Risk of Alzheimer's disease and duration of NSAID use

Increasing evidence indicates that inflammation is involved in the pathogenesis of Alzheimer's disease (AD). Neuritic plaques, a cardinal neuropathologic marker of AD, are composed of amyloid peptides and numerous other proteins indicative of an inflammatory response. ^[1-8] The latter includes chronically activated microglia, complement cascade and defense proteins, cytokines, acute phase reactants, and protease inhibitors. If inflammation is part of the causal pathway leading to AD, anti-inflammatory medications may be effective in slowing disease progression or preventing onset of AD.

In population-based case-control and cross-sectional studies, evidence consistently supports an association between reported use of NSAIDs and a lower risk of AD. ^[9-14] However, previous studies have relied on retrospective recall or surrogate interviews for information on drug use. Moreover, the cross-sectional design of most studies raises concerns that reporting bias (e.g., cases or informants underreport use of NSAIDs) or selection bias (e.g., healthy survivor effect, higher education, and use of medical care) may explain the observed association. Finally, there is no evidence to date to support a dose-response relationship between use of NSAIDs and risk of AD.

To determine whether duration of NSAID use is associated with a reduced risk of AD, we used self-reported prospective data on medications from the Baltimore Longitudinal Study of Aging/National Institute on Aging (BLSA/NIA). ^[15] We also examined the association between AD risk and use of acetaminophen, a pain-relief medication with little or no anti-inflammatory activity, to assess the specificity of the association between AD risk and self-reported medications.

Methods.

The BLSA is a prospective study of "normal aging" conducted at the Gerontology Research Center/National Institute on Aging. ^[15] The cohort consists of volunteers who have been continuously recruited from the Baltimore/Washington area since 1958. To date, 1,532 (64.6%) men and 839 (35.4%) women have been enrolled. The study was initially limited to men; recruitment of women began in 1978. Subjects are predominantly white (93%) and from middle to upper-middle socioeconomic brackets. Over one-half the participants attained at least a college degree. As of September 1995, 1,165 of the 2,357 enrollees (49%) were active (i.e., had an examination within 2 years), 370 (16%) were inactive (i.e., alive but no visit in 2 years), 172 (7%) had withdrawn from the study, and 650 (28%) were deceased. This report is limited to BLSA participants who were actively followed for 1 year or more between 1980 and 1995.

Participants return biennially to the Gerontology Research Center for 2.5 days of multidisciplinary evaluations. ^[15] All subjects over age 65 had a standardized neurologic examination and neuropsychological testing in addition to the usual BLSA protocols. ^[15] Neuropsychological evaluations included the Blessed IMC Test, ^[16] Mini-Mental State Examination, ^[17] Immediate and Delayed Cued Recall, ^[18] Boston Naming Test, ^[19] Controlled Verbal Fluency, ^[20] Trail Making Tests A and B, ^[21] Clock Drawing ^[22] and other constructions, ^[23] Center for Epidemiologic Studies Depression Scale, ^[24] Pfeffer Functional Activities Questionnaire, ^[25] and the National Adult Reading Test. ^[26] The Blessed IMC Test was administered to all subjects 55 to 64 years of age. Those with a Blessed IMC score greater than 3 were clinically examined and administered the above test battery.

Results.

A total of 1,686 BLSA participants were followed for 1 or more years between 1980 and 1995 (Table 1). Eighty-one individuals were diagnosed with AD during the 16-year follow-up period. Of the 1,686 participants, 58% were men, 75% completed college, 45% were less than 55 years at enrollment, and 45% enrolled before 1980. About 80% of the subjects were followed for 5 or more years, and 90% completed two or more visits between 1980 and 1995. A total of 52% of the person-time between 1980 and 1995 occurred when participants were below age 65 (data not shown).

Since 1980, reported use of aspirin has been relatively high, ranging from 38.5 to 43.8%. Frequency of NSAID use also increased substantially after 1980 when ibuprofen was available over the counter (Table 2). Reported use of NSAIDs more than tripled between 1980-84 and 1990-95. Ibuprofen accounted for 49.7% of the reported use of drugs in this category, followed in order by naproxen (15.3%), indomethacin (8.2%), fenoprofen (1.6%), flurbiprofen (1.4%), and a variety of prescription NSAIDs. Women reported use of NSAIDs on one or more visits more often than men (53% versus 31% of visits). No difference in NSAID use was found by level of education.

Overall, a statistically significant lower risk of AD (RR = 0.50; 95% confidence interval [CI]: 0.30-0.85) was associated with reported use of NSAIDs defined as a binary variable (Table 3). Although the risk of AD was reduced among aspirin users, the RR was not significantly different from 1.0 (RR = 0.81; 95% CI: 0.52-1.28). There was no evidence of reduced risk of AD among reported users of acetaminophen (RR = 1.23; 95% CI: 0.73-2.07).

Lagging had little or no effect on relative risks for binary medication use variables (see Table 3). In contrast, the RRs by duration of use of NSAIDs varied by lagging (Table 4). In particular, the RRs for the two durations of use groups (i.e., <2 years, 2+ years) were similar when no lag was employed. In contrast, with a 2-year lag, the RRfor those with less than 2 years of reported NSAID use increased to 0.65 (95% CI: 0.33-1.29); the RR for those with 2 or more years of NSAID use decreased to 0.40 (95% CI: 0.19-0.84). The results were essentially the same for lagging up to 5 years (data not shown).

Duration of aspirin use and the risk of AD were also examined (see Table 4), but no statistically significant RRs were observed. Moreover, in contrast to NSAIDs, there was no trend of decreasing risk with increasing duration of use (Figure 1). Lagging did not affect the duration of exposure risk relationship.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1. Relative risk of AD (2-year lag) by type and duration of medication use (Baltimore Longitudinal Study of Aging, 1996).

No association was observed between risk of AD and duration of acetaminophen use, regardless of how the variable was defined or the amount of lagging. The RR for users- versus nonusers was 1.23 (1.35 for a 2-year lag). No meaningful trend was observed by duration of use (see Figure 1).

Discussion.

Evidence from this study supports an association between NSAID use and reduced risk of AD. In the past, the relationship between NSAID use and AD was assessed in either the cross-sectional or case-control design. ^[9-14] The primary concern with past studies has been that the observed protective effect of NSAIDs might be due to underreporting of NSAIDs among AD cases or their surrogates compared with non-cases. In this longitudinal study of BLSA participants, information on drug use was reported prospectively, thus avoiding the limitations of past studies. To our knowledge, this is the first study to demonstrate that increasing duration of NSAID use is associated with a decreasing risk of AD.

To determine whether the observed association between AD risk and NSAIDs was specific and not simply a product of self-reported use of medications (e.g., differential reporting between AD cases in the early stages of disease and non-cases), we examined the use of acetaminophen, a drug with little or no anti-inflammatory properties. No association was found with use of acetaminophen and risk of AD, nor was there a difference in the RR by duration of use of acetaminophen.

Although aspirin use may confer some protection against AD, the RR estimate from our study was not statistically significant and no trend of decreasing risk with increasing duration of use was observed. The less consistent evidence for aspirin use may be explained by an increasing proportion of elderly individuals taking a prophylactic dose (i.e., 65 to 85 mg/day) to prevent heart disease. Prophylactic use of aspirin for venous and arterial thrombosis was first examined in clinical trials in the late 1960s and early 1970s. Aspirin use was shown to reduce the risk of recurrent myocardial infarction ^[33] in the early 1980s, and to reduce the risk of first-time myocardial infarction in the late 1980s. ^[34] Clinical recommendations for prophylactic aspirin use may have begun before 1980. Although low-dose aspirin may be sufficient to affect platelet aggregation, a low dose of 65 to 80 mg may not be adequate for anti-inflammatory effects in the CNS.

In this study, use of NSAIDs was reported by 26% of AD cases and 48% of non-cases. In general, the observed protective effect of NSAIDs is consistent with recent evidence regarding inflammatory activity in the pathophysiology of AD, although the specific mode of action is not clear. Neuritic plaques contain complement proteins, activated microglia, cytokines, and acute phase proteins. ^[7,8,34,35] Complement proteins may induce migration of microglia and, in turn, the synthesis or release of a number of inflammatory intermediaries including interleukin-1 and prostaglandins. Activated microglia may cause neurodegeneration through the chronic release of cytotoxic host defense factors and possibly other agents that may mediate this process (e.g., glutamate, nitric oxide, superoxide). NSAIDs may influence inflammation through a number of mechanisms such as interference with activation of complement proteins and the formation and release of chemical mediators that inhibit cyclooxygenase and, in turn, prostaglandin production.

One limitation of our study is that we were unable to accurately determine when and for how long a medication was used relative to the time that use was reported during a visit. We interpreted an affirmative report of medication use to mean that exposure occurred, on average, from the midpoint between the previous visit and the visit of the report and continued to the midpoint between the visit of the report and the next visit. Given this assumption, we are likely to overestimate use in some individuals and misclassify individuals with little drug use as exposed. Information on duration of drug use was not consistently obtained from subjects during each visit. Consequently, we could not examine the validity of this assumption. Moreover, we cannot state how reduction in risk is related to actual duration of use of NSAID. Rather, duration of use is a surrogate variable that allowed us to categorize individuals by their relative duration of use of NSAIDs and other drugs.

Nonetheless, we found that AD risk decreased with increasing duration of NSAID use with a lag of 2 years. In this analysis, the 2-year lag effectively ignores exposure to medications in the 2-year period before diagnosis. There are two reasons why lagging may have helped. First, to be effective, NSAIDs may have to be taken a number of years before AD diagnosis. That is, NSAIDs may not be effective in the late preclinical phase of AD (e.g., in the 2 years before diagnosis) because the disease process has either advanced beyond the inflammatory phase or deterioration is too advanced to alter the course of the disease. Misclassification of exposure may have been introduced by including information on exposure to NSAIDs near the time of diagnosis because NSAIDs used during this time may not have been effective.

Lagging may also have reduced the potential problem of reporting differences between AD cases and non-cases. Individuals in the preclinical phase of AD may tend to underreport recent drug use compared with age- and gender-matched cognitively normal individuals. In the absence of lagging, reporting bias of this type would result in a greater apparent protective benefit from NSAID use. We found no support for this type of reporting bias (see RRs in Table 3 with and without lagging).

The conclusions in this study are essentially in line with those reported in a recent abstract. ^[36] The earlier analysis differs, however, in that it encompassed all subjects, regardless of the calendar year of follow-up. In the earlier analysis, we were concerned about the known correlation between calendar year of follow-up (before 1980 versus on or after 1980) and older age and greater risk of AD, and the inverse correlation between calendar year of follow-up and use of NSAIDs. These correlations were an inherent problem when all data were used and introduced confounding which could not be adequately controlled. The problem of confounding was resolved by limiting the analysis to follow-up time occurring on or after 1980.

Although NSAID prophylaxis offers promise as a strategy for the prevention or delayed onset of AD, it is not possible to conclude that a protective benefit is attributable to any one particular NSAID. Randomized trials will be essential to prove that a specific NSAID confers protection. In addition, trials will be critical to identify an appropriate point of intervention based on age, cognitive status, or genetic susceptibility, as well as a medication and dose that confer protective benefits with minimal adverse events. Because chronic use of NSAIDs is associated with potentially serious adverse events (e.g., gastric irritation and bleeding, peptic ulcer disease, impaired renal function, nephrotoxicity, and in rare instances renal failure), the risk of problems associated with long-term prophylactic use of these drugs must be balanced by the risk of AD. ^[34,35,37] Depending on the frequency and severity of adverse events, prophylactic oral administration of some of the currently available NSAIDs may only be suitable for clinical trials of high-risk individuals.

Acknowledgments

The authors gratefully acknowledge the participants of the Baltimore Longitudinal Study of Aging/National Institute on Aging (BLSA/NIA) who made this study possible. Special thanks are due to the staff of the Longitudinal Studies Branch of the NIA and particularly Mr. Neil Gittings for his help and guidance in assembling the BLSA data used in this study.