Plasma proteins are commonly used as biomarkers to profile inflammation and tissue injury in acute respiratory distress syndrome (ARDS). Whether baseline variation in these biomarkers relates to body size in individuals without acute illness is not well characterized. This analysis tested whether biomarkers previously associated with ARDS risk or mortality vary across the range of body mass index (BMI) in ambulatory adults.
The study analyzed 8,680 ambulatory adults from the All of Us Research Program. Participants had proteomic profiling covering 5,420 protein assays. From the literature, the authors prespecified a panel of 35 proteins that have been associated with ARDS risk or mortality. These 35 proteins were grouped into five pathophysiological compartments according to assignments made in a registered systematic review completed before analysis.
Body mass index was modeled as a restricted cubic spline to capture nonlinearity. For each of the 35 ARDS-related proteins, the authors reported the difference in protein level between BMI 35 and BMI 25 kg/m2, expressed in standard deviations. Models adjusted for age, sex, smoking status, thirteen chronic conditions, and ten genetic principal components. Each protein estimate was ranked against all 5,420 assays to contextualize effect sizes.
Across the full set of 5,420 assays, associations with BMI were predominantly positive, with a median effect of 0.065 standard deviations when comparing BMI 35 to 25 kg/m2. However, this overall trend masked heterogeneity among proteins and compartments; several lung epithelial proteins were lower at higher BMI while key inflammatory proteins were higher.
Four epithelial proteins previously implicated in ARDS were lower at higher BMI. The largest decrease was observed for the soluble receptor for advanced glycation end products (sRAGE), which was lower by 0.350 standard deviations between BMI 35 and 25 kg/m2. Three additional epithelial markers—surfactant protein A2 (SFTPA2), surfactant protein D (SFTPD), and club cell secretory protein 16 (CC16/SCGB1A1)—were lower by 0.081 to 0.084 standard deviations. Each of these epithelial protein differences ranked below at least 92% of all assays, indicating relative specificity of the negative association with BMI among the broader proteome.
Markers of systemic inflammation showed marked positive associations with higher BMI. Interleukin-6 (IL-6) increased by 0.534 standard deviations and E-selectin by 0.465 standard deviations comparing BMI 35 to 25 kg/m2. These effect sizes ranked above 99% of all 5,420 assays, indicating that IL-6 and E-selectin are among the most BMI-responsive proteins in the dataset.
When the 35 proteins were considered by their prespecified compartments, the compartment assignments explained 49% of the variance in the protein estimates. This contrasted with 10% variance explained under random reassignment of proteins to compartments, a difference that was statistically significant (P < 0.001). The analysis also noted that BMI-associated differences extended to lung-enriched proteins outside the prespecified 35-protein panel.
At a given BMI, common cardiometabolic comorbidities—type 2 diabetes, hypertension, and dyslipidemia—showed smaller differences in the ARDS biomarker panel than BMI itself. These comorbidities were not associated with the lower epithelial proteins observed at higher BMI, indicating that body size had a distinct association independent of these conditions in this cohort of ambulatory adults.
The study demonstrates that plasma proteins used as biomarkers of inflammation and injury in ARDS vary with body size even in the absence of acute illness. Consequently, differences in biomarker levels measured during ARDS may not reflect injury alone; some observed biomarker variation in patients with obesity may represent antecedent, BMI-associated differences. This has implications for interpretation of plasma biomarkers in clinical and research settings involving ARDS, suggesting that body size should be considered when using these markers to infer lung injury or inflammation.
The analysis used the All of Us Research Program Controlled Tier Dataset C2025Q4R6; individual-level data are available to authorized users on the Researcher Workbench and cannot be redistributed per the program's data use agreement. The authors report that analysis code and the estimates for all 5,420 assays are available from the corresponding author on reasonable request for authorized use within the Researcher Workbench. The Institutional Review Board of Mass General Brigham waived ethical approval for this secondary analysis of de-identified data; the All of Us Research Program IRB provided ethical approval for the parent protocol under which participants consented. The work was supported by a National Heart, Lung, and Blood Institute award (5K23HL175128) as declared in the source.