FormalPara Key Points

Real-world evidence obtained through routine clinical practice is a key source of data on the clinical effectiveness of COVID-19 treatments such as sotrovimab and supplements efficacy data from clinical trials.

Significant differences in demographic and clinical characteristics were observed among individuals with COVID-19 receiving sotrovimab and those not treated, which highlights the need to control for these differences in statistical analyses.

Sotrovimab, as compared with individuals not treated, was effective in preventing hospitalisation and death during the study period that included Delta, Omicron BA.1 and early Omicron BA.2 variants.

1 Introduction

Coronavirus disease 2019 (COVID-19) results from infection with severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). The rapid and widespread transmission of this virus around the world led the World Health Organization (WHO) to declare a pandemic in March 2020 [1]. Globally, as of August 2023, there have been approximately 770 million reported cases of COVID-19, including nearly 7 million deaths [2]. In the USA, just over 1.1 million deaths have been attributed to COVID-19 to date [2].

Critical disease can present as respiratory failure, acute respiratory distress syndrome, sepsis, shock, thromboembolism and multiorgan failure [3, 4]. Despite vaccination, certain groups of individuals, including older adults and those with immunosuppression, cancer, obesity, diabetes or chronic kidney, lung, liver or cardiovascular disease, continue to be at higher risk of developing severe COVID-19, which may lead to hospitalisation or death [5,6,7,8,9,10]. As of May 2023, more than 80% of the US population had received at least one vaccine dose (91% of those aged 12 years and over) and 70% had completed a primary vaccination series (78% of those aged 12 years and over), but only 17% (19% of those aged 12 years and over) had received an updated booster dose [11]. This trend to lower vaccination rates, and the rapidly evolving variant landscape, potentially exposes more people to future infection with SARS-CoV-2.

The emergence of the COVID-19 pandemic led to the rapid introduction of several treatments in the USA, which were granted Emergency Use Authorization (EUA) by the US Food and Drug Administration (FDA). These included antiviral agents such as remdesivir, nirmatrelvir/ritonavir and molnupiravir, and several monoclonal antibodies (mAbs), including sotrovimab, bamlanivimab, bamlanivimab–etesevimab, casirivimab–imdevimab and bebtelovimab [12]. These EUAs were based on data from randomised controlled trials conducted early in the pandemic [13,14,15,16]. At the time, these mAbs were effective early treatments for COVID-19; however, the number of cases far exceeded the capacity to treat patients, given levels of mAb production and availability. In the USA, mAbs were federally procured by the Department of Health and Human Services’ Assistant Secretary for Preparedness and Response (ASPR) and distributed to individual states. Initially (November 2020–February 2021), the distribution model was based on the number of mAb doses available (from all manufacturers) and the number of COVID-19 cases and hospitalisations in state [17]. From February 2021, the distribution model was revised to allow hospitals and other healthcare sites to order mAbs directly from the supplier. Subsequently (in September 2021), following a rapid rise in COVID-19 cases (and associated increase in mAb utilisation) due to the Delta variant, the ASPR reverted to the original distribution model (based on COVID-19 cases and hospitalisations) [17]. Of note, EUAs for all of the mAbs have subsequently been suspended or revoked, based on in vitro studies reporting reduced neutralisation activity against new SARS-CoV-2 variants [18].

Sotrovimab is an engineered, dual-action human immunoglobulin G1κ mAb derived from the parental mAb S309, a potent neutralising mAb directed against the spike protein of SARS-CoV-2 [19,20,21,22]. In a randomised clinical trial (COMET-ICE; NCT04545060) conducted during the period of the pandemic dominated by the original Wuhan or ‘wild-type’ variant, a single intravenous infusion of sotrovimab 500 mg significantly reduced the risk of all-cause > 24-h hospitalisation or death compared with placebo in high-risk patients with mild-to-moderate COVID-19 [13, 23]. Sotrovimab has been granted marketing authorisation in the European Union, Norway, Iceland, Australia (conditional), Great Britain (conditional), Saudi Arabia (conditional) and Switzerland (conditional). In Japan, a Special Approval in Emergency has been granted. In addition, sotrovimab currently has a temporary/emergency authorisation in Bahrain, Canada and the United Arab Emirates. An EUA was issued by the US FDA in May 2021. However, the sotrovimab EUA was deauthorised in all US regions on 5 April 2022 owing to decreased in vitro neutralisation of sotrovimab against circulating Omicron BA.2 SARS-CoV-2 variants relative to wild-type SARS-CoV-2 [24, 25].

It remains uncertain how well in vitro data reflect the clinical effectiveness of dual-action mAbs such as sotrovimab [26], particularly given its ability to mediate potent Fc-effector functions, which include antibody-dependent cellular cytotoxicity and antibody-dependent cellular phagocytosis [22, 27,28,29]. Given the ever-evolving variant landscape, real-world evidence obtained through routine clinical practice has become a key source of data on the clinical effectiveness of sotrovimab [30,31,32,33,34,35,36,37]. In this study, we used electronic health record (EHR) data from the US National COVID Cohort Collaborative (N3C) to assess the clinical effectiveness of sotrovimab versus no treatment in reducing the risk of progression to severe COVID-19 in a high-risk population with mild-to-moderate COVID-19.

2 Methods

2.1 Study Design and Data Source

In this retrospective cohort study, EHRs from the N3C (limited dataset) were used to identify non-hospitalised patients diagnosed with COVID-19. In the limited dataset, the original data were stripped of 17 direct identifiers as listed within the Health Insurance Portability and Accountability Act Privacy Rule, although dates were preserved to allow identification of SARS-CoV-2 variant periods. N3C is a centralised, secure data resource created by the National Institutes of Health following the emergence of SARS-CoV-2 that aggregates EHR data from across more than 65 clinical organisations in the USA [38], including the Clinical and Translational Science Awards (CTSA) Program hubs (60 institutions), the National Center for Advancing Translational Science (NCATS), the Center for Data to Health (CD2H) and the community [39, 40].

2.2 Study Population

Non-hospitalised patients were eligible for inclusion if they were aged ≥ 12 years, had a confirmed COVID-19 diagnosis (positive polymerase chain reaction or antigen test or ICD-10: U07.1 between 1 June 2021 and 30 April 2022), fulfilled at least one of the EUA criteria for sotrovimab associated with high risk of progression of COVID-19 to severe illness (Table S1) and had sufficient data to determine vital status at Day 29 (deemed as either continuous follow-up through Day 29 or death occurring in the acute period). As the earliest record of sotrovimab administration occurred on 27 September 2021, the study period was truncated to begin on this date during analysis.

Patients were excluded if they had been: previously administered COVID-19 treatments or prophylaxis, including a mAb (e.g. bamlanivimab, bamlanivimab–etesevimab, casirivimab–imdevimab or bebtelovimab), an antiviral (nirmatrelvir/ritonavir, molnupiravir or remdesivir) or tixagevimab/cilgavimab within 12 months prior to the index date; administered a mAb, an antiviral or tixagevimab/cilgavimab in a non-hospitalised setting [i.e., in an outpatient or emergency room (ER) setting] during the acute observation period; administered sotrovimab > 10 days after COVID-19 diagnosis date; or hospitalised or receiving critical care [intubation or mechanical ventilation, extracorporeal membrane oxygenation (ECMO), critical care services or intensive care services] in the 14 days prior to and including the index date.

Included patients were assigned to either the sotrovimab or untreated cohort. For the sotrovimab cohort, patients had received sotrovimab in an outpatient or ER setting within 10 days of COVID-19 diagnosis; the index date was defined as the date of first infusion of sotrovimab. Patients in the untreated cohort (no evidence of sotrovimab, any other mAb or antiviral treatment for COVID-19 in an outpatient or ER setting during the acute observation period, or of pre-exposure prophylaxis against SARS-CoV-2) were assigned an imputed index date based on the time distribution between diagnosis and sotrovimab treatment in the sotrovimab cohort as well as calendar month of COVID-19 diagnosis. In the untreated cohort, patients could have received sotrovimab or any other mAb or antiviral treatment after inpatient hospitalisation.

The baseline period was defined as the 12 months prior to the index date for assessment of most demographic and clinical characteristics. However, all available patient data from the start of the observation period were used to assess vaccination status; the vaccine assessment period ended 14 days prior to the index date (as immunity is not expected to develop until 14 days after vaccination). The primary and secondary study outcomes were evaluated during the acute period, defined as the 29-day period starting from the index date. End of follow-up was defined as either death or the acute period end date (Day 29), whichever came first.

Patients were considered fully vaccinated if they had received two or more vaccinations with an mRNA vaccine [Pfizer-BioNTech (BNT162b2) or Moderna (mRNA-1273)] or a single dose of viral vector [Johnson & Johnson [(NJ-784336725)] during the assessment period. Patients were considered partially vaccinated if they had received one vaccination with an mRNA vaccine [Pfizer-BioNTech (BNT162b2) or Moderna (mRNA-1273)] during the assessment period.

2.3 Study Outcomes

The primary endpoint was all-cause hospitalisation or death within 29 days of the index date in the sotrovimab and untreated cohorts. All-cause hospitalisations were defined from N3C's algorithm mapping visits to macrovisits [38]. Specifically, hospital visits were built from recorded Observed Medical Outcomes Partnerships (OMOP) visits by first filtering to inpatient (OMOP code 9201), inpatient hospital (8717), intensive care (32037), ER and inpatient (262), inpatient critical care facility (581379) visits of any duration, ER visits (9203) spanning at least 2 calendar days, or outpatient visits (9202) spanning exactly 2 calendar days. These visits were then merged such that any visits with overlapping calendar days would end up in the same hospital stay. Finally, all merged visits that did not contain at least one inpatient or ER visit were unmerged. This process results in combined hospital stays separated by a period of at least 1 calendar day; visits of any type that occurred during a combined hospital stay were added to the hospital stay.

Secondary endpoints included the proportion of patients hospitalised due to COVID-19, the proportion of patients receiving critical care due to any cause (e.g. intubation or mechanical ventilation, ECMO, critical care services, intensive care services), the proportion of patients admitted to an intensive care unit due to any cause, and the proportion of patients presenting at the ER for all- or COVID-19-related causes, all during the 29-day acute period. OMOP codes for critical care were identified using templates from the N3C Knowledge Store, a resource in the N3C Data Enclave created and validated by N3C domain experts [41, 42].

Subgroup analyses included rates of all-cause hospitalisation and/or death stratified by: age at index (12–54, 55–64, and ≥ 65 years), vaccination status (fully vaccinated versus unknown) and periods of SARS-CoV-2 variant predominance (Delta, 1 September 2021–30 November 2021; Omicron BA.1, 1 December 2021–28 February 2022; and Omicron BA.2, 1 March 2022–30 April 2022). Given the lack of patient-level variant genotyping data, a proxy for viral variant based on the most prevalent strains in the USA during the time of each patient’s COVID-19 diagnosis was used, as determined by the Centers for Disease Control and Prevention COVID Data Tracker [43]. Subgroup analyses were also conducted for patients at high risk of progression to severe COVID-19 (specifically in patients with the following during the baseline period: Hodgkin’s lymphoma, non-Hodgkin’s lymphoma, leukaemia, solid cancer, human immunodeficiency virus, autoimmune disease, solid organ transplant and/or allogenic stem cell transplant or receipt of immunosuppressive treatment), for individual EUA criteria (Table S1) and for patients meeting a certain number of EUA criteria (i.e. 1, 2, 3, 4, and ≥ 5).

2.4 Data Analysis

The primary endpoint of all-cause hospitalisation and/or death was reported as descriptive rate, and unweighted and weighted [via inverse probability of treatment weighting (IPTW)] odds ratio (OR) and 95% confidence interval (CI).

For descriptive analyses, continuous variables (e.g. age) were summarised using mean, standard deviation (SD) and interquartile range. Categorical variables (e.g. sex) were described using frequencies and percentages. p-Values for descriptive analyses were calculated from Pearson chi-square tests.

IPTW based on propensity scores (PS) was used to adjust for imbalances in the distribution of selected baseline characteristics between the sotrovimab-treated and untreated cohorts (Table S2). The baseline covariates considered in the PS model were defined a priori and were determined on the basis of the 12-month pre-index baseline period (except for vaccination status; see Sect. 2.2). These included age, sex, ethnicity, US state of patients’ primary care provider, antibody testing, vaccination status, components of the EUA criteria for sotrovimab, Quan–Charlson Comorbidity Index, number of healthcare visits, number of inpatient visits and number of healthcare visits requiring critical care. Differences in baseline characteristics between the sotrovimab-treated and untreated cohorts were assessed using standardised differences (a standardised difference of > 10% was considered to indicate a meaningful imbalance by convention). The IPTW-weighted logistic regression model was further adjusted for baseline covariates with a standardised difference of > 10% after weighting, which is a doubly robust approach [44]. Secondary endpoints were reported as descriptive data only.

3 Results

3.1 Patient Demographics and Baseline Characteristics

Of nearly 2.9 million patients diagnosed with COVID-19 in the N3C dataset between 27 September 2021 and 30 April 2022, 4992 met the inclusion criteria for the sotrovimab cohort, and 541,325 were included in the untreated cohort (Fig. 1). Over half of patients (3000/4992, 60.1%) in the sotrovimab cohort initiated sotrovimab on the day of (Day 0) or the day after (Day 1) COVID-19 diagnosis.

Fig. 1
figure 1

Patient selection. (1) Patients were identified from the National Center for Advancing Translational Science (NCATS) National COVID Cohort Collaborative (N3C), a centralised, secure clinical data resource created by the National Institutes of Health. (2) Positive COVID-19 diagnoses were determined from laboratory and condition data by using the phenotype algorithm identified by N3C: https://github.com/National-COVID-Cohort-Collaborative/Phenotype_Data_Acquisition/wiki/Latest-Phenotype. (3) Patient's first confirmed diagnosis during the study period is included. (4) For patients who received sotrovimab within 10 days from the date of a positive COVID-19 diagnosis, the first sotrovimab infusion was assigned as the index date. To avoid immortal time bias, index dates were imputed for patients who did not receive sotrovimab (i.e. untreated controls). Index dates were imputed using the distribution for number of days from a positive COVID diagnosis to the infusion date of sotrovimab among sotrovimab users. Since the first instance of sotrovimab infusion in this dataset was on 27 September 2021, prior to 30 September 2021, index dates for the untreated control cohort were imputed using the overall distribution of index dates in sotrovimab cohort patients during the study period. For the time period of 30 September 2021 onwards, the untreated controls were matched to sotrovimab patients on the basis of the calendar month of diagnosis when imputing the index date. These index dates were generated at this step since they were needed for the application of the remaining inclusion/exclusion criteria. (5) Patients were further excluded if they did not have sufficient birth information or were older than 100 years at their index date. (6) Patients fulfilling any one or more of the Emergency Use Authorization criteria, at index date, were classified as being at high risk of developing severe COVID-19. (7) Vital status is deemed as either continuous follow-up through Day 29 or death occurring in the acute observation period. COVID-19 coronavirus disease 2019, ER emergency room, mAb monoclonal antibody, RT-PCR reverse transcription-polymerase chain reaction

Full size image

Table 1 presents the baseline demographic and clinical characteristics of sotrovimab-treated and untreated patients. Prior to weighting, absolute standard differences of > 10% were observed for most variables, indicating imbalance between the sotrovimab and untreated cohorts. After weighting, these imbalances were addressed for most characteristics with the exception of Black or African American race, Quan–Charlson Comorbidity Index, pregnancy and immunosuppressive disease.

Table 1 Baseline demographic and clinical characteristics prior to and after weighting
Full size table

3.2 Primary Endpoints

The overall observed proportions of 29-day all-cause hospitalisation or death were 3.5% (176/4992) and 4.5% (24,163/541,325) in the sotrovimab and untreated cohorts, respectively (Table 2). The difference in proportions between the cohorts was statistically significant (p = 0.002). In adjusted analysis, sotrovimab was associated with a 25% reduction in the odds of all-cause hospitalisation or death compared with the untreated cohort (IPTW-adjusted OR: 0.75; 95% CI: 0.65, 0.87; p < 0.001; Table 3).

Table 2 Descriptive analysis of primary and secondary endpoints among the sotrovimab and untreated cohorts at Day 29
Full size table
Table 3 Comparative analysis of all-cause hospitalisation or death and hospitalisation alone among the sotrovimab and untreated cohorts at Day 29
Full size table

The overall observed all-cause hospitalisation-only rates in the sotrovimab and untreated cohorts were 3.3% (167/4992) and 4.2% (22,572/541,325), respectively (Table 2). Again, the difference between the cohorts was statistically significant (p = 0.004). The unadjusted OR was 0.80 (95% CI: 0.68, 0.93; p = 0.004). Following adjustment, sotrovimab was associated with a 22% reduction in the odds of all-cause hospitalisation compared with the untreated cohort (IPTW-adjusted OR: 0.78; 95% CI: 0.67, 0.90; p = 0.001).

The overall (unadjusted) all-cause death rate was 0.3% (< 20/4992) in the sotrovimab cohort and 0.5% (2811/541,325) in the untreated cohort (Table 2). The difference between the cohorts was statistically significant (p = 0.041).

3.3 Secondary Endpoints

The overall 29-day COVID-19-related hospitalisation rates were 1.5% (75/4992) and 1.6% (8811/541,325) in the sotrovimab and untreated cohorts, respectively (p = 0.522; Table 2). The rate of COVID-19-related ER visits was 2.0% (102/4992) in the sotrovimab cohort and 2.2% (12,051/541,325) in the untreated cohort (p = 0.410).

All-cause ER visit rates were 4.4% (220/4992) and 6.1% (33,208/541,325) in the sotrovimab and untreated cohorts, respectively (p < 0.001; Table 2). Other unadjusted all-cause clinical outcomes were similar across cohorts (Table 2).

3.4 Subgroup Analyses

3.4.1 Age

Among patients aged ≥ 65 years, the all-cause hospitalisation or death rate was lower in the sotrovimab cohort [2.8% (67/2374)] than in the untreated cohort [6.0% (11,105/185,181); p < 0.001; Table 2]. No significant differences were observed in patients aged 55–64 years [4.0% (33/830) versus 3.9% (3498/89,986); p = 0.967], nor in patients aged 12–54 years [4.3% (76/1788) versus 3.6% (9560/266,258); p = 0.154]. A similar pattern was observed across age groups for all-cause hospitalisation only (Table 2).

Among patients aged ≥ 65 years, those in the sotrovimab cohort were statistically significantly less likely to have an event of all-cause hospitalisation or death compared with those in the untreated cohort (OR: 0.46; 95% CI: 0.36, 0.58; p < 0.001; Fig. 2).

Fig. 2
figure 2

Forest plot summarising the overall likelihood of all-cause hospitalisation or mortality among sotrovimab versus untreated controls (unadjusted subgroup analyses). CI confidence interval, OR odds ratio. aHad received two or more vaccinations with an mRNA vaccine [Pfizer-BioNTech (BNT162b2) or Moderna (mRNA-1273)] or a single dose of viral vector [Johnson & Johnson (JNJ-784336725)] during the assessment period. bPatients who had at least one of the following during the baseline period: Hodgkin's lymphoma, non-Hodgkin lymphoma, leukaemia, solid cancer, human immunodeficiency virus, autoimmune disease, solid organ transplant, allogenic stem cell transplant and/or immunosuppressive treatment (two or more drug exposure events for systemic corticosteroid therapy in the past 365 days, or one or more systemic non-corticosteroid immunosuppressant drug exposures in the past 365 days). *Denotes statistical significance (p < 0.05)

Full size image

3.4.2 Vaccination Status

In patients with unknown vaccination status, the all-cause hospitalisation or death rate [3.7% (92/2477) versus 5.2% (17,193/329,005); p < 0.001] and all-cause hospitalisation rate [3.5% (86/2477) versus 4.8% (15,952/329,005); p = 0.002] were statistically significantly lower in the sotrovimab cohort versus the untreated cohort (Table 2). In a post hoc analysis, no differences across cohorts were observed in the descriptive or comparative analyses in patients who were fully vaccinated (Tables 2 and 4).

Table 4 Comparative analysis of all-cause hospitalisation or death and hospitalisation alone among the sotrovimab and untreated cohorts at Day 29 in the subgroup of fully vaccinated patients
Full size table

Among patients with unknown vaccination status, those in the sotrovimab cohort were statistically significantly less likely to have an event of all-cause hospitalisation or death compared with those in the untreated cohort (OR: 0.70; 95% CI: 0.57, 0.86; Fig. 2).

3.4.3 Periods of Variant Predominance

The all-cause hospitalisation or death rate was lower in the sotrovimab cohort versus the untreated cohort across both the Delta [3.6% (43/1188) versus 5.1% (11,530/225,283); p = 0.023] and the Omicron BA.2 [1.0% (6/576) versus 4.9% (1730/35,005); p < 0.001] periods of predominance; however, no difference was observed in the Omicron BA.1 period of predominance [sotrovimab cohort, 3.9% (127/3228); untreated cohort, 3.9% (10,903/281,037); p = 0.909; Table 2]. A similar pattern was observed across the variant periods of predominance for all-cause hospitalisation only.

Sotrovimab-treated patients were significantly less likely to have an event of all-cause hospitalisation or death compared with untreated patients in both the Delta (OR: 0.70; 95% CI: 0.51, 0.94; p = 0.02) and Omicron BA.2 (OR: 0.20; 95% CI: 0.09, 0.45; p < 0.001) periods of predominance (Fig. 2).

3.4.4 High-risk Subgroup

The all-cause hospitalisation or death rate was lower in the sotrovimab cohort versus the untreated cohort [4.1% (86/2093) versus 5.1% (6066/118,948); p = 0.046]; however, no difference was observed for all-cause hospitalisations only (Table 2). Patients in the sotrovimab cohort were significantly less likely than those in the untreated cohort to have an event of all-cause hospitalisation or death (OR: 0.80; 95% CI: 0.64, 0.99; p = 0.04; Fig. 2).

3.4.5 EUA Criteria

Figure 3 shows the overall likelihood of all-cause hospitalisation or mortality among those receiving sotrovimab versus untreated controls for each EUA criteria. The OR favoured sotrovimab for most categories, with statistical significance reached for age ≥ 65 years, cardiovascular disease or hypertension, history of chronic kidney disease (CKD; any stage), history of CKD stage ≥ 3, history of type 2 diabetes, immunosuppressive disease, heart failure, acquired heart disease, two EUA criteria met, four EUA criteria met and five or more EUA criteria met. The OR favoured the untreated control cohort for immunosuppressive treatment and neurodevelopmental disease, with statistical significance reached for the former.

Fig. 3
figure 3

Forest plot summarising the overall likelihood of all-cause hospitalisation or mortality (unadjusted) among sotrovimab versus untreated controls (EUA criteria). CI confidence interval, CKD, chronic kidney disease, COPD chronic obstructive pulmonary disease, EUA Emergency Use Authorization, ICS inhaled corticosteroid, OR odds ratio. *Denotes statistical significance (p < 0.05)

Full size image

4 Discussion

It is now more than 3 years since the WHO declared the spread of the SARS-CoV-2 virus as a global pandemic. In the USA, the federal Public Health Emergency declared in response to COVID-19 ended on 11 May 2023. The global response to the pandemic witnessed the rapid development and introduction of novel vaccines against the SARS-CoV-2 virus as well as treatments for COVID-19, all in the face of an evolving landscape as new variants of concern emerged. Most of the randomised clinical trials to demonstrate the safety and efficacy of mAbs in COVID-19 were conducted early in the pandemic [13,14,15,16], raising questions about their effectiveness against subsequent variants. In lieu of data from randomised clinical trials, in vitro neutralisation testing was used to infer changes in effectiveness. This ultimately led to suspension or revocation of mAb EUAs due to reported reduced in vitro activity against new variants compared with original ‘wild-type’ virus [18]; however, the link between in vitro neutralisation and clinical outcomes is uncertain. In this context, real-world studies are crucial for assessing the ongoing effectiveness of vaccines and treatments. This study was conducted to describe the real-world effectiveness of sotrovimab in US-based patients with mild-to-moderate COVID-19 at high risk for progression to severe COVID-19. A particular strength of the study is the large sample size. Our results indicate that, during the study period (27 September 2021–30 April 2022), sotrovimab was effective in preventing all-cause hospitalisation or death compared with no treatment.

Our data also demonstrate that there are distinct differences in the baseline characteristics of patients diagnosed with COVID-19 that warrant careful consideration when designing comparative effectiveness studies. Even after application of IPTW, imbalances between the cohorts remained for several baseline characteristics that confer higher risk of severe COVID-19, including Black/African American race, pregnancy and immunosuppressive disease. Hence, the overall protective effect of sotrovimab occurred despite these imbalances favouring lower risk of severe outcomes in the untreated cohort.

It is well documented that older age is an important risk factor for development of severe COVID-19 [45]. In the subgroup of patients aged ≥ 65 years of age, we observed a substantial reduction in the risk of all-cause hospitalisation or death in the sotrovimab cohort compared with the untreated cohort. No differences were observed between the sotrovimab and untreated cohorts for the other age subgroups analysed (55–64 years and 12–54 years).

It is interesting to note the relative impact of vaccination status on the observed effectiveness of sotrovimab in the current study. Among the subgroup with unknown vaccination status, patients in the sotrovimab cohort had a significantly lower risk of all-cause hospitalisation or death than those in the untreated cohort. However, in the overall cohort there was no benefit of sotrovimab when only those with confirmed fully vaccinated status were considered. One possible explanation is that vaccination reduces the risk of severe COVID-19 among people who are subsequently infected with SARS-CoV-2. This is supported by other studies that have shown hospitalisation to be rare among patients with breakthrough COVID-19 infections [46], and similar findings have been reported for other infectious diseases such as influenza [47]. An observational cohort study conducted in England after the emergence of the Omicron variant demonstrated that the clinical effectiveness of sotrovimab was maintained in fully vaccinated patients [37]. However, this study involved fewer patients than the present study and had an active comparator group (molnupiravir or nirmatrelvir/ritonavir) but no untreated control group.

We observed no benefit of sotrovimab treatment in the subgroups of patients in receipt of immunosuppressive treatment and those with neurodevelopmental disease; the OR favoured the untreated cohort in both of these subgroups, reaching statistical significance in the former. It should be noted that these were unadjusted analyses that involved a relatively small proportion of the overall study population in both cases (approximately 3%); therefore, confounding cannot be excluded. Several other studies have indicated that sotrovimab is effective in patients who are immunocompromised [48, 49], while an ongoing clinical study (LUNAR; NCT05305651) will provide additional data for sotrovimab in this patient population.

We observed significant benefit of sotrovimab treatment compared with no treatment during periods of the study when Delta and Omicron BA.2 were the predominant variants in the US, albeit based on a low number of events in the BA.2 period. The finding in relation to the BA.2 period is of particular interest, given that the sotrovimab EUA was deauthorised in all US regions due to decreased in vitro neutralisation of sotrovimab against circulating BA.2 variants relative to wild-type SARS-CoV-2 [24, 25]. There is no validated model available to consistently and reliably corelate in vitro neutralisation to predicted clinical efficacy. It should also be noted that direct neutralisation of virus is not the only antiviral mechanism expected for sotrovimab in vivo, since it has also been shown to mediate antibody-dependent cellular cytotoxicity and antibody-dependent cellular phagocytosis [29]; these effector functions are not measured in standard neutralisation assays. Our results are in agreement with those from an observational study conducted from October to December 2021 (when Delta was the predominant circulating variant) in non-hospitalised adult patients infected with SARS-CoV-2 [50]. Compared with no mAb treatment, sotrovimab was associated with 63% (OR: 0.37; 95% CI: 0.19, 0.66) lower odds of 28-days all-cause hospitalisation and 89% (OR: 0.11; 95% CI: 0.0, 0.79) lower odds of all-cause mortality, similar to our findings during the September 2021–December 2021 time period. Also, in our study, no benefit of sotrovimab was observed during the period when Omicron BA.1 was the predominant variant (1 December 2021–28 February 2022). This period immediately followed the US Thanksgiving holiday and also covered holidays associated with religious festivals such as Christmas, when travel around the country and the opportunity for large gatherings are increased. It also coincided with a shortage in supply of sotrovimab and other mAbs [51], which may have altered the patient population receiving sotrovimab.

Other studies, in the USA and elsewhere, have shown a real-world clinical benefit of sotrovimab during periods of Omicron BA.1 predominance, as well as other emerging variants. A retrospective analysis of data from the FAIR Health National Private Insurance Claims database included patients diagnosed with COVID-19 between 1 September 2021 and 30 April 2022 (a period when the prevalence of circulating variants in the USA changed between Delta and Omicron) [31]. Sotrovimab (n = 15,633) was associated with reduced risk of 30-day all-cause hospitalisation and mortality versus no mAb treatment (n = 1,514,868), and clinical effectiveness persisted during Delta and early Omicron variant waves and also among all high-risk subgroups assessed. A second study used a retrospective cohort from the US Department of Veterans Affairs healthcare system [36]. Veterans aged ≥ 18 years diagnosed with COVID-19 between 1 December 2021 and 4 April 2022 were included. During BA.1 predominance, compared with matched controls (n = 11,250), sotrovimab-treated patients (n = 2816) had a 70% lower risk of 30-day hospitalisation or mortality [hazard ratio (HR) 0.30; 95% CI: 0.23, 0.40], a 66% lower risk of 30-day hospitalisation (HR 0.34; 95% CI: 0.25, 0.46) and a 77% lower risk of 30-day all-cause mortality (HR 0.23; 95% CI: 0.14, 0.38). During BA.2 predominance, sotrovimab-treated patients had a 71% (HR 0.29; 95% CI: 0.08, 0.98) lower risk of 30-day COVID-19-related hospitalisation, emergency or urgent care [36]. Conversely, an analysis of state-wide EHR data from Colorado reported no difference between sotrovimab-treated patients (n = 1542) and matched controls (n = 3663) in 28-day hospitalisation or mortality during the period of BA.1 predominance (December 2021–March 2022) [52].

The effectiveness of sotrovimab during the Omicron BA.2 subvariant predominance period has been evaluated in a systematic literature review of published observational studies [32]. The review included data from five eligible studies and concluded that the effectiveness of sotrovimab is maintained against Omicron BA.2, either through the demonstration of low and comparable rates of severe clinical outcomes between the Omicron BA.1 and BA.2 predominance periods, or by comparison with active treatments or no treatment within the Omicron BA.2 predominance period. A more recent study reported that sotrovimab (compared with no treatment) was associated with reduced risk of adverse COVID-19 outcomes in the Omicron BA.1 predominance period, but there was weaker evidence of a benefit in the Omicron BA.2 predominance period [53].

Another study assessed the characteristics and outcomes of patients with COVID-19 at high risk of disease progression receiving sotrovimab, oral antivirals or no treatment in England [34]. Low hospitalisation rates were observed for all treated cohorts, and results were consistent among subgroups of patients with advanced renal disease, among patients aged 18–64 years and ≥ 65 years, and across periods of Omicron subvariant (BA.1, BA.2 and BA.5) predominance. Another study using data from the National Health Service in England reported low rates of COVID-19-attributable hospitalisations and deaths in patients presumed to be treated with sotrovimab [35]. Results were consistent for patients with severe renal disease and active cancer, and there was no evidence of differences in hospitalisation rates during periods when Omicron BA.1 and BA.2 or BA.5 subvariants were predominant.

Findings from the present study should be considered in light of some limitations. First, due to the retrospective observational design of the current study, caution should be exercised when interpreting results, due to the potential for bias from residual or unmeasured confounding. Certain attributes that influence the choice of treatment and endpoints of interest (such as patient preference, baseline lung function, oxygen levels, symptoms and severity, and availability of sotrovimab) were not captured in the N3C data source. There may also be a component of ‘healthy user’ bias, such that included patients treated with sotrovimab were more amenable than untreated patients to other interventions that improved their overall health and/or COVID-19 symptoms. We were also unable to account for potential socioeconomic factors that might influence ability to receive treatment with sotrovimab. Such confounding may influence study results in favour of the sotrovimab (e.g. healthy user bias) or the untreated (e.g. severity of symptoms) cohort. In addition, the retrospective nature of the study meant that the untreated cohort had to be defined based on patients who did not receive treatment with other mAbs or antivirals in an outpatient setting during the acute observation period; this may have led to selection bias, as the reason for not receiving treatment was not available in the data. This risk of selection bias was mitigated by including patients who received other treatments for COVID-19 after hospitalisation. Another potential limitation is the absence of variant sequencing data, which precludes any definitive conclusions on the effectiveness of sotrovimab against the different variants in circulation during the study period. Also, the rapid evolution of SARS-CoV-2 variants, and the varying protocols of sotrovimab distribution, created challenges in identifying an untreated cohort that was contemporaneous with the sotrovimab cohort as well as being demographically and clinically similar, although IPTW helped to create pseudo-populations that were similar in terms of observed baseline characteristics. Finally, all-cause mortality, one of the study outcomes, may not be specifically coded in the EHRs and therefore under-represented in the data.

The World Health Organization announced in May 2023 that COVID-19 no longer constituted a Public Health Emergency of International Concern, but acknowledged remaining uncertainties as a result of the constantly evolving SARS-CoV-2 variant landscape [54]. In this context, generating efficacy and safety data for COVID-19 therapies through randomised, controlled trials (RCTs) has proven extremely challenging; hence, data from other sources will remain important for assessing their benefit–risk ratio. In vitro data can be informative, although there is currently no validated model that can consistently and reliably correlate in vitro neutralisation activity of mAbs to clinical outcomes. Notwithstanding the limitations outlined above, real-word data will continue to provide insights on the effectiveness of COVID-19 therapies, and help to contextualise in vitro data where the link to clinical outcomes is unclear. Given the challenges around conducting RCTs, it may also be necessary to consider new regulatory pathways to evaluate future COVID-19 therapies.

5 Conclusions

The findings of this study demonstrate that there are distinct differences in baseline characteristics of patients diagnosed with COVID-19 that warrant careful consideration when designing comparative effectiveness studies. In addition, sotrovimab demonstrated (in descriptive and adjusted analyses) clinical effectiveness in preventing acute-phase all-cause hospitalisation and/or mortality for the duration of the study period (September 2021–April 2022).