Introduction: A phenotype-tailored lifestyle intervention (PLI) has shown superior outcomes when compared to a standard lifestyle intervention (SLI) in a 12-week trial (PMID: 37007741). However, phenotype testing for obesity is time-consuming, expensive, and not widely accessible. We aimed to compare the outcomes of PLI and SLI in participants with high calories to satiation (CTS) as defined by both phenotyping (CTS_PHE) and a novel surrogate genetics-based biomarker (CTS_GRS).
Methods: We conducted a post-hoc analysis of a 12-week non-randomized clinical trial in adults with obesity assigned to receive either: A) SLI, consisting of a low-calorie mediterranean diet or a B) PLI, consisting of an a priori defined diet for each of the four phenotypes. Participants underwent validated obesity phenotyping at baseline, including an ad libitum meal test in which the number of calories consumed to achieve satiation was recorded (CTS_PHE). The abnormal satiation - hungry brain - phenotype was defined as a CTS of more than 894 kcal in females and 1376 kcal in males. Among the PLI arm, participants with this phenotype were assigned to a time-restricted, volumetric, high fiber, low-calorie diet designed to induce satiation (PLI_STN). We developed a machine-learning assisted gene risk score (GRS) for calories to satiation (CTS_GRS) in a cohort of 483 adults with obesity that underwent genotyping and obesity phenotyping. We validated the CTS_GRS in an independent cohort of 57 patients with an AUROC of 0.82. We used forward feature selection to identify the most informative weighted gene risk scores within a set of 40 expert-guided genes and employed a support vector machine as a classifier to predict patients as having low or high CTS. The primary endpoint was comparing weight loss outcomes in response to SLI vs. PLI_STN for patients with high CTS_PHEN and high CTS_GRS at 12 weeks. Continuous data are presented as mean ± SD.
Results: There were no significant differences in baseline characteristics among groups (Table 1). At 12 weeks, patients with high CTS_PHE had superior total body weight loss percent (TBWL%) with the PLI (n=25, -8.3% ± 4.6) as compared to the SLI (n=28, -2.0% ± 3.6, p<0.001) (Figure 1A). Patients with high CTS_GRS had superior TBWL% with the PLI (n= 12, -8.9% ± 4.9) as compared to the SLI (n=23, -3.9% ± 5.1, p<0.05) (Figure 1B). A higher proportion of patients with high CTS_PHEN and CTS_GRS achieved a TBWL of least 5% and 10% in response to the PLI as compared to the SLI (Figure 1C).
Conclusion: High CTS, measured by ad libitum meal or a surrogate biomarker, was associated with better outcomes to a satiation tailored PLI. Further randomized studies with a two-by-two factorial design are needed to better assess the role of these parameters in predicting best responders to different lifestyle interventions.

Figure 1. Weight loss outcomes for patients with A) high calories to satiation during an ad libitum meal test at baseline (CTS<sub>PHE</sub>) and B) a high machine-learning assisted gene risk score for satiation (CTS<sub>GRS</sub>) in response to a standard lifestyle intervention (SLI) or a phenotype-tailored lifestyle intervention for abnormal satiation (PLI<sub>STN</sub>). C) Proportion of participants with high CTS<sub>PHE </sub>and high CTS<sub>GRS </sub>achieving a total body weight loss (TBWL) percent of at least 5% or 10% by treatment allocation. Continuous data were compared with independent t tests and categorical data with Fisher's exact test. ns, non-significant; *, p<0.05; **, p<0.01; ***, p<0.001. Whiskers represent SEM.

Figure 1. Weight loss outcomes for patients with A) high calories to satiation during an ad libitum meal test at baseline (CTS_PHE) and B) a high machine-learning assisted gene risk score for satiation (CTS_GRS) in response to a standard lifestyle intervention (SLI) or a phenotype-tailored lifestyle intervention for abnormal satiation (PLI_STN). C) Proportion of participants with high CTS_PHEand high CTS_GRSachieving a total body weight loss (TBWL) percent of at least 5% or 10% by treatment allocation. Continuous data were compared with independent t tests and categorical data with Fisher's exact test. ns, non-significant; *, p<0.05; **, p<0.01; ***, p<0.001. Whiskers represent SEM.

Donald Hensrud

Appearances