Received: 16 June 2020 Revised: 20 October 2020 Accepted: 24 October 2020 DOI: 10.1002/alz.12237 F E AT U R E D A R T I C L E Multimorbidity burden and dementia risk in older adults: The role of inflammation and genetics Giulia Grande1,* Alessandra Marengoni1,2,* Davide L. Vetrano1,3 Albert Roso-Llorach4,5 Debora Rizzuto1,7 Alberto Zucchelli6 Chengxuan Qiu1 Laura Fratiglioni1,7 Amaia Calderón-Larrañaga1 1 Aging Research Center, Department of Neurobiology, Care Sciences and Society, Karolinska Institutet and Stockholm University, Stockholm, Sweden 2 Department of Clinical and Experimental Sciences, University of Brescia, Brescia, Italy 3 Centro di Medicina dell’Invecchiamento, IRCCS Fondazione Policlinico “A. Gemelli” and Università Cattolica del Sacro Cuore, Rome, Italy 4 Fundació Institut Universitari per a la recerca a l’Atenció Primària de Salut Jordi Gol i Gurina (IDIAPJGol), Barcelona, Spain 5 Universitat Autònoma de Barcelona, Bellaterra (Cerdanyola del Vallès), Spain 6 Department of Information Engineering, University of Brescia, Brescia, Italy 7 Stockholm Gerontology Research Center, Stockholm, Sweden Correspondence Giulia Grande, Aging Research Center, Karolin- Abstract ska Institutet, Tomtebodavägen 18A, 171 65 Solna, Sweden. Introduction: We investigate dementia risk in older adults with different disease pat- Email: giulia.grande@ki.se terns and explore the role of inflammation and apolipoprotein E (APOE) genotype. * Methods: A total of 2,478 dementia-free participants with two or more chronic Contributed equally diseases (ie, multimorbidity) part of the Swedish National study on Aging and Care in Funding information Kungsholmen (SNAC-K) were grouped according to their multimorbidity patterns and Swedish National study on Aging and Care; Ministry of Health and Social Affairs; Swedish followed to detect clinical dementia. The potential modifier effect of C-reactive protein Research Council, Grant/Award Number: (CRP) and apolipoprotein E (APOE) genotype was tested through stratified analyses. 2016-00981; Swedish Research Coun- cil for Health, Working Life and Welfare, Results: People with neuropsychiatric, cardiovascular, and sensory impairment/cancer Grant/Award Number: 2017-01764; Italian multimorbidity had increased hazards for dementia compared to the unspecific (Hazard Ministry of Health, Grant/Award Number: PE-2016-02364885 ration (HR) 1.66, 95% confidence interval [CI] 1.13-2.42; 1.61, 95% CI 1.17-2.29; 1.32, 95% CI 1.10-1.71, respectively). Despite the lack of statistically significant interaction, high CRP increased dementia risk within these patterns, and being APOE ε4 carriers heightened dementia risk for neuropsychiatric and cardiovascular multimorbidity. Discussion: Individuals with neuropsychiatric, cardiovascular, and sensory impair- ment/cancer patterns are at increased risk for dementia and APOE ε4, and inflam- mation may further increase the risk. Identifying such high-risk groups might allow tailored interventions for dementia prevention. KEYWORDS dementia, genetics, inflammation, multimorbidity patterns This is an open access article under the terms of the Creative Commons Attribution-NonCommercial License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited and is not used for commercial purposes. © 2020 The Authors. Alzheimer’s & Dementia published by Wiley Periodicals LLC on behalf of Alzheimer’s Association 768 wileyonlinelibrary.com/journal/alz Alzheimer’s Dement. 2021;17:768–776. GRANDE ET AL . 769 1 INTRODUCTION RESEARCH IN CONTEXT As a consequence of the worldwide aging of the population, demen- 1. Systematic review. After conducting a systematic review tia will increasingly become a challenge faced by individuals and of the literature (PubMed and Embase) we found a limited societies.1 Owing to the current lack of curative drugs, dementia pre- number of studies investigating either cross-sectional vention has become an especially relevant strategy.2 As such, the iden- associations or the effect of the number of co-existing dis- tification of individuals at higher risk of dementia is a clinical and public eases on dementia. No studies have been conducted to health priority.3 date on the impact of specific multimorbidity patterns on Several longitudinal studies have identified specific diseases (eg, dia- incident dementia. betes, atrial fibrillation, depression) as at-risk conditions for increased 2. Interpretation. Over a 12-year period, people with dementia incidence.4,5 However, it is relatively uncommon for older neuropsychiatric, cardiovascular, and sensory impair- adults to be affected by only one chronic disease. In fact, up to ment/cancer multimorbidity showed the highest hazards 90% of adults 60 years of age or older adults have two or more of dementia, and the presence of inflammation and APOE chronic diseases,6 a condition referred to as multimorbidity. Multimor- ε4 allele further increased this association. Identifying bidity has been associated with incident mild cognitive impairment7 multimorbidity patterns in the presence of inflammation, and greater brain pathology burden.8 However, there is scanty and and among APOE ε4 carriers, might allow for more tai- contrasting evidence of how clusters of diseases in individuals with lored clinical interventions aimed at dementia prevention. multimorbidity are differentially associated with the occurrence of 3. Future directions. Future studies specifically designed dementia.9 Some diseases tend to aggregate in the same person due to evaluate the interplay between aging-related inflam- to similar pathophysiological mechanisms and/or shared risk factors,10 matory processes, genetic predisposition, multimorbidity which leads to multimorbidity patterns characterized by chronic disor- patterns, and dementia are needed to explore the biolog- ders that systematically cluster together beyond chance. The identifi- ical mechanisms behind these associations. cation of such patterns has provided useful clinical insight,11 suggest- ing a potentially differential impact of specific multimorbidity patterns on dementia development. One of the factors plausibly implicated in the association between specific multimorbidity patterns and dementia is inflammation. Sys- pation rate: 73%). Younger participants (age <78 years) were followed temic inflammation may accelerate the progression of neurodegen- up every 6 years, and the older (age ≥78 years) every 3 years. In the eration and vascular pathology in the brain,12 and an imbalance present study, we identified a dementia-free cohort of 2622 persons between pro- and anti-inflammatory responses is described in older affected by multimorbidity (two or more chronic diseases) at baseline individuals.13 Likewise, as the role of the apolipoprotein E gene vari- (Figure S1). ant (APOE ε4) on dementia risk is well known, it is plausible to The Regional Ethical Review Board in Stockholm, Sweden approved hypothesize that the combination of specific multimorbidity patterns the protocols of SNAC-K. All participants, or next of kin for cognitively and APOE ε4 allele puts individuals at even greater risk of dementia impaired individuals, provided written informed consent. development. The results of this study are reported following the strengthening In this study, we aimed to quantify the impact of specific pat- the reporting of observational studies in epidemiology (STROBE) rec- terns of multimorbidity on dementia risk, and to explore the role ommendations. played by inflammation and APOE genotype in such an association over 12 years. 2.2 Data collection 2 METHODS At each study wave, SNAC-K participants underwent a comprehensive clinical and functional assessment by trained physicians, nurses, and 2.1 Study population psychologists, following standard procedures. Home visits were car- ried out for those who agreed to participate but were unable to travel Data were gathered from the Swedish National study on Aging and Care in to the research center. Kungsholmen (SNAC-K), an ongoing population-based study.14 SNAC- K includes a random sample of community-dwelling and institutional- ized adults aged 60 years and older, living in the Kungsholmen district 2.2.1 Covariates of Stockholm, Sweden. At baseline (2001 to 2004), a random sample of 5111 people from 11 age cohorts were invited to participate in the Participants’ sociodemographic information (ie, age, sex, education, study. Of the 4590 eligible individuals, 3363 were examined (partici- and civil status) were collected during nurses’ interviews. Educational 770 GRANDE ET AL . attainment was categorized as elementary, high school, and university 2.3 Statistical analysis or higher, indicating the highest level of education attained. Civil status was categorized as unmarried, married, divorced, and widowed. Body 2.3.1 Patterns identification mass index (BMI) was obtained by dividing participants’ weight by their squared height (kg/m2 ). A BMI <18.5 kg/m2 was considered as a proxy Multimorbidity patterns at baseline were identified using a fuzzy of malnutrition.15 The Mini-Mental State Examination (MMSE) score c-means cluster analysis algorithm, which allows for the cluster- was used as a measure of global cognition. ing individuals based on their underlying combinations of chronic diseases.10 Thirty-five of 60 individual chronic disease categories were included in the cluster analysis; those with a prevalence of <2% were 2.2.2 Potential modifiers excluded to avoid statistical noise, and therefore, spurious findings in the models (Table S1). To further characterize the multimorbidity Serum C-reactive protein (CRP) was measured at Karolinska Hospital, patterns, observed/expected ratios were calculated by dividing the Stockholm, through an assay with a lower detection limit of 5 mg/L (ie, prevalence of a given disease within a cluster by its prevalence in low-sensitivity test). Consequently, such cut-off was used to define low the overall population. Disease exclusivity, defined as the fraction of versus high CRP concentrations.16 DNA was extracted from peripheral participants with the disease included in the cluster over the total blood samples and APOE alleles were genotyped. Based on the APOE number of participants with the disease, was also calculated. Diseases alleles, participants were categorized as ε4 carriers (at least one ε4 with both observed/expected ratio ≥2 and exclusivity ≥25% were allele) versus non-carriers. Serum CRP levels and APOE genotype were considered as characterizing a given cluster.17,18 Such criteria were considered at baseline. used to name the multimorbidity patterns based on the diseases that characterized them. The fuzzy c-means algorithm assigns a probability of cluster membership for all individuals within each cluster; however, 2.2.3 Chronic disease assessment participants were finally allocated a single cluster based on their highest membership probability. Details on the methodology are Details on definition and classifications of diseases are reported reported in the Supplementary Text and Table S3. The association elsewhere.6 Briefly, physicians collected information on diagnoses between multimorbidity patterns and dementia was tested through via physical examination, medical history, self-reported information, Cox regression models, and adjusted hazard ratios (HRs) with 95% and/or proxy interviews. Clinical parameters, lab tests, medication, and confidence intervals (CIs) were obtained. The proportional hazard inpatient and outpatient care data were also used to identify specific assumption was assessed by regressing the scaled Schoenfeld resid- conditions. All diagnoses were coded according to the International uals against survival time. No deviation from the proportional hazard Classification of Diseases, Tenth Revision (ICD-10) and classified into assumption was detected. Follow-up time was estimated as the time 60 chronic disease categories in accordance with a clinically driven from study entry until dementia diagnosis, death, or last examination, methodology. Drugs were classified and coded in accordance with using whichever occurred first. Interactions between APOE genotype, the Anatomical Therapeutic Chemical (ATC) classification. Chronic dis- CRP levels, and multimorbidity patterns were tested and stratified eases were considered at baseline. analyses were conducted. 2.2.4 Dementia diagnosis 2.3.2 Sensitivity analyses The diagnosis of clinical dementia was made in accordance with the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition To test the strength and consistency of our results, we repeated the (DSM-IV) criteria, following a three-step procedure. First, a preliminary Cox regression models, excluding those people who developed demen- diagnosis was made by the examining physician, followed by a second tia within the first 6 years of follow-up. In addition, to investigate the preliminary diagnosis by a reviewing physician also involved in data col- association of multimorbidity clusters with dementia in comparison lection. This process was blinded to the diagnosis of the first physician. with non-multimorbid participants, we repeated the survival analysis In cases of disagreement between the first and the second diagnoses, including multimorbidity-free individuals (N = 420) as reference group. the final diagnosis was made by a neurologist who was external to the The proportion of individuals with missing data was 5.5% for serum data collection process (LF, GG). For those people who died between CRP, 8.6% for APOE genotype, and 1% for civil status and education. the follow-up assessments, we gathered additional clinical informa- A complete case data analysis was carried out based on 88.0% of the tion for the diagnosis of dementia from the clinical charts and medical cohort. Those with missing data in these variables were more likely to records, the discharge diagnoses of the National Patient Register, and be older, female, and less educated than those with no missing data at the Swedish National Cause of Death Register. baseline. GRANDE ET AL . 771 TA B L E 1 Baseline characteristics of study sample by multimorbidity patterns Total sample Neuro- psychiatric Cardiovascular Sensory impairment/ Respiratory/metabolic/ Unspecific (N = 2,478) (n = 183) (n = 243) cancer (n = 384) musculoskeletal (n = 503) (n = 1,165) Age, mean (SD) 75.0 (10.4) 72.5 (10.0) 82.4 (8.5) 83.4 (8.8) 71.6 (8.9) 72.6 (9.5) Female, % 64.3 70.5 60.0 67.2 59.6 65.3 Education, % Elementary 17.0 13.7 26.0 21.5 18.3 13.4 High school 51.6 48.6 55.4 53.9 51.8 50.6 University 31.4 37.7 18.6 24.6 29.9 36.0 Civil status, Partnered, % 44.2 39.6 40.0 30.3 47.7 49.7 APOE ε4 allele, % 29.2 27.3 23.9 24.8 27.7 32.6 Serum CRP > 5 mg/L, % 21.9 26.0 30.6 25.1 28.8 15.6 Malnutrition*, % 3.2 3.3 5.8 5.7 1.4 2.6 Number of diseases, 4.4 (2.2) 5.3 (2.1) 7.7 (2.4) 5.8 (1.9) 4.5 (1.8) 3.2 (1.2) mean (SD) MMSE score, mean (SD) 28.5 (1.9) 28.7 (1.6) 27.9 (2.3) 27.6 (2.3) 28.9 (1.4) 28.8 (1.6) Abbreviations: APOE ε4, apolipoprotein epsilon 4 (at least 1 allele); CRP, C-reactive protein; MMSE, Mini Mental State Examination. The distribution of all variables was significantly different across multimorbidity patterns (P < 0.05). *Defined as body mass index <18.5 kg/m2 . Missing data: 8 in educational level; 6 in civil status; 189 in APOE genotype, 136 in CRP, 5 in MMSE. A two-tailed P-value < 0.05 was considered statistically significant likely to belong to the unspecific multimorbidity pattern (50% vs in all analyses. All analyses were performed using Stata version 15 46%). (StataCorp, Texas, USA). Of the 2478 participants with complete data, 506 (20%) devel- oped dementia; among them, 383 were identified during SNAC-K assessments and 123 based on clinical charts and the Cause of Death 3 RESULTS Register. Participants who developed dementia were older and more frequently women, with lower educational level, a greater number of 3.1 Patterns identification chronic conditions, lower MMSE score, and more likely to be APOE ε4 allele carriers (P < 0.001 for all). At baseline, the following multimorbidity patterns were identified: neuropsychiatric (7.6%), cardiovascular (9.5%), sensory impairment/cancer (14.8%), and respiratory/metabolic/musculoskeletal multimorbidity 3.1.1 Baseline characteristics (20.8%). The remaining 47.3% of the population was grouped in a cluster containing diseases that were not overexpressed in relation At study entry, the mean (SD) age of the population was 75.0 years to the overall population prevalence, and thus was named “unspe- (10.4); 64.3% were women; 83.2% had high school or university level cific multimorbidity.” This pattern was considered as reference group education; and 56.0% were unmarried, divorced, or widowed. Baseline because it includes individuals with the lowest mean number of chronic sample characteristics by multimorbidity pattern are shown in Table 1. diseases and CRP level. In addition, in previous studies by our group, we showed that individuals in the unspecific pattern were also less disabled that those in the other patterns10 and presented the lowest 3.1.2 Multimorbidity patterns and dementia risk 6- and 12-year crude mortality.19 (Table S2 reports the list of chronic diseases included in each pattern.) Table 2 shows the association between the multimorbidity patterns During the 12-year follow-up (mean 8.4 years ± 3.9) of the 2622 and dementia risk. In the fully adjusted model (by age, sex, educa- participants, 1003 died and 466 dropped out (Figure S1). Those tion, civil status, malnutrition, baseline MMSE score, APOE genotype, who dropped out were more likely to be younger (difference: −1.49 and CRP level), individuals with neuropsychiatric multimorbidity pre- years; 95% CI −2.35, −0.46), and with a lower burden of chronic sented with a 66% higher hazard of dementia (HR 1.66; 95% CI 1.13, diseases (difference: −0.34; 95% CI −1.54, −0.14), than those who 2.42) as compared to those with unspecific multimorbidity. Similarly, participated in at least two assessments. In addition, they were more participants with cardiovascular multimorbidity showed a 61% higher 772 GRANDE ET AL . TA B L E 2 Incidence rates (IRs) per 100 person-years and adjusted 95% CI 0.78,1.45 for the respiratory/metabolic/musculoskeletal pattern). hazard ratios (aHRs) of dementia with 95% confidence intervals (CIs) Similar findings have been obtained when we excluded those people by baseline multimorbidity patterns (N = 2478) who developed dementia during the follow-up also from the cluster Incident dementia analysis (data not shown). Multimorbidity patterns Cases IR (95% CI) aHR (95% CI) Consistent results were obtained when we included the multimorbidity-free individuals as reference group in the survival Unspecific (N = 1165) 208 1.9 (1.7-2.2) 1.00 (Ref.) analysis (Table S4). Neuropsychiatric (N = 183) 38 2.5 (1.8-3.5) 1.66 (1.13-2.42) Cardiovascular (N = 243) 65 4.7 (3.7-6.0) 1.61 (1.17-2.29) Sensory impairment/cancer 122 4.7 (4.0-5.7) 1.32 (1.10-1.71) 3.1.3 Role of CRP and APOE genotype (N = 384) Respiratory/metabolic/ 75 1.6 (1.7-2.2) 0.97 (0.73-1.29) No statistically significant interaction was detected between APOE musculoskeletal (N = 503) genotype, CRP levels, and multimorbidity patterns. Figure 1 reports stratified analyses by CRP level (Panel A) and APOE ε4 allele (Panel B). Cox models are adjusted for age, sex, education, civil status, malnutrition, C-reactive protein levels, APOE genotype, and baseline Mini-Mental State The association between the neuropsychiatric, cardiovascular, and sen- Examination score. sory impairment/cancer multimorbidity and dementia was even stronger in the group with high levels of CRP (HR 2.69; 95% CI 1.19, 6.09; HR 2.34; 95% CI 1.08, 5.09; HR 1.92; 95% CI 1.02, 3.63, respectively) hazard of dementia (HR 1.61; 95% CI 1.17, 2.29), as compared to (Figure 1). No statistically significant association was detected the reference group (ie, unspecific multimorbidity). Finally, the sensory between the respiratory/metabolic/musculoskeletal pattern and demen- impairment/cancer pattern was associated with a 32% increased haz- tia development in either group. ard of dementia compared with the unspecific cluster (HR 1.32; 95% CI In the analyses stratified by APOE genotype, we detected a stronger 1.10, 1.71). No statistically significant association was found between association between the neuropsychiatric (HR 2.79; 95% CI 1.43, the respiratory/metabolic/musculoskeletal pattern and dementia devel- 5.43) and cardiovascular (HR 2.58; 95% CI 1.39, 4.76) multimorbidity opment. Table 2 also shows the incidence rate (IR) of dementia per 100 patterns and dementia development in ε4 carriers with respect to person-years by multimorbidity pattern. non-carriers. No statistically significant association with dementia Consistent, but slightly attenuated, results were obtained when we development was detected for either the sensory impairment/cancer excluded participants who developed dementia within the first 6 years or the respiratory/metabolic/musculoskeletal pattern in APOE ε4 of follow-up (HR1.67; 95% CI 1.07, 2.62 for the neuropsychiatric pat- carriers. tern; HR 1.53; 95% CI 1.04, 2.26 for the cardiovascular pattern; HR 1.31; Further adjustments (for use of drugs with anticholinergic action, 95% CI 0.98, 1.78 for the sensory impairment/cancer pattern; HR 1.05; corticosteroids, and polypharmacy) did not modify the results, as well F I G U R E 1 Incidence rates (IRs) per 100 person-years and hazard ratios (HRs) of dementia with 95% confidence intervals (CIs) by baseline multimorbidity pattern, according to serum C-reactive protein (CRP) levels and apolipoprotein E (APOE) genotype. HRs are plotted on a logarithmic scale. P for interaction: CRP: Neuropsychiatric MM#CRP: P = 0.151. Cardiovascular MM#CRP: P = 0.147. Sensory impairment/cancer MM#CRP: P = 0.833. Respiratory/metabolic/MSK MM#CRP: P = 0.419. APOE: Neuropsychiatric MM#APOE: P = 0.227. Cardiovascular MM#APOE: P = 0.895. Sensory impairment/cancer MM#APOE: P = 0.241. Respiratory/metabolic/MSK MM#APOE: P = 0.983. Models are adjusted for: Panel A: age, sex, education, civil status, malnutrition, and baseline Mini-Mental State Examination score, and APOE ε4 allele. Panel B: age, sex, education, civil status, malnutrition, and baseline Mini-Mental State Examination score, and CRP levels. GRANDE ET AL . 773 as when we use age, instead of calendar time, as time scale (data not chiatric symptoms are common features of dementia, and can be part shown). of its clinical picture, even in the early stages. Still, longitudinal studies with long follow-ups have shown that depressive episodes are linked with incident dementia, reinforcing the idea of depression as a risk fac- 4 DISCUSSION tor for Alzheimer’s disease.30 Regardless of the direction of this asso- ciation, depression seems to lead to an imbalance in stress hormones, Older adults with neuropsychiatric and cardiovascular multimorbidity reduced hippocampal volume, and increased systemic inflammation— experienced the highest hazard of dementia throughout 12 years, in turn, further impairing cognitive function. followed by those with sensory impairment and cancer. Of interest, Of interest, in our sample, sensory impairment (eg, visual impair- even if the interaction was not statistically significant, high serum CRP ment and deafness) and cancer clustered together and were associ- levels seemed to amplify this risk in all three multimorbidity patterns, ated with a 30% increased hazard of dementia. Previous studies have whereas the presence of APOE ε4 appeared to increase dementia risk independently investigated both sensory impairments and cancer with only in the neuropsychiatric and cardiovascular patterns. incident dementia, finding conflicting results.31–34 Among others, hear- Although the role of specific chronic diseases as risk factors ing loss has been linked with an increased dementia risk,35 but such for dementia has been largely acknowledged,4,5 studies addressing an association calls for further clarification, as to whether it can be dementia risk in relation to multimorbidity are limited to few reports. considered a risk factor, or rather a prodromal sign of the disease. Multimorbidity has been associated with different degrees and sever- Greater debate exists surrounding the cognitive impact of cancer, with ity of cognitive dysfunction, such as steeper cognitive decline,20,21 some studies pointing toward an inverse association with dementia33 increased risk of mild cognitive impairment7 and dementia,22 and and others finding higher dementia risk in persons with neoplasms.36 faster dementia progression.23 However, the majority of these studies Previous studies have reported greater cancer risk in people with have operationalized multimorbidity in quantitative terms through dis- visual impairment37 and, likewise, that age-related macular degener- ease count, which is regarded by clinicians as uninformative.24 Indeed, ation occurs more frequently among smokers, which in turn increases given a group of individuals with the same number of chronic dis- the risk of different cancer types.38,39 Furthermore, many cancer treat- eases, understanding of how and which diseases cluster together and ment regimens combine multiple ototoxic medications.40 Finally, we increase dementia risk compared to others may help to gain insight into can hypothesize that these two conditions might share common risk the pathophysiology of dementia. factors and pathophysiological mechanisms, and that they tend to co- To date, the evidence relating multimorbidity patterns and dementia occur as an expression of independent biological degenerative changes is extremely scarce,25,26 and to the best of our knowledge, no studies occurring with age. This is also witnessed by the fact that, in our study, have been conducted between multimorbidity patterns and incident the cluster sensory impairment/cancer includes the oldest, more likely dementia. An observational study carried out in The Netherlands eval- to be women, malnourished and with lower levels of cognitive perfor- uated the effects of groups of diseases on cognitive performance in a mance. This profile identifies a frail proportion of our study popula- healthy population,25 finding that heart diseases and cancer were asso- tion, and a number of studies have recently pointed out that cognitive ciated with lower verbal memory performance, whereas cerebrovas- impairment and frailty frequently co-occur and that frail older adults cular diseases, heart diseases, cancer, and movement disorders were present a higher risk of dementia.41 Notably, the risk of dementia asso- related to worse processing speed. ciated with this pattern was even higher in the presence of high CRP Multimorbidity has been found previously to be associated with levels, but was not modified by the APOE ε4 allele. As such, dementia biomarkers of neurodegeneration and amyloid deposition8 ; however, onset in this group of participants might be linked to systemic health specific multimorbidity patterns may increase dementia risk through deterioration, as a consequence of underlying complex diseases or drug different pathways. The detrimental role of cardiovascular diseases regimens. on the brain is well established, being people with heart and cere- Of interest, we were not able to detect any association between the brovascular diseases at high risk of developing dementia and with a respiratory/metabolic/musculoskeletal cluster and dementia, although poor prognosis.4,27 Indeed, accumulating evidence points toward a sleep disorders and diabetes (two known risk factors for dementia)42,43 close interplay between body and mind in dementia development,2,28 are part of this cluster. However, in the current study, we are not inves- and our study aligns with this literature. It is notable that we also add tigating the effect of individual diseases on dementia risk but the risk weight to such knowledge by reporting that multiple, co-occurring conveyed by belonging to a cluster of individuals displaying a given cardiovascular diseases are linked to dementia. The existence of a disease combination. Moreover, the comparison with the unspecific heart-brain connection is supported by the fact that those with cardio- cluster—still composed by individuals with multimorbidity—may have vascular diseases tend to be at higher dementia risk due to a number of diluted the effect. different mechanisms, including but not limited to cerebral hypoperfu- A similar increased dementia risk was found in individuals with sion and infarcts, cardio-embolization, and chronic inflammation.29 neuropsychiatric and cardiovascular multimorbidity and high levels of In our study, we also found increased dementia risk among persons systemic inflammation. Although these results should be interpreted with neuropsychiatric diseases, which can be interpreted both as a risk cautiously due to a lack of statistically significant interaction and factor or a prodromal manifestation of dementia. Notably, neuropsy- therefore wide, overlapping confidence intervals in the stratified 774 GRANDE ET AL . analyses, they indicate that inflammation might play a role in dementia acterized by a high proportion of cardiovascular risk factors such as development in those with neuropsychiatric and cardiovascular chronic hypertension and dyslipidemia. In addition, early dementia diagnosis conditions. Despite that the link between chronic conditions, inflam- could be achieved by the close monitoring of persons affected by mation, and dementia is yet far from being entirely understood, several neuropsychiatric multimorbidity. Finally, the demographic and clinical hypotheses can be made. It is recognized that a low-grade chronic characteristics of those belonging to the sensory impairment/cancer proinflammatory state characterized by high levels of serum cytokines pattern calls for a close follow-up of frail individuals. The identification is very frequent in older individuals44 and that older persons with high of those patterns of multimorbidity in the presence of inflammation inflammatory markers have a higher number of chronic diseases as and among APOE ε4 carriers might help tailor clinical interventions well as a steeper increase in multimorbidity over time.45 At the same aimed at dementia prevention, even to a greater extent. time, it is under debate if neuroinflammation is a driving force in the development of dementia46 or an epiphenomenon of brain dysfunction and neuronal loss. Furthermore, the imbalance between inflammatory 6 STRENGTHS AND LIMITATIONS and anti-inflammatory agents (a frequent finding in older individuals, termed ‘‘inflammaging’’)47 due to several chronic diseases can lead To our knowledge, this is the first prospective observational study to a systemic and chronic proinflammatory state, which ultimately assessing dementia risk across different patterns of multimorbidity. may affect the brain. The exact chain of events linking chronic dis- These findings are derived from a large, well-characterized population- eases/inflammation/dementia remains a matter of debate, but based based cohort, with a long follow-up and an extensive individual-level on the findings from our study we can speculate that the presence of clinical evaluation of diseases, including dementia. Of note, we were inflammation further increases the risk of dementia within specific able to identify dementia cases, even among those individuals who multimorbidity patterns. The possible role as mediator of inflammation died during follow-up, by performing a comprehensive review of clin- has not been tested in the present study, since multimorbidity and CRP ical records and death registers. This comprehensive identification of were assessed at the same time point, thus preventing us from having cases minimized the competing influence of death on dementia inci- a proper chain of events as required in mediation analyses. dence. Notwithstanding, we acknowledge that some mild cases might Because APOE ε4 is the strongest genetic risk factor for dementia, not have been captured, resulting in a possible underestimation of the it is unsurprising that we found stronger associations with dementia effect. Finally, even if the presence of residual confounding cannot be among APOE ε4 carriers. It is notable that APOE has been linked to ruled out completely, our findings appeared robust to various sensitiv- accelerated atherosclerosis and has been implicated in the modulation ity analyses and accounted for several relevant confounders. of proinflammatory and anti-inflammatory cytokines.48,49 It is worth Some limitations must be mentioned. First, the stratified analyses noting that despite a lack of statistically significant interaction, we should be interpreted with caution due to the non-statistically sig- detected such an increased risk in the cardiovascular and neuropsychi- nificant interaction, which is also reflected in the wide and overlap- atric multimorbidity patterns, but not in the sensory impairment/cancer ping confidence intervals. Notwithstanding, effect modification can be pattern. Again, it is plausible to hypothesize that dementia develop- present also in the absence of a statistical interaction,51 as witnessed ment differs when related to the presence of sensory impairment/cancer by our findings that indicate that an interplay might exist and thus versus cardiovascular and neuropsychiatric conditions, and might even that inflammation and genetic predisposition might play a role. Sec- follow different pathways. This hypothesis is in agreement with the ond, despite the 12-year follow-up period, we cannot completely rule greater dementia risk observed in the participants who had high sys- out the possibility of subclinical dementia. This has prevented us from temic inflammation. making a conclusion on whether the neuropsychiatric multimorbidity pattern is more likely to be an at-risk condition or a clinical manifesta- tion of the disease. However, the results from the sensitivity analysis, 5 CLINICAL AND PUBLIC HEALTH excluding participants who developed dementia 6 years after baseline, IMPLICATIONS were only slightly attenuated. Third, SNAC-K is a population-based study where the diagnosis of dementia is clinical, lacking biological The timely identification of neuropsychiatric, cardiovascular, and sensory markers for the characterization of the subtypes. Thus future studies impairment/cancer multimorbidity in older adults might have important with biological data are needed to confirm our results and move a step implications in terms of prevention and clinical practice. forward in the understanding of the mechanisms behind the associa- Our study adds weight to the evidence that cardiovascular diseases tions we observed. Fourth, serum CRP was measured in a single time play a role in the development of adverse health-related outcomes not point, which does not capture chronic inflammation. Further evalua- only individually, but also as co-occurring entities, and confirms the tions on persistent levels of inflammation during follow-up time are importance of assessing them holistically.50 Dementia prevention may needed to confirm our findings. In addition, more fine markers of sys- be optimized by controlling cardiovascular risk factors and enhancing temic inflammation exist (including tumor necrosis factor α [TNFα] and the treatment of concurrent cardiovascular diseases (eg, improving some interleukins), which were not available in the data set. Future medication adherence). It is worth mentioning that half of the study studies are needed to better evaluate the role of inflammation in the participants were part of the unspecific multimorbidity cluster, char- studied association using more sensitive measures. Finally, being that GRANDE ET AL . 775 our study was observational, we cannot draw any causal relationship 4. Shang Y, Fratiglioni L, Marseglia A, et al. Association of diabetes with between exposures and outcome. stroke and post-stroke dementia: a population-based cohort study. Alzheimers Dement. 2020. 5. Ding M, Fratiglioni L, Johnell K, et al. Atrial fibrillation, antithrombotic treatment, and cognitive aging: a population-based study. Neurology. 7 CONCLUSIONS 2018;91:e1732-e40. 6. Calderon-Larranaga A, Vetrano DL, Onder G, et al. Assessing and mea- suring chronic multimorbidity in the older population: a proposal for its Older adults with neuropsychiatric, cardiovascular, and sensory operationalization. J Gerontol A Biol Sci Med Sci. 2017;72:1417-1423. impairment/cancer multimorbidity are at high risk of dementia; this 7. Vassilaki M, Aakre JA, Cha RH, et al. Multimorbidity and risk of mild risk is further elevated by concurrent inflammation and genetic pre- cognitive impairment. J Am Geriatr Soc. 2015;63:1783-1790. disposition. Future studies, specifically designed to evaluate the inter- 8. Vassilaki M, Aakre JA, Mielke MM, et al. Multimorbidity and neu- roimaging biomarkers among cognitively normal persons. Neurology. play between aging-related inflammatory processes, genetic predispo- 2016;86:2077-2084. sition, multimorbidity patterns, and dementia are needed to explore 9. Calderon-Larranaga A, Vetrano DL, Ferrucci L, et al. Multimorbidity such mechanisms. Identifying older adults affected by specific mul- and functional impairment-bidirectional interplay, synergistic effects timorbidity patterns might be helpful in tailoring interventions for and common pathways. J Intern Med. 2019;285:255-271. dementia prevention and targeting individuals who would benefit most 10. Marengoni A, Roso-Llorach A, Vetrano DL, et al. Patterns of multimor- bidity in a population-based cohort of older people: sociodemographic, from such interventions. lifestyle, clinical, and functional differences. J Gerontol A Biol Sci Med Sci. 2020;75:798-805. ACKNOWLEDGEMENTS 11. Prados-Torres A, Calderon-Larranaga A, Hancco-Saavedra J, Poblador- The authors are grateful to all the participants in the SNAC-K project as Plou B, van den Akker M. Multimorbidity patterns: a systematic review. J Clin Epidemiol. 2014;67:254-266. well as staff in the SNAC-K Study Group for their collaboration in data 12. Sankowski R, Mader S, Valdes-Ferrer SI. Systemic inflammation and collection and management. This work was supported by the funders the brain: novel roles of genetic, molecular, and environmental cues as of the Swedish National study on Aging and Care (SNAC): the Ministry drivers of neurodegeneration. Front Cell Neurosci. 2015;9:28. of Health and Social Affairs, Sweden; the participating county councils 13. Franceschi C, Campisi J. Chronic inflammation (inflammaging) and its potential contribution to age-associated diseases. J Gerontol A Biol Sci and municipalities and the Swedish Research Council. Specific grants Med Sci. 2014;69(Suppl 1):S4-9. were obtained from the Swedish Research Council (2016-00981) and 14. Lagergren M, Fratiglioni L, Hallberg IR, et al. A longitudinal study inte- the Swedish Research Council for Health, Working Life and Welfare grating population, care and social services data. The Swedish National (2017-01764). The Italian Ministry of Health (PE-2016-02364885). study on Aging and Care (SNAC). Aging Clin Exp Res. 2004;16:158-168. 15. Cederholm T, Bosaeus I, Barazzoni R, et al. Diagnostic criteria for mal- The financial sponsors played no role in the design, execution, analysis nutrition - an ESPEN Consensus Statement. Clin Nutr. 2015;34:335- and interpretation of data, or writing of the study 340. 16. Kravitz BA CM, Kawas CH. Elevated C-reactive protein levels are CONFLICT OF INTEREST associated with prevalent dementia in the oldest-old. Alzheimers Dement. 2009;5(4):318-323. https://doi.org/10.1016/j.jalz.2009.04. No conflict of interest to declare. 1230. PMID: 19560102; PMCID: PMC2740472. 17. Guisado-Clavero M, Roso-Llorach A, Lopez-Jimenez T, et al. Multimor- AUTHORS CONTRIBUTION bidity patterns in the elderly: a prospective cohort study with cluster AM, GG, DLV, LF, and ACL contributed to the conception and design analysis. BMC Geriatr. 2018;18:16. of the study. GG, ARL, and ACL conducted the statistical analyses. GG 18. Violan C, Roso-Llorach A, Foguet-Boreu Q, et al. Multimorbidity pat- terns with K-means nonhierarchical cluster analysis. BMC Fam Pract. and AM conducted the literature search. All the authors contributed 2018;19:108. to interpretation of the results. GG and AM drafted the first version 19. Vetrano DL, Roso-Llorach A, Fernandez S, et al. Twelve-year clinical of the manuscript. All the authors critically revised the manuscript for trajectories of multimorbidity in a population of older adults. Nat Com- important intellectual content. All the authors made a significant con- mun. 2020;11:3223. 20. Wei MY, Levine DA, Zahodne LB, Kabeto MU, Langa KM. Multimorbid- tribution to the research and the development of the manuscript and ity and Cognitive decline over 14 years in older Americans. J Gerontol approved the final version for publication. SNAC-K personnel collected A Biol Sci Med Sci. 2020;75:1206-1213. the data for the study. 21. Fabbri E, An Y, Zoli M, et al. Association between accelerated multi- morbidity and age-related cognitive decline in older baltimore longi- tudinal study of aging participants without dementia. J Am Geriatr Soc. REFERENCES 2016;64:965-972. 1. Shah H, Albanese E, Duggan C, et al. Research priorities to reduce 22. Bauer K, Schwarzkopf L, Graessel E, Holle R. A claims data-based com- the global burden of dementia by 2025. Lancet Neurol. 2016;15:1285- parison of comorbidity in individuals with and without dementia. BMC 1294. Geriatr. 2014;14:10. 2. Grande G, Qiu C, Fratiglioni L. Prevention of dementia in an ageing 23. Haaksma ML, Rizzuto D, Leoutsakos JS, et al. Predicting cognitive world: evidence and biological rationale. Ageing Res Rev. 2020:101045. and functional trajectories in people with late-onset dementia: 2 3. Winblad B, Amouyel P, Andrieu S, et al. Defeating Alzheimer’s dis- population-based studies. J Am Med Dir Assoc. 2019;20:1444-1450. ease and other dementias: a priority for European science and society. 24. Whitty CJM, MacEwen C, Goddard A, et al. Rising to the challenge of Lancet Neurol. 2016;15:455-532. multimorbidity. BMJ. 2020;368:l6964. 776 GRANDE ET AL . 25. Aarts S, van den Akker M, Tan FE, Verhey FR, Metsemakers JF, van Box- 41. Grande G, Haaksma ML, Rizzuto D, et al. Co-occurrence of cognitive tel MP. Influence of multimorbidity on cognition in a normal aging pop- impairment and physical frailty, and incidence of dementia: systematic ulation: a 12-year follow-up in the Maastricht Aging Study. Int J Geriatr review and meta-analysis. Neurosci Biobehav Rev. 2019;107:96-103. Psychiatry. 2011;26:1046-1053. 42. Sindi S, Kareholt I, Johansson L, et al. Sleep disturbances and dementia 26. Wang JH, Wu YJ, Tee BL, Lo RY. Medical comorbidity in Alzheimer’s risk: a multicenter study. Alzheimers Dement. 2018;14:1235-1242. Disease: a nested case-control study. J Alzheimers Dis. 2018;63:773- 43. Marseglia A, Darin-Mattsson A, Kalpouzos G, et al. Can active life mit- 781. igate the impact of diabetes on dementia and brain aging?. Alzheimers 27. Qiu C, Fratiglioni L. A major role for cardiovascular burden in age- Dement. 2020. related cognitive decline. Nat Rev Cardiol. 2015;12:267-277. 44. Ferrucci L, Fabbri E. Inflammageing: chronic inflammation in ageing, 28. Grande G, Rizzuto D, Vetrano DL, et al. Cognitive and physical markers cardiovascular disease, and frailty. Nat Rev Cardiol. 2018;15:505-522. of prodromal dementia: a 12-year-long population study. Alzheimers 45. Fabbri E, An Y, Zoli M, et al. Aging and the burden of multimorbidity: Dement. 2020;16:153-161. associations with inflammatory and anabolic hormonal biomarkers. J 29. Vicario A, Cerezo GH. At the heart of brain disorders - preventing cog- Gerontol A Biol Sci Med Sci. 2015;70:63-70. nitive decline and dementia. Eur Cardiol. 2015;10:60-63. 46. Ardura-Fabregat A, Boddeke E, Boza-Serrano A, et al. Target- 30. Singh-Manoux A, Dugravot A, Fournier A, et al. Trajectories of depres- ing neuroinflammation to treat Alzheimer’s disease. CNS Drugs. sive symptoms before diagnosis of dementia: a 28-year follow-up 2017;31:1057-1082. study. JAMA Psychiatry. 2017;74:712-718. 47. Furman D, Campisi J, Verdin E, et al. Chronic inflammation in the etiol- 31. Livingston G, Sommerlad A, Orgeta V, et al. Dementia prevention, ogy of disease across the life span. Nat Med. 2019;25:1822-1832. intervention, and care. Lancet. 2017;390:2673-2734. 48. Zhang H, Wu LM, Wu J. Cross-talk between apolipoprotein E and 32. Driver JA, Beiser A, Au R, et al. Inverse association between cancer and cytokines. Mediators Inflamm. 2011;2011:949072. Alzheimer’s disease: results from the Framingham Heart Study. BMJ. 49. Attems J, Jellinger KA. The overlap between vascular disease and 2012;344:e1442. Alzheimer’s disease–lessons from pathology. BMC Med. 2014;12:206. 33. Musicco M, Adorni F, Di Santo S, et al. Inverse occurrence of cancer 50. Vetrano DL, Rizzuto D, Calderon-Larranaga A, et al. Trajectories and Alzheimer disease: a population-based incidence study. Neurology. of functional decline in older adults with neuropsychiatric and 2013;81:322-328. cardiovascular multimorbidity: a Swedish cohort study. PLoS Med. 34. van der Willik KD, Schagen SB, Ikram MA. Cancer and dementia: two 2018;15:e1002503. sides of the same coin?. Eur J Clin Invest. 2018;48:e13019. 51. VanderWeele TJ. On the distinction between interaction and effect 35. Lin FR, Metter EJ, O’Brien RJ, Resnick SM, Zonderman AB, Ferrucci L. modification. Epidemiology. 2009;20:863-871. Hearing loss and incident dementia. Arch Neurol. 2011;68:214-220. 36. Ganguli M. Cancer and dementia: it’s complicated. Alzheimer Dis Assoc Disord. 2015;29:177-182. 37. Pukkala E, Verkasalo PK, Ojamo M, Rudanko SL. Visual impairment and SUPPORTING INFORMATION cancer: a population-based cohort study in Finland. Cancer Causes Con- Additional supporting information may be found online in the Support- trol. 1999;10:13-20. ing Information section at the end of the article. 38. Mitchell P, Liew G, Gopinath B, Wong TY. Age-related macular degen- eration. Lancet. 2018;392:1147-1159. 39. Sasco AJ, Secretan MB, Straif K. Tobacco smoking and cancer: a brief review of recent epidemiological evidence. Lung Cancer. How to cite this article: Grande G, Marengoni A, Vetrano DL, 2004;45(Suppl 2):S3-9. 40. Frisina RD, Wheeler HE, Fossa SD, et al. Comprehensive audio- et al. Multimorbidity burden and dementia risk in older adults: metric analysis of hearing impairment and tinnitus after cisplatin- The role of inflammation and genetics. Alzheimer’s Dement. based chemotherapy in survivors of adult-onset cancer. J Clin Oncol. 2021;17:768–776. https://doi.org/10.1002/alz.12237 2016;34:2712-2720.