Among the 67 participants who underwent MBS, met the study inclusion criteria, and provided their consent, 59 (88.06%) and 8 (11.94%) were female and male, respectively. The average age of patients was 41.94 (± 9.28) years. The participant pool was predominantly Caucasian, accounting for 73.13% of the total, with Black or African American individuals making up 20.90%. Patients identifying with more than one race, Hispanic or Latino, and Native Hawaiian/Pacific Islander comprised 2.99%, 1.49% and 1.49%, respectively (Table 1). Out of the total number of participants, 41.79% (or 28 individuals) underwent MBS at Clinic 1. Out of these participants, 13 underwent Roux-en-Y Gastric Bypass (RYGB) and 15 underwent sleeve gastrectomy. Clinic 2 accounted for 58.21% (or 39 individuals) of participants. Out of these, 36 participants underwent Roux-en-Y Gastric Bypass surgery while 3 underwent sleeve gastrectomy (Table 2). Out of all the 67 patients, fecal energy density was measured in 67 patients at 0M, 56 patients at 1M, 60 patients at 6M, 67 patients at 12M, 44 patients at 18M and 47 patients at 24M. Weight loss outcomes including change in BMI, %EWL, % total body weight loss, and total body weight loss were summarized (Table 3). In the first 12 months following surgery, there was a consistent decline in the average weight across all patients. However, the rate of weight loss decelerated thereafter, and by the 24-month mark post-MBS, there was an observed increase in average weight, which is considered expected and normal among post-MBS patients(13,14)(Figure 1a). Following surgery, there was a notable rise in average REC, which then reached its lowest point at the 12-month mark. After that, the average REC slightly increased (Figure 1b). When analyzed by type of surgery, this trend remained consistent. +-Additionally, a significant difference in REC was observed between the gastric bypass group and the sleeve gastrectomy group at the 1-month post-MBS mark (Figure S1).
- Relative fecal energy content is associated with successful weight loss changes post-MBS
Participants were classified as achieving “optimal weight loss” or “sub-optimal weight loss” post-MBS based on the median %EWL at each time point. The differences in fecal REC between patients that achieved optimal weight loss compared to those that did not were investigated (Figure 2, each point represents the fecal REC value of an individual). Specifically, the fecal REC between patients that achieved optimal weight loss compared to those that did not at one-month post-MBS were compared at baseline and 1-month timepoints (Figure 2a). Similarly, the fecal REC between patients that achieved optimal weight loss compared to those that did not at 6 months post-MBS were compared at baseline,1-month, and 6-month timepoints (Figure 2b). The same number of participants was included in all comparisons (Figure 2a-2e). No significant differences in fecal REC between the optimal and sub-optimal weight loss groups were observed at 1-month post-MBS (Figure 2a). Using 6-month optimal vs. sub-optimal weight loss categories, a significant difference in fecal REC was found between groups at 1 month (p<0.01, Figure 2b) post-MBS—that is, participants at 6 months post-MBS within the optimal weight loss group had a significantly different fecal REC at a prior time point (1 month) to those in the sub-optimal weight loss group. Using 12-month optimal vs. sub-optimal weight loss categories, a significant difference in fecal REC was found between groups at 1 month (p=0.02), 6 months (p=0.01) and 12 months (p=0.02) post-MBS (Figure 2c). Using 18-month optimal vs. sub-optimal weight loss categories, a significant difference in fecal REC was found between groups at 1 month post-MBS (p<0.01, Figure 2d). Using 24-month optimal vs. sub-optimal weight loss categories, a significant difference in fecal REC was found between groups at 1 month (p=0.03) and 12 months post-MBS (p=0.05) (Figure 2e). The p-values remained statistically significant after applying the Benjamini-Hochberg FDR correction for multiple hypothesis testing(15). A supplementary analysis was conducted on participants who had complete fecal REC and weight data at every timepoint (0, 1, 6, 12, 18, and 24M, Supp. Figure 2). Among the 28 participants examined, we noted a similar pattern to our initial analysis with the exception of a notable deviation observed at 12M, where no significant distinction between the “optimal” and “sub-optimal” groups was found. It is hypothesized that this discrepancy stems from the relatively small size of the sample (n=28).
Multivariable linear regression models were constructed to identify variables that had a significant effect on the weight loss outcomes among participants who received MBS. Using a stepwise selection method, we identified relative energy content, surgical type, baseline BMI, clinical site, and sex as statistically significant for weight loss outcome for at least one time point (Table S1). Specifically, energy harvest from a previous timepoint was selected at the 6-month, 18-month, and 24-month timepoints. To rule out the influence of the type of surgery on differences in energy harvest outcomes, a sub-analysis was conducted where the variable "surgery type" was intentionally kept in the regression models. Besides surgery type, baseline BMI, BMI post-MBS, sex and protein intake were selected as statistically significant factors influencing weight loss outcome for at least one time point. However, energy harvest was not selected in any of the models, implying that the surgery type could significantly impact intestinal energy harvest.
Random forest models were utilized to differentiate between optimal and suboptimal weight loss groups post-MBS, using 5-fold cross validation. Area under the ROC curve (AUC) values were calculated for each prediction model and labeled in Figure 3. In the model for 12 months, the inclusion of energy harvest variables (from 0M, 1M, and 6M) improved the predictive accuracy of the baseline model—which consisted of factors sex, age, surgery type, and clinical site—from 84% to 89%. Similarly, in the 24-month model, the inclusion of energy harvest variables (from 0M, 1M, and 6M) improved the predictive accuracy from 60% to 78% (Figure 3).
- Absolute fecal energy density shows limited association with macronutrient consumption
We investigated the relationship between absolute fecal energy density and dietary intake of macronutrients, such as proteins, carbohydrates, and lipids, given that these macronutrients contribute to overall caloric intake, which in turn is a factor in estimating fecal REC. After applying the Benjamini-Hochberg FDR correction, Spearman correlation tests revealed no significant correlation between macronutrient consumption and absolute fecal energy density (Table 4). These findings suggest that the influence of macronutrient consumption on fecal energy density measurements is relatively minimal.