Study population and setting
This study linked national data sources to compare disease severity by SARS-CoV-2 variant type among 43,338 individuals in England with whole-genome sequencing confirmed Alpha or Delta variant infection between March 29, 2021 and May 23, 2021. Approximately 60% of all positive PCR tests in the national database were successfully sequenced by the second month of the study period. Confirmed Alpha or Delta variant cases were excluded from the study population if they had a prior SARS-CoV-2 infection or if they lacked an associated national health service (NHS) identification number needed to link cases to data on outcomes and covariates. The primary severity outcome was COVID-19-related hospitalization, and a secondary outcome was COVID-19-related hospitalization or emergency department visits, both ascertained through national routine health information systems. Hospitalizations 1-14 days after the positive SARS-CoV-2 specimen collection were defined as COVID-19 related. Hospitalizations which began on the same date as the positive test were only included if the patient’s symptoms began 1-7 days prior to hospitalization, or if the patient died in the hospital and had COVID-19-relevant ICD-10 codes. Emergency department visits were identified as COVID-19 related using the same timeframes relative to specimen collection and additionally excluded visits with ICD-10 codes related to injuries. The following variables that could confound the association between variant type and risk of hospitalization were identified by linking each sequenced case to additional data sources with their NHS identification number: age, sex, area or residency, level of socioeconomic status as measured by the index of multiple deprivation (IMD), ethnicity, international travel within 14 days of specimen collection, and date of specimen collection. The relationship between the variant type and hospital admission/emergency care visit was assessed using Cox proportional hazard regression models, adjusted and stratified for these confounders. An interaction term with vaccination status was included in the model to determine whether an association between the Delta variant and hospitalization differed between vaccinated and unvaccinated groups.
Summary of Main Findings
A total of 43,338 cases were included in the study; 34,656 (80%) had the Alpha variant and 8,682 (20%) had the Delta variant. Over the eight-week study period, the prevalence of the Alpha variant dropped from 99% to 35% of weekly cases. Patients with the Delta variant were younger (median age 29 vs 31 years old) and were more likely to be Asian and live in London or the northwest England. Individuals with the Delta variant had a higher risk of hospitalization than those with the Alpha variant, with an adjusted hazard ratio (aHR) of 2.26; 95%CI 1.32-3.89. The risk of the combined endpoint of hospitalization or an emergency department visit was also higher among those with the Delta compared to the Alpha variant (aHR of 1.45; 95%CI 1.08-1.95). These inferences were robust to a series of sensitivity analyses that varied the definitions of outcomes and covariates and used alternative analytic approaches. Among unvaccinated patients and those less than 21 days from their first vaccinated dose, Delta variant was associated with a higher risk for hospitalization with an adjusted HR of 2.32; 95%CI 1.29-4.16. Among individuals with at least partial vaccination, the risk of hospital admission was higher among patients with Delta variant compared to those with Alpha variant (aHR 1.94, 95%CI 0.47-8.05); however, the estimate was not statistically significant with low precision due to the small number of vaccinated individuals who were hospitalized.
Whereas other studies of Delta variant severity inferred the variant type based on S-gene target detection, a strength of this study was the measurement of variant type by whole-genome sequencing during a period when the national coverage was relatively high in England (60% of all PCR-confirmed cases were successfully sequenced by the second month of the study). The analysis accounted for most strong confounders through stratification and regression adjustment. Several sensitivity analyses were done to reduce the possibility of a spurious association between the Delta variant and hospitalization risk.
Since the study covered only the first few weeks that the Delta variant accounted for a significant proportion of COVID-19 cases, it is possible that infections with faster durations or more severe courses are over-represented among the Delta cases and under-represented among the Alpha cases, overestimating the hazard ratios. Studies over longer periods will provide more confidence that these early trends continued as Delta became the dominant variant in England and other countries. Second, the use of whole-genome sequencing, while the gold standard for variant measurement, may not be possible on PCR test results with lower viral loads, thus the sample likely underrepresents individuals with milder cases. An additional 11% of cases with genomic sequencing were excluded from the analysis due to lacking NHS number to link to the other data sources, and this was more likely to be missing for Delta cases, and for Black or Asian individuals or those who travelled internationally. Finally, the Cox regression analysis was described as being stratified on several variables simultaneously, but the paper was unclear as to how the authors arrived on the single hazard ratio reported for hospitalization risk by variant. While sensitivity analyses suggested that the Delta hospitalization risk associated with the Delta variant was robust to more standard multivariable regression methods and varying definitions, the study was not able to account for confounding by comorbidity status, a strong predictor of hospitalization.
This study was among the first and largest to establish a link between the Delta variant and disease severity with variants confirmed by whole-genome sequencing.
This review was posted on: 15 December 2021