Does aspirin protect against Alzheimer's dementia? A study in a Swedish population-based sample aged ≥80 years

Abstract

Objective

It has been reported that aspirin and other non-steroidal anti-inflammatory drugs (NSAID) may protect against dementia of Alzheimer's type and/or vascular dementia. However, co-morbidity and the dose of aspirin may be critical. A major indication for low-dose aspirin is prophylaxis after stroke and transient ischaemic attacks, conditions that may obscure an anti-dementia effect by the drug. Alternatively, low-dose aspirin may be insufficient if the protective effect is due to an anti-inflammatory mechanism. The aim of this study was to assess whether high-dose or low-dose aspirin may protect against Alzheimer's dementia in subjects aged ≥80 years. For comparison, effects of (other) NSAID, paracetamol and d-propoxyphene were studied.

Methods

Global, cross-sectional, and longitudinal (1991–2000) epidemiological analyses of clinical, cognitive and drug treatment data on 702 individuals 80 years old or more (351 twin pairs of same sex), all alive at inclusion: mean age 83.9 years (80–99 years). Calculations were made with logistic regression of associations between use of various analgesics and cognitive function, after adjustment for age, gender, and cardiovascular and cerebrovascular diseases.

Results

Users of high-dose aspirin had significantly lower prevalence of Alzheimer's dementia and better-maintained cognitive function than non-users. There were numerically similar but not significant associations with use of low-dose aspirin and other NSAID. There were no such associations with use of either paracetamol or d-propoxyphene.

Conclusion

Aspirin might protect against Alzheimer's disease, but controlled trials are warranted.

Introduction

It has been speculated that aspirin (acetylsalicylic acid) and other non-steroidal anti-inflammatory drugs (NSAID) may prevent dementia of Alzheimer's type and/or vascular dementia [1, 2, 3, 4, 5]. A Cochrane report [6] has stated that there is no evidence of a protective effect of aspirin in vascular dementia. However, co-morbidity and the dose of aspirin may be critical for the prognosis on cognitive decline. A major indication for low-dose aspirin treatment is prophylaxis after stroke and transient ischaemic attacks (TIAs). These conditions may obscure a neuroprotective effect by low-dose aspirin. Alternatively, if the mechanism of the anti-dementia effect is anti-inflammatory, low-dose aspirin may be insufficient to produce a discernable treatment effect.

The current observational study compared the possible neuroprotective effects of different dose regimens of aspirin and of non-aspirin NSAID, paracetamol (acetaminophen) and d-propoxyphene, in a Swedish population-based cohort of individuals 80 years old or more.

Subjects and methods

Data on study subjects were available from the on-going Swedish longitudinal study "Origins of Variance in the Old-Old: Octogenarian Twins" (OCTO-Twin) [7]. The current analyses used data from 1991 through the end of March 2000. When completed, this study will include five measurement occasions with 2-year intervals. The OCTO-Twin sample was drawn from the population-based Swedish twin register [8]. All registered twins 80 years old or more from same-sex pairs, both members being alive during the initial observation period (February 1991 through March 1994), were potential study subjects. At baseline, 702 individuals in 351 twin pairs were examined by the use of a broad-based bio-behavioural assessment, carried out by registered research nurses (RNs) at the places of residence of the subjects. The mean age of subjects at inclusion was 83.9 years (80–99 years). A comparison of subsequent mortality of those included in the study and those not included revealed no difference, supporting the assumption that the sample was representative for elderly subjects in Sweden. Similarly, a comparison of the prevalence of malignancy with that in the Swedish Cancer Register 1960–1998 revealed no difference. Also socio-economic conditions were similar to those of a representative sample of Swedish singletons [9].

The impact of mortality on a possible association between cognitive decline and drug therapy regimens is hard to evaluate. For that reason, we defined two separate groups, both without any mortality influence: the first group giving pre-inclusion morbidity for the baseline cases, and the second including those being alive at the end of the observation period.

There was a significant but low correlation between the frequencies of each analgesic drug at subsequent registrations during the observation period (Table 1). This indicates that the pre-inclusion medication might have influenced the risk of cognitive decline up to the start of the observation period. Because of that, we decided to test whether the analgesic drug regimens at baseline were associated with pre-inclusion dementia. Of a total of 702 individuals, 91 turned out to have a diagnosis of dementia of any kind (55 of them had Alzheimer's disease), established earlier or by tests of memory and other cognitive abilities at the first measurement occasion. The follow-up from 1991 through 2000 comprised 315 individuals. Among them were 88 cases with dementia of any kind (50 cases of Alzheimer's disease) diagnosed during the observation period, including four measurement occasions.

Table 1. Coefficients of correlation between specific analgesics at subsequent periods ("waves") of registration (interval 2 years). All correlations significant (P<0.001) except d-propoxyphene 1 versus 3 (P<0.05)

Collection of data

Information on diseases and other health problems was obtained from medical records, self-reports and registration of the use of medication. RNs conducted standardised tests of memory and other cognitive abilities as well as an interview on health conditions. The complete assessment lasted about 3.5 h. All subjects were asked about their current medication, and they were also asked to show all their prescribed drugs. If either non-prescribed aspirin or paracetamol was not initially reported, specific inquiries on such use were made. Drugs were coded according to the Anatomical Therapeutic Chemical (ATC) classification system [10]. Prescription and dosage were registered at each occasion.

Information on diagnoses was available from medical records in 99% of all cases, and in 93% from self-reports and interviews with subjects, relatives and caregivers.

Diagnoses

A consensus diagnosis of dementia was determined for each individual on a lifetime basis using various psychometric tests (see below), interviews and medical records. The diagnosis of dementia was based on the Diagnostic and Statistic Manual (DSM)-III-R criteria [11], and that of Alzheimer's disease according to the US National Institute for Neurological and Communicative Disorders and Stroke, and the Alzheimer's disease Related Disorders Association (NINCDS/ARDRA) [12]. Vascular dementia was diagnosed according to the National Institute of Neurological Disorders and Stroke and the Association Internationale pour la Recherche et l'Enseignement en Neurosciences (NINDS/AIREN) [13]. An intact cognitive level was established by various cognitive tests (Appendix 1) including a Mini-Mental State Examination (MMSE) [14] indicating a cognitive level of at least 26.

Other diagnoses were mainly based on medical records [15]. Except for dementia, all diagnoses were classified according to ICD-10 [16]. Appendix 2 gives criteria of diagnoses other than dementia. Data on prevalence (for occasional diseases cumulated incidence) of the diseases in focus are given in Table 2.

Table 2. Disease prevalence (cumulated incidence* for occasional diseases) and gender ratio

Aspirin dosage

In Sweden, 75-mg tablets of aspirin are dispensed only by prescription for use in the prevention of cardiac and cerebrovascular diseases, whereas 250-mg and 500-mg tablets of aspirin and 250, 500 and 1000-mg tablets of paracetamol can be obtained at a pharmacy without a prescription. However, both kinds of drug sales are registered in the national drug register (ACS) maintained by the National Corporation of Pharmacies which owns all Swedish pharmacies. Individuals taking prophylactic aspirin may prefer to obtain 500-mg aspirin tablets and split them in minor fractions instead of using 75-mg tablets. Others prefer 500-mg aspirin tablets two or three times a week. In the current analysis, subjects on aspirin were grouped according to dosage. In some analyses, all subjects on aspirin less than 3500 mg per week formed a combined low-dose aspirin group.

The various analgesic regimens used during the study period are shown in Table 3. Wave 1 refers to the use at baseline (1991–1994) and waves 2, 3 and 4 to the use at the subsequent 2-year follow-up periods. With time, there was an increased use of low-dose aspirin, reflecting that this kind of treatment was officially recommended as secondary—and, later, also as primary—prophylaxis of coronary and cerebrovascular diseases. During 1998–2001, about 27% were prescribed 75 mg aspirin daily. Of 131 cases on treatment with non-aspirin NSAID in the study period, 42 used naproxen, 24 diclofenac, and 18 ibuprofen.

Table 3. Prevalence of various analgesic regimens over the study period. Aspirin total (all types of aspirin regimens), aspirin <3500 mg/week (a non-prescribed weekly dose <3500 mg), aspirin tabl 75 mg/day (prescribed 75 mg a day), aspirin low-dose total (merged low-dose regimens), aspirin prescribed 500 mg (prescribed once or more a day), NSAID (all non-aspirin NSAID)

For aspirin, non-aspirin NSAID, and paracetamol, twin-ness did not influence the use of drugs in focus.

Statistics

Differences between users of various types of analgesics and non-users of these drugs were analysed using independent t-tests for age, gender, mortality, time of observation, and diseases (myocardial infarction, angina pectoris, congestive heart failure, cardiac arrhythmia, venous thrombo-embolism, TIA, stroke, peptic ulcer, dorsalgia/osteporosis, gonarthrosis/coxrthrosis), as well as for various levels of cognitive function.

To assess the relative risks of dementia and cognitive decline associated with the use of various analgesics, logistic regression analyses were conducted with Alzheimer's disease, all dementia and intact cognitive level as the dependent variables. Each model was controlled in step 1 for age and gender. Age was entered as a continuous variable and gender as dichotomous variable (male=0, female=1). In step 2, coronary diseases, congestive heart failure, cardiac arrhythmia, and cerebrovascular disease were entered as dichotomised variables (0 = not prevalent, 1 = prevalent). The type of analgesic drug in question (0 = not prevalent, 1 = prevalent) was added to the model in step 3.

Logistic regression analyses were carried out separately for the total sample (wave 1), for a sample including only those with a clinical diagnosis of dementia before inclusion, and for a sample surviving from wave 1 through wave 3. Only data from the first three waves in the survival sample were used because of the small sample size in the fourth wave. In the survival sample, the continuous use of analgesics was also studied with a dichotomous variable (0= <2 registrations, 1= 2 or 3 registrations).

The statistical package SPSS 10.0 was used for all analyses.

Results

Description of the sample

Sample characteristics are described in Table 4, in which users of different types of analgesic drugs are compared with non-users by t-tests for independent sample. Only in ten cases (1.3%) was no analgesic drug at all used during the time of observation. The mean age of those taking 75 mg aspirin was lower than in non-users, resulting in a somewhat longer time of observation. Aspirin was more frequently used by males with a diagnosis of coronary and/or cerebrovascular diseases. Use of aspirin was avoided by individuals who had had a peptic ulcer. For treatment of arthritis, non-aspirin NSAID and d-propoxyphene were preferred.

Table 4. Participant characteristics by use of aspirin and various other types of analgesics. t-Tests for independent sample comparing users and non-users of specific drugs and doses

Use of analgesics in association with dementia, and intact cognitive function

The separate impact from various analgesic drug regimens on Alzheimer's disease, dementia of any type and intact cognitive function, respectively, was analysed by logistic regression in the total sample (Table 5), in the sample including only subjects with a clinical diagnosis of dementia before inclusion (Table 6), and in a survival sample (Table 7). In all analyses, age, gender and various vascular diseases were controlled. Because some of the diseases showed no association, they were deleted from the analysis. Of the analgesics, only aspirin in doses over 75 mg, even when given occasionally, turned out to be a significant factor. In the total sample, there was a significant association between the use of aspirin and a lower frequency of Alzheimer's disease and all dementia, respectively. In addition, it was significantly more likely that an intact cognitive function was maintained among those taking aspirin. Use of low-dose aspirin alone did not affect the risk ratio. Nor were there any significant effects by other types of analgesics.

Table 5. Logistic regressions for aspirin (total and occasional, respectively) and other factors influencing cognitive state and risk of dementia (n=702)

Table 6. Logistic regression of the risk of dementia before inclusion (n=702)

Table 7. Logistic regression of factors influencing cognitive state and risk of a new diagnosis of dementia in a follow-up from 1991 through 2000 (n=437). Exposure period aspirin use >5 years

The results on Alzheimer's disease were confirmed in the analysis of the pre-inclusion cases of dementia, as well in the survival sample, and so were the results on intact cognitive function in the latter. The results from this survival sample were reported using data from the first three waves of registration. The β values were the same also using four waves but were not significant, probably because of the reduced sample size at wave four. The effects of NSAID, paracetamol, and d-propoxyphene on Alzheimer's dissease, dementia and intact cognitive function, respectively, were tested in separate models. All of them turned out to be non-significant even though the risk ratios for non-aspirin NSAID were similar to those of aspirin.

Discussion

A recent Dutch study [5] reported a reduced risk of Alzheimer's disease in subjects treated with non-aspirin NSAID but not in those treated with aspirin. In the present study, both the cross-sectional and the longitudinal analyses indicated that aspirin might reduce the development of Alzheimer's disease. In those treated with non-aspirin NSAID, who were much fewer than those on aspirin, there was a numerically similar but insignificant reduction of risk. The discrepancy concerning aspirin between the Dutch study and the present one may relate to the fact that the Dutch study only analysed prescribed aspirin, thus excluding aspirin bought over the counter, which at least in Sweden constitutes a considerable proportion of total aspirin use. Moreover, the Dutch study did not control for the cardiac and vascular indications for prophylactic use of aspirin.

In the present study, the association between use of aspirin and reduced risk of Alzheimer's disease seemed more obvious after correction for stroke and TIA, as well as for myocardial infarction, angina pectoris and congestive heart failure. After such correction, even use of low-dose aspirin, predominantly 75 mg a day, was associated with a numerical, albeit insignificant, reduction of cognitive decline and Alzheimer's disease.

In addition to indication bias, some other aspects warrant comment. Demented and non-demented subjects may not have the same access to pharmacotherapy, which could account in part for findings of a lower risk in the total sample. Moreover, the association of a higher risk of dementia and lower analgesic use may be influenced by other factors, e.g. differences in pain sensitivity between demented and non-demented subjects. Also, demented persons may be less able to establish medical contacts and to describe their sensations. However, the absence of an influence by paracetamol use, most of which like aspirin is bought over the counter, lessens the objection that the factors above in the demented could account for the findings.

The present study also confirmed previous findings of a twofold-increased risk of mortality in demented old-old individuals [17]. To overcome this bias, we limited the follow-up to those who survived the first nine years of observation. We analysed the subgroup that remained cognitively intact, and the analogous findings in this group strengthen the assumption that there may be a genuine association between aspirin use and preserved cognitive function, unbiased by mortality differences.

Alzheimer's disease may at least in part be secondary to inflammatory processes in cognitive brain centres, the microglia being of central importance [18]. NSAID are well known to inhibit the activity of cyclooxygenases 1 and 2 (COX-1 and COX-2), whereby the formation of thromboxanes and prostaglandins is reduced. As some prostaglandins promote inflammation, COX inhibition may be the anti-inflammatory mechanism through which NSAID influence Alzheimer development, including pre-clinical cognitive decline. As selective COX-2 inhibitors appear ineffective against Alzheimer's disease [19], it seems probable that the anti-Alzheimer effect resides in COX-1 inhibition.

In conclusion, an inhibitory effect of aspirin and other NSAID on the inflammatory brain processes contributing to Alzheimer's disease and cognitive decline seems plausible, even if convincing evidence from controlled intervention trials still is lacking. The possibility that a well-documented and non-expensive agent such as aspirin might function to maintain cognitive function and reduce the development of Alzheimer's disease makes it worth testing aspirin as an agent against cognitive decline.

Appendices

Appendix 1

Evaluation of cognitive status and dementia in the OCTO Twin Study

A rating of cognitive status (Cognitive Rating, CR) was conducted on all subjects using the scale below. The CR represents a composite score and was used as a convenient way to summarise performance across the tests and to provide an evaluation of the person's overall memory and cognitive function, taking overall health and functioning into account.

• 1. Intact—"normal"—memory and cognitive functioning (corresponding MMSE ≥26), sensory and motor function, overall health, and physical functioning were taken into account

• 2. Mild dysfunction/questionable impairment: evidence of compromised memory and/or cognitive functioning, not meeting DSM-III-R criteria for dementia

• 3. Mild impairment/dementia: meeting behavioural criteria for mild dementia [20]

• 4. Moderate impairment/dementia: meeting behavioural criteria for moderate dementia DSM-III-R criteria [20]

• 5. Severe impairment/dementia: meeting behavioural criteria for severe dementia [20]

The CR was based on a thorough review of memory and cognitive performance. The test battery encompassed:

• 1. Orientation to own person (name, year of birth, current age)

• 2. Self-ratings of memory and thinking (six questions)

• 3. The Mini-Mental State Examination (MMSE), [14]

• 4. The Information Test (A Swedish version of the WAIS Information Task [21]

• 5. Figure Logic/Inductive Reasoning (SRB:2) [22]

• 6. Block Design—Koh's Block Test ( SRB 3) [22]

• 7. Verbal Meaning—Synonyms (SRB 1) [22]

• 8. Digit—Symbol (A modified version of the speeded Symbol-Digit Substitution Test from the Wechsler Adult Intelligence Scale, WAIS, where subjects were instructed to give a verbal response, instead of a written) [23]

• 9. Digit—Span, Forward and Backward test (WAIS) [23]

• 10. Perceptual Speed—Psif [22]

• 11. Thurstone's Picture Memory Test [24]

• 12. Prose Recall (A Swedish language prose recall task similar to the prose passages in the Wechsler Memory Test (WMS) [23]

• 13. The MIR Test including naming, free recall, recogntion and correspondence tasks [25]

• 14. The Swedish Clock Test (including the three subtests: clock drawing, set the hands of a wooden clock with no numbers on the face to certain standard times, and set time to certain standard times [26]

• 15. The Coin Test [26]

The CR was made in the context of the subject's overall health and physical functioning, including, e.g., sensory functioning and motor handicaps. Also, observations of test-taking behaviour and everyday functioning during the in-person testing sessions were presented and taken into account before determining a CR score. In this respect, the CR represent a clinical evaluation based on multiple sources of information.

Individuals rated with mild to severe impairment (3–5) and tentatively diagnosed as dementia or demented, were routinely given a diagnostic work-up, including a detailed interview with a key informant about memory and cognitive problems (onset, course and symptomatology), and a review of medical records. Findings were presented and discussed at a consensus diagnosis conference. Diagnoses were assigned following DSM-III-R criteria for dementia [20], the NINCDS/ADRDA criteria for Alzheimer's disease [27], and the NINDS-AIREN criteria for vascular dementia [13].

Appendix 2

Definitions and distinctions for examined diagnoses (codes within parentheses refer to ICD-10)

• Myocardial infarction (I21–22): the diagnosis requires an increase of specific enzymes in conjunction with the typical clinical manifestations and/or an ECG-pattern with specific QRS-aberrations.

• Angina pectoris (I20): prescription of nitrates is considered to be a valid evidence for the diagnosis. In cases where angina is observed in temporal connection with a myocardial infarction, only infarction is registered.

• Congestive heart failure (I50) means dyspnea and oedema referable to heart disorders, and typically treated with diuretics and/or digitalis.

• Arrhythmia (I44–49): a diagnosis of cardiac arrhythmia is determined only when medical records explicitly use arrhythmia as a diagnosis. That means that fibrillation and other rhythm disturbances in connection with ongoing congestive heart failure are not included. The chief indications are paroxysmal attacks and arrhythmia in conjunction with heart blocks or Adams-Stokes attacks.

• TIA (transitory ischaemic attack) (G45): acute neurological symptoms <24 h. Diagnosis only on the basis of medical records.

• Stroke (I61;I63–65) is accepted as a diagnosis from information in records as well as from self-reports, although the few cases diagnosed only from the latter source may include TIA attacks.

• Venous thromboembolism (I121;I180) includes deep thrombosis of lower extremities and/or pulmonary embolism.

• Peptic ulcer of stomach and duodenum (K25–26) is clinically diagnosed with fibre endoscopy or X-ray support. Self-reports are accepted as the base for diagnosis.

• Dorsal pain (M54): the diagnosis is mainly based on self-reports. In cases where the diagnosis is derived from medical records it requires a specific treatment or X-ray documented spondylarthritis. For cases in which the X-ray confirms demineralisation and/or spontaneous fractures, the diagnosis is osteoporosis (M80–81).

• Coxarthritis (M16) and gonarthritis (M17) as clinical diagnoses require an X-ray confirmation and/or surgical intervention. As for dorsal pain, coxarthritis also includes mild and less well-documented cases derived from self-reports. For gonarthritis only record information was available.