Burden and impact of metabolically healthy obesity on in-hospital outcomes of acute ischemic stroke: A nationwide study

Abstract

BACKGROUND

Metabolically healthy obesity (MHO) is defined as obesity without coded hypertension, diabetes, and hyperlipidemia. The impact of MHO on acute ischemic stroke (AIS) outcomes is uncertain. We examined the burden of MHO on AIS hospitalization outcomes.

AIM

To evaluate whether MHO is associated with differences in in-hospital outcomes among adult patients hospitalized with AIS across various age groups (18-44 years and 45-64 years), sexes, and racial/ethnic subgroups, using a nationwide propensity score-matched analysis.

METHODS

In this observational study, the National Inpatient Sample (2020) was queried for AIS patients with vs without MHO. Adult AIS patients were analyzed across age groups (18-44 years and 45-64 years), sexes, and racial/ethnic subgroups to determine if MHO was associated with in-hospital outcomes. A propensity score-matched analysis (1:1) was used to compare baseline characteristics, comorbidities, in-hospital outcomes, and healthcare resource utilization between the MHO and non-MHO AIS cohorts.

RESULTS

Among 62675 AIS hospitalizations, 5135 (8.2%) had MHO. After propensity matching across the groups, there were 4935 patients with and without MHO. The MHO cohort had more males (38.8% vs 37.4%) and fewer rural hospital admissions (5.8% vs 7.0%, P = 0.02) compared with the non-MHO cohort. MHO patients exhibited higher rates of baseline comorbidities such as hypothyroidism (9.9% vs 6.9%) and chronic obstructive pulmonary disease (16.6% vs 14.2%). However, in-hospital mortality was similar between the groups (13.5% in MHO vs 14.5% in non-MHO patients). Compared with non-MHO AIS patients, there was no significant difference in all-cause mortality in MHO patients (odds ratio = 0.93, 95%CI: 0.70-1.24). Subgroup analyses stratified by age, sex, and race did not demonstrate statistically significant differences in in-hospital mortality, and multivariable models did not identify significant associations between MHO and in-hospital mortality.

CONCLUSION

This nationwide propensity-matched analysis found that metabolically healthy obesity was not associated with worse short-term in-hospital outcomes among patients hospitalized with acute ischemic stroke. However, these findings should be interpreted cautiously given the claims-based definition of metabolically healthy obesity. Further studies with more granular metabolic and longitudinal data are needed to clarify the long-term clinical implications of metabolically healthy obesity in stroke.

Key Words: Acute ischemic stroke; Metabolically healthy obesity; In-hospital mortality; Risk factors; Stroke

Core Tip: Metabolically healthy obesity (MHO) is increasingly recognized as a distinct obesity phenotype, but its implications in acute ischemic stroke (AIS) remain uncertain. In this nationwide propensity score-matched study using the 2020 National Inpatient Sample, MHO was not associated with worse short-term in-hospital outcomes or higher in-hospital mortality among patients hospitalized with AIS. These findings support a more phenotype-based interpretation of obesity-related risk in stroke, while also highlighting the need for further studies with more granular metabolic and longitudinal data to clarify long-term clinical implications.

INTRODUCTION

Obesity has been linked to an increased risk of developing cardiovascular disease (CVD) and, in turn, an increased risk of stroke. It is the major contributor to the global epidemic of hypertension (HTN), diabetes mellitus (DM), CVD, coronary artery disease (CAD), and heart failure. The prevalence of obesity has increased by 27% globally over the last three decades, accounting for 2.1 billion people and contributing to 1 in every 5 deaths[1]. Its prevalence among stroke patients ranges from 18%-44%, as defined by body mass index (BMI)[2]. However, studies evaluating the influence of obesity on cardiovascular outcomes have yielded conflicting findings, particularly after recognition of the clinical phenotype known as “metabolically healthy obesity” (MHO)[3].

MHO is characterized by the predominance of subcutaneous adipose tissue, which maintains relatively preserved adipocyte function, lower inflammatory signaling, and greater insulin sensitivity without dyslipidemia and HTN[4]. In contrast, metabolically unhealthy obesity (MUO) is marked by excess visceral and abdominal adiposity, where adipocytes exhibit cellular stress, chronic low-grade inflammation, impaired adipokine secretion, and insulin resistance. Visceral adipocyte hypertrophy contributes to lipotoxicity, mitochondrial dysfunction, and systemic metabolic derangements, which underlie the higher cardiometabolic risk associated with MUO. These biological differences are believed to influence the atherosclerotic burden and vascular outcomes in stroke patients. The true prevalence of MHO is unknown because of the lack of broadly accepted diagnostic criteria[5]. Despite this fact, based on the diagnostic criteria currently used, the prevalence of MHO ranges from 3.3%-32.1% in men and 11.4%-43.3% in women[6].

Acute ischemic stroke (AIS) is a significant health burden and is considered an important manifestation of CVD. The relationship between CVD risk factors and ischemic stroke has been extensively documented. HTN and DM stand out as the most prevalent risk factors among patients with stroke, accounting for approximately 64%[7] and 14%-46%[8], respectively. Although obesity is a recognized modifiable risk factor for stroke, its clinical significance may depend more on associated metabolic dysfunction than on excess body weight alone[9]. Prior studies evaluating MHO and cardiovascular outcomes have shown conflicting findings[10], and data specifically examining in-hospital outcomes among patients hospitalized with AIS remain limited and inconclusive.

Given the increasing prevalence of obesity and the uncertainty surrounding the clinical implications of MHO, better characterization of this phenotype in patients with AIS is necessary. Therefore, we performed a nationwide analysis using the 2020 National Inpatient Sample (NIS) to evaluate the association between MHO and in-hospital outcomes among adults hospitalized with AIS.

MATERIALS AND METHODS

Source of data

Hospital records of AIS patients aged > 18 years, with and without obesity, identified using appropriate ICD-10 codes from the 2020 NIS were included in the study. The NIS is the largest publicly available all-payer inpatient care database in the United States, with estimated data on more than 35 million hospital stays nationally. The Agency for Healthcare Research and Quality sponsored the Healthcare Cost and Utilization Project, which was established through a Federal-State-Industry partnership[11]. Its large sample size is ideal for developing national and regional estimates and enables physicians to analyze rare conditions, uncommon treatments, and special populations. Because all of the study’s hospitalizations were de-identified, institutional review board approval was not required. Records were processed using IBM SPSS Statistics version 25. ICD-10-CM codes were used to define AIS (I63.x), obesity (E66.x), HTN (I10.x-I15.x), diabetes (E08.x-E13.x), and hyperlipidemia (E78.x). MHO was defined as patients with obesity in the absence of coded HTN, DM, and hyperlipidemia. Data were cleaned for duplicates, and any records with missing outcome data were excluded.

Study outcomes and variables

AIS hospitalizations with HTN, DM, and hyperlipidemia were excluded from the study population, and patients were categorized into those with MHO and those in the non-MHO group. Cases with missing sex and race variables were also excluded before propensity score matching. Because this was a retrospective observational study, we performed 1:1 nearest-neighbor propensity score matching with a caliper width of 0.01 based on age, sex, and race after excluding cases with missing sex and race variables. This approach achieved an absolute standardized difference of less than 10% between the matched variables pre- and post-matching, suggesting improved balance, reduced confounding, and better comparability between the matched groups. Several baseline characteristics of AIS patients were considered in the analysis, including mean age at hospitalization, sex, ethnicity, median household income, primary payer, type of hospital admission, region, and hospital location. These variables were carefully stratified for patients with and without MHO to provide a comprehensive comparison (Figure 1).

Figure 1 Flow diagram of adult acute ischemic stroke admissions from the 2020 National Inpatient Sample database. After exclusion of patients with metabolic abnormalities and missing sex or race data, 4935 patients with metabolically healthy obesity were matched 1:1 to 4935 patients in the non-metabolically healthy obesity group using propensity score matching based on age, sex, and race. MHO: Metabolically healthy obesity.

Statistical analysis

The Pearson χ² test for categorical variables and the Mann-Whitney U test for continuous variables (non-normal distribution) were used to compare baseline demographics, clinical comorbidities, substance-use variables, and hospital characteristics of AIS patients between the two groups (without and with MHO). A significance level of P < 0.05 was set for statistical tests using a two-tailed approach. Categorical and continuous variables were quantified in percentages and median and interquartile range, respectively. A multivariable logistic regression model was performed after adjusting for confounding factors (age, sex, race, median household income national quartile, payer type, weekend/weekday admissions, hospital location/teaching status, peripheral vascular disease, tobacco use disorder, prior myocardial infarction, prior transient ischemic attack or stroke without neurologic deficit, prior sudden cardiac arrest, prior venous thromboembolism, cancer, chronic kidney disease, acquired immune deficiency syndrome, alcohol abuse, drug abuse, cocaine, cannabis, depression, chronic pulmonary disease, hypothyroidism, other thyroid disorders, valvular disease, autoimmune conditions, and coronavirus disease 2019) to determine the odds of in-hospital mortality in AIS patients with or without MHO in the overall sample and in subgroups by age (18-44 years, 45-64 years, and ≥ 65 years), sex (male and female), and race (White, Black, Hispanic, and Asian or Pacific Islander). SPSS v25 software (IBM Corp., Armonk, NY, United States) and complex sample modules were used to perform all statistical analyses following strata and cluster designs.

RESULTS

Among the 62675 AIS-related hospitalizations, 5135 patients (8.2%) had MHO. After propensity matching, there were 4935 patients in each of the MHO and non-MHO groups. The median age of respondents with MHO was 56 years [interquartile range (IQR): 50-62], while the median age of those without MHO was 61 years (IQR: 54-66). A higher proportion of females was seen in both cohorts, although females were slightly more prevalent in the non-MHO cohort (62.6% vs 61.2%) and males were more prevalent in the MHO cohort (38.8% vs 37.4%, P = 0.146). Patients in the first income quartile constituted a larger proportion of the MHO cohort than the second income quartile (32.3% vs 29.8%). Additionally, the MHO group had a higher proportion of patients with private insurance compared with the non-MHO group (64.4% vs 59.7%, P = 0.01). AIS admissions with MHO were more common in urban teaching hospitals, followed by urban non-teaching and rural hospitals (79.0% vs 15.2% vs 5.8%, P = 0.01) and were associated with higher hospital expenses ($88297 vs $78270). MHO patients exhibited higher rates of selected baseline comorbidities, including chronic pulmonary disease (16.6% vs 14.2%) and hypothyroidism (9.9% vs 6.9%); small residual differences were also noted in prior myocardial infarction and prior transient ischemic attack/stroke (Table 1). However, the in-hospital mortality rate was similar between the two groups (13.5% vs 14.5%, P = 0.613). The odds of all-cause mortality were not significantly different between the two groups (OR = 0.93, 95%CI: 0.70-1.24, P = 0.613) (Table 2). Subgroup analyses by age, sex, and race did not identify statistically significant differences in in-hospital mortality between the groups. Among the age subgroups, patients aged 18-44 years had an OR of 0.85 (95%CI: 0.42-1.73), those aged 45-64 years had an OR of 0.98 (95%CI: 0.60-1.61), and those aged > 65 years had an OR of 1.02 (95%CI: 0.65-1.61). In terms of sex, males had an OR of 1.04 (95%CI: 0.64-1.70), while females had an OR of 0.96 (95%CI: 0.65-1.39). Regarding racial groups, Asians or Pacific Islanders had an OR of 2.93 (95%CI: 0.46-18.82), Hispanics had an OR of 1.01 (95%CI: 0.34-3.01), Black respondents had an OR of 0.77 (95%CI: 0.33-1.80), and White respondents had an OR of 0.92 (95%CI: 0.66-1.28). None of these subgroup estimates reached statistical significance (Figure 2).

Figure 2 Forest plot showing adjusted odds ratios for in-hospital mortality in patients with acute ischemic stroke and metabolically healthy obesity compared with patients in the non-metabolically healthy obesity group, stratified by age, sex, and race/ethnicity. The X-axis was truncated at 5 for readability; the confidence interval for the Asian/Pacific Islander subgroup extends beyond the plotted range. AIS: Acute ischemic stroke; MHO: Metabolically healthy obesity; CI: Confidence interval.

Table 1 Baseline characteristics, comorbidities, substance use variables, and outcomes following acute ischemic stroke in patients with and without metabolically healthy obesity, 2020.

Variables		Non-MHO (n = 4935)	MHO (n = 4935)	P value
Demographic characteristics
Age (years) at admission	Median [IQR]	61 [54-66]	56 [50-62]	0.115
	18-44	29.9	29.0
	45-64	36.5	38.6
	≥ 65	33.5	32.4
Sex	Male	37.4	38.8	0.146
	Female	62.6	61.2
Race	Whites	67.2	69.3	0.083
	Blacks	19.2	18.6
	Hispanics	11.3	10.3
	Asian/PI	1.4	1.2
	Nat Ame	0.9	0.6
Socioeconomic & hospital characteristics
Median household income national quartile for patient ZIP code	0-25th	32.0	32.3	< 0.0001
	26-50th	25.9	29.8
	51-75th	23.0	23.3
	76-100th	19.1	14.6
Payer type	Medicaid	40.3	35.6	< 0.0001
	Private	59.7	64.4
Hospital location & teaching status	Rural	7.0	5.8	< 0.0001
	Urban non-teaching	12.5	15.2
	Urban teaching	80.5	79.0
Hospital region	Northeast	20.1	16.2	< 0.0001
	Midwest	17.9	21.9
	South	42.6	42.1
	West	19.4	19.7
Clinical comorbidities
Peripheral vascular disease		10.6	9.7	0.182
Prior myocardial infarction		2.3	1.7	0.032
Prior TIA/stroke without neurologic deficit		7.8	6.7	0.033
Prior venous thromboembolism		5.2	4.9	0.489
Valvular disease		2.4	1.0	< 0.0001
Cancer		11.2	7.3	< 0.001
Chronic kidney disease		4.9	5.2	0.489
Acquired immune deficiency syndrome		1.2	0.2	< 0.0001
Depression		9.8	10.6	< 0.244
Chronic pulmonary disease		14.2	16.6	< 0.001
Hypothyroidism		6.9	9.9	0.0001
Other thyroid disorders		1.9	1.6	0.253
Substance use variables
Alcohol abuse		7.2	4.7	< 0.0001
Drug abuse		10.1	5.6	< 0.0001
Cocaine		2.5	1.7	< 0.0005
Tobacco use disorder		17.8	16.8	0.182
Outcomes
All-cause mortality		14.5	13.5	0.613
Disposition of patient	Routine	44.8	43.3	0.021
	Transfer to short-term Hospital	4.2	4.3
	Transfer other: SNF, ICF, etc.	35.9	38.8
	Home health care	15.2	13.6
Length of stay in days	Median [IQR]	5 [2-13]	5 [2-12]	0.032
Health care expenses (in United States dollars)	Median	$78270	$88297	0.001

P < 0.05 was considered statistically significant. PI: Pacific Islanders; Nat Ame: Native Americans; TIA: Transient ischemic attack; IQR: Interquartile range; SNF: Skilled nursing facility; ICF: Intermediate care facility; MHO: Metabolically healthy obesity.

Table 2 Multivariable logistic regression assessing the odds of in-hospital mortality following acute ischemic stroke in patients with and without metabolically healthy obesity.

In-hospital mortality	Metabolically healthy obesity	Adjusted OR	95%CI		P value
			Lower Limit	Upper limit
Overall mortality	Yes vs no	0.93	0.70	1.24	0.613
In-hospital mortality by individual subgroup
18-44 years	Yes vs no	0.85	0.42	1.73	0.662
45-64 years	Yes vs no	0.98	0.60	1.61	0.930
≥ 65 years	Yes vs no	1.02	0.65	1.61	0.930
Male	Yes vs no	1.04	0.64	1.70	0.880
Female	Yes vs no	0.96	0.65	1.39	0.812
Whites	Yes vs no	0.92	0.66	1.28	0.618
Blacks	Yes vs no	0.77	0.33	1.80	0.546
Hispanics	Yes vs no	1.01	0.34	3.01	0.981
Asian/Pacific Islanders	Yes vs no	2.93	0.46	18.82	0.256

P < 0.05 indicates statistical significance. Multivariable regression models were adjusted for baseline demographics, hospital level characteristics and peripheral vascular disease, tobacco use disorder, prior myocardial infarction, prior transient ischemic attack, prior sudden cardiac arrest, prior venous thromboembolism, cancer, chronic kidney disease, acquired immune deficiency syndrome, alcohol abuse, drug abuse, cocaine, cannabis, depression, chronic pulmonary disease, hypothyroidism, other thyroid disorders, valvular disease, autoimmune conditions, coronavirus disease 2019. CI: Confidence interval; OR: Odds ratio.

DISCUSSION

This population-based observational study evaluated the association between metabolically healthy obesity and short-term in-hospital outcomes among patients hospitalized with acute ischemic stroke. Specifically, we investigated the impact of MHO on AIS hospitalization outcomes, including in-hospital mortality, length of stay, and hospital expenditure, and whether disparities existed based on age, sex, or race. Our nationwide analysis did not identify a statistically significant difference in in-hospital mortality between the MHO and non-MHO cohorts. However, lower adjusted odds were observed in females and younger age groups (18-44 years, 45-65 years) compared with males and older age groups (65 years and above), while higher odds were observed among Asians and Hispanics relative to Black and White respondents. Nevertheless, these disparities were not statistically significant (P > 0.05), indicating that MHO was not associated with an increased risk of AIS-related in-hospital mortality or poorer short-term in-hospital outcomes.

The prevalence of overweight and obesity in the global adult population continues to rise, with projections estimating 2.16 billion overweight and 1.12 billion obese individuals by 2030[12]. Obesity is a recognized independent risk factor for CVD. A large 2018 United States meta-analysis involving more than 1 million participants further strengthened the association between obesity and the risk of CAD and other CVDs, although its effect on disease outcomes remained inconclusive[13]. Similarly, evidence regarding different obesity phenotypes, particularly MHO, remains conflicting. Some studies have reported increased stroke prevalence in the MHO group compared with metabolically healthy non-obese individuals[5,14] as well as higher all-cause mortality[15], whereas others have suggested that MHO does not increase the risk of stroke[10,16,17] or all-cause mortality[18]. Our primary hypothesis aligns with the latter observations.

This absence of increased risk has been attributed in some studies to a possible protective effect of obesity on stroke outcomes, including in-hospital mortality and early readmissions[19], functional outcomes[20], and recurrence[21], when compared with non-obese individuals. This protective phenomenon is often referred to as the “obesity-stroke paradox”. However, this effect may not extend to all post-stroke neurological deficits[22], and not all studies have supported the existence of an “obesity-stroke paradox”[23]. Several metabolic and immunological mechanisms have been proposed to explain this phenomenon in AIS. In MHO, preserved insulin sensitivity, higher adiponectin levels, and lower pro-inflammatory cytokine activity may contribute to a more vasculoprotective profile[24]. In contrast, MUO is characterized by visceral adiposity, oxidative stress, endothelial dysfunction, and immune activation[25]. Additional mechanisms, including leptin resistance, altered lipid partitioning, and cardiometabolic resilience, may also play a role[4,24,25]. However, these hypotheses require further validation in well-phenotyped prospective cohorts with integrated metabolic and inflammatory biomarker data.

The obesity-stroke paradox may reflect the combined influence of demographic factors, cardiorespiratory fitness, genetic predisposition, body composition, and socioeconomic status in patients with MHO. Ethnic and racial differences may also contribute to these disparities, as non-White individuals have been associated with poorer discharge disposition[15], and some Asian populations, including Filipinos, have been reported to have higher mortality rates compared with White individuals, partially attributed to hemorrhagic transformation[26]. In our study, similar directional differences were observed across racial groups, although these did not reach statistical significance. In addition, patients with MHO have been reported to demonstrate healthier behavioral profiles, including lower rates of smoking and alcohol use[27], lower fat mass and trunk adiposity compared with MUO[28], and greater physical activity[16], all of which may contribute to a more favorable cardiometabolic profile. Inflammation may also play an important role, as some studies have shown similar inflammatory profiles in MHO[29], whereas others have reported a more favorable and less inflammatory state compared with MUO[30]. Despite these potentially protective features, MHO should not be considered a stable or benign phenotype. Patients with MHO have a higher likelihood of transitioning over time to a metabolically unhealthy state[31], and longitudinal studies suggest that both metabolically healthy non-obese and metabolically healthy obese phenotypes may progressively shift toward metabolic dysfunction[15,18]. Although BMI remains the most widely used screening tool for obesity, with current guideline definitions classifying overweight as a BMI of 25 kg/m² to 30 kg/m² and obesity as a BMI ≥ 30 kg/m²[32], it does not fully capture body composition or fat distribution. More precise approaches, such as dual-energy X-ray absorptiometry and android-to-gynoid fat ratio assessment, may better characterize adiposity, although they are less practical in routine or resource-limited settings. Therefore, BMI should be interpreted cautiously and not used as the sole marker of obesity-related risk.

Obesity should not be viewed as a single uniform risk state. With growing recognition of distinct obesity phenotypes, differentiating MHO from MUO is increasingly important for understanding disease risk and tailoring management strategies. Our findings support a more phenotype-based interpretation of obesity in AIS, as patients with MHO did not demonstrate worsened in-hospital outcomes despite having obesity. This distinction may help clinicians avoid overgeneralizing risk on the basis of anthropometric measures alone and instead focus on the presence or absence of metabolic dysfunction when evaluating prognosis and planning follow-up.

These findings also carry important clinical implications. While patients with MHO may not necessarily require intensified glycemic or blood pressure management solely based on obesity in the absence of metabolic abnormalities, they should still undergo regular surveillance of blood pressure, glycemic status, and lipid profiles, given the dynamic nature of metabolic health and the possibility of transition to MUO over time. Importantly, the administrative-data definition of MHO used in this study is not suitable for clinical phenotyping at the individual patient level, and rigorous metabolic assessment remains necessary in practice.

Lifestyle counseling, preventive care, and routine follow-up remain important for all overweight and obese individuals, but patients with metabolic abnormalities may warrant closer monitoring and more aggressive risk-factor modification. A personalized risk-stratification approach based on metabolic phenotype rather than obesity status alone may help optimize long-term vascular prevention and post-stroke care.

Limitations

Several limitations of this study should be acknowledged. First, the cross-sectional nature of the NIS precludes longitudinal follow-up and causal inference. In addition, the NIS does not provide modified Rankin Scale scores or other detailed functional outcome measures at discharge, limiting assessment of short-term and long-term disability, post-discharge outcomes, and readmission. Secondly, due to the administrative nature of the database, detailed clinical metrics such as laboratory values, BMI, lipid panels, and inflammatory markers were unavailable, and metabolically healthy obesity was defined using the absence of coded hypertension, diabetes mellitus, and hyperlipidemia rather than direct metabolic assessment, limiting phenotype granularity and introducing the possibility of phenotypic misclassification.

Furthermore, this analysis is primarily applicable to the United States population as it is based solely based on data from United States hospitals, which may limit generalizability to non-United States populations with differing health systems, risk factor profiles, and access to care. Importantly, obesity phenotypes may differ significantly across racial and ethnic groups, particularly among Asian populations, where normal BMI individuals may exhibit visceral adiposity or insulin resistance (“thin-fat phenotype”). The lack of granular racial stratification in our dataset further restricts our ability to explore phenotype-outcome interactions across populations. BMI categories were not available in the NIS database, which limited our ability to report BMI-specific distributions.

Finally, the identification of comorbidities and stroke subtype relied on ICD-10 coding, which is susceptible to misclassification or coding inconsistencies, although such codes have been validated for large-scale stroke epidemiological research.

CONCLUSION

In this nationwide propensity-matched study, patients with MHO hospitalized with AIS had a higher prevalence of selected baseline comorbidities, including hypothyroidism and chronic obstructive pulmonary disease. However, no statistically significant difference in in-hospital mortality was observed between MHO and non-MHO patients. Key takeaways from our study are as follows: (1) In this nationwide cohort, MHO was not associated with worse short-term in-hospital outcomes among patients with AIS; (2) These findings suggest that metabolic phenotype may be an important consideration when interpreting obesity-related risk in AIS, but the results should be interpreted cautiously because long-term functional recovery, recurrence, and longitudinal transition to metabolically unhealthy obesity could not be assessed; and (3) Further studies are needed to better define the long-term clinical implications of MHO in patients with stroke and to guide phenotype-based risk stratification and management.