Once again, clinical trial data suggest that non-steroidal anti-inflammatories (NSAIDs) are non-starters when it comes to preventing dementia, and once again the data seem at odds with observational studies. In the May 12 Archives of Neurology online, Barbara Martin and colleagues from the Alzheimer’s Disease Anti-inflammatory Prevention Trial (ADAPT) Research Group report that neither celecoxib nor naproxen prevents cognitive decline. In fact, naproxen may even hasten it slightly. In contrast, data from one of the largest observational studies of NSAIDs and dementia to date suggest that the anti-inflammatories, particularly ibuprofen, do reduce the incidence of dementia. Those results, published in the May 6 Neurology, were based on Veterans’ Affairs (VA) medical records of just under 250,000 patients, 20 percent of whom went on to develop Alzheimer disease. The two studies typify the back-and-forth between observation and trial data—not only for NSAIDs but other potential treatments for AD, such as estrogen and statins.

The ADAPT report follows early data published last year indicating that the two used anti-inflammatories had no effect on the incidence of dementia (The ADAPT Research Group, 2007). “As you would expect, the results are fairly concordant. But the [new] cognitive function results do allow us to look a little more broadly in the population, not just at the incidence of developing dementia but what is happening over time with cognition,” said Martin, who is deputy director of the ADAPT coordinating center at the Johns Hopkins Bloomberg School of Public Health, Baltimore, and primary author on the paper. That closer look suggests that by a variety of cognitive measures these anti-inflammatories have no benefit.

Martin and colleagues measured cognitive function with a battery of seven different tests, including verbal and visuospatial learning and memory trials, in normal older adults aged 70 and up who had a family history of AD-like dementia. Not surprisingly, scores in all seven measures and a global summary score worsened for those on placebo. The bad news was that participants on the anti-inflammatories fared no better. Scores for those taking celecoxib (200 mg twice daily) and naproxen sodium (220 mg twice daily) also worsened from baseline out to three years. In all cases the declines were small and not statistically significant. The greatest decline was in the 3MS-E, a modified Mini-Mental State Examination. In that test, the decline in both treatment groups was worse than in those taking placebo. In fact, for those taking naproxen, declines across all seven scales were worse than in those taking no drug. “The findings indicate that celecoxib and naproxen did not have a positive effect on cognitive function, and that, if anything, there is some weak evidence of a detrimental effect with naproxen,” said Martin.

The results support earlier findings from the Alzheimer's Disease Cooperative Study (ADCS), which found that neither naproxen nor rofecoxib slowed cognitive decline in those with mild to moderate AD. That study also suggested that rofecoxib may, in fact, make things worse (see ARF related news story and Aisen et al., 2008).

Results from both of these trials contrast with observational data from such cohorts as the Rotterdam study (see ARF related news story and also meta-analysis of epidemiological data) and now the VA study. Researchers led by Steven Vlad at Boston University looked through publicly available VA records to correlate AD incidence with NSAID use. Because of the size of the study (49,349 AD cases and 196,850 controls), the researchers had enough power to look at individual drug use rather than NSAIDs as an entity. They were also able to parse the data by duration. The results indicate that taking any NSAID for more than four years reduced the odds of getting AD to 0.76, while between one and four years’ duration reduced the odds to around 0.9. Less than one year on the anti-inflammatories had little effect.

Looking at individual drugs, Vlad and colleagues found that ibuprofen had the most robust effect, reducing the odds ratio for AD by almost half (0.56, 95 percent confidence interval 0.42-0.75). The results for the other drugs were inconsistent and not statistically robust. Naproxen, for example, seemed to reduce the incidence of AD when taken for three to four years (odds ratio of 0.79; 0.66-0.95 95 percent CI), but over five years or more the confidence interval expanded to between 0.55 and 1.11.

It is not clear why the observation and the trial data are at odds. Vlad suggests it might be because the trials are not exactly looking at the same thing. “The big epidemiology studies that have been done so far that show an effect all looked at NSAIDs as a group because there wasn’t enough detail to look at individual drugs, whereas the trials all look at individual drugs,” he said in an interview with ARF. He added that the drugs that have been tested in random controlled trials may not be the best ones to use because only some NSAIDs, in addition to being anti-inflammatories, also lower levels of amyloid-β (see review of mechanisms by Kukar and Golde, 2008). Indeed, discussion about whether ibuprofen could have been included in the ADAPT trial arose even while the trial was still underway (see ARF related news story). “Trials used naproxen, which might have only a weak Aβ42-lowering effect, and rofecoxib, which we don’t think has any effect, and other similar drugs,” said Vlad. He noted that one small trial used indomethacin, which is an Aβ-lowering NSAID. The VA data suggests a trend toward protection with that drug, though it is not frequently used in the VA system.

Paul Aisen is not convinced. He was one of the principal investigators in the ADCS trial, and now heads the ADCS at the University of California, San Diego. “This [VA study] is an example of a study that shows an association between NSAID use and reduced risk of dementia, but no difference between those that lower Aβ and those that do not,” he told ARF. “Keep in mind that the VA studies suggest this benefit from huge numbers, but are generally limited by poor reliability of data,” he added.

Martin agreed that other subtleties, such as choice of drug, time, or duration of treatment, may be at play. “In some observational studies that have suggested an effect with longer term treatment, two years is the cutoff point for when an effect shows up,” she said. The ADAPT trial was suspended in December 2004 for safety reasons (see ARF related news story), which means that only about half of the volunteers had been on the drugs for two years or more.

But Martin offered a different explanation, as well. “It is possible that the trial data are giving us a better picture of the effect of anti-inflammatories, making sure that it is not due to some confounding effect that might be present in the observational studies,” she said. Aisen agreed. “My own feeling is that the epidemiological data is flawed by bias,” he told ARF. “Those biases are shared by the estrogen study, where we had the same discrepancy. The epidemiological studies strongly indicated a protective effect and yet all of the clinical trials have been negative, even leaning toward an adverse effect,” he said. “One could say treat earlier, or longer, or you could say that we didn’t choose the right dose or drug, but I believe in the context of all of the studies, there is simply no evidence that anti-inflammatories are helpful for Alzheimer disease—for either treatment or prevention,” he said.

Vlad acknowledged that bias is a possible explanation but suggested that the VA study takes care of one major source of bias, called “confounding by indication.” “The idea is that physicians recognize patients who have cognitive decline or memory loss, realize they are at greater risk, and avoid prescribing them NSAIDs, which makes it look like people who used NSAIDs are protected, when it is really that physicians are more comfortable giving them the drugs,” he said. His study chose patients who were on the anti-inflammatories before they were diagnosed, effectively eliminating this potential confound.

We may never know if NSAIDs, as anti-inflammatories, are truly beneficial. Martin said that as far as she is aware no other trials to test anti-inflammatories in AD are planned. “These are large, long-term trials, and if you don’t get it right the first time and you see either no effect, or worse, some sort of detrimental effect, then it makes it harder to tinker and say ‘well, let’s try ibuprofen,’ or ‘let’s make sure we continue treatment for longer or start people earlier,’” she said.

Aisen is even less optimistic. “If you consider the results here, along with every other trial of an anti-inflammatory drug for the prevention or treatment of AD, then I believe you have to conclude that we have seen no evidence at all of any beneficial effect. I believe it is time to focus on other approaches to treatment, be it anti-amyloid or anti-tau,” he said.

Ironically, the next drug in Phase 3 due to show its colors may be the R-enantiomer of flurbiprofen, which has little NSAID activity but does appear to modulate production of amyloid-β (see ARF related news story and Wilcock et al., 2008).—Tom Fagan


  1. The ADAPT trial was a bold attempt to determine if cyclooxygenase (COX) inhibitors could prevent cognitive decline, but it was flawed from the start by the unfortunate choice of celecoxib and naproxen as the agents to test. Celecoxib should never have been chosen for this or any other AD clinical trial. There is no epidemiological evidence to support such a choice, and there are strong contraindications from basic research against its use. The same applies to all selective COX-2 inhibitors. COX-2, in contrast to COX-1, is an essential enzyme for normal physiological functioning. That was revealed when COX-2 knockout mice died prematurely from kidney failure and cardiac fibrosis (1), while COX-1 knockout mice had a normal life span. This information was known before the coxibs were introduced clinically and presaged the later withdrawal of Vioxx due to cardiac complications.

    Brain is one of the few areas in the body that constitutively expresses high levels of COX-2, again signifying an important physiological function (2). It is concentrated in the very pyramidal neurons that are vulnerable in AD (3). Despite these facts, four COX-2 inhibitor trials in AD have been carried out, all of them predictably ending in failure. Only recently has it come to light that in the rofecoxib trials, there was a 2.85-fold increase in mortality in AD patients compared to controls (4).

    The negative effects of naproxen may have been less predictable at the outset of the ADAPT trial, but there was no epidemiological evidence to favor its selection over other traditional NSAIDs where such evidence did exist (5). To date, 25 epidemiological studies have been published showing a protective effect of traditional NSAIDs against AD, and nine have been published showing a beneficial treatment effect in transgenic mouse models of the disease. For the latest published review, see (6). Better treatments for AD are urgently needed, as are better strategies for taking basic science to the bedside.


    . Renal abnormalities and an altered inflammatory response in mice lacking cyclooxygenase II. Nature. 1995 Nov 23;378(6555):406-9. PubMed.

    . Pharmacological and biochemical demonstration of the role of cyclooxygenase 2 in inflammation and pain. Proc Natl Acad Sci U S A. 1994 Dec 6;91(25):12013-7. PubMed.

    . Distribution of cyclooxygenase-1 and cyclooxygenase-2 mRNAs and proteins in human brain and peripheral organs. Brain Res. 1999 Jun 5;830(2):226-36. PubMed.

    . Reporting mortality findings in trials of rofecoxib for Alzheimer disease or cognitive impairment: a case study based on documents from rofecoxib litigation. JAMA. 2008 Apr 16;299(15):1813-7. PubMed.

    . Reduced prevalence of AD in users of NSAIDs and H2 receptor antagonists: the Cache County study. Neurology. 2000 Jun 13;54(11):2066-71. PubMed.

    . NSAIDs and Alzheimer disease: epidemiological, animal model and clinical studies. Neurobiol Aging. 2007 May;28(5):639-47. Epub 2006 May 11 PubMed.

    View all comments by Pat McGeer
  2. I wonder if exercise is a confounding variable in the epidemiological studies? People who exercise are probably more likely to take NSAIDs. In which case the protective effect could be from the exercise, not the NSAID.

Make a Comment

To make a comment you must login or register.


News Citations

  1. Naproxen/Rofecoxib Trial Results Published
  2. Large Prospective Study Finds NSAIDs Reduce Risk of Developing AD
  3. NSAIDs and AD—Prevention Better Than Cure?
  4. Trials and Tribulations: Does ADAPT Have to Adapt?
  5. Safety Concerns Halt ADAPT Trial
  6. News Brief: Flurizan Still in the Running

Paper Citations

  1. . Naproxen and celecoxib do not prevent AD in early results from a randomized controlled trial. Neurology. 2007 May 22;68(21):1800-8. Epub 2007 Apr 25 PubMed.
  2. . Rofecoxib in patients with mild cognitive impairment: further analyses of data from a randomized, double-blind, trial. Curr Alzheimer Res. 2008 Feb;5(1):73-82. PubMed.
  3. . Possible mechanisms of action of NSAIDs and related compounds that modulate gamma-secretase cleavage. Curr Top Med Chem. 2008;8(1):47-53. PubMed.
  4. . Efficacy and safety of tarenflurbil in mild to moderate Alzheimer's disease: a randomised phase II trial. Lancet Neurol. 2008 Jun;7(6):483-93. PubMed.

Further Reading

Primary Papers

  1. . Protective effects of NSAIDs on the development of Alzheimer disease. Neurology. 2008 May 6;70(19):1672-7. PubMed.
  2. . Cognitive function over time in the Alzheimer's Disease Anti-inflammatory Prevention Trial (ADAPT): results of a randomized, controlled trial of naproxen and celecoxib. Arch Neurol. 2008 Jul;65(7):896-905. Epub 2008 May 12 PubMed.