Neurology · Medical Literature
Semaglutide showed limited improvements in patients with Alzheimer's disease: Revisiting the evoke and evoke + clinical trials.
Abstract
BackgroundSemaglutide is a glucagon-like peptide-1 analog that is on the market to treat type 2 diabetes and weight loss (Ozempic, Wegovy). Two phase 3 clinical trials have been conducted, Evoke and Evoke+, testing the drug in patients with Alzheimer's disease. The trial management presented results of the intermediate readout at week 104 of the CDR-SB scores, which were negative. On the basis of that, the management decided to declare the trials a failure. However, data from week 130 and 156 had not been statistically analyzed.ObjectiveWhen evaluating time points 130 and 156, several results show a separation between drug group and placebo group with semaglutide showing better results.MethodsUsing the means, converting the SEMs to SDs and numbers of patients per group, I analyzed the results using the Welch T-test (two-tailed), which does not assume equal SD.ResultsThe ADCS-ADL-MCI test, Evoke trial, week 130, did show a significant difference, p = 0.0039. Other test such as the ADAS-cog-13 results show trends towards improvement by semaglutide at week 156. Cerebrospinal fluid biomarker analyses showed significant differences in some AD markers, too.ConclusionsThe results did show some limited drug effects at later time points of the trials. However, Semaglutide has been designed to stay in the blood for a long time and therefore does not cross the blood-brain barrier readily. Novel GLP-1 type drugs that can cross the blood-brain barrier easily may show superior protective effects in AD patients.
Clinical relevance
Do the week-130 and week-156 data from the Evoke and Evoke+ phase 3 trials — time points not subjected to formal pre-specified statistical analysis by the sponsor — provide evidence of clinical or biomarker benefit of oral semaglutide over placebo in patients with early-stage symptomatic Alzheimer's disease? __SNAPSHOT__ {"studyDesign":"Post-hoc re-analysis of pooled data from two phase 3, randomized, double-blind, placebo-controlled trials (Evoke [NCT04777396] and Evoke+ [NCT04777409]); author-performed Welch t-tests (two-tailed) applied to previously reported group means and SEMs converted to SDs","setting":"566 sites across 40 countries (multinational); analysis conducted and published independently by a single academic author","sampleSize":"Evoke: n=1,855; Evoke+: n=1,953 (total enrolled ~3,808); exact subgroup n per time point used in this re-analysis not reported in abstract","population":"Adults aged 55–85 years with amyloid-confirmed early-stage symptomatic Alzheimer's disease (mild cognitive impairment or mild dementia due to AD); evoke+ additionally included participants with small vessel pathology","intervention":"Oral semaglutide up to 14 mg once daily (flexible dosing) added to standard of care","comparator":"Matched oral placebo","primaryOutcome":"The pre-specified primary endpoint of the original trials was change in CDR-SB score from baseline to week 104; the current re-analysis focuses on the ADCS-ADL-MCI score at week 130, ADAS-cog-13 trends at week 156, and CSF AD biomarker changes — none of which were pre-specified primary endpoints","followUp":"156 weeks (3 years) planned in the original trials; the sponsor's pre-specified primary analysis was at week 104. The 1-year extension (weeks 104–156) was discontinued by Novo Nordisk after the week-104 primary analysis was negative; data from weeks 130 and 156 were available but not formally analyzed by the sponsor","fundingSource":"Original trials funded by Novo Nordisk; this re-analysis appears independently authored with no stated external funding (not reported in abstract)"}
Key findings
The pre-specified primary endpoint of both Evoke and Evoke+ — change in CDR-SB from baseline to week 104 — was negative: mean CDR-SB changes were 2.3 (SE 0.1) with semaglutide vs. 2.3 (0.1) with placebo in Evoke (estimated difference −0.08, 95% CI −0.35 to 0.20, p=0.57), and 2.2 (0.1) vs. 2.1 (0.1) in Evoke+ (estimated difference 0.10, 95% CI −0.17 to 0.38, p=0.46). The ADCS-ADL-MCI treatment differences at the pre-specified week-104 time point were also non-significant (Evoke: 0.43; Evoke+: −0.21). In this post-hoc re-analysis by Hölscher, applying Welch t-tests to reported group means with SEMs converted to SDs, the ADCS-ADL-MCI score in the Evoke trial at week 130 reached nominal statistical significance (p=0.0039). The ADAS-cog-13 showed trends favoring semaglutide at week 156, but statistical significance for this endpoint is not reported in the abstract. In a CSF biomarker substudy (n=199; 98 semaglutide, 101 placebo), nominally significant reductions of approximately 10% were observed for pTau181, pTau217, npTau181, and npTau205, as well as other neuroinflammation markers, in the semaglutide group; however, these were described as statistically significant but not large enough to have clinical impact. Peripheral inflammation markers (e.g., CRP) were reduced approximately 30% with semaglutide but did not translate into cognitive or functional benefit. No new safety signals were identified; treatment-emergent adverse events were reported in 91.2% of semaglutide recipients vs. 84.8% of placebo recipients in the original trials, with gastrointestinal disorders being most common. NNT/NNH are not calculable from the abstract data alone.
Practice implications
This paper is a single-author post-hoc re-analysis of data from trials whose sponsor declared failure based on a pre-specified primary analysis at week 104. The re-analysis identifies one nominally significant finding (ADCS-ADL-MCI at week 130, Evoke trial only, p=0.0039) and biomarker trends at week 156 that were not formally analyzed by the trial team. Several critical limitations must be acknowledged before interpreting these results: (1) The week-104 primary endpoint was definitively negative in both trials, with near-zero estimated treatment differences on the CDR-SB; (2) The week-130 and week-156 analyses are unplanned, exploratory, and not corrected for multiple comparisons, making the reported p-values highly susceptible to type I error inflation; (3) The analytical method (converting publicly reported SEMs to SDs and applying Welch t-tests to aggregate means) is not equivalent to an individual-patient-data analysis and may introduce additional error; (4) The extension period was discontinued by Novo Nordisk after the negative week-104 readout, raising questions about sample size, completeness of data, and potential informative censoring at later time points; (5) The author is a known proponent of GLP-1 therapy for neurodegeneration, which may introduce advocacy bias in interpretation. The population studied — amyloid-confirmed, symptomatic early AD aged 55–85 — is representative of the target clinical population, but the findings from this re-analysis do not apply to prevention or asymptomatic at-risk groups. The CSF biomarker improvements are mechanistically interesting but have not been shown to correlate with clinical outcomes in this population. Practice should not change on the basis of this analysis. __LIMITATIONS__ Post-hoc, unplanned re-analysis of data from trials whose pre-specified primary endpoint (CDR-SB at week 104) was clearly and consistently negative in both Evoke and Evoke+.,No correction for multiple comparisons across many time points and outcome measures, substantially elevating false-positive risk for the reported p=0.0039 finding.,Analysis relies on publicly reported aggregate means and SEMs rather than individual patient data; conversion of SEM to SD and application of Welch t-tests to group-level summaries may introduce analytical inaccuracies.,The 1-year extension (weeks 104–156) was prematurely discontinued by Novo Nordisk; data completeness, dropout patterns, and potential informative censoring at weeks 130 and 156 are unknown.,Single-author analysis without independent peer validation of statistical methods or raw data access.,Potential for advocacy bias: the author has an established academic and commercial interest in GLP-1 therapies for neurodegeneration.,The significant ADCS-ADL-MCI finding was present only in Evoke, not replicated in Evoke+, suggesting a chance finding.,CSF biomarker improvements (~10% reductions in tau species) were characterized by independent experts as statistically significant but not clinically meaningful in magnitude.,Semaglutide has limited CNS penetration, predominantly targeting peripheral inflammation; the biological plausibility of clinical benefit via this mechanism remains unestablished in humans.,Only the abstract was available for this review; the full statistical methods, exact n per group at each time point, and any sensitivity analyses cannot be evaluated. __BOTTOM_LINE_CATEGORY__ low_certainty __BOTTOM_LINE__ The pre-specified primary endpoints of both Evoke and Evoke+ were negative: oral semaglutide did not slow clinical progression in early Alzheimer's disease at week 104, and Novo Nordisk appropriately declared the trials a failure. This post-hoc, exploratory re-analysis by a single author identifies a nominally significant ADCS-ADL-MCI difference at week 130 in one trial and biomarker trends at week 156, but these findings carry very high risk of type I error, are not replicated across both trials, and are insufficient to alter clinical practice or justify off-label semaglutide use for Alzheimer's disease.