Your privacy, your choice

We use essential cookies to make sure the site can function. We also use optional cookies for advertising, personalisation of content, usage analysis, and social media.

By accepting optional cookies, you consent to the processing of your personal data - including transfers to third parties. Some third parties are outside of the European Economic Area, with varying standards of data protection.

See our privacy policy for more information on the use of your personal data.

for further information and to change your choices.

Skip to main content

L-shaped association of body mass index with prognosis in individuals with sepsis: a multicenter cohort study

Abstract

Objective

The relationship between body mass index (BMI) and sepsis prognosis remains highly controversial and uncertain. This study investigated the association between BMI and prognosis in patients with sepsis.

Methods

This retrospective observational cohort study included adult patients admitted to the intensive care unit (ICU) with sepsis from Medical Information Mart for Intensive Care-IV version 2.2 (MIMIC-IV V2.2) and eICU Collaborative Research Database (eICU-CRD). The cut-off value of BMI was identified by the restricted cubic spline (RCS) curve and included patients were categorized into two groups: the low BMI group (< 28 kg/m2) and the high BMI group (≥ 28 kg/m2). The primary outcome was ICU mortality, and secondary outcomes were in-hospital and 28-day mortality. We performed the log-rank test to detect whether there is a difference in prognosis among different groups in two different cohorts. Multiple distinct models were used to validate the robustness of the results.

Results

There were 18,385 and 38,713 patients in the MIMIC-IV 2.2 and eICU-CRD cohorts, respectively. An L-shaped relationship was observed between BMI and ICU mortality in the primary analysis from MIMIC-IV 2.2. Similar relationships were found in eICU-CRD. When BMI was less than the cut-point, the risk of ICU mortality increased rapidly with decreasing BMI. When BMI was greater than the cut-point, the risk of ICU mortality levelled off as BMI increased. Sepsis patients with higher BMI values exhibited decreased ICU all-cause mortality rates (MIMIC-IV cohort: HR: 0.81, 95% CI 0.75–0.88, p < 0.001; eICU-CRD cohort: HR: 0.75, 95% CI 0.71–0.80, p < 0.001). Consistent trends were observed for both in-hospital mortality and 28-day mortality rates. The results remained robust in multiple distinct models and subgroup analyses.

Conclusion

An L-shaped relationship was observed between BMI and prognosis in septic patients, indicating that lower BMI values are significantly linked to increased mortality. Targeted nutritional interventions and close monitoring for patients with low BMI could potentially enhance their prognosis. Therefore, BMI can also be utilized to categorize the risk levels of patients with sepsis and effectively predict their prognosis.

Introduction

Sepsis is a life-threatening syndrome of organ dysfunction resulting from a dysregulated host response to infection [1]. In 2017, the global incidence of sepsis was estimated at 48.9 million cases, and the number of sepsis-associated deaths was 11 million, accounting for 19.7% of all global deaths [1]. Although antibiotic therapy and supportive care have significantly reduced sepsis mortality, regional disparities remain significant and represent a major global public health burden. As understanding of sepsis has increased, there has been growing interest in factors associated with sepsis prognosis. Among these, the relationship between body mass index (BMI) and sepsis prognosis has become a significant area of research.

As living standards continue to improve worldwide, the proportion of people with a high BMI is increasing, leading to obesity reaching pandemic proportions [2]. Obesity significantly increases the risk of developing chronic diseases such as type 2 diabetes, hypertension, fatty liver disease, and cardiovascular disease, which can reduce both quality of life and life expectancy [3, 4]. Nonetheless, recent studies have indicated that obese patients may have a better prognosis in certain conditions, such as sepsis [5, 6]. In a prospective, population-based, continuous observational study conducted in Sweden, which included 2,196 patients with severe sepsis and septic shock, researchers found that obesity was associated with increased survival rates in severe bacterial infections [7]. A meta-analysis supports this, showing that overweight patients had a reduced risk of mortality following sepsis. In contrast, severely obese patients also did not exhibit an increased risk of death [8]. However, a systematic review reports uncertain associations between obesity and sepsis prognosis, highlighting the complexity of this relationship and the need for further research [9]. This so-called "obesity paradox" has garnered widespread attention. A study from 90,760 hospitalized sepsis patients across the United States revealed an inverse J-shaped relationship between BMI and prognosis [10]. A study by Lin et al. [11] analyzed data from 7967 septic patients in the MIMIC-III database and identified a U-shaped relationship between BMI and short-term prognosis. However, the exclusion of patients with BMI values below 10 and above 60 may restrict the applicability of these findings to the full range of BMI values. Additionally, the lack of external validation raises concerns regarding their broader applicability. Although these studies have initially explored the relationship between BMI and sepsis prognosis, the findings remain controversial. In clinical practice, we often observe that patients with a higher BMI tend to have better prognoses. Meawhile, there is still a lack of evidence from multicenter, large-scale studies to conclusively establish the relationship between BMI and sepsis prognosis. Therefore, our study utilized two large publicly available databases (Medical Information Mart for Intensive Care-IV version 2.2 and eICU Collaborative Research Database [eICU-CRD]) to further investigate the relationship between BMI and the prognosis of patients with sepsis. Specifically, we hypothesize that BMI is non-linearly associated with sepsis outcomes, with both underweight and obesity potentially influencing mortality in different populations. To test this hypothesis, subgroup analyses will be conducted based on BMI and clinical factors such as age, gender, race, and comorbidities. Through this investigation, we aim to not only enhance our understanding of the prognostic implications of BMI in sepsis but also to inform clinical practices that could mitigate the risk of mortality associated with this critical condition.

Materials and methods

Data source

This multicenter retrospective observational cohort study utilized data from two different cohorts. The original cohort for analysis was sourced from the MIMIC-IV V2.2 database, while the validation cohort was drawn from the eICU-CRD database. MIMIC-IV is a longitudinal, single-center database of adult patients admitted to Beth Israel Deaconess Medical Center between 2008 and 2019. Its data-sharing initiative has been approved by the Institutional Review Board of Beth Israel Deaconess Medical Center with a waiver of informed consent. Users are required to pass an exam to register and gain access after receiving approval from the MIMIC-IV database administrator [12]. The eICU-CRD database is a multicenter ICU resource developed by Philips Healthcare that includes more than 200,000 ICU admissions from 208 different ICUs in the United States between 2014 and 2015. All data tables are de-identified in accordance with the Safe Harbor provisions of the US Health Insurance Portability and Accountability Act (HIPAA). All protected health information, including hospital and unit identifiers, has been removed [13]. Kunping Cui was certified to extract data from the two databases mentioned above for research after passing the “Collaborative Institutional Training Initiative (CITI)” training course on the National Institutes of Health (NIH) website (certification number: 59589454). We developed detailed data extraction steps and conducted pilot extractions prior to the official data extraction phase to test and refine the clarity and operationalization of these steps. In addition, we used a variety of validation measures to ensure the accuracy of the data. This included independent reviews of key data points and consistency checks using statistical software to identify and correct possible input errors or inconsistencies.

Study population

The inclusion criteria met the definition of Sepsis 3.0 criteria [14], which was defined as a suspected infection combined with an acute increase in Sequential Organ Failure Assessment (SOFA) score ≥ 2. The exclusion criteria were as follows: ICU LOS < 24 h (Mitigate the potential of including individuals whose clinical conditions are either too severe, leading to rapid mortality, or too mild to genuinely represent the cohort of interest) [15, 16], patients < 18 years old, missing BMI data or incomplete height and weight measurements for BM calculation, ICD information was not available (Given that comorbidities are crucial variables to consider in our analysis, we made the decision to exclude the patients with missing ICD codes to ensure the integrity and robustness of our results), survival duration information was incorrect or unavailable. We included only the first ICU stay for analysis in patients with multiple ICU admissions. There were 32,970 and 48,797 ICU patients based on sepsis 3.0 criteria from the MIMIC-IV v2.2 and eICU database. A total of 18,385 and 38,713 patients from MIMIC-IV and eICU were ultimately included in the final analysis (Fig. 1).

Fig. 1
figure 1

Inclusion and exclusion flowchart of the study

Post-hoc power analysis demonstrated that our sample size (MIMIC-IV cohort: N = 18,385; eICU cohort: N = 38,713) provided more than 80% power (MIMIC-IV cohort: 84%; eICU cohort: 99%) to detect the observed difference in mortality between BMI groups (MIMIC-IV cohort: effect size h = 0.044, α = 0.05; eICU cohort: effect size h = 0.054, α = 0.05). This exceeds the conventional 80% threshold, confirming adequate statistical power for our primary analysis.

Data extraction

In order to obtain the patient data necessary for the study, we collected demographic information, vital signs, scoring systems, and laboratory results within the first 24 h of ICU admission. To maximize the completeness of our BMI data, we prioritized extracting it from the bmi omr table, even if height or weight data were missing from the database. For patients without BMI data in the bmi omr table, we then retrieved height and weight data from bmi cal table to calculate BMI. Structured Query Language (SQL) codes were developed and tested using DBeaver Community version 23.1.2. The DBI package was then utilized to execute these codes, creating relevant tables in the MIMIC-IV and eICU databases, along with associated variables in the R global environment for data collection.

Exposure and clinical outcomes

The nonlinear relationship between BMI and ICU mortality was examined using a restricted cubic spline (RCS) curve based on the Cox regression model using rms package, determining the optimal threshold for patient grouping as the exposure factor. In the present investigation, the primary outcome was set as all-cause ICU mortality, with the secondary outcome being all-cause in-hospital and 28-day all-cause mortality (not available in the eICU-CRD database). The follow-up period started with the patients' admission and concluded upon their discharge or death.

Covariates

In the current study, we included demographic and admission data [e.g., age, gender, body mass index (BMI), race (Asian, Black, White, and Other), service unit, smoking (not available in the eICU-CRD database), alcohol (not available in the eICU-CRD database), geriatric nutritional risk index (GNRI) [17], SOFA score, simplified acute physiological score II (SAPS II), charlson score)], therapeutic interventions [e.g., mechanical ventilation use (MV), time of MV initiation, vasopressor use, time of vasopressor use, renal replacement therapy, sedative use], pre-existing comorbid conditions [e.g., heart failure (HF), hypertension, diabetes, atrial fibrillation (AFIB), renal disease, liver disease, chronic obstructive pulmonary disease (COPD), coronary artery disease (CAD), stroke], vital signs [e.g., mean arterial pressure (MAP), heart rate, temperature, respiratory rate], along with laboratory tests [e.g., white blood cell (WBC) count, hemoglobin, platelet counts, sodium, potassium, chloride, bicarbonate, blood urea nitrogen (BUN), lactate, pH, partial pressure of oxygen (PO2), partial pressure of carbon dioxide (PCO2)] during the first 24 h in the ICU admission.

Feature selection

To address the risk of potential overfitting, we used the Variance Inflation Factor (VIF) to assess multicollinearity among covariates. Variables with a VIF > 5 were considered to be highly correlated with the exposure factors and required further adjustment. Three widely used methods, univariate logistic regression, Cox regression, and Boruta algorithm were used to identify which factors might influence the ICU mortality in sepsis. These rigorous methods are crafted to validate the significance of the selected variables in augmenting the explanatory power of the research model.

The univariate logistic regression and Cox regression approaches leverage P-values as a sieve for variable significance. Specifically, within the univariate logistic regression framework, a predictor variable is deemed statistically significant with a P-value threshold of less than 0.1, thereby warranting its progression to subsequent analytical phases. The univariate Cox regression mirrors this criterion, employing P-values to preemptively filter variables for their prognostic relevance.

Divergent from the regression-based approaches, the Boruta algorithm emerges as a sophisticated feature selection mechanism, originating from the ensemble of random forest algorithms. This algorithm quantitatively evaluates the intrinsic importance of each feature through an iterative process of constructing multiple random forest models, which meticulously estimate feature importance scores. A novel aspect of the Boruta algorithm is the incorporation of 'shadow features,' which serve as a comparative baseline to accurately discern the substantial impact of each feature. Integral to the Boruta algorithm is the imposition of the Bonferroni correction, a statistical adjustment designed to mitigate the inflated risk of Type I errors inherent in multiple hypothesis testing scenarios. By recalibrating the P-value threshold post-Bonferroni correction, features with a P-value inferior to 0.01 are accorded significant status, thereby earning their integration into the final research model as variables of consequence [18].

Statistical analysis

Descriptive statistics were presented as median (interquartile range [IQR]) or number (percentage) where appropriate. Continuous variables were compared using Kruskal–Wallis tests (data with skewed distributions) and categorical variables were compared using the chi-squared test. Variables with missing data are common in the MIMIC-IV and eICU databases. Variables with missing values exceeding 40% in the two databases will not be included as covariates in the model for analysis. Multiple imputations were performed using the “mice” package in R software to address the missing values [19]. The unadjusted log-rank test was employed with the survival package to plot survival curves.

The statistical analysis of this study consisted mainly of the following three steps to assess the relationship between BMI and prognosis in sepsis patients. First, after adjustment for all covariates, the non-linear association of BMI with ICU mortality was modelled using 3-knot (the number of knots was chosen based on the sample size and the need to prevent overfitting) restricted cubic spline (RCS) (the 25th, 50th, and 75th percentiles) with multivariable logistic regression analysis and multivariable Cox regression analysis [20]. According to the cut-off value of BMI determined by the RCS results, patients were divided into low-BMI and high-BMI groups. Kaplan–Meier methods were used to plot survival curves. Differences in survival between groups were compared using the log-rank test. Second, we ran seven different association inference models based on variables selected by different methods to assess the robustness of the study findings. Finally, potential modifications to the association between BMI and the prognosis of sepsis were assessed, including the following variables: age (< 65 vs. ≥ 65 years), gender (female vs. male), race (Asian, Black, White,and Other), SOFA scores (< 5 vs, ≥ 5), SAPS II score (< 49 vs. ≥ 49), MV use (yes vs. no), vasopressors use Cox regression analysis and interactions between the subgroups and BMI were examined using likelihood ratio testing.

All analyses were performed using a statistical programming language (R, version 4.3.2; R Development Core Team). Statistical significance was determined at the level of 0.05, and all P values were based on two-tailed tests.

Results

Non-linear relationship between BMI and ICU mortality in sepsis patients

After adjusting for all covariates, an L-shaped relationship was observed between BMI and ICU mortality in sepsis patients through RCS with multivariate COX regression or multivariate logistic regression analysis in the MIMIC-IV cohort (nonlinear p < 0.01, Fig. 2A and B). Similar results were consistently observed in the eICU cohort (Fig. 2C and D), indicating that a lower BMI may be a risk factor for poor prognosis in sepsis patients. Interestingly, the cut-off value of BMI was set at 28 kg/m2, which was consistently identified by the RCS curve in both the MIMIC-IV and eICU cohorts. When BMI was less than 28 kg/m2, the risk of ICU mortality increased rapidly with decreasing BMI. When BMI was greater than 28 kg/m2, the risk of ICU mortality decreased slowly with increasing BMI. Patients were categorized into two groups based on this cut-off value: the low BMI group (< 28 kg/m2) and the high BMI group (≥ 28 kg/m2).

Fig. 2
figure 2

The nonlinear relationship between BMI and the risk of ICU mortality was modeled using multivariate Cox regression and multivariate logistic regression with 3-knot RCS analyses. A and B The MIMIC-IV cohort. C and D The eICU-CRD cohort

Baseline characteristics

The baseline characteristics are listed in detail in Table 1. The low BMI groups of the MIMIC-IV and eICU-CRD database included 9540 and 20,699 patients, respectively, and the high BMI groups included 8845 and 18,014 patients. Patients were younger in the high BMI group than in the low BMI group in the MIMIC-IV database (66.00 [56.00, 75.00] vs. 68.00 [57.00, 79.00] years) and in the eICU-CRD (65.00 [55.00, 75.00] vs. 69.00 [56.00, 80.00] years). In the MIMIC-IV database and eICU-CRD, female patients in the high BMI group accounted for 40.51% and 50.02%, respectively. While the low BMI group accounted for 39.25% and 44.91%; those in the high BMI group had a higher GNRI than those in the high BMI group in the MIMIC-IV database (67.14 [61.80, 76.06] vs 50.12 [45.45, 54.04]) and in the eICU-CRD (68.36 [61.69, 79.03] vs 48.12 [43.13, 52.57]).

Table 1 Basic demographic characteristics of the enrolled patients stratified by BMI values in the MIMIC-IV and eICU-CRD cohorts

In both the MIMIC-IV and eICU cohorts, we consistently observed that a higher proportion of patients in the high BMI group required MV, renal replacement therapy, and sedation compared to the low BMI group. Moreover, the high BMI group exhibited a greater prevalence of comorbidities, including HF, hypertension, diabetes, renal disease, and CAD, while demonstrating a significantly lower incidence of stroke.

Primary and secondary outcome

The ICU mortality rates in the MIMIC-IV cohort and eICU-CRD cohort were 12.49% and 9.36%, respectively. The in-hospital mortality rates in the MIMIC-IV cohort and eICU-CRD cohort were 16.85% and 15.82%, respectively. The 28-day mortality rate was 19.13% in MIMIC-IV cohort (Table 1).

The incidence of ICU mortality was 13.19% (1,258 patients) in the low BMI group and 11.75% (1039 patients) in the high BMI group in the MIMIC-IV cohort. In the eICU cohort, the incidence was 10.08% (2087 patients) for the low BMI group and 8.53% (1536 patients) for the high BMI group. The Kaplan–Meier survival curve showed that the probability of ICU and in-hospital mortality of sepsis patients in the high BMI group was lower than that of patients in the low BMI group (Fig. 3). The unadjusted log-rank test further revealed that the high BMI group exhibited reduced all-cause mortality rates, including ICU mortality (MIMIC-IV cohort: HR: 0.81, 95% CI 0.75–0.88, p < 0.001; eICU-CRD cohort: HR: 0.75, 95% CI 0.71–0.80, p < 0.001), in-hospital mortality (MIMIC-IV cohort: HR: 0.80, 95% CI 0.75–0.86, p < 0.001; eICU-CRD cohort: HR: 0.77, 95% CI 0.73–0.81, p < 0.001) and 28-day mortality (MIMIC-IV cohort: HR: 0.74, 95% CI 0.69–0.79), p < 0.001) (Table S1-S5).

Fig. 3
figure 3

Unadjusted Kaplan–Meier survival curve for primary and secondary outcome of original cohort. A Unadjusted Kaplan–Meier survival curve for ICU mortality in the MIMIC-IV cohort. B Unadjusted Kaplan–Meier survival curve for ICU mortality in the eICU-CRD cohort. C Unadjusted Kaplan–Meier survival curve for in-hospital mortality in the MIMIC-IV cohort. D Unadjusted Kaplan–Meier survival curve for in-hospital mortality in the eICU-CRD cohort

High BMI is correlated with low risk of mortality

To validate the robustness of our findings, we employed a series of models to perform the sensitivity analysis. Besides, to address the risk of potential overfitting, we used the VIF to assess multicollinearity among covariates. All variables with a VIF < 5 in our study (Table S6-S10). These models comprise the following: adjusted multivariate Cox model and adjusted multivariate Logistic model [covariates selected by all covariates, covariates selected by univariable-selected, and covariates selected by random forest algorithm (Fig. 4)]. As summarized in Table 2, all estimated models converge towards the same conclusion: high BMI group exhibit the lower mortality risk (including ICU, in-hospital, and 28-day mortality). Detailed estimates of covariates for each model can be found in Table S11-40.

Fig. 4
figure 4

Importance of each variable for ICU mortality in sepsis according to the Boruta feature selection. A MIMIC-IV cohort; B eICU-CRD cohort

Table 2 Primary and secondary outcome analyses with different models for the MIMIC-IV and eICU-CRD cohorts

Subgroup analysis

Subgroup analyses were meticulously conducted to assess the uniformity of the correlation between BMI and ICU mortality rates across diverse demographic and clinical subpopulations. As illustrated in Fig. 5, within the MIMIC-IV cohort, individuals with a higher BMI exhibited a reduced risk of ICU mortality compared to their lower BMI counterparts. This trend was consistently observed across most subgroups, with the exception of those aged < 65 years, individuals of Asian descent, and those without hypertension. Notably, no significant interactive effects were detected across these subgroups (All p > 0.05). Similarly, the eICU-CRD cohort demonstrated analogous trends in ICU mortality rates. The high BMI group consistently presented a lower risk profile compared to the low BMI group, reinforcing the robustness of the observed association across different patient populations.

Fig. 5
figure 5

Forest plot of subgroup analysis for ICU mortality of sepsis. A MIMIC-IV cohort; B eICU-CRD cohort

Discussion

In this study, we investigated the association BMI and prognosis in patients with sepsis using two international cohorts. In the primary analysis, we observed an L-shaped relationship between the BMI and ICU mortality in patients with sepsis. A higher BMI was significantly associated with better short-term prognosis of sepsis, even after adjusting for potential confounders. Given the single-center study of the original cohort (MIMIC-IV), we sought to strengthen the robustness of our findings by validating them with a multicenter eICU in the USA. These further analyses confirmed our initial observations that higher BMI decreased the mortality risk in sepsis patients. These validations not only reinforced our initial observations but also underscored the potential of BMI as a critical factor for clinicians in the early prediction of sepsis prognosis.

Obesity is becoming an increasingly prevalent health condition as the global socioeconomic development. A recent meta-analysis found that a 21–26% decreased risk of mortality in overweight and obese BMIs patients with sepsis or septic shock [21]. A multicenter prospective national cohort study including 6,424 Asian patients with sepsis found that obesity is associated with better in-hospital survival and lower Clinical Frailty Scale in septic patients for Asian populations [22]. Another study including 55,038 adult sepsis cases in the United States also demonstrated the protective role of obesity in sepsis, which lower short-term mortality in patients with higher BMI compared with those with normal BMI and higher short-term mortality in those with low BMI [23]. Additionally, an experimental animal study confirmed that during critical illness, premorbid obesity can improve prognosis by optimizing the utilization of stored lipids and mitigating muscle atrophy and weakness [24]. Similarly, our findings are consistent with these previous observations and further support the contention that obesity is a protective mortality factor for sepsis patients. These studies consistently confirm that a lower BMI is associated with a higher risk of death from sepsis. This association is likely due to significant changes in immune function associated with cachexia, primarily through the release of pro-inflammatory cytokines such as IL-6 and TNF-α. These cytokines not only promote muscle catabolism but also contribute to systemic inflammation, further compromising the immune response. In addition, cachexia impairs the functionality of key immune cells, including T cells and macrophages, which has a negative impact on disease prognosis [25, 26].

Previous studies utilizing the MIMIC III database explored the relationship between BMI and sepsis prognosis, suggesting a U-shaped association with short-term outcomes [11]. However, the generalizability and statistical robustness of these findings require further validation due to variations in sample size and the exclusion of patients with BMI values below 10 and above 60, as well as differing medical standards and awareness of weight management. Additionally, the MIMIC III data, covering 2001 to 2012, may limit the applicability of its conclusions. In contrast, our study analyzed the MIMIC IV database, spanning 2008 to 2019, adopting a more inclusive approach by validating the BMI-sepsis prognosis relationship through a multicenter eICU database. Our findings indicate an L-shaped relationship between BMI and sepsis prognosis, suggesting that a high BMI does not increase mortality risk and may instead serve as a protective factor in this context. The factors contributing to the inconsistencies in research findings may be complex. First, Sepsis leads to an increased basal metabolic rate, resulting in significantly increased energy expenditure, particularly during the early stages of infection [27, 28]. This metabolic alteration not only affects the body's energy supply, but also results in muscle protein breakdown and abnormal fat metabolism, ultimately contributing to weight loss and malnutrition [29]. Additionally, sepsis can impair cellular energy synthesis, especially in vital organs such as the heart and kidneys, further compromising their function [30]. However, obese patients tend to have higher nutritional reserves, which may provide more energy and nutritional support to the body during sepsis, thus contributing to the recovery of patients [24]. Second, there is a close relationship between adipose tissue and the inflammatory response. The development of sepsis is accompanied by the release of numerous inflammatory mediators, particularly cytokines such as tumor necrosis factor (TNF) and interleukins (IL-1, IL-6). These inflammatory mediators not only trigger systemic inflammatory responses but also lead to endothelial dysfunction, increased microvascular permeability, and organ dysfunction [31]. Adipose tissue secretes various anti-inflammatory mediators, including leptin, interleukin-10 (IL-10), and soluble tumor necrosis factor receptor 2 (sTNFR2). Among these mediators, IL-10 is particularly important, as it stimulates the release of soluble TNF-α and IL-1 receptor antagonists. Furthermore, IL-10 effectively reduces the levels of pro-inflammatory cytokines such as IL-6, gamma interferon (IFN-γ), and TNF-α, thereby inhibiting the activity of Th1 cells and limiting the amplification of the inflammatory response [32, 33]. Third, adipose tissue is a complex endocrine organ capable of producing and releasing various biologically active molecules, including cytokines and adipokines, which play a crucial role in the regulation of metabolism and immunity [34]. Finally, although obese people are at higher risk of metabolic and cardiovascular diseases, which may lead to adverse outcomes, improvements in economic and medical standards have effective managed chronic diseases related to obesity. As a result, the coexistence of these conditions has not further increased the risk of mortality.

Our study has several important clinical implications. First, by monitoring BMI, clinicians can identify high-risk patients with low BMI, allowing early intervention and treatment to optimise management strategies and reduce the risk of mortality. Our previous research has confirmed that low BMI is an independent risk factor for the prognosis of tuberculosis combined with severe community-acquired pneumonia [35]. In our sepsis study, we observed consistent results, further emphasizing the importance of stratified management based on BMI. Second, personalised nutritional support plans should be tailored according to BMI. Our research showed that patients with a BMI below 28 kg/m2 have a higher risk of mortality, highlighting the need to focus on underweight patients while not overlooking sepsis patients with a normal BMI. A study showed that higher protein intake in the first week of sepsis was associated with lower in-hospital mortality, and higher energy intake was also associated with lower 30-day mortality [36]. In addition, another study of 2,088 hospitalised, non-critically ill patients found that personalised, protocol-guided nutritional support significantly reduced mortality compared with standard care [37].

This study has several limitations that warrant thorough discussion. First, as with all observational studies, uncontrolled confounding variables may exist. Despite our efforts to adjust for covariates to establish a robust association between BMI and mortality, we acknowledge the potential for residual confounding factors that may not have been measured. This limitation is particularly important given the complex nature of sepsis and its many influencing variables. Second, our study's retrospective nature and the assessment of included parameters only at baseline may weaken our findings. Additionally, the decision to exclude patients with ICU stays of less than 24 h was made to avoid including individuals with rapid mortality or minimal clinical significance; however, we recognize that this exclusion could introduce survivorship bias, limiting the generalizability of our results. Furthermore, the presence of significant missing data raises concerns about selection bias, which cannot be entirely avoided. We emphasize the need for further validation of our findings through prospective multicenter studies to enhance the reliability and applicability of our results. Lastly, while we presented our findings within the context of U.S. populations, we acknowledge that the generalizability of our results to non-ICU populations or settings outside the U.S. is limited. Specifically, the applicability of our findings to low- and middle-income countries with different BMI distributions remains uncertain and warrants further investigation. Moreover, we recognize that the MIMIC-IV and eICU-CRD databases, despite their rich detail, are confined to specific healthcare environments. This confinement potentially restricts the diversity of the patient populations included, which may not accurately reflect the broader global demographic spectrum.

Conclusion

In conclusion, we found that baseline BMI independently predicts mortality in patients with sepsis, with lower BMI significantly associated with a higher risk of mortality, while higher BMI is linked to a better prognosis. Baseline BMI may assist clinicians in identifying patients at high risk of mortality. Therefore, we propose incorporating baseline BMI into risk stratification systems for sepsis.

Data availability

The data supporting the findings of this study are available from the first author, Kunping Cui, upon reasonable request.

References

  1. Rudd KE, Johnson SC, Agesa KM, Shackelford KA, Tsoi D, Kievlan DR, et al. Global, regional, and national sepsis incidence and mortality, 1990–2017: analysis for the Global Burden of Disease Study. Lancet. 2020;395(10219):200–11. https://doi.org/10.1016/S0140-6736(19)32989-7.

    Article  PubMed  PubMed Central  Google Scholar 

  2. Swinburn BA, Sacks G, Hall KD, McPherson K, Finegood DT, Moodie ML, et al. The global obesity pandemic: shaped by global drivers and local environments. Lancet. 2011;378(9793):804–14. https://doi.org/10.1016/S0140-6736(11)60813-1.

    Article  PubMed  Google Scholar 

  3. Bluher M. Obesity: global epidemiology and pathogenesis. Nat Rev Endocrinol. 2019;15(5):288–98. https://doi.org/10.1038/s41574-019-0176-8.

    Article  PubMed  Google Scholar 

  4. Khan SS, Ning H, Wilkins JT, Allen N, Carnethon M, Berry JD, et al. Association of body mass index with lifetime risk of cardiovascular disease and compression of morbidity. JAMA Cardiol. 2018;3(4):280–7. https://doi.org/10.1001/jamacardio.2018.0022.

    Article  PubMed  PubMed Central  Google Scholar 

  5. Pepper DJ, Sun J, Welsh J, Cui X, Suffredini AF, Eichacker PQ. Increased body mass index and adjusted mortality in ICU patients with sepsis or septic shock: a systematic review and meta-analysis. Crit Care. 2016;20(1):181. https://doi.org/10.1186/s13054-016-1360-z.

    Article  PubMed  PubMed Central  Google Scholar 

  6. Wacharasint P, Boyd JH, Russell JA, Walley KR. One size does not fit all in severe infection: obesity alters outcome, susceptibility, treatment, and inflammatory response. Crit Care. 2013;17(3):R122. https://doi.org/10.1186/cc12794.

    Article  PubMed  PubMed Central  Google Scholar 

  7. Alsio A, Nasic S, Ljungstrom L, Jacobsson G. Impact of obesity on outcome of severe bacterial infections. PLoS ONE. 2021;16(5): e0251887. https://doi.org/10.1371/journal.pone.0251887.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  8. Wang S, Liu X, Chen Q, Liu C, Huang C, Fang X. The role of increased body mass index in outcomes of sepsis: a systematic review and meta-analysis. BMC Anesthesiol. 2017;17(1):118. https://doi.org/10.1186/s12871-017-0405-4.

    Article  PubMed  PubMed Central  Google Scholar 

  9. Trivedi V, Bavishi C, Jean R. Impact of obesity on sepsis mortality: a systematic review. J Crit Care. 2015;30(3):518–24. https://doi.org/10.1016/j.jcrc.2014.12.007.

    Article  PubMed  Google Scholar 

  10. Lebovitz S, Rozen G, Abu Ghosh Z, Korem M, Elinav H, Zayyad H, et al. The relationship between body mass index and in-hospital mortality in bacteremic sepsis. J Clin Med. 2023. https://doi.org/10.3390/jcm12113848.

    Article  PubMed  PubMed Central  Google Scholar 

  11. Lin S, Ge S, He W, Zeng M. Association between body mass index and short-term clinical outcomes in critically ill patients with sepsis: a real-world study. Biomed Res Int. 2020;2020:5781913. https://doi.org/10.1155/2020/5781913.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  12. Johnson AEW, Bulgarelli L, Shen L, Gayles A, Shammout A, Horng S, et al. Author Correction: MIMIC-IV, a freely accessible electronic health record dataset. Sci Data. 2023;10(1):31. https://doi.org/10.1038/s41597-023-01945-2.

    Article  PubMed  PubMed Central  Google Scholar 

  13. Pollard TJ, Johnson AEW, Raffa JD, Celi LA, Mark RG, Badawi O. The eICU Collaborative Research Database, a freely available multi-center database for critical care research. Sci Data. 2018;5:180178. https://doi.org/10.1038/sdata.2018.178.

    Article  PubMed  PubMed Central  Google Scholar 

  14. Singer M, Deutschman CS, Seymour CW, Shankar-Hari M, Annane D, Bauer M, et al. The third international consensus definitions for sepsis and septic shock (Sepsis-3). JAMA. 2016;315(8):801–10. https://doi.org/10.1001/jama.2016.0287.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  15. Dou J, Guo C, Wang Y, Peng Z, Wu R, Li Q, et al. Association between triglyceride glucose-body mass and one-year all-cause mortality of patients with heart failure: a retrospective study utilizing the MIMIC-IV database. Cardiovasc Diabetol. 2023;22(1):309. https://doi.org/10.1186/s12933-023-02047-4.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  16. Wang J, Li H, Luo H, Shi R, Chen S, Hu J, et al. Association between serum creatinine to albumin ratio and short- and long-term all-cause mortality in patients with acute pancreatitis admitted to the intensive care unit: a retrospective analysis based on the MIMIC-IV database. Front Immunol. 2024;15:1373371. https://doi.org/10.3389/fimmu.2024.1373371.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  17. Bouillanne O, Morineau G, Dupont C, Coulombel I, Vincent J-P, Nicolis I, et al. Geriatric Nutritional Risk Index: a new index for evaluating at-risk elderly medical patients. Am J Clin Nutr. 2005;82(4):777–83.

    Article  CAS  PubMed  Google Scholar 

  18. Degenhardt F, Seifert S, Szymczak S. Evaluation of variable selection methods for random forests and omics data sets. Brief Bioinform. 2019;20(2):492–503. https://doi.org/10.1093/bib/bbx124.

    Article  PubMed  Google Scholar 

  19. Jakobsen JC, Gluud C, Wetterslev J, Winkel P. When and how should multiple imputation be used for handling missing data in randomised clinical trials—a practical guide with flowcharts. BMC Med Res Methodol. 2017;17(1):162. https://doi.org/10.1186/s12874-017-0442-1.

    Article  PubMed  PubMed Central  Google Scholar 

  20. Liu H, Wang L, Chen C, Dong Z, Yu S. Association between dietary niacin intake and migraine among American adults: National Health and Nutrition Examination Survey. Nutrients. 2022. https://doi.org/10.3390/nu14153052.

    Article  PubMed  PubMed Central  Google Scholar 

  21. Bai L, Huang J, Wang D, Zhu D, Zhao Q, Li T, et al. Association of body mass index with mortality of sepsis or septic shock: an updated meta-analysis. J Intensive Care. 2023;11(1):27. https://doi.org/10.1186/s40560-023-00677-0.

    Article  PubMed  PubMed Central  Google Scholar 

  22. Yeo HJ, Kim TH, Jang JH, Jeon K, Oh DK, Park MH, et al. Obesity paradox and functional outcomes in sepsis: a multicenter prospective study. Crit Care Med. 2023;51(6):742–52. https://doi.org/10.1097/CCM.0000000000005801.

    Article  PubMed  PubMed Central  Google Scholar 

  23. Pepper DJ, Demirkale CY, Sun J, Rhee C, Fram D, Eichacker P, et al. Does obesity protect against death in sepsis? A retrospective cohort study of 55,038 adult patients. Crit Care Med. 2019;47(5):643–50. https://doi.org/10.1097/CCM.0000000000003692.

    Article  PubMed  PubMed Central  Google Scholar 

  24. Goossens C, Marques MB, Derde S, Vander Perre S, Dufour T, Thiessen SE, et al. Premorbid obesity, but not nutrition, prevents critical illness-induced muscle wasting and weakness. J Cachexia Sarcopenia Muscle. 2017;8(1):89–101. https://doi.org/10.1002/jcsm.12131.

    Article  PubMed  Google Scholar 

  25. Baazim H, Antonio-Herrera L, Bergthaler A. The interplay of immunology and cachexia in infection and cancer. Nat Rev Immunol. 2022;22(5):309–21. https://doi.org/10.1038/s41577-021-00624-w.

    Article  CAS  PubMed  Google Scholar 

  26. Wu Q, Liu Z, Li B, Liu YE, Wang P. Immunoregulation in cancer-associated cachexia. J Adv Res. 2024;58:45–62. https://doi.org/10.1016/j.jare.2023.04.018.

    Article  CAS  PubMed  Google Scholar 

  27. Blangy-Letheule A, Vergnaud A, Dupas T, Rozec B, Lauzier B, Leroux AA. Spontaneous sepsis in adult horses: from veterinary to human medicine perspectives. Cells. 2023. https://doi.org/10.3390/cells12071052.

    Article  PubMed  PubMed Central  Google Scholar 

  28. He FF, Wang YM, Chen YY, Huang W, Li ZQ, Zhang C. Sepsis-induced AKI: from pathogenesis to therapeutic approaches. Front Pharmacol. 2022;13:981578. https://doi.org/10.3389/fphar.2022.981578.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  29. Murata S, Yamashita H, Kido S, Harada D, Ohtsuru S, Sato N. Dynamic metabolic changes observed in an LPS-induced systemic inflammation rat model using continuous long-term indirect calorimetry experiments. Shock. 2023;60(1):130–6. https://doi.org/10.1097/SHK.0000000000002144.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  30. Aziz M, Wang P. Glucocorticoid resistance and hyperlactatemia: a tag team to worsen sepsis. Cell Metab. 2021;33(9):1717–8. https://doi.org/10.1016/j.cmet.2021.08.007.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  31. Cross D, Drury R, Hill J, Pollard AJ. Epigenetics in sepsis: understanding its role in endothelial dysfunction, immunosuppression, and potential therapeutics. Front Immunol. 2019;10:1363. https://doi.org/10.3389/fimmu.2019.01363.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  32. Taylor EB. The complex role of adipokines in obesity, inflammation, and autoimmunity. Clin Sci. 2021;135(6):731–52. https://doi.org/10.1042/CS20200895.

    Article  CAS  Google Scholar 

  33. Yeung ST, Ovando LJ, Russo AJ, Rathinam VA, Khanna KM. CD169+ macrophage intrinsic IL-10 production regulates immune homeostasis during sepsis. Cell Rep. 2023;42(3):112171. https://doi.org/10.1016/j.celrep.2023.112171.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  34. Karampela I, Christodoulatos GS, Dalamaga M. The role of adipose tissue and adipokines in sepsis: inflammatory and metabolic considerations, and the obesity paradox. Curr Obes Rep. 2019;8(4):434–57. https://doi.org/10.1007/s13679-019-00360-2.

    Article  PubMed  Google Scholar 

  35. Cui K, Mao Y, Feng S, Luo H, Yang J, Bai L. Development and validation of a risk mortality prediction model for patients with pulmonary tuberculosis complicated by severe community-acquired pneumonia in the intensive care unit. Infect Drug Resist. 2024;17:3113–24. https://doi.org/10.2147/IDR.S459290.

    Article  PubMed  PubMed Central  Google Scholar 

  36. Cha JK, Kim HS, Kim EJ, Lee ES, Lee JH, Song IA. Effect of early nutritional support on clinical outcomes of critically ill patients with sepsis and septic shock: a single-center retrospective study. Nutrients. 2022. https://doi.org/10.3390/nu14112318.

    Article  PubMed  PubMed Central  Google Scholar 

  37. Schuetz P, Fehr R, Baechli V, Geiser M, Deiss M, Gomes F, et al. Individualised nutritional support in medical inpatients at nutritional risk: a randomised clinical trial. Lancet. 2019;393(10188):2312–21. https://doi.org/10.1016/S0140-6736(18)32776-4.

    Article  PubMed  Google Scholar 

Download references

Acknowledgements

We thank the Ascetic Practitioners in Critical Care (APCC) team, and the easy Data Science for Medicine (easyDSM) team for sharing their knowledge and codes in big data of critical care, along with the cross-platform Big Data Master of Critical Care (BDMCC) software (https://github.com/ningyile/BDMCC_APP). We especially appreciate the MIMIC-IV and eICU official team's efforts to open-source the database and codes.

Funding

This research received no external funding.

Author information

Authors and Affiliations

Authors

Contributions

All authors made a significant contribution to the work reported, whether that is in the conception, study design, execution, acquisition of data, analysis and interpretation, or in all these areas; took part in drafting, revising or critically reviewing the article; gave final approval of the version to be published; have agreed on the journal to which the article has been submitted; and agree to be accountable for all aspects of the work.

Corresponding authors

Correspondence to Shanling Xu or Lang Bai.

Ethics declarations

Ethics approval and consent to participate

Kunping Cui was certified to extract data from the MIMIC-IV and eICU-CRD databases for research after passing the " Collaborative Institutional Training Initiative (CITI)" training course on the National Institutes of Health (NIH) website (certification number: 59589454). This study used public deidentification databases, so there is no need to obtain the informed consent and approval of the Institutional Review Board.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.

Additional information

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary Information

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Cui, K., Teng, X., liu, W. et al. L-shaped association of body mass index with prognosis in individuals with sepsis: a multicenter cohort study. Diabetol Metab Syndr 17, 43 (2025). https://doi.org/10.1186/s13098-025-01607-w

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://doi.org/10.1186/s13098-025-01607-w

Keywords