Sex Difference in Effects of Low-Dose Aspirin on Prevention of Dementia in Patients With Type 2 Diabetes: A Long-term Follow-up Study of a Randomized Clinical Trial

The number of patients with dementia is growing rapidly and is estimated to reach 82 million in 2030 and 152 million in 2050. Dementia has significant social and economic implications, and the total global societal cost for managing dementia was estimated to be 818 billion USD in 2015 (1). However, prophylaxis for dementia has not been established.

Several factors have been reported to be risk factors for dementia, such as hypertension, tobacco use, and obesity (2,3). Type 2 diabetes (T2D) is one of the important risk factors for the development of both vascular and nonvascular dementia (4,5). Patients with T2D are considered to be at a high risk of developing dementia due to multiple risk factors such as hypertension and obesity. Since there is a huge worldwide burden of T2D (6), the establishment of strategies for prevention of dementia in T2D patients is a pressing issue.

Since both vascular dementia and Alzheimer dementia (AD) share several pathophysiological factors with atherosclerotic disease, aspirin has been attracting interest as a preventive agent for dementia (7). While some studies have suggested that low-dose aspirin may protect against AD by improving platelet and endothelium functions (7), the findings have not been consistent. Meta-analysis of randomized controlled trials (RCTs) with median follow-up of 5 years evaluating the effects of low-dose aspirin on cognitive functions has reported no improvement (8). On the other hand, some observational studies have reported an association between low-dose aspirin and a lower risk of dementia or cognitive decline (9,10). However, no studies have evaluated the long-term association between low-dose aspirin and the risk of dementia in patients with T2D using real world data, even though the pathological change of typical dementia develops slowly over decades.

Furthermore, differences have been suggested in the risk of developing dementia between the two sexes (11–14). For instance, the genotype of apolipoprotein E (ApoE), which plays a role in the prevention of dementia by nonsteroidal anti-inflammatory drugs (NSAIDs) (15), seems to have a stronger effect in women compared with men (13). However, only a few studies have evaluated the difference in the effects of low-dose aspirin in the prevention of dementia between men and women.

Therefore, in this cohort study, we evaluated the effects of low-dose aspirin on the prevention of dementia among patients with T2D. We also compared the effects of low-dose aspirin between men and women.

The JPAD2 study is a prospective follow-up cohort study of the JPAD trial (Japanese Primary Prevention of Atherosclerosis With Aspirin for Diabetes), which was a randomized, open-label, standard care–controlled trial that examined the effects of low-dose aspirin on cardiovascular events in 2,539 Japanese patients with T2D (Supplementary Figs. 1 and 2). A detailed description of both the studies has previously been published (16,17). Briefly, patient enrollment in the JPAD trial started in December 2002 and was completed in May 2005 at 163 institutions throughout Japan (Supplementary Data). After the trial was completed in April 2008, we continued to follow up all the patients and constructed the JPAD2 cohort. Patients were followed until the day of any fatal event or July 2017, regardless of the occurrence of any cardiovascular event. Both the JPAD trial and JPAD2 cohort study were carried out according to the Declaration of Helsinki and were approved by the ethics committee of each participating hospital (the Graduate School of Medical Science, Kumamoto University, ethics committee and Nara Medical University ethics committee). All the participants provided written informed consent. The JPAD trial was registered at ClinicalTrials.gov.

The details of the inclusion and exclusion criteria for the JPAD trial have previously been published (16,17). Briefly, the inclusion criteria were 1) diagnosis of T2D, 2) age 30–85 years, and 3) ability to provide informed consent. The exclusion criteria included a history of cardiovascular disease (CVD) including cerebral vascular disease and the use of antiplatelet or antithrombotic therapy. A total of 2,536 patients were enrolled in the JPAD2 cohort study (Supplementary Figs. 1 and 2). None of the patients had dementia at the baseline of the JPAD2 study.

Patients were randomly allocated (1:1) to receive low-dose aspirin (81 mg or 100 mg daily) or no aspirin in the JPAD trial. All participants were allowed to receive other concurrent treatments. The details of the intervention have previously been described (16). At the end of the JPAD trial, the participants were administered low-dose aspirin based on the decision of each physician during the follow-up period. We checked whether the participants were administered low-dose aspirin during the follow-up.

Dementia was ascertained based on 1) prescription of antidementia drugs or 2) admission to a hospital/nursing home due to dementia. All events were reviewed with a case report form by an independent event adjudication committee that was blinded to the assigned intervention. The event adjudication committee requested detailed information from each participating hospital when additional information was necessary to adjudicate the events. When there were disagreements on the diagnosis of dementia, all committee members participated in the discussion and made final agreements.

We present the baseline characteristics as mean and SD for continuous variables and number and percentage for categorical variables. We used the t test or Wilcoxon rank sum test for continuous variables and χ² tests for categorical variables to compare the baseline characteristics between groups. The follow-up time was computed from baseline until the occurrence of dementia, death, or the date of last known contact. Cumulative incidence of dementia in each assigned group was estimated using the Kaplan-Meier method, and differences between groups were assessed using the log-rank test. We used the Cox proportional hazards models to estimate the hazard ratio (HR) with the corresponding 95% CI for the efficacy of low-dose aspirin in lowering the incidence of dementia. The initial model (model 1) was adjusted for age and sex, while the second model was additionally adjusted for BMI, hypertension, dyslipidemia, smoking status (current and past), and HbA_1c (model 2). We then evaluated the association and interaction between low-dose aspirin and the risk of dementia based on sex. We also performed sensitivity analyses to assess the robustness of our findings by repeating Cox proportional hazards modeling using the per-protocol cohort.

All analyses were conducted using JMP 12.2 and SAS 9.4 (SAS Institute, Cary, NC). Two-tailed P values <0.05 were considered statistically significant.

The mean ± SD age of patients at baseline was 65 ± 10 years, and the median duration of diabetes was 7.0 years (interquartile range 2.9–12.3). The baseline characteristics were similar between patients in the two groups (Table 1). Among the patients initially enrolled in JPAD, 2,121 of them (84%) retained their original allocation until July 2017. During a median follow-up of 11.4 years, 128 patients developed dementia (5.4/1,000 person-years). The incidence of dementia was not significantly different between the low-dose aspirin and no-aspirin groups (log rank, P = 0.73) (Fig. 1). In addition, the Cox proportional hazards model adjusted for age, sex, and other established risk factors for dementia did not show a significant association between the low-dose aspirin group and risk of dementia (HR 0.82, 95% CI 0.58–1.16) (Table 2). Among the covariates included in the multivariate Cox proportional hazards model, only age was significantly associated with an increased risk of dementia, with an HR of 1.16 (95% CI 1.11–1.20, P < 0.001; 1-year increase of age).

The incidence of dementia was greater in women compared with men (women 6.5/1,000 person-years vs. men 4.5/1,000 person-years, P < 0.001). Women in the low-dose aspirin group had a lower incidence of dementia compared with those in the no aspirin group (Fig. 2A). However, no such difference was seen between men in the two groups (Fig. 2B). The adjusted HR for the risk of dementia among women in the low-dose aspirin group was 0.58 (95% CI 0.36–0.95), while that in men was 1.27 (95% CI 0.75–2.13) (Table 2). The interaction between the groups and sexes was statistically significant (P_interaction = 0.03).

Sensitivity analysis using data from the per-protocol cohort consisting of 2,121 patients who retained their original allocation showed the robustness of the primary findings. The low-dose aspirin group did not have decreased risk of dementia overall in sensitivity analysis (HR 0.70, 95% CI 0.46–1.06). On the other hand, the risk of dementia among women in the low-dose aspirin group was statistically significant (HR 0.47, 95% CI 0.25–0.86), but not in men (HR 1.12, 95% CI 0.62–2.02).

This study suggests that low-dose aspirin may be effective in the prevention of dementia in women with T2D. Our study is the first to prospectively evaluate the long-term association between low-dose aspirin and the risk of dementia in patients with T2D using real-world data.

A nationwide retrospective study from Taiwan reported that a mean daily dose of aspirin within 40 mg was associated with a lower risk of AD in patients with T2D (HR 0.5, 95% CI 0.27–0.97) (18). However, this study used a health insurance database, and thus the dose of aspirin as well as the patients’ diagnosis and clinical status did not necessarily conform with reality. In fact, the dose of <40 mg/day aspirin reported in this study, which was calculated by dividing the cumulative dose by the total observational days, is far from the optimal dose used in general practice. Some prospective observational studies have also reported a significant association between the use of aspirin and a lower risk of AD or cognitive impairment. Yet, since the participants in these studies were elderly (age >70 years), the focus was not on patients with T2D (7,9,10). On the other hand, several observational studies have reported no significant association between the use of aspirin and risk of dementia (19,20). However, many of these studies did not clarify the definite dose of aspirin used, and there is a possibility that the effects of aspirin on the prevention of dementia could be dose dependent (18). Besides the dose, the duration of low-dose aspirin use is also important. Recently, the Aspirin in Reducing Events in the Elderly (ASPREE) trial, which evaluated the effects of low-dose aspirin on a composite of death and dementia prevention, reported no significant effects of aspirin on dementia prevention in an aged population (21). Even though ASPREE is a well-designed large RCT, the median follow-up period of 4.7 years of this trial is much shorter than that of our study. Since the development and progression of dementia usually span over a long period of time, long-term use of low-dose aspirin, rather than short- or intermediate-term use, is needed to see the preventive effects. In fact, to the best of our knowledge, among the observational studies that have evaluated the association between aspirin and risk of dementia, our study with a median follow-up of 11.4 years has been the longest, while most of the previous studies followed up participants for <8 years. Interestingly, in the current study, the cumulative incidence of dementia dramatically increased 8 years after the initial enrollment. In addition, it is suggested that the preventive effects of aspirin on dementia may be enhanced in subjects with cardiovascular risk (18,22,23). Since the current study consisted of patients with a high cardiovascular risk profile, the long-term use of low-dose aspirin could partly explain our findings.

Dementia and CVD share several pathophysiological factors, and some previous studies suggested potential sex difference in the effects of aspirin on CVD prevention (24,25). A meta-analysis of six trials evaluating the effects of aspirin in CVD prevention in patients with diabetes reported that aspirin reduced the risk of myocardial infarction (MI) only in men and not in women (24). Also, a meta-analysis of 23 trials that evaluated whether sex might play a role in explaining the variation of aspirin efficacy in MI prevention trials suggested that sex accounted for a substantial proportion of the variability in the efficacy of aspirin in reducing MI rates (25). On the contrary, recent meta-analysis of 12 RCTs evaluating the efficacy of aspirin for the primary prevention of CVD in patients with diabetes revealed no sex difference in the effects of aspirin (26). The reason for inconsistent results is uncertain. However, other than the condition diabetes, age, obesity, and kidney function may influence the effects of aspirin on CVD.

Several studies have reported a favorable association between the use of aspirin and better cognitive function in women (10,23), but the underlying mechanism of the difference between sexes was uncertain. The ApoE genotype could to be a potential factor for this difference. Carriers of the E4 allele of ApoE are more likely to develop AD compared with noncarriers (13). Some studies have suggested effect modification by the ApoE genotype in the association between NSAIDs and AD (15,27). Users of NSAIDs carrying the E4 allele of ApoE are known to have 50–65% lower risk of developing AD compared with nonusers, and this effect is more pronounced in women compared with men, even though this genotype is equally prevalent in women and men (13). Nonetheless, in this study, we did not obtain any genetic information of the participants, which makes it impossible to evaluate the influence of the ApoE genotype. In addition, anti-inflammatory effects of aspirin should play an important role in the prevention of dementia in addition to CVD (28), and such effects might be different between sexes. Thus, further studies evaluating the sex differences in the effects of low-dose aspirin on inflammation cytokines in conjunction with development of dementia should be conducted.

We also need to consider gender difference in sociocultural factors that may play some role in the association of low-dose aspirin and risk of dementia by sex in our study. For instance, low education level and low occupation history are important risk factors for the development of dementia (29), which can be different by sex. Also, women are reported to admit to hospital/nursing homes more often than men, as men tend to be cared for by their spouse at home in Japan (30). Because we ascertained dementia as admission due to dementia or prescription of antidementia drugs, it is possible that higher incidence of dementia in women compared with men in our study was influenced by this sociocultural aspect. For the possible difference in the prescription rate of antidementia drugs by sex in Japan, we could not find any previous report. However, one study from Spain reported that there was no sex difference in the prescription rate of the central nervous system drugs in dementia patients (31).

Further studies that include genetic and sociocultural information are needed to assess the sex-based difference in the effects of low-dose aspirin on the prevention of dementia.

Our study has several strengths, including its large sample size and long duration of follow-up. However, it also has potential limitations. First, the diagnosis of dementia was not based on a cognitive function test and diagnostic imaging, but the ascertainment of dementia was based on admission due to dementia or prescription of antidementia drugs, and thus, the cases of dementia in our study could tend to be advanced dementia and some cases of mild dementia may not be accounted for in our study. And, thus, there can be concern that the “undiagnosed mild dementia” group primarily consists of subjects who received low-dose aspirin. However, the disproportion of undiagnosed dementia in each intervention group may not theoretically occur in this study, since the JPAD trial was designed as an RCT originally. Second, we could not differentiate the subtypes of dementia. Because the pathophysiological features of each subtype of dementia were different and the effects of low-dose aspirin might be different by each subtype of dementia, combining all the types could offset the effects of low-dose aspirin. On the other hand, AD and vascular dementia, the major subtypes of dementia, share several pathophysiological factors with atherosclerotic disease. In addition, the differentiation of subtype of dementia is sometimes very difficult in clinical settings. Third, the JPAD trial was designed as an RCT to evaluate the efficacy of low-dose aspirin for the prevention of CVD but not for dementia. Thus, the sample size was insufficient to evaluate the latter. Fourth, the residual confounding risk factors (i.e., physical activity, education) and genetic variations may have impacted our findings, though we have comprehensively controlled for other risk factors for dementia. In addition, since the low-dose aspirin was randomly assigned in the original RCT, and the allocation of treatment was not changed for the majority of the patients, the effect of the residual confounders on the findings should be small if any.

Nevertheless, no large-scale RCT to date has evaluated the efficacy of the long-term use of low-dose aspirin for the prevention of dementia. Additionally, none of the previous observational studies have evaluated the association of low-dose aspirin with the risk of dementia, especially in a population at high risk, using real-world data. Our study, therefore, could provide some important leads in the prevention of dementia.

In conclusion, low-dose aspirin might reduce the risk for dementia in women with T2D but not in men. These findings should be further validated by studies with larger sample sizes and longer follow-up periods including genetic and sociocultural evaluation of the participants.

Acknowledgments. The authors are indebted to the participants of the JPAD trial and the JPAD2 cohort study for their outstanding commitments and cooperation. The authors thank M. Ohtorii (Institute for Clinical Effectiveness, Kyoto, Japan) for her roles in data management and statistical analyses. The authors also thank M. Nagahiro, M. Okamoto, and M. Aoyama (Kumamoto University) and Y. Wada, Y. Kamada, and M. Miyagawa (Nara Medical University) for secretarial work.

Funding. JPAD was supported by the Ministry of Health, Labour and Welfare of Japan (H16-Junkanki-004, H26-Iryo-Ippan-012, H27-Junkanki-Ippan-001, and H28-ICT-Ippan-004). The JPAD2 cohort study was supported by the Japan Heart Foundation and Japan Society for the Promotion of Science KAKENHI grants 26293159, 16H05297, 17K18278, and 18H03032.

Duality of Interest. C.M. reports research grants from Morinaga. H.O. reports research grants from Shionogi, Daiichi Sankyo, Chugai, Novartis, and Bayer; nonpurpose research grants from Abbott, Eisai, Ono, Otsuka, Johnson & Johnson, Daiichi Sankyo, Sumitomo Dainippon, Takeda, Teijin, Terumo, Nihon Kohden, Bayer, Fukuda Denshi, Boston Scientific, Mochida, Mitsubishi Tanabe, Medtronic, Teijin Pharma Home Healthcare, and Boehringer Ingelheim; and lecturer’s fees from Merck Sharp & Dohme (MSD), Daiichi Sankyo, Takeda, Bayer, AstraZeneca, Eisai, Otsuka, and Teijin. Y.S. reports research grants from Novartis, Ono, Shionogi, Teijin, St. Jude Medical, and Mitsubishi Tanabe; nonpurpose research grants from Astellas, Boston Scientific, Chugai, Daiichi Sankyo, Sumitomo Dainippon, Eisai, Fuji Yakuhin, Kyowa Hakko Kirin, Medtronic, Mitsubishi Tanabe, MSD, Nihon Medi-Physics, Ono, Otsuka, Pfizer, Sanofi, Shionogi, Takeda, Teijin, and Zeria; lecturer’s fees from Asahi Kasei, Astellas, Bayer, Daiichi Sankyo, Sumitomo Dainippon, Fuji Yakuhin, Kowa, Kyowa Hakko Kirin, Mitsubishi Tanabe, MSD, Boehringer Ingelheim, Novartis, Ono, Otsuka, Pfizer, Sanofi, Taisho Toyama, Takeda, and Toa Eiyo; a manuscript fee from Pfizer; service on advisory boards of Novartis, Pfizer, Mitsubishi Tanabe, Ono, and Boehringer Ingelheim; and a sponsored office from MSD. S.O. reports lecturer’s fees from Novo Nordisc, Mitsubishi Tanabe, Sumitomo Dainippon, MSD, Bayer, Eli Lilly, Boehringer Ingelheim, Ono, AstraZeneca, Sanofi, Takeda, and Arkray. H.S. reports a research grant from Boehringer Ingelheim and lecturer’s fees from Boehringer Ingelheim, Sumitomo Dainippon, and MSD. M.N. reports lecturer’s fees from Bayer, Shionogi, Takeda, Daiichi Sankyo, Sanofi, Boehringer Ingelheim, Sumitomo Dainippon, Fujifilm Medical, Kowa, and Pfizer. N.D. reports lecturer’s fees from Daiichi Sankyo, Mitsubishi Tanabe, Takeda, Otsuka, Astellas, Boehringer Ingelheim, Abbott, Bayer, Medtronic, and Pfizer. H.J. reports research grants from MSD, Boehringer Ingelheim, Novo Nordisk, Daiichi Sankyo, Takeda, Taisho Toyama, Astellas, Chugai, Bayer, Sanofi, GlaxoSmithKline, Sanwa Kagaku Kenkyusho, Ono, Eli Lilly, AstraZeneca, Pfizer, and Shionogi; lecturer’s fees from MSD, AstraZeneca, Astellas, Abbott, Sanofi, Terumo, Novo Nordisk, Bayer, Sanwa Kagaku Kenkyusho, Kyowa Hakko Kirin, Taisho Toyama, Daiichi Sankyo, Teijin, Mitsubishi Tanabe, Eli Lilly, Boehringer Ingelheim, and Takeda; and manuscript fees from Novo Nordisk and Taisho Toyama. M.W. reports a research grant from Sanofi and lecturer’s fees from MSD, Astellas, Amgen Astellas BioPharma, AstraZeneca, Otsuka, Ono, Kowa, Kyowa Hakko Kirin, Novartis, Sanofi, Sanwa Kagaku Kenkyusho, Johnson & Johnson, Daiichi Sankyo, Taisho Toyama, Sumitomo Dainippon, Takeda, Mitsubishi Tanabe, Teijin, Eli Lilly, Novo Nordisk, Bayer, Pfizer, and Boehringer Ingelheim. T.M. reports a research grant from Nexis; lecturer’s fees from AbbVie, AstraZeneca, Daiichi Sankyo, Kyorin, Mitsubishi Tanabe, and Pfizer; a manuscript fee from Pfizer; and service on advisory boards of Asahi Kasei, Bristol-Myers Squibb, and Boston Scientific. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. C.M., H.O., Y.S., and T.M. contributed to study design and conception. C.M. and T.M. contributed to data access, responsibility, and analysis. C.M. and T.M. contributed to drafting the manuscript. H.O., Y.S., S.O., H.S., M.S., I.M., M.N., N.D., H.J., and M.W. contributed to manuscript review for scientific content. H.O., Y.S., and T.M. contributed to study supervision. T.M. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.