [35, 36] examined the interaction with apolipoprotein E (APOE) status and found discrepant results. One study [36] reported that smoking at midlife, especially among the APOE4 carriers, is associated with the risk of AD. Another study [35] found
X.-F. Meng et al. / Midlife Vascular Risk Factors and the Risk of Alzheimer’s Disease
1305
Fig. 3. Association of three vascular risk factors (hypercholesterolemia (3A); obesity (3B); diabetes mellitus (3C) of midlife and AD risk in cohort studies. Squares indicate study-specific odds ratio (size of the square reflects the study-specific statistical weight); horizontal lines indicate 95% CIs; diamond indicates the combined odds ratio with its 95% CI. The summary estimates were obtained using a random-effects model. CI, confidence interval; COR, combined odds ratio; AD, Alzheimer’s disease.
Table 4 Combined odds ratio and measures of heterogeneity and bias Vascular risk factors High blood pressure Hypercholesterolemia Obesity Diabetes Mellitus
n of studiesa 5 4 5 4
COR 1.31 1.72 1.88 1.40
95% CI 1.01–1.7 1.32–2.24 1.32–2.69 1.25–1.57
p .039 .000 .000 .000
Heterogeneity
Begg’s Test
Q
p
I2
z
p
14.72 3.28 14.67 5.59
.065 .351 .023 .348
45.7% 8.5% 59.1% 10.6%
.00 1.7 .6 0.00
1.000 .089 .548 1.000
CI, confidence interval; n, number; COR, combined odds ratio. a In some studies, authors listed results for more than one subgroup (men or women) for the given risk factor, which were included separately in the meta-analysis (Kimm et al. [8]), Beydoun et al. [14], Whitmer et al. [15], Wang et al. [16]). Papers that addressed more than one vascular risk factor were included in multiple risk factor sections in this analysis (Kivipelto et al. [7], Kimm et al. [8]).
Table 5 Risk of AD attributable to vascular risk factors Midlife (40–65 years) Risk Factor Assessment Hypertension Hypercholesterolemia Obesity Diabetes mellitus a We
relative risk for AD
Estimated prevalencea (%)
Estimated population attributable risk (%)
1.31 1.72 1.88 1.4
30–40 20–25 35–40 2–8
9–11 13–15 24–26 0.8–3.1
estimated prevalence of each risk factor using published prevalence rates in United States (National Center for Health Statistics, 2012. http://www.cdc.gov/nchs/data/hus/hus12.pdf).
1306
X.-F. Meng et al. / Midlife Vascular Risk Factors and the Risk of Alzheimer’s Disease
that current smokers without the APOE4 allele had an increased risk of AD ± cerebrovascular disease (AD with or without cerebrovascular disease). Furthermore, no association was found between smoking status and AD ± cerebrovascular disease among those current smokers with the 4 allele. The results of other studies [38, 39] are consistent with the latter. Their results indicated non-association or beneficial effects between smoking and AD in APOE4 carriers. One potential explanation is that smoking may be harmful through alternative mechanisms, but beneficial in APOE4 carriers. Previous findings that APOE4 carriers with AD have fewer nicotinic receptor binding sites and lower activity of choline acetyltransferase compared with non-carriers [40, 41] have supported this hypothesis. The relationship between midlife hyperhomocysteinemia and AD has not been explored to the same extent as between VRFs per se and AD, and we obtained only one paper which met our criteria [42] (Tables 1 and 2). This recently published high-quality study followed a group of women for 35 years and revealed that midlife high homocysteine level is an independent risk factor for the development of late-stage AD. Irrespective of whether analyzing tertiles of total homocysteine or total homocysteine as a continuous variable, and whether analyzing alone or in the full model, the results were consistent. Ultimately, it is also possible that higher total homocysteine is a predictor or marker for late-life AD rather than a causal factor. Clinical trials with better study design and good animal models of hyperhomocysteinemia which more closely reflect the human condition are urgently needed in order to establish or deny a mechanistic link between AD and midlife hyperhomocysteinemia in a conclusive manner. DISCUSSION To our knowledge, this paper provides the first systematic review and meta-analysis that examines the association of VRFs which specifically measured at midlife with AD in late-life. No association was found between high SBP and AD, whereas high DBP was strongly associated with AD. The combined effect for high BP which just reached significance is likely attributable mostly to the studies of high SBP. Hypercholesterolemia, obesity, and DM of midlife were all significantly associated with AD. Smoking and hyperhomocysteinemia (although only one highquality paper) were also associated with an increased risk of AD generally. Limitations of this meta-analysis must be considered. First, we cannot ignore the limitations of perform-
ing a meta-analysis on observational studies, especially with long-term follow-up. Design issues, such as selective survival, incident AD, and the diluted effect of VRFs on AD during follow-up to very old ages can have substantial effects. Moreover, observational primary studies cannot prove causality. There are possible noncausal explanations for this association. Genetic predisposition could also have a modulating role in the association between VRFs and AD, but thus far only the involvement of the APOE genotype has been examined. Very few of the studies adjusted for presence of the APOE4 allele, in addition to various VRFs, which together could contribute to a noncausal association between AD and VRF. Furthermore, most studies did not have information on the treatment of VRFs. Second, we also cannot ignore the limitations of clinical diagnostic criteria in the classification of dementia by pathological subtype, especially in a complex disorder such as DM and hypertension. These studies, in general, did not employ postmortem neuropathological analyses to confirm diagnoses of AD. Finally, the heterogeneity among most pooled studies was low (
Journal of Alzheimer’s Disease 42 (2014) 1295–1310 DOI 10.3233/JAD-140954 IOS Press
Midlife Vascular Risk Factors and the Risk of Alzheimer’s Disease: A Systematic Review and Meta-Analysis Xiang-Fei Menga , Jin-Tai Yua,b,c,∗ , Hui-Fu Wangb , Meng-Shan Tanc , Chong Wanga , Chen-Chen Tana and Lan Tana,b,c,∗ a Department
of Neurology, Qingdao Municipal Hospital, School of Medicine, Qingdao University, China of Neurology, Qingdao Municipal Hospital, Nanjing Medical University, China c Department of Neurology, Qingdao Municipal Hospital, College of Medicine and Pharmaceutics, Ocean University of China, China b Department
Accepted 8 May 2014
Abstract. Background/Objective: We examine whether midlife vascular risk factors (VRFs) are associated with increased risk of incident Alzheimer’s disease (AD) in a systematic review and meta-analysis of published cohort studies. Methods: Original cohort studies were included if they reported adjusted combined odds ratio (COR) and corresponding 95% confidence intervals (CIs) or enough information to quantify the association between risk for AD in late-life and baseline VRFs of midlife. Results: There were positive and significant associations between high blood pressure (COR 1.31; 95% CI: 1.01–1.70), hypercholesterolemia (COR 1.72; 95% CI: 1.32–2.24), obesity (COR 1.88; 95% CI: 1.32–2.69), and diabetes mellitus in midlife (COR 1.4; 95% CI: 1.25–1.57). Smoking and hyperhomocysteinemia (although only one high-quality paper) were also associated with an increased risk of AD generally. Conclusions: These results strengthen the epidemiological evidence that VRFs of midlife significantly increase risk for AD. Keywords: Alzheimer’s disease, diabetes mellitus, high blood pressure, hypercholesterolemia, hyperhomocysteinemia, metaanalysis, midlife vascular risk factors, obesity, smoking, systematic review
INTRODUCTION Alzheimer’s disease (AD) is the most common cause of dementia and has been identified as a research priority. A cure for AD appears unlikely when significant ∗ Correspondence to: Lan Tan and Jin-Tai Yu, Department of Neurology, Qingdao Municipal Hospital, School of Medicine, Qingdao University, No. 5 Donghai Middle Road, Qingdao, Shandong Province 266071, PR China. Tel.: +86 532 8890 5659; Fax: +86 532 85968434; E-mails: [email protected] (Lan Tan); [email protected] (Jin-Tai Yu).
cognitive loss has occurred because the neuronal networks that controlled the perturbed cognitive abilities are either dead or irreversibly damaged and replacing them, even if it were technically possible, would not reconstruct the intellectual identity of the host. Prevention of risk factors to sporadic AD appears as a more realistic stratagem, and treatment, when indicated, ideally should begin in cognitively intact individuals. The vascular hypothesis of AD, first proposed in 1993 [1] and refined over the years as more data became available [2–5], provides substantial evidence that suggests vascular risk factors (VRFs) play
ISSN 1387-2877/14/$27.50 © 2014 – IOS Press and the authors. All rights reserved
1296
X.-F. Meng et al. / Midlife Vascular Risk Factors and the Risk of Alzheimer’s Disease
a critical role in the onset and development of AD during aging. This represents an opportunity for interventions that could prevent or postpone the onset of AD with a possible major effect on public health. A number of primary studies have assessed the associations between VRFs and incidence of AD, suggesting that VRFs could be independently associated with AD; VRFs may therefore be modifiable risk factors for AD. The mass screening and management of traditional VRFs in the midlife group are general public health strategies against the onset of mild to severe cognitive impairment in advanced age. Better insights into the strength and the nature of the associations between these midlife risk factors or diseases and AD are of great importance. We therefore conducted a systematic review and meta-analysis of cohort studies describing the associations between midlife modifiable risk factors (obesity, smoking, and hyperhomocysteinemia) or midlife treatable medical conditions (high blood pressure, diabetes mellitus (DM), hypercholesterolemia) and the incidence of AD in late-life. We aimed to quantify the extent to which VRFs or vascular diseases might be associated with or might be risk factors for AD in late life, and to discuss underlying mechanisms and relevant neuropathological studies. We also discuss the limitations and gaps in current studies. METHODS AND MATERIALS Literature search We conducted a systematic literature search of PubMed, EMBASE for studies using key words including Alzheimer, dementia, weight, overweight, adiposity, obesity, diabetes mellitus, insulin resistance, metabolic syndrome, body mass index (BMI), hypertension, blood pressure, homocysteine, hyperhomocysteinemia, tobacco smoking, cigarette smoking, smoking, cholesterol, hypercholesterolemia, and dyslipidemia. We also searched the reference lists of relevant studies and previous meta-analysis as well as systematic review for additional studies. We required that the articles were published in English (1966 to January 2014). Inclusion criteria and data extraction We systematically searched for studies that investigated the associations between mid-aged VRFs and AD in late life. For inclusion into the meta-analysis, papers therefore had to meet the following criteria: studies had to be original cohort studies (prospec-
tive cohort or historical cohort) with the population level (i.e., not hospital based) which were recruited at midlife (defined here as 40 to 65 years of age and with follow-up duration of at least 10 years) and exposed to aforementioned VRFs with an appropriate comparison group (a comparison group that was not exposed to VRFs per se); the study reported separate risk statistics (relative risks (RRs), odds ratios (ORs), or hazard ratios (HRs) with 95% confidence intervals (95% CIs)) for AD (excluding familial AD) by contrasted in smoking habits as well as BMI, systolic blood pressure (SBP), diastolic blood pressure (DBP), fasting blood glucose value, serum cholesterol consistent with World Health Organization (WHO) criteria for obesity, i.e., BMI ≥30 kg/m2 or for hypertension, i.e., SBP/ DBP ≥160/95 mm Hg, or for DM, i.e., >126 mg/dl (6.99 mmol/l), or for hypercholesterolemia, i.e., >251 mg/dl (6.49 mmol/l). We searched all studies evaluating the association between midlife homocysteine level and incident AD but identified only one cohort study which reported adjusted results for hyperhomocysteinemia using threshold of >12.6 mol/l. AD was mostly diagnosed according to the National Institute of Neurological and Communicative Diseases and Stroke and the Alzheimer’s Disease and Related Disorders Association (NINCDSADRDA) criteria [6], including both cases of probable and possible AD. Studies on people with cognitive impairments or other types of dementia but not AD were excluded. The titles and abstracts of all articles identified by the search were screened and potentially relevant articles were retrieved and assessed according to the criteria. Papers that addressed more than one VRF were included in multiple risk factor sections in this analysis [7–9]. When more than one paper reported on the same population (e.g., FINMONICA Study, CAIDE study) and they provided the same information [10, 11], only the paper with the largest sample size and/or the most detailed information on that risk factor and/or AD was included (the approach which was used to identify eligible studies of meta-analysis is detailed in Fig. 1). The search yielded 4,142 hits for high BP, 3,012 hits for hypercholesterolemia, 1,461 hits for obesity, 4,683 hits for diabetes, 1,080 hits for smoking, and 942 hits for hyperhomocysteinemia. 16 fit the inclusion criteria and were included for meta-analysis. A formal meta-analysis of the AD-risk estimates from the different studies investigating smoking or hyperhomocysteinemia in the middle-aged was not done because of limited number of studies (only one study of hyperhomocysteinemia and five studies of smoking) and the
X.-F. Meng et al. / Midlife Vascular Risk Factors and the Risk of Alzheimer’s Disease
1297
Fig. 1. Flowchart describing the approach used to identify all eligible studies of meta-analysis. VRFs, vascular risk factors; BP, blood pressure.
differences between studies in smoker types (e.g., former versus never smokers; heavy versus light). Some results which measured changes in these VRFs are also shown in Tables 1 and 2 but could not be used for meta-analysis. First author’s name, year of publication, participants’ age, gender rate, country of origin, follow-up time in years, numbers of participants and events, adjusted ORs, RRs, and HRs, and corresponding 95% CIs, as well as covariates adjusted in the statistical analysis were extracted from the included studies and are presented in Tables 1 and 2. Assessment of study quality and statistical analysis The qualities of reporting of all included studies were appraised with checklists developed for longitudinal observational study designs [12, 13] (Table 3). A maximum score of two points were given for each of the following five parameters: sample representatives; participation at follow-up; VRFs measurement; outcome measurement; confounding factors. We conducted meta-analyses combining RRs, HRs, and ORs for AD. We reported separate meta-analytic results for high BP, for hypercholesterolemia, for obesity, and for DM using the most fully adjusted risk statistics for AD available. In some studies, authors listed results for more than one subgroup
for the given risk factor, which were included separately in the meta-analysis [8, 14–16]. Moreover, we estimated prevalence of each risk factor in meta-analysis using published prevalence rates in United States (National Center for Health Statistics, 2012, http://www.cdc.gov/nchs/data/hus/hus12.pdf). We calculated the population attributable risk using the formula [(RR–1)PF] / [1 + (RR–1)PF] in which RR is the adjusted relative risk for a particular risk factor and PF the population fraction with that risk factor [17]. Meta-analysis was performed using Stata version 12.0 (StataCorp, College Station, TX). We used the “metan” command in Stata with the DerSimonian and Laird method [18]. Forest plots were used to visually assess the individual study odds ratio and corresponding 95% confidence intervals (CIs) across studies. A random-effects model was used to calculate the combined odds ratio (COR). To assess for heterogeneity of COR across studies, the Cochrane Q statistic (significance level of p ≤ 0.10) and the I2 statistic (25%, 50%, and 75% were considered to represent low, medium, and high heterogeneity, respectively) were calculated [19, 20]. We assessed publication bias using the method of Begg’s test and visual inspection of a funnel plot [21, 22]. Subgroup analyses were performed to estimate the relationships between midlife high BP and risk of AD because of varied definitions of high BP.
Country
FU (Y)a N Total Baseline/FU
ND/870 ND/ND
VRFs Baseline/FU
74 men 446 women 371
nb AD
477
ND/1027
Obesity Studies that compare the risk of AD in subjects above a cut-off for obesity to subjects with normal weight Whitmer et al., [15]f USA 36 ND/10136 ND/891
30
31
Japanese American
ND/ND
Stewart et al., 2007 [79]
Systolic blood pressure Studies that compare the risk of AD in subjects above a cut-off for high SBP to subjects with normal SBP Launer et al., [24] Japanese American 27 8006/3734 ND/228 118 Morris et al., [25] USA 13.6 ND/378 ND/ND 41 Kivipelto et al., [7] Finland 21 2293/1449 ND/ND 48 Ronnemaa et al., [9] Sweden 29 2268/551 ND/ND 127 Joas et al., [26] Sweden 37 1462/707 ND/ND 103 Studies that express the results per 10 mm Hg increase or per SD increase Morris et al., [25] USA 13.6 ND/378 ND/ND 41 Yamada et al., [77] Japan 30 ND/1774 ND/ND 51 Diastolic blood pressure Studies that compare the risk of AD in subjects above a cut-off for high DBP to subjects with normal DBP Launer et al., [24] Japanese American 27 8006/3734 ND/393 118 Morris et al., [25] USA 13.6 ND/378 ND/ND 41 Kivipelto et al., [7] Finland 21 2293/1449 ND/ ND 48 Studies that express the results per 10 mm Hg increase or per SD increase Morris et al., [25] USA 13.6 ND/378 ND/ND 41 Hypercholesterolemia Studies that compare the risk of AD in subjects above a cut-off for high cholesterol to subjects with normal cholesterol levels Notkola et al., [29]d Finland 23 1711/444 ND/260 27 Kivipelto et al., [7] Finland 21 2293/1449 ND/ND 48 Solomon et al., [27]e USA 30 ND/9844 ND/3147 469 Mielke et al., [28] Sweden 32 1462/648 ND/ND 46 Ronnemaa et al., [9] Sweden 29 2268/551 ND/ND 127 Studies that express the results per mmol/l increase or per SD increase Tan et al., [78] USA 30 5209/853 ND/ND 60
Hypertension Studies that compare the risk of AD in subjects with hypertension to subjects without hypertension Stewart et al., [23] Japanese American 32 ND/1890 Kimm et al., [8] c Korean 14 ND/ND
Study
Table 1 Population characteristics in studies of AD and midlife vascular risk factors
100 ND 62 ND
100 62 46 0 100 45
53 (45–68) 60 50 60
40–59 50 40–45 40–60 50 40 (Mean of first 15 measurements (1948–1978) Mean age 50 (Recruited between 1965–1968)
42.56 (40–45)
ND 27
60 43
45
100
100 ND 62 100 0
100 ND
Gender (% men)
53 (45–68) 60 estimated 50 50 45
46–68 men 52 women 54
Age Baseline
8
8
8
6 9 8 8 7
8
6 8 9
8 7
6 8 9 7 8
9 8
Quality Rating
1298 X.-F. Meng et al. / Midlife Vascular Risk Factors and the Risk of Alzheimer’s Disease
9 9 9 7 8 7 8 8
5 7 6 8 7 9
8 7 7
8
8
9
37 40.9 31 100 39 27 100 39
100 27 58.1 ND 43 50
100 40 100
43
ND
0
n, number of persons; ND, not determined; FU, follow-up; AD, Alzheimer’s disease; SD, standard deviation; HP, hypertension; SBP, systolic blood pressure; DBP, diastolic blood pressure; DM, diabetes mellitus; BMI, body-mass index; VRFs, vascular risk factors; DSM-III, Diagnostic and Statistical Manual of Mental Disorders-III; DSM- IV, Diagnostic and Statistical Manual of Mental Disorders- IV; ICD-9, International Statistical Classification of Diseases and Related Health Problems, 9th Revision; ICD-10, International Statistical Classification of Diseases and Related Health Problems, 10th Revision; NINCDS-ADRDA, National Institute of Neurological and Communicative Disease and Stroke-Alzheimer’s Disease and Related Disorders Association. a Follow-up duration of all studies of meta-analysis was at least 10 years except Wang et al. [16]. b AD diagnostic criteria according to NINCDS-ADRDA and dementia diagnostic criteria according to DSM-III or DSM-IV unless otherwise indicated. c,h AD diagnostic criteria according to ICD 10. d Dementia was diagnosed according to DSM-III-R; AD was diagnosed, if there was a gradual onset and progressive cognitive decline without other conditions known to cause irreversible dementia. e,f,g AD diagnostic criteria according to ICD 9.
Beydoun et al., [14] USA 23.4 ND/2322 ND/32 187 30 at enrollment Fitzpatrick et al., [32] USA 30.1 ND/2798 ND/261 236 50 Hassing et al., [30] Sweden 30 1465/1152 ND/ND 181 45–65 Ronnemaa et al., [9] Sweden 29 2268/551 ND/ND 127 50 Tolppanen et al., [31] Finland 26 3559/1304 ND/199 77 50.4 Studies that express the results in terms of weight change or per BMI point or SD increase Yamada et al., [77] Japan 30 ND/1774 ND/ND 51 43 Stewart et al., [80] Japanese American 32 ND/1890 ND/ND 74 45–68 Tolppanen et al., [30] Finland 26 3559/1304 ND/199 77 50.4 Diabetes Mellitus (The studies did generally not distinguish between type 1 and type 2 diabetes mellitus, but given the age of populations involved, the vast majority of the participants is likely to have type 2 DM) Curb et al., [34] Japanese American 25 8006/3734 ND/259 ND 53 (45–68) Yamada et al., [77] Japan 30 ND/1774 ND/ND 51 43 Xu et al., [33] Swedish 14 ND/ND ND/ND 16 90 mmHg)
Men HR 1.3 – (1.0–1.5) Women HR 1.0 (0.8–1.3) Studies that compare the risk of AD in subjects above a cut-off for high SBP to subjects with normal SBP. Launer et al., [24] SBP – OR 1.2 (0.4–4.0) Morris et al., [25] SBP OR 1.13 – (0.24–5.37) Kivipelto et al., [7] SBP – OR 2.6 (1.1–6.6) Ronnemaa et al., [9] SBP HR 1.0 (0.8–1.2) – Joas et al., [26]d SBP RR 4.92 – (1.35–8.48) Studies that express the results per 10 mm Hg increase or per SD increase of high SBP. Morris et al., [25] SBP OR 1.03 – (0.80–1.32) Yamada et al., [77] SBP OR 1.03 – (p < 0.001) Studies that compare the risk of AD in subjects above a cut-off for high DBP to subjects with normal DBP. Launer et al., [24]e DBP – OR 4.5 (1.5–13.1) Morris et al., [25] DBP OR 1.56 – (0.46–5.32) Kivipelto et al., [7] DBP – OR 2.0 (0.9–4.6) Studies that express the results per 10 mm Hg increase or per SD increase of high DBP. Morris et al., [25] DBP OR 1.16 – (0.75–1.81) Studies that compare the risk of AD in subjects above a cut-off for high cholesterol to subjects with normal cholesterol levels Notkola et al., [29] Hypercholesterolemia OR 2.6 (1.0–6.6) OR 3.1 (1.2–8.5) Kivipelto et al., [7] Hypercholesterolemia – OR 2.8 (1.2–6.7) Solomon et al., [27] Hypercholesterolemia – HR 1.57 (1.23–2.01) Mielke et al., [28] Hypercholesterolemia HR 1.5 HR 1.48 (0.76–2.99) (0.73–2.96) Ronnemaa et al., [9] Hypercholesterolemiaf HR 1.0 (0.9–1.2) –
Kimm et al., [8]
Studies that compare the risk of AD in subjects with hypertension to subjects without hypertension (High SBP or High DBP). Stewart et al., [23] HP (SBP >160 mmHg or – RR 0.89 DBP >95 mmHg) (0.57–1.4)
Study
Table 2 Risk of AD in patients with vascular risk factors—longitudinal studies with late-life assessment
–
BMI, DBP
APOE APOE, smoking, alcohol race/ethnic group, BMI, DM, HP, VaD
–
APOE, smoking, alcohol
VaD, APOE, smoking, alcohol –
–
–
APOE, smoking, alcohol – –
VaD, APOE, smoking, alcohol –
His stroke, his hyp, DM, smoking, BMI, impaired physical function, depressive symptoms –
Additional adjustmentsc
1300 X.-F. Meng et al. / Midlife Vascular Risk Factors and the Risk of Alzheimer’s Disease
Obesity Obesityk Obesity
BMI (decrease)
Tolppanen et al., [31]
OR = 0.989 p = 0.809 −0.01 (−0.13 to +0.10) kg/year –
– HR 1.1 (0.9–1.3) –
Studies that express the risk of AD associated with type 2 diabetes mellitus. Curb et al., [34] DM RR 0.98 (0.48–1.99) Yamada et al., [77] DM OR 4.4 (p = 0.07) Xu et al., DM – Ronnemaa et al., [8] DM HR 1.0 (0.8–1.2) Kimm et al., [9] DM Men HR1.6 (1.2–2.1) Women HR 1.2 (0.8–1.7)
BMI (decrease)
Stewart et al., [80]
Studies that express the results of obesity in terms of weight change. Yamada et al., [77] BMI (increase)
Hassing et al., [30] Ronnemaa et al., [9] Tolppanen et al., [31]
– – VaD, HP, BMI. – –
– OR 2.25 (1.29–3.92) – –
Vascular factors, impaired physical function, depression HP, DM, VaD, APOE, total cholesterol level, smoking.
−0.01 (−0.13 to +0.10) kg/year HR1.20 (1.09–1.33)
–
–
Race, C-reactive protein level, interleukin 6 level, HP, DM, VaD, APOE, smoking, kilocalories expended per week HP, VaD, DM, smoking, alcohol habits – SBP, DM, VaD, APOE, total cholesterol level, smoking.
ethnicity, smoking
HP, DM, VaD, smoking, hyperlipidemia, race
vascular risk factors, weight change, alcohol intake, use of lipid-lowing agents
APOE, VaD, therapy to lower lipid levels, BMI
–
OR 1.68 (1.21–2.33) – HR 1.57 (0.75–3.29)
HR each 10 mg/dL increase = 0.95 (0.87–1.04)g Stewart et al., [79] Cholesterol –  = −0.33; p = 0.03 (change in cholesterol over timeh , fully adjusted for covariates, as a predictor of incident AD)i Studies that compare the risk of AD in subjects above a cut-off for obesity to subjects with normal weight Whitmer et al., [15] Obesity HR 1.39 Total HR 3.10 (2.19, 4.38) (1.00–1.94) Men HR 2.60(1.44–4.69) Women HR3.38(2.20–5.19) Beydoun et al., [14] Obesityj – Men HR 0.52 (0.13–2.17) Women HR 2.43 (0.84–7.02) Fitzpatrick et al., [32] Obesity – HR 1.25 (0.74–2.11)
Studies that express the results per mmol/l increase or per SD increase of high cholesterol. Tan et al., [78] Cholesterol –
X.-F. Meng et al. / Midlife Vascular Risk Factors and the Risk of Alzheimer’s Disease 1301
DM
Wang et al., [16]
Results (95% CI)a
Additionally adjusted resultsb
BMI, SBP, DBP, cholesterol, triglycerides, smoking, creatinine and vitamin B12
HR 2.13 (1.22–3.73)
Studies are listed according to year of publication, data are presented as odds ratios (OR), relative risks (RR), and hazard ratios (HR). ND, not determined; FU, follow-up; AD, Alzheimer’s disease; CI, confidence interval. a All studies adjusted for age (or plus gender, education). b Analyses additionally adjusted for other vascular risk factors are listed in this column, the variables that were adjusted for are listed in the last column. Moreover, most studies did not have information on treatment of those midlife treatable medical conditions (When data on untreated participants were available (i.e., Launer et al., [24]) these were included in the tables.). c Additional adjustments: HP, hypertension; SBP, systolic blood pressure; DBP, diastolic blood pressure; his hyp, history of hypertension; BMI, body-mass index; VaD, vascular disease (including cerebrovascular disease, stroke, cardiac disease, myocardial infarction, coronary artery disease, congestive heart failure, angina pectoris, ankle-brachial index); DM, diabetes mellitus; BMI, body-mass index; APOE, apolipoprotein E status. d Definition of high SBP was not shown in this paper. e Investigators (Launer et al., [24]) also found significantly elevated odds ratios for AD in those with untreated borderline high DBP between 90 and 94 mm Hg (OR = 3.49; 95% CI, 1.28 - 9.52), but not untreated mixed DBP (OR = 1.33; 95% CI,.54 –3.26). f Serum cholesterol > 7.0 mmol/l. g Cox proportional hazards regression. h There was a greater decline in cholesterol level from baseline in subjects with AD at follow-up. i Random effects model. j Obesity at age 50 years; Beydoun et al. [14] also reported HRs for AD for obesity at age 30 years, 35 years, 40 years, and 45 years. k Obesity determined by BMI ≥ 28 kg/m2 . l The diagnosis of diabetes was based on medical history or medication use only. m No association was found between very heavy smoking and AD which probably explained by the hardy survivor bias effect. n Data of AD were not shown in this paper.
–
– – HP, DM, BMI, VaD, high cholesterol, alcohol drinking –
OR 1.37 (0.53–3.44)
OR 1.08 (0.43–2.63)
APOE, VaD, SBP, DBP, use of antihypertensive medication, respiratory (forced expiratory volume in 1 s adjusted for height) variables
VaD, HP, DM, hyperlipidemia (except for age and gender)
Additional adjustmentsc
– – HR 2.57 (1.63–4.03)
OR 2.93 (1.37–6.53)
OR 2.40 (1.16–5.17)
– HR 1.0 (0.7–1.4) HR 2.36 (1.54–3.61)
OR 2.55 (1.22–5.58)
OR 1.17 (0.69–1.98)
Men HR1.40 (1.16–1.68) Women HR 1.34, (1.15–1.56)
OR 2.18 (1.07–4.69)
OR 1.00 (0.61–1.63)
–
AD
Men HR 1.3 (1.0–1.7) Women HR 1.4 (0.9–2.1) Studies that compare the risk of AD in subjects above a cut-off for the highest quartile to the lowest quartile. Zylberstein et al., [42] Hyperhomocysteinemia –
Kimm et al., [8]
Kivipelto et al., [36]n Ronnemaa et al., [9] Rusanen et al., [37]
medium versus light (>26.7–40.5 pack-years versus ≤26.7 pack-years) heavy versus light (>40.5–55.5 pack-years versus ≤26.7 pack-years) very heavy versus light (>55.5–156 pack-years versus ≤26.7 pack-years)n current versus never current versus never and former heavy versus nonsmokers (≥2 packs daily versus not smoking) current versus never
Studies that express the risk of AD associated with smoking Tyas et al., [35]m current versus never
Risk factors
Study
Table 2 (Continued)
1302 X.-F. Meng et al. / Midlife Vascular Risk Factors and the Risk of Alzheimer’s Disease
X.-F. Meng et al. / Midlife Vascular Risk Factors and the Risk of Alzheimer’s Disease
1303
Table 3 Quality checklist Quality parameters
Score 2
Sample representatives
Participation at follow-up
70% response or more
Vascular risk factors measurement
Recognized international criteria (the most standard method, e.g., fasting glucose/oral glucose tolerance test for diabetes mellitus) Recognized international criteria by a central consensus committee Age (or plus gender and education) and pertinent independent risk factors (vascular diseases or other closely related vascular risk factors)
Outcome measurement
Confounding factors
RESULTS
1
0
Defined geographic area or baseline response rate 50% or more Participation rate at follow-up 60–69.9% Self-report, examination, blood tests, medical records (combination of those, e.g., random glucose measurement for diabetes mellitus) Active screening with ad-hoc criteria
Only self-report or one off measurement of blood pressure or computed from insufficient data Based on medical records
Only sociodemographic variables (age or plus gender and education)
random-model COR is likely attributable mostly to the studies of hypertension as previously noted (Table 4).
Results of meta-analysis Factors considered under this heading include high BP, hypercholesterolemia, obesity, and DM of midlife. Midlife high BP We included five studies in meta-analysis that examined the association between midlife high BP and AD [7, 8, 23–25]. We performed subgroup analysis of BP to explore the association between marked differences of the method of analysis in studies and their results because of varied definitions of high BP (high SBP, or high DBP). Meta-analysis yielded a randommodel COR for AD of 1.1 (95% CI. 88–1.37; z = 0.83; p = 0.405) with hypertension (high SBP or high DBP) (Fig. 2A), of 1.77 (95% CI. 93–3.37; z = 1.74; p = 0.082) with high SBP (Fig. 2B), and a pooled COR for AD of 2.38 (95% CI 1.34–4.23; z = 2.95; p = 0.003) with high DBP (Fig. 2C). Furthermore, the COR was 1.31 (95% CI, 1.01–1.7; z = 2.06, p = 0.039; Fig. 2) for the combined results of these three subgroups on the association between high BP and AD risk. The studies of high SBP (Q = 1.47; p = .48; I2 = .0%), and of high DBP (Q = 1.96; p = .375; I2 = .0%) were homogeneous, but the studies were moderately heterogeneous of hypertension (Q = 3.89; p = .143; I2 = 48.6%). Medium heterogeneity (Q = 14.72; p = .065; I2 = 45.7%) in the
Midlife hypercholesterolemia We identified four studies on the relationship of hypercholesterolemia measured for the midle-aged to AD [7, 27–29]. After pooling these four studies, the random-model COR was 1.72 (95% CI 1.32–2.24; z = 4.05; p = .000). The low I2 statistic showed low heterogeneity for the pooled studies (Q = 3.28; p = .351; I2 = 8.5%) (Fig. 3A; Table 4). Midlife obesity Five studies of midlife obesity [14, 15, 30–32] were included in our meta-analysis. The meta-analysis also indicated that midlife obesity was associated with higher risk of incident AD (COR, 1.88; 95% CI 1.32–2.69; z = 3.49; p = .000). Again, there was evidence of medium heterogeneity (Q = 14.67; p = 0.023; I2 = 59.1%) (Fig. 3B; Table 4). Midlife diabetes mellitus We only included four studies in our analysis for the association between midlife DM and AD [8, 16, 33, 34]. The meta-analysis also demonstrated a positive and clear effect of midlife DM on the rates of AD (COR, 1.4; 95% CI 1.25–1.57; z = 5.79; p = .000). Heterogeneity was low for analysis of midlife DM. (Q = 5.59; p = .348; I2 = 10.6%) (Fig. 3C; Table 4).
1304
X.-F. Meng et al. / Midlife Vascular Risk Factors and the Risk of Alzheimer’s Disease
Fig. 2. Association of high blood pressure (HP (2A); High SBP (2B); High DBP (2C); Combined results (2) of midlife and AD risk in cohort studies. Squares indicate study-specific odds ratio (size of the square reflects the study-specific statistical weight); horizontal lines indicate 95% CIs; diamond indicates the combined odds ratio with its 95% CI. The summary estimates were obtained using a random-effects model. CI, confidence interval; COR, combined odds ratio; HP, hypertension; SBP, systolic blood pressure; DBP, diastolic blood pressure; AD, Alzheimer’s disease.
Publication bias Begg’s test did not find significant evidence for publication bias in the meta-analysis for hypertension (z = .00, p = 1.000), for hypercholesterolemia (z = 1.7, p = .089), for obesity (z = .6, p = .548), or for DM (z = 0.00, p = 1.000). Although Begg’s test did not find significant evidence for publication bias in the final COR of the four VRFs respectively, the possibility of publication bias is not fully excluded by this method. Population attributable risk of AD Population attributable risks of AD for these four VRFs are presented in Table 5. The population attributable risk was highest for midlife obesity, up to 25% of cases of late life AD. These estimated population attributable risks should be interpreted with some caution since the results are highly dependent on the data that are entered in the equation. The population attributable risks presented in Table 5 cannot simply be added up across risk factors to obtain an estimate
of the combined effects of the risk factors, because it is unlikely that these risks are independent. Nevertheless, these data demonstrate that a substantial number of AD cases can be attributed to these VRFs, indicating an evident target for therapeutic intervention. Other midlife VRFs (Smoking and Hyperhomocysteinemia) Five publications reported the risk of midlife smoking for AD, which are described in Table 1 and Table 2 [8, 9, 35–37]. The association between the amount of smoking in midlife and the risk of AD is deserved to note. Two studies [35, 37], which specifically examined this association, were consistent indicating that heavier smoking is associated with a greater AD risk. Upon autopsy in this study, it was found that the number of amyloid plaques increased with amount smoked. Two studies [35, 36] examined the interaction with apolipoprotein E (APOE) status and found discrepant results. One study [36] reported that smoking at midlife, especially among the APOE4 carriers, is associated with the risk of AD. Another study [35] found
X.-F. Meng et al. / Midlife Vascular Risk Factors and the Risk of Alzheimer’s Disease
1305
Fig. 3. Association of three vascular risk factors (hypercholesterolemia (3A); obesity (3B); diabetes mellitus (3C) of midlife and AD risk in cohort studies. Squares indicate study-specific odds ratio (size of the square reflects the study-specific statistical weight); horizontal lines indicate 95% CIs; diamond indicates the combined odds ratio with its 95% CI. The summary estimates were obtained using a random-effects model. CI, confidence interval; COR, combined odds ratio; AD, Alzheimer’s disease.
Table 4 Combined odds ratio and measures of heterogeneity and bias Vascular risk factors High blood pressure Hypercholesterolemia Obesity Diabetes Mellitus
n of studiesa 5 4 5 4
COR 1.31 1.72 1.88 1.40
95% CI 1.01–1.7 1.32–2.24 1.32–2.69 1.25–1.57
p .039 .000 .000 .000
Heterogeneity
Begg’s Test
Q
p
I2
z
p
14.72 3.28 14.67 5.59
.065 .351 .023 .348
45.7% 8.5% 59.1% 10.6%
.00 1.7 .6 0.00
1.000 .089 .548 1.000
CI, confidence interval; n, number; COR, combined odds ratio. a In some studies, authors listed results for more than one subgroup (men or women) for the given risk factor, which were included separately in the meta-analysis (Kimm et al. [8]), Beydoun et al. [14], Whitmer et al. [15], Wang et al. [16]). Papers that addressed more than one vascular risk factor were included in multiple risk factor sections in this analysis (Kivipelto et al. [7], Kimm et al. [8]).
Table 5 Risk of AD attributable to vascular risk factors Midlife (40–65 years) Risk Factor Assessment Hypertension Hypercholesterolemia Obesity Diabetes mellitus a We
relative risk for AD
Estimated prevalencea (%)
Estimated population attributable risk (%)
1.31 1.72 1.88 1.4
30–40 20–25 35–40 2–8
9–11 13–15 24–26 0.8–3.1
estimated prevalence of each risk factor using published prevalence rates in United States (National Center for Health Statistics, 2012. http://www.cdc.gov/nchs/data/hus/hus12.pdf).
1306
X.-F. Meng et al. / Midlife Vascular Risk Factors and the Risk of Alzheimer’s Disease
that current smokers without the APOE4 allele had an increased risk of AD ± cerebrovascular disease (AD with or without cerebrovascular disease). Furthermore, no association was found between smoking status and AD ± cerebrovascular disease among those current smokers with the 4 allele. The results of other studies [38, 39] are consistent with the latter. Their results indicated non-association or beneficial effects between smoking and AD in APOE4 carriers. One potential explanation is that smoking may be harmful through alternative mechanisms, but beneficial in APOE4 carriers. Previous findings that APOE4 carriers with AD have fewer nicotinic receptor binding sites and lower activity of choline acetyltransferase compared with non-carriers [40, 41] have supported this hypothesis. The relationship between midlife hyperhomocysteinemia and AD has not been explored to the same extent as between VRFs per se and AD, and we obtained only one paper which met our criteria [42] (Tables 1 and 2). This recently published high-quality study followed a group of women for 35 years and revealed that midlife high homocysteine level is an independent risk factor for the development of late-stage AD. Irrespective of whether analyzing tertiles of total homocysteine or total homocysteine as a continuous variable, and whether analyzing alone or in the full model, the results were consistent. Ultimately, it is also possible that higher total homocysteine is a predictor or marker for late-life AD rather than a causal factor. Clinical trials with better study design and good animal models of hyperhomocysteinemia which more closely reflect the human condition are urgently needed in order to establish or deny a mechanistic link between AD and midlife hyperhomocysteinemia in a conclusive manner. DISCUSSION To our knowledge, this paper provides the first systematic review and meta-analysis that examines the association of VRFs which specifically measured at midlife with AD in late-life. No association was found between high SBP and AD, whereas high DBP was strongly associated with AD. The combined effect for high BP which just reached significance is likely attributable mostly to the studies of high SBP. Hypercholesterolemia, obesity, and DM of midlife were all significantly associated with AD. Smoking and hyperhomocysteinemia (although only one highquality paper) were also associated with an increased risk of AD generally. Limitations of this meta-analysis must be considered. First, we cannot ignore the limitations of perform-
ing a meta-analysis on observational studies, especially with long-term follow-up. Design issues, such as selective survival, incident AD, and the diluted effect of VRFs on AD during follow-up to very old ages can have substantial effects. Moreover, observational primary studies cannot prove causality. There are possible noncausal explanations for this association. Genetic predisposition could also have a modulating role in the association between VRFs and AD, but thus far only the involvement of the APOE genotype has been examined. Very few of the studies adjusted for presence of the APOE4 allele, in addition to various VRFs, which together could contribute to a noncausal association between AD and VRF. Furthermore, most studies did not have information on the treatment of VRFs. Second, we also cannot ignore the limitations of clinical diagnostic criteria in the classification of dementia by pathological subtype, especially in a complex disorder such as DM and hypertension. These studies, in general, did not employ postmortem neuropathological analyses to confirm diagnoses of AD. Finally, the heterogeneity among most pooled studies was low (
