Obesity Paradox Emerges in Urban COVID-19 Cohort as Diabetes and Hyperglycemia Drive Mortality

When the first wave of COVID-19 swept through Philadelphia in the spring of 2020, clinicians at Temple University Hospital, an academic medical center serving one of the city’s most underserved urban communities, faced a patient population already burdened by metabolic disease. A retrospective study of 478 adults hospitalized with confirmed COVID-19 between March and May 2020 has now revealed a striking and counterintuitive pattern: while diabetes and hyperglycemia were firmly linked to greater disease severity, intensive care utilization, and death, patients with higher body mass index paradoxically survived more often than those of normal weight. The findings, published in Frontiers in Clinical Diabetes and Healthcare, add a complex urban dimension to one of the most debated phenomena in metabolic medicine, the obesity paradox.

The research team, led by endocrinologists Sanah Sadiq and Rashelle Ripa together with biostatistician Huaqing Zhao and endocrinologist Kristin Criner, set out to disentangle the combined influence of diabetes, glycemic control, and body composition on COVID-19 outcomes. Earlier studies had typically examined obesity and diabetes separately, and the authors hypothesized that diabetes would increase morbidity and mortality in hospitalized patients, with obesity amplifying that risk. What they found instead challenged the second half of that hypothesis. Among the 478 patients included after careful chart review, 45 percent were classified as obese with a body mass index of at least 30 kilograms per square meter, and another 29 percent were overweight, reflecting the heavy metabolic burden of the North Philadelphia population the hospital serves.

The demographic profile of the cohort underscores why the study is distinctive. The mean age was 55.7 years, 60 percent of patients were men, and Black patients made up 38 percent of the group, with Hispanic patients accounting for 23 percent. By the end of follow-up in August 2022, 65 patients, or 13.6 percent of the cohort, had died. The differences between survivors and nonsurvivors were sharp. Deceased patients were significantly older, with a mean age of 67.9 years compared with 53.8 years among survivors. They spent far longer in the hospital, 14.0 days versus 6.5 days, and required dramatically more intensive care, with a mean ICU stay of 106.2 hours compared with 20.4 hours. Mechanical ventilation was needed by 26.2 percent of deceased patients but only 1.5 percent of survivors.

Racial disparities in mortality were also evident and statistically significant. Black patients comprised 38.3 percent of the overall cohort but accounted for 46 percent of the deaths, while White patients, 12.8 percent of the cohort, represented 24.6 percent of the deceased. Hispanic patients were proportionally underrepresented among deaths, at 16.9 percent versus 26.6 percent of the cohort. The authors note that these patterns reflect the structural inequities of the communities served by urban safety-net hospitals, where comorbidities such as hypertension, diabetes, and kidney disease are highly prevalent and access to care is often limited.

Diabetes emerged as a central driver of severity. Of the 478 patients, 173, or 36 percent, had preexisting diabetes, the overwhelming majority of it type 2. Patients with diabetes were significantly older, with a mean age of 61.8 years versus 52.3 years for those without the condition. More tellingly, 8.7 percent of patients with diabetes required mechanical ventilation compared with 2.6 percent of those without, and 73.4 percent of patients with diabetes developed hyperglycemia during hospitalization, defined as blood glucose of at least 200 milligrams per deciliter on two or more occasions, compared with just 11.5 percent of patients without diabetes. Among patients with available hemoglobin A1c measurements, those with diabetes who developed cytokine storm, the unregulated release of inflammatory cytokines that can precipitate acute respiratory distress syndrome and shock, had significantly higher mean HbA1c than those who did not, with a median of 7.8 percent among patients with diabetes versus 6.3 percent among those without.

Hyperglycemia itself, whether related to preexisting diabetes or to corticosteroid therapy, was a powerful marker of critical illness. Deceased patients had a significantly higher incidence of hyperglycemia, 50.8 percent versus 31.2 percent among survivors, and a much higher incidence of steroid-induced hyperglycemia, 15.4 percent versus 6.1 percent. In the intensive care unit, the pattern was even more pronounced: 61.3 percent of ICU patients developed hyperglycemia and 21.5 percent developed steroid-induced hyperglycemia, compared with 26.5 percent and 3.9 percent respectively among patients who never required ICU care. Multivariable logistic regression identified steroid-induced hyperglycemia, with an odds ratio of 3.532, and obstructive sleep apnea, with an odds ratio of 4.346, as independent predictors of ICU admission, alongside higher cumulative methylprednisolone use.

The corticosteroid findings require careful interpretation, and the authors are explicit about this. Every patient at the institution received glucocorticoids as part of a hospital-wide COVID-19 treatment protocol, with dosing scaled to disease severity. Deceased patients received a mean initial daily dose of 100.6 milligrams of methylprednisolone compared with 42 milligrams among survivors, and a total mean steroid dose over admission of 1,010 milligrams versus 404.8 milligrams. ICU patients received a mean of 1,401.5 milligrams compared with 264.3 milligrams among non-ICU patients. The authors caution that these associations reflect confounding by indication, since sicker patients received more intensive steroid therapy, rather than a causal effect of corticosteroid exposure. The same logic applies to the higher rates of hyperglycemia among critically ill patients, who received both higher steroid doses and were more severely ill to begin with.

Against this backdrop of metabolic dysfunction driving mortality, the body mass index results were startling. Mean BMI was significantly lower among deceased patients, 28.1 kilograms per square meter, than among survivors, 31.0 kilograms per square meter. Kaplan-Meier survival analysis demonstrated significant differences in survival across BMI categories, with lower survival probability in the normal weight group compared with the overweight and obese groups. In a multivariable Cox proportional hazards model adjusting for age, sex, race, smoking status, and diabetes diagnosis, BMI remained significantly associated with mortality, with a Wald chi-square of 10.24 and a p-value of 0.006. Age and smoking status were also significant in the adjusted model, while race, sex, and diabetes status were not. In a separate hazard ratio model, higher body weight in kilograms was associated with a lower hazard of death, with a hazard ratio of 0.981, while heart failure and former smoking were associated with higher mortality risk.

The obesity paradox, the observation that overweight or obesity by BMI classification is linked to higher survival in certain acute and chronic illnesses, has been documented in conditions ranging from myocardial infarction and stroke to heart failure, sepsis, and community-acquired pneumonia. Proposed mechanisms include greater metabolic reserves in the form of large energy stores, potentially protective adipokines such as adiponectin, and greater muscle mass supporting cardiorespiratory fitness. Prior studies in urban populations have found similar patterns, including one showing that overweight and obese patients admitted to an urban medical center with community-acquired infections had significantly lower 90-day and long-term mortality, and another demonstrating one-year survival benefits among obese older Medicare beneficiaries with severe sepsis. A large population-based cohort from England also found a U-shaped association between BMI and COVID-19 mortality among people with diabetes, with the lowest risk at a BMI of 25.0 to 29.9 kilograms per square meter.

Yet the authors resist a simple protective interpretation. The paradox is increasingly understood as partly an artifact of statistical biases. Survival or selection bias may allow healthier individuals with overweight or obesity to be overrepresented among hospitalized cohorts, and patients with higher BMI may receive more aggressive clinical management, particularly respiratory support. At the institution, patients with overweight or obesity were more likely to develop respiratory distress, more commonly placed on high-flow oxygen, bilevel positive airway pressure, or the prone position, and had a lower threshold for ICU or step-down care. BMI itself is an imperfect measure that does not distinguish fat mass from lean mass, and severe illness can cause unintentional weight loss that lowers BMI and further complicates the picture. The study’s retrospective design, single-center setting, modest sample size for some analyses, and the fact that most patients lacked admission HbA1c measurements all limit generalizability. Still, the authors believe this is the first study examining the obesity paradox in an urban population with both diabetes and COVID-19, and they emphasize that close monitoring of patients with diabetes or hyperglycemia during COVID-19 hospitalization remains essential, while the long-term risks of obesity are never in question.

Subject of Research: Body mass index, diabetes, inflammation and survival in COVID-19: a retrospective study evaluating the obesity paradox in an urban hospital center

Article Title: Body mass index, diabetes, inflammation and survival in COVID-19: a retrospective study evaluating the obesity paradox in an urban hospital center

Article References: Sadiq, S., Ripa, R., Zhao, H., & Criner, K. (2026). Body mass index, diabetes, inflammation and survival in COVID-19: a retrospective study evaluating the obesity paradox in an urban hospital center. Frontiers in Clinical Diabetes and Healthcare, 7, Article 1946918. https://doi.org/10.3389/fcdhc.2026.1946918

Image Credits: AI Generated

DOI: 10.3389/fcdhc.2026.1946918

Keywords: Body, mass, index, diabetes, inflammation, survival, COVID-19, retrospective, evaluating, obesity, paradox, urban