SARS-CoV-2 viral load is associated with increased disease severity and mortality

Study design and participants: Hospitalized and non-hospitalized participants with COVID-19 were enrolled in a longitudinal sample collection study at Brigham and Women’s Hospital and Massachusetts General Hospital. Cohorts included 88 hospitalized patients with COVID-19, symptomatic outpatients evaluated at a respiratory infection clinic (including those with positive and negative SARS-CoV-2 nasopharyngeal tests), and 53 recovered individuals previously diagnosed with COVID-19. Clinical data (oxygenation status, demographics, comorbidities, hospitalization outcomes) were abstracted from electronic medical records. IRB approval was obtained and informed consent was collected.

Sample collection and processing: From hospitalized participants, plasma (blood), nasopharyngeal (NP) and oropharyngeal (OP) swabs, sputum, and urine were collected by nurses. NP/OP swabs were placed in 3 mL phosphate-buffered saline. Samples were processed within 48 hours at the Ragon Institute and stored at −80 °C. Inflammatory markers including C-reactive protein (CRP) and interleukin-6 (IL-6) were quantified from plasma.

Viral load quantification: SARS-CoV-2 RNA levels were quantified using US CDC reagents and protocols for RT-qPCR. RNA was extracted per manufacturer instructions, and qPCR was performed to determine viral RNA copies/mL. Viral loads were analyzed as continuous measures and also categorically (detectable vs. undetectable) across multiple specimen types (NP, OP, sputum, plasma, urine).

Statistical analysis: Group comparisons used non-parametric tests (Kruskal–Wallis, Wilcoxon rank-sum, Wilcoxon signed-rank). Associations were assessed with Spearman correlation. Categorical comparisons used chi-square and Fisher’s exact tests. Logistic regression models evaluated associations between continuous viral loads (by compartment) and risk of death. Odds ratios (ORs) and P values were reported. Longitudinal analyses compared viral loads between first and second available time points in each compartment.

• Participant characteristics: Hospitalized participants (N=88) were older than symptomatic outpatients and recovered individuals (Kruskal–Wallis P < 0.001). Outpatients had shorter time from symptom onset to sampling (median 5 days) than hospitalized (median 13 days) and recovered (median 27 days).
• Detection rates at initial sampling in hospitalized patients: NP 50% detectable, OP 67% detectable, sputum 85% detectable, plasma viremia 27% detectable, urine 10% detectable.
• Plasma viremia among outpatients and recovered: 13% of outpatients diagnosed with COVID-19 had detectable plasma SARS-CoV-2 RNA; 0% in NP-negative symptomatic outpatients and 0% in recovered individuals.
• Viral load levels: In those with detectable plasma viremia, median plasma viral load was 2.4 log10 RNA copies/mL (range 1.8–3.8), significantly lower than sputum viral load (median 4.4 log10 RNA copies/mL, range 1.8–9.0; Wilcoxon P < 0.001).
• Correlations among compartments: NP vs OP Spearman rho = 0.34 (P = 0.03); NP vs sputum r = 0.39 (P = 0.003); OP vs sputum r = 0.56 (P = 0.001). Plasma viral load was associated with NP (r = 0.32, P = 0.02) and OP (r = 0.36, P = 0.049) viral loads. No significant associations were observed with urine viral load.
• Association with disease severity: Detectable plasma viremia prevalence increased with respiratory disease severity among hospitalized patients: 44% in the most severe group vs 19% in those on nasal cannula vs 0% in those not requiring supplemental oxygen (chi-square P = 0.006).
• Inflammation and immune markers: Higher plasma viral loads correlated with lower absolute lymphocyte counts (r = −0.31, P = 0.008), higher CRP (r = 0.40, P = 0.003), and higher IFN-γ (r = 0.50, P < 0.0001). Presence of detectable plasma, NP, and sputum viral loads was associated with lower lymphocyte counts and higher CRP and IL-6. NP and/or OP viral loads were also associated with elevated IL-6, IP-10, MCP-1, IFN-γ, and IL-1RA.
• Mortality: Eleven hospitalized participants died (primary cause: respiratory failure). Those who died had higher initial plasma viremia (median 2.0 vs 1.0 log10 RNA copies/mL; P = 0.009), with sampling a median of 11 days before death. Mortality was higher with initial detectable plasma viremia: 32% vs 8% without viremia (OR 5.5, P = 0.02). Among patients on ventilatory support at sampling, mortality was 43% with viremia vs 17% without (OR 3.8, P = 0.11). Older age was strongly associated with mortality (mean 76 vs 55 years; P < 0.001); in those ≥70 years, plasma viremia was strongly associated with death (OR 21, P = 0.01). Logistic regression treating viral load as continuous showed plasma, OP, and sputum viral loads associated with increased risk of death.
• Longitudinal trends: Plasma and respiratory viral loads declined between first and second time points in most participants across compartments.

This study demonstrates that SARS-CoV-2 viral load, particularly plasma viremia, is relatively common among hospitalized patients with COVID-19 and correlates with markers of systemic inflammation, lymphopenia, and clinical severity. Importantly, higher plasma viral loads and the presence of detectable viremia at initial sampling were associated with increased risk of in-hospital mortality, supporting a role for viral burden as a prognostic indicator. Correlations across respiratory compartments and with plasma suggest a systemic component of infection. The associations between viral load and inflammatory mediators (CRP, IL-6, IFN-γ, IP-10, MCP-1, IL-1RA) are consistent with virus-driven hyperinflammation in severe disease. Potential mechanisms include spillover of virus from pulmonary tissue into the vasculature and/or direct endothelial infection, aligning with reports of endotheliitis and extrapulmonary involvement. These findings indicate that quantifying viral RNA levels across compartments can aid risk stratification and may inform therapeutic strategies targeting viral replication and inflammation.

SARS-CoV-2 viral load measurement across respiratory samples and plasma reveals strong associations with disease severity, systemic inflammation, and mortality in COVID-19. Plasma viremia, in particular, identifies patients at higher risk of adverse outcomes and may serve as a useful biomarker for risk stratification. Future research should clarify the mechanisms underlying plasma viremia and its relationship to endothelial and systemic involvement, determine whether plasma RNA represents infectious virus, and evaluate whether antiviral and/or immunomodulatory treatments guided by viral load levels improve outcomes. Larger, prospective studies with standardized sampling times are warranted to validate these findings and refine clinical use of viral load metrics.

• Mechanistic uncertainty: The source and significance of plasma viremia are not fully defined; it is unclear whether plasma RNA represents infectious virus.
• Timing variability: Sampling occurred at different times relative to symptom onset across cohorts, potentially affecting viral load comparisons.
• Power limitations: Subgroup analyses (e.g., patients on ventilatory support) did not always reach statistical significance, suggesting limited power.
• Multifactorial inflammation: Some participants exhibited elevated inflammatory markers without detectable viremia, indicating multiple drivers of inflammation beyond viral load.
• Generalizability: Single health system setting and modest number of deaths (n=11) may limit generalizability of mortality analyses.