Epidemiological data suggest that non-steroidal anti-inflammatory drugs (NSAIDs) stave off dementia, but multiple clinical trials have failed to bear that out. Now, the most recent and, the authors say, final effort has come up short. In the INTREPAD trial, led by John Breitner, McGill University in Montreal, asymptomatic people in their 60s at high risk of Alzheimer’s disease took daily naproxen for two years. As published online April 5 in Neurology, the regimen did not affect their cognition or imaging and CSF markers, and may have even made the primary outcome, a composite of all these measures, a bit worse. What’s more, and as predicted, the drugs caused significant side effects, including gastrointestinal and cardiovascular problems.

  • Epidemiology has suggested NSAIDs curb risk of AD.
  • New study finds naproxen does not slow cognitive decline in people at high risk.
  • Previous failures, dangerous side effects, doom NSAID use for prevention.

Taken together with previous trials, INTREPAD should mark the end of the road for NSAIDs and AD prevention, the authors and several commentators agreed. “This is a good and important study that allows us to conclude that NSAIDs are just not going to be effective in this condition,” said Lon Schneider, University of Southern California Keck School of Medicine, Los Angeles. Eric Larson of the Kaiser Permanente Washington Health Research Institute in Seattle expressed similar sentiments. “The drugs caused enough harm to be detectable in a trial, and they produced no benefit. School is out for NSAIDs for prevention,” Larson said.

Breitner sees it the same way. “I do not think there should be any more clinical trials of NSAIDs for prevention or treatment of AD,” he said. “There was a lot of hope, based on epidemiological data, that these drugs would prove to be beneficial, but we’ve learned over the last 10 to 15 years that when you test NSAIDs in trials, they are not beneficial and invariably, you make people sick,” he added.

A two-decade stream of epidemiological studies linked the use of NSAIDs for chronic inflammatory conditions to a lower incidence of Alzheimer’s disease. The data fit a biological theory that inflammation stokes AD pathogenesis, and led to a series of trials of various NSAIDs for prevention. In the largest of those, ADAPT, people at high risk of AD due to a family history of the disease took daily doses of naproxen or celecoxib, and were monitored for age-related cognitive decline or onset of AD.  The trial was supposed to run for five to seven years, but was stopped early due to concerns about cardiovascular safety. The truncated trial revealed no evidence for efficacy, but the data held out the possibility that naproxen could stem cognitive decline in people treated in the earliest presymptomatic stages (Breitner et al., 2011). 

INTREPAD, described by co-first authors Pierre-Francois Meyer and Jennifer Tremblay-Mercier, was designed specifically to test naproxen in this earliest-stage, at-risk, but cognitively normal group. To join the study, people had to have a strong family history of AD, including a parent or multiple siblings, or both, with the disease. The researchers rigorously screened participants, mostly in their 60s, for what Breitner termed “cognitive fragility,” or any signs of decline or incipient dementia. The participants took 220 mg naproxen twice daily, or placebo.

For the trial, the investigators developed a new composite primary endpoint to track disease progression in asymptomatic volunteers. The Alzheimer Progression Score (APS) comprised the RBANS battery to measure cognitive decline, along with assessments of olfactory function, brain structure, cerebral blood flow, and, in half the volunteers, cerebrospinal fluid total tau and Aβ42. An increase in the APS signaled disease progression.

Over the two-year treatment period, the APS composite worsened in the group as a whole, but there was no difference in the rate of change between treatment and placebo groups. On the contrary, there was a trend for naproxen to promote progression, as reported for people with memory impairment in earlier studies. None of the individual endpoints benefitted from treatment.

Naproxen did cause more side effects, and more severe adverse events. Gastrointestinal complaints, constipation, shortness of breath, high blood pressure, and petechiae—small lesions caused by bleeding under the skin—were the most common. Serious vascular or cardiac events were also more common in the naproxen group.

The trial had limitations. The CSF measures were expected to enhance its power to detect changes, but did not. As a result, the trial was underpowered. In addition, the baseline composite scores of the treatment and placebo groups differed. In an accompanying editorial, Linda Hershey at the University of Oklahoma Health Center in Oklahoma City and Richard Lipton, Albert Einstein College of Medicine, Bronx, New York, raised the prospect that another trial with a different dose of naproxen, another NSAID, or a different high-risk group might produce a different result. Breitner and colleagues disagree. They believe that due to the lack of apparent benefit, and the demonstrated harm, further trials with NSAIDs would raise ethical concerns.

This is not the first time AD epidemiological data has not panned out in subsequent clinical trials. Other examples include estrogen replacement, rosiglitazone, statins, and others. It is unclear why the trial results are at odds with the epidemiological data. In the case of NSAIDs, perhaps their use is merely an indicator for an underlying immune response that spares people from AD. “I don’t think we can exclude that possibility, based on everything we’ve done up to this point,” Breitner said.

Gregory Cole, University of California, Los Angeles, pointed out that the dosing regimen could be critical: Using NSAIDS intermittently, as many people do, versus the constant exposure in the trial, could modulate the immune system differently (see comment below). 

Piet Eikelenboom, Vrije University in Amsterdam, has pioneered research into inflammatory processes in aging and Alzheimer’s. He argues that the original rationale for NSAID treatment may no longer hold, and noted a meta-analysis suggesting that the reported NSAID benefits may have resulted from recall, prescription, and publication bias (de Craen et al., 2005). In addition, Eikelenboom said, a growing understanding of the positive effects of inflammation on limiting AD pathogenesis supports the idea that NSAIDs would be detrimental early in disease (see comment below). See also ibuprofen, Flurizan.—Pat McCaffrey


  1. In this carefully performed study, Meyer and colleagues show that in cognitively intact individuals at risk, naproxen does not reduce the progression of presymptomatic AD. The authors conclude that these findings left them with extreme pessimism regarding any possible role of NSAIDs in AD prevention. In an accompanying editorial, Hershey and Lipton discussed the question of whether this is the end or shall we try again. They choose the last option and suggest that trials with a different dose of naproxen, a more CNS-penetrant NSAID, or a different high-risk group of patients, could have led to different study results. I disagree with this view and believe that, seeing the negative study of Meyer et al., it is now time for us to reconsider the initial biological and epidemiological arguments for the view that NSAIDs could prevent AD.

    The roots of these arguments are based in research performed 20 to 30 years ago. In the ’90s, it was a widely accepted idea that activated microglia could kill neurons by bystander lysis, and that we had to hamper microglia activation by anti-inflammatory drugs such as NSAIDS. Research of the last two decades have changed this view, and both beneficial and detrimental effects of microglia activation have been described depending on factors such as age, disease stage, etc. (Wyss-Coray and Mucke, 2002; Hoozemans et al., 2008). In the early, preclinical stage of AD, mild microglial activation could play a beneficial role. If this is the case, then we have to take into account possible negative consequences with the use of anti-inflammatory drugs in presymptomatic AD persons.

    The epidemiological arguments for the use of NSAIDs in prevention trials come from observational data obtained from large population-based studies such as the most-quoted Rotterdam Study (in 't Veld et al., 2001). However, similar population-based studies were unable to confirm that the use of NSAIDs could prevent cognitive impairment or incidental AD. De Craen and colleagues have performed a meta-analysis of the epidemiological evidence for the association between NSAIDs and the risk of dementia, and they conclude that most of the reported beneficial effects of NSAIDs may result from various forms of bias: recall bias, prescription bias, and publication bias (de Craen et al., 2005). So, we need to discuss whether the initial biological and epidemiological arguments for the use of NSAIDs in prevention studies can still be considered a valid and solid basis for prevention trials with NSAIDs.

    The study of Meyer did not suggest that we can close the book on inflammation-based therapies, but we can, in my opinion, turn the page on the use of NSAIDs in prevention of AD.


    . Inflammation in neurodegenerative disease--a double-edged sword. Neuron. 2002 Aug 1;35(3):419-32. PubMed.

    . Cyclooxygenase-1 and -2 in the different stages of Alzheimer's disease pathology. Curr Pharm Des. 2008;14(14):1419-27. PubMed.

    . Nonsteroidal antiinflammatory drugs and the risk of Alzheimer's disease. N Engl J Med. 2001 Nov 22;345(21):1515-21. PubMed.

    . Meta-analysis of nonsteroidal antiinflammatory drug use and risk of dementia. Am J Epidemiol. 2005 Jan 15;161(2):114-20. PubMed.

  2. The INTREPAD trial is notable as a follow-up to questions raised by the ADAPT naproxen trial, e.g., if starting patients earlier, prior to any decline, and treating for a full two years with 400 mg naproxen daily can impact rates of decline and/or tau/Aβ42 in CSF, as suggested by some results of ADAPT. The answer is no, but with some limitations concerning the lack of statistical power for the cognitive endpoints and the sub-optimal randomization of baseline CSF tau and Aβ42, which John Breitner and the writers of an accompanying editorial very reasonably discuss.

    In response to this NSAID clinical trial failure despite observational evidence of NSAID protection, Breitner suggests that either you have to start treatment too early to be tested in trials, or the underlying peripheral inflammatory diseases prompting NSAID use might turn out to be the actual protective factor explaining the association of long-term NSAID use with risk reduction for dementia. While the latter is certainly possible, other observational data argue that higher C-reactive protein or TNFα and other markers of peripheral inflammation predispose to dementia.

    As the genetic data strongly implicate altered immune function in AD risk, there remains a very strong argument for trying to develop an efficacious immunomodulatory approach to AD. And as with immunomodulatory amyloid reduction approaches, such as the recently failed Biogen aducanumab antibody, you may have to intervene significantly earlier to be effective. 

    One limitation of chronic NSAID trials in capturing the apparent protection observed in groups with over-the-counter NSAID use, is that most non-prescription NSAID use by patients is intermittent and not continuous. The cyclooxygenase targets of conventional NSAIDs such as naproxen initially produce pro-inflammatory PGE2 in first responder innate immune cells, but subsequently they produce products that contribute to the resolution of inflammation (Ricciotti and FitzGerald, 2011). Thus, patients who take lower doses or take NSAIDS only as needed to control intermittent symptoms are less likely to exert a block for resolution of inflammation and the useful innate immune functions that are involved. Resolution-stage functions include phagocytic clearance and neurotrophic factor production.

    Subjects compelled to take a daily NSAID dose sufficient to increase risk for GI side effects are also at risk for losing the beneficial functions of innate immune cells. For example, our recent paper in preclinical models finds that low but not high doses of curcumin can increase expression of TREM2, TYROBP, and CD11c and apparent phagocytic amyloid clearance while suppressing pro-inflammatory cytokines and the neurodegenerative microglia driver miR-155 (Teter et al., 2019). In contrast, chronic high doses of curcumin also suppressed inflammation, but the high doses were no longer permissive for TREM2/TYROBP increases or M2 indices of resolving-phase innate immune differentiation. The high dose also failed to promote amyloid clearance, arguing for benefits from permissive immunomodulation, rather than immunosuppression.

    What the “anti-inflammatory” naproxen dose and regimen chosen for the INTREPAD trial was actually doing to CNS immune functions is unknown. Unfortunately, like most other “immunomodulatory trials,” whether with antibodies or NSAIDs, to date the INTREPAD trial doesn’t offer us information about innate immune cell target engagement, differentiation status, and downstream effects on soluble mediators. Further, the TSPO PET probes we have now are relatively poorly refined indices of innate immune cell “activation” and don’t discriminate the more specific functional status of “activated” innate immune cells. We need better feedback, probably from the CSF. Beyond “don’t do this again” to learn from our failures, this trial and the field in general need more specific biomarkers that reveal the effects of immunomodulatory interventions on specific immune functions, in particular CNS innate immune functions. 

    A few other comments. The INTREPAD trial shows that two years of naproxen failed to impact cognitive decline. While this seems to run counter to the epidemiology, consider the results of one of the big Veterans Administration data sets from Vlad et al., 2008, with 49,349 cases and 196,850 controls. In Figure 2 of that paper, you see little or no effect of naproxen exposure of one to three years on the odds ratio for dementia, and the protection is only emerging at three to four and five years of exposure, most clearly with ibuprofen.

    This and other observational data would argue that an NSAID intervention trial should probably use ibuprofen, run for five years, and be powered based on the odds ratio of 0.73. The INTREPAD trial design expected that more consistent naproxen exposure in the trial might improve upon the epidemiology data. On the other hand, Breitner and Larson have also published data showing that consistent heavy NSAID use was not protective (Breitner et al., 2009), and Sonnen and colleagues have even reported increases in amyloid with heavy NSAID use consistent with a suppression of amyloid clearance (Sonnen et al., 2010). 

    Based on biomarker data from ADAPT, the INTREPAD trial hoped to see an impact on CSF tau/Aβ42 but didn’t. While this negative finding with bona fide AD diagnostic CSF biomarkers is disappointing, at a mechanistic level we don’t really know that heavy naproxen use should have been predicted to increase CSF Aβ42 or reduce CSF tau.


    . Prostaglandins and inflammation. Arterioscler Thromb Vasc Biol. 2011 May;31(5):986-1000. PubMed.

    . Curcumin restores innate immune Alzheimer's disease risk gene expression to ameliorate Alzheimer pathogenesis. Neurobiol Dis. 2019 Apr 2;127:432-448. PubMed.

    . Protective effects of NSAIDs on the development of Alzheimer disease. Neurology. 2008 May 6;70(19):1672-7. PubMed.

    . Risk of dementia and AD with prior exposure to NSAIDs in an elderly community-based cohort. Neurology. 2009 Jun 2;72(22):1899-905. PubMed.

    . Nonsteroidal anti-inflammatory drugs are associated with increased neuritic plaques. Neurology. 2010 Sep 28;75(13):1203-10. PubMed.

  3. The observation that NSAIDs do not significantly slow disease progression when administered to asymptomatic people in their 60s at high risk of Alzheimer’s disease does not debunk any of the potential benefits mediated by reducing neuroinflammation over the course of an entire lifetime. I tend to think of neurodegenerative disease like a snowball rolling down a hill. By the time the snowball nears the bottom of the hill it has become a very large mass with a great deal of momentum and thus, the energy required to counteract this mass must be very powerful indeed. On the contrary, a much smaller effort is required to slow the snowball when it is just forming. Such an effort can be mediated at this point in time by removing snow from its path so this cannot be added to the mass of the newly forming snowball and/or by raising the ambient temperature slightly so the snowball melts as fast as it forms. In short, we need to start thinking about neurodegenerative disease prevention much differently if we are going to achieve any clinically meaningful impacts on age at onset.

  4. One weakness of epidemiological evidence is the association does not permit cause-effect relationships to be determined. There are physiological reasons some individuals elect to take NSAIDs and others do not. Instead of the NSAIDs themselves providing the protection, their use may identify a subpopulation that has a physiological trait which both increases the need for NSAIDs, but protects against Alzheimer's disease, as Dr. Breitner alludes to in his comments above. Were this the case, to the extent the NSAIDs entered the brain, they might have the opposite effects on cognition to those desired, especially after random assignment rather than self-selection.

  5. This is clearly an important study. However, prevention in an epidemiology study is likely delay of initial pathology, whereas prevention in this type of study would likely be delay of symptoms arising from existing pathology. Thus, this negative NSAID finding does not necessarily contradict the earlier epidemiology.

  6. We understand Dr. Rebeck’s comment and his (and editorialists Drs. Hershey’s and Lipton’s) evident desire to consider all possibilities before abandoning NSAIDs altogether for prevention of AD. We offer two caveats, however: 1) We question whether it’s useful to speculate about something that, with today’s technology at least, can’t be measured. We designed INTREPAD to “look as far upstream” as was reasonably practical in human trials—even going to the point of developing a novel endpoint to reveal early changes with greater power than any single known endpoint. As Dr. Morgan (and many before) comment, one cannot rely on observational data to infer causality. 2) In relation to the last, in the cited study (Breitner et al., 2009) we had Group Health Cooperative pharmacy-dispensing records going back more than 20 years in many instances, and more than 30 in a fair few. That paper, which reported an odds ratio of >1.5 (highly significant) for incident AD with prior heavy NSAID use, included data going back to the years when NSAIDs achieved widespread usage. An exposure history of >10 years was required for consideration in those analyses. Furthermore, a sensitivity analysis for that work was described in its “Results” section thus:

    “Splitting heavy users into distant and recent users yielded (adjusted hazard ratios) for all-cause dementia of 1.51 (95% CI, 1.03–2.20) for distant users and 1.85 (95% CI, 1.28–2.66) for recent users. Results for probable AD, and for all other sensitivity analyses, were similar …”

    This, along with the persistent (and, we suggest, now irrefutable) evidence that the risks of NSAID use outweigh their potential for benefit, constitute much of the reason for the “extreme pessimism” described in this new report.


    . Risk of dementia and AD with prior exposure to NSAIDs in an elderly community-based cohort. Neurology. 2009 Jun 2;72(22):1899-905. PubMed.

  7. While this study demonstrated the lack of benefit of a nonselective NSAID for chronic use, targeting inflammation as a contributor to cognitive loss in the context of accumulation of amyloid remains a viable goal. The ability to slow cognitive decline with a more selective inhibition of the detrimental inflammation in AD while preserving repair mechanisms remains consistent with multiple reports in human populations. New, highly specific drugs to prevent toxic inflammation and promote the appropriate homeostatic responses to injury are still to be tested.

    Furthermore, with more precise screening for trial participants, the potential for detecting a successful therapeutic will be enhanced. With each completed study we learn, our ideas evolve, and our conceptual framework for the next trial is refined.

  8. How about we just close the book on naproxen? NSAIDs is a broad class of drugs, and it hardly seems appropriate to castigate the whole lot of them based on INTREPAD and ADAPT. If one were to simply follow the strongest epidemiological signal, then testing ibuprofen (not other NSAIDs) for its ability to prevent or delay AD onset (not progression) could still be construed as reasonable, to the degree that patients and researchers have the stomach for it.

Make a Comment

To make a comment you must login or register.


Therapeutics Citations

  1. Naproxen
  2. Ibuprofen
  3. Flurizan™

Paper Citations

  1. . Extended results of the Alzheimer's disease anti-inflammatory prevention trial. Alzheimers Dement. 2011 Jul;7(4):402-11. PubMed.
  2. . Meta-analysis of nonsteroidal antiinflammatory drug use and risk of dementia. Am J Epidemiol. 2005 Jan 15;161(2):114-20. PubMed.

Further Reading


  1. . Effects of rofecoxib or naproxen vs placebo on Alzheimer disease progression: a randomized controlled trial. JAMA. 2003 Jun 4;289(21):2819-26. PubMed.
  2. . Results of a follow-up study to the randomized Alzheimer's Disease Anti-inflammatory Prevention Trial (ADAPT). Alzheimers Dement. 2013 Nov;9(6):714-23. Epub 2013 Apr 3 PubMed.

External Links

  1. Naproxen

Primary Papers

  1. . INTREPAD: A randomized trial of naproxen to slow progress of presymptomatic Alzheimer disease. Neurology. 2019 Apr 5; PubMed.
  2. . Naproxen for presymptomatic Alzheimer disease: Is this the end, or shall we try again?. Neurology. 2019 Apr 5; PubMed.