We recently published a study examining how brain lipid pathways related to inflammation are associated with cognitive decline late in life. Using postmortem brain tissue from participants in the Religious Orders Study and Rush Memory and Aging Project, we measured lipids related to calcium-dependent phospholipase A2, or cPLA2, and arachidonic acid metabolism and connected those measurements to years of annual cognitive testing.
Higher brain EPA relative to arachidonic acid was associated with better cognition and slower cognitive decline. In a complementary transcriptomic analysis, greater activation of the arachidonic acid pathway was associated with worse cognition. [1]
These findings fit with a pathway that our group has studied for years. But before interpreting them as evidence that inflammation causes cognitive decline, it is worth considering a major cardiovascular trial that recently produced a sobering result.
It illustrates one of the central problems in inflammation research:
An inflammatory pathway can be strongly associated with disease, a drug can successfully suppress that pathway, and the clinical disease may still not improve.
That lesson applies to our own work as much as it does to cardiovascular disease.
A major test of the inflammation hypothesis
For decades, higher circulating inflammatory markers have been associated with cardiovascular disease. Among the most prominent are interleukin-6, or IL-6, and C-reactive protein, or CRP.
One important test of whether inflammation was more than an association came from CANTOS.
CANTOS enrolled more than 10,000 people who had experienced a myocardial infarction and continued to have elevated hsCRP despite standard treatment. Participants received canakinumab, an antibody targeting IL-1β, or placebo.
The 150 mg dose reduced the primary cardiovascular endpoint by about 15 percent, with a hazard ratio of 0.85. This was statistically significant, but the clinical benefit was modest. All cause mortality was not reduced, and fatal infections were more common. Canakinumab was ultimately not developed as a cardiovascular prevention therapy. [2]
CANTOS was therefore important proof of concept. Manipulating an inflammatory pathway could affect cardiovascular events. But it was far from demonstrating that broadly targeting inflammation would transform cardiovascular prevention.
Then came ZEUS.
ZEUS: the biomarker moved, the disease did not
The phase 3 ZEUS trial studied ziltivekimab, an antibody targeting the IL-6 ligand.
The population seemed particularly well suited to an anti-inflammatory strategy. More than 6,300 participants had established atherosclerotic cardiovascular disease, chronic kidney disease, and evidence of systemic inflammation, defined by hsCRP of at least 2 mg/L.
The rationale was strengthened by the earlier phase 2 RESCUE trial, in which ziltivekimab produced striking reductions in hsCRP, reaching approximately 77 to 92 percent across tested doses, along with reductions in several other inflammatory biomarkers.
ZEUS asked the question that ultimately matters: would those biological effects translate into fewer cardiovascular events?
They did not.
According to headline results released in July 2026, ziltivekimab produced the expected inhibition of IL-6 signaling and reductions in free IL-6 and hsCRP. Yet the hazard ratio for cardiovascular death, nonfatal myocardial infarction, or nonfatal stroke was 0.99, with a 95 percent confidence interval of 0.88 to 1.11.
In other words, cardiovascular outcomes were essentially identical to placebo. Serious infections were also more frequent with ziltivekimab. [3]
An important caveat is that, as of this writing, these are headline results. The complete ZEUS results have not yet been published, so mechanistic and subgroup conclusions should wait for the full dataset.
Still, the central result is difficult to ignore.
The inflammatory pathway was engaged. The biomarkers fell. The cardiovascular endpoint did not.
Association, target engagement, and causation are different things
ZEUS provides an unusually clear illustration of a problem that runs through biomedical research.
High IL-6 and CRP predict cardiovascular events.
That does not necessarily mean that IL-6 itself is the process that directly causes a vulnerable plaque to rupture and form a thrombus.
IL-6 participates in a large inflammatory network. It may contribute to atherosclerosis. It may amplify other responses. But circulating IL-6 may also reflect macrophage activity, renal inflammation, adipose inflammation, infection, or other processes occurring alongside vascular disease.
A biomarker can therefore be highly predictive without being the critical therapeutic lever.
There is another important limitation to how trials such as CANTOS and ZEUS approached inflammation.
They targeted defined molecular pathways, so the drugs themselves were not nonspecific. But patient selection was relatively blunt from the perspective of tissue biology.
Participants were selected because they had cardiovascular disease and evidence of systemic inflammation, primarily elevated hsCRP. That does not demonstrate that the IL-1β or IL-6 pathway is particularly active in the coronary plaque that will later rupture. Nor does lowering CRP tell us whether the vascular wall actually became biologically more stable.
This distinction may be crucial.
The question is not simply:
Is this person inflamed?
It is:
Which inflammatory pathway is active, where is it active, what tissue process is it affecting, and is that process actually causing the clinical event?
Alzheimer disease has the same problem
We frequently describe Alzheimer disease as having a major inflammatory component.
That is true, but “neuroinflammation” is an extraordinarily broad label.
Microglia, astrocytes, endothelial cells, pericytes, neurons, infiltrating immune cells, cytokines, complement proteins, reactive oxygen species, eicosanoids, and many other pathways can all participate.
More importantly, inflammation is not inherently pathological.
An inflammatory response may be:
A driver, actively producing tissue injury.
A response, appearing because amyloid, tau, vascular injury, infection, or dying neurons have triggered it.
Or compensatory, representing an attempt to clear damaged tissue and restore homeostasis.
The same pathway might even move between these roles over time.
This is especially important in a disease that evolves over decades.
It is difficult to imagine that the brain simply enters a continuously destructive inflammatory state in middle age and remains there unchanged for 30 years.
A more plausible model is dynamic.
Inflammatory pathways may turn on in response to an insult and subsequently resolve. Repeated activation may become increasingly difficult to terminate. With aging, amyloid accumulation, vascular disease, oxidative stress, or metabolic dysfunction, an initially adaptive response may become persistent and damaging.
And once tissue injury begins, the damaged tissue itself can generate new inflammatory stimuli.
The critical issue may therefore be not merely how much inflammation exists, but whether an inflammatory response resolves appropriately.
APOE4 does not simply mean “more inflammation”
This is where APOE4 becomes particularly interesting.
APOE4 is the strongest common genetic risk factor for late-onset Alzheimer disease, but it does not inevitably lead to dementia.
If APOE4 produced lifelong chronic inflammation, one might expect cognitively healthy APOE4 carriers to have consistently greater inflammation.
Human data suggest a more complicated picture.
In the Framingham Heart Study, APOE4 carriers actually had lower circulating CRP concentrations than APOE2 and APOE3 carriers across repeated examinations (Table).
Yet when persistent systemic inflammation was present, its association with Alzheimer disease was much stronger among APOE4 carriers. [4] Fig 1
That combination is important.
APOE4 itself was not associated with higher systemic inflammation in those observations. But inflammation occurring on the background of APOE4 appeared more consequential.
Our previous brain lipidomics work pointed in a similar direction.
Cognitively unimpaired APOE4 carriers did not show the exaggerated lipid inflammatory phenotype seen in APOE4 carriers with dementia. As a matter of the fact, the opposite. APOE4 carriers who died without a diagnosis of dementia had less inflammatory lipids in their brain. In the dementia brains, however, APOE4 was associated with larger elevations in both inflammatory and pro-resolving lipid mediators. [5]
This has led us toward a different model:
APOE4 may create an inflammatory vulnerability or fingerprint rather than a state of continuously elevated inflammation.
A second insult may be required.
That insult might be amyloid accumulation, vascular injury, metabolic stress, infection, trauma, oxidative stress, or some combination.
Once activated, the APOE4 brain may have greater difficulty bringing the response back under control.
Inflammation has to resolve
Resolution is not simply what happens when pro-inflammatory molecules disappear.
It is an active biological process.
Lipid mediators participate in switching inflammatory responses off, clearing damaged material, and restoring tissue homeostasis. Lipoxins can be derived from arachidonic acid, while DHA and EPA provide substrates for several families of specialized pro-resolving mediators.
This is why the simultaneous presence of inflammatory and pro-resolving lipid mediators in Alzheimer brain tissue is so interesting.
A high pro-resolving signal does not necessarily mean that inflammation is successfully resolving. It could mean that the tissue is mounting a compensatory response that is being overwhelmed.
The biological state may therefore be better described as unresolved inflammation.
cPLA2 and arachidonic acid provide a more specific pathway
Our recent study focused on one pathway that may help us move beyond the generic label of inflammation.
cPLA2 releases arachidonic acid, or AA, from membrane phospholipids.
AA can subsequently enter multiple metabolic pathways producing prostaglandins, leukotrienes, thromboxanes, lipoxins, and other oxylipins.
Some of these products promote inflammation. Others participate in resolution.
This is why AA itself should not simply be described as a “bad” or pro-inflammatory fatty acid.
The important question is how actively the pathway is turning over and what metabolic products are being generated.
cPLA2 can be activated by intracellular calcium, MAP kinase signaling, amyloid related stress, oxidative stress, and other cellular stimuli. Experimental APOE4 models show increased activation through pathways that include p38 MAP kinase.
And this pathway has another feature that may be just as important as inflammation.
It can amplify oxidative stress.
Our new study: linking the pathway to cognitive trajectories
In our newly published study, we analyzed postmortem dorsolateral prefrontal cortex tissue from participants who had undergone annual cognitive testing during life.
Brain lipidomics were available in 191 individuals, while transcriptomic data were analyzed in 411. [1]
Rather than simply comparing brains from people with and without dementia, we asked whether these lipid pathways were associated with the trajectory of cognition over time.
One of the clearest findings involved EPA relative to AA.
Higher brain EPA/AA was associated with better global cognition close to death and a slower rate of cognitive decline. A one standard deviation higher EPA/AA was associated with better late-life cognitive performance and slower terminal decline.
We then examined transcriptomic signatures related to AA metabolism.
Greater AA pathway activation was associated with worse cognition, including lower global cognition near death and impairment across several cognitive domains.
These observations support an association between the balance of cPLA2-related lipid metabolism and cognitive decline.
But the ZEUS lesson applies directly to us.
Our associations are not proof of causation either
Our findings do not prove that increased cPLA2 activity caused cognitive decline.
cPLA2 could contribute directly to neuronal or vascular injury.
But it could also be activated because amyloid, tau, oxidative stress, vascular disease, or neurodegeneration is already occurring.
And some lipid products may represent compensatory efforts to repair that injury.
Our study therefore identifies a candidate pathway, not a proven therapeutic target.
This distinction is not a weakness. It tells us what experiment needs to come next.
Looking beyond inflammation to oxidative and vascular injury
There is an additional reason cPLA2 interests us.
Lowering cPLA2 activation may do more than decrease inflammatory signaling.
Activated cPLA2 increases AA metabolism and downstream oxylipin production. PUFA metabolism and lipid peroxidation can generate reactive oxygen species and oxidized lipid products. Oxidative stress can, in turn, activate cPLA2 further.
This creates the possibility of a feed-forward cycle:
cellular stress → cPLA2 activation → PUFA release and metabolism → oxylipins and lipid oxidation → oxidative stress → further cPLA2 activation
Experimental work shows that inhibiting cPLA2 can reduce both inflammatory signaling and oxidative stress markers.
That may be particularly important at the neurovascular unit, the integrated system of endothelial cells, pericytes, astrocytic endfeet, neurons, and extracellular matrix that maintains the blood brain barrier and regulates delivery of nutrients to brain tissue.
Our recent Perspective led by Ambreen Kanwal and Bilal Kerman proposed that cPLA2 may sit at the intersection of AA signaling, oxylipin production, oxidative stress, membrane lipid depletion, cerebral amyloid angiopathy, and blood brain barrier dysfunction (Figure 5). [6] The paper also shows activated cPLA2 in and around cerebral vessels containing amyloid, providing a tissue based rationale for considering cPLA2 as part of a neurovascular inflammatory phenotype, rather than simply a generic brain inflammatory marker.
The proposed biology therefore extends beyond:
cPLA2 → inflammation
to something closer to:
cPLA2 → abnormal lipid signaling + oxidative stress + membrane vulnerability + neurovascular dysfunction
That distinction could matter therapeutically.
Can we see the pathway in the living brain?
This brings us to one of the biggest differences between the type of precision trial we envision and previous inflammation trials.
Blood CRP can tell us that systemic inflammation exists.
It cannot tell us whether cPLA2 is abnormally active in the brain or where that activity is greatest.
Fatty acid PET imaging offers a potential way to begin answering that question.
When cPLA2 releases AA from membrane phospholipids, the released fatty acid eventually needs to be replaced. Greater cPLA2-mediated AA turnover therefore creates increased demand for AA incorporation from the circulation.
Dynamic PET imaging can measure this incorporation.
Studies using [11C]AA PET have reported higher brain AA incorporation in Alzheimer disease, and the incorporation coefficient has been used as an indirect surrogate for cPLA2 activity. [7]
Carbon-11 has a short radioactive half-life of about 20 minutes. Newer 18F-labeled AA probes, with a half-life of approximately 110 minutes, may make this approach more practical for multicenter studies.
DHA PET provides complementary information about brain omega-3 uptake and metabolism. We have reported that young cognitively normal APOE4 carriers have shown higher DHA incorporation coefficients than noncarriers, possibly reflecting greater utilization or compensatory uptake.
PET will not give us microscopic cellular resolution. An AA PET signal cannot, by itself, tell us whether cPLA2 is active specifically in an endothelial cell, pericyte, astrocyte, microglial cell, or neuron.
But it can potentially provide something that a blood inflammatory marker cannot:
regional evidence that an abnormal lipid metabolic pathway is active in the organ we intend to treat. Tissue studies can then help us understand which cells and structures contribute to that signal.
We are actively developing 18-F DHA PET tracers to measure this inflammation in the living brain.
From inflammation trials to precision inflammation trials
This may be the larger lesson.
CANTOS and ZEUS targeted specific cytokines, but patients were selected using clinical disease and circulating evidence of inflammation. They did not establish that the targeted inflammatory pathway was active in the particular arterial lesion responsible for the future event.
For Alzheimer disease, we should try to go further.
We imagine a trial in which we:
identify a biologically susceptible population, potentially including APOE4 carriers;
demonstrate evidence of excessive cPLA2-related AA metabolism in the living brain;
determine where in the brain the abnormality is occurring;
administer a selective, brain-penetrant cPLA2 inhibitor;
use PET to establish that brain pathway activity actually changed;
measure independent consequences of that intervention, including oxidative stress and neurovascular function;
and only then ask whether cognitive decline slows.
For me, this is a precision inflammation trial.
The critical difference is that we would not treat someone simply because they carry APOE4 or have an elevated systemic inflammatory biomarker.
We would attempt to show that the specific pathway we intend to inhibit is abnormal in the target organ before treatment begins.
Even that may fail.
We could lower DHA/AA PET uptake, reduce biochemical markers of pathway activity, and still discover that cognition does not improve. If so, we would have our own version of ZEUS. That is how science goes!!
Perhaps cPLA2 activation is a response to injury rather than the driver.
Perhaps we intervened too late.
Perhaps another pathway is more important.
But that would still be a far more informative than a blunt, expensive negative large trial because we would know that the drug reached the relevant organ and changed the biology we intended to change.
Take Home Messages
Association is not causation. ZEUS demonstrated that profound suppression of an inflammatory pathway and its biomarkers does not guarantee improvement in clinical outcomes.
Inflammation is not one biological state. An inflammatory signal may cause injury, respond to injury, or represent an attempt to repair it. Its role may also change over the course of disease.
APOE4 may confer an inflammatory vulnerability rather than simply causing lifelong inflammation. Human data suggest that the relevant phenotype may emerge when APOE4 is combined with additional inflammatory, metabolic, vascular, or neurodegenerative stress.
Our new study links cPLA2-related AA metabolism to cognitive trajectories, but it does not prove that the pathway causes cognitive decline. That is the hypothesis that now needs experimental testing.
cPLA2 may matter for more than inflammation. Its activation connects lipid signaling with PUFA oxidation, oxidative stress, membrane injury, and potentially dysfunction of the neurovascular unit.
Where a pathway is active matters. Blood inflammatory markers cannot tell us whether the pathway being targeted is active in the brain. AA and DHA PET imaging may provide regional metabolic information that helps bridge this gap.
Future trials should become precision trials. The ideal study would select patients with evidence that the targeted pathway is active in the relevant organ, demonstrate that treatment changes that pathway, measure downstream tissue biology, and then determine whether the clinical disease changes.
The goal is therefore not simply to develop a stronger anti-inflammatory drug.
It is to identify the right pathway, in the right patient, in the right tissue, at the right stage of disease, and then prove that changing that pathway changes the disease.
That is a much harder experiment. It is also the experiment that inflammation research increasingly needs.
Full reference
Yaskolka Meir A, Wang X, Li B, Wilson RS, Bennett DA, Louie SG, Yassine HN, Arvanitakis Z. Brain pro-resolving-to-pro-inflammatory lipids and cPLA2 activation associations with late-life cognitive trajectories. eBioMedicine. 2026;131:106421. doi:10.1016/j.ebiom.2026.106421.
Acknowledgements
We thank the ROS and MAP participants for their tireless altruism towards the pursuit of scientific knowledge about ageing human conditions.
This research was supported by National Institutes of Health Grants: R01AG076124, R01NS084965, RF1AG059621, R01AG074549, P30AG10161, P30AG072975, R01AG15819, R01AG017917, U01AG61356, and P30AG066530. A.Y.M. was supported by the Council for Higher Education-Zuckerman support program for outstanding postdoctoral female researchers.
References
Yaskolka Meir A, Wang X, Li B, et al. Brain pro-resolving-to-pro-inflammatory lipids and cPLA2 activation associations with late-life cognitive trajectories. eBioMedicine. 2026;131:106421.
Ridker PM, Everett BM, Thuren T, et al. Antiinflammatory therapy with canakinumab for atherosclerotic disease. N Engl J Med. 2017;377:1119-1131.
Novo Nordisk. Update on the ZEUS phase 3 trial in people with ASCVD, CKD and inflammation. July 31, 2026. Full scientific results pending at the time of writing.
Tao Q, Ang TFA, DeCarli C, et al. Association of chronic low-grade inflammation with risk of Alzheimer disease in ApoE4 carriers. JAMA Netw Open. 2018;1:e183597.
Ebright B, Asante I, Poblete RA, et al. Eicosanoid lipidome activation in postmortem brain tissues of individuals with APOE4 and Alzheimer’s dementia. Alzheimers Res Ther. 2022;14:152.
Kanwal A, Kerman BE, Wang S, et al. A perspective: PLA2G4A as drug target for vascular inflammation in Alzheimer’s disease. Alzheimers Dement. 2026;22:e71320.
Duro MV, Ebright B, Yassine HN. Lipids and brain inflammation in APOE4-associated dementia. Curr Opin Lipidol. 2022;33:16-24.
Disclosure: This post contains partially AI-assisted text. All references were reviewed for accuracy. I have no relevant conflicts of interest. These opinions are mine and do not reflect the opinions of my employer.
Of course, Zeus is not a done deal. I've only read the press release. But perhaps picking CKD patients with CV disease is just the wrong patient population. They may be just too sick to benefit. Or maybe the true target is IL-1.
I doubt this but timewill tell.
Noncontrast CT (Caristo, FDA approved) is a cheaper technology. It could screen Apo E4 carriers for NONSPECIFIC brain inflammation in targeted areas. You Ca+PLA2 AA Pet could then be used in a more specific fashion. At least that's my thinking.