Association between treatment failure and hospitalization after receipt of neutralizing monoclonal antibody treatment for COVID-19 outpatients

We completed a secondary analysis of a multicenter observational cohort study collaborating with leaders from the University of Colorado Hospital, University of Colorado Health (UCHealth), and the Colorado Department of Public Health and Environment (CDPHE). The study was approved by the Colorado Multiple Institutional Review Board (COMIRB) with a waiver of informed consent. We previously reported methods for data collection [7]. In brief, we accessed patient data from the electronic health record (EHR; Epic, Verona, WI) of UCHealth, the largest health system in Colorado. UCHealth consists of 13 hospitals across the state and accounts for approximately 141,000 annual hospital admissions. Data from the EHR were merged with statewide data on vaccination status from the Colorado Comprehensive Immunization Information System and mortality from Colorado Vital Records.

We included adult patients (≥ 18 years of age) diagnosed with SARS-CoV-2 between November 20, 2020, and December 9, 2021, who received mAb treatment within 10 days of a positive SARS-CoV-2 test (n = 7974). We identified patients using EHR-based date of SARS-CoV-2 positive testing (by polymerase chain reaction or antigen tests) or date of administration of mAb treatment (if no SARS-CoV-2 test result date was available). The decision to utilize mAb treatment was made by patients and clinicians. The CDPHE established a statewide referral system to facilitate patient referrals to facilities for mAb infusion [13]. We excluded patients who died before their SARS-CoV-2 test results, patients missing both a mAb administration date and a SARS-CoV-2 positive date, and patients who tested positive either while in the hospital or on the same day as admission as these patients were not eligible to receive mAbs and not at risk for the primary study outcome. We also excluded patients with a positive SARS-CoV-2 test after December 9, 2021, to avoid comparing patients infected with the Delta variant with those infected with the Omicron variant, which had a much lower rate of hospitalization or death [14]. We also excluded patients who had more than 10 days between their SARS-CoV-2 positive test and the date of mAb administration, in accordance with the FDA’s Emergency Use Authorization (EUA) criteria (Fig. 1). For patients who were missing a documented SARS-CoV-2 test date, a test date was imputed based on the distribution of observed times between SARS-CoV-2 positive test to mAb administration.

The primary outcome was mAb treatment failure, defined as hospitalization or death within 28 days of a positive SARS-CoV-2 test obtained from EHR data. Hospitalization was defined as an inpatient or observation encounter documented in the EHR. Death was defined as all-cause mortality whether the patient was hospitalized or not.

The presence and status of comorbid conditions (cardiovascular disease, hypertension, pulmonary disease, renal disease, diabetes) were determined using the Charlson and Elixhauser comorbidity indices [15] for EHR data and a 90-day lookback period. Immunocompromised status was further validated by manual chart reviews and was categorized as “Not Immunocompromised,” “Mild,” or “Moderate/Severe,” based on chronic medications or specific conditions. Mild criteria included the administration of tumor necrosis factor-alpha inhibitors, or azathioprine, as well as human immunodeficiency virus (HIV) without acquired immunodeficiency syndrome (AIDS). Moderate/Severe patients were NIH tier 1 immunocompromised individuals, including those receiving chemotherapeutic agents or antirejection medications, and HIV with AIDS (Additional file 5: Table S1). If a patient met at least one defining element for “Moderate/Severe,” they were categorized as “Moderate/Severe” immunocompromised. Systemic corticosteroids, excluding dexamethasone, were classified as “Mild” immunocompromised. The number of comorbid conditions was created by summing all conditions excluding immunocompromised status or obesity because these two were the key pre-specified conditions we chose to evaluate. Pandemic phase was divided based on SARS-CoV-2 positive date and in accordance with the prevalent variant in Colorado. Phases included Pre-Alpha (November 2020–February 2021), Alpha (March 2021–June 2021), and Delta (June 2021–December 2021). Virus sequencing results were not available at an individual patient level. Vaccination status at the time of SARS-CoV-2 positive date was categorized by the number of vaccine doses received (zero, one, two, or more) at least 2 weeks before infection. mAb treatments included bamlanivimab (Eli Lilly), casirivimab + imdevimab (Regeneron), bamlanivimab + etesevimab (Eli Lilly), and sotrovimab (GlaxoSmithKline).

Patient characteristics by treatment failure status were compared using t-tests and Chi-squared test statistics, as appropriate. Due to the infrequent nature of our primary outcome, Firth’s bias-reduced multiple logistic regression model was used to investigate the association between patient risk factors and treatment failure [16]. For each of the risk factors in the model, we computed the adjusted odds ratio (OR) and 95% confidence intervals (95% CI) via penalized profile likelihood. A separate model was fitted for each risk factor along with a final multivariable model including age, sex, race/ethnicity, insurance status, presence of obesity, insurance status, immunocompromised status, number of comorbidities, pandemic phase, and vaccination status to adjust association estimates for potential confounding effects between risk factors. In addition, we computed the absolute risk difference along with 95% CI. Kaplan–Meier curves were estimated to visually assess temporal trends in cumulative incidence by treatment status for secondary outcomes.

We performed two sensitivity analyses. For the first, we employed a more conservative imputation method for missing SARS-CoV-2 positive test dates by assuming all missing positive test dates were 10 days prior to the mAb administration date (the maximum time difference allowed by the EUA). Second, we included only patients with complete dates for both SARS-CoV-2 positive test and mAb administration dates verified by EHR data. All statistical analyses were performed using R Statistical Software (version 3.6.0; R Foundation for Statistical Computing, Vienna, Austria).

A total of 7406 patients with confirmed SARS-CoV-2 infection received mAbs between November 20, 2020, and December 9, 2021. A CONSORT flow diagram of applied exclusion criteria is presented in Fig. 1. We excluded 47 patients who died before their SARS-CoV-2 test results, 377 patients missing both a mAb administration date and a SARS-CoV-2 positive date, and 109 patients who tested positive for SARS-CoV-2 either while in the hospital or on the same day as admission. We also excluded 35 patients who had more than 10 days between their SARS-CoV-2 positive test and the date of mAb administration, in accordance with EUA criteria. Treatment failure within 28 days of a positive SARS-CoV-2 test occurred in 258 (3.5%) patients who received mAbs. Ten patients (0.1%) died within 28 days of their positive test, and all but one of these patients were hospitalized prior to death. The maximum level of oxygenation support required during the index hospitalization for the 258 patients who experienced treatment failure is displayed in Additional file 1: Figure S1.

Patients who experienced treatment failure were older (47% vs. 33% were age ≥ 65 years), more likely to be male (55% vs. 44%), more likely to be obese (46% vs. 27% with body mass index [BMI] ≥ 30 kg/m²), more likely to be at least mildly immunocompromised (37% vs. 23%) and more likely to have two or more additional comorbidities (63% vs. 30%) (Table 1).

All patient risk factors were significantly associated with treatment failure during initial (unadjusted) analysis (Table 1). After accounting for potential confounding of insurance status and different comorbid conditions, the following risk factors remained as significant independent predictors of treatment failure: age ≥ 65 years old compared to those younger than 45 years old (adjusted OR 1.62, 95% CI 1.13–2.35, p = 0.008), male versus female sex (OR 1.56, 95% CI 1.21–2.02, p = 0.001), Non-Hispanic black race versus non-Hispanic white race (OR 2.21, 95% CI 1.20–3.82, p = 0.013), and obesity (OR 1.79, 95% CI 1.36–2.34, p < 0.001) (Fig. 2). The cumulative number of additional comorbid conditions was also significantly associated with hospitalization: patients with one additional comorbidity had OR 1.68 (95% CI 1.11–2.57, p = 0.014), and those with two or more additional comorbidities had OR 3.71 (95% CI 2.52–5.56, p < 0.001) compared to those with no additional comorbidities. Compared to patients who received two or more doses of vaccines, those who received no SARS-CoV-2 vaccine had a nearly three times higher likelihood of experiencing treatment failure (adjusted OR 2.73, 95% CI 2.01–3.77, p < 0.001).

Presence of at least mild immunocompromised status was significantly more likely in hospitalized patients compared to non-hospitalized patients in univariate analysis (37% vs. 23% in Table 1; unadjusted OR 2.02, 95% CI 1.55–2.61, p < 0.001). However, after adjusting for other risk factors, neither mild immunocompromised status (OR 1.39, 95% CI 0.94–1.99) nor moderate/severe immunocompromised status (OR 1.36, 95% CI 0.97–1.91) was significantly associated with treatment failure, compared to non-immunocompromised patients (Fig. 2). Pandemic phase and type of mAb received were also not associated with treatment failure. The cumulative incidence of hospitalization stratified by immunocompromised status is displayed in Additional file 2: Figure S2, by sex in Fig. 3, and by patient age in Fig. 4. After adjusting for relevant characteristics, increasing age was significantly associated with treatment failure; however, age 45–64 years was not (Fig. 2).

We first employed a more conservative imputation method for missing SARS-CoV-2 positive test dates by assuming all missing positive test dates were 10 days prior to the mAb administration date (the maximum time difference allowed by the EUA). In that sensitivity analysis, we found that Hispanic ethnicity (adjusted OR 1.63, 95% CI 1.13–2.31, p = 0.01) and moderate/severe immunocompromised status (OR 1.42, 95% CI 1.02–1.97; p = 0.04) were significantly associated with hospitalization (Additional file 3: Figure S3). We did not observe any other differences between our primary analysis and this analysis for other risk factors. In this analysis, the overall sample size changed slightly because SARS-CoV-2 positive test date was used in defining our initial eligibility criteria. For the second sensitivity analysis, we included only patients with confirmed dates for both SARS-CoV-2 positive test and mAb administration. In this cohort, we found that age ≥ 65 years was not associated with treatment failure (OR 1.65, 95% CI 0.98–2.82), and patients who received only one dose of SARS-CoV-2 vaccine had a higher risk of experiencing treatment failure compared to those who had two or more doses of vaccine (OR 2.55, 95% CI 1.19–5.16, p = 0.02) (Additional file 4: Figure S4). No other differences were observed between this analysis and our primary analysis.