Metabolic & Cardiometabolic
PCSK9 inhibition does not impair cognition
In plain terms: Are PCSK9 inhibitors safe for memory and thinking?
Strong support Metabolic & Cardiometabolic 💰 Industry COI noted🔬 Includes disconfirming
Part of: 💊 PCSK9 inhibitors
RefutedContestedStrong support
consensus score 0.88
Yes, on the evidence so far. The worry that driving LDL cholesterol very low with PCSK9 inhibitors might harm the brain has not held up: dedicated trials tracking memory and cognition found no meaningful decline. Very low LDL from these drugs looks cognitively safe.
📅 Last reviewed: 2026-07-14 ⓘ
Evidence ladder
How far up the ladder this claim has climbed. A high consensus on a low rung means "consistent so far," not "proven in people."
Top evidence so far: All trials, pooled (Meta-analysis)
MechanismIn-vitroAnimalObservationalRCTMeta-analysis
How the studies fall
7 support 1 contradict 0 tested null 0 mixed · 8 sources, 8 independent groups
The evidence (11)
| Source | Grade | Stance | Quality | Finding |
|---|---|---|---|---|
| Zimerman 2025 (FOURIER-OLE) 2025 · NEJM Evid | observational # was RCT — FOURIER-OLE extension: "This prospective study", no randomised contrast, all on evolocumab (reread #169, precedent k) | supports | moderate | FOURIER-OLE cognitive follow-up — the long-term open-label extension of the EBBINGHAUS substudy, not a new cohort: 'A total of 473 patients out of the 1974 patients in the parent EBBINGHAUS study were enrolled and additionally followed for a median of 5.1 years (maximum follow-up since original random assignment 7.2 years).' Median LDL cholesterol was 35 mg/dl (IQR 21-55) at 12 weeks of the extension, and annual CANTAB testing found no loss of executive function: 'Treatment with evolocumab was not associated with a change in executive function during the open-label extension in either patients who were originally randomly assigned to and continued evolocumab (mean±standard deviation of 0.1±2.8, P=0.49) or patients originally randomly assigned to placebo who then started on evolocumab (-0.1±2.5, P=0.64)', and 'At the final study visit, executive function scores were similar between randomly assigned groups (17.5±3.7 and 17.3±3.7, respectively)'. Conclusion: 'Exposure to very low levels of LDL cholesterol, achieved via PCSK9 inhibition and statin therapy, was not associated with cognitive impairment through long-term follow-up.' DUPLICATE POPULATION on this claim, hence zero-weighted: every patient here is a strict subset of the 1974 EBBINGHAUS randomisees already counted through s28813214 (the cognitive anchor) and of the 27,564 FOURIER patients counted through s32381158 — zero independent enrolment, which is the CONVENTIONS §5 open-label-extension/rollover archetype. The zero weight is about not re-counting the same bodies, NOT a discount of the finding: this is the only multi-year (up to 7.2 y) objective cognitive testing on the claim, and it is retained here in full for exactly that reason. Quality lowered high→moderate: during the extension there is no concurrent control (everyone received evolocumab, so the primary endpoint is a within-arm change from baseline and the surviving randomised contrast is longer- vs shorter-exposure), only 473 of 1974 parent-substudy patients (24%) enrolled with no enrolment-flow given, the design is open-label, and the cohort is low-dementia-risk — 'The median age was 62 years; 70% were male, and 91% were White' — with the authors declining to generalize: 'Further studies are needed to assess the generalizability to adults at higher risk of dementia.' No funding statement is readable, but the byline carries two 'Global Development, Amgen' coauthors and one from Cambridge Cognition, vendor of the CANTAB instrument that defines the endpoint. Corrects the prior extract's '~8 yr', which the paper does not say. Text basis: PubMed abstract only (3,498 bytes); NEJM Evid full text and supplement are subscription-only. |
| Giugliano RP et al (FOURIER Investigators) 2017 · Lancet. 2017 Oct 28;390(10106):1962-1971 (Epub 2017 Aug 28) | RCT | supports | high | Prespecified SECONDARY analysis of FOURIER (NCT01764633), stratified by ACHIEVED LDL-C at 4 weeks rather than by randomised assignment: 'In this prespecified secondary analysis of 25 982 patients from the randomised FOURIER trial ... ten prespecified safety events of interest was examined over a median of 2.2 years of follow-up', with '2669 (10%) of 25 982 patients' reaching LDL-C <0.5 mmol/L, and 'no significant association was observed between achieved LDL cholesterol and safety outcomes, either for all serious adverse events or any of the other nine prespecified safety events'. DUPLICATE POPULATION on this claim, hence zero-weighted: these 25,982 are 94% of the same 27,564 FOURIER patients, so they strictly contain all 1,204 patients of the EBBINGHAUS cognitive substudy (s28813214, the group's anchor, which measured cognition prospectively with the CANTAB battery against a pre-specified randomised noninferiority endpoint) and overlap the whole-cohort FOURIER cognitive-event analysis (s32381158); this paper adds no independent cognitive information over them. Read honestly: the abstract never enumerates the ten safety events, so the readable text does not itself confirm that neurocognitive events were among them and reports no cognitive event counts - the cognition-specific reading rests on the cohort's other publications, not on this one. Also note the safety contrast is post-randomisation and confounder-adjusted ('We used multivariable modelling to adjust for baseline factors associated with achieved LDL cholesterol'), and follow-up is a median of 2.2 years. Text basis: PubMed abstract only; Lancet full text paywalled. |
| Giugliano 2017 (EBBINGHAUS) 2017 · N Engl J Med | RCT | supports | high | EBBINGHAUS, the dedicated prospective cognitive substudy of FOURIER (n=1204 of 27,564, median 19 months, CANTAB battery, pre-specified NONINFERIORITY primary endpoint with margin = 20% of the placebo SD): change in the spatial working memory strategy index of executive function was '-0.21+/-2.62 in the evolocumab group and -0.29+/-2.81 in the placebo group (P<0.001 for noninferiority; P=0.85 for superiority)', with no significant between-group differences in working memory, episodic memory or psychomotor speed, and 'no associations between LDL cholesterol levels and cognitive changes' in an exploratory analysis. Strongest instrument on this claim: cognition was measured, not adverse-event-reported. Limits kept visible: funded by Amgen with five Amgen-employee coauthors and a coauthor from Cambridge Cognition (the vendor of the endpoint instrument); 19 months cannot address multi-year neurodegeneration; and the primary result is bounded noninferiority, not a demonstrated null. |
| Gencer 2020 · J Am Coll Cardiol | RCT | supports | high | FOURIER whole-cohort patient-reported cognition: 22,655 patients completed a 23-item Everyday Cognition (ECog) self-survey at the final visit after a median 2.2 years. Self-reported cognitive decline (ECog >=2) was similar on placebo vs evolocumab -- 'both for total score 3.6% versus 3.7% (p = 0.62) and for subdomains (memory,' '5.8% vs. 6.0%; total executive, 3.6% vs. 3.7%).' Decline was also similar in the 2,338 patients achieving LDL-C <20 mg/dl vs the 3,613 'patients with LDL-C >=100 mg/dl (3.8% vs. 4.5%, p = 0.57).' Evolocumab 'had no impact on patient-reported cognition after an average of 2.2 years of' treatment. NOTE: this analysis is self-report only -- adjudicated cognitive events and objective neurocognitive testing (EBBINGHAUS, n=1,204) are reported elsewhere, not here. |
| Hirsh Raccah 2021 · Int J Cardiol | meta-analysis | supports | high | Meta-analysis of 21 RCTs (59,733 patients, 31,611 on a PCSK9 inhibitor) found no excess reported neurocognitive adverse events: 'No significant difference in the incidence of neurocognitive adverse effects between the groups was identified (RR = 1.01, 95% CI: 0.86-1.19, I2 = 3%).' Outcome is reported adverse events, not administered cognitive testing; the meta-regression concerned study DURATION for evolocumab, not achieved LDL-C. |
| Benn 2017 · BMJ | observational | supports | moderate | Drug-target MR in 111,194 Danes (1001 Alzheimer's, 2154 any dementia, 460 Parkinson's): risks 'were not statistically significantly different among participants scores of 2-4 PCSK9 LDL cholesterol lowering alleles ... compared with participants with scores of -1 or 0' — genetics against harm. Quality moderate, not high: the authors note 'the power to exclude a genetic association is limited for some of the endpoints' (AD RR 0.57, 0.27-1.17; PD RR 1.02, 0.26-4.00), the PCSK9/HMGCR-specific two-sample estimates disagreed by method (IVW 2.04, Egger 0.24, weighted-median 4.66), and they caution that 'the present results should not be used to predict effects of specific lipid lowering drugs.' |
| Harvey 2018 · Eur Heart J | meta-analysis | supports | moderate | Sponsor-pooled IPD from 14 alirocumab Phase 2/3 RCTs (alirocumab n=3340 vs control n=1894; 4029 patient-years): neurocognitive TEAEs 22 (0.9%) vs 9 (0.7%) placebo, HR 1.24 (95% CI 0.57-2.68), and 10 (1.2%) vs 8 (1.3%) ezetimibe, HR 0.81 (95% CI 0.32-2.08); no excess in the 839 patients reaching two consecutive LDL-C <25 mg/dL (HR 0.39, 95% CI 0.14-1.10). A real randomized null, but underpowered on its own terms: the authors write that 'the overall number of neurocognitive events ... is too small to draw definitive conclusions' and the 'CIs for the HRs are wide and do not exclude HRs >2', the largest constituent trial (LONG TERM) ran HR 2.28 (95% CI 0.77-6.75), 'No formal neurocognitive testing was included in trials' (patient self-report only, no baseline screening), longest follow-up 2 years, and the analysis was 'funded by Sanofi and Regeneron Pharmaceuticals, Inc.' with sponsor-employee co-authors. |
| Schwartz GG, Steg PG, Szarek M, Bhatt DL, ODYSSEY OUTCOMES Investigators 2018 · N Engl J Med 2018 Nov 29;379(22):2097-2107 | RCT | supports | moderate | COGNITION IS NOT AN OUTCOME IN THE AVAILABLE TEXT. The trial's only safety sentence is aggregate: 'The incidence of adverse events was similar in the two groups, with the exception of local injection-site reactions (3.8% in the alirocumab group vs. 2.1% in the placebo group).' No neurocognitive event count, rate, hazard ratio or comparison appears anywhere in the abstract, and PMID 30403574 has no PMC full text on this box, so both halves of the former extract - 'neurocognitive adverse events were rare and balanced' and 'very low achieved LDL' - are UNVERIFIED. A general adverse-event similarity is not a cognition-specific test. Marked off-scope pending the full-text adverse-event table; restore as supports (RCT, n=18,924, median 2.8 y, double-blind) the moment that table is read and shown to carry a neurocognitive row with its numbers quoted here. |
| Li 2022 · Heart | meta-analysis | supports | high | Systematic review and random-effects meta-analysis of 32 RCTs / 65 861 participants (median follow-up 40 weeks): 'PCSK9 inhibitors do not increase the risk of new-onset diabetes mellitus, neurocognitive events, cataracts or gastrointestinal haemorrhage with high certainty evidence.' The only harm identified was severe injection-site reaction (15 events per 1000 persons over 5 years). Read from abstract only; the neurocognitive-specific risk ratio is not recoverable without the paywalled full text. |
| Williams 2020 2020 · Ann Neurol | observational | contradicts | moderate | Drug-target Mendelian randomization, 24,718 AD cases / 56,685 controls (IGAP + PGC): 'genetically instrumented exposure to PCSK9 inhibitors was predicted to increase AD risk in both of the AD samples (combined odds ratio per standard deviation lower LDL-C inducible by the drug target = 1.45, 95% confidence interval = 1.23-1.69)', and the authors conclude 'pharmacovigilance for AD risk among users of PCSK9 inhibitors may be warranted'. The post hoc circulating-PCSK9 model points the same way but is underpowered (AD meta IVW OR 1.19, 95% CI 0.94-1.51, 3 SNPs). No analysis in this paper is protective — the prior extract's 'if anything protective' has no basis in the text. Disconfirming but INDIRECT: lifelong genetic proxy, not drug exposure; conservative-clumping CI crosses 1 (OR 1.44, 0.94-2.20) and the most functionally relevant variant (rs11591147) was individually non-significant, so the authors state it 'in no way confirm[s]' harm. |
| Nordestgaard 2025 · Alzheimers Dement | observational | supports | moderate | [FT-verified 2026-08-19] CORRECTION to prior extract, which said PCSK9 was 'associated with LOWER dementia risk' - the paper's own MR results are NULL, not lower. One-sample MR OR 0.97 (0.70-1.35) for all-cause dementia; two-sample MR OR 1.03 (0.90-1.17); authors: 'In one-sample and two-sample Mendelian randomization, we did not find evidence of an effect of PCSK9 variants on risk of all-cause dementia, but results from Cox regression showed a protective effect' (Cox HR 0.35, 0.23-0.54). Highlights concede 'An effect of PCSK9, ANGPTL4, and LPL variants on dementia risk cannot be excluded.' This is nonetheless an INFORMATIVE null for a no-harm claim: n=1,091,775, three strong instruments (F=238/428/661), and the IHD positive control WORKED for PCSK9 (OR 0.61 and 0.63, both significant), so the dementia null is not a power failure. No harm signal in any arm; upper CI 1.35/1.17. Supports 'does not impair cognition'. Note COI: senior author reports consultancies for Sanofi, Amgen and Novartis (PCSK9-lowering manufacturers). |
Disagree, or know a study we missed?
We grade by evidence, not opinions. The way to weigh in is to point us to a study we haven't cited (check the evidence table above first), or to flag a problem with one we have. Every submission is reviewed; if it holds up, the grade updates and shows in Science Changes Its Mind.
Opens a short form. You'll sign in with Google so submissions are tied to a real account — we don't display your identity, and we only accept a link we can verify (PubMed, DOI, ClinicalTrials.gov).
Educational only, not medical advice. Grades and scores reflect published evidence weighted by study design and quality; see the methodology.