Table 1.

Characteristics of prospective studies investigating the association of HDL and apolipoproteins with dementia

First author, year, name of the study, country	Study characteristics	Exposure, average (baseline) concentration	Outcome (no. of cases), diagnostic criteria	Main findings	Adjustments, notes
Ancelin 2013, Three-City (3C) study, France [18]	n = 7,053 participants. Multi-site cohort study of community-dwelling persons (mean age 74 years, male 39%). Follow-up: 7y. Baseline: 1999–2001.	Fasting serum HDL-C (mmol/L): Men: overall geometric mean: 1.44 Q1: <1.19 Q2+Q3: 0.19-<1.63 Q4: ≥ 1.63 Women: overall geometric mean: 1.73. Quartile values not reported.	All cause dementia (men: n = 184; women: n = 297): Neuropsychological and neurological examination by trained psychologists and neurologists, MRI or PET if available, consensus of neurologists committee according to DSM-IV revised criteria; AD: NINCDS-ADRDA.	A tendency for higher risk of dementia in men with lowest HDL-C. HR (95% CI) for all-cause dementia in all men: Q1: 1.45 (1.03–2.04) Q2+Q3: reference Q4: 0.97 (0.67–1.39). Women: no results shown. Reported that HDL-C was not significantly associated with all-cause dementia or AD.	Age (y), sex (m; f), center (Bordeaux; Dijon; Montpellier), years of education (5; 9; 12; >12 y). Results were slightly attenuated with adjustment for additional covariates (HR in Q1 was 1.36 [0.97–1.93] after including hypertension, diabetes, depression and genotype). When restricted to men without vascular pathologies, the HR for all-cause dementia was 1.64 (1.02–2.66) among men in lowest HDL-C (Q1). Participants had to show up at exam year 2, 4 and 7 to be diagnosed with dementia.
Apostolova 2015, Alzheimer’s Disease Neuroimaging Initiative (ADNI), USA and Canada [19**]1	n = 298 subjects with MCI (mean age 75 y, male 64%). Follow up: 36 months. Baseline: 2005–2008.	Mean (SD) plasma apoE (log10 μg/mL): Individuals progressing to AD: 1.66 (0.18) Individuals not progressing to AD: 1.71 (0.18). Mean (SD) plasma apoJ (log10 μg/mL): individuals progressing to AD: 2.49 (0.07) Individuals not progressing to AD: 2.48 (0.07).	AD (n = 161): NINCDS-ADRDA	Lower baseline apoE concentration (p = 0.007) but no difference in apoJ concentration (p=0.6) in individuals progressing from MCI to AD compared to individuals not progressing from MCI to AD.	None.
Gatz 2010, Study of Dementia in Swedish Twins (SATSA), Sweden [20]	n = 30 same-sex twin pairs enrolled in Swedish Twin Registry (mean age 71 y, male n = 7 pairs). Follow-up: 3 years. Baseline: 1987	Serum HDL-C, apoA-I, and apoB concentration. Baseline levels not reported.	Incident dementia (AD: n = 18, VD: n = 4, mixed AD and vascular n = 2, secondary cause dementia n = 2, unspecified dementia n = 4) in one twin whereas the other twin remains non-dement	No difference in HDL-C (p=0.3), or apoA1 (p=0.3) concentration comparing twins. Higher apoB concentration (p=0.0062) and apoB to apoA1 ratio (p=0.006) comparing the twins developing dementia to non-dement twins.	Implicit adjustment for genetics but potentially shared familial risk factors for dementia.
Mielke 2012, Women’s Health and Aging Study II, USA [24]	n = 99 women free of dementia at baseline (mean age 74 y), Follow-up 9 years. Baseline: 1994–1995	Nonfasting serum HDL-C. Baseline level not reported.	Dementia (n = 17): DSM-IV, Probable and possible AD (n = 18): NINCDS-ADRDA.	HDL-C not significantly associated with dementia. HR (95% CI) comparing extreme tertiles was: 0.7 (0.3–1.8) for T3. In analysis restricted to probable and possible AD cases, the HR (95% CI) for T3 was T3: 1.6 (0.5–5.5).	Age (y), BMI (kg/m2), blood glucose at baseline (continuous). Non-fasting blood samples.
Rasmussen 2015, Copenhagen General Population Study (CGPS) and Copenhagen City Heart Study (CCHS), Denmark [25**]	n = 75,708 participants from two combined studies in Copenhagen. Baseline: CGPS=2003; CCHS=At blood sampling in 1991–1994 or 2001–2003. Follow-up: 2011 (mean of 4 years). No overall mean age at baseline, but mean age of non-cases was 57 years (45% male) and 73 years (40% male) for future cases.	Mean (SD) plasma HDL-C (mmol/L): All dementia: 1.2 (0.02) AD: 1.7 (0.03) No dementia: 1.6 (0.002) Median plasma apoE (mg/dL) in age and sex stratified tertiles: T1: 3.1 T2: 4.1 T3: 5.8	Register data only, based on World Health Organization International Classification of Disease, 8th revision and 10th revision. n = 1,060 progressed to dementia (443 of which to AD).	Lower HDL-C concentration in participants who developed dementia, compared to the non-cases (p< 0.001). The lowest risk of dementia was found among those with highest apoE levels. Adjusted HR (95% CI) for all dementia comparing apoE tertiles: T1: 1.22 (1.01–1.47) T2: 1.09 (0.91–1.30) T3 (highest apoE): reference HR (95% CI) for AD: T1: 1.53 (1.13–2.08) T2: 1.31 (0.98–1.76) T3 (highest apoE): reference.	Age (y), sex (m; f), body mass index (kg/m2), hypertension (yes; no), diabetes mellitus (yes; no), smoking (yes; no), alcohol consumption (yes; no), physical inactivity (yes; no), menopause status (yes; no), hormonal replacement therapy, lipid-lowering therapy (yes; no), education (<8; ≥8y), total cholesterol (mmol/L), low-density lipoprotein cholesterol (mmol/L), triglycerides (mmol/L), and HDL-C (mmol/L), APOE genotype (e4/-; e4/e4). APOE genotype was a strong confounder. Results were not modified by APOE genotype. No verification or validation of endpoints.
Reitz 2010, Northern Manhattan Study (NOMAS), USA [14]	n = 1130 elderly individuals sampled from Medicare recipients free of cognitive impairment at baseline: 1999–2001 (mean age 76 y, male: 34%). Follow-up: 4 y.	Fasting plasma HDL-C (mgl/dL) range in HDL-C quartiles: Q1: ≤38.00 Q2: 38.01–46.00 Q3: 46.01–56.00 Q4: >56.00	Follow-up assessments every 18 months and medical records. Probable and possible AD (n = 101); Probable AD (n = 98), NINCDS-ADRDA.	Lowest risk of AD among those with highest HDL-C levels. Adjusted HR (95% CI) for probable and possible AD: Q1: reference Q2: 0.8 (0.4–1.5) Q3: 1.1 (0.6–1.9) Q4: 0.4 (0.2–0.9), pfor trend=0.10 HR (95% CI) for probable AD: Q1: reference Q2: 0.8 (0.4–1.6) Q3: 0.9 (0.5–1.9) Q4: 0.4 (0.2–0.9), pfor trend=0.06.	Age (y), sex (m; f), education (y), ethnic group (White/non-Hispanic; Black/non-Hispanic; Hispanic), APOE genotype (e4/-; e4/e4), diabetes mellitus (yes;no), hypertension (yes;no), heart disease (yes;no), BMI (kg/m2), lipid lowering treatment (yes; no).
Schrijvers 2011, Rotterdam study, the Netherlands [26]1	Case-cohort among non-demented participants. Random subcohort n= 926 [n = 61 developed dementia (n = 52 AD)] and additional cases of parent cohort [n = 178 developed dementia (n = 156 AD)] during follow-up. Baseline: 1997–1999. Mean follow-up: 7y.	Fasting mean (SD) plasma apoJ (μg/mL): Subcohort at risk for dementia: 115 (25)	Incident dementia (n = 237, including n = 208 AD during follow-up: register linkage and diagnosed by a panel of neurologists, neuropsychologists at study exams every 3–4 years using NINCDS-ADRDA criteria.	No association between plasma apoJ and incident AD. All values per 1 SD increase in apoJ at baseline: HR (95% CI) per 1 SD higher apoJ: 1.00 (0.85–1.17) for AD and 0.96 (0.82–1.12) for all-cause dementia.	Age (y), sex (m; f), education level (primary education; more than primary education), APOE genotype (e4/-; e4/e4), diabetes (yes; no), smoking (currently smoking; not currently smoking), coronary heart disease (yes; no), hypertension (yes; no). In a cross-sectional analysis comparing 60 participants with AD at baseline to the non-demented, higher plasma apoJ was significantly associated with greater odds of having AD (p trend over quartiles =0.004) as well as greater severity of AD.

¹Additional reports from the ADNI study [16, 21, 22, 23, 38] and Rotterdam study [27] identified are not included in the table.

AD, Alzheimer’s disease; BMI, Body mass index; DSM-IIIR, Diagnostic and Statistical Manual of Mental Disorders, third edition, text revision; HR, Hazard Ratio; MCI, Mild cognitive impairment; NINCDS-ADRDA, National Institute of Neurological and Communicative Disorders and Stroke-Alzheimer’s Disease and Related Disorders Association; Q: Quintile. T, Tertile.