The coronavirus disease 2019 (COVID-19) pandemic led to multiple waves of infection with the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), especially with the emergence of new, more transmissible, and sometimes immune-evading variants of concern (VOCs) of the virus. Among the deadliest of these variants is the Delta VOC, which fueled an intense surge of cases, hospitalizations, and deaths throughout the world.
In the United Kingdom, the SARS-CoV-2 Delta variant was first found in imported cases in April 2021, while cases in India were quickly rising into the high thousands. Soon after, in May 2021, the Delta variant was declared a VOC and studies were rapidly initiated to determine how different this new strain of SARS-CoV-2 was from the previously dominant Alpha variant. A new preprint published on the medRxiv* server describes the early results of one such study.
Study: Estimating the Increase in Reproduction Number Associated with The Delta Variant Using Local Area Dynamics in England. Image Credit: Adam Vilimek / Shutterstock.com
The researchers recycled an approach previously used with the Alpha variant to quickly arrive at estimates of whether the Delta variant would lead to an increase in the effective reproduction number of the virus. The aim of the study was to compare the time-varying effective reproduction number (Rt) by each upper-tier local authority (UTLA), with the percentage of positive cases confirmed by the reverse transcriptase-polymerase chain reaction (RT-PCR) that were positive for the spike (S) gene.
The scientists used more than one model and explored several different scenarios in order to accommodate the many unknown factors. While reports have already been examined by experts such as the Scientific Pandemic Influenza Group on Modeling (SPI-M) and by the Scientific Advisory Group for Emergencies (SAGE) in June 2021, this latest version presents more or less the same approach and results to ensure that the evidence that was available at the time of the study is the basis of the presented conclusions arrived.
Data on cases came from PCR tests carried out at local authority levels and were provided by Public Health England (PHE), Google mobility data classified by context, and the changing restrictions on mobility and social interactions over time. The mobility data were analyzed for mean changes per week.
S gene positivity was calculated in terms of weekly percentages using sequencing data. The data was applied to the preceding week to account for the delay from infection to sequencing, so as to find out how much of each week’s total cases were potentially due to the Delta variant rather than the Alpha.
The scientists found that between February 23, 2021, to May 25, 2021, the percentage of S gene positivity was associated with increased Rt estimates at UTLA level and strengthened over time. The S gene positivity varied across regions.
The confidence limits ranged from a decrease of 10% to an increase of over 110% during April and May, varying with the implemented generation time and the model in use. The best fit appeared to be with the model that compensated for residual variation over time, both at national and regional levels.
Interestingly, all models showed that Rt was increasing, with higher proportions of S gene positivity; however, the bounds varied with the generation times used in the model. The lower bounds varied from 20% to 28%, depending on whether they adjusted for both national-level and regional-level variation over time, only national-level, or neither, with a short generation time.
Conversely, the lower bounds ranged from 27% to 38% for these models if the generation time was longer. As the generation time was changed, the Rt values also increased from 33% to 55%.
If residual variation was not adjusted for, the model reproduced estimated reproduction numbers quite well over time. In March, mass testing was carried out in schools, the results of which were subsequently entered the case data and caused unexpected variations to arise in the data. The same outliers were seen when the Delta variant was causing a rising number of cases later in the study period.
The model also showed that different stages of the reopening process were associated with different impacts on the Rt. The effect at each step was small in size, about 5% to 20%. Altogether, the Rt came close to one by the time the third phase of relaxations was begun, even when the Alpha variant was assumed to be circulating.
Using the S gene positivity percentage as a stand-in for infection with the Delta variant, the researchers found that the Rt estimates at UTLA showed a direct relationship with it in all four tested models that compensated for variation at the national or regional level, both, or none.
The greatest rise in the Rt was seen in the model that linked variation over time with public health interventions, mobility, and SGTF. In fact, almost 50% increases occurred if a long generation time was assumed, whereas an increase by a third was recorded if a short generation time was assumed.
This decreased to a quarter and just over a third, with short and long generation times, respectively, if the only regional variation was adjusted for. This fitted the data best but could probably reflect the differences within each region at UTLA being affected by factors that are not implemented in the model.
In the non-adjusted model, secondary attack rates increased from about 8% with the Alpha variant to over 12% with the Delta, thus indicating the Rt to be raised by around 50%. Other models also produced estimates close to the data, which could be a signal that the variant itself caused the Rt to cross one.
The researchers chose to potentially enhance the contribution of stochastic effects by giving equal weight to all local areas when estimating the growth of the outbreak, whereas most earlier studies gave greater weight to local areas that saw larger numbers of cases. This could bias the study if the local outbreak was driven by specific conditions.
The S gene positivity was ultimately found to be consistently associated with higher Rt values in all models under a variety of assumptions. The conclusion is that the interventions applied from April to June 2021 could not effectively reduce the Rt of the Delta variant to less than one.
Despite the various limitations of the study, such as assuming equal generation times for the Delta and Alpha variants and a uniform effect for the lockdown all over the country, this analysis represents a real-time estimate of the Rt. Future research may show greater consideration of the uncertainty inherent in the modeling approach to transmission and sequencing data while considering the spatial variation.
This could lead to producing an approach that can be used to all future scenarios requiring a rapid study to put together evidence that will be of help to policymakers.
medRxiv publishes preliminary scientific reports that are not peer-reviewed and, therefore, should not be regarded as conclusive, guide clinical practice/health-related behavior, or treated as established information.