Hospital-treated infectious diseases and the risk of dementia: a large, multicohort, observational study with a replication cohort

Abstract

Background: Infections have been hypothesised to increase the risk of dementia. Existing studies have included a narrow range of infectious diseases, relied on short follow-up periods, and provided little evidence for whether the increased risk is limited to specific dementia subtypes or attributable to specific microbes rather than infection burden. We aimed to compare the risk of Alzheimer's disease and other dementias across a wide range of hospital-treated bacterial and viral infections in two large cohorts with long follow-up periods.

Methods: In this large, multicohort, observational study, the analysis was based on a primary cohort consisting of pooled individual-level data from three prospective cohort studies in Finland (the Finnish Public Sector study, the Health and Social Support study, and the Still Working study) and an independent replication cohort from the UK Biobank. Community-dwelling adults (≥18 years) with no dementia at study entry were included. Follow-up was until Dec 31, 2012, in the Health and Social Support study, Dec 31, 2016, in the public sector study and the Still Working study, and Feb 7, 2018, in the replication cohort. Through record linkage to national hospital inpatient registers, we ascertained exposure to 925 infectious diseases (using the International Classification of Diseases 10th Revision codes) before dementia onset, and identified incident dementia from hospital records, medication reimbursement entitlements, and death certificates. Hazard ratios (HRs) for the associations of each infectious disease or disease group (index infection) with incident dementia were assessed by use of Cox proportional hazards models. We then repeated the analysis after excluding incident dementia cases that occurred during the first 10 years after initial hospitalisation due to the index infection.

Findings: From March 1, 1986, to Jan 1, 2005, 260 490 people were included in the primary cohort, and from Dec 19, 2006, to Oct 1, 2010, 485 708 people were included in the replication cohort. In the primary cohort analysis based on 3 947 046 person-years at risk (median follow-up 15·4 years [IQR 9·8-21·0]), 77 108 participants had at least one hospital-treated infection before dementia onset and 2768 developed dementia. Hospitalisation for any infectious disease was associated with increased dementia risk in the primary cohort (adjusted HR [aHR] 1·48 [95% CI 1·37-1·60]) and replication cohort (2·60 [2·38-2·83]). The association remained when analyses were restricted to new dementia cases that occurred more than 10 years after infection (aHR 1·22 [95% CI 1·09-1·36] in the primary cohort, the replication cohort had insufficient follow-up data for this analysis), and when comorbidities and other dementia risk factors were considered. There was evidence of a dose-response association between the number of episodes of hospital-treated infections and dementia risk in both cohorts (p_trend=0·0007). Although the greatest dementia risk was seen for central nervous system (CNS) infections versus no infection (aHR 3·01 [95% CI 2·07-4·37]), excess risk was also evident for extra-CNS infections (1·47 [1·36-1·59]). Although we found little difference in the infection-dementia association by type of infection, associations were stronger for vascular dementia than for Alzheimer's disease (aHR 2·09 [95% CI 1·59-2·75] versus aHR 1·20 [1·08-1·33] in the primary cohort and aHR 3·28 [2·65-4·04] versus aHR 1·80 [1·53-2·13] in the replication cohort).

Interpretation: Severe infections requiring hospital treatment are associated with long-term increased risk of dementia, including vascular dementia and Alzheimer's disease. This association is not limited to CNS infections, suggesting that systemic effects are sufficient to affect the brain. The absence of infection specificity combined with evidence of dose-response relationships between infectious disease burden and dementia risk support the hypothesis that increased dementia risk is driven by general inflammation rather than specific microbes.

Funding: UK Medical Research Council, US National Institute on Aging, Wellcome Trust, NordForsk, Academy of Finland, and Helsinki Institute of Life Science.

Figure 1

Classification of hospital-treated infectious diseases and the number of cases in the primary cohort The number of cases of different infectious diseases add up to more than the total number of infection cases, because some participants were admitted to hospital for more than one infectious disease. *Too rare to be analysed separately.

Figure 2

Risk of dementia associated with hospital-treated infectious diseases in the full follow-up and after 10 years or more from the onset of infection in the primary cohort Error bars are 95% CIs. HRs are adjusted for sex and socioeconomic status, and age is the timescale. aHR=adjusted hazard ratio. *Difference in the association of bacterial and viral infections with dementia. †Difference in the association of herpes virus infections and other persistent viral infections with dementia. ‡Difference in the association of acute viral infection and that of herpes and other persistent viral infections with dementia.

Figure 3

Multivariable-adjusted associations between hospital-treated infections and dementia by dementia type in the replication cohort Data are adjusted HRs (95% CIs), unless otherwise specified. Error bars are 95% CIs. Model 1 was adjusted for age (as the timescale), sex, and socioeconomic status. Numbers of participants, dementia cases, and the forest plot are for this model. Model 2 used the same adjustment criteria as model 1 and excluded participants with HIV infection; it was based on 485 453 participants (2131 [0·4%] with dementia) with complete information for analysis of any infection, 471 511 participants (2044 [0·4%] with dementia) for analysis of bacterial infections, and 400 502 participants (1322 [0·3%] with dementia) for analysis of viral infections. Model 3 used the same criteria as model 2 and additionally adjusted for alcohol drinking, smoking, body-mass index, hypertension, and diabetes; it was based on 480 842 participants (2080 [0·4%] with dementia) with complete information for analysis of any infection, 467 058 participants (1995 [0·4%] with dementia) for analysis of bacterial infections, and 397 333 participants (1304 [0·3%] with dementia) for analysis of viral infections. Model 4 used the same criteria as model 3 and additionally adjusted for apolipoprotein E genotype; it was based on 470 551 participants (2025 [0·4%] with dementia) with complete information for analysis of any infection, 457 104 participants (1945 [0·4%] with dementia) for analysis of bacterial infections, and 389 067 participants (1267 [0·3%] with dementia) for analysis of viral infections. aHR=adjusted hazard ratio. *The number of patients with dementia and viral infection was less than five for frontotemporal dementia and Parkinson's disease dementia.

Figure 4

Risk of dementias associated with infection burden, simultaneous infections, and CNS vs extra-CNS infections in the primary cohort and replication cohort Infections are hospital-treated. Error bars are 95% CIs. HRs are adjusted for sex and socioeconomic status, and age is the timescale. aHR=adjusted hazard ratio. NA=not applicable. *For subgroup difference. †If the participant was admitted to hospital several times for exactly the same diagnosis, only the first counted towards infection burden because the UK Biobank data included only the first hospitalisation for each diagnosis. ‡Fewer than five Alzheimer's disease and vascular dementia cases among those exposed to CNS infections.