TL;DR
Wastewater surveillance is a powerful tool for tracking viral evolution, but concerns about differential shedding between variants could bias estimates of their selection advantage. This study demonstrates, through mathematical modeling and real-world data, that wastewater-based estimates of variant selection advantage are remarkably robust to these shedding differences, maintaining their utility for public health.
What Problem Does This Paper Address?
The COVID-19 pandemic spurred the widespread adoption of wastewater-based epidemiology (WBE) as a cost-effective and population-level method for tracking viral dynamics and the emergence of new variants. A key epidemiological parameter of interest is the "selection advantage" of a variant, which reflects its increased transmissibility or immune evasion compared to existing strains. However, a potential pitfall for WBE is the variability in how much viral RNA individuals shed over time (shedding profiles). If different SARS-CoV-2 variants have distinct shedding profiles, this could distort the relationship between the viral load detected in wastewater and the actual number of infections in a community. This discrepancy could lead to biased estimates of variant selection advantage, undermining the reliability of WBE for informing public health decisions. This paper aims to rigorously assess whether these differential shedding profiles indeed bias estimates of variant selection advantage and the effective reproduction number ($R_e$), and to what extent.
Methods
The researchers employed a multi-pronged approach:
- Mathematical Modeling: They derived closed-form expressions to analytically quantify the potential bias introduced by shedding profiles on estimates of selection advantage and $R_e$. This involved developing models that account for viral incidence, transmission rates, and shedding load profiles.
- Simulations: They conducted extensive simulations to test their analytical findings under various scenarios, including different shedding profiles (same shedding, scaled shedding, different distribution), varying transmission rates (constant, rate-limiting growth, intervention), and different generation times.
- Real-World Data Analysis: They analyzed wastewater surveillance data from Switzerland during the Omicron BA.1 wave, comparing it with clinical data to empirically validate their theoretical findings. This involved using both next-generation sequencing (NGS) and digital PCR (dPCR) for variant detection.
Datasets
- Swiss Wastewater Surveillance Data: Collected from six wastewater treatment plants (WWTPs) across Switzerland between November 2021 and January 2022. This data included viral loads measured by NGS and dPCR assays targeting specific SARS-CoV-2 mutations.
- Clinical Sequencing Data: Obtained from the LAPIS API of CoV-Spectrum, representing sequenced PCR-positive clinical samples from cantons surrounding the WWTPs. This data was used as a benchmark for comparison.
- Geneva University Hospital SGTF Data: Used as an additional clinical data source for the Geneva sewershed.
The study did not introduce new benchmark datasets but rather leveraged existing surveillance data to validate its theoretical models.
Key Insights and Ideas
- Robustness of Selection Advantage Estimates: The most significant finding is that estimates of variant selection advantage derived from wastewater data are remarkably unbiased and robust to differences in shedding profiles, even under a wide range of assumptions. This means that even if one variant sheds more or less virus than another, our ability to estimate its competitive edge remains largely unaffected.
- Transient Impact on $R_e$: While selection advantage estimates are robust, the effective reproduction number ($R_e$) estimates can experience a transient bias due to differential shedding. However, this bias diminishes over time as the variant's prevalence stabilizes.
- Wastewater's Enduring Advantage: Despite potential biases in direct incidence estimation, WBE's strengths in estimating selection advantage and $R_e$ are preserved, making it a valuable tool for public health.
- Conditions for Robustness Breakdown: The study identifies specific, albeit less common, conditions where robustness might be compromised, such as when shedding load distributions differ significantly in shape and transmission rates are varying rapidly.
Key Results
- Analytical derivations and simulations showed that estimates of selection advantage are unbiased and robust to arbitrary changes in shedding profiles when common model assumptions (e.g., constant transmission rates) hold.
- Even when these assumptions are relaxed, estimates of selection advantage remain robust to differences in total shedding, which is the typical way variants differ.
- Real-world data from Switzerland during the Omicron BA.1 wave confirmed that wastewater-based estimates of BA.1's selection advantage were highly concordant with those derived from clinical sequencing data, despite suspected differential shedding.
- Estimates of $R_e$ are shown to be transiently affected by shedding differences, with an upper bound on the bias derived.
Limitations and Future Work
- Compounding Effects: The authors acknowledge that they did not explore scenarios where both the shape of the shedding profile and the shape of the generation time distribution vary simultaneously, suggesting these effects could compound.
- Variance vs. Bias: While the study focuses on bias, it notes that differential shedding could potentially increase the variance of estimates, leading to less precise results. This aspect warrants further investigation.
- Early Estimation Precision: The paper suggests that changes in shedding might delay the early estimation of parameters due to a loss of precision, which could be relevant for rapid response scenarios.
- Generalizability: The authors propose that their findings are generalizable to other pathogens and variants, highlighting the broad applicability of WBE. Future work could focus on applying these robust estimation methods to other infectious diseases.
Summary updated