Variations in Weight Loss and Glycemic Outcomes After Sleeve Gastrectomy by Race and Ethnicity

Abstract

Objective:

This study examined racial/ethnic differences in percent total weight loss (%TWL) and glycemic improvement following sleeve gastrectomy (SG) and explored the role of socioeconomic and psychosocial factors in post-surgical outcomes.

Methods:

This longitudinal study included patients who underwent SG between 2017 and 2020, with follow-up visits over 24 months.

Results:

Non-Hispanic Black (NHB) participants had lower %TWL at 3-, 12-, and 24-M compared to Hispanic (H) and Non-Hispanic White (NHW) participants. Fat mass index (FMI) was initially lower in NHB, with smaller reductions over time and significant group differences persisting at 24-M. NHB participants had higher baseline Fat-free Mass Index (FFMI), by 24-M FFMI was lower in H participants. A1c decreased across all groups but remained consistently higher in NHB and H compared to NHW at 24-M. NHB participants reported higher perceived discrimination, sleep disturbance, and perceived stress than H and NHW participants at all time points. Employment status predicted %TWL at 12-M. There was a significant interaction between race ethnicity and employment status, observed at 12- and 24-M, suggesting that employment-related disparities could impact surgical outcomes.

Conclusion:

NHB participants experienced less favorable outcomes following SG, emphasizing the need for tailored interventions addressing socioeconomic and psychosocial disparities.

Introduction

Obesity affects 42.4% of U.S. adults and poses significant health risks, necessitating effective interventions.^{1 2,3} On average, bariatric surgery has been reported to achieve weight loss of ~25% (~25% for sleeve gastrectomy and ~30% for gastric bypass) and has longer follow-up data than pharmacologic treatments. Semaglutide, a GLP-1 receptor agonists, results in ~15% weight loss, while dual-incretin therapy with tirzepatide has demonstrated weight loss of ~20%. ^{4, 5 6–8, 9} These findings suggest that some pharmacologic approaches may achieve weight loss comparable to bariatric surgery, though long-term durability remains to be established. Furthermore, there is limited knowledge on how race/ethnicity, socioeconomic, and other demographic variables might influence outcomes in these interventions.

Data from the Centers for Disease Control (CDC) show non-Hispanic Black (NHB) adults have higher obesity rates and poorer bariatric outcomes than non-Hispanic White (NHW) and Hispanic (H) adults, with disparities in body mass index (BMI), body fat mass index (FMI), fat-free mass index (FFMI), and hemoglobin (A1c).^{1, 10} Notably, minoritized NHB adults experience poorer outcomes across these health indicators compared to their NHW and H counterparts.¹⁰ NHB individuals experience lower percentages of total weight loss (TWL) compared to NHW and H counterparts. These disparities reflect biological and socio-cultural factors.

We established a sleeve gastrectomy (SG) cohort to investigate outcomes and mechanisms of SG in diverse patient populations.¹¹ Racial and ethnic disparities in bariatric surgery outcomes, may reflect socioeconomic and psychosocial differences, such as eating restraint, body image, stress, depression, sleep, and discrimination.¹² This study compared SG outcomes and explored factors contributing to disparities among NHB, H, and NHW patients in New York City.

Methods

Study Design

This longitudinal study analyzed predictors of weight loss by race and ethnicity in a SG cohort. This study was conducted across two major sites: New York University Langone Health (NYULH) Tisch Hospital and NYC Health + Hospital Bellevue Hospital. Participants were recruited from June 2017 to March 2020.¹¹ Institutional Review Board approval was obtained from NYULH (IRB #16–01995). All participants provided written informed consent prior to participation and received compensation for their time. Eligible participants were English- or Spanish- speaking patients aged 18–65, scheduled for SG surgery who agreed to be part of a longitudinal cohort. A detailed account of the cohort study design and baseline characteristics is published.¹¹

Primary and Secondary Outcomes

Body measurements and clinical laboratory data were obtained pre-surgery and at 1.5, 3-, 12-, and 24-months (M) following surgery. Psychosocial data were obtained through questionnaires at the same time points, with the exception of the 1.5-M assessment. These data were collected either at the bariatric clinics or at the NYU Clinical and Translational Science Institute Clinical Research Center (CTSI-CRC).

The primary outcomes included TWL, A1c, Homeostasis Model Assessment of Insulin Resistance (HOMA-IR), and fasting plasma concentrations of glucose and insulin. Secondary outcomes included body composition outcomes (waist-to-height ratio [WHtR], FMI, FFMI); and socioeconomic and psychosocial outcomes (eating restraint, body shape/weight image, discrimination, depression severity, perceived stress, sleep disturbance, education, marital status, and income). Changes were assessed to 24-M.

Anthropometric measurements

Height was measured with a stadiometer during the pre-surgical visit. Weight and waist circumference were measured using standard techniques. Body composition was assessed using bioelectrical impedance analysis (BIA) with either the InBody 270 (InBody, Inc., Cerritos, CA, USA) or the Bodycomp Scale Elite (Valhalla Scientific Inc, Poway, CA, USA). Percent body fat from BIA measurements were used to calculate FFMI, and FMI as follows:

Fat Mass Index (FMI):

1. Fat Mass (FM) Calculation:

   $$\mathrm{F}\mathrm{M}\left(\mathrm{k}\mathrm{g}\right)=\mathrm{B}\mathrm{o}\mathrm{d}\mathrm{y}\mathrm{W}\mathrm{e}\mathrm{i}\mathrm{g}\mathrm{h}\mathrm{t}\left(\mathrm{k}\mathrm{g}\right)\times (\mathrm{B}\mathrm{o}\mathrm{d}\mathrm{y}\mathrm{F}\mathrm{a}\mathrm{t}\mathrm{P}\mathrm{e}\mathrm{r}\mathrm{c}\mathrm{e}\mathrm{n}\mathrm{t}\mathrm{a}\mathrm{g}\mathrm{e}/100)$$
2. FMI Calculation:

   $$\mathrm{F}\mathrm{M}\mathrm{I}=\mathrm{F}\mathrm{a}\mathrm{t}\mathrm{M}\mathrm{a}\mathrm{s}\mathrm{s}\left(\mathrm{k}\mathrm{g}\right)/{\mathrm{H}\mathrm{e}\mathrm{i}\mathrm{g}\mathrm{h}\mathrm{t}}^2\left({\mathrm{m}}^2\right)$$

Fat-Free Mass Index (FFMI):

1. Fat-Free Mass (FFM) Calculation:

   $$\mathrm{F}\mathrm{F}\mathrm{M}\left(\mathrm{k}\mathrm{g}\right)=\mathrm{B}\mathrm{o}\mathrm{d}\mathrm{y}\mathrm{W}\mathrm{e}\mathrm{i}\mathrm{g}\mathrm{h}\mathrm{t}\left(\mathrm{k}\mathrm{g}\right)\times (1-\mathrm{B}\mathrm{o}\mathrm{d}\mathrm{y}\mathrm{F}\mathrm{a}\mathrm{t}\mathrm{P}\mathrm{e}\mathrm{r}\mathrm{c}\mathrm{e}\mathrm{n}\mathrm{t}\mathrm{a}\mathrm{g}\mathrm{e}/100)$$
2. FFMI Calculation:

   $$\mathrm{F}\mathrm{F}\mathrm{M}\mathrm{I}=\mathrm{F}\mathrm{a}\mathrm{t}-\mathrm{F}\mathrm{r}\mathrm{e}\mathrm{e}\mathrm{M}\mathrm{a}\mathrm{s}\mathrm{s}\left(\mathrm{k}\mathrm{g}\right)/{\mathrm{H}\mathrm{e}\mathrm{i}\mathrm{g}\mathrm{h}\mathrm{t}}^2\left({\mathrm{m}}^2\right)$$

All performed in fasted state, measured twice; the average measurements were used for analysis.

Blood biochemistry

Fasting blood for A1c, glucose, and insulin was analyzed by NYU clinical labs or obtained from medical records. Glucose and insulin concentrations were used to calculate HOMA-IR using the formula: HOMA-IR= (Fasting Insulin x Fasting Glucose)/22.5.¹³

Questionnaires

Questionnaires assessed demographics, eating restraint, body image, sleep, discrimination, depression, and stress. The Eating Disorder Examination Questionnaire (EDE-Q), validated for use in bariatric populations,¹⁴ was used to assess characteristics of eating restraint and body image concerns. Its subscales encompass Restraint, Shape Concern, and Weight Concern. The Experiences of Discrimination Scale (EOD-revised) questionnaire assessed the situation and frequency of discrimination based on race, ethnicity, or color across 9 life domains, such as school, work, and medical care, using a 9-item scale. This scale has been validated for use in patients with obesity.^{15, 16} The Pittsburgh Sleep Quality Index (PSQI) assessed sleep dysfunction and disturbances, evaluating various sleep characteristics such as duration, latency, efficiency, and disturbances.¹⁷ The Perceived Stress Scale (PSS) was incorporated to gauge individuals’ perceptions of the degree of stress in their lives. Additionally, the Patient Health Questionnaire (PHQ9) was used to measure depression severity.¹⁸ Scores are reported as means and standard deviations.

Statistical analysis

Data from participants who underwent surgery, self-identified as H, NHB, or NHW, and completed assessments over 24-M were included in the analyses. All outcomes were assessed quantitatively and graphically for normality. For the descriptive analysis, categorical data were presented as frequencies and percentages whereas continuous data were summarized using means and standard deviations. Unadjusted comparisons of categorical variables were performed using Chi-square test (or Fisher’s Exact test for counts <5), while means of continuous variables were compared using generalized linear regression. Trend analysis for changes in restraint, shape concern, weight concern, sleep quality, perceived stress, experiences of discrimination, and depression severity were compared using general linear regression. Anthropometric and socio-economic variables, as well as blood biochemistry outcomes, were explored using generalized linear models in SAS software (version 9.4, SAS Institute Inc., Cary, NC, USA). Models explored race/ethnicity associations, adjusted for gender, age, BMI, and site.

Linear-mixed models (LMMs) assessed race/ethnicity’s effect on outcomes. For each of the specified outcomes of interest and corresponding independent variables, we regressed repeated measured of the outcome on (1) the independent variable, (2) race/ethnicity (3) interactions between race/ethnicity and the independent variable, (4) time, and (5) the interaction between race/ethnicity and time. In addition, each model adjusted for the effects of age, gender, and site. This model allowed us to isolate the association between the independent variable and the outcome of interest within each race/ethnicity group. Missing data was not imputed as our analyses were exploratory rather than confirmatory except for LMMs analysis which accounted for missing data and assumed missing at random. We conducted additional sensitivity analyses to assess whether the use of two BIA devices to measure body composition may have impacted results. These sensitivity analyses replicated the original analysis but included additional adjustments for measurement device. Significance was set at p<0.05.

Results

We included 297 participants in the analyses (Figure 1). A majority of minoritized Hispanic participants categorized themselves as “other” for race. Consequently, we categorized race/ethnicity as H, NHB, and NHW. Retention rates were 81%, 78%, 71%, and 59% at 1.5-M to 24-M. Baseline data (Table 1) showed 69% H, 20% NHB, 11% NHW participants; 80% female, with similar ages. Baseline BMI differed significantly, with NHB participants having the highest BMI (48.8 ± 7 kg/m²) compared to H (42.9 ± 6 kg/m²) and NHW participants (45.4 ± 7 kg/m²; p < 0.0001). Type 2 diabetes status did not differ significantly across groups, with 18% of the overall cohort classified as having type 2 diabetes. Educational attainment varied significantly, with a higher proportion of NHW participants having completed graduate or professional school (34%) compared to NHB (13%) and H participants (4%; p < 0.0001). Income levels also varied significantly, with NHW participants more likely to earn over $50,000 annually (72%) compared to NHB (30%) and H participants (7%; p < 0.0001). Marital status was similar across groups.

Figure 1:

Flow diagram of participants.

Table 1:

Baseline characteristics of participants.

Characteristic	Mean ± SD, or n	Range or %	Mean ± SD, or n	Range or %	Mean ± SD, or n	Range or %	Mean ± SD, or n	Range or %	P value
	All		Hispanic		NHB		NHW
Participants	297		205		60		32
Sex, F/M	238/59		164/41		54/6		20/12
Age, y	38 ± 12	18–68	38 ± 12	18–68	37 ± 11	18–66	43 ± 13	22–62	0.068
BMI (kg/m2)	44 ± 7	33–83	42 ± 6	33–83	48 ± 7	35–70	45 ± 7	33–81	<.001
HbA1c (%)	5.9 ±1.1		6.1 ±1.2		6.0 ±0.9		5.8 ±1.3		0.484
Diabetes status, n									0.144
Normal glycemia	171	58%	119	58%	31	50%	21	63%
Prediabetes	72	24%	53	26%	17	30%	2	9%
Diabetes	54	18%	33	16%	12	20%	9	28%
Education, n									<.001
Grad/Prof school	27	9%	8	4%	8	13%	11	34%
4 yr college	64	22%	36	18%	17	28%	11	34%
< 4 yr college	80	27%	60	29%	13	22%	7	22%
High School	72	24%	56	27%	15	25%	1	3%
Some High School	54	18%	45	22%	7	12%	2	6%
Income, n									<.001
<$10,000	71	24%	50	24%	16	27%	5	16%
$10,000 – $30,000	121	41%	101	49%	18	30%	2	6%
$30,000 – $50,000	48	16%	39	19%	7	12%	2	6%
>$50,000	55	18%	14	7%	18	30%	23	72%
Missing	2	1%	1	1%	1	2%	0	0%
Marital status, n									0.555
Married	92	31%	68	33%	14	23%	10	32%
Single	150	50%	98	48%	33	55%	19	59%
Divorced	11	4%	9	4%	1	2%	1	3%
Separated	21	7%	14	7%	7	12%	0	0%
Widowed	6	2%	4	2%	2	3%	0	0%
Common-law	17	6%	12	6%	3	5%	2	6%
Employment, n									<.001
Working full time	76	26%	43	21%	12	20%	21	66%
Working part time	39	13%	30	15%	6	10%	3	9%
Seeking work	18	6%	9	4%	8	13%	1	3%
Going to school	29	10%	23	11%	5	8%	1	3%
Caring for family	44	15%	36	18%	8	13%	0	0%
Caring for family & seeking work	65	22%	46	22%	16	27%	3	9%
Recovering illness	20	7%	14	7%	4	7%	2	6%
Missing	6	2%	4	2%	1	2%	1	3%

HbA1c, hemoglobin A1c; NHB, non-Hispanic Black; NHW, non-Hispanic White General linear regression analysis was used to test differences in baseline BMI and AGE (*=p<0.05). Chi-Square test was used for categorical data. Fisher’s Exact test was used when expected counts were less than 5 (*=p<0.05).

Percent total weight loss over 24-M is shown in Figure 2A. NHB participants experienced significantly lower TWL at all time-points (p<0.05), except at 1.5M. At 12M, TWL difference between NHB and NHW participants was 6.2% (95% CI, 2.1% - 10.3%), and 4.9% between NHB and H participants (95% CI, 1.9% - 7.8%). At 24-M, the TWL difference was 6.0% between NHB and NHW (95% CI, 1.0% - 11.0%) and 4.8% between NHB and H (95% CI, 1.5% - 8.2%). No significant differences in TWL were observed between the H and NHW participants. In addition, we report body composition changes (TWL, WHtR, FMI, and FFMI) in Figures 2B, 2C, and 2D. At baseline, participants across all race/ethnic groups had a WHtR above 0.6, indicating a high risk for metabolic complications and cardiovascular diseases. Baseline WHtR was slightly lower for H compared to both NHW (0.74 vs. 0.80; p=0.016) and NHB participants (0.74 vs. 0.81; p<0.001). Over24-M, WHtR decreased for all, with no differences between H, NHW, or NHB. FMI decreased less in NHB, persisting at 24-M (p=0.029). At baseline, NHB participants had a significantly higher FFMI compared to both H (24.8 vs. 24.3 kg/m², p=0.017) and NHW participants (24.8 vs. 23.9 kg/m², p=0.007) (Figure 2D). No significant difference in FFMI was observed between H and NHW. Although FFMI decreased at 1.5-M, 3-M, and 12-M for all participants, these changes were not significant different between race/ethnicity groups. However, at 24-M, H participants showed a decrease in FFMI compared to NHW (19.4 vs. 20.8 kg/m², p=0.022) and NHB participants (19.4 vs. 20.8 kg/m², p<0.001).

Figure 2:

() Hispanic, () Non-Hispanic Black, () Non-Hispanic White.

Racial group differences in the trajectories of body composition over time. (A) total Weight loss (TWL), (B) waist-to-height ratio (WHtR), (C) fat mass index (FMI), (D) fat-free mass index (FFMI). Data were analyzed to assess differences between race/ethnic groups at each timepoint, with adjustments made for age, gender, BMI, site. Additionally, WHtR, FMI, and FFMI were adjusted for baseline values. *p<0.05.

The trajectories of glycemic metrics, including A1c, fasting glucose, fasting insulin, and HOMA-IR, were analyzed at 1.5-, 3-, 12- and 24-M post-surgery across the different race/ethnicity groups. At 24-M, A1c decreased but remained higher in NHB vs. NHW (5.3% vs. 4.9%, p=0.022) and H vs. NHW (5.3% vs. 4.9%, p=0.026) (Figure 3A). Fasting glucose, insulin, and HOMA-IR decreased initially and stabilized (Figure 3B-D). These findings suggest that while all race/ethnicity groups experience significant early improvements in glycemic control post-surgery, the magnitude and persistence of these improvements, particularly in A1c, varied among groups.

Figure 3:

() Hispanic, () Non-Hispanic Black, () Non-Hispanic White.

Racial group differences in the trajectories of glycemic metrics over time. (A) Hemoglobin A1c, (B) fasting glucose, (C) fasting insulin, and (D) HOMA-IR. Data were analyzed to assess differences between race/ethnic groups at each timepoint, with adjustments made for age, gender, BMI, site. *p<0.05.

The analysis of restraint, body shape concern, body weight concern, sleep quality, perceived stress scores across, and experiences of discrimination in different racial/ethnic groups revealed significant variations in these metrics over time (Figure 4). At baseline, restraint scores were relatively similar across all groups, indicating comparable levels of restraint-related eating behaviors among NHW, NHB, and H participants (Figure 4A). Restraint scores decreased across groups, without statistically significant differences between groups. Shape concern scores differed significantly at baseline and 12-M, with NHW participants reporting higher shape concern scores overall. At 12-M, NHW participants had higher shape concern scores compared to H participants (2.96 vs. 1.65; p=0.007) but not compared to NHB participants (2.96 vs. 2.19; p=0.112) (Figure 4B). Weight concern score differed significantly at 12-M and 24-M, with NHW participants consistently reporting the highest scores and H participants the lowest scores (Figure 4C). At 24-M, NHW participants had higher weight concern scores compared to both H participants (3.56 vs. 1.96; p=0.005) and NHB participants (3.56 vs. 2.02; p=0.008). NHB participants had poorer sleep quality (PSQI) than H at 12-M (8.17 vs. 5.93; p=0.002) and 24-M (8.31 vs. 6.38; p=0.020). (Figure 4D).

Figure 4:

() Hispanic, () Non-Hispanic Black, () Non-Hispanic White.

Racial group differences in the trajectories of restraint, shape concern, weight concern, sleep disturbance, stress, discrimination, and depression over time. (A) Restraint, (B) Shape (EDEQ) concern, (C) Weight (EDEQ) concern, (D) Sleep Quality Index, (E) Perceived Stress, (F) Experiences of Discrimination (EOD), (G) Patient Health Questionnaire. Data were analyzed to assess differences between race/ethnic groups at each timepoint, with adjustments made for age, gender, BMI, site. *p<0.05.

Perceived stress levels were significantly higher among NHB participants at baseline. Although stress levels decreased across all race/ethnicity groups by 3-M, they increased again by 24-M, remaining below baseline. At 12-M, NHB participants had significantly higher stress levels compared to H participants (14.9 vs. 11.9; p=0.041) but not compared to NHW participants (14.9 vs. 14.8; p=0.960). No significant differences were observed at 24-M (Figure 4E). The analysis of the discrimination (EOD) scores over time revealed significant race ethnic differences among participants (Figure 4F). At baseline, NHB participants reported the highest EOD scores, reflecting greater perceived discrimination compared to H participants (6.99 vs. 2.71; p<.0001) and NHW participants (6.99 vs. 2.07; p=0.002). This trend persisted at all time points, with NHB participants reporting significantly higher EOD scores at 24-M compared to H (5.68 vs. 2.26; p=0.004) and NHW participants (5.68 vs. 0.03; p=0.002). H and NHW participants had similar EOD scores, both consistently lower than those of NHB participants. Depression severity (PHQ-9), did not differ significantly between race/ethnicity groups at baseline and decreased over time, with a significant difference at 3-M between NHB participants and H participants only (3.36 vs. 1.82; p=0.023) Figure 4G.

General linear model analyzing the association between TWL and education, income, and employment, while controlling for age, BMI, gender, and hospital site, revealed that race/ethnicity was likely a factor influencing TWL at 12-M (p=0.053) and was a significant factor at 24-M (p=0.039), shown in Table 2. Income was not associated with weight loss across the various time points. Education categories showed significant differences in weight loss at 1.5-M (p=0.025); however, the effect of education attainment on weight loss interaction by race/ethnicity did not reach statistical significance. In contrast, employment status became a likely predictor of weight loss at 1.5-M (p=0.054) and 12-M (p=0.033), with race/ethnicity interactions at 12-M and 24-M.

Table 2:

General linear model for sociodemographic predictors of total weight loss.

	1.5M				3M				12M				24M
Independent Variable:	General Linear Model (R2=0.38 p<0.0001)				General Linear Model (R2=0.34 p=0.0001)				General Linear Model (R2=0.33 p=0.0010)				General Linear Model (R2=0.38 p=0.0014)
	df	MS	F	p	df	MS	F	p	df	MS	F	p	df	MS	F	p
Race/ethnicity	2	6.0	0.9	0.429	2	26.3	1.8	0.168	2	308.2	2.99	0.053	2	187.6	3.3	0.039
Education	4	20.2	2.8	0.025	4	7.8	0.5	0.712	4	25.0	0.48	0.747	4	4.7	0.1	0.987
Income	3	1.3	0.2	0.908	3	14.7	1.0	0.392	3	67.2	1.3	0.275	3	19.5	0.4	0.792
Employment	6	15.0	2.1	0.054	6	12.9	0.9	0.507	6	121.5	2.36	0.033	6	108.4	1.9	0.081
Interactions:
Race/Ethnicity * Education	8	13.0	1.8	0.074	7	8.7	0.6	0.759	6	45.5	0.88	0.509	6	59.1	1.1	0.396
Race/Ethnicity * Income	6	7.0	1.0	0.434	5	5.6	0.4	0.860	4	55.7	1.08	0.368	3	16.8	0.3	0.827
Race/Ethnicity * Employment	11	7.0	1.0	0.458	11	12.2	0.8	0.603	9	105.5	2.05	0.037	8	115.8	2.1	0.044

DF, degrees of freedom; MS, mean square; F, F value. Adjusted for gender, age, BMI, site.

We utilized linear mixed models (LMMs) to explore the associations between TWL, body composition, glycemic control, and psychosocial factors over time (Table 3). In our LMM analysis, we identified a significant difference in the relationship between sleep disturbance and TWL comparing NHW and NHB. Specifically, NHB individuals experience less weight loss compared to NHW individuals for the same level of sleep disturbance (β = 0.518, p = 0.045). As opposed to NHW participants, we observed significant evidence that both NHB and H participants experienced a more positive relationship between PHQ9 score and FMI (β = 0.155, p = 0.041 and β = 0.147, p = 0.016), suggesting that worse depression is associated with higher FMI in NHB and H compared to NHW. In terms of FFMI, poor sleep quality was significantly associated with lower FFMI among NHB (β = 0.15, p = 0.016) and H (β = 0.209, p < 0.001) participants compared to HNW participants.

Table 3:

Linear mixed model analysis: estimated effects of specified independent variables, race/ethnicity, and their interaction on the change in select dependent variables over time.

Variables		Main Effects						Interactions
		NHW		NHB		H		NHB vs NHW		H vs NHW
Dependent	Independent	β	p	β	p	β	p	β	p	β	p
TWL	RESTRAINT (EDEQ)	0.476	0.247	0.583	0.057	0.17	0.265	0.107	0.834	−0.306	0.485
	SHAPE (EDEQ)	0.344	0.522	0.073	0.829	−0.022	0.893	−0.27	0.668	−0.366	0.516
	WEIGHT (EDEQ)	0.633	0.186	0.334	0.363	−0.129	0.448	−0.299	0.617	−0.763	0.133
	PSQI	−0.328	0.136	0.19	0.157	0.047	0.531	0.518	0.045	0.375	0.107
	PHQ-9	−0.254	0.066	0.013	0.906	0.031	0.528	0.266	0.132	0.285	0.052
	PSS	0.019	0.857	0.092	0.256	0.033	0.458	0.072	0.59	0.013	0.907
	EOD	0.165	0.237	0.056	0.425	0.063	0.303	−0.109	0.486	−0.102	0.502
FFMI	RESTRAINT (EDEQ)	0.169	0.085	0.042	0.583	0.024	0.541	−0.127	0.308	−0.145	0.169
	SHAPE (EDEQ)	0.137	0.445	0.013	0.881	0.039	0.342	−0.124	0.531	−0.098	0.596
	WEIGHT (EDEQ)	0.165	0.235	0.018	0.848	0.035	0.416	−0.147	0.377	−0.13	0.369
	PSQI	−0.176	0.001	−0.026	0.463	0.033	0.079	0.15	0.016	0.209	0
	PHQ-9	−0.075	0.02	0.015	0.595	0.01	0.438	0.089	0.037	0.084	0.015
	PSS	−0.032	0.209	−0.004	0.848	0.005	0.661	0.028	0.396	0.037	0.182
	EOD	−0.022	0.489	−0.003	0.875	0.003	0.842	0.019	0.613	0.025	0.483
FMI	RESTRAINT (EDEQ)	−0.016	0.927	−0.003	0.983	−0.028	0.69	0.012	0.956	−0.012	0.948
	SHAPE (EDEQ)	−0.485	0.127	−0.014	0.927	0.058	0.424	0.47	0.178	0.543	0.098
	WEIGHT (EDEQ)	0.035	0.886	0.041	0.806	0.004	0.958	0.006	0.983	−0.031	0.903
	PSQI	−0.037	0.687	0.034	0.596	0.05	0.142	0.07	0.529	0.086	0.372
	PHQ-9	−0.121	0.031	0.033	0.511	0.026	0.256	0.155	0.041	0.147	0.016
	PSS	−0.013	0.77	0.07	0.059	0.011	0.586	0.082	0.155	0.024	0.619
	EOD	0.065	0.264	0.024	0.482	0.004	0.882	−0.041	0.545	−0.061	0.343
A1C	RESTRAINT (EDEQ)	−0.01	0.868	0.021	0.622	0.009	0.676	0.031	0.668	0.019	0.756
	SHAPE (EDEQ)	0.122	0.102	−0.064	0.162	0.041	0.073	−0.186	0.033	−0.081	0.302
	WEIGHT (EDEQ)	0.129	0.053	−0.075	0.134	0.049	0.038	−0.205	0.014	−0.08	0.255
	PSQI	−0.037	0.175	−0.023	0.174	0.01	0.297	0.014	0.662	0.047	0.106
	PHQ-9	−0.001	0.962	−0.018	0.229	0.008	0.244	−0.017	0.471	0.008	0.674
	PSS	0.007	0.61	−0.004	0.721	−0.001	0.87	−0.011	0.552	−0.008	0.595
	EOD	0.004	0.85	0.003	0.761	0.014	0.108	−0.001	0.958	0.011	0.625
HOMA IR	RESTRAINT (EDEQ)	0.296	0.448	0.302	0.29	0.128	0.379	0.006	0.991	−0.168	0.687
	SHAPE (EDEQ)	0.179	0.729	−0.284	0.371	−0.254	0.1	−0.464	0.443	−0.434	0.424
	WEIGHT (EDEQ)	0.923	0.047	−0.242	0.485	−0.265	0.097	−1.166	0.044	−1.188	0.016
	PSQI	−0.123	0.563	−0.11	0.394	−0.109	0.128	0.014	0.956	0.014	0.95
	PHQ-9	0.042	0.748	−0.111	0.279	0.018	0.698	−0.153	0.359	−0.024	0.861
	PSS	0.103	0.303	0.03	0.69	−0.058	0.161	−0.073	0.56	−0.162	0.136
	EOD	−0.017	0.901	0.087	0.213	−0.014	0.812	0.105	0.502	0.004	0.981

TWL, percent total Weight loss; FFMI, fat free mass index; FMI, fat mass index; A1c, hemoglobin. A1c; HOMA IR, Homeostasis Model Assessment of Insulin Resistance; NHW, non-Hispanic White NHB, non-Hispanic Black; H, Hispanic; PSQI, Pittsburg Sleep Quality Index; PHQ-9, Patient Health Questionnaire; PSS; Perceived Stress Score; EOD, Experiences of Discrimination Score.

For glycemic variables, higher shape concern and weight concern among NHB participants were significantly associated with higher A1c levels when compared with NHW participants, β = −0.186, p = 0.033 and β = −0.2205, p = 0.014. Additionally, weight concern was significantly associated with HOMA-IR among NHB and H participants compared to NHW participants, β = −1.166, p = 0.044 and β = −1.188, p = 0.016, respectively. The sensitivity analyses exploring use of two different BIA devices, yielded results and conclusions that were similar to those of the primary analyses (Supplemental Table 1).

Discussion

This study aimed to investigate racial/ethnic disparities in outcomes following sleeve gastrectomy, while exploring the role of socioeconomic and psychosocial factors. NHB participants had lower TWL post-surgery than NHW and H, consistent with prior studies.^19–22 In contrast, no significant differences in TWL were observed between H and NHW participants. For H patients, fewer and more inconsistent findings have been reported. Some studies have reported that H patients experience significantly less weight loss compared to NHW patients.^23–26 However, aligned with a previous two-year post Roux-en-Y gastric bypass (RYGB) or gastric band surgery study,²⁷ we have found that weight loss was similar between NHW and H patients and significantly greater than that of NHB patients. Elli and colleagues reported comparable findings up to three years post RYGB and SG surgery among NHB, H, and NHW groups.²⁸ More recently, Samaan and colleagues reported similar patterns in six-month weight loss outcomes among NHB, H, and NHW groups in a cohort of patients who underwent RYGB or SG surgery.²⁹ Our study focuses on SG, the most widely performed weight loss surgery today. In addition to TWL differences, we observed a difference in A1c and HOMA-IR, with mainly the NHW participants faring better after surgery. Admiraal et al. found no difference in the prevalence of having an A1c ≥6.5% between NHW and NHB in a meta-analysis that included 3 studies.²¹ However, in the 3 studies included, the cut off for A1c levels was heterogenous and HOMA-IR was not included. Their metanalysis study included various types of bariatric surgeries.

To our knowledge, this is the first study to examine the association between socioeconomic and psychosocial factors and weight loss differences among H, NHB, and NHW participants following weight loss surgery. By including variables such as education, income, and employment status, our analysis provides a more comprehensive understanding of how these factors may influence weight loss trajectories across different racial and ethnic groups. Race-employment interactions at 12- and 24-M suggest employment disparities impact outcomes. We observed different trends in restraint, shape/weight concerns, sleep disturbance, stress, depression severity, and experiences of discrimination between our different patient populations. NHB participants consistently reported higher sleep disturbance, perceived stress, and striking differences in experiences of discrimination. Elevated EOD scores indicate greater discrimination-related stress for NHB participants. ³⁰ We observed an association between body composition and depression severity between the NHB and H compared to NHW and shape and weight concern with glycemic control variables. These different associations further highlight how different experiences and self-perceptions between race ethnic populations influence surgical outcomes. However, it is important to note that BIA, while practical and commonly used, is not considered the gold standard for body composition assessment. Additionally, the use of two different BIA machines in this study may introduce some variability in body composition estimates; however, sensitivity analyses yielded similar results and conclusions to those resulting from the main analyses.

Beyond socioeconomic and psychosocial factors, physiological differences may also contribute to racial/ethnic differences in weight loss and glycemic outcomes following SG. Prior research suggests that NHB individuals tend to have a greater fat-free mass and lower resting energy expenditure compared to NHW individuals, which may influence post-surgical weight trajectories.^{31, 32} Additionally, differences in insulin dynamics, including higher insulin resistance and blunted postprandial insulin response in NHB individuals, may impact glycemic improvement following bariatric surgery.^{33, 34} Our study did not assess these physiological mechanisms directly; thus, future research should explore how body composition, energy expenditure, and metabolic adaptations interact with socioeconomic and psychosocial factors to shape long-term post-surgical outcomes across various populations.

Traditional SES markers may not capture challenges of diverse patients. For instance, factors like job security, access to healthcare, work-related stress, and the intersection of race/ethnicity with socioeconomic status may play more influential roles than income or educational attainment alone. Cultural differences in body image may inform adapted surgical care plans.^{35, 36} Our findings, therefore, reinforce the importance of adopting a holistic approach to patient care, taking into account not just individual socioeconomic indicators but also their interaction with race/ethnicity and other social determinants of health. Developing more targeted interventions and support systems that address these specific challenges may be key to improving weight loss outcomes for underrepresented populations undergoing bariatric surgery.

Limitations

In our analysis some interaction terms, particularly those involving specific combinations of race/ethnicity and employment categories, resulted in non-estimable effects. This reflects limited data for combinations. We also did not control for multiple comparisons; therefore, our results should be considered exploratory. Additionally, while we controlled for several key socioeconomic and demographic factors, other unmeasured variables not accounted for in our models, such as access to healthcare, social support, or the impact of systemic inequalities, may also play a role in weight loss outcomes. Our conclusions about these effects should be interpreted with caution, as they are based on the available data and may not capture all potential dynamics in the broader population. Studies with more comprehensive, ethnicity race balanced datasets and inclusion of additional relevant variables are needed to better understand the complex interplay of race/ethnicity, socioeconomic status, psychological factors, and weight loss surgical outcomes.

We found that NHB black patients had less favorable outcomes after SG than H and NHW patients. While no single socioeconomic or psychosocial factor fully explained differences in outcomes, certain psychosocial stressors, such as perceived discrimination and sleep disturbances, were more prevalent among NHB participants and may have contributed to these differences. Additional research is needed to explore how these factors interact with other determinants of postoperative outcomes. Future studies need to further investigate the complex interplay between additional socioeconomic factors, such as healthcare access, and post-surgical support in shaping weight loss trajectories, particularly for underrepresented racial and ethnic groups. Additionally, the study of larger, more diverse cohorts will be necessary to fully capture these relationships and guide the development of targeted interventions that address both medical and socioeconomic challenges for improved long-term outcomes.

Study Importance.

What is already known?

• Obesity disproportionately affects racial and ethnic minority populations in the United States, with non-Hispanic Black and Hispanic adults experiencing higher rates of obesity and related comorbidities.

• Sleeve gastrectomy is an effective treatment for obesity; however, few studies have explored outcome differences between non-Hispanic Black, Hispanic and non-Hispanic White patient populations, and even fewer have explored reasons for outcome disparities.

What does this study add?

• Non-Hispanic Black patients experienced less favorable total weight loss compared to Hispanic and non-Hispanic White patients.

• Non-Hispanic White patients had more favorable glycemic improvement compared to the other groups.

• Psychosocial stressors, including perceived discrimination and sleep disturbances, were more prevalent among non-Hispanic Black participants and may partially contribute to disparities in surgical outcomes, although no single factor fully explained these differences.

How might these results change the direction of research or the focus of clinical practice?

• Clinicians will be made more aware of potential outcome differences in their non-Hispanic Black and Hispanic patients so that they can potentially provide more individualized support.

• This study highlights the need for future research with larger cohorts to investigate reasons for these outcome disparities including additional socioeconomic, psychosocial, and biologic factors to guide the development of interventions that address the challenges and improve long-term outcomes in vulnerable populations.

Data availability statement:

Data are available upon reasonable request. Researchers interested in accessing the data from this cohort study can request it by directly contacting the corresponding author at melanie.jay@nyulangone.org. De-identified data will be available to qualified researchers for approved analyses, and access will be granted following review and approval of a study proposal, data use agreement, and ethical approval. Restrictions may apply depending on the specific nature of the data requested to ensure participant privacy and compliance with institutional policies. We encourage collaboration on projects that align with the goals of our study, particularly those focused on metabolic and psychosocial outcomes following sleeve gastrectomy.