A landmark study published in Nature in April 2026 has exposed just how poorly understood GLP-1 drug efficacy truly is. Researchers found that genetics, ancestry, age and existing health conditions all shape how a person responds to medications such as Ozempic and Mounjaro, yet three quarters of what drives individual outcomes remains completely unexplained. For the millions of people now turning to these drugs, that is a deeply uncomfortable finding.

The study, conducted by researchers at 23andMe Research Institute, analysed data from 27,885 people who had taken semaglutide or tirzepatide. It is one of the largest investigations into GLP-1 medication response ever carried out, and its results paint a picture that is far more complex, and far more uncertain, than the public conversation around these drugs tends to acknowledge.

GLP-1 Drug Efficacy Is Not What the Headlines Suggest

The numbers tell a sobering story. In clinical trials of semaglutide, the headline average weight reduction sits at around 10.2%. But that figure masks enormous variation. Around 32.2% of patients lost less than 5% of their body weight or even gained weight. Nearly one in three people saw little to no benefit at all.

In the 23andMe study cohort, participants self-reported a median BMI drop of 11.7% after 8.3 months on treatment. Yet when researchers checked electronic health records, the actual median BMI change was just 5.79%. That gap between what people believe they have lost and what medical records show is significant. It raises real questions about how these drugs are being experienced and discussed.

GLP-1 drug efficacy, in other words, is far more variable and far less certain than the marketing around these medications suggests.

The Genetic Lottery Behind Every Prescription

Scientists identified a variant in the GLP1R gene that links to modestly better weight loss outcomes. People carrying the effect allele lost an average of 0.76 kg more per copy of that allele. The researchers consider this variant likely causal, though its real-world impact is small.

More concerning is what the genetic data reveals about side effect risk. A separate variant in the GIPR gene significantly raises the risk of vomiting for people taking tirzepatide. People carrying the highest-risk genetic profile at both the GLP1R and GIPR loci faced a 14.8-fold increase in odds of severe vomiting. Nobody taking these medications routinely knows their genetic risk before they begin.

Even with all genetic and non-genetic data combined, the researchers’ predictive model explained only 25% of the variation in weight loss outcomes. The vast majority of what determines whether these drugs help or harm a given individual is still unknown.

The Side Effect Burden Nobody Warns You About

The side effect data in this study deserves far more attention than it typically receives. Among all 27,885 participants, 58.9% reported loss of appetite as a moderate to severe side effect. More than one in four, 25.1%, reported significant nausea. Constipation affected 27.3% of participants. Loss of pleasure in eating affected 25.7%. Diarrhoea affected 12.4%.

For Wegovy users specifically, nausea rates reached 29%. For tirzepatide users, loss of appetite affected 64.1%.

These are not rare or minor inconveniences. They are widespread, consistent experiences affecting the majority of people who take these medications. The body’s response to appetite suppression through pharmaceutical means carries real physical costs that extend well beyond the scale.

Who Bears the Greatest Risk

The study found that GLP-1 drug efficacy and side effect burden are not evenly distributed. People with type 2 diabetes lost significantly less weight, around 2.87 percentage points less in BMI change on average. People of African American ancestry responded least well to treatment. People with hypertension or non-alcoholic fatty liver disease also saw weaker outcomes.

In short, some of the people most likely to be prescribed these medications are also the least likely to benefit from them in terms of weight loss, while still carrying the full burden of side effects.

Age plays a role too. Every additional ten years of age corresponds to a 0.5% reduction in weight loss. Older patients, who already face greater health vulnerabilities, may gain the least while risking the most.

What Science Still Does Not Know

Perhaps the most important finding in this study is what researchers could not explain. Even the most comprehensive model, combining drug type, dosage, treatment duration, genetic variants, ancestry, age, sex and disease history, accounted for only 25% of weight loss variance. The remaining 75% is simply not understood.

The predictive model for nausea scored an AUC of just 65.4%. The vomiting model scored 68.0%. In practical terms, these models cannot reliably tell any individual what their experience will be before they start treatment.

The study population itself introduces further uncertainty. Participants were 78.3% European and 82.4% female, recruited from a commercial genetic testing company’s customer base. This is a self-selected group that does not reflect the diversity of people now using these medications globally.

A Cultural Phenomenon Built on Incomplete Evidence

According to a 2024 KFF Health Tracking Poll, an estimated one in eight people in the United States has now used a GLP-1 medication. Their rapid uptake has far outpaced the clinical understanding of long-term safety, optimal candidates and realistic outcomes.

This study does not resolve that uncertainty. It deepens it. What it confirms is that these drugs carry meaningful risks that are genetically and biologically embedded, that results vary enormously from person to person, and that the majority of what drives those differences remains beyond the reach of current science.

Weight management is one of the most personal and complex challenges a person can face. The evidence increasingly suggests that pharmaceutical shortcuts carry unpredictable costs, and that sustainable, lasting change rooted in lifestyle, behaviour and genuine support remains the most reliable path forward.

Source: dbrecoveryresources