Which patient subgroups (tau burden, APOE4 status, clinical stage) show the largest benefit from donanemab or lecanemab in trial subgroup analyses?
Executive summary
Donanemab shows its clearest, largest subgroup benefit in patients with low-to-medium baseline tau burden (the “earlier” tau stage), with subgroup analyses reporting roughly one-third slower clinical decline and higher proportions with no progression versus placebo (TRAILBLAZER‑ALZ2) [1] [2]. Lecanemab produced an overall modest CDR‑SB advantage in CLARITY‑AD but subgroup signals are weaker and underpowered; APOE ε4 status has not been shown convincingly to increase clinical benefit and instead mainly predicts higher ARIA risk for both drugs [3] [4] [5].
1. Low/medium tau burden: the clearest “who benefits most” signal
Across the trial publications and secondary analyses, the most consistent pattern is that people with lower tau PET at baseline—interpreted as an earlier pathologic stage—experience larger cognitive/functional slowing with donanemab: TRAILBLAZER‑ALZ2 reported ~35% slowing on iADRS and ~36% on CDR‑SB in the low/medium‑tau subgroup and other summaries estimate ~33% slowing in that “ideal” subgroup [1] [2] [5]. Lecanemab analyses and extension data similarly suggest greater preservation among participants with lower tau burden, though the CLARITY investigators did not power definitive tau‑by‑treatment subgroup claims [3] [4].
2. APOE4: safety signal dominates; efficacy differences small, inconsistent, underpowered
APOE ε4 carriers face a substantially higher risk of ARIA (edema and hemorrhagic findings) on both drugs—multiple meta‑analyses and trial reports document markedly increased ARIA rates in carriers, especially homozygotes [5] [6] [7]. Regarding efficacy, evidence is mixed: some reports note numerically smaller effects in APOE4 homozygotes—possibly owing to dose interruptions from ARIA—while formal subgroup analyses and Bayesian reanalyses have not established a robust APOE4‑dependent treatment effect difference and in some cases the absence of a difference was more likely [8] [9]. Regulatory labeling therefore emphasizes APOE testing primarily to inform ARIA risk rather than to select for superior efficacy [4].
3. Clinical stage: earlier (MCI/mild dementia) is the tested and favored population
Both pivotal programs enrolled people with early symptomatic AD (prodromal/MCI or mild dementia) with biomarker‑confirmed amyloid; the approvals and trial designs focus on this early stage because that is where clinical signals were measured [1] [4]. Analytical comparisons across later stages are limited—the trials were not designed to evaluate moderate or severe AD—so the available data point to greatest measured benefit in the early symptomatic window alongside low/medium tau burden [3] [1].
4. Tradeoffs: larger subgroup effects often pair with higher risk and caveats
Where subgroup benefit appears largest (low/medium tau, earlier disease), the drugs also carry nontrivial ARIA risks and bleeding concerns that rise with APOE4 carriage; meta‑analyses and reanalyses highlight that the clinical magnitude of benefit is modest in absolute terms and must be weighed against elevated ARIA‑E and ARIA‑H rates [5] [7] [10]. Industry and academic summaries emphasize that some subgroup contrasts (for example, sex, race/ethnicity, or APOE stratification) were underpowered or limited by small subgroup sizes, meaning apparent differences may reflect chance, dosing interruptions, or selection effects [11] [4] [9].
5. Bottom line and uncertainties
The best‑supported conclusion from trial subgroup analyses is that patients with lower baseline tau (an earlier tau stage) and overall earlier clinical stage obtain the largest observed treatment benefit—most clearly for donanemab’s low/medium tau subgroup and with suggestive supporting data for lecanemab—while APOE4 status primarily predicts higher ARIA risk and does not consistently predict greater benefit [1] [2] [4] [5]. Important uncertainties remain: subgroup analyses were sometimes post hoc or underpowered, APOE4 homozygote samples were small and affected by dose pauses, and racial/ethnic subgroup data are too sparse to draw firm conclusions [9] [8] [11].