Red Light Therapy Beats Meds in Dementia? It Depends.

Red light therapy delivers pulses of near-infrared light across the skull

Non-pharmacological interventions finally get head-to-head data

STUDY: Zhao Y et al, J Prev Alzheimers Dis 2026;13:100564

STUDY TYPE: Network meta-analysis

FUNDING: Independent

Background

Drugs like donepezil and memantine slow down the progression of Alzheimer’s, but their small benefits and side effects have placed them on the list of drugs to consider deprescribing. This network meta-analysis compares meds to non-pharmacological interventions.

The Study

Fifty-seven randomized controlled trials of 6,737 patients with mild-to-moderate Alzheimer’s were included. Six interventions were ranked against controls. Cognition was measured with standard dementia scales (MMSE, MoCA, or ADAS-Cog).

Here they are in order of ranking, with effect sizes listed in parentheses as a range (confidence intervals):

1. Red light therapy (effect size 0.29-1.02): Also called transcranial photobiomodulation, it is a noninvasive, therapeutic use of red and near-infrared light (600–1000 nm). It stimulate cellular repair, reduces inflammation, and boosts cellular energy production (ATP) without generating significant heat.
2. Meds (0.17-0.55): Donepezil, memantine, or sodium oligomannate.
3. Cognitive stimulation therapy (0.11-0.55): Socializing through group activities like puzzles, games, and conversation.
4. Exercise therapy (0.06-0.51): Mild aerobics (walking or dancing), chair yoga, resistance band exercises, or a mix of these.

Sounds promising, but wait. When I compare the trials by size, the ranking flips:

• Meds: 15 trials, 11 large
• Cognitive stimulation: 14 trials, 3 large
• Exercise: 8 trials, 1 large
• Red light therapy: 7 trials, all small

Why? Most likely, it’s because small trials bias the data with artificially inflated effect sizes. Large trials tend to get registered or published, regardless of outcome, while publication bias shines a light on small positive trials and hides small negative ones. This is a problem in meta-analyses which lump them all together, as happened in this famous example where mirtazapine looked like a great augmentation option in depression. The reality: All its small trials were negative, all its large ones were positive.

Besides these meta-analytic limitations, the network method used here is even more prone to error, as it assumes that all trials enrolled similar patients and had comparable placebo effects.

In this analysis, two therapies came up promising but not clearly positive:

• Enriched environment (0.08-1.31): Enhancing surroundings with stimulating activities like nature, novelty, puzzles, or sensory change. It’s range came close to the zero mark (0.08), and removing one outlier where it had an outsized effect brought it down.
• Music therapy (-0.08, 0.52): Passive listening, reflecting on songs, and singing or playing instruments (guided by a therapist).

Other analyses have raised doubts about the Mozart effect in this population, but some find modest benefits in specific cognitive domains in mild to moderate dementia.

Photobiomodulation is not related to bright light therapy, which delivers white light through the eye. Bright light therapy was not included here as it doesn’t improve cognition, but it did improve agitation, mood, and sleep in 6/9 small trials.

The trial also left out recently approved dementia meds, like lecanemab (Leqembi) (2023) and donanemab (Kisunla) (2024). These have generated excitement by addressing a known mechanism, amyloid plaques, but they are risky (brain swelling) and expensive. They have not been compared directly to the older meds, but in my back-of-the-envelope analysis they have a similar, small effect. Learn more in our interview with Dr. Collier.

Practice Implications

We get different results depending on how we carve the data. But that’s what happens when the bottom line is this:

1. All of these treatments have small effects
2. Non-med approaches need larger trials, but should not be overlooked in favor of pharmacotherapy