Immune-mediated diseases are associated with a higher incidence of dementia: a prospective cohort study of 375,894 individuals

The participants and data resources involved in this study were from a population-based prospective cohort UKB, under application number 19542. Between 2006 and 2010, over 500,000 participants aged 40–69 years were recruited in UKB with a comprehensive baseline assessment of health status at 22 centers throughout the UK [16]. Follow-up is conducted chiefly through linkages to routinely available national datasets, including death register data, cancer register data, hospital inpatient data, and primary care data (https://​www.​ukbiobank.​ac.​uk/​). Environmental, genetic, disease status, and other clinical data were obtained from questionnaires, physical measures, sample assays, and linked electronic health data. Ethics approval for the UKB study was obtained from the North West Multicenter Research Ethical Committee. All participants provided written informed-consent paper at baseline. We restricted analyses to individuals with hospital inpatient data where the diagnosis information based on International Classification of Diseases (ICD)-10 codes were available to identify immune-mediated diseases patients. The control comparators were defined as the individuals with hospital inpatient data but none of the immune-mediated diseases was diagnosed. Prevalent dementia patients at baseline and those without traceable outcome data were excluded.

We identified immune-mediated diseases' status and diagnosis date by matching the corresponding ICD-10 codes [17] from hospital inpatient data (HESIN). If two or more different diagnosis dates were recorded for the same disease, the earliest one was used as the primary diagnosis date. Only the diseases with at least 500 affected individuals were kept for further analyses because the dementia incidence was low in some rare immune-mediated diseases. A total of 20 immune-mediated diseases were ascertained (Supplementary Table 1). We assessed the longitudinal associations of any immune-mediated diseases and the individual immune-mediated diseases with dementia. The immune-mediated diseases were required to be diagnosed at least one year before the dementia diagnosis.

Incident all-cause dementia (ACD), AD, and vascular dementia (VD) were ascertained and classified according to the ICD codes and Read codes (Supplementary Table 2). ACD covered all the types of dementia including AD, VD, FTD, dementia with Lewy bodies (DLB), Parkinson’s disease dementia (PDD), dementia in corticobasal degeneration (CBD), dementia in other neurodegenerative, and specified diseases. The outcome records were extracted from algorithmically defined (Fields 42,018–42,025), first occurrences data reported (Fields 131,036–131,037, 130,836–130,843), death register data documented (Fields 40,001–40,002), HESIN summarized (Fields 41,270–41,271, 41,280–41,281), and primary care data recorded (Field 42,040) dementia. Follow-up visits began from the date of attending the assessment center (Field 53) to the earliest incident dementia diagnosis, date of death, the last data collection date by the general practitioner, or the last time of hospital inpatient admission, whichever occurred first.

Blood samples collected in ethylenediaminetetraacetic acid (EDTA) vacutainers of the UKB participants were analyzed at the UKB central laboratory within 24 h of blood draw. Differential leukocytes counts were acquired from the automated quantitative analyzer, the Beckman Coulter LH750 Hematology Analyzer. We extracted the blood count data of neutrophils and lymphocytes collected after immune-mediated diseases diagnosis and before dementia onset.

Baseline characteristics of participants were summarized for those with and without incident dementia status as mean and standard deviation (SD) for continuous variables, number, and percentage for categorical variables. Longitudinal associations of immune-mediated diseases with incident dementia were examined with multivariable time-varying Cox proportional hazard regression models and presented with hazard ratio (HR) and 95% confidence interval (CI). Participants who died, withdrew from the study before the end of follow-up, or had not developed dementia by the end of follow-up (April 7, 2021) were censored. In the primary analysis, we performed the model unadjusted, then adjusted for age, sex, ApoE-ε4, education, ethnicity, body mass index (BMI), socioeconomic status indicated by Townsend deprivation score, smoking, and alcohol consumption. The p values were further adjusted to control the false discovery rate (FDR) at 5% using the Benjamini–Hochberg procedure (labeled as Q values) when analyzing the associations of 20 individual immune-mediated diseases with dementia. In secondary analyses, we explored the effect of time since immune-mediated diseases on dementia. Time since immune-mediated diseases was split into overlapping periods (1 to 2 years, 1 to 3 years, 1 to 4 years, 1 to 5 years, 1 to 10 years, 1 to 15 years, 1 to 20 years, and 1 to over 20 years). We also estimated the effect differences of immune-mediated diseases on dementia across sex groups at recruitment. To evaluate the potential reverse causation biases, we also performed the sensitivity analyses by excluding the participants with follow-up time less than 5 or 10 years as suggested in work conducted by Sipila et al. [18]. Finally, the mediation analyses based on the methods proposed by Baron and Kenny [19] with the significance determined using 10,000 bootstrapped iterations were conducted to investigate whether the associations of immune-mediated diseases with dementia were mediated by peripheral immune cells which were proven to be related to dementia incidence in our previous study based on UKB. In addition, the interaction analyses were performed between sex and peripheral immune cells in contributing to the dementia incidence.

Among 502,493 participants at baseline in the UKB, 375,894 participants with HESIN ICD-10 data were eligible for Cox analysis after excluding 888 participants with prevalent dementia at baseline and 33,527 participants without follow-up data (Supplementary Fig. 1). Baseline characteristics of the included participants are presented by incident dementia status (Table 1) and immune-mediated diseases status (Supplementary Table 3). Overall, the mean age of participants was 57.32 (± 7.99) years, and 205,575 (54.69%) of them were females. Among these participants, 58,589 were diagnosed with at least one of the 20 immune-mediated diseases 1 year before dementia onset. During a median follow-up of 9.08 years, 5291 participants developed dementia, of which 2301 cases were AD and 1249 cases were VD.

Immune-mediated diseases in whole were associated with increased risk of ACD (HR, 1.10; 95% CI, 1.00–1.21) and VD (HR, 1.43; 95% CI, 1.19–1.73) while not associated with AD after multivariable adjustment for age, sex, education, ApoE-ε4, ethnicity, BMI, Townsend deprivation score, smoking, and alcohol consumption (Fig. 1 and Supplementary Table 4). When split the time since immune-mediated diseases into overlapping periods, the results were still statistically significant for ACD and VD in all periods. A similar risk efficacy was also observed for AD in three years before immune-mediated diseases diagnosis (Fig. 2). The risk of dementia was highest between 1 and 2 years after immune-mediated diseases onset for ACD (HR, 2.74; 95% CI, 1.86–4.04), AD (HR, 2.15; 95% CI, 1.12–4.16), and VD (HR, 5.39; 95% CI, 2.95–9.84). Subgroup analyses of sex stratification indicated that females with immune-mediated diseases were more prone to AD in the early years after immune-mediated disease (Fig. 2), while males were more susceptible to VD during a long follow-up time period (Fig. 2 and Supplementary Table 4). Sensitivity analyses showed that the risk associations with ACD and VD remained when immune-mediated diseases occurred more than 5 years before the diagnosis of dementia (Supplementary Table 5). While when dementia cases occurred from 10 years onward, the associations for ACD only persisted in models unadjusted or adjusted for age, sex, ApoE-ε4, and education, and the associations for VD were consistently robust in all models (Supplementary Table 5).

For individual immune-mediated diseases, participants were at higher risk of developing ACD when suffering from type I diabetes mellitus (HR, 2.49; 95% CI, 1.97–3.15), rheumatic fever or rheumatic heart diseases (HR, 1.36; 95% CI, 1.05–1.77), multiple sclerosis (HR, 2.87; 95% CI, 1.92–4.30), and necrotizing vasculopathies (HR, 1.71; 95%CI, 1.03–2.85) (Fig. 3 and Supplementary Table 6). However, the effect on subtypes of dementia varies in different immune-mediated diseases. Type I diabetes mellitus was the only disease consistently related to a higher risk of both AD (HR, 2.21; 95% CI, 1.49–3.28) and VD (HR, 4.25; 95% CI, 2.90–6.23). Necrotizing vasculopathies were related to a higher incidence of AD (HR, 2.13; 95% CI, 1.06–4.27), while ulcerative colitis was related to a lower incidence of AD (HR, 0.37; 95% CI, 0.17–0.83). Rheumatic fever or rheumatic heart diseases (HR, 1.36; 95% CI, 1.05–1.77) and psoriasis (HR, 1.36; 95% CI, 1.05–1.77) were proven to be risk factors for VD. In further FDR adjusting, the Q values remained statistically significant for associations of type I diabetes mellitus with ACD, AD and VD, for multiple sclerosis with ACD, for rheumatic fever or rheumatic heart diseases and psoriasis with VD (Supplementary Table 6). Sensitivity analyses showed that the results remained significant for associations of ACD with type I diabetes mellitus and multiple sclerosis, VD with type I diabetes mellitus and rheumatic fever or rheumatic heart diseases when immune-mediated diseases occurred more than 5 or 10 years before the diagnosis of dementia (Supplementary Table 7).

Mediation analyses revealed that the associations between any immune-mediated diseases and dementia was partially mediated by peripheral immune cells, including neutrophils with the proportion of mediation 5.09% (p < 0.001) and lymphocytes with the proportion of mediation 1.55% (p = 0.024) respectively (Fig. 4). As for individual immune-mediated diseases, the risk effects on dementia of type I diabetes mellitus (Supplementary Fig. 2A), multiple sclerosis (Supplementary Fig. 2B), rheumatic fever or rheumatic heart diseases (Supplementary Fig. 2C), and necrotizing vasculopathies (Supplementary Fig. 2D) were partially mediated by neutrophils while not by lymphocytes. What is more, the interaction analyses results indicated that there was an interaction between sex and peripheral immune cells in contributing to the dementia incidence (Supplementary Table 8).