Association of a wide range of chronic diseases and apolipoprotein E4 genotype with subsequent risk of dementia in community-dwelling adults: A retrospective cohort study.

This analysis was based on data from the UK Biobank, which is a population-based prospective cohort of more than 500,000 participants aged 38–73 years at baseline between 2006 and 2010. 19 These participants attended one of the 22 assessment centers throughout the UK. 19 The details of design and population have been described elsewhere. 19 Briefly, of approximately 9·2 million eligible invited people aged 38–73 years who were registered with the National Health Service, 502,505 individuals were assessed at baselinewith a participation rate of 5·5%. The UK Biobank Study's ethical approval has been granted by the National Information Governance Board for Health and Social Care and the NHS North West Multicenter Research Ethics Committee. All participants provided informed consent through electronic signature at recruitment. Datasets of the main UK Biobank (questionnaires, physical examinations, blood tests), genetics, inpatient hospital, and death register were used in the analysis. The main UK Biobank and inpatient hospital data were used to define individual diseases (predictors) of interest. APOE4 was defined using genetic data. Outcome (dementia) cases were identified using inpatient hospital and death register data. Covariates were assessed using the main UK Biobank data. These datasets were combined using unique patient identifiers. Individuals who had both main UK Biobank and inpatient hospital/death register data, those who were of European ancestry, and those without prevalent dementia or cognitive impairmentat baseline were included in the analysis. Data on admissions and diagnoses were used to identify dementia (all dementia, AD, and vascular dementia [VD]) cases with a primary/secondary diagnosis using the international classification diseases (ICD) coding system (detailed in Table S1). The Hospital Episode Statistics database, the Scottish Morbidity Record, and the Patient Episode Database were used to document inpatient hospital records in England, Scotland and Wales. Additional cases were defined as underlying/contributory causes of death being dementia through linkage to death register data (ICD codes). Individuals with dementia diagnosed <65 years were classified as young-onset dementia, and those diagnosed ≥65 years were classified as late-onset dementia. 18 The earliest recorded date regardless of sources was used as the onset date of dementia. The follow-up period was between March 16, 2006, and January 31, 2021. Person-years for each participant were calculated from the date of baseline assessment to the date of onset dementia, date of death, or the end of follow-up (December 31, 2020 for England and Wales and January 31, 2021 for Scotland), whichever came first. Incident rates were computed as the number of events per 1000 person-years. Self-reported outcomes were assessed using a touchscreen questionnaire. Participants were asked, “Has a doctor ever told that you have a disease (for example, diabetes)”. Participants who reported that they had ever been told by a doctor that they had a disease were classified as having the specific disease. We included 62 major diseases such as cardiovascular disease, cancer, diabetes, dementia, CKD in the analysis (Field code is listed in Table S2). Additional disease cases at baseline were defined using inpatient data (initial diagnosis date before baseline interview date). Inpatient hospital data for the UK Biobank participants were available since 1997. 19 ICD codes for each of the 62 diseases are listed in Table S3. Body mass index (BMI) was computed based on measured weight and height, and obesity was defined as BMI ≥ 30 kg/m 2 . Multimorbidity was defined by two or more out of chronic diseases that were statistically significantly associated with dementia risk. We computed a multimorbidity risk score for dementia based on the estimates for individual diseases that were statistically significantly associated with dementia. The multimorbidity risk score was calculated using the formula: ∑ i 1 β i , where β i is the coefficient (log (hazard ratio [HR]) for incident dementia associated with the i disease. The multimorbidity risk score incorporating age/APOE4 was also computed. We then calculated Cardiovascular Risk Factors, Aging, and Incidence of Dementia (CAIDE) risk score, 20 and Framingham Heart Study (FRS) score 21 for dementia, which were shown to have good prediction performance. 22 The prediction performance of multimorbidity risk score (plus age/APOE4), CAIDE risk score, and FRS risk score was compared. Genotyping was conducted by Affymetrix using a bespoke BiLEVE Axiom array or the UK Biobank Axiom array. 23 All genetic data were quality controlled and imputed by UK Biobank. APOE genotype was directly genotyped based on two SNPs (rs7412 and rs429358). APOE4+ dominant model of E3/E4 and E4/E4 was used to define the presence of APOE4. Given the genetic difference between ethnicities, those of non-European ancestry were excluded from the analysis. Information on age, sex, education, ethnicity, and income was self-reported. Sleep duration was assessed with the survey item “About how many hours sleep do you get in every 24 h?” A short form of the International Physical Activity Questionnaire was used to assess physical activity. A healthy diet score was computed based on seven commonly eaten food groups following recommendations on dietary priorities for cardiometabolic health with a higher score representing a healthier diet. A higher healthy diet score has been associated with decreased dementia risk. 24 Cholesterol, high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C), and triglyceride were measured by direct enzymatic methods (Konelab, Thermo Fisher Scientific, Waltham, Massachusetts). Glycosylated haemoglobin (HbA1c) was measured using high-performance liquid chromatography. Data of baseline characteristics were expressed as frequency (percentage), means ± standard deviations (SDs), or median (interquartile range [IQR]) by multimorbidity/APOE4/age/incident dementia. Cox proportional hazard regression models were used to examine associations of individual diseases of interest with incident dementia and its phenotypes (AD, and VD). The following models were tested: (1) age and gender; (2) model 1 plus education, income, diet score, smoking, alcohol consumption, sleep, physical activity, and BMI (BMI was not included in the analysis for obesity); 3) model 2 plus HDL-C, LDL-C, triglyceride, HbA1c, and blood pressure at baseline. The clustering effect of participants within assessment centers as a random effect was controlled for in all the models. Benjamin-Hochberg's procedure was used to control the false discovery rate at a 5% level for multiple comparisons. 25 The number of diseases was then computed based on those that were statistically significantly associated with incident dementia. The association of individual diseasesand the number of diseases with incident dementia stratified by APOE4 was then tested. Population attributable risk (PAR) of individual diseases, multimorbidity, and APOE4 for dementia was computed based on multivariable-adjusted HR. The receiver operating characteristic curves (ROCs) were plotted and the area under the curve (AUC) was calculated for risk scores of multimorbidity (plus age/APOE4), CAIDE, and FRS. The confounding due to socioeconomic factors for the association between multimorbidity and incident dementia could not be adequately corrected by adjustment. We conducted the analysis for the association between multimorbidity and incident dementia stratified by socioeconomic factors. Similarly, multimorbidity and dementia were both highly related to age, and the adjustment for age might not be able to address the confounding. We conducted the analysis for the association between multimorbidity and incident dementia stratified by age. As the prodromal period of dementia can last many years, dementia developed in the first several years of follow-up might have occurred before the diagnosis of chronic conditions but not be diagnosed. We conducted a sensitivity analysis to examine the association between major diseases and incident dementia by excluding those who developed dementia in the first five years of follow-up. The diagnosis of dementia in 2020 might be impacted by COVID-19. We conducted a sensitivity analysis to test whether the associations between multimorbidity/APOE4 and incident dementia might be impacted by COVID-19 (set the end of follow-up on December 31, 2019). The percentage of participants with missing values in physical activity, income, HbA1c, and blood pressure was 19·5%, 14·5%, 6·7%, and 6·0%, respectively. A much lower missing rate was seen for other covariates. Missing values for categorical variables were assigned as a new single category. Multiple imputations for missing data in continuous variables were conducted for baseline covariates, and we included all covariates in the imputation models to create 10 imputed datasets. Data analyses were conducted using SAS 9·4 for Windows (SAS Institute Inc.) and all P values were two-sided with statistical significance set at <0·05. The funder of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report. All authors had full access to all the data in the study and had final responsibility for the decision to submit for publication.

Baseline data were collected among 502,505 participants. After excluding individuals who were not linked to inpatient hospital/death register data ( n  = 29), those who were non-European ancestry ( n  = 30,378), those with prevalent dementia ( n  = 345) or cognitive impairment ( n  = 233), or with incident dementia that developed in the first year of follow-up ( n  = 36), 471,485 adults (54·5% females) aged 38–73 years (mean ± SD: 56·8 ± 8·0) were included in the final analysis (Fig. S1). Participants with dementia occurred in the first year of follow-up were excluded from the analysis because those participants were more likely to have preexisting dementia at baseline which may confound the association between major diseases and incident dementia. Individuals with multimorbidity were more likely to be older, male, lowly educated, and to have low income and unhealthier lifestyle habits compared to those with zero disease. Multimorbidity was associated with higher BMI, HbA1c, and blood pressure ( Table 1 ). Compared with individuals who did not develop dementia during follow-up, those who developed dementia were more likely to be older, males, APOE4 carriers, and lowly educated and to have higher HbA1c and systolic blood pressure. Adults aged 65 years or over were more likely to be males and lowly educated, and to have higher systolic blood pressure compared with those aged young than 65 years (Table S4). APOE4 carriers were more likely to have higher total cholesterol, LDL-C, and triglycerides compared with non-APOE4-carriers (Table S5). Table 1 Baseline characteristics of participants. Table 1 Number of major diseases at baseline 0 1 2 3 4 5 6 Age (years) 53·60 ± 7·97 55·53 ± 8·00 57·24 ± 7·81 58·79 ± 7·43 59·87 ± 7·06 60·69 ± 6·68 61·19 ± 6·41 Age (years), median (IQR) 53·0 (47·0–60·0) 56·0 (49·0–62·0) 59·0 (51·0–64·0) 60·0 (54·0–65·0) 61·0 (56·0–65·0) 62·0 (57·0–66·0) 63·0 (58·0–66·0) Gender  Females 59,754 (63·2) 75,982 (56·8) 54,493 (53·1) 31,800 (49·4) 17,404 (47·3) 9135 (45·6) 8238 (42·3)  Males 34,797 (36·8) 57,691 (43·2) 48,131 (46·9) 32,527 (50·6) 19,365 (52·7) 10,909 (54·4) 11,259 (57·7) APOE4  No 69,533 (73·5) 98,809 (73·9) 75,964 (74·0) 47,751 (74·2) 27,181 (73·9) 14,895 (74·3) 14,478 (74·3)  Yes 22,343 (23·6) 31,290 (23·4) 24,051 (23·4) 15,006 (23·3) 8637 (23·5) 4585 (22·9) 4442 (22·8)  Missing 2675 (2·8) 3574 (2·7) 2609 (2·5) 1570 (2·4) 951 (2·6) 564 (2·8) 577 (3·0) Education  0–5 years 9177 (9·7) 17,347 (13·0) 17,399 (17·0) 13,766 (21·4) 9495 (25·8) 6276 (31·3) 7341 (37·7)  6–12 years 47,044 (49·8) 66,585 (49·8) 51,431 (50·1) 32,070 (49·9) 17,999 (49·0) 9389 (46·8) 8749 (44·9)  ≥13 years 36,734 (38·9) 47,605 (35·6) 32,165 (31·3) 17,463 (27·1) 8658 (23·5) 4024 (20·1) 3049 (15·6)  Missing 1596 (1·7) 2136 (1·6) 1629 (1·6) 1028 (1·6) 617 (1·7) 355 (1·8) 358 (1·8) Household income (pounds) <18,000 11,034 (11·7) 19,523 (14·6) 19,155 (18·7) 14,998 (23·3) 10,375 (28·2) 6665 (33·3) 8068 (41·4)  18,000–30,999 18,093 (19·1) 28,222 (21·1) 23,390 (22·8) 15,122 (23·5) 8722 (23·7) 4794 (23·9) 4200 (21·5)  31,000–51,999 23,821 (25·2) 32,859 (24·6) 23,558 (23·0) 13,409 (20·8) 6720 (18·3) 3159 (15·8) 2298 (11·8)  52,000–100,000 22,642 (23·9) 27,473 (20·6) 17,298 (16·9) 8745 (13·6) 4016 (10·9) 1596 (8·0) 1027 (5·3)  >100,000 6913 (7·3) 7464 (5·6) 4297 (4·2) 1988 (3·1) 817 (2·2) 308 (1·5) 161 (0·8)  Unknown 2567 (2·7) 4401 (3·3) 4054 (4·0) 2979 (4·6) 1922 (5·2) 1195 (6·0) 1422 (7·3)  Not answered 9481 (10·0) 13,731 (10·3) 10,872 (10·6) 7086 (11·0) 4197 (11·4) 2327 (11·6) 2321 (11·9) Diet score, median (IQR) 4·0 (3·0–5·0) 4·0 (3·0–5·0) 4·0 (3·0–5·0) 4·0 (3·0–5·0) 4·0 (3·0–5·0) 4·0 (3·0–5·0) 4·0 (3·0–5·0) Physical activity (MET-minutes/week), median (IQR) 2505·0 (1152·0–3207·5) 2478·0 (1078·0–3093·0) 2462·0 (1011·0–2994·0) 2445·0 (942·0–2862·0) 2479·5 (904·0–2730·0) 2373·0 (813·0–2651·9) 2160·0 (634·0–2651·9) Smoking  Never 58,108 (61·5) 76,811 (57·5) 54,511 (53·1) 31,802 (49·4) 16,771 (45·6) 8433 (42·1) 7045 (36·1)  Former 27,056 (28·6) 42,993 (32·2) 37,126 (36·2) 25,324 (39·4) 15,771 (42·9) 9170 (45·7) 9570 (49·1)  Current 9167 (9·7) 13,494 (10·1) 10,613 (10·3) 6950 (10·8) 4042 (11·0) 2330 (11·6) 2733 (14·0)  Missing 220 (0·2) 375 (0·3) 374 (0·4) 251 (0·4) 185 (0·5) 111 (0·6) 149 (0·8) Sleep duration (hours/day)  <7 18,662 (19·7) 29,634 (22·2) 25,266 (24·6) 16,951 (26·4) 10,236 (27·8) 5805 (29·0) 6227 (31·9)  7–9 74,929 (79·2) 102,181 (76·4) 75,318 (73·4) 45,469 (70·7) 25,062 (68·2) 13,136 (65·5) 11,595 (59·5)  >9 695 (0·7) 1375 (1·0) 1531 (1·5) 1491 (2·3) 1138 (3·1) 846 (4·2) 1322 (6·8)  Missing 265 (0·3) 483 (0·4) 509 (0·5) 416 (0·6) 333 (0·9) 257 (1·3) 353 (1·8) Alcohol consumption  Never 2482 (2·6) 3682 (2·8) 3147 (3·1) 2348 (3·7) 1464 (4·0) 982 (4·9) 1181 (6·1)  Previous 1837 (1·9) 3309 (2·5) 3327 (3·2) 2658 (4·1) 2011 (5·5) 1370 (6·8) 1964 (10·1)  Current 90,189 (95·4) 126,600 (94·7) 96,069 (93·6) 59,256 (92·1) 33,231 (90·4) 17,657 (88·1) 16,300 (83·6)  Missing 43 (0·0) 82 (0·1) 81 (0·1) 65 (0·1) 63 (0·2) 35 (0·2) 52 (0·3) BMI (kg/m 2 ) 24·85 ± 2·70 26·23 ± 3·88 27·75 ± 4·58 29·02 ± 4·94 30·15 ± 5·19 30·92 ± 5·27 31·99 ± 5·59 HbA1c (mmol/mol) 34·39 ± 3·99 35·01 ± 4·75 35·78 ± 5·44 36·87 ± 6·61 38·28 ± 8·26 39·66 ± 9·44 41·93 ± 11·32 Cholesterol (mmol/L) 5·83 ± 1·02 5·86 ± 1·04 5·80 ± 1·08 5·63 ± 1·14 5·41 ± 1·18 5·22 ± 1·19 4·97 ± 1·18 HDL-C (mmol/L) 1·54 ± 0·36 1·49 ± 0·35 1·44 ± 0·35 1·40 ± 0·34 1·35 ± 0·33 1·32 ± 0·33 1·28 ± 0·33 LDL-C (mmol/L) 3·63 ± 0·79 3·67 ± 0·80 3·64 ± 0·82 3·53 ± 0·87 3·36 ± 0·89 3·22 ± 0·89 3·03 ± 0·87 Triglycerides (mmol/L) 1·51 ± 0·84 1·65 ± 0·93 1·80 ± 0·99 1·90 ± 1·04 1·99 ± 1·08 2·06 ± 1·12 2·15 ± 1·18 Diastolic blood pressure (mmHg) 79·97 ± 9·32 81·51 ± 9·71 83·11 ± 9·88 83·86 ± 9·85 83·71 ± 9·78 83·09 ± 9·91 81·94 ± 10·18 Systolic blood pressure (mmHg) 133·08 ± 17·20 136·09 ± 17·86 139·34 ± 18·17 141·33 ± 18·12 141·95 ± 17·79 141·75 ± 17·78 140·55 ± 17·96 APOE4, apolipoprotein E4; BMI, body mass index; HbA1c, glycatedhaemoglobin; HDL-C, high-density lipoprotein cholesterol; IQR, interquartile range; LDL-C, low-density lipoprotein cholesterol. *APOE4+ dominant model of E3/E4 and E4/E4 was used to define the presence of APOE4. †Diet score was computed based on seven commonly eaten food groups following recommendations on dietary priorities for cardiometabolic health with higher score representing healthier diet. Baseline characteristics of participants. APOE4, apolipoprotein E4; BMI, body mass index; HbA1c, glycatedhaemoglobin; HDL-C, high-density lipoprotein cholesterol; IQR, interquartile range; LDL-C, low-density lipoprotein cholesterol. *APOE4+ dominant model of E3/E4 and E4/E4 was used to define the presence of APOE4. †Diet score was computed based on seven commonly eaten food groups following recommendations on dietary priorities for cardiometabolic health with higher score representing healthier diet. Over 5624,248 person-years of follow-up (median [interquartile range] length of follow-up: 11·9 [11·2–12·6] years), 6189 cases of incident all-cause dementia, 2597 incident AD cases, and 1381 incident VDcases were documented. Among 377,214 participants who were aged younger than 65 years at baseline, 503 developed young-onset dementia. Whilst 5686 individuals developed late-onset dementia among those who were 65 years or older at the end of follow-up. Of 63 diseases of interest, 36 were associated with an increased risk of all-cause dementia after controlling the false discovery rate. The HR (95% CI) ranged from 1·12 (1·06–1·19) for obesity to 14·12 (12·33–16·18) for Parkinson's disease. In addition to Parkinson's disease, there were 9 diseases including depression (2·12 (1·94–2·31)), anxiety (2·06 (1·78–2·39)), diabetes (2·12 (1·97–2·30)), stroke (2·39 (2·14–2·66)), alcohol problems (2·71 (2·18–3·37)), psychoactive substance abuse (3·28 (2·44–4·42)), epilepsy (2·89 (2·47–3·39)), multiple sclerosis (2·20 (1·61–3·02)), and schizophrenia (3·77 (3·05–4·66)) with HR over 2. Associations for 33 diseases remained statistically significant after adjustment for age, gender, education, income, diet score, smoking, alcohol consumption, sleep, physical activity, BMI, HDL-C, LDL-C, triglyceride, HbA1c, and blood pressure at baseline. The association between stomach cancer and incident dementia was attenuated to be non-statistically significant after adjustment for covariates. The HR (95% CI) for all-cause dementia associated with prostate cancer in men was reversed from 1·54 (1·22–1·94) to 0·75 (0·59–0·94) after adjustment for age. Notably, the association of hypertension (HR (95% CI): 1·29 (1·22–1·36)) and high cholesterol (1·30 (1·22–1·38)) with dementia remained statistically significant after even adjustment for blood pressure, HDL-C, LDL-C, and triglyceride at baseline ( Table 2 ). Table 2 Risk for all-cause dementia associated with a wide range of diseases. Table 2 Participants Events Person-years Incidence rate * HR (95% CI), HR (95% CI), HR (95% CI), Model 1 † Model 2 Model 3 Hearing impairment ǂ  No 270,224 2756 3,232,267 0·85 Reference Reference Reference  Yes 201,261 3433 2,388,864 1·44 1·28 (1·22–1·35) 1·25 (1·19–1·32) 1·25 (1·19–1·32) Obesity ǂ  No 357,994 4534 4,267,577 1·06 Reference Reference Reference  Yes 113,491 1655 1,353,554 1·22 1·12 (1·06–1·19) 1·12 (1·03–1·23) 1·11 (1·01–1·21) Hypertension ǂ  No 350,404 3485 4,180,853 0·83 Reference Reference Reference  Yes 121,081 2704 1,440,278 1·88 1·38 (1·31–1·45) 1·33 (1·26–1·40) 1·29 (1·22–1·36) High cholesterol ǂ  No 410,568 4532 4,901,284 0·92 Reference Reference Reference  Yes 60,917 1657 719,847 2·30 1·43 (1·35–1·52) 1·36 (1·28–1·44) 1·30 (1·22–1·38) Coronary heart disease ǂ  No 447,000 5204 5,330,836 0·98 Reference Reference Reference  Yes 24,485 985 290,295 3·39 1·84 (1·72–1·97) 1·64 (1·53–1·76) 1·55 (1·44–1·67) Atrial Fibrillation ǂ  No 464,197 5916 5,534,962 1·07 Reference Reference Reference  Yes 7288 273 86,169 3·17 1·55 (1·38–1·76) 1·49 (1·32–1·68) 1·45 (1·28–1·64)  Heart failure ǂ  No 469,301 6098 5,595,321 1·09 Reference Reference Reference  Yes 2184 91 25,810 3·53 1·85 (1·50–2·28) 1·57 (1·27–1·93) 1·45 (1·18–1·79) Stroke ǂ  No 464,276 5841 5,535,558 1·06 Reference Reference Reference  Yes 7209 348 85,573 4·07 2·39 (2·14–2·66) 2·05 (1·84–2·29) 1·92 (1·72–2·15) Peripheral vascular disease  No 468,564 6097 5,586,101 1·09 Reference Reference Reference  Yes 2921 92 35,030 2·63 1·54 (1·25–1·89) 1·29 (1·05–1·59) 1·18 (0·96–1·45) Other cardiac problem ǂ  No 465,758 6013 5,552,476 1·08 Reference Reference Reference  Yes 5727 176 68,654 2·56 1·64 (1·41–1·90) 1·46 (1·26–1·70) 1·39 (1·19–1·62) Diabetes ǂ  No 452,227 5448 5,392,832 1·01 Reference Reference Reference  Yes 19,258 741 228,298 3·25 2·12 (1·97–2·30) 1·95 (1·80–2·12) 1·72 (1·56–1·89) COPD ǂ  No 468,160 6054 5,581,889 1·08 Reference Reference Reference  Yes 3325 135 39,242 3·44 1·87 (1·58–2·22) 1·48 (1·24–1·75) 1·44 (1·21–1·71) Asthma ǂ  No 415,916 5358 4,959,066 1·08 Reference Reference Reference  Yes 55,569 831 662,064 1·26 1·25 (1·16–1·35) 1·20 (1·12–1·30) 1·20 (1·11–1·29) Bronchiectasis  No 470,096 6155 5,604,625 1·10 Reference Reference Reference  Yes 1389 34 16,505 2·06 1·24 (0·88–1·73) 1·20 (0·86–1·68) 1·17 (0·84–1·64) Depression ǂ  No 443,780 5610 5,291,854 1·06 Reference Reference Reference  Yes 27,705 579 329,276 1·76 2·12 (1·94–2·31) 1·88 (1·73–2·05) 1·90 (1·74–2·07) Anxiety ǂ  No 463,534 6005 5,526,650 1·09 Reference Reference Reference  Yes 7951 184 94,481 1·95 2·06 (1·78–2·39) 1·82 (1·57–2·11) 1·84 (1·59–2·13) Schizophrenia ǂ  No 469,245 6102 5,594,612 1·09 Reference Reference Reference  Yes 2240 87 26,519 3·28 3·77 (3·05–4·66) 3·09 (2·50–3·82) 3·13 (2·53–3·87) Parkinson's disease ǂ  No 470,626 5972 5,611,768 1·06 Reference Reference Reference  Yes 859 217 9363 23·18 14·12 (12·33–16·18) 13·49 (11·77–15·46) 13·40 (11·68–15·37) Multiple Sclerosis ǂ  No 469,653 6150 5,599,274 1·10 Reference Reference Reference  Yes 1832 39 21,856 1·78 2·20 (1·61–3·02) 2·02 (1·47–2·77) 2·05 (1·50–2·82) Alcohol problems ǂ  No 468,855 6107 5,589,767 1·09 Reference Reference Reference  Yes 2630 82 31,364 2·61 2·71 (2·18–3·37) 1·82 (1·46–2·27) 1·82 (1·46–2·28) Psychoactive substance abuse ǂ  No 470,250 6145 5,606,554 1·10 Reference Reference Reference  Yes 1235 44 14,577 3·02 3·28 (2·44–4·42) 2·33 (1·73–3·15) 2·34 (1·74–3·16) Migraine  No 456,894 6042 5,446,086 1·11 Reference Reference Reference  Yes 14,591 147 175,045 0·84 0·98 (0·84–1·16) 0·96 (0·82–1·13) 0·99 (0·84–1·16) Epilepsy ǂ  No 466,909 6031 5,566,886 1·08 Reference Reference Reference  Yes 4576 158 54,245 2·91 2·89 (2·47–3·39) 2·45 (2·09–2·87) 2·47 (2·11–2·90) Meniere disease  No 469,814 6156 5,600,936 1·10 Reference Reference Reference  Yes 1671 33 20,195 1·63 1·17 (0·83–1·65) 1·11 (0·78–1·56) 1·12 (0·80–1·58) Chronic sinusitis  No 467,067 6128 5,568,195 1·10 Reference Reference Reference  Yes 4418 61 52,935 1·15 1·03 (0·80–1·32) 1·03 (0·80–1·32) 1·04 (0·81–1·34) Traumatic brain injury ǂ  No 466,073 6084 5,556,434 1·09 Reference Reference Reference  Yes 5412 105 64,697 1·62 1·62 (1·34–1·96) 1·44 (1·18–1·74) 1·44 (1·18–1·74) Other brain problem  No 470,435 6169 5,608,617 1·10 Reference Reference Reference  Yes 1050 20 12,514 1·60 1·48 (0·96–2·30) 1·31 (0·84–2·03) 1·32 (0·85–2·05) Painful conditions ǂ  No 409,559 5074 4,889,420 1·04 Reference Reference Reference  Yes 61,926 1115 731,710 1·52 1·33 (1·25–1·42) 1·25 (1·17–1·33) 1·25 (1·17–1·33) Chronic fatigue syndrome ǂ  No 469,332 6157 5,595,323 1·10 Reference Reference Reference  Yes 2153 32 25,808 1·24 1·53 (1·08–2·16) 1·43 (1·01–2·02) 1·45 (1·02–2·05) Connective tissue disorders ǂ  No 459,751 5944 5,481,184 1·08 Reference Reference Reference  Yes 11,734 245 139,946 1·75 1·30 (1·14–1·47) 1·19 (1·05–1·36) 1·19 (1·04–1·35) Osteoporosis ǂ  No 426,388 5106 5,083,699 1·00 Reference Reference Reference  Yes 45,097 1083 537,431 2·02 1·29 (1·21–1·38) 1·20 (1·13–1·29) 1·21 (1·13–1·29) Fracture ǂ  No 469,381 6135 5,596,575 1·10 Reference Reference Reference  Yes 2104 54 24,555 2·20 1·63 (1·25–2·13) 1·51 (1·16–1·98) 1·50 (1·15–1·96) Anorexia  No 471,100 6187 5,616,511 1·10 Reference Reference Reference  Yes 385 2 4619 0·43 1·18 (0·30–4·73) 1·01 (0·25–4·04) 1·01 (0·25–4·05) Dyspepsia ǂ  No 417,998 5103 4,985,899 1·02 Reference Reference Reference  Yes 53,487 1086 635,232 1·71 1·25 (1·17–1·34) 1·15 (1·08–1·23) 1·15 (1·08–1·23) Treated constipation ǂ  No 466,486 6049 5,562,139 1·09 Reference Reference Reference  Yes 4999 140 58,992 2·37 1·91 (1·61–2·26) 1·66 (1·40–1·96) 1·63 (1·38–1·93) Pernicious anemia  No 469,972 6153 5,603,054 1·10 Reference Reference Reference  Yes 1513 36 18,077 1·99 1·44 (1·04–2·00) 1·29 (0·93–1·79) 1·25 (0·90–1·73) Diverticular disease ǂ  No 458,797 5851 5,471,108 1·07 Reference Reference Reference  Yes 12,688 338 150,023 2·25 1·28 (1·14–1·42) 1·21 (1·09–1·36) 1·22 (1·09–1·36) Chronic kidney disease ǂ  No 468,838 6109 5,589,325 1·09 Reference Reference Reference  Yes 2647 80 31,806 2·52 1·69 (1·35–2·10) 1·48 (1·19–1·85) 1·42 (1·14–1·77) Inflammatory bowel disease  No 466,585 6103 5,562,747 1·10 Reference Reference Reference  Yes 4900 86 58,384 1·47 1·11 (0·95–1·29) 1·06 (0·91–1·23) 1·07 (0·92–1·25) Irritable bowel syndrome  No 458,672 6018 5,467,588 1·10 Reference Reference Reference  Yes 12,813 171 153,543 1·11 1·26 (1·02–1·56) 1·19 (0·96–1·47) 1·18 (0·96–1·46) Viral hepatitis  No 470,078 6166 5,604,122 1·10 Reference Reference Reference  Yes 1407 23 17,009 1·35 1·33 (0·88–2·00) 1·32 (0·88–1·99) 1·31 (0·87–1·98) Chronic liver disease  No 467,881 6118 5,575,708 1·10 Reference Reference Reference  Yes 3604 71 45,423 1·56 1·24 (0·98–1·57) 1·12 (0·88–1·41) 1·11 (0·88–1·40) Prostate disorders ǂ , §  No 203,626 2920 2,427,426 1·20 Reference Reference Reference  Yes 11,053 348 130,782 2·66 1·16 (1·04–1·30) 1·14 (1·02–1·27) 1·14 (1·01–1·27) Endometriosis ¶  No 250,301 2876 2,985,552 0·96 Reference Reference Reference  Yes 6505 45 77,371 0·58 0·96 (0·72–1·30) 0·98 (0·73–1·32) 0·99 (0·74–1·33) Polycystic ovary ¶  No 256,123 2916 3,054,857 0·95 Reference Reference Reference  Yes 683 5 8066 0·62 0·66 (0·28–1·59) 2·56 (1·07–6·17) 2·54 (1·06–6·12) Thyroid disorders  No 442,824 5706 5,280,710 1·08 Reference Reference Reference  Yes 28,661 483 340,421 1·42 1·15 (1·05–1·27) 1·10 (1·00–1·21) 1·09 (0·99–1·19) Eczema ǂ  No 453,444 5930 5,404,867 1·10 Reference Reference Reference  Yes 18,041 259 216,264 1·20 1·18 (1·04–1·33) 1·16 (1·02–1·31) 1·16 (1·02–1·31) HIV  No 471,128 6187 5,616,939 1·10 Reference Reference Reference  Yes 357 2 4192 0·48 0·80 (0·20–3·22) 0·69 (0·17–2·78) 0·71 (0·18–2·85) Glaucoma  No 466,299 6070 5,560,029 1·09 Reference Reference Reference  Yes 5186 119 61,102 1·95 1·05 (0·88–1·26) 1·04 (0·87–1·25) 1·03 (0·86–1·24) Cataract ǂ  No 460,633 5873 5,494,928 1·07 Reference Reference Reference  Yes 10,852 316 126,202 2·50 1·26 (1·13–1·42) 1·22 (1·09–1·37) 1·18 (1·06–1·33) AMD  No 470,564 6162 5,610,361 1·10 Reference Reference Reference  Yes 921 27 10,769 2·51 1·30 (0·89–1·90) 1·28 (0·88–1·87) 1·24 (0·85–1·81) Lung cancer  No 471,062 6176 5,616,120 1·10 Reference Reference Reference  Yes 423 13 5010 2·59 1·36 (0·79–2·35) 1·10 (0·64–1·89) 1·09 (0·63–1·88) Skin cancer  No 466,470 6079 5,561,744 1·09 Reference Reference Reference  Yes 5015 110 59,387 1·85 1·05 (0·87–1·27) 1·04 (0·86–1·25) 1·04 (0·86–1·25) Melanoma  No 467,477 6113 5,573,346 1·10 Reference Reference Reference  Yes 4008 76 47,785 1·59 1·10 (0·88–1·38) 1·13 (0·90–1·41) 1·14 (0·91–1·42) Stomach cancer  No 471,165 6176 5,617,252 1·10 Reference Reference Reference  Yes 320 13 3878 3·35 1·98 (1·15–3·41) 1·69 (0·98–2·91) 1·63 (0·95–2·82) Oesophageal cancer  No 469,800 6146 5,600,911 1·10 Reference Reference Reference  Yes 1685 43 20,219 2·13 1·12 (0·83–1·52) 1·00 (0·74–1·35) 1·00 (0·74–1·35) Colon cancer  No 469,422 6140 5,596,470 1·10 Reference Reference Reference  Yes 2063 49 24,661 1·99 0·81 (0·57–1·15) 0·80 (0·56–1·14) 0·80 (0·56–1·14) Rectal cancer  No 470,490 6169 5,609,133 1·10 Reference Reference Reference  Yes 995 20 11,997 1·67 0·86 (0·55–1·33) 0·79 (0·51–1·23) 0·79 (0·51–1·22) Prostate cancer §  No 211,378 3194 2,519,178 1·27 Reference Reference Reference  Yes 3301 74 39,030 1·90 1·54 (1·22–1·94) 0·73 (0·58–0·92) 0·75 (0·59–0·94) Ovarian cancer ¶  No 255,894 2908 3,052,049 0·95 Reference Reference Reference  Yes 912 13 10,874 1·20 0·91 (0·53–1·58) 0·88 (0·51–1·52) 0·88 (0·51–1·53) Breast cancer ¶  No 245,944 2747 2,933,849 0·94 Reference Reference Reference  Yes 10,862 174 129,074 1·35 0·99 (0·85–1·15) 1·00 (0·86–1·17) 1·00 (0·86–1·16) Uterine cancer  No 470,098 6156 5,604,394 1·10 Reference Reference Reference  Yes 1387 33 16,737 1·97 1·26 (0·89–1·78) 1·22 (0·87–1·72) 1·22 (0·86–1·71) Other cancers  No 447,035 5777 5,327,435 1·08 Reference Reference Reference  Yes 24,450 412 293,695 1·40 1·03 (0·93–1·14) 1·01 (0·91–1·12) 1·01 (0·92–1·12) APOE4, apolipoprotein E4; CI, confidence interval; COPD, chronic obstructive pulmonary disease; HIV, human immunodeficiency virus; HR, hazard ratio. ⁎ Refers to number of cases per 1000 person-years. † Cox proportional hazard regression models were used to examine the association between each of the 61 major diseases at baseline and incident dementia. Model 1 was adjusted for age and gender (Model 1 was unadjusted model for prostate cancer); Model 2 was adjusted for Model 1 plus education, income, BMI, smoking, physical activity, alcohol consumption, sleep duration, and diet; Model 3 was adjusted for blood pressure, HDL-C, LDL-C, triglycerides, and HbA1c. ǂ Refers to statistically significant association in the multivariable-adjusted model after controlling false discovery rate.There are four steps to run the Benjamini-Hochberg procedure: (1) Rank p- values from the smallest to the largest; (2) Assign rank to the p- values; (3) Compute Benjamini-Hochberg critical value for each individual p- value: critical value =  i m × Q , where i refers to the rank of the p- value, m refers to the total number of tests, and Q refers to the false discovery rate (0.05 in the present study); 4) Compare the original p- values to corresponding Benjamini-Hochberg critical values and find the largest p- value that is smaller than the critical value. § Analysis was conducted among males only. ¶ Analysis was conducted among females only. Risk for all-cause dementia associated with a wide range of diseases. APOE4, apolipoprotein E4; CI, confidence interval; COPD, chronic obstructive pulmonary disease; HIV, human immunodeficiency virus; HR, hazard ratio. Refers to number of cases per 1000 person-years. Cox proportional hazard regression models were used to examine the association between each of the 61 major diseases at baseline and incident dementia. Model 1 was adjusted for age and gender (Model 1 was unadjusted model for prostate cancer); Model 2 was adjusted for Model 1 plus education, income, BMI, smoking, physical activity, alcohol consumption, sleep duration, and diet; Model 3 was adjusted for blood pressure, HDL-C, LDL-C, triglycerides, and HbA1c. Refers to statistically significant association in the multivariable-adjusted model after controlling false discovery rate.There are four steps to run the Benjamini-Hochberg procedure: (1) Rank p- values from the smallest to the largest; (2) Assign rank to the p- values; (3) Compute Benjamini-Hochberg critical value for each individual p- value: critical value =  i m × Q , where i refers to the rank of the p- value, m refers to the total number of tests, and Q refers to the false discovery rate (0.05 in the present study); 4) Compare the original p- values to corresponding Benjamini-Hochberg critical values and find the largest p- value that is smaller than the critical value. Analysis was conducted among males only. Analysis was conducted among females only. Well-known risk factors including hearing impairment (PAR(95% CI): 9·6% (7·5–12·0%)), hypertension (6·9% (5·4–8·5%)), depression (5·0% (4·2–5·9%)), high cholesterol (3·7% (2·8–4·7%)), diabetes (2·9% (2·2–3·5%)), and obesity (2·6% (0·2–4·8%)) were among leading contributors to dementia. Painful conditions (3·2% (2·2–4·2%)) and CHD (2·8% (2·2–3·4%)) were also among leading contributors to dementia (Fig. S2). Similar results were seen for AD (Table S6) and VD (Table S7). HRs associated with 29 individual diseases were larger for young-onset dementia than for late-onset dementia. For example, the HR (95% CI) associated with hearing impairment was 1·64 (1·38–1·97) for young-onset dementia and 1·22 (1·16–1·28) for late-onset dementia. Obesity was predictive of young-onset dementia (1·43 (1·05–1·95)) but not late-onset dementia (1·09 (0·99–1·19)). Similarly, eczema was predictive of young-onset dementia (1·58 (1·11–2·26)) but not late-onset dementia (1·11 (0·97–1·27)). In contrast, atrial fibrillation was a risk factor for late-onset dementia (1·45 (1·28–1·65)) but not young-onset dementia (0·91 (0·41–2·05)). Diverticular disease was associated with an increased risk of late-onset dementia (1·21 (1·08–1·35)), but not young-onset dementia (1·09 (0·61–1·94), Figure 1 ). These diseases were greater contributors to young-onset dementia than to late-onset dementia (Fig. S1). Figure 1 Risk for young-onset and late-onset dementia associated with individual major diseases CI, confidence interval; CKD, chronic kidney disease; COPD, Chronic obstructive pulmonary disease. Young-onset dementia was defined as dementia diagnosed <65 years, and the analysis was conducted among individuals aged younger than 65 years at baseline. Late-onset dementia was defined as dementia diagnosed ≥65 years, and the analysis was conducted among individuals aged older than 65 years at the end of follow-up. Cox proportional hazard regression models were used to examine the association between individual diseases and the incidence of young-onset and late-onset dementia. Analysis was adjusted for age, gender, education, income, BMI, smoking, physical activity, alcohol consumption, sleep duration, diet, blood pressure, HDL-C, LDL-C, triglycerides, and HbA1c. Horizontal lines indicate the ranges of the 95% CIs and the vertical dash lines indicate the hazard ratio of 1·0. Fig 1 Risk for young-onset and late-onset dementia associated with individual major diseases CI, confidence interval; CKD, chronic kidney disease; COPD, Chronic obstructive pulmonary disease. Young-onset dementia was defined as dementia diagnosed <65 years, and the analysis was conducted among individuals aged younger than 65 years at baseline. Late-onset dementia was defined as dementia diagnosed ≥65 years, and the analysis was conducted among individuals aged older than 65 years at the end of follow-up. Cox proportional hazard regression models were used to examine the association between individual diseases and the incidence of young-onset and late-onset dementia. Analysis was adjusted for age, gender, education, income, BMI, smoking, physical activity, alcohol consumption, sleep duration, diet, blood pressure, HDL-C, LDL-C, triglycerides, and HbA1c. Horizontal lines indicate the ranges of the 95% CIs and the vertical dash lines indicate the hazard ratio of 1·0. Thirty-three diseases with statistically significant association with dementia (listed in Table 2 ) were used to identify multimorbidity. Individuals with zero disease had a higher survival probability from dementia during follow-up compared to those with one or more diseases (Fig. S3). In the multivariable analysis, a larger number of diseases was associated with larger HR for incident dementia, AD, and VD. The HR associated with multimorbidity was larger for young-onset dementia than for late-onset dementia ( Figure 2 ). PAR (95% CI) of multimorbidity was 51·2% (45·5–56·3%) for all dementia, 64·7% (49·9–75·2%) for young-onset dementia, and 47·4% (40·8–53·2%) for late-onset dementia (data not shown). Figure 2 Risk for dementia associated with the number of diseases at baseline. CI, confidence interval. Young-onset dementia was defined as dementia diagnosed <65 years, and the analysis was conducted among individuals aged younger than 65 years at baseline. Late-onset dementia was defined as dementia diagnosed ≥65 years, and the analysis was conducted among individuals aged older than 65 years at the end of follow-up. Cox proportional hazard regression models were used to examine the association between the number of diseases at baseline (statistically significantly associated with dementia in multivariable-adjusted analysis in Table 2 ) and incident dementia. The analysis was adjusted for age, gender, education, income, BMI, smoking, physical activity, alcohol consumption, sleep duration, diet, blood pressure, HDL-C, LDL-C, triglycerides, and HbA1c. Reference was the group with no disease of interest at baseline. Horizontal lines indicate the ranges of 95% CIs. Fig 2 Risk for dementia associated with the number of diseases at baseline. CI, confidence interval. Young-onset dementia was defined as dementia diagnosed <65 years, and the analysis was conducted among individuals aged younger than 65 years at baseline. Late-onset dementia was defined as dementia diagnosed ≥65 years, and the analysis was conducted among individuals aged older than 65 years at the end of follow-up. Cox proportional hazard regression models were used to examine the association between the number of diseases at baseline (statistically significantly associated with dementia in multivariable-adjusted analysis in Table 2 ) and incident dementia. The analysis was adjusted for age, gender, education, income, BMI, smoking, physical activity, alcohol consumption, sleep duration, diet, blood pressure, HDL-C, LDL-C, triglycerides, and HbA1c. Reference was the group with no disease of interest at baseline. Horizontal lines indicate the ranges of 95% CIs. Non-APOE4 carriers had a higher survival probability from dementia during follow-up compared with APOE4 carriers (Fig. S4). As shown in Table 3 , HRs (95% CIs) for all-cause dementiaassociated with APOE4 was 2·73 (2·59–2·87). The association was stronger for AD (HR (95% CI): 3·69 (3·41–4·00)) than for VD (2·52 (2·26–2·81)). APOE4 was a stronger predictor for late-onset dementia (HR (95% CI): 2·90 (2·75–3·06)) but not for young-onset dementia (1·26 (1·03–1·54)). Table 3 Risk for dementia associated with APOE4. Table 3 Events/person-years Incidence rate * HR (95% CI), HR (95% CI), HR (95% CI), Non-APOE4 Carriers APOE4 carriers Non-APOE4 carriers APOE4 carriers Model 1 † Model 2 Model 3 All-cause dementia  All 3281/4,155,062 2667/1,311,270 0·79 2·03 2·65 (2·52–2·79) 2·67 (2·54–2·81) 2·73 (2·59–2·87)  Young-onset ǂ 349/3,335,485 131/1,053,453 0·10 0·12 1·19 (0·97–1·45) 1·23 (1·00–1·50) 1·26 (1·03–1·54)  Late-onset § 2932/2,798,517 2536/874,175 1·05 2·90 2·83 (2·69–2·99) 2·85 (2·70–3·01) 2·90 (2·75–3·06) AD  All 1172/4,161,795 1322/1,315,407 0·28 1·01 3·64 (3·37–3·94) 3·64 (3·36–3·93) 3·69 (3·41–4·00)  Young-onset 97/3,335,485 47/1,053,453 0·03 0·04 1·53 (1·08–2·17) 1·55 (1·09–2·20) 1·58 (1·11–2·24)  Late-onset 1075/2,798,517 1275/874,175 0·38 1·46 3·88 (3·58–4·21) 3·87 (3·57–4·20) 3·93 (3·62–4·27) VD  All 758/4,162,776 565/1,317,422 0·18 0·43 2·41 (2·16–2·68) 2·45 (2·19–2·73) 2·52 (2·26–2·81)  Young-onset 56/3,335,485 20/1,053,453 0·02 0·02 1·14 (0·68–1·89) 1·19 (0·71–1·98) 1·27 (0·76–2·13)  Late-onset 702/2,798,517 545/874,175 0·25 0·62 2·55 (2·28–2·85) 2·60 (2·32–2·91) 2·67 (2·39–2·99) AD, Alzheimer's disease; APOE4, apolipoprotein E4; CI, confidence interval; HR, hazard ratio; VD, vascular dementia. ⁎ Refers to number of cases per 1000 person-years. † Cox proportional hazard regression models were used to examine the association between APOE4 and incident dementia. Model 1 was adjusted for age and gender; Model 2 was adjusted for Model 1 plus education, income, BMI, smoking, physical activity, alcohol consumption, sleep duration, and diet; Model 3 was adjusted for blood pressure, HDL-C, LDL-C, triglycerides, and HbA1c. ǂ Young-onset dementia was defined as dementia diagnosed <65 years. The analysis was conducted among individuals aged younger than 65 years at baseline. § Late-onset dementia was defined as dementia diagnosed ≥65 years. The analysis was conducted among individuals aged older than 65 years at the end of follow-up. Risk for dementia associated with APOE4. AD, Alzheimer's disease; APOE4, apolipoprotein E4; CI, confidence interval; HR, hazard ratio; VD, vascular dementia. Refers to number of cases per 1000 person-years. Cox proportional hazard regression models were used to examine the association between APOE4 and incident dementia. Model 1 was adjusted for age and gender; Model 2 was adjusted for Model 1 plus education, income, BMI, smoking, physical activity, alcohol consumption, sleep duration, and diet; Model 3 was adjusted for blood pressure, HDL-C, LDL-C, triglycerides, and HbA1c. Young-onset dementia was defined as dementia diagnosed <65 years. The analysis was conducted among individuals aged younger than 65 years at baseline. Late-onset dementia was defined as dementia diagnosed ≥65 years. The analysis was conducted among individuals aged older than 65 years at the end of follow-up. As shown in Fig. S5, multimorbidity, FRS, and CAIDE risk scores showed similar prediction performance. When age/APOE4 was incorporated into the risk score calculation, any multimorbidity score performed better than conventional ones. Multimorbidity+age+APOE4 risk score had the highest prediction performance (AUC (95% CI): 82·2% (81·7–82·7%)) followed by multimorbidity plus age (80·9% (80·5–81·4%)) and multimorbidity plus APOE4 scores (73·0% (72·4–73·7%), Table S8). Statistically significant interaction was observed between APOE4 and painful conditions, depression, psychoactive substance abuse, Parkinson's disease, obesity, diabetes, stroke, or high cholesterol for incident dementia. The association between all these individual diseases and incident dementia was stronger among non-APOE4 carriers ( Figure 3 ). Figure 3 Interaction between individual major diseases and APOE4 for incident dementia. APOE4, apolipoprotein E4; CI, confidence interval; CKD, chronic kidney disease; COPD, Chronic obstructive pulmonary disease. Cox proportional hazard regression models were used to examine the association between individual diseases and incident dementia stratified by APOE4. Analysis was adjusted for age, gender, education, income, BMI, smoking, physical activity, alcohol consumption, sleep duration, diet, blood pressure, HDL-C, LDL-C, triglycerides, and HbA1c. Horizontal lines indicate the ranges of the 95% CIs and the vertical dash lines indicate the hazard ratio of 1·0. Fig 3 Interaction between individual major diseases and APOE4 for incident dementia. APOE4, apolipoprotein E4; CI, confidence interval; CKD, chronic kidney disease; COPD, Chronic obstructive pulmonary disease. Cox proportional hazard regression models were used to examine the association between individual diseases and incident dementia stratified by APOE4. Analysis was adjusted for age, gender, education, income, BMI, smoking, physical activity, alcohol consumption, sleep duration, diet, blood pressure, HDL-C, LDL-C, triglycerides, and HbA1c. Horizontal lines indicate the ranges of the 95% CIs and the vertical dash lines indicate the hazard ratio of 1·0. The association between multimorbidity and incident dementia was stronger among non-APOE4-carriers ( P -value for interaction<0·0001). In non-APOE4 carriers, individuals with one to ≥six diseases were 1·43 (1·21–1·68), 1·85 (1·58–2·17), 2·30 (1·96–2·71), 2·53 (2·12–3·00), 3·24 (2·70–3·89), 5·01 (4·21–5·96) times more likely to develop dementia compared to those with no observed disease. The corresponding number for APOE4 carriers were 1·12 (0·95–1·30), 1·38 (1·18–1·61), 1·51 (1·28–1·77), 1·69 (1·42–2·01), 2·47 (2·06–2·97), and 2·84 (2·36–3·41), respectively ( Figure 4 ). Figure 4 Interaction between APOE4 and multimorbidity with incident dementia. APOE4, apolipoprotein E4; CI, confidence interval; PAR, population attributable risk. P -value for interaction between APOE4 and multimorbidity was less than 0·0001 and stratified analysis was conducted. Cox proportional hazard regression models were used to examine the association between the number of diseases at baseline and incident dementia. The analysis was adjusted for age, gender, education, income, BMI, smoking, physical activity, alcohol consumption, sleep duration, diet, blood pressure, HDL-C, LDL-C, triglycerides, and HbA1c. Reference was the group with no disease of interest at baseline. Horizontal lines indicate the ranges of 95% CIs. Fig 4 Interaction between APOE4 and multimorbidity with incident dementia. APOE4, apolipoprotein E4; CI, confidence interval; PAR, population attributable risk. P -value for interaction between APOE4 and multimorbidity was less than 0·0001 and stratified analysis was conducted. Cox proportional hazard regression models were used to examine the association between the number of diseases at baseline and incident dementia. The analysis was adjusted for age, gender, education, income, BMI, smoking, physical activity, alcohol consumption, sleep duration, diet, blood pressure, HDL-C, LDL-C, triglycerides, and HbA1c. Reference was the group with no disease of interest at baseline. Horizontal lines indicate the ranges of 95% CIs. APOE4 was a predictor for all-cause dementia across subgroups of disease number, but HR decreased with the accumulation of diseases. PAR of APOE4 for dementia decreased from 39·6% (95% CI: 33·4–45·7%) among individuals with zero disease to 19 · 6% (15·3–23·9%) among those with ≥6 diseases (Fig. S6). The association between multimorbidity and incident dementia was not moderated by education ( P -value for interaction = 0·15) or household income ( P -value for interaction = 0·0568). The association was statistically significant across subgroups of these socioeconomic factors (Table S9). There was a statistically significant interaction between multimorbidity and age for incident dementia ( P -value < 0·0001). Although multimorbidity was more predictive of dementia in younger than in older individuals, this association was statistically significant among older adults (Table S10). We repeated the analysis for associations between individual diseases and dementia by excluding those who developed dementia in the first five years of follow-up. In the multivariable-adjusted analysis, 32 diseases were statistically significantly associated with the risk of dementia. Of these diseases, 30 were overlapped with the 33 diseases identified in the main analysis. Obesity (HR (95% CI): 1·08 (0·99–1·19)) was not statistically significantly associated with dementia. Whilst stomach cancer (HR (95% CI): 1·84 (1·07–3·17)) was associated with an increased risk of dementia in the sensitivity analysis where this association was not statistically significant in the main analysis (Table S11). The number of diagnosed dementia cases dropped off in 2020 (Table S12). The association between multimorbidity and incident all-cause dementia did not substantially change whether the dementia cases diagnosed in 2020 were included in the analysis or not (Table S13). Similarly, the difference in HRs for dementia associated with APOE4 was minimal between different cut-offs for follow-up (Table S14).

In this large population of community-dwelling adults, we found 33 out of 63 diseases were associated with an increased risk of incident dementia. Traditional risk factors including obesity, diabetes, hypertension, hearing impairment, and depression were leading contributors to dementia. In addition to these conventional diseases, cardiovascular disease (coronary heart disease, heart failure, and stroke), respiratory disorders (COPD and asthma), musculoskeletal disorders (osteoarthritis, fracture, and connective tissue disorders), digestive disorders (diverticulitis and dyspepsia), painful conditions, and CKD were found to be risk factors of dementia. The larger number of diseases the larger HR for incident dementia was observed. The dementia risk score of multimorbidity incorporating age/APOE4 had better prediction performance than conventional scores. These associations were stronger among non-APOE4 carriers than in APOE4 carriers. Individual diseases and multimorbidity were more predictive of young-onset dementia than of late-onset dementia. The present study has several limitations. Firstly, there may be recall bias for the definition of major diseases using self-reported and inpatient hospital data, which may result in an underestimated prevalence of these diseases. However, this is more likely to bias the association towards the null. Secondly, some cases of all-cause dementia might not be captured in the medical records or death registers. However, previous research has shown that there is a good agreement between dementia case ascertainment with primary care records. 26 Thirdly, because of the observational design of our study, causal relationships cannot be established based on our findings. Fourthly, as the prodromal period of dementia can last many years, 27 dementia might have occurred before the diagnosis of chronic conditions. However, our sensitivity analysis by excluding individuals with dementia developed in the first five years of follow-up showed similar results compared with the main findings. This may have helped rule out reverse causation to some extent. Fifthly, many individuals may have had mild cognitive impairment at baseline. This might have biased the associations, although those with cognitive impairment were excluded from the analysis in a subgroup with corresponding cognitive tests. Sixthly, although multiple risk factors for dementia were adjusted for in the analysis, potential confounding from other sources such as levels of beta-amyloid was unpreventable. Seventhly, our findings might be impacted by COVID-19. However, the sensitivity analysis demonstrated the results by setting the cut-off of follow-up at the end of 2019 were similar to those for the main analysis. Finally, the analysis was restricted to European ancestry participants, thus, our findings especially the multimorbidity risk score for dementia need to be validated in other ethnic groups. Our research together with previous studies have confirmed the importance of conventional diseases for the development of dementia. 28 In addition to conventional cardiometabolic disorders, stroke, high cholesterol, coronary heart disease, heart failure, and other cardiac problem conferred risk of dementia. This is not surprising as cardiometabolic disorders and dementia may share pathology. 29 Long-term exposure to high blood pressure and cholesterol may increase motor stiffness and reduce compliance, 29 , 30 and reduce cerebral blood flow and increase cerebrovascular reactivity 28 , 30 thus resulting in brain damage. Besides depression, numerous psychological and mental disorders including anxiety, schizophrenia, Parkinson's disease, multiple sclerosis, alcohol problems, psychoactive substance abuse, and epilepsy were linked to dementia risk. It could be expected that so many psychological and mental disorders were associated with an increased risk of dementia considering that they shared pathology and numerous risk factors. 18 , 31 , 32 In line with previous study, 2 although hearing impairment is not among the diseases with the largest HRs, it is the leading contributor to dementia given its high prevalence in our study. Our study underlines the importance of cardiometabolic disorders, psychological and mental disorders, and hearing impairment on the development of dementia. Less is known regarding whether other diseases confer risk for the development of dementia. In addition to conventional risk factors, multiple diseases have been linked to dementia in a recent prospective study 13 ; however, this study is limited by investigating only 22 hospitalized physical diseases and identifying 8 physical diseases associated with increased dementia risk. We found multiple musculoskeletal disorders (osteoporosis, fracture, and connective tissue disorders), and painful conditions (headache, joint pain) were associated with increased dementia risk, which is consistent with previous studies. 9 , 13 These associations might be attributed to the fact that individuals with musculoskeletal disorders or painful conditions were less likely to exercise and participate in social activity resulting in an increased risk of dementia. The association between fracture and dementia may be contributed to shared risk factors including APOE4, low levels of physical activity, and vitamin D. 33 Whilst fracture as an independent risk factor of dementia has been reported in several studies. 34 , 35 Respiratory disorders (COPD and asthma) were also linked to increased dementia in our study. Although the underlying mechanisms are unclear, our study is supported by recent evidence that poor pulmonary function might accelerate cognitive decline. 10 , 11 Our study is consistent with a previous study demonstrating that digestive disorders (diverticulitis and dyspepsia) were associated with a higher risk of dementia. 13 Inflammation in those with digestive disorders might increase the risk of dementia. In line with previous study, 12 we found CKD was a risk factor for dementia, which might be attributed to shared vascular risk factors and inflammation between CKD and dementia. 36 Our findings suggest respiratory, musculoskeletal, and digestive disorders, painful conditions, and CKD may also play an important role in the development of dementia. In line with our research, several studies have shown that eczema was associated with an increased risk of dementia and decreased cognitive function. 37 , 38 Eczema is associated with itch, pain, sleep disturbance, depression, and psychological distress, 39 all of which may contribute toward cognitive dysfunction/dementia. Multimorbidity should not be overlooked given its high prevalence and detrimental effect on health in adults. 40 We found 51·6% of participants had multimorbidity and a larger number of diseases was associated with larger HR for dementia. Individuals with ≥6 diseases were around four times more likely to develop dementia, and around 51·2% of incident dementia was attributed to one or more observed diseases. The score of multimorbidity incorporating with age was more accurate in dementia prediction than conventional risk scores suggesting a tool based on self-reported information only might be greatly useful for targeting intervention priorities. Only few studies have investigated the association between multimorbidity and incident dementia. 41 , 42 These studies agree with our research highlighting the importance of multimorbidity on the development of dementia. Further analysis shows, with the accumulation of disease, the development of dementia is becoming less determined by genetics (APOE4). One recent study has examined the role of genetics on the association between multimorbidity and dementia but did not find a statistically significant association. 41 Two previous studies showed that APOE4 might modify the association of diabetes and depression with dementia or cognitive decline. 43 , 44 A prospective study of 3444 older adults from the USA has reported that stroke was associated with greater cognitive decline among non-APOE4 carriers but not APOE4 carriers. 45 Likely, we found the association of multimorbidity, hypertension, high cholesterol, diabetes, stroke, depression, Parkinson's disease, and painful conditions with dementia was stronger in non-APOE4 carriers than in APOE4 carriers. Our findings suggest the importance of the prevention or delay the development of these diseases and their multimorbidity for minimizing dementia risk, especially in non-APOE4 carriers although APOE4 carriers with multimorbidity had a higher risk of dementia. Although the prevalence was low, 18 young-onset dementia might be more detrimental as it affects individuals in the midst of their careers. We found young-onset dementia than late-onset dementia was less determined by APOE4. This is consistent with previous studies showing that APOE4 plays a more important role in AD in older than in younger individuals. 14 , 15 The relative risk for young-onset dementia associated with 29 diseases was larger than that for late-onset dementia. Besides these diseases, chronic liver disease was associated with an increased risk of young-onset dementia but not late-onset dementia. Similarly, a larger relative risk for young-onset dementia associated with multimorbidity was observed for late-onset dementia. Although several diseases have been associated with increased young-onset dementia risk, 46 our study provides evidence on a wide range of diseases as modifiable risk factors for young-onset dementia. This is important as genetics (APOE4) is less predictive of young-onset dementia. We found a wide range of major chronic diseases (some have not been investigated previously) that might play an important role in the development of dementia. This may help identify individuals at higher risk of dementia. Notably, we created a multimorbidity risk score (plus age and APOE4) with better prediction performance of dementia than conventional risk scores, which would help clinicians to make plans for the prevention of dementia among patients with specific chronic diseases/multimorbidity. To our knowledge, this is the first study to examine the interaction between multiple diseases and multimorbidity and APOE4 for incident dementia. This provides evidence on priority for the emerging anti-APOE4 therapy. 16 This study also uniquely examined associations between diseases/APOE4 and young-onset and late-onset dementia, which is important for a better understanding of the underlying mechanisms of dementia. In conclusion, 33 well-known and emerging diseases are associated with an increased risk of dementia. A larger number of these diseases is associated with a larger excessive relative risk for dementia. The multimorbidity risk score may be useful for the prediction and screening of dementia. These associations are stronger among non-APOE4 carriers. APOE4 is a strong predictor of dementia, but this association is attenuated with the accumulation of diseases. A larger relative risk associated with individual diseases and their multimorbidity is observed for young-onset dementia than for late-onset dementia, whereas APOE4 is a strong predictor for late-onset but not young-onset dementia. These findings may facilitate the identification of individuals at higher risk of dementia. XS, and MH conceived and designed the study. XS, ZZ, and WW verified the data. XS conducted data analysis, data interpretation and drafted the initial manuscript. XS, ZZ, XLZ, YH, XYZ, JL, HY, WW, ST, ZG, XY, and MH made critical revisions of the manuscript for important intellectual content. All authors read the manuscript and approved the final draft. All authors should confirm that they had full access to all the data in the study and accept responsibility to submit for publication. Data are available in a public, open access repository ( https://www.ukbiobank.ac.uk/ ). The present study was conducted under application number 62,443 of the UK Biobank resource.

Dementia was the seventh leading cause of mortality accounting for 1·6 million deaths globally in 2019. 1 Although age-specific incidence rates of dementia are lower in more recent cohorts compared with cohorts from previous decades in high-income countries including the UK, 2 a modelling study suggests the number of people with dementia in the UK would increase from 0·77 million in 2016 to 1·20 million in 2040 given the increasing ageing population. 3 It is critical to identify important determinants for dementia, as there is no effective way to stop the progression of dementia yet. 2 , 4 Chronic conditions including hypertension, obesity, diabetes, hearing loss, depression, and traumatic brain injury have been associated with an increased risk of dementia. 2 , 5 These diseases together with other well-known risk factors accounted for less than 40% of the development of dementia. 4 A growing number of studies have linked eye diseases to dementia, 6 , 7 but it is unknown whether the importance of these eye diseases is comparable to that of conventional diseases. There are conflicting results for the association between cancers and Alzheimer's disease (AD) between previous studies. 8 Little is known regarding the association between types of cancers and incident dementia. It is also of great importance to confirm the associations of emerging risk factors for dementia including musculoskeletal disorders, painful conditions, 9 respiratory disorders (such as chronic obstructive pulmonary disease [COPD]), 10 , 11 chronic kidney disease (CKD), 12 and digestive disorders 13 with dementia in large prospective cohort studies. A recent prospective study has identified several physical diseases including erysipelas, hypothyroidism, duodenal ulcer, gastritis, and duodenitis (in addition to conventional risk factors), that may increase the long-term risk of dementia. 13 However, this study is limited by investigating a small number of chronic diseases. Strong evidence has demonstrated the importance of apolipoprotein E4 (APOE4) on the development of dementia especially AD. 14 , 15 APOE4 has also been considered as a therapeutic target for AD, 16 given that APOE4 may modify the association of multiple biomarkers including vasculature, insulin, and inflammation with AD. 16 , 17 Investigating the interaction between genetics and individual diseases and multimorbidity for the development of dementia may provide evidence on priority for anti-APOE4 therapy. However, limited data from such studies are available. Although dementia usually develops in older people, young-onset dementia also deserves scrutiny as its more detrimental effect on life. 18 APOE4 has been shown to play a more important role in the development of late-onset AD than in young-onset AD, 14 , 15 but less is known regarding associations of non-genetic factors such as major diseases with young-onset and late-onset dementia. Using the data of the UK Biobank study, the main aim of the present study was to investigate associations of a wide range of major diseases and their multimorbidity with incident dementia. Meanwhile, we aimed to create a multimorbidity risk score based on the associations between individual diseases and incident dementia. Thirdly, the analysis for young-onset and late-onset dementia was also conducted. Finally, we examined the interaction between APOE4 and diseases with dementia.

The authors have no conflicts of interest to declare that are relevant to the content of this article.