Skip to main content

Our take —

This study, available as a preprint and thus not yet peer-reviewed, described the identification of a new SARS-Cov-2 variant (B.1.220 with the E484K mutation) in Upstate New York in early 2021. The E484K mutation has been identified in other variants of concern, including 501Y.V2/B.1.351 identified first in South Africa, and P.1. identified first in Brazil. Some studies suggest that this mutation may lead to reduced vaccine efficacy as well as antibody-based treatment and prevention of COVID-19; recent clinical data indicate that the Pfizer, Moderna, and J&J vaccines are still protective against other variants with the E484K mutation. The independent emergence of E484K in different SARS-CoV-2 variant backgrounds is concerning, as it implies that this mutation provides a fitness advantage to the virus, even in the context of increasing vaccine coverage. This study illustrates the key role of genomic surveillance in the early identification of new SARS-CoV-2 variants, which is important for limiting their spread.

Study design

Retrospective Cohort

Study population and setting

This study described the identification of a SARS-CoV-2 B.1.220 variant carrying the E484K mutation in the Finger Lakes region of New York. The E484K mutation in the spike protein, which is present in other variants of concern (501Y.V2/B.1.351, P.1, and B.1.526), may lead to immune escape due to reduced antibody neutralization activity. Since June 2020, SARS-CoV-2-positive samples from the Rochester Regional Health System have been sent to the Multidrug-Resistant Organism Repository and Surveillance Network (MRSN) for whole genome sequencing as part of a joint surveillance effort. This study included sequences collected at five hospitals in Upstate New York between January 27, and February 7, 2021 (n=20); additional sequences of interest were identified using the Microreact and GISAID databases (n=12). B.1.220 variant sequences with the E484K mutation (n=17) were used for phylogenetic analysis and single nucleotide polymorphism-based comparisons.

Summary of Main Findings

The E484K mutation was identified in five sequences from four unique patients (one male, three females; mean age of 85 years; all recovered) with no evidence of epidemiologic linkage. All five sequences were assigned to the B.1.220 lineage and had additional lineage-defining mutations (R203K, G204R, P314L and D614G). The samples collected from the same patient were genetically identical. The Microreact and GISAID databases contained 12 additional B.1.220 variant sequences with the E484K mutation collected from six different New York counties between December 2020 and February 2021. Phylogenetic analysis suggested that the E484K mutation arose in at least two independent events within this sample set. At the time of submission, no B.1.1.220 variants with the E484K mutation had been identified outside of New York.

Study Strengths

Routine genomic surveillance with whole genome sequencing was used to identify a new SARS-CoV-2 variant with a mutation of known concern for immune escape.

Limitations

The sample size in this study was quite small (n=20); phylogenetic analysis was based on only 16 unique variant sequences, 12 of which were identified using database searches. Additional sequences are needed to better assess the emergence and transmission dynamics of this potential variant of concern. No experiments were presented assessing the functional impact of the E484K mutation (in the context of the B.1.220 genetic background) on the efficacy of vaccines or antibody-based treatments.

Value added

This study describes the emergence of a new SARS-CoV-2 B.1.220 variant with the E484K mutation, identified by whole genome sequencing as part of an ongoing genetic surveillance project. When present in other variants of concern, this mutation has been implicated in immune escape.

Our take —

The study sought to examine the rate of reinfection following initial infection using a retrospective cohort in Ohio and Florida from March 20, 2020 to February 24, 2021. The study found that, of 150,325 people tested during this period in the health system before August 30, 2020, 1,278 patients who initially tested positive later were tested again, 63 (4.9%) showed evidence of reinfection. Protection from reinfection due to prior infection was estimated at 81.8%. The study’s primary limitation was the use of infection occurrence within 6-month time blocks to infer re-infection, rather than having an initial positive test, a subsequent negative test, and a later positive test to determine that individuals cleared the infection. Therefore, the study may have incorrectly estimated the true reinfection rate. It should also be noted that there may have been unmeasured underlying behavioral differences in those who were and were not previously infected however, the contribution of such differences to risk of reinfection was not assessed.

Study design

Retrospective Cohort

Study population and setting

The study investigated the risk of re-infection among individuals previously infected with SARS-CoV-2. Patients within a single health system in Ohio and Florida who were tested for COVID-19 between March 12, 2020 and February 24, 2021 were included in the retrospective cohort, regardless of test results. Within individuals, positive SARS-CoV-2 tests prior to August 30, 2020 were considered an initial infection. Following this initial diagnosis, re-infection was defined as an individual’s second positive PCR test from September 1, 2020 onwards. Individuals who tested negative before August 30, 2020 and later positive from September 1 onwards were used as the comparison group to determine whether there were differences in the risk of infection in September onwards due to prior infection history. Reinfection was determined by the rate of second infection among all previously diagnosed patients tested 4 to 5 months after initial test, 6 to 7 months after, and 8 or more months after. Protection due to prior infection was defined as (1 – the ratio of infection among initially positive vs. initially negative patients). Repeated tests within 90 days of an initial test were ignored. Patients who tested negative at their initial test and then later positive within 90 days were excluded from the analysis. Symptoms were recorded by the ordering provider at the time of test order. Sensitivity analyses included examining only symptomatic infections.

Summary of Main Findings

Overall, 612,611 tests were conducted among 386,336 people with an overall positivity rate of 9.9% during the study period. Before August 30, 38.9% (N=150,325) patients were tested and included in the analysis. Of these, 8,845 (5.9%) tested positive and 141,480 (94.1%) tested negative. After 90 days, 1,278 positive patients were retested and 63 were identified as possible reinfections (4.9%). Of these, 31 were symptomatic (49.2%). The average time to reinfection was 138.9 days (range: 90.2 to 294.9 days). Protection from reinfection due to prior infection was estimated at 81.8% (95% CI: 76.6 – 85.8%) and 84.5% against symptomatic infection (95% CI: 77.9 – 89.1%). Protection was lowest at 4 to 5 months and greatest at 8 months after initial infection.

Study Strengths

The study included a large number of patients, which allowed them to make inferences around reinfection, a rare outcome. They also had multiple tests on individuals, which allowed them to compare individuals testing positive vs. negative at baseline to examine whether there was a potentially protective benefits for initial infection. They were also able to examine symptomatic infection and asymptomatic infection rates, which provides important granularity into the severity of disease following first infection.

Limitations

The major limitation was that they inferred a probable reinfection due to temporality, however it was not possible to confirm a reinfection with a negative test between the initial positive and later positive test. While it is expected that infections would clear within 3 months, there are cases of long-haul COVID-19 that may continue to test positive due to the same initial infection. This may have overestimated the actual reinfection rate in their sample. Additionally, the study did not have access to consumer testing data, such as at pharmacies or ordered online, or tests from other health facilities. The study also did not conduct any genetic testing to estimate if the infections were new or reflect novel variants. The use of timing as the cutoff to determine reinfection may also reflect temporal trends in different risk behaviors (such as less socializing outdoors during the summer than during September and the winter), and it is unclear if this confounded any results. Additionally, any re-tests that occurred in August were also ignored, which may have decreased their sample size unnecessarily and potentially underestimated reinfections. Finally, there may be behavioral differences between those previously infected with SARS-CoV-2 and those not previously infected which are difficult to predict (some may assume immunity, others may take further pre-cautions following an infection) which may lead to differences in infection rates between those previously infected and uninfected, but may be partially due to behavioral differences rather than just biological effects.

Value added

This is one of the largest studies of reinfection to date, though the study may have biased estimates of reinfection risk.

Our take —

S gene target failure (SGTF) is a hallmark of the B.1.1.7 variant. Among 2,245,263 individuals who tested positive for SARS-CoV-2 in the community in England between November 1, 2020 and February 14, 2021, 0.9% died, and 51% had conclusive SGTF test results (59% of whom had SGTF). SGTF was associated with >50% increased hazard of death across all analyses, accounting for covariates, missing data, and misclassification, but bias due to non-random missing data and residual confounding cannot be ruled out.

Study design

Retrospective Cohort

Study population and setting

This study includes 2,245,263 (54% female, 79% < 55 years old) individuals who had a positive SARS-CoV-2 test at a community testing site (“Pillar 2” testing) in England between November 1, 2020 and February 14, 2021. Data were combined across three Public Health England datasets: 1) specimen date and demographic information from community testing, 2) cycle threshold values for ORF1ab, N( nucleocapsid), and S (spike) genes from three national laboratories, and 3) deaths due to COVID-19 in England reported to the Office for National Statistics. S gene target failure (SGTF), a hallmark of the B.1.1.7 variant, was defined as a SARS-CoV-2 test result with cycle threshold (CT) values <30 for both ORF1ab and N targets, but no detectable S gene (CT>40). Complete case and inverse-probability weighted models (to account for missing SGTF status) estimated the hazard of death associated with SGTF were adjusted for several covariates, including age, sex, ethnicity, index of multiple deprivation, place of residence, national health service (NHS) England region, and test date. Absolute risks of death were estimated from the adjusted hazard ratios from complete case analysis and compared to “baseline” estimates of absolute risk of death for sex and age groups using data from individuals tested during August-October 2020. SGTF was used as a proxy for B.1.1.7 variant detection, and misclassification analyses explored potential biases resulting from this.

Summary of Main Findings

Fewer than 1% of people (N=19,615) in the study had a death recorded, and, overall, only 51% of had conclusive SGTF results; of those, 59% had SGTF (30% of the total study population). Prevalence of SGTF increased throughout the study period from 5.8% to 94.3%. For the 51% of individuals with known SGTF status, the death rate was higher among those with SGTF than those without (1.86 vs. 1.42 per 10,000 person-days), and this elevated death rate was consistently seen in nearly all (94%) analyses stratified for covariates. In adjusted models, SGTF was associated with increased risk of death in complete case analysis (hazard ratio (HR): 1.55, 95% confidence interval (CI): 1.39-1.72) and when using inverse probability weighting to account for missing SGTF status (HR: 1.58, 95% CI: 95% CI: 1.40-1.78). There was evidence of non-proportional hazards, indicating that as time since positive test increased, the hazard ratio for death comparing SGTF and non-SGTF groups also increased. For individuals <70, absolute risk of death was <1% for baseline and SGTF groups, though slightly elevated among individuals with SGTF, whereas absolute risk of death for SGTF vs. baseline was markedly increased among older individuals (females 70-84: 4.4% vs. 2.9%, females 85+: 19% vs. 13%, males 70-84: 7.2% vs. 4.7%, males 85+: 25% vs. 17%). Results were consistent in analyses that accounted for potential misclassification due to using SGTF as a proxy for the variant of concern B.1.1.7.

Study Strengths

This was a very large study, and included recent data on this variant of concern. Analyses were robust to different methods of handling missing data, non-proportionality of hazards over time, and misclassification of SGTF. Analyses were adjusted for several covariates that may skew the observed association between SGTF (a proxy for B.1.1.7) and death.

Limitations

Almost half of eligible participants were missing data on SGTF and were excluded from the analysis. The missingness was not random (i.e. related to several important covariates, including age and location of residence) which may lead to biased results. Additionally, the study population included only individuals who were tested for SARS-CoV-2 in the community, and thus may not be generalizable to individuals who first tested positive in the hospital, which is where most severe cases, including majority of cases leading to mortality, were first tested. Other variants lead to SGTF, and B.1.1.7 prevalence among SGTF samples was not confirmed via sequencing; results should be cautiously interpreted as B.1.1.7 specific.

Value added

This was a very large and robust study with data linked across Public Health England evaluating the impact of S gene target failure (as a proxy for the B.1.1.7 variant of concern) on mortality.

Our take —

This study was a preprint and thus was not yet peer-reviewed. Similar to previous studies, this large retrospective cohort study in England from November 2020 through January 2021 found that S gene target failure, the testing anomaly associated with the B.1.1.7 variant of concern, was associated with >50% risk of mortality and critical care admission among individuals tested in the community. Among a subset of individuals receiving critical care, SGTF was not associated with mortality or duration of organ support. However, these results should be interpreted with caution given that about 50% of persons with SARS-CoV-2 infections were excluded because their SGTF status was unknown.

Study design

Retrospective Cohort

Study population and setting

This retrospective cohort study included two cohorts (a primary care cohort and a critical care cohort) of patients diagnosed with COVID-19 in England who had available data on S gene target failure (SGTF) status, a proxy for the B.1.1.7 mutation. Results from COVID-19 testing in England are reported to Public Health England, including molecular diagnosis of SGTF, which was defined as non-detectable S gene and cycle threshold less than or equal to 30 for N and ORF1ab target. The primary care cohort included 198,420 who tested positive for SARS-CoV-2 in the community between November 1, 2020 and January 26, 2021 who were followed to investigate the association between SGTF and receipt of critical care or death due to COVID-19 (confirmed or suspected) within 28 days of the positive COVID-19 test. The critical care cohort included 3,452 individuals who tested positive for SARS-CoV-2 in the community between November 1, 2020 and January 27, 2021 and who received critical care. This cohort investigated the association between SGTF and duration of organ support, (e.g. respiratory, cardiovascular, renal, liver, or neurological support), duration of critical care, and mortality at the end of critical care.

Summary of Main Findings

59.4% of individuals in the primary care cohort had SGTF. In analyses adjusting for demographics, comorbidities, and date of positive test, the hazard ratio for mortality for the SGTF group vs. the non-SGTF group was 1.59 (95% CI: 1.25-2.03) and for critical care admission was 1.99 (95% CI: 1.59-2.49). Results were consistent among individuals with at least 28 days of follow-up (mortality endpoint) or 20 days of follow-up (critical care endpoint). Among the 3,452 individuals included in the critical care cohort, 58.5% had SGTF and 822 (24%) died during the study period. After adjusting for confounders, SGTF was not associated with mortality or other indicators of COVID-19 severity among those already in clinical care.

Study Strengths

This was a large study drawing on validated and broadly representative data sources in England.

Limitations

The study population included in each analysis was not well described, and only about half of potentially eligible participants (people tested in the community) had SGTF status available. If SGTF results were not missing randomly, this could bias the results. SGTF was used as a proxy for the B.1.1.7 mutation: though the authors report that >99% of SGTF during the study period were due to B.1.1.7, this was not confirmed with sequencing data. The study population included only individuals who were tested for COVID-19 in the community, and it is unknown whether the association between SGTF and outcomes among individuals in critical care who were first tested at the hospital differs from those tested in the community.

Value added

This study adds to growing literature about increased severity of COVID-19 due to SGTF, a proxy for the B.1.1.7 SARS-CoV-2 variant, and is the first to explore disease outcomes related to SGTF among patients receiving critical care.

Our take —

The US CDC recently updated its guidelines for schools to recommend at least three feet of spacing between students in classroom settings, half of the six feet it had previously recommended. This study of 251 Massachusetts school districts provides some of the only evidence from a head-to-head comparison of different distancing requirements. The authors found no statistically significant difference in the incidence rates of SARS-CoV-2 infection comparing districts requiring at least 6 feet of distance to districts requiring only 3 feet of distance. Actual classroom spacing was not observed or measured; district-level requirements were used as a proxy for spacing. The study was conducted during a time when most schools had lower attendance than normal and required near-universal masking, and could not determine whether infections were acquired in school or elsewhere.

Study design

Retrospective Cohort

Study population and setting

This study compared COVID-19 case counts in 251 Massachusetts school districts with different requirements for physical distancing between students (at least three feet vs. at least six feet) from September 24, 2020 to January 27, 2021. Districts were classified by the minimum distance permitted across grades (e.g., a district with a minimum requirement of at least 3 feet for some grades, even if other grades required at least 6 feet or greater distances were described as “preferred,” would be classified as at least 3 feet). Districts with intermediate distance requirements were excluded from the primary analysis. Data on distance requirements in district-level infection control plans were taken from publicly available sources, and were abstracted and classified by three investigators. District-level case counts were taken from the website of the Massachusetts Department of Elementary and Secondary Education, to which districts were required to report cases on a weekly basis. Cases were defined by a positive laboratory test for SARS-CoV-2 in a student or staff member who had been physically present in a school building during the seven days preceding the positive test. Weekly incidence was calculated by district for students and staff separately, and tested for association with distancing requirements (at least 6 feet vs. at least 3 feet) with negative binomial regression. Analyses were adjusted for community SARS-CoV-2 incidence (at the county level weighted by zip code), proportion of children living in poverty in the district, and the racial/ethnic distribution of enrolled students in the district.

Summary of Main Findings

In the 251 school districts during the study period, 537,336 students (6,400,175 person-weeks) and 99,390 staff members (1,342,574 person-weeks) were present in school buildings. There were 4,226 COVID-19 cases among students and 2,382 cases among staff members. Incidence rates among students and staff exhibited a high correlation with community incidence rates but were generally lower than those in the community. Most (77%) districts required at least 6 feet of distance between students, while 19% required at least 3 feet, and 4% had intermediate requirements. Most (64%) districts opened with less than 80% of students on campus during the study period. All districts required universal masking for all staff and for students in grade 2 and above, while 70% of districts required masking for younger grades. During the study period, most districts (>90%) implemented student cohorting, enhanced disinfection, and various ventilation interventions. The unadjusted incidence rate ratio (IRR) comparing districts requiring at least 6 feet of distancing to those requiring at least 3 feet of distancing was 0.89 (95% CI: 0.59 to 1.34) for students and 1.02 (95% CI: 0.75 to 1.37) for staff. After adjustment for community incidence and district-level demographic variables, the IRR was 0.79 (95% CI: 0.53 to 1.18) for students and 0.92 (95% CI: 0.67 to 1.25) for staff.

Study Strengths

Classification of distance requirements was performed carefully by three reviewers with several reliability checks. There appears to have been near-uniformity in masking requirements, which minimizes the possibility for confounding by this intervention.

Limitations

The study examined distance requirements at the district level, not actual distancing behavior in classrooms. There may have been considerable variation in the actual distance between students within a given classification, and this misclassification may have biased results toward the null. Most districts were operating with significantly lower student attendance than normal, which likely permitted greater distancing than would be possible otherwise. Thus, it may be the case that some districts requiring only 3 feet of distance were able in practice to achieve greater than 3 feet of distance. There were several differences in other interventions (e.g., ventilation, cohorting) between the two types of districts that the authors did not adjust for, though it appears from Table 2 that the interventions were more common in districts requiring 6 feet of distance. Although the authors adjusted for district-level demographics and county-level infection rates, there may have been heterogeneity of transmission determinants within districts that led to residual confounding. No information was available about whether new cases were acquired in schools or elsewhere. Data were not disaggregated by grade level; any effects of distancing may differ across age groups. Because of likely effect modification by mask use, results may not apply to environments in which masks are not universally mandated.

Value added

This is the first published comparison of different distance requirements in public schools in the context of SARS-CoV-2 transmission.

Our take —

In spring of 2020, Swedish upper-secondary schools moved to remote-only learning, while lower-secondary schools remained in person. This study reported on SARS-CoV-2 infections among parents with school-aged children as well as  teachers and their partners. Using a country-wide data source, the authors found that the odds of infection were 17% higher for parents of in-person vs. remote-only students and the odds of infection were double for teachers of in-person students vs. teachers of remote-only students. This corresponded to a relatively small number of additional infections (1 per 1,000) among parents of in-person students and a more substantive increase (5 per 1,000) among teachers who were exposed to in-person schooling. Although these increases were modest, the study did not measure impacts of in-person schooling in the wider community. Therefore, the paper’s claim that school reopening had a “minor impact on the overall spread of SARS-CoV-2 in society” is not directly substantiated.

Study design

Retrospective Cohort

Study population and setting

This study examined reported SARS-CoV-2 infections among parents, teachers, and teachers’ partners in Sweden from March 18 to June 15, 2020. During the study period, upper-secondary schools (school years 10-12, age 17-19 years) were engaged in online instruction, while lower-secondary schools (school years 7-9, age 14-16 years) continued to meet in person. The authors compared infections among parents/teachers/partners who were differentially exposed to in person school. The primary analysis included all parents in Sweden whose youngest child was in lower-secondary or upper-secondary school, excluding parents born outside the EU and Nordic countries (16% of all parents) out of a concern for selection bias (students with non-EU backgrounds are more likely to repeat upper-secondary years in preparatory programs). The analysis focused on the comparison between parents of the youngest children in the final year of lower-secondary or in the first year of upper-secondary school to minimize age-related confounding. Linear probability and logistic regression models were fit separately for the three populations under consideration (parents, teachers, and partners of teachers), with the outcome of interest being a positive SARS-CoV-2 test or a COVID-19 diagnosis. Models for parents were adjusted for age, sex, occupation, education, income, location of residence, and location of origin. Models for teachers and their partners were adjusted for the same covariates, and for partner and household characteristics.

Summary of Main Findings

Parents with a youngest child in the final year of lower-secondary school (in person) had an estimated 17% higher odds of SARS-CoV-2 infection (95% CI: 3% to 32%) relative to parents whose youngest child was in the first year of upper-secondary school (remote). No statistically significant association was seen in the odds of receiving a COVID-19 diagnosis after seeking healthcare. Lower-secondary school teachers had twice the odds (OR: 2.01, 95% CI: 1.52 to 2.67) of SARS-CoV-2 infection relative to upper-secondary school teachers. Partners of lower-secondary teachers also had higher odds of SARS-CoV-2 infection (OR: 1.29, 95% CI: 1.00 to 1.67). These estimates corresponded to an additional 1.05 infections per 1,000 parents, an additional 2.81 infections per 1,000 teachers, and an additional 1.47 infections per 1,000 partners of teachers. The results for parents, if assumed to be causal, imply that among the ~450,000 parents of lower-secondary students, closing schools would have decreased infections by 17% during the study period, excluding subsequent effects on transmission and epidemic growth. The authors conducted several robustness checks, and with some variations in magnitude of estimated effect, results were broadly similar.

Study Strengths

Emphasizing the comparison between parents of children in adjacent school years reduced the possibility of age-related confounding. The availability of data from national registries meant that nearly all members of the target population were included in the analysis, though SARS-CoV-2 testing detected only a fraction of actual infections. The authors performed a broad array of tests to examine effect estimates under alternate assumptions, such as excluding health care workers who may have been occupationally exposed.

Limitations

Because of limited testing capacity, estimates of additional reported infections may represent a small proportion of actual infections. Exclusion of parents from non-EU or Nordic countries resulted in better covariate balance between exposure groups, and the authors also fit models to the full parent population, but these families appear to have had higher infection risks overall, and in-person school may have had differential impacts in this group. The stated conclusion that overall impacts on transmission were minor and or/limited is not well substantiated in the absence of any estimation of subsequent extra-household transmission effects in the community.

Value added

By directly comparing parents and teachers of children differentially exposed to in-person school during the same time period, this study provides a reasonable estimate of increased reported SARS-CoV-2 infections associated with in-person secondary school.

Our take —

This cohort study compared the likelihood SARS-CoV-2 infection, COVID-19 hospitalization, and in-hospital mortality by race and ethnicity among 629,953 individuals tested for SARS-CoV-2 from March 1 to December 31, 2020 in a large health system on the West coast of the United States. Unsurprisingly, they found higher odds of SARS-CoV-2 infection, COVID-19 hospitalization, and in-hospital mortality in individuals who identified as members of historically marginalized racial and ethnic groups. While their analysis was not able to control for individual-level social determinants of health that may have blunted the relationship between race/ethnicity and adverse outcomes, it provides further evidence that the United States has a tremendous amount of work to do to address systemic and structural inequities that continue to result in persistent racial and ethnic disparities in health outcomes.

Study design

Retrospective Cohort

Study population and setting

This cohort study included 629,953 individuals who were tested for SARS-CoV-2 between March 1 and December 31, 2020 at Providence St. Joseph Health facilities in California, Oregon, and Washington state in the United States. The authors collected data from electronic medical records on participants’ age, sex, race, ethnicity, insurance, zip code, underlying medical conditions (by ICD-10-CM codes from January 1, 2019 to date of testing), Charlson Comorbidity Index, and body mass index. If participants were admitted to the hospital, authors also collected the participants’ presenting vital signs, laboratory results, supplemental oxygen use, hospital diagnoses, length of say, intensive care unit transfer, and disposition at discharge. Participants with more than one SARS-CoV-2 test during the study period were followed until their first positive test. Authors assessed demographic and clinical characteristics by participant race/ethnicity as well as the relationship between race/ethnicity and SARS-CoV-2 infection, COVID-19 hospitalization, and in-hospital mortality (through January 31, 2021) using a mixed-effects logistic regression model including random intercepts for each state and SARS-CoV-2 test month. Multiple imputation was conducted for covariates with less than 20% missingness.

Summary of Main Findings

Of the 49,081 (8.6%) individuals with a known race/ethnicity (n = 570,298) who tested positive for SARS-CoV-2, participants who identified as Hispanic (34.3% vs. 13.4%) or Native Hawaiians/Pacific Islanders (1.4% vs. 0.6%) were the most likely to test positive as compared to their likelihood of being tested. Among participants hospitalized with COVID-19, white participants were older and had higher Charlson Comorbidity Indices, but were less likely to have diabetes, hypertension, asthma, kidney disease, and obesity compared to those who identified as Hispanic, Black, Asian, Native Hawaiian/Pacific Islanders, or American Indians/Alaska Natives. After controlling for clinical characteristics, health insurance status, and zip code-derived neighborhood characteristics (median income, crowded housing, English proficiency), patients who identified as members of marginalized communities were more likely to test positive for SARS-CoV-2, be hospitalized with COVID-19, and, in the case of Hispanic participants, die in the hospital from COVID-19 when compared to white participants.

Study Strengths

This large cohort study from a large health system on the west coast of the United States presents SARS-CoV-2 infection data by race/ethnicity that aligns with State-level data, suggesting that the findings are likely generalizable to the region.

Limitations

The authors correctly note that these results are not causal and instead reflect underlying structural forces that likely drive COVID-19-related health inequities. Since they collected data through electronic medical records, information on other potential risk factors for COVID-19, such as occupation, or individual-level data on experienced discrimation or other sequelae of systemic racism in the United States were not included. Although they used zip code-derived neighborhood characteristics in their model, their random effects accounted for state-level, rather than zip code-level interdependence, which may have led to artificially narrow confidence intervals, overestimating the precision and significance of their estimates. Electronic medical record data are also prone to misclassification (for example, missed comorbidities) and, although they allowed for age to be non-linear in their final model, they categorized other linear covariates (body mass index and laboratory values), which complicates effect estimate interpretation and makes some dubious assumptions. Finally, approximately 9.4% of participants had missing race/ethnicity data.

Value added

This large study provides further evidence that individuals who belong to racial and ethinic groups that have been historically marginalized in the United States were disproportionately impacted by the COVID-19 pandemic.

Our take —

Patients with severe COVID-19 requiring mechanical ventilation meet the criteria for acute respiratory distress syndrome (ARDS), but it is not clear whether symptom severity and clinical outcomes differ between ARDS patients with and without COVID-19. This study, from a single center at the University of Michigan, compared 130 COVID-19 patients who required mechanical ventilation to 382 non-COVID-19 ARDS patients, and found similar degrees of lung impairment, extrapulmonary organ damage, and clinical outcomes between the two groups. Among patients with available laboratory measurements, COVID-19 patients had higher fibrinogen levels and platelet counts, suggesting differences in coagulopathy. The data do not suggest large differences in clinical features of severe COVID-19 relative to non-COVID-19 ARDS, but the study was fairly small and limited in its ability to control sources of bias. Moreover, study design choices may have artificially reduced disease severity in the cohort of COVID-19 patients.

Study design

Retrospective Cohort

Study population and setting

This study compared 130 COVID-19 patients who required mechanical ventilation to 382 patients with non-COVID-19 acute respiratory distress syndrome (ARDS) on mechanical ventilation. COVID-19 patients were consecutively admitted between March 1 and June 30, 2020 to a single center in the United States (University of Michigan); non-COVID-19 ARDS patients were consecutively admitted to the same center between January 1, 2016 and December 31, 2017. Patients were excluded from analysis if they received extracorporeal membrane oxygenation (ECMO) or were intubated for multiple days prior to transfer. Measures of hypoxemia, ventilation, lung function, and the highest sequential organ failure assessment score (SOFA) were extracted from electronic health records from the first 24 hours of mechanical ventilation. Laboratory values were taken from the nearest time within 48 hours of intubation. Lung compliance, ventilatory ratio, and oxygenation index were compared between cases and controls, both unmatched and matched by age, sex, BMI, and the first positive end expiratory pressure (PEEP) level. Time to unassisted breathing after intubation (up to 28 days) was assessed via competing risks survival analysis; patients who died before 28 days were assigned a time-to-event of more than 28 days. Patient status at 28 days and laboratory values (among patients with nonmissing values) were also compared between patients with COVID-19 and patients without COVID-19.

Summary of Main Findings

Respiratory compliance at time of ventilation was low among both COVID-19 (median 34.6 mL/cm H2O) and non-COVID-19 ARDS patients (median 30.0 mL/cm H2O). In a matched subgroup analysis (n=82 in each cohort), there were no statistically significant differences in median respiratory compliance or median oxygenation index between cohorts, while median ventilatory ratio was lower in COVID-19 patients than in non-COVID-19 patients (difference -0.21, 95% CI: -0.35, -0.07). At 28 days after ventilation, 30% of COVID-19 ARDS patients had died, while 38% of non-COVID-19 ARDS patients had died; neither the unadjusted risk ratio nor an adjusted risk ratio was statistically significant. When examining time to unassisted breathing, treating death as a competing risk in multivariable regression, the subdistribution hazard ratio comparing COVID-19 patients to non-COVID-19 patients was null (0.97, 95% CI: 0.73, 1.29). In exploratory analyses among the subset of patients with laboratory values available, measures of extrapulmonary organ injury at initiation of mechanical ventilation were similar between the two cohorts. Total white blood cell counts and neutrophil counts were lower among COVID-19 patients than among non-COVID-19 patients. Some measures of coagulopathy (platelet counts and fibrinogen) were higher among COVID-19 patients than among non-COVID-19 patients.

Study Strengths

The authors examined a range of functional and clinical outcomes, and performed competing risks survival analysis to appropriately examine potential differences in incidence of unassisted breathing after mechanical ventilation.

Limitations

Exclusion of patients who received ECMO or were intubated for several days prior to transfer may have biased results, particularly since the proportion of such patients was much higher in the COVID-19 cohort than in the non-COVID-19 ARDS cohort (70/214 vs. 10/414). Matching on PEEP may have been inappropriate if it was associated with ARDS severity. There were only 130 COVID-19 patients in the analysis, which limited generalizability and the ability to adjust for covariates. Because this was a retrospective cohort study, many patients were missing laboratory parameters that had not been measured at time of mechanical ventilation. The non-COVID-19 ARDS patients were treated during 2016-17, so clinical practices such as prone positioning may have varied between cohorts. COVID-19 patients were far more likely to be Black than non-COVID-19 ARDS patients (42% vs. 10%), and several analyses were not adjusted for self-reported race. With the exception of death and time to unassisted breathing, all indices were compared at time of initiation of mechanical ventilation; longitudinal differences in lung function and laboratory parameters following mechanical ventilation may also be of interest.

Value added

This is the largest direct comparison to date between mechanically ventilated COVID-19 patients and patients with non-COVID-19 ARDS.

Our take —

This manuscript, available as a preprint and thus not yet peer reviewed, shows that infection with the B.1.1.7 variant of SARS-CoV-2 is associated with a longer duration of infection, which may help explain increased transmissibility of the B.1.1.7 variant. However, only a limited number (n=65) of individuals were included in this study, of which only 7 were infected with the B.1.1.7 variant. The conclusions should be interpreted with caution until further studies with larger, more representative cohorts can be conducted.

Study design

Retrospective Cohort

Study population and setting

This study compared the duration of SARS-CoV-2 infection in individuals infected with the B.1.1.7 variant to those infected by non-B.1.1.7 virus. The study population was selected from an initial pool of 298 individuals who were affiliated with a professional sports league in the United States (players and staff) and who first tested positive for SARS-CoV-2 between November 28, 2020 and January 20, 2021. From this initial pool, a convenience sample was chosen of 65 individuals who had at least 5 positive PCR tests (nasopharyngeal swabs), of which at least one had a cycle threshold (Ct) value of less than 35. Of the 65 included individuals, 7 were found to have been infected with the B.1.1.7 variant. Daily surveillance testing of the 65 included individuals allowed the authors to track the duration of infection in each person, and to perform comparisons between individuals infected with the B.1.1.7 variant versus non-B.1.1.7 virus.

Summary of Main Findings

The authors found that acute infection with B.1.1.7 was associated with more sustained nasopharyngeal viral concentrations than infection with non-B.1.1.7 virus. Specifically, they found that individuals infected with B.1.1.7 had a mean proliferation time (time from first detectable virus to peak viral concentration) of 5.3 days, and a mean clearance time (time from peak viral concentration to initial return to limit of detection) of 8.0 days, resulting in a mean overall duration of infection of 13.3 days. Individuals infected with non-B.1.1.7 had a mean proliferation time of 2.0 days, and a mean clearance time of 6.2 days, resulting in a total duration of infection of 8.2 days. They also found a slight difference in the mean peak viral concentration: 19 Ct (B.1.1.7) versus 20.2 Ct (non-B.1.1.7).

Study Strengths

Daily surveillance of all infected individuals allowed the authors to accurately determine the beginning, maximum, and end of detectable viral infection.

Limitations

This study has a number of limitations: (1) The samples included were a convenience sample from within a specific population (individuals associated with a professional sports league) and therefore may not be representative of the general population. Similarly, no information was provided about the age or underlying conditions of these individuals (both of which are known to affect SARS-CoV-2 manifestation), and the final population was 90% male. (2) The study was conducted on a small number of samples (65), with only 7 belonging to the group of interest (B.1.1.7). Additional research is needed to determine if the findings presented hold true in a larger (and more diverse) population and if the observed differences (e.g., mean peak viral concentration of 19 versus 20.2 Ct) are statistically significant.

Value added

This paper is the first to suggest that there may be differences in duration of infection between individuals infected with B.1.1.7 versus non-B.1.1.7 SARS-CoV-2 variants. A difference in duration of infection may explain increased transmissibility of B.1.1.7, the reason for which is currently unknown.

Our take —

This manuscript, available as a preprint and therefore not yet peer-reviewed, demonstrated how SARS-CoV-2 emerged in, and then spread throughout Louisiana at the beginning of 2020. The authors showed that SARS-CoV-2 was likely introduced domestically from a single source (likely Texas), that SARS-CoV-2 was likely circulating in New Orleans before the Mardi Gras festival in February 2020, and that Mardi Gras likely acted as a super-spreading event, greatly increasing the number of cases and deaths, and likely also seeding local outbreaks in nearby cities and states. Although sampling bias can affect our understanding of transmission patterns, the integration of genomic, epidemiological, and mobility data improves our confidence in these findings. Additionally, this paper contributes to the growing body of evidence that large events without precautions early in the pandemic likely contributed to the rapid rise of COVID-19 cases, and that these types of events should be approached with caution in the future.

Study design

Retrospective Cohort

Study population and setting

This study aimed to understand the early COVID-19 epidemic in New Orleans, Louisiana, and explored the connection between the annual Mardi Gras festival (which occurred, without precautions, in New Orleans in February 2020) and the subsequent rapid spread of SARS-CoV-2 in the Southern US and the rest of the country. The study used aggregated COVID-19 case data to analyze reported cases and deaths during the first wave of the epidemic in Louisiana, which lasted from the first detection of COVID-19 in the state on March 9, to May 15, 2020. The authors also analyzed 235 SARS-CoV-2 virus genomes collected from patients in New Orleans, Shreveport, and other parishes in Louisiana during this time period, alongside a representative sample of 1,263 publicly available global sequences. Finally, the authors used domestic and foreign air travel records to understand transmission into the state, as well as human mobility data collected from weekly cell phone records available as part of the SafeGraph database.

Summary of Main Findings

Using genomic and epidemiological data, the authors found that SARS-CoV-2 was likely introduced into Louisiana via domestic, not international, travel. Specifically, they found that there was very little genetic diversity within Louisiana sequences from the first wave (also suggesting that SARS-CoV-2 transmission within Louisiana was predominantly seeded from a single introduction), and that all lineages observed closely resembled sequences from other parts of the US. This matched airline data, which showed that the vast majority of airline travel during that period was from domestic locations. Using the genetic data, the authors also estimated the date of these introductions and found that SARS-CoV-2 likely emerged in the state prior to the Mardi Gras festival (around February 13, 2020, with a 95% posterior density interval of January 24 to February 27, 2020). Coupled with the rapid increase in cases in early March, this suggests that the Mardi Gras festival in late February likely acted as a super-spreader event for SARS-CoV-2 transmission. Finally, the authors used genetic, mobility, and epidemiological data to conclude that SARS-CoV-2 in Louisiana likely came from Texas, and that the Mardi Gras-related outbreak in Louisiana likely caused localized outbreaks in nearby states.

Study Strengths

Combining genomic, epidemiological, and human mobility data allowed the authors to draw conclusions from one dataset and support those findings with additional data. This significantly improves our confidence in understanding the early transmission patterns of the epidemic in Louisiana.

Limitations

Phylogeographic analyses are inherently limited by sampling bias. While confirmation from other datasets helps, conclusions involving pinpointing the source of a particular SARS-CoV-2 lineage are particularly difficult to make with certainty. For example, export risk from New Orleans is likely biased towards states with larger populations, as these states will have more sequence data as well as more mobility records. Additionally, the limited genetic data observed in Louisiana makes it difficult to pinpoint an exact time of introduction.

Value added

This manuscript improves our understanding of how COVID-19 spread in Louisiana at the beginning of the outbreak. Identification of Mardi Gras as a super-spreading event also highlights the contribution of large events to increases in COVID-19 cases, which both aids our understanding of observed cases and deaths in New Orleans, while acting as a cautionary tale for other large gatherings without appropriate precautions.