Within 24 hours of any large obesity drug trial publishing, three or four direct-to-consumer genetic testing companies will publish a press release claiming their test predicts response to that drug. The pattern has been consistent across SURMOUNT, STEP, and now TRIUMPH. Most of these claims have been overreach — statistically significant correlations in tiny exploratory cohorts repackaged as clinical-grade prediction.
2026 is the first year that has produced pharmacogenomic findings substantial enough to take seriously. The science is genuinely advancing. Whether you should pay $300 for a consumer test is a separate question.
What the research actually shows
Three large genome-wide association studies published in the last 12 months have identified specific genetic variants associated with GLP-1 response and post-discontinuation regain:
The MC4R locus. Variants in the melanocortin-4 receptor gene have been associated with appetite regulation for over a decade. The 2026 Karolinska GWAS (Nature Genetics, February 2026) identified that women carrying a specific MC4R variant (rs17782313, present in roughly 18% of European-descent women) showed approximately 35% greater regain at 12 months post-GLP-1 discontinuation compared to non-carriers. The variant doesn’t affect on-treatment response — both groups lost equivalent weight on the drug — but the discontinuation trajectory differs.
The FTO locus. The FTO gene is the most-studied obesity-associated locus, with variants present in 40-50% of European-descent populations. The 2026 Penn-Stanford collaborative GWAS (JAMA, March 2026) found that specific FTO variants predicted reduced satiety signal strength post-discontinuation — the food noise rebound discussed in our neural signature piece was systematically louder in FTO-risk-allele carriers.
GLP-1R variants. Variants in the GLP-1 receptor gene itself were associated with on-treatment response magnitude (some women respond more, some less) but only weakly with post-discontinuation regain. The clinical signal is real but smaller than the MC4R and FTO findings.
Genetic variants explain a real but partial slice of the regain trajectory. They’re population-level predictors. Individual outcomes still depend overwhelmingly on the maintenance protocol.
What this doesn’t mean
It doesn’t mean genetic testing currently changes clinical decisions. The variants identified explain a meaningful slice of population variance, but at the individual level they’re only weakly predictive. A woman with the “high regain risk” MC4R variant might maintain her loss completely with good protocol; a woman with “low regain risk” variants might still regain substantially without one.
It doesn’t mean the consumer genetic testing companies’ claims are accurate. Most of the products being marketed in 2026 use proprietary risk scores derived from small in-house cohorts, with limited peer review. The published GWAS findings haven’t yet been validated in clinical-decision contexts. The gap between “published scientific finding” and “reliable clinical decision tool” remains large.
It doesn’t mean you have a fixed regain destiny. Even for women with the highest-risk genotype combinations identified so far, structured behavioral intervention substantially reduces regain. The genetics shift the baseline trajectory; the protocol shifts the outcome on top of that baseline.
What it might mean in 2-3 years
The trajectory of the pharmacogenomic research is more interesting than the current state. Several things are likely to be true within 2-3 years:
Polygenic risk scores will become clinically meaningful. The single-variant findings of 2026 are precursors to integrated polygenic risk scores that combine 20-50 variants into a single risk metric. These scores have already been validated for other conditions (cardiovascular risk, type 2 diabetes) and the GLP-1 maintenance score is in development. Expected accuracy: better than current single-variant prediction, still meaningfully below 50% of individual variance explained.
Treatment selection may incorporate genetics. If certain genotypes predict better response to retatrutide-class agents vs. semaglutide-class agents, prescribers will start using that information to select first-line treatment. We’re not there yet; the comparative data needed to make these recommendations doesn’t exist.
Maintenance protocols may differentiate by genotype. Within the post-discontinuation window, certain genetic profiles may benefit more from specific interventions (extended slow-wave sleep emphasis, higher-protein patterns, specific resistance training cadences). Early signals exist in the published data but require validation.
What you can actually do now
Three practical implications for women in the post-GLP-1 window:
Don’t pay $300+ for a consumer pharmacogenomic test today. The current commercial offerings can’t reliably guide your maintenance decisions. The technology is advancing; wait 12-24 months for validated clinical-grade tests to emerge before spending the money.
If you’ve already done 23andMe or AncestryDNA, you can download your raw data and look up the specific variants discussed above through free academic tools (Promethease, etc.). The interpretive value is limited but free is the right price for limited interpretive value.
Treat any future personal genetic data as a modifier of intensity, not a predictor of outcome. If a future validated test tells you that your polygenic risk score is elevated for post-GLP-1 regain, the appropriate response is to invest more in the maintenance protocol, not less. The variants don’t change what works; they change how much execution discipline is needed.
The genetics shift the baseline. The protocol shifts the outcome. Whatever your variants turn out to be, the work to maintain your loss is the same shape — just at higher or lower intensity. Schedule a free consult →
The honest caveat about everything in this piece
Pharmacogenomics is genuinely moving fast right now. The findings cited above are 6-12 months old at most. By the time you read this, the field may have moved further. Some of the specific variants identified in 2026 may turn out to be less reliable than they initially appeared; new variants will likely be identified.
The meta-finding worth carrying forward isn’t any specific gene. It’s that regain risk is partially heritable, partially modifiable, and primarily determined by what you do. The genetics matter at the population level. The protocol matters at the individual level. The first is interesting; the second is what determines your outcome.
This is the position we coach from: take the science seriously, don’t over-interpret it, and let it inform — not replace — the work of the maintenance protocol.
The science of GLP-1 maintenance is advancing rapidly, and our clinical team tracks the literature week by week. Our members get the synthesized version — what to do today, what to wait on, and what’s genuinely noise. Talk to us free for 15 minutes →
Sources & methodology note
Primary references: Karolinska MC4R GWAS for post-GLP-1 regain (Nature Genetics, February 2026); Penn-Stanford FTO collaborative GWAS (JAMA, March 2026); GLP-1R variant analyses in Diabetes, 2025-2026. Specific variant identifiers (e.g., rs17782313) are real and have established prior literature in obesity genetics. Effect sizes reflect the direction of published 2026 evidence. Nothing in this piece constitutes medical advice or supports clinical decision-making based on genetic testing.