Emulation of a Target Trial From Observational Data to Compare Effectiveness of Casirivimab/Imdevimab and Bamlanivimab/Etesevimab for Early Treatment of Non-Hospitalized Patients With COVID-19

Abstract

Objectives: Comparative analysis between different monoclonal antibodies (mAbs) against SARS-CoV-2 are lacking. We present an emulation trial from observational data to compare effectiveness of Bamlanivimab/Etesevimab (BAM/ETE) and Casirivimab/Imdevimab (CAS/IMD) in outpatients with early mild-to-moderate COVID-19 in a real-world scenario of variants of concern (VoCs) from Alpha to Delta.

Methods: Allocation to treatment was subject to mAbs availability, and the measured factors were not used to determine which combination to use. Patients were followed through day 30. Viral load was measured by cycle threshold (CT) on D1 (baseline) and D7.Primary outcome was time to COVID-19-related hospitalization or death from any cause over days 0-30. Weighted pooled logistic regression and marginal structural Cox model by inverse probability weights were used to compare BAM/ETE vs. CAS/IMD. ANCOVA was used to compare mean D7 CT values by intervention. Models were adjusted for calendar month, MASS score and VoCs. We evaluated effect measure modification by VoCs, vaccination, D1 CT levels and enrolment period.

Results: COVID19-related hospitalization or death from any cause occurred in 15 of 237 patients in the BAM/ETE group (6.3%) and in 4 of 196 patients in the CAS/IMD group (2.0%) (relative risk reduction [1 minus the relative risk] 72%; p=0.024). Subset analysis carried no evidence that the effect of the intervention was different across stratification factors. There was no evidence in viral load reduction from baseline through day 7 across the two groups (+0.17, 95% -1.41;+1.74, p=0.83). Among patients who experienced primary outcome, none showed a negative RT-PCR test in nasopharyngeal swab (p=0.009) and 82.4% showed still high viral load (p<0.001) on D7.

Conclusions: In a pre-Omicron epidemiologic scenario, CAS/IMD reduced risk of clinical progression of COVID-19 compared to BAM/ETE. This effect was not associated with a concomitant difference in virological response.

Keywords: COVID-19; SARS-COV-2; bamlanivimab/etesevimab; casirivimab/imdevimab; early treatment for COVID-19; monoclonal antibodies.

Figure 1

Assumptions regarding the underlying causal link between measured factors. According to our assumptions, month of infusion and type of VoC were identified as main time-fixed confounders of our comparison of interest. MASS score was an important predictor and used in some of the models to increase efficiency. Serology test was performed in all participants on the day of the infusion. As a consequence, the result of the test was not known to the prescriber when the intervention was initiated. Consequently, serology cannot be a confounder for the association of interest.

Figure 2

Study flow chart.

Figure 3

Unweighted (A) and Weighted (B) Kaplan Meier curves of time to primary endpoint by intervention group -weights include month of enrolment and type of VoC.

Figure 4

Forest plot of the effect of the intervention by subsets - Primary endpoint. *p-value corresponds to the test for interaction between intervention and each subgroup unadjusted for multiplicity. **aHR1– (vaccination, serology and CT strata) adjusted for calendar month and VoCs; **aHR2 – (VoCs stratum) adjusted for MASS score because of a positivity issue leading to abnormally high weights in the stratum of participants infected with the Delta VoC; **aHR3 – (enrolment period stratum) adjusted for MASS score and VoCs. PDFU, person days of follow-up.

Figure 5

SARS-CoV-2 RNA levels at D1 and D7 in patients treated with CAS/IMD and BAM/ETE. (A) Box-and-whiskers plot showing the comparison of viral loads detected at D1 and D7 in patients treated with BAM/ETE (n=170 and 154, respectively) or CAS/IMD (n=183 and 174, respectively), and the variation of RNA levels observed between the two time-points by intervention (BAM/ETE, n=145; CAS/IMD, n=165). Viral RNA levels are expressed as CT of Orf1ab gene amplification. Median CT values and IQR are shown. Statistical analysis of the comparisons between treatment groups was performed by Mann–Whitney test: D1, p=0.29; D7, p=0.87; and Variation, p=0.71. CT values at D1 and D7 within each group were compared using paired Wilcoxon sign-rank test, ****p<0.0001. (B, C) Spaghetti-plot showing the Orf1ab CT values measured at D1 and D7 in each patient treated with BAM/ETE (n= 145, B) or CAS/IMD (n=165, C). Horizontal dashed line represents the limit of detection (CT: 40.0), values ≥40 are considered negative.