Diet and Hair Loss: What a Large Genetic Study Really Found

A large 2025 genetic study found signals linking dietary preferences with both androgenetic alopecia (AGA) and alopecia areata (AA). Its alcohol finding deserves attention. Its melon, onion, and tea findings do not yet justify changing what you eat.

The distinction matters because this was not a trial in which people ate specific foods and researchers tracked their hair. It was a Mendelian randomization (MR) study: a sophisticated genetic approach intended to test whether an exposure may cause a disease. The analysis was ambitious, transparent, and much stronger than another cross-sectional nutrition survey. But its central exposure was food liking, not food intake. That gap leaves the study unable to tell readers how much alcohol, tea, onion, or melon changes hair-loss risk—or whether consuming any of them changes it at all.

What the genetic study found

Pan et al. (2025) tested 187 genetically derived food-preference patterns against two non-scarring hair-loss outcomes from FinnGen R11: AGA and AA. The food-preference data came from 161,625 UK Biobank participants, who rated 139 foods on a nine-point liking scale. The researchers then used genetic correlations and structural equation modeling to construct 187 non-redundant patterns.

They found 18 associations that passed a 5% Benjamini-Hochberg false-discovery-rate threshold across 374 tests: 187 exposures multiplied by two outcomes. Reported signals included higher AGA risk with alcohol consumption and lower risk signals for melon, onions, tea, and several composite sweet, bitter, and savory preference patterns.

The study used standard MR safeguards. It selected genetic instruments at genome-wide significance, removed correlated variants with LD clumping, and reported F-statistics above 10. Its primary analysis used inverse-variance weighting, with MR-Egger, weighted-median, constrained maximum-likelihood, and MR-PRESSO analyses as checks. Those choices reduce some common sources of bias. They cannot establish that a gene linked to liking melon changes melanoma-free scalp biology through melon consumption.

Why alcohol is the finding to watch

Alcohol is different from the other reported signals because the authors tested it with instruments drawn from three independent GWAS: beverage-specific intake among 141,145 people who drank at least once per week; drinks per week in 941,280 people; and grams of ethanol per day in 480,842 people. The convergent AGA signal across those datasets gives this result more support than the single-GWAS food-preference findings.

It also fits established AGA biology better than the individual-food claims do. AGA involves androgen-sensitive follicles, where conversion of testosterone to dihydrotestosterone (DHT) through 5α-reductase contributes to progressive follicle miniaturization. In dermal papilla cell experiments, Muangsanguan et al. (2025) described DHT-related 5α-reductase and TGF-β1 signaling that suppresses follicle proliferation. This is mechanistic context, not evidence that a drink causes hair loss in a person. Still, alcohol-related disruption of androgen metabolism, oxidative stress, and depletion of nutrients such as zinc and folate form a biologically plausible route to AGA.

The practical conclusion remains modest: people with AGA or a strong family history have a reasonable reason to reduce heavy alcohol use for overall health, with this study adding a possible hair-related rationale. The analysis does not provide a clinical dose-response curve, an absolute risk increase, or proof that reducing alcohol slows hair loss.

Why liking a food is not the same as eating it

The non-alcohol results rest on a harder assumption. The genetic variants predict preference ratings, not servings, grams, nutrient exposure, or long-term adherence to a diet. A person genetically inclined to like onions may eat more onions, eat the same amount as someone else, or substitute another food with a similar nutritional profile. The MR design cannot distinguish those possibilities.

The authors note that food-preference scores have test-retest reliability of 0.8 to 0.9, higher than the 0.5 to 0.8 often reported for food-frequency questionnaires. That means preference ratings are fairly stable. It does not prove that they are valid proxies for intake of quercetin, beta-carotene, kaempferol, or any other candidate compound.

The way the patterns were built adds another layer of uncertainty. The researchers clustered foods using genetic correlations in liking. Shared taste-receptor genetics could place nutritionally dissimilar foods in the same pattern because people experience bitterness or sweetness similarly, rather than because they eat those foods together or receive comparable nutrients. A pattern based on taste architecture is not automatically a dietary pattern in the ordinary nutritional sense.

The AGA–AA problem

AGA and AA are both hair-loss conditions, but they do not share the same primary biology. AGA is androgen-dependent. AA is an autoimmune disease involving loss of immune privilege in the follicle and CD8+ T-cell activity. Research on follicle cycling, including Han et al. (2022), shows that signaling pathways such as FGF can affect follicle regeneration in experimental systems; that does not turn AA into an androgen disorder.

When a single sweep of food-preference instruments produces specific associations across two diseases with such different drivers, two explanations compete. Food preferences could affect separate androgen and immune pathways at once. Or the variants may influence hair-loss risk through pathways unrelated to food consumption—a problem called horizontal pleiotropy. The latter is a serious concern here, not a technical footnote.

MR-Egger and MR-PRESSO can detect some pleiotropy, but neither method reliably solves correlated pleiotropy, where the genetic pathway affecting the putative exposure also tracks the pathway affecting disease. A clean-looking sensitivity analysis cannot rescue an instrument that never measured the exposure readers care about.

Multiple testing keeps the food claims exploratory

Testing 374 hypotheses creates opportunities for chance findings. The false-discovery-rate procedure used by Pan et al. is appropriate for exploratory work and implies about 0.9 expected false positives among the 18 reported associations. It is less stringent than a Bonferroni threshold of p<1.34×10⁻⁴. Because the publication does not report all raw p-values, readers cannot determine which of the 18 results would survive that stricter correction.

None of the 18 associations has been replicated in an independent alopecia outcome cohort. The instruments were also developed largely in British and European UK Biobank participants and applied to Finnish FinnGen outcomes. Differences in linkage disequilibrium, dietary culture, and genetic background could affect how well those instruments travel between populations.

Biology can justify research without justifying a menu

Plant compounds can affect AGA-relevant pathways under laboratory conditions. Iltaf et al. reported that Ficus benghalensis extracts inhibited 5α-reductase type II in vitro and evaluated hair growth in rabbits. That finding supports a research question: can plant-derived compounds modify an enzyme that generates DHT? It does not show that ordinary dietary amounts of tea, onions, or melon inhibit scalp 5α-reductase in humans.

The closest human intervention literature concerns concentrated supplements, not whole foods. A 2026 systematic review and network meta-analysis of dietary supplements for AGA found evidence of benefit for some supplements, with variable effect sizes and evidence quality. It did not establish that any whole-food pattern prevents or treats AGA. No controlled dietary intervention has shown efficacy for AA.

This is why diet claims in better-studied conditions look different. Dietary interventions for cardiovascular outcomes and glycemic control have been examined in randomized trials and syntheses such as a 2024 network meta-analysis of cardiovascular dietary interventions and a network meta-analysis of dietary approaches for type 2 diabetes. Dietary alopecia research has not yet built that base of intake measurements, follow-up, dose-response data, and clinical endpoints.

Do not add melon, onions, or tea to your diet as a hair-loss treatment on the basis of this study. If heavy alcohol use is part of your routine and AGA concerns you, reducing it is a defensible health decision—but the study cannot tell you what change to expect in your hair.

Pan et al. established that genetic causal-inference methods can be applied to dietary alopecia. The next studies need to measure what people actually consume, replicate findings outside FinnGen, test dose-response relationships, and connect dietary exposure to scalp androgen or immune biology in humans. Until then, the alcohol signal is a hypothesis with unusual support. The food-specific claims remain leads, not prescriptions.