Can GV-971 Really Rank With Donepezil for Alzheimer’s?

A ranking cannot create evidence that the trials do not contain

GV-971, also called sodium oligomannate, has been placed in comparative rankings beside donepezil, galantamine, rivastigmine, and memantine. That placement can look like a verdict: another Alzheimer's drug, measured and sorted. It is not. The problem is that GV-971 enters the comparison with a far thinner and more geographically concentrated trial record than the older drugs it is ranked against.

A 2024 network meta-analysis by Chen et al. compared cholinesterase inhibitors, memantine, and GV-971 across cognition (ADAS-cog), daily function (ADCS-ADL), neuropsychiatric symptoms (NPI), and global clinical impression (CIBIC-plus). It searched trials through December 2022 and aimed to compare treatment performance across disease stages. That is a useful map of a limited literature. It does not turn two GV-971 trials from Chinese clinical settings into an internationally replicated estimate.

The authors acknowledge the issue themselves: GV-971 has relatively little clinical evidence and needs more randomized trials. That admission should govern how readers interpret its rank. A league table can make evidence look evenly distributed when it is not. Here, it is not.

The GV-971 estimate rests on a narrow foundation

The GV-971 evidence in Chen et al. consists of the GREN Phase 3 trial, which supported conditional approval in China, plus a randomized controlled trial by Zhang and colleagues. The Zhang et al. trial evaluated sodium oligomannate in people with Alzheimer's disease and reported ADAS-cog and ADCS-ADL outcomes across three groups. Both available sources come from Chinese clinical settings.

This does not prove the drug fails outside China. It means the published evidence cannot establish how reproducible its observed effects would be in geographically and ethnically diverse populations or under independent regulatory review. Those are different questions from whether a signal appeared in the original studies.

Chen et al. also used fixed-effects models, arguing that the included studies had nearly identical designs and eligibility criteria. Yet the trials spanned baseline Mini-Mental State Examination ranges of 10 to 27 and lasted 20 to 48 weeks. The paper does not report per-drug I² statistics that would let readers judge its homogeneity assumption. Fixed-effects pooling can yield estimates that appear more precise than the underlying variation warrants when studies differ in population, duration, or setting.

There is a second asymmetry. Chen et al. added molecular docking analyses for donepezil, galantamine, rivastigmine, and memantine against beta-amyloid targets. GV-971 was excluded because it is a heterogeneous polysaccharide mixture without a discrete PubChem compound identifier. That exclusion is reasonable for the docking exercise, but it makes the comparison uneven. In-silico binding scores also do not establish clinical effectiveness, and they cannot validate a pooled difference on ADAS-cog or ADCS-ADL.

Older symptomatic drugs have a different kind of evidence base

Donepezil, galantamine, and rivastigmine are not interchangeable drugs, but they belong to a class with decades of multicenter clinical development. Their benefits in mild-to-moderate Alzheimer's disease are generally modest and symptomatic: they can improve cognitive and functional outcomes on average, but they do not restore lost function or reliably halt disease progression.

The evidence behind each drug has more depth than a single broad ranking can show. A 2024 Cochrane review of galantamine synthesized evidence on cognitive, functional, and neuropsychiatric outcomes in Alzheimer's disease and mild cognitive impairment. A separate 2024 systematic review and meta-analysis of acetylcholinesterase inhibitors assessed efficacy and safety across donepezil, galantamine, and rivastigmine. For donepezil, an individual-participant-data meta-analysis pooled eight randomized controlled trials and examined which patient characteristics and disease stages predicted different treatment responses.

That last point matters in practice. A population average can tell a clinician what tends to happen across a trial population. It cannot tell a particular patient how much benefit they will experience, or whether side effects will outweigh it. Participant-level evidence gives donepezil a more refined evidence base than an aggregate ranking alone can provide.

Memantine occupies a different place in care. The comparative review supports its established role as a symptomatic treatment for moderate-to-severe Alzheimer's disease. It should not be treated as a direct substitute for cholinesterase inhibitors without considering disease stage, goals of care, coexisting illness, and tolerability.

Safety rankings can miss the harms that matter most

Chen et al. assessed gastrointestinal adverse events and trial withdrawals. Those outcomes matter, but they are not a complete safety assessment for cholinesterase inhibitors. A 2024 systematic review and meta-analysis of donepezil found a signal for proarrhythmic major adverse cardiac events. A 2023 systematic review and meta-analysis also associated cholinesterase inhibitor use in cognitively impaired patients with falls, syncope, and related injuries.

These findings do not mean every patient should avoid a cholinesterase inhibitor. They mean a favorable gastrointestinal profile or low trial-withdrawal rate does not settle the real-world risk-benefit calculation, especially for someone with conduction disease, arrhythmia risk, recurrent falls, fainting, or medications that affect heart rate. A prescribing decision needs the patient's medical history, not only a ranking column.

There is also a 2025 meta-analysis suggesting acetylcholinesterase inhibitors may reduce hippocampal atrophy rates. The review describes the long-term effect on structural brain degeneration as unclear. Hippocampal volume is a structural outcome, not proof that the drugs modify clinical disease progression. Replicated, adequately powered trials would need to confirm that finding before it changes the usual description of these drugs as symptomatic treatments.

Amyloid antibodies change the question, not the trade-off

Anti-amyloid monoclonal antibodies such as lecanemab and donanemab aim to modify disease biology rather than primarily relieve symptoms. A 2025 systematic review and meta-analysis of passive anti-amyloid immunotherapies found statistically significant but modest cognitive and functional effects in carefully selected patients with early-stage Alzheimer's disease.

Statistical significance does not answer the personal question: is the average clinical gain meaningful enough for this patient to justify treatment? These drugs also carry a substantial safety and logistical burden because amyloid-related imaging abnormalities, or ARIA, require active MRI surveillance. Their trial results should not be casually generalized to people who would not have met the trials' selection criteria or could not receive the same monitoring.

They therefore sit in a separate category from GV-971. Anti-amyloid antibodies have demonstrated disease-modifying intent and meta-analytic evidence of modest clinical benefit, but their benefit-risk balance remains contested. GV-971 has a more basic unresolved problem: whether its evidence can be replicated beyond the setting in which it was developed.

A drug-only chart leaves out part of Alzheimer's care

A 2025 network meta-analysis of pharmacological, non-pharmacological, and combined interventions found that combined approaches may improve cognitive and functional outcomes more than drug monotherapy. The finding is emerging rather than settled, but it exposes a limitation of drug-only comparisons: families do not experience Alzheimer's disease as an ADAS-cog score alone.

Physical activity also provides context for interpreting the scale of drug effects. A meta-analysis of randomized trials compared physical activity in people with Alzheimer's disease or mild cognitive impairment with donepezil as a reference point. It was not a head-to-head claim that exercise beats donepezil; it helps place drug effect sizes in a broader treatment context.

Melatonin has another, narrower evidence signal. A network meta-analysis of placebo-controlled trials reported dose- and duration-dependent associations between melatonin treatment and overall cognition in Alzheimer's dementia. This is not enough to treat melatonin as a replacement for established therapy, and the review itself notes divergent trial results. It does show why a supposedly complete drug landscape rarely stays complete for long.

The practical verdict

Chen et al. provide a useful survey of comparative trial data through December 2022. The analysis cannot support treating GV-971 as equally established with donepezil, galantamine, rivastigmine, or memantine. The difference is not a matter of taste or geography; it is evidence depth. The older drugs have geographically broader, repeatedly synthesized trial programs. GV-971 has two trials from one clinical and regulatory context, and the ranking paper itself calls for more evidence.

For mild-to-moderate Alzheimer's disease, cholinesterase inhibitors remain the best-established symptomatic medicines, with side effects that require individualized review. Memantine remains an established symptomatic option for moderate-to-severe disease. Anti-amyloid antibodies offer modest disease-modifying clinical effects in selected early-stage patients but require specialized safety monitoring. GV-971 remains a hypothesis awaiting independent replication.

For patients and families, the useful question is not which drug wins a meta-analysis table. It is which option fits the disease stage, expected benefit, cardiovascular and fall risk, treatment burden, and access to monitoring. Alzheimer's treatment has no clean winner because the evidence describes several different clinical decisions.