Maybe It’s Not Just in Your Genes - How Everyday Behaviors Could Enhance Predictive response to Obesity Treatment

What if we could predict who would respond to which medication—or, better yet, who wouldn’t?

That’s been one of the holy grails in obesity medicine: moving from “trial-and-error prescribing” to personalized, predictive treatment. We’ve made some progress—but not enough.

Beyond Clinical Phenotypes

Historically, obesity phenotyping has relied on what we can measure in the clinic: symptoms, biomarkers, and health outcomes. In research, the tools have become increasingly sophisticated—multi-omics datasets spanning genomics, proteomics, metabolomics, and epigenetics promise to decode the biological underpinnings of obesity. But these approaches have practical shortcomings: beyond their current cost, limited accessibility, and lack of scalability, they are largely non-functional measurements—capturing static molecular signatures rather than dynamic processes or behaviours. Even large electronic medical record (EMR) datasets—often heralded as “real-world evidence”—are mostly single snapshots, aggregated through provider entries or periodic surveys. They miss what happens between visits, which is where most chronic disease outcomes really unfold.

So here’s the catch: these are rich biological datasets, not behavioural ones. They tell us what’s happening inside the body, not how that body behaves day-to-day. If epigenetics determines our susceptibility to obesity, then understanding the environment will ultimately help us predict who will activate and express that risk.

What if the most deterministic layer of data isn’t molecular at all?

What if it’s the millions of micro-decisions patients make every day—when to eat, what to eat, how they react to hunger cues, how they interpret side effects, when they skip a dose, or when motivation falters and returns that actually predict success of obesity treatment?

More than 99% of a person’s chronic-disease life is spent outside the healthcare system. Yet that’s the data we’ve never truly captured.

With the help of advances in technology, we can finally start to map this invisible terrain.

By quantifying digital behaviours and emotional cues—how people engage, adapt, and self-regulate—we can identify digital phenotypes that may be even more predictive than traditional clinical classifications. Patterns that once seemed random begin to show rhythm. Associations that repeat across thousands of lives begin to look less like coincidence and more like signal.

Of course, correlation isn’t causation—but with scale and consistency, weak signals can become strong predictors.

This same principle has inspired initiatives such as the ICHOM Obesity Standard Set—a global effort to standardize the measurement of patient-reported outcomes in obesity care, which I had the privilege of contributing to alongside a team of international experts. The vision was clear: if we measured lived experience as consistently as we do lab values, what new truths could emerge?

The challenge, as always, is scaling and operationalizing it in the real world.

Clinical Phenotypes: A Step Forward

One of the most compelling efforts to date came from Dr. Andreas Acosta and colleagues at the Mayo Clinic, who pioneered a pragmatic, clinic-friendly model of obesity phenotyping.

They identified four main groups based on behavioural and physiological patterns:

• Hungry brain (abnormal satiation): delayed fullness, needing larger meals to feel satisfied.

• Hungry gut (abnormal satiety): hunger returns soon after eating.

• Emotional hunger: eating triggered by emotion or stress rather than caloric need.

• Slow burner: reduced resting energy expenditure and metabolic efficiency.

In their randomized clinical trial, patients whose treatments were phenotype-guided—meaning therapy was matched to their dominant phenotype—achieved a 16.7% average weight loss, compared with 9% in those treated with standard care. Nearly two-fold greater efficacy, achieved simply by aligning mechanism with the behaviour. It was an elegant demonstration of what precision obesity medicine could look like.

But even Acosta’s data revealed the complexity we all see in clinic: most patients straddle multiple phenotypes. Obesity is not a single disease—it’s a network of overlapping biological and behavioural drivers. And no one pathway—or drug—can capture that fully.

Shifting from Phenotypes to Health Outcomes

The Canadian Obesity Clinical Practice Guidelines published an update in August 2025 to the pharmacotherapy chapter (of which I was an author for full transparency) and we took a different stance - what if, instead of classifying by obesity phenotype, we prescribed by health outcome?

If a patient’s dominant obesity related complication is obstructive sleep apnea, tirzepatide is a rational first choice—it’s the only medication to date shown to improve moderate sleep apnea.

If the dominant risk is cardiovascular disease, semaglutide should be considered first line—it remains the only anti-obesity pharmacotherapy with proven cardiovascular benefit.

But what if they have both? Or both plus fatty liver disease, joint pain, and prediabetes?

That’s not hypothetical—that’s just reality. People and their organs don’t live in silos, and neither should their treatment. Yet that’s exactly where we are today in obesity medicine, constrained by the fragmented way our data are collected and interpreted.

So here is the real question: can we integrate all these outcomes simultaneously, and let behavioural data guide which mechanism matters most for a given individual at a given time?

Harnessing New Digital + Biological Data: A Paradigm Shift

Recent advances hint at a major leap forward. In oncology, for example, there has been a strategic partnership merging deep molecular data (genomics, liquid biopsies) with real-world clinical data and advanced AI analytics to predict drug response and identify novel biomarkers.

This model is fascinating because it shows how merging molecular + behavioural + real-world data can lead to actionable predictions of which patients will respond to which therapy.

So why not apply the same logic to obesity medicine?

What if we combined:

• behavioural micro-decisions (daily eating patterns, dose adherence, motivation shifts)

• biological data (genetics, epigenetics, metabolic markers)

• real-world context (comorbidities, environment, patient-reported experiences)

At nymble, this is exactly the challenge we’ve embraced.

We’re exploring whether conversational data—the words, sentiments, and engagement patterns that patients naturally share—can be transformed into objective obesity phenotypes that help predict:

• Risk of medication intolerance

• Likelihood of efficacy

• Patterns of adherence and relapse

The goal isn’t to monitor—it’s to empower. Imagine if the subtle cues in your conversations—your side-effect experiences, tone of motivation, and rhythm of engagement—could inform and improve a more personalized, compassionate treatment plan.

All of this must be built with a privacy-by-design mindset: transparent, secure, and ensuring that the patient is always the primary benefactor of the technology.

To make this possible, a new phenotyping framework must be:

• Scalable across populations and care settings

• Standardized so signals can be compared across contexts

• Culturally adaptive, recognizing that behavior and language vary

• Longitudinal, learning not just who you are, but how you change

That’s what we’re building toward at nymble: a bridge between the biological and behavioural layers of chronic disease using real-world conversational data as the missing link.

We believe predictive, personalized obesity treatment is not a distant ideal.

It’s an inevitable evolution if we choose to measure what truly matters.

---

📍 Meet us in person at the Patient Support Summit in Toronto on November 4–5th, or at Obesity Week in Atlanta November 4-7th.

📧 Or reach out to us directly at hello@nymble.health to explore how nymble can fit into your organization.