key: cord-0992532-kh53z9ni
authors: Aggarwal, N.; Beatty, L.; Bennett, T. D.; Carlson, N.; Davis, C.; Kwan, B.; Mayer, D.; Ong, T.; Russell, S.; Steele, J.; Wogu, A.; Wynia, M.; Zane, R.; Ginde, A. A.
title: Real-World Evidence of the Neutralizing Monoclonal Antibody Sotrovimab for Preventing Hospitalization and Mortality in COVID-19 Outpatients
date: 2022-04-05
journal: medRxiv : the preprint server for health sciences
DOI: 10.1101/2022.04.03.22273360
sha: a10182adf0a99ef4a681fb3884db19abcf1812d8
doc_id: 992532
cord_uid: kh53z9ni

Background: It is not known whether sotrovimab, a neutralizing monoclonal antibody (mAb) treatment authorized for early symptomatic COVID-19 patients, is effective against the SARS-CoV-2 Delta variant to prevent progression to severe disease and mortality. Methods: Observational cohort study of non-hospitalized adult patients with SARS-CoV-2 infection from October 1st 2021 - December 11th 2021, using electronic health records from a statewide health system plus state-level vaccine and mortality data. We used propensity matching to select 3 patients not receiving mAbs for each patient who received outpatient sotrovimab treatment. The primary outcome was 28-day hospitalization; secondary outcomes included mortality and severity of hospitalization. Results: Of 10,036 patients with SARS-CoV-2 infection, 522 receiving sotrovimab were matched to 1,563 not receiving mAbs. Compared to mAb-untreated patients, sotrovimab treatment was associated with a 63% decrease in the odds of all-cause hospitalization (raw rate 2.1% versus 5.7%; adjusted OR 0.37, 95% CI 0.19-0.66) and an 89% decrease in the odds of all-cause 28-day mortality (raw rate 0% versus 1.0%; adjusted OR 0.11, 95% CI 0.0-0.79), and may reduce respiratory disease severity among those hospitalized. Conclusion: Real-world evidence demonstrated sotrovimab effectiveness in reducing hospitalization and all-cause 28-day mortality among COVID-19 outpatients during the Delta variant phase.

Patients were identified using an EHR-based date of SARS-CoV-2 positive test 108 (by polymerase chain reaction or antigen) or date of administration of mAb treatment (if no 109 SARS-CoV-2 test result date available). The decision to seek mAb treatment was made by 110 patients and clinicians. [11] We did not exclude patients solely for lack of EUA eligibility based 111 on EHR data, because not all eligibility criteria were consistently available in the EHR. We 112 excluded patients who tested positive for SARS-CoV-2 on the same day of or during 113 hospitalization because they were not eligible for mAb treatment. We also excluded patients 114 missing both a positive test date and a sotrovimab administration date (n=708), or if it had been 115 more than 10 days between the positive test date and sotrovimab administration (n=26), resulting 116 in a cohort of sotrovimab (N = 566) or mAb untreated (N = 9,470) patients. 117

Nearest neighbor propensity matching was conducted using logistic regression with 118 treatment status as the outcome. Approximately three untreated patients (N=1,563) were matched 119 to each sotrovimab-treated patient (N=522). [12, 13] Only 42 sotrovimab treated patients were 120 lost due to incomplete covariate data. The propensity model included categorial age, sex, 121 race/ethnicity, obesity status, immunocompromised status, number of comorbid conditions other 122 than obesity and immunocompromised status, number of vaccinations at time of infection, and 123 insurance status. We assessed effectiveness of matching using standardized mean differences 124 (SMDs) with a threshold of 0.1 with results shown in Appendix The primary outcome was all-cause hospitalization within 28 days of a positive SARS-CoV-2 128 test, obtained from EHR data. Secondary outcomes included 28-day all-cause mortality, 129 emergency department (ED) visit within 28 days, in-hospital disease severity based on maximum 130 level of respiratory support, hospital and intensive care unit (ICU) lengths of stay (LOS) in survivors, rates of ICU admission, and in-hospital mortality. For both hospitalization and ED 132 visits, the index visit was used. When mAb treated patients were missing a SARS- (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. Due to the small number of hospitalized participants, descriptive statistics including 168 counts and raw rates were calculated for all secondary outcomes among hospitalized participants, 169 including disease severity, hospital LOS, ICU visit, and ICU LOS. 170

Kaplan-Meier curves were estimated to visually assess cumulative incidence patterns by 171 treatment status for 28-day hospitalization. 172

Two sensitivity analyses were performed. First, we repeated the above analysis using 173 only EUA-eligible patients as verified by available EHR data. Second, we repeated the above 174 analysis with a more conservative SARS-CoV-2 imputation approach where all missing positive 175 test dates were imputed as ten days prior to the mAb administration date (the maximum time 176 difference allowed by the EUA). All statistical analyses were performed using R Statistical 177 Software (version 3.6.0; R Foundation for Statistical Computing, Vienna, Austria).[19] 178 179 All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 5, 2022. ; https://doi.org/10.1101/2022.04.03.22273360 doi: medRxiv preprint

Of 10,036 patients with SARS-CoV-2 infection in the full cohort, 566 subjects received mAbs 182 and 9,470 patients did not (Appendix Table 1 , Supplement). In the full cohort, the sotrovimab-183 treated group generally reflects EUA criteria for use of mAbs, with many being older (33.0% 184 were age ≥ 65 years vs. 11.4% in mAb-untreated group), more likely to be obese (24.0% vs. 185 16.8%), or having one or more comorbidities (49.1% vs. 36.5%). Propensity matching eliminated 186 clinically meaningful differences in matching variables between groups ( The characteristics of sotrovimab-treated and mAb-untreated patients in the matched 190 cohort are presented ( Table 1 ). The age distribution was similar, with 34% aged ε65. The cohort 191 was 56% female, 81% Non-Hispanic white, and 56% had private/commercial insurance. 192

Hypertension (32%) and pulmonary disease (25%) were the most common comorbid conditions. 193

Notably, 54% had received at least two vaccinations at the time of infection, and 39% had not 194 received any vaccine doses. The mean time from positive SARS-CoV-2 test to administration of 195 sotrovimab treatment was 3.7 days (SD 1.8) in those who did not have an imputed positive test 196 date. 197

Sotrovimab treatment was associated with a lower rate of 28-day hospitalization compared to 199 matched mAb-untreated controls (11 [2.1%] vs. 89 [5.7%]), representing a 63% decrease in the 200 adjusted odds of hospitalization (adjusted OR 0.37, 95% CI 0.19-0.68; p < 0.001) ( Table 2 ). The 201 unadjusted NNT for hospitalization for the untreated group was 28. Based on a time-to-event 202 analysis, the benefits associated with reduced hospitalization were largely accrued within 12 203 All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 5, 2022. ; https://doi.org/10.1101/2022.04.03.22273360 doi: medRxiv preprint days of the positive SARS-CoV-2 test date (Figure 1) , or an average of 9 days after sotrovimab 204 treatment. Other factors associated with hospitalization are in Supplemental material (Appendix 205 Table 3 , Supplement). Covariates that were associated with increased odds of 28-day 206 hospitalization included age ≥ 65 (p = 0.049), obesity (p < 0.001), and one (p = 0.013) or two or 207 more (p < 0.001) comorbid conditions other than obesity or immunocompromised status 208 (Appendix Table 3 Importantly, all-cause 28-day mortality in the sotrovimab-treated group was 0 (0%) 212 compared to 15 (1.0%) among the mAb-untreated group, equating to an 89% decrease in the 213 mortality odds (adjusted OR 0.11, 95% CI 0.0-0.79 (Table 2) . There was not a significant 214 association between sotrovimab treatment and the odds of visiting the ED (adjusted OR 1.12, 215 95% CI 0.77-1.60). 216

Among hospitalized patients, 0 of 11 (0%) in the sotrovimab-treated group required invasive 218 mechanical ventilation (IMV) or died in the hospital, compared to 19 of 89 (21.3%) mAb-219 untreated group (Table 2) . We also observed that a higher proportion of sotrovimab-treated 220 patients required no supplemental oxygen or only required standard (low-flow) oxygen in 221 comparison to mAb-untreated patients (72.7% vs. 48.3%). The average hospital length of stay 222 (LOS) for sotrovimab patients was 5.3 (+/-5.9) days in comparison to 9.4 (+/-10.6) days in the 223 untreated group. Collectively, these data suggest a lower severity of disease among hospitalized 224 sotrovimab-treated patients, although statistical inference was not performed (Figure 2 ). 2 of 9 225 (18.2%) sotrovimab-treated patients required ICU level of care, a similar percentage compared to 226 mAb-untreated patients (21.3%). 227 All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. This study has several limitations. The setting was a single health system and geographically 261 limited to one US state with relatively low racial and ethnic minority representation, though it 262 serves both urban and rural populations through academic and community hospitals. Even 263 though we used statewide data for mortality and vaccination status, hospitalizations were 264 collected only within one single health system. If mAb-untreated patients were less likely to be 265 seen in this health system, hence more likely to be hospitalized elsewhere, this may bias our 266 results toward the null. We also relied on EHR data, including manual chart reviews, which may 267 have missing or inaccurate information about the presence of chronic conditions.[23] These 268 factors might have limited our ability to detect the impact of sotrovimab treatment. 269

We only collected 28-day hospitalization and mortality data, and therefore we do not 270 know whether sotrovimab effectiveness extends to a longer period after SARS-CoV-2 infection. 271

However, our prior study would suggest that 28-day and 90-day data are similar with respect to 272 hospitalization and mortality endpoints. [7] In this study, propensity scoring achieved excellent 273 matching between mAb-treated and mAb-untreated patient groups across multiple variables, but 274 All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. Finally, this study was conducted during a limited portion of the Delta variant-dominant 278 period after sotrovimab drug distribution and infusion had been well-established in Colorado. In 279 addition, hospitalization rates were lower over this same period, precluding our ability to 280 perform inferential statistics on severity of illness among the hospitalized sub-cohort. Because 

This study demonstrated real-world evidence for effectiveness of sotrovimab treatment in 292 reducing hospitalizations among COVID-19 outpatients during the Delta variant phase, as well 293 as a remarkable 89% overall reduction in mortality at 28 days, compared to matched mAb-294 untreated patients. For hospitalized patients, prior outpatient sotrovimab treatment may reduce 295 respiratory disease severity, hospital length of stay, and death, but a larger cohort is necessary to 296 further examine this observation. When access to mAbs is limited, prioritizing patients at highest 297 All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 5, 2022. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 5, 2022. ; https://doi.org/10.1101 https://doi.org/10. /2022 (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. Comparing severity of hospitalizations for n=11 sotrovimab-treated and n=89 mAb-untreated 425 patients, the maximum level of respiratory support appeared lower for sotrovimab-treated 426 patients, but inferential statistics were not able to be performed. 427

Abbreviations: mAb, monoclonal antibody; HFNC, high-flow nasal cannula oxygen; NIV, non-428 invasive ventilation; IMV, invasive mechanical ventilation 429 All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 5, 2022. ; https://doi.org/10.1101/2022.04.03.22273360 doi: medRxiv preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 5, 2022. ; https://doi.org/10.1101/2022.04.03.22273360 doi: medRxiv preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 5, 2022. ; https://doi.org/10.1101 https://doi.org/10. /2022 

SARS-CoV-2 infection-340 induced and vaccine-induced immunity

COVID-19 Treatment Guidelines Panel. Coronavirus Disease

Bamlanivimab plus Etesevimab in Mild or 347

Moderate Covid-19

Effect of Sotrovimab on Hospitalization or 349

Death Among High-risk Patients With Mild to Moderate COVID-19: A Randomized 350

Antibody Cocktail, in Outpatients with Covid-19

Helpful Lessons and Cautionary 354 Tales: How Should COVID-19 Drug Development and Access Inform Approaches to 355

Monoclonal Antibodies for Preventing Hospitalization and Mortality in COVID-19 358 9

Optimizing the Trade-off Between Learning and Doing in a Pandemic

Colorado Department of Public Health and Environment. Treatments for Covid-19

An Introduction to Propensity Score Methods for Reducing the Effects of 371

Confounding in Observational Studies

Matching as Nonparametric 373

Preprocessing for Reducing Model Dependence in Parametric Causal Inference. Political 374

Balance diagnostics for comparing the distribution of baseline covariates 376 between treatment groups in propensity-score matched samples

CoV-2 infection

logistf: Firth's Bias-Reduced Logistic Regression

A solution to the problem of separation in logistic regression