In this retrospective cohort study, 19,311 critically ill patients were included from the MIMIC-III database and then classified into four BMI-based categories: underweight, healthy weight, overweight, and obese. Notably, the proportion of underweight patients (2.3%) was the lowest, while overweight (34.8%) and obese (34.5%) patients constituted a much larger proportion of critically ill patients. The prevalence of obesity in our study was much greater than that in recent reports, in which the proportion was approximately 20%14. A positive association was observed between glucose and BMI as well as between the frequency of hypertension and BMI, which is consistent with previous studies20–22. Briefly, we found that underweight was associated with a significantly greater risk for 28-day mortality and 60-day mortality, while overweight was associated with a lower risk for 28-day mortality and 60-day mortality than healthy weight in critically ill patients. Additionally, we observed a typical L-shaped association between BMI and mortality, with the lowest mortality occurring at approximately 32 kg/m². Moreover, patients with a higher BMI had a greater likelihood of receiving mechanical ventilation therapy and RRT. A large amount of research has shown that BMI has J-shaped associations with overall mortality and most specific causes of death23. However, a small amount of research has estimated a lower risk among overweight patients than among healthy weight patients24. The discrepant results between different studies might be explained by reverse causality and residual confounding. Most studies have shown clear heterogeneity in the associations between BMI and different causes of death, which means that the association of all-cause mortality with BMI will be affected by cause-of-death distributions in the source population. Currently, the impact of obesity on ICU mortality is still debated. Although obesity seems to be associated with morbidity, the relationship between the two is not yet clear 13,14. Thus, this study provides evidence for the possibility of using BMI to predict the prognosis of critically ill patients, highlighting the possibility of employing BMI as a novel, easily obtained risk marker.
The prevalence of obesity has tripled worldwide and has been increasing since 1975. It has become a nonnegligible health issue that affects all age groups, populations, and countries of all income levels. Identifying the complex relationship between BMI and mortality is challenging. This challenge becomes increasingly difficult in complex, dynamic environments such as the ICU. In general, innovations that have been found to be beneficial in specific disease populations may be applied to other populations in which those findings have not been adequately analyzed, potentially exposing patients to harm without commensurate benefit 25. Large cohort studies in the general population demonstrated that BMI had J-shaped associations with overall mortality and most specific causes of death, except for mental and behavioral, neurological, and external causes, which means that a high BMI was associated with an elevated risk of mortality for most patients 23,26. More recent data in hospitalized patients or patients with chronic illnesses have shown that overweight and moderate obesity are associated with lower mortality than a normal BMI or more severe obesity 14,18. Such a scenario has been observed not only in multiple chronic diseases, such as heart failure 27, end-stage kidney disease 28 and coronary artery disease 29 but also in acute conditions, such as pneumonia 30, acute respiratory distress syndrome 31, sepsis 32 or critical illness in general 33. In addition, increasing BMI was not associated with increased mortality in patients with COVID-19 admitted to intensive care 34. The phenomenon known as the “obesity paradox” is that obesity increases the risk of obesity-related disease but is paradoxically associated with increased survival in patients with these conditions. Our results are derived from an analysis of a real-world database and provide additional evidence for understanding the relationship between BMI and mortality in critically ill patients.
Although a large number of studies have revealed the obesity survival paradox, the actual protective effect of adipose tissue has been challenged. The reason might be that the protective effect of adipose tissue could be influenced by factors that are difficult to control. First, some young and healthy obese patients may be admitted to the ICU at a low threshold to avoid complications, thereby leading to selection bias. Second, the obesity survival paradox may be affected by therapeutic factors. Obese patients prefer to receive a lower weight-based dosage of fluids and vasopressors, which potentially reduces the side effects of these therapies 35. Finally, there are increasing criticisms of the use of BMI as a measure of obesity.
In this study, we systematically analyzed the relationship between outcomes and BMI based on data from a large sample size extracted from a real-world database. We employed a novel methodology to address continuous covariates and reverse causality and conducted a wide range of sensitivity analyses.
However, our research has several limitations. First, approximately 50% of individuals who were otherwise eligible were excluded, mostly because they had no BMI records available. This would induce selection bias, which is difficult to avoid. Second, the lack of relevant information about therapy may limit the generalizability of these results. However, sensitivity analyses for subgroups with diabetes or liver disease who might receive a nutritional approach reached similar conclusions. Third, the MIMIC-III population is predominantly white, so these findings may not be generalizable to other populations.