Excess-Weight Metrics Misrepresent Weight Loss and No Metric Predicts the Metabolic Outcome After Metabolic and Bariatric Surgery

Technical Summary: Excess-Weight Metrics Misrepresent Weight Loss and No Metric Predicts the Metabolic Outcome After Metabolic and Bariatric Surgery

Problem Statement
The reporting of outcomes following Metabolic and Bariatric Surgery (MBS) has long relied on weight-loss metrics such as Percent Excess Weight Loss (%EWL), Percent Excess BMI Loss (%EBMIL), and Percent Total Weight Loss (%TWL). While the American Society for Metabolic and Bariatric Surgery (ASMBS) recommended %TWL in 2015 due to its statistical independence from baseline BMI, excess-weight metrics remain prevalent in surgical literature. A critical, untested assumption underpinning this debate is that weight loss metrics serve a dual function: accurately describing the magnitude of weight lost and predicting metabolic improvement (specifically glycemic control). This study investigates whether these metrics, including the recommended %TWL, accurately represent weight loss across different baseline BMIs and, more importantly, whether any weight metric can predict the 6-month metabolic outcome in patients with Type 2 Diabetes (T2D).

Methodology
The authors conducted a retrospective analysis of the 2023 Metabolic and Bariatric Surgery Accreditation and Quality Improvement Program (MBSAQIP) Participant Use File, a national audited registry.

• Cohort: The study included 182,826 adults undergoing initial sleeve gastrectomy (SG) or Roux-en-Y gastric bypass (RYGB). The primary weight-metric analysis utilized 136,406 patients with paired baseline and 6-month BMI data. A subset of 27,517 patients with paired baseline and 6-month HbA1c and BMI data formed the head-to-head metabolic analysis sample.
• Metrics: Four weight-loss metrics were computed in BMI space to ensure unit invariance:
  1. %TWL: $(BMI_0 - BMI_6) / BMI_0 \times 100$
  2. %EBMIL: $(BMI_0 - BMI_6) / (BMI_0 - 25) \times 100$
  3. %EBWL (Devine formula): $(BMI_0 - BMI_6) / (BMI_0 - BMI_{IBW}) \times 100$
  4. %EBWL (Robinson formula): Alternative calculation for ideal body weight.
• Endpoints:
  • Primary Metabolic: Change in HbA1c ($\Delta$HbA1c) at 6 months.
  • Secondary Metabolic: Early glycemic response (HbA1c < 7.0% at 6 months) among patients uncontrolled at baseline.
  • Concordance: Direct comparison of weight response (%TWL $\ge$ 20%) versus glycemic response.
• Statistical Analysis:
  • Dependence: Spearman correlation and Ordinary Least Squares (OLS) slopes were used to test metric dependence on baseline BMI.
  • Prediction: Linear models (for $\Delta$HbA1c) and logistic models (for early glycemic response) were fitted. Models adjusted for baseline HbA1c, age, sex, procedure, and insulin use.
  • Benchmarking: A "baseline-severity" benchmark (using only baseline HbA1c, insulin use, and age) was established to compare the predictive power of weight metrics against baseline metabolic status.
  • Discordance: Cohen's kappa and discordance rates were calculated to assess the alignment between weight loss and glycemic improvement.

Key Contributions and Results

1. Metric Dependence on Baseline BMI:

   • %TWL was found to be essentially independent of baseline BMI (Spearman $\rho = -0.016$) and never exceeded 100%.
   • %EBMIL and %EBWL were strongly dependent on baseline BMI ($\rho = -0.458$ and $-0.337$, respectively).
   • Uninterpretable Values: %EBMIL yielded values exceeding 100% (mathematically uninterpretable as a percentage of loss) in 1.46% of the national cohort. %EBWL variants yielded negative values (indicating weight gain) in approximately 0.5% of patients.

2. Predictive Power for Metabolic Outcomes:

   • Weak Prediction: No single weight metric significantly predicted the 6-month metabolic outcome. Adding any weight metric to a model containing baseline covariates increased the explained variance ($R^2$) for $\Delta$HbA1c by only 0.07% to 1.5%.
   • Poor Discrimination: For early glycemic response, the Area Under the Curve (AUC) for all weight metrics was poor, ranging from 0.596 (%EBMIL) to 0.630 (%TWL).
   • Baseline Severity Dominance: A model based solely on baseline severity (HbA1c, insulin use, age) achieved an AUC of 0.768, significantly outperforming any weight metric. Adding a weight metric to this baseline model only marginally improved discrimination to ~0.79.

3. Decoupling of Weight and Glycemic Response:

   • Among patients with uncontrolled baseline diabetes, weight response (%TWL $\ge$ 20%) and glycemic response (HbA1c < 7.0%) were discordant in 47.4% of patients (Cohen's $\kappa = 0.119$).
   • Specifically, 41.0% of patients achieved the glycemic target without meeting the weight threshold, while 6.4% met the weight threshold but failed to reach the glycemic target.
   • The group that lost weight but failed to improve glycemia was characterized by the highest baseline insulin use, suggesting that metabolic severity, not weight loss magnitude, drives the outcome.

4. Procedure Independence:

   • Splitting the analysis by procedure (SG vs. RYGB) reproduced these findings. Both procedures showed weak predictive power for weight metrics and high discordance between weight and glycemic responses.

Significance and Claims
The paper concludes that while %TWL is the superior metric for describing weight loss (due to its independence from baseline BMI), no weight metric, including %TWL, predicts the metabolic outcome after MBS.

• Reframing the Outcome: The authors argue that the metabolic outcome (glycemic improvement) is not a byproduct of weight change but is driven primarily by baseline metabolic severity.
• Reporting Standards: The study asserts that weight metrics should not be used as proxies for metabolic success. In patients with Type 2 Diabetes, the metabolic outcome should be measured and reported directly, prioritized alongside or above weight metrics.
• Clinical Implication: Relying solely on the "scale" risks missing patients whose diabetes does not improve despite weight loss, or conversely, overlooking patients who achieve metabolic remission without meeting arbitrary weight-loss thresholds. The authors suggest that the field must move toward standardized reporting of metabolic endpoints independent of weight-loss metrics, aligning with the mechanistic understanding that MBS improves glycemia through pathways (e.g., incretin secretion, bile acid signaling) that are distinct from fat loss.