Background
It is estimated that around 50 million people are diagnosed with dementia worldwide [
1]. The development of dementia is influenced by genetic, lifestyle, and medical risk factors [
2]. So far, dementia can be neither cured nor reversed by available pharmacological interventions. However, a recent report from the Lancet Commission declared that 40% of dementia could be prevented or delayed if some major risk factors would be modified, including dietary factors [
3]. Emerging evidence showed that diet quality and dietary nutrients appear to play important roles in preventing cognitive decline [
4,
5]. For instance, dietary intakes of vitamin D, omega-3, and omega-6 fatty acids were found to be inversely associated with cognitive decline in the elderly [
6,
7].
As a common nutritional supplement, glucosamine is widely recommended to use for managing osteoarthritis or joint pain in European [
8,
9]. Meanwhile, it is regarded as one of the popular nutraceuticals in the USA and Australia [
10,
11]. In vitro and in vivo experiments have revealed that glucosamine exerts beneficial effects by suppressing inflammation and antioxidant activities [
12]. Brain and neural cells are highly sensitive to oxidative stress, which may cause profound cell damages and provoke neurodegenerative disorders [
13]. Nevertheless, epidemiological evidence about glucosamine use and dementia is unclear to date. Previous longitudinal studies suggested that glucosamine use was associated with lower risk of type 2 diabetes (T2D) [
14]. We hypothesized that T2D may play an essential role in the association between habitual glucosamine use and dementia. In this large-scale cohort study in UK Biobank, we aimed to investigate the association between habitual glucosamine use and dementia and explore the roles of T2D in the association.
Methods
Study population
UK Biobank is a large-scale biomedical study designed to contribute to modern medicine and improve public health [
15,
16]. Participants aged 37 to 73 years were recruited across 22 assessment centers in the UK during 2006 to 2010. Details of its study design have been reported previously [
17]. Participants were invited to provide information on lifestyle and medical history, take part in physical measurements, and provide biological samples regularly in the assessment centers at baseline. At the same time, health-relevant records of them would be linked to the UK National Health Service System throughout the study. The North West Multi-Centre Research Ethics Committee (reference 11/NW/0382) had approved the UK Biobank as a Research Tissue Bank.
We excluded participants (1) with prevalent dementia at baseline, (2) who did not complete the self-report assessments of habitual glucosamine use, and (3) who requested to be removed from the UK Biobank dataset. Our final analysis samples included 495,942 participants for disease outcomes. Flowchart of participant enrolment can be seen in Supplementary Fig.
1.
Assessment of glucosamine use
At the assessment centers, participants were asked “Do you regularly take any of the following?” through a touchscreen questionnaire. Of all options, six dietary supplements could be selected, including glucosamine use. They can choose more than one answer in the questionnaire. We defined the use of glucosamine as: “0 = no” and “1 = yes”.
Ascertainment of outcomes
The primary outcome in the current study was the incidence of dementia, and the secondary outcomes included the incidence of Alzheimer’s disease (AD) and vascular dementia (VD). Record linkage containing admissions and diagnoses information was linked to the Hospital Episode Statistics, Scottish Morbidity Record data, and the Patient Episode Database. Dementia events during the follow-up were ascertained from hospital inpatient records according to the International Classification of Disease version 10 (ICD-10) codes: F00-F03, and G30-G31 (Supplementary Table
1) [
18]. Follow-up of this study was censored at the date of incident dementia, death, or the end of follow-up (September 30, 2020), whichever occurred first.
Covariates
We collected the potential confounders at baseline, including age, sex (female, male), ethnicity (White, Black, south Asian, and mixed background), socioeconomic status (Townsend deprivation index), education attainment (college or university degree, professional qualifications, and others), body mass index (BMI, calculated as weight divided by the square of height, kg/m2), self-reported smoking status (never, previous, current), alcohol consumption (g/day), total cholesterol (mmol/L), family history of dementia (no or yes), aspirin use (no or yes), and mineral and vitamin supplements use (vitamin A, B, C, D, E; multivitamin; or folic acid).
The Townsend deprivation index was assigned as a continuous measure on the basis of postal codes, which were derived from census data on housing, employment, social class, and car availability; a higher index indicated more deprivation [
19]. A healthy diet score was also summarized by using the following food categories: ≥ 4.5 servings total fruit and vegetable intake consumption per week, ≥ 2 fish intake per week, ≤ 2 times intake of processed meat per week, and ≤ 5 times red meat intake per week [
20]. A healthy diet was ascertained if an individual met at least two healthy food items, as described in the previous study. Prevalent hypertension was defined as systolic blood pressure ≥ 140 mm Hg, or diastolic blood pressure ≥ 90 mm Hg, or reported use of antihypertensive drugs. In the main analyses, missing information on covariables was coded as a missing indicator category for categorical variables such as smoking status and with mean values for continuous variables. Detailed information of the missing variables can be seen in Table
1.
Table 1
Baseline characteristic of participants stratified by glucosamine use
Age (years), mean (SD) | 56.54 (8.09) | 55.94 (8.20) | 59.08 (7.07) | <0.001 |
Female | 270,044 (54.45) | 210,902 (52.54) | 59,142 (62.59) | <0.001 |
Ethnicity | | | | <0.001 |
White | 467,865 (94.34) | 377,256 (93.97) | 90,609 (95.88) | |
Black | 7906 (1.59) | 6866 (1.71) | 1040 (1.10) | |
South Asian | 11,123 (2.24) | 9627 (2.40) | 1496 (1.58) | |
Mixed background | 7344 (1.48) | 6275 (1.56) | 1069 (1.913) | |
Missing | 1704 (0.34) | 1420 (0.35) | 284 (0.30) | |
Townsend deprivation index | | | | <0.001 |
1 (Least deprived) | 99,244 (20.01) | 77,476 (19.30) | 21,768 (23.04) | |
2 | 99,620 (20.09) | 78,450 (19.54) | 21,170 (22.40) | |
3 | 99,304 (20.02) | 79,407b (19.78) | 19,897 (21.06) | |
4 | 98,711 (19.90) | 80,915 (20.16) | 17,796 (18.83) | |
5 (Most deprived) | 99,063 (19.97) | 85,196 (21.22) | 13,867 (14.67) | |
Missing | 618 (0.12) | 538 (0.13) | 80 (0.08) | |
Education attainment | | | | <0.001 |
College or university degree | 160,792 (32.42) | 129,576 (32.28) | 31,216 (33.03) | |
Professional qualifications | 245,272 (49.46) | 197,114 (49.10) | 48,158 (50.96) | |
Others | 84,679 (17.07) | 70,392 (17.53) | 14,287 (15.12) | |
Missing | 5199 (1.05) | 4362 (1.09) | 837 (0.89) | |
Body mass index (kg/m2), mean (SD) | 27.43 (4.79) | 27.45 (4.82) | 27.36 (4.64) | <0.001 |
Smoking status | | | | <0.001 |
Never | 270,649 (54.57) | 218,693 (54.48) | 51,956 (54.98) | |
Previous | 171,298 (34.54) | 135,197 (33.68) | 36,101 (38.20) | |
Current | 52,132 (10.51) | 46,012 (11.46) | 6120 (6.48) | |
Missing | 1863 (0.38) | 1542 (0.38) | 321 (0.34) | |
Alcohol consumption (g/day), mean (SD) | 14.54 (18.25) | 14.66 (18.66) | 14.05 (16.37) | <0.001 |
Healthy dieta | | | | <0.001 |
Yes | 249,104 (50.23) | 192,905 (48.05) | 56,199 (59.47) | |
No | 242,069 (48.81) | 204,286 (50.89) | 37,783 (39.98) | |
Missing | 4769 (0.96) | 4253 (1.06) | 516 (0.55) | |
APOE ε4 carrier | 137,781 (27.78) | 111,708 (27.83) | 26,073 (27.59) | <0.001 |
Carrier | 137,781 (27.78) | 111,708 (27.83) | 26,073 (27.59) | |
Non-carrier | 346,783 (69.92) | 280,342 (69.83) | 66,441 (70.31) | |
Missing | 11,378 (2.29) | 9394 (2.34) | 1984 (2.10) | <0.001 |
Family history of dementia | 65,161 (13.14) | 50,620 (12.61) | 14,541 (15.39) | <0.001 |
Total cholesterol (mean, mmol/L), mean (SD) | 5.69 (1.11) | 5.67 (1.11) | 5.81 (1.09) | <0.001 |
Mineral and vitamin supplements use | 157,450 (31.75) | 104,927 (26.14) | 52,523 (55.58) | <0.001 |
Aspirin use | 69,318 (13.698) | 56,018 (13.95) | 13,300 (14.07) | 0.338 |
Hypertension | 345,726 (69.71) | 278,411 (69.35) | 67,315 (71.23) | <0.001 |
Arthritis | 52,214 (10.53) | 32,973 (8.21) | 19,241 (20.36) | <0.001 |
Statistical analyses
We used the mean and standard deviation (SD) to describe continuous variables with normal distribution, and numbers (percentages) to describe categorical variables. Baseline characteristics were compared based on whether participants used or did not use glucosamine. We used the χ2 test for categorical variables, or analysis of variance or Mann-Whitney U test for continuous variables, as appropriate.
Cox proportional hazards models with age as timescale were used to calculate hazard ratios (HRs) and 95% confidence intervals (CIs) of dementia. The proportional hazards assumptions were tested based on Schoenfeld residuals [
21]. Participants who reported themselves as glucosamine non-users were considered as the reference group. Three models were estimated in the following analyses. Model 1 was adjusted for age (timescale) and sex. Model 2 was additionally adjusted for ethnicity, BMI, education attainment, Townsend deprivation index, smoking status, alcohol consumption, family history of dementia, hypertension, arthritis, total cholesterol, healthy diet, and
APOE genotype. Model 3 was finally adjusted for ethnicity, BMI, education attainment, Townsend deprivation index, smoking status, alcohol consumption, family history of dementia, hypertension, arthritis, total cholesterol,
APOE genotype, healthy diet, use of mineral and vitamin supplements, and use of aspirin.
We assessed the additive- and multiplicative-scale interaction measures to examine the interaction effect of glucosamine use and
APOE ε4 genotype on the risk of dementia onset. We calculated the attributable proportion due to interaction (AP), relative excess risk due to interaction (RERI), and synergy index (SI) to derive the additive interaction. Above mentioned indices were used to assess whether the risk due to having both exposures is greater than the sum of the risks due to each condition. The multiplicative-scale interaction has been widely used to examine the interaction effects by identifying whether the risk due to having both exposures is greater than the product of the risks due to each exposure alone [
22].
In subsequent analyses, Markov multi-state models were used to explore the role of T2D over the follow-up in the association between glucosamine use and risk of dementia. The Markov multi-state models were semi-parametric models which could facilitate data preparation and flexible estimation of different types of covariate effects in Cox regression models [
23]. It is a useful way of describing a process in which an individual moves through a series of states in continuous time. These models allow simultaneous estimation of the association of glucosamine use with risk of incident T2D over the follow-up (Transition 1), the association of glucosamine use with risk of dementia without incident T2D over the follow-up (Transition 2), and the association of glucosamine use with risk of dementia following diagnosis of T2D (Transition 3). The models were adjusted for the covariates mentioned above.
Several sensitivity analyses were finally performed to confirm the robustness of our results. First, those who developed dementia events within 2 years of follow-up were excluded to reduce the possibility of spurious association due to reverse causation. Second, we conducted the complete-case analyses and multiple imputation using chained equations with 5 imputations. Third, the competing risk of death on the association between glucosamine use and dementia was investigated using flexible parametric competing risk model [
24]. Finally, we also calculated the
E-value to evaluate the risk of unmeasured confounding. The
E-value estimates the minimum strength that an unmeasured confounding variable would have to have to nullify the observed association between glucosamine use and dementia while considering all other measured covariates [
25].
We used Stata to conduct all the statistical analyses (version 15, StataCorp). The statistical significance was set as P < 0.05 (two-sided test). Bonferroni correction (significance level 0.05/16 = 0.003) was conservatively corrected when we tested the effect modifications.
Discussion
In this large prospective cohort study of 495,942 UK adults, we found that glucosamine use was associated with a 10% lower risk of dementia, there were significant additive and multiplicative interactions between glucosamine use and APOE ε4 genotype. Moreover, the association was mediated by T2D, suggesting that the protective benefits of glucosamine use on risk of dementia might be partly explained by preventing or delaying the incidence of T2D over the follow-up.
Our current study found a decreased risk of dementia in glucosamine users from a perspective of epidemiology. The underlying mechanism of the association is poorly understood. It has been found that glucosamine is associated with decreased inflammation in cell culture studies [
26]. Existing animal studies have also reported that the anti-inflammatory properties of glucosamine may play a preventive role in the pathophysiology of adverse health events [
27‐
29]. A previous animal study revealed that glucosamine could mimic a low carbohydrate diet, which were characterized by decreasing glycolysis and improved amino acid catabolism [
30]. This may also partly explain the anti-inflammatory effect of glucosamine because low carbohydrate diets were significantly associated with a lower risk of dementia as reported from a randomized clinical trial [
31]. Other biological plausibility for the potential protective effect of glucosamine on dementia included its antioxidant activities in brain tissue. Glucosamine was involved into antioxidant activities by scavenging the superoxide and hydroxyl radicals and protecting the macromolecules. As oxidative stress had been consistently determined to link to increased dementia risk [
32], the antioxidant properties of glucosamine may thus help with interpreting its potential anti-dementia mechanism. In a preclinical test, glucosamine can penetrate the blood-brain barrier and have a positive impact on the ability to complete cognitive tasks [
33]. Glucosamine was also shown to improve spatial learning and memory in rats [
34]. In small mammals and invertebrates, glucosamine promotes the generation of mitochondria, thereby providing energy for cells [
30].
We found that the association between glucosamine use and dementia differed by
APOE genotype. Stronger association between glucosamine use and decreased risk of dementia was observed in participants without
APOE ε4 carriers. One of the possible explanations is that the
ε4 allele of the apolipoprotein E gene is the strongest genetic risk factor for late-onset AD [
35], and it has been identified as a risk factor for other dementia sub-type [
36]. Moreover, the evidence from the Rotterdam study suggested that a healthy lifestyle can attenuate the risk of dementia [
37]. A UK Biobank study found that a favorable lifestyle was associated with a lower risk of dementia, even in those who are at high genetic risk [
38]. In our study, it is of great significance to find that glucosamine users were associated with a lower risk of dementia, regardless of adherence to a healthy lifestyle (such as a healthy diet, no current smoking, and normal BMI) or not. Moreover, previous studies have reported that glucosamine was often used in conjunction with other supplements [
8], and our study consistently showed that glucosamine users tended to take vitamin and mineral supplements in combination than non-users. Similarly, we found that glucosamine use was associated with a lower risk of dementia, independent of mineral and vitamin supplement use.
We also observed that glucosamine user had a lower risk of subsequent dementia among those with developing T2D during the follow-up period. This means that the protective benefits of glucosamine use on risk of dementia might be partly explained by preventing or delaying the incidence of T2D over the follow-up. It has been suggested that T2D is an independent risk factor of all-cause dementia, including AD and VD [
39,
40]. Prior longitudinal studies have ascertained the protective effects of glucosamine use on risk of incident T2D [
12]. Suggested pathways that underpin the links between T2D and dementia may include systemic insulin resistance, and increased levels of circulating pro-inflammatory markers, which lead to defects in the insulin signaling pathway and changes in brain synaptic plasticity [
41].
In analyses of secondary outcomes, we found a reverse association between glucosamine use and VD rather than AD, which may result from small sample size in the dementia subtype, and different mechanisms of dementia. The main pathophysiology of AD is beta-amyloid peptide deposition that leads to a cascade of neuronal cell apoptosis [
42,
43]. However, VD occurs mainly through ischemia of the brain parenchyma caused by atherosclerotic disease. It is assumed that glucosamine use may exert on a beneficial effect against VD by influencing cardiovascular health. Future studies are warranted to elucidate the precise pathophysiological pathways from glucosamine to AD and VD.
To the best of our knowledge, the association between glucosamine use and dementia in such a large sample has not been reported. Besides large sample size, our study has the advantage of prospective design, longer follow-up time, and ability to adjust for potential confounders. Several sensitivity analyses also supported the robustness of our findings. Nevertheless, there are some unavoidable limitations of this study. First, the results we found were from an observational cohort study rather than a clinical trial, so conclusions regarding causality cannot be made. Second, dementia diagnoses in this study were obtained from electronic health records; the numbers of AD and VD are relatively small compared to the number of dementia cases, since the presence of large amount of other and unspecified dementia cases. We will have inevitably omitted undiagnosed conditions and less severe dementia cases that are largely dealt with in electronic health records. However, validation studies have suggested that these are reliable for ascertaining dementia, with a positive predictive value of 84.5% in UK Biobank when compared with expert clinical adjudication [
44,
45]. Third, habitual glucosamine use was collected by self-reported at baseline, and information on form, duration, or dosage was not fully taken into account. Therefore, we could not explore the nature of potential associations between dosage and duration of glucosamine use with the defined outcomes. Finally, taking glucosamine could represent a healthy lifestyle. Although lifestyle-related factors were adjusted as well as
E-values were calculated, it is difficult to disentangle the effects of a healthy lifestyle from the habitual glucosamine supplements. It is possible that unmeasured confounders and reverse causation remain in an observational study.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.