Genetic variation in clusterin and risk of dementia and ischemic vascular disease in the general population: cohort studies and meta-analyses of 362,338 individuals

Information on diagnoses of Alzheimer’s disease, all dementia, vascular dementia, ischemic cerebrovascular disease, and ischemic heart disease was collected from the national Danish Patient Registry and the national Danish Causes of Death Registry. The national Danish Patient Registry has information on all patient contacts with all clinical hospital departments in Denmark since 1977, including emergency wards and outpatient clinics from 1995. The national Danish Causes of Death Registry contains data on causes of all deaths in Denmark, as reported by hospitals and general practitioners since 1977. In the World Health Organization’s (WHO) International Classification of Diseases (ICD), Alzheimer’s disease is ICD8 290 and ICD10 F00 and G30. The validity of Alzheimer’s disease ICD codes was ensured by the presence of the well-known association with ε4 in the present cohort [18], with risk estimates similar to those reported elsewhere [3, 22]. Vascular dementia is ICD10 F01. All dementia includes Alzheimer’s disease, vascular dementia, and unspecified dementia (ICD8 290.18 and ICD10 F03). The quality of these registry-based diagnoses has previously been validated [23]. For ischemic cerebrovascular disease (ICD8 433-435 and ICD10 I63, I64, G45), validation of ICD codes has been described previously [17]. In brief, information on diagnoses of ischemic cerebrovascular disease (transitory ischemic attacks, amaurosis fugax, and ischemic stroke) was collected by reviewing all hospital admissions and diagnoses entered in the national Danish Patient Registry and the national Danish Causes of Death Registry. Possible cerebrovascular events (hospitalized as well as non-hospitalized) were validated by trained physicians using the WHO definition of cerebrovascular disease. Diagnoses of ischemic heart disease (ICD 8410-414 and ICD10 I20-I25) were collected and verified by reviewing all hospital admissions and diagnoses entered in the national Danish Patient Registry, all causes of death entered in the national Danish Causes of death Registry, and medical records from hospitals and general practitioners. Ischemic heart disease was fatal or nonfatal myocardial infarction or characteristic symptoms of angina pectoris, including revascularization procedures [24].

The follow-up period began at the time of blood sampling (2003 and onward for CGPS and 1991–1994 or 2001–2003 for CCHS). Follow-up ended at occurrence of an event, death, or emigration, or on 14 November 2014 (last update of the registry), whichever came first. Median follow-up times were 6 years (range 0–22 years) for Alzheimer’s disease, all dementia, and vascular dementia, 6 years (range 0–23 years) for ischemic cerebrovascular disease, and 5 years (range 0–22) for ischemic heart disease. No individuals were lost to follow-up.

An ABI PRISM 7900HT Sequence Detection System (Applied Biosystems Inc., Foster City, CA, USA) and Taqman-based assays were used to genotype for rs9331896 and APOE (rs7412 and rs429358). rs7412 and rs429358 defines the six common APOE genotypes (ε22, ε32, ε42, ε33, ε43, and ε44), as previously described [18].

Plasma total cholesterol, HDL cholesterol, triglycerides, apolipoprotein B, and apolipoprotein AI were measured using standard hospital assays (Boehringer Mannheim GmbH, Mannheim, Germany and Konelab, Thermo Fischer Scientific, Waltheim, MA, USA). LDL cholesterol was calculated with the Friedewald equation [25] when plasma triglycerides were ≤4.0 mmol/L (≤352 mg/dL) and otherwise measured directly (Konelab). All assays were assessed daily for precision using internal controls and four to 12 times yearly for accuracy with a Scandinavian external quality control program. ApoE was measured by nephelometry as previously described [18] with a BNII autoanalyzer using goat anti-human apoE polyclonal antibodies (OQDLG09, Dade Behring, Deerfield, Illinois, USA) or by turbidimetry with a Kone autoanalyzer (Konelab) using rabbit anti-human apoE polyclonal antibodies (A0077, Dako, Glostrup, Denmark). A human serum apoE calibrator (Apolipoprotein Standard Serum, OUPGG07, Siemens Healthcare Diagnostics, Ballerup, Denmark) was used for both nephelometry and turbidimetry.

Body mass index is the measured weight in kilograms divided by the measured height in meters squared. Hypertension was defined as systolic blood pressure ≥140 mmHg, diastolic blood pressure ≥90 mmHg, and/or use of anti-hypertensive medication. Diabetes mellitus was defined as a self-reported disease, use of insulin or oral hypoglycemic agents, and/or non-fasting plasma glucose levels of more than 11 mmol/L (>198 mg/dL). Smoking, alcohol consumption, physical inactivity, education, use of lipid-lowering therapy as well as postmenopausal status and hormonal replacement therapy in women were all self-reported and dichotomized. Smoking was current smoking. High alcohol consumption was >14/21 units of alcohol per week for women/men (1 unit alcohol ~12 g). Physical inactivity was ≤4 h of light physical exercise in leisure time weekly. Low educational level was less than eight years in school.

Cumulative incidences of Alzheimer’s disease and all dementia were plotted against age and genotype, using the method of Fine and Gray [17, 26], to account for the possibility of death as a competing event. P values were calculated using the “stcrreg” command in Stata. Similar cause-specific (censoring at death) Cox proportional hazards regression models with age as time scale and left truncation (delayed entry) were used to estimate hazard ratios for Alzheimer’s disease, all dementia, vascular dementia, ischemic cerebrovascular disease, and ischemic heart disease per rs9331896 T allele or as a function of the rs9331896 genotype. Age was accounted for using age as the time scale, which implies that age is automatically adjusted for. For Cox regression models, the proportionality of hazards over time was assessed by plotting -ln(−ln[survival]) versus ln(analysis time). There was no suspicion of non-proportionality. Cox regression models were multifactorially adjusted for age (as time scale), sex, body mass index, hypertension, diabetes mellitus, smoking, alcohol intake, physical inactivity, postmenopausal status and hormonal replacement therapy in women, lipid-lowering therapy, and educational level. Missing values (0.7%) were imputed (continuous covariates) or assigned a dummy value (categorical covariates). However, if only individuals with complete data were included, the results were similar to those reported. The interaction between rs9331896 and rs429358 (defining the ɛ4 allele) relating to risk of Alzheimer’s disease, all dementia, vascular dementia, ischemic cerebrovascular disease, and ischemic heart disease was evaluated by the inclusion of two-factor interaction terms in the Cox regression model, using a likelihood ratio test between models excluding and including the interaction term. To exclude that our results were related to the APOE genotype, adjustment for the APOE genotype as well as a restricted analysis for APOE ε33 individuals only were performed. Odds ratios in meta-analyses were calculated based on beta coefficients from CCHS, CGPS, IGAP, and the CARDIoGRAMplusC4D Consortium using the “metan” command in Stata.

In CCHS and CGPS, the cumulative incidence of Alzheimer’s disease and all dementia increased stepwise from zero to two T alleles of rs9331896 (trend test P = 0.001 and P = 0.009) (Fig. 1), whereas the cumulative incidence of vascular dementia, ischemic cerebrovascular disease, and ischemic heart disease did not differ as a function of an increasing number of rs9331896 T alleles (Additional file 2). In CGPS and CCHS, multifactorially adjusted hazard ratios for Alzheimer’s disease, all dementia, and vascular dementia were 1.18 (95% confidence interval 1.07–1.30), 1.09 (1.02–1.17) and 0.96 (0.80–1.17) per T allele, respectively. For ischemic cerebrovascular disease and ischemic heart disease, the hazard ratios were 1.02 (0.97–1.07) and 0.97 (0.93–1.01) per T allele. No interactions between rs9331896 and rs429358 (defining the ε4 allele) relating to risk of Alzheimer’s disease, all dementia, vascular dementia, ischemic cerebrovascular disease, or ischemic heart disease were observed (P = 0.39 for Alzheimer’s disease, P = 0.21 for all dementia, P = 0.81 for vascular dementia, P = 0.06 for ischemic cerebrovascular disease, and P = 0.71 for ischemic heart disease). Multifactorially adjusted hazard ratios were 1.24 (1.00–1.53) for the rs9331896 TC versus CC and 1.42 (1.15–1.76) for TT versus CC genotypes for Alzheimer’s disease. Corresponding hazard ratios for all dementia were 1.07 (0.93–1.23) for the TC versus CC and 1.18 (1.03–1.36) for TT versus CC genotypes. These associations remained after adjustment for the APOE genotype, as well as in an analysis restricted to ε33 carriers for Alzheimer’s disease. No associations between rs9331896 and vascular dementia or ischemic cerebrovascular disease were observed (Fig. 2). A trend toward reduced risk was observed for ischemic heart disease, most evident when the analysis was restricted to ε33 carriers (P = 0.01). Similar estimates were observed when CGPS and CCHS were analyzed separately (Additional files 3 and 4).

We had 80% statistical power to detect hazard ratios above 1.28 for Alzheimer’s disease, 1.18 for all dementia, 1.49 for vascular dementia, 1.11 for ischemic cerebrovascular disease, and 1.10 for ischemic heart disease for TC heterozygotes. Similar estimates for TT homozygotes were 1.29 for Alzheimer’s disease, 1.19 for all dementia, 1.51 for vascular dementia, 1.12 for ischemic cerebrovascular disease, and 1.11 for ischemic heart disease.

Results for rs9331896 are described above. Multifactorially adjusted hazard ratios for Alzheimer’s disease and all dementia were 2.72 (2.45–3.01) and 2.21 (2.05–2.38) per APOE rs425358 C allele (ɛ4 allele). Frequencies were 15%, 47%, and 37% for the rs9331896 CC, TC, and TT genotypes, respectively. Frequencies were 69%, 28%, and 3% for the APOE TT, CT, and CC genotypes, respectively (Fig. 3).

Meta-analyses on risk of Alzheimer’s disease as a function of risk-increasing T alleles in CLU included four independent studies. The overall fixed- and random-effects odds ratios were 1.16 (1.13–1.18) per risk-increasing allele (T allele) (I ² = 0.0%; P for heterogeneity = 0.89) (Fig. 4). Meta-analyses on risk of ischemic heart disease as a function of risk-increasing T alleles in CLU included three independent studies. The overall fixed- and random-effects odds ratios were 0.99 (95% confidence interval 0.96–1.02) (I ² = 0.0%; P for heterogeneity = 0.77) (Fig. 4).

The rs9331896 variant was associated with slightly higher plasma levels of apolipoprotein B from CC to TC to TT (P = 0.04) (Additional file 5), which however, is no longer significant after a Bonferroni correction for seven parallel tests (required P < 0.05/7 = 0.007). Otherwise no associations were observed between CLU genotypes and lipid, lipoprotein, or apolipoprotein levels.