Clinical Study Protocol of the 'Biomarkers of Severity of COVID-19 Patients' (BIOMARCOVID) Project

Technical Summary of the BIOMARCOVID Clinical Study Protocol

Problem Statement
The COVID-19 pandemic exposed critical limitations in existing tools for predicting disease severity, particularly regarding the sensitivity and specificity of standard clinical features (e.g., age, BMI) and routine blood biomarkers (e.g., CRP, IL-6, platelet count). While these factors provide some prognostic value, a significant "grey area" remains where clinically fit or young patients without known comorbidities may still develop severe forms of the disease. Current untargeted metabolomics research has shown promise in assessing severity but lacks sufficient data to validate novel metabolite biomarkers that could significantly improve predictive accuracy when combined with standard medical biology data.

Methodology
The BIOMARCOVID project is a monocentric, retrospective, observational cohort study conducted at the Centre Hospitalier Universitaire Grenoble Alpes (CHUGA). The study focuses on patients admitted between March and December 2020 with PCR-confirmed SARS-CoV-2 infection.

• Cohort Selection: Eligible participants were adults (>18 years) admitted for suspected COVID-19 who tested positive. Exclusion criteria included age <18, transfer from another hospital, expressed opposition to research, direct admission to the ICU on day 1, or blood sampling occurring more than 48 hours after admission.
• Data Collection:
  • Clinicome: Demographic data, medical history (comorbidities), and disease progression metrics (e.g., ICU admission, oxygen initiation, mortality) were recorded. Patients were classified retrospectively using the WHO 10-category ordinal scale, defining "mild" as scores 4–5 and "severe" as scores >5.
  • Biologicome: 62 routine blood tests were analyzed, performed according to standard clinical procedures upon admission.
  • Metabolome: Residual EDTA plasma samples, originally collected for IL-6 assays, were repurposed for untargeted metabolomics. Samples were processed within 4 hours of collection, aliquoted, and stored at −80°C. Analysis was conducted using multiplatform LC-MS/MS across four national platforms (CEA Saclay, INRAE Clermont, MetaToul, GEMELI) to ensure broad metabolome coverage.
• Data Processing and Statistical Analysis:
  • Metabolomics data underwent rigorous processing, including conversion to mzXML, peak detection via XCMS, and filtering based on QC correlation, repeatability (CV <30%), and biological-to-blank ratios.
  • Primary Analysis: A logistic regression model with LASSO variable selection was constructed to predict severity outcomes using clinical and routine biochemical parameters. Model performance was evaluated via accuracy, sensitivity, specificity, and predictive values.
  • Secondary Analysis: Individual and combined testing of routine blood tests and metabolites was performed to identify predictive variables. Logistic regressions adjusted for sex, age, and obesity were utilized, with p-values corrected using the Benjamini-Hochberg method.
  • Multi-block Integration: The study employs a multi-block chemometric approach to integrate three heterogeneous data blocks (Metabolome, Biologicome, and Clinicome), aiming to provide a systems-level view of COVID-19 severity and assess the relative contribution of each data source.

Key Contributions

• High-Quality Pre-analytical Data: The study leverages blood samples collected within 48 hours of admission and prior to severe symptom onset, with aliquoting within 4 hours, ensuring high pre-analytical quality.
• Multiplatform Metabolomics: The use of four complementary analytical platforms provides extensive metabolome coverage, reducing platform-specific biases.
• Integrated Systems Approach: By combining untargeted metabolomics with 62 routine blood tests and clinical data, the study moves beyond single-modality analysis to explore complex interactions between metabolic pathways and standard clinical markers.
• Biobank Creation: The project establishes a biobank of clinical, biological, and metabolomic data to support future sub-studies and validation efforts.

Results
As this document is a study protocol, it does not present final experimental results or specific biomarker discoveries. The text outlines the planned analytical framework and notes that the study is a pilot with no formal sample size calculation, meaning findings are intended to be exploratory and hypothesis-generating. The protocol details the statistical methods (e.g., LASSO regression, multi-block integration) that will be applied to the collected data to identify differential metabolites and predictive models.

Significance and Claims
The paper positions the BIOMARCOVID project as a methodological advancement for predicting COVID-19 severity. Its primary significance lies in the potential to enhance the sensitivity and specificity of outcome prediction by integrating metabolite biomarkers with standard medical biology data. The authors claim that this approach could:

1. Improve patient stratification for hospitalized individuals.
2. Aid in the anticipation of ICU bed requirements.
3. Identify early therapeutic targets by revealing specific metabolic pathways dysregulated in early disease stages.
4. Provide a transferable methodology adaptable to future emerging infectious diseases.

The authors explicitly acknowledge limitations, including the monocentric and retrospective nature of the study, the lack of a formal sample size calculation, and potential generalizability issues due to the single epidemic wave and setting. They emphasize that the study is designed to generate hypotheses for future validation rather than to provide definitive clinical guidelines at this stage.