Waning of vaccine effectiveness against moderate and severe covid-19 among adults in the US from the VISION network: test negative, case-control study…

Study design

The VISION network has been described previously.15 We applied a test negative design to estimate vaccine effectiveness of mRNA vaccines using retrospectively collected data. We focused on mRNA vaccines because they comprise more than 95% of covid vaccines administered in the US.16 Separate analyses were done of patients who were admitted to hospital (hospital sample) and patients who received care in an emergency department or urgent care clinic (emergency department or urgent care sample).

The study population included adults (18 years) who received care for covid-like illness at a VISION network hospital or emergency department or urgent care center and had molecular testing for SARS-CoV-2 at least 14 days after vaccines became locally available for their age group (17 January to 3 May 2021). The last contact included in this study period occurred on 12 July 2022. We excluded individuals who received any vaccine other than the BNT162b2 or mRNA-1273 vaccine, individuals who received more than four doses of an mRNA vaccine before the index medical contact, individuals who received only one dose of an mRNA vaccine less than 14 days before the index contact or who had a third or fourth dose less than seven days before the index contact, individuals known to have a positive laboratory test result for a SARS-CoV-2 infection more than 14 days before the index contact, and individuals with a positive SARS-CoV-2 test result but no diagnoses or symptoms suggesting covid-19 illness.

Vaccination status was categorized by the number of doses received and the number of months between the most recent vaccine dose and the index contact date (referred to as time since vaccination). Patients were considered partially vaccinated if they received only one dose at least 14 days prior to the index contact date or had received a second dose less than 14 days previously. Patients with no record of vaccination before the index contact date were considered unvaccinated. Patients with three doses were those who received a third dose in a primary vaccination series (eg, among immunocompromised individuals) or a booster dose after a primary series of two doses. Aligning with recommendations for receipt of a fourth dose, we examined the effectiveness of four doses among adults aged 50 years or older and among immunocompromised adults of any age. Vaccination status was ascertained from immunization registries, electronic health records, and insurance claims.

The primary outcome was a positive or negative molecular SARS-CoV-2 result for a test done within 14 days before a medical contact to less than 72 h after among patients presenting with covid-like illness, as identified from ICD-9 and ICD-10 (international classification of diseases, ninth and 10th revision, respectively) diagnostic codes (supplemental methods; supplemental table S1). The index date for each contact was the earlier of either the contact date or the date of the closest SARS-CoV-2 molecular assay. An individual could be included as a case once in the emergency department or urgent care sample and once in the hospital sample. Individuals could be included as a control multiple times.

We used a test negative case-control design in which cases were patients with covid-like illness with laboratory confirmed covid-19 and controls were patients with covid-like illness and negative SARS-CoV-2 test results (controls could have had positive test results for other respiratory viruses such as influenza). We compared cases with controls in the hospital sample, and separately compared cases with controls in the emergency or urgent care sample. Cases were not individually matched to controls.

Conditional logistic regression was used to examine case-control status in relation to vaccination status categorized as vaccinated with four, three, or two doses, or partially vaccinated; unvaccinated individuals were used as the reference group. To examine waning of vaccine effectiveness, we categorized people who were vaccinated using time specific indicators defined by two month intervals of time since vaccination; unvaccinated individuals were used as the reference group. We exponentiated the regression coefficient of each vaccination status indicator to yield an odds ratio, subtracted the odds ratio from 1 to estimate vaccine effectiveness, and multiplied by 100 to scale vaccine effectiveness as a percentage. In several analyses, a sparse bimonthly interval for which the vaccine effectiveness estimate had a confidence interval wider than 50 percentage points was combined with the previous bimonthly interval to provide a more precise estimate of vaccine effectiveness (see supplemental methods). Vaccine effectiveness estimates (and confidence limits) were scaled to a range of 100% to 100%.17

Logistic regression models were conditioned by calendar week and geographical area such that we compared cases with controls tested during the same week in the same region (supplemental table S2). Covariates included in the models were those determined through bivariate analyses to be statistically significantly associated with both the outcome and vaccination status, as well as those specified a priori as established confounders, including age, race, ethnicity, presence of respiratory and non-respiratory comorbidities, immunocompromised status, and local viral circulation. Cubic splines were used for age, seven day average positivity of SARS-CoV-2 test in the area of the contact, and the propensity to be vaccinated; others were indicator variables. Propensity scores (supplemental methods) predicted vaccination (any versus none) based on demographics, comorbidities (supplemental table S3), and characteristics of the facility (supplemental table S4), and were derived independently for each period of variant dominance (supplemental table S5). Patients who were immunocompromised were identified by ICD-9 and ICD-10 diagnostic codes (supplemental methods).18 We conducted separate analyses for three periods based on when a variant accounted for 50% or more of sequenced isolates in each site: before delta was predominant, when delta was predominant, and when omicron was predominant (supplemental table S6). We assessed the magnitude of waning as the difference in vaccine effectiveness between patients who had recently been vaccinated (defined as less than two months) and patients at a specified level of time since vaccination (eg, four to five months from dose 3), and we examined waning by age (18-44 years, 45-64 years, 65 years), vaccine product, and immunocompromised status. Bootstrapping was used to estimate a 95% confidence interval around the difference between vaccine effectiveness at less than two months and vaccine effectiveness at four to five months.

We conducted several sensitivity analyses. First, we added to the study population patients with a known prior infection to assess the sensitivity of results to whether previously infected patients are included or excluded.. Second, we wanted to distinguish results between patients who had been admitted to hospital and patients who had been admitted to an emergency department or to urgent care. Therefore, we examined vaccine effectiveness in the emergency department or urgent care sample and omitted patients admitted to hospital within 30 days. Third, we investigated a negative control exposure19 by examining vaccine effectiveness in patients who received their first dose less than 14 days before the index date of contact. These patients were not expected to have substantial vaccine induced protection, and a vaccine effectiveness estimate substantially more than zero would be evidence of residual confounding.20

Analyses were conducted with SAS version 9.4 and R version 4.1.2. All confidence limits are 95% intervals. Confidence intervals excluding the null value were considered statistically significant.

Study participants contributed in important ways to this research by supplying the underlying data on which the study is based. It was not, however, feasible to involve them in the design, conduct, reporting, or dissemination of this study because the study was conducted under the CDCs covid-19 incident response structure and limited to analysis of retrospectively collected electronic data only, with no patient interaction.

Original post:

Waning of vaccine effectiveness against moderate and severe covid-19 among adults in the US from the VISION network: test negative, case-control study...