Efficacy of COVID-19 Oral antivirals in hospitalised oldest-old with high morbidity burden: a target trial emulation study

Abstract

Background

Molnupiravir and nirmatrelvir-ritonavir are orally administered pharmacotherapies for mild to moderate COVID-19. However, the effectiveness of these drugs among very old (≥80 years), hospitalised patients remains unclear, limiting the risk–benefit assessment of these antivirals in this specific group. This study investigates the effectiveness of these antivirals in reducing mortality among this group of hospitalised patients with COVID-19.

Methods

Using a territory-wide public healthcare database in Hong Kong, a target trial emulation study was conducted with data from 13 642 eligible participants for the molnupiravir trial and 9553 for the nirmatrelvir-ritonavir trial. The primary outcome was all-cause mortality. Immortal time and confounding bias was minimised using cloning-censoring-weighting approach. Mortality odds ratios were estimated by pooled logistic regression after adjusting confounding biases by stabilised inverse probability weights.

Results

Both molnupiravir (HR: 0.895, 95% CI: 0.826–0.970) and nirmatrelvir-ritonavir (HR: 0.804, 95% CI: 0.678–0.955) demonstrated moderate mortality risk reduction among oldest-old hospitalised patients. No significant interaction was observed between oral antiviral treatment and vaccination status. The 28-day risk of mortality was lower in initiators than non-initiators for both molnupiravir (risk difference: −1.09%, 95% CI: −2.29, 0.11) and nirmatrelvir-ritonavir (risk difference: −1.71%, 95% CI: −3.30, −0.16) trials. The effectiveness of these medications was observed regardless of the patients’ prior vaccination status.

Conclusions

Molnupiravir and nirmatrelvir-ritonavir are moderately effective in reducing mortality risk among hospitalised oldest-old patients with COVID-19, regardless of their vaccination status.

Key points

• This study provides valuable data on the real-world efficacy of molnupiravir and nirmatrelvir-ritonavir among the oldest-old.

• There is good effectiveness in mortality risk reduction following the use of molnupiravir and nirmatrelvir-ritonavir.

• COVID-19 antiviral drugs for hospitalised old people should be recommended regardless of vaccination status.

Introduction

Since May 2023, the COVID-19 pandemic is no longer regarded as a public health emergency [1]. Nevertheless, there is an ongoing circulation of SARS-CoV-2 in the communities [2] that will plausibly continue to cause disease. Molnupiravir and nirmatrelvir-ritonavir were authorised for emergency use by the Food and Drug Administration (United States) in December 2021 as treatment options for the management of mild to moderate COVID-19 disease which may progress to severe disease [3]. The use of these medications are supported by robust evidence of tolerable safety and good efficacy profiles from clinical trials [4, 5]. For instance, molnupiravir has been shown to be associated with 33% and 88% of risk reduction for hospitalisation and mortality respectively [4], while nirmatrelvir-ritonavir is associated with 77% and 90% such risk reduction [5].

In relation to clinical practice, the evidence from the clinical trials is, nonetheless, limited in several ways. First, the trials were conducted mainly among people with no pre-existing immunity acquired from vaccination or a previous infection [6–8]. The evaluation of the benefits in addition to the existing immunity is needed for an informed prescription decision in terms of the risk–benefit ratio. Pharmacoepidemiologic real-world evidence has indeed been generated to add to the evidence base. For instance, Bajema et al. conducted a retrospective cohort study to demonstrate an effectiveness of 77% in terms of mortality following the use of molnupiravir among 167 461 patients, who were mostly vaccinated, in the United States [9]. In another study, Schwartz et al. examined data from 177 545 patients in Ontario, Canada and showed an effectiveness of 51% for nirmatrelvir-ritonavir [10]. Second, the clinical trials were focused on non-hospitalised patients, despite the fact that hospitals remain as one of the most common settings to manage patients with COVID-19 in many countries.

Our team has previously conducted a target trial emulation study to use real-world data to assimilate the findings that could have been yielded should a randomised trial on the two medications be conducted among hospitalised patients [11]. We showed that both medications are effective in reducing mortality to a consistent extent among hospitalised patients regardless of vaccination status.

However, the current body of evidence is still limited about such effectiveness among the oldest-old, who are most likely hospitalised for COVID-19 following an infection and have a significantly elevated risk of death [12]. Notably, the pharmacokinetic profile of the oldest-old is likely different from younger groups [13]. Moreover, with increased preexisting morbidity burden among most of the oldest-old, drug–drug interaction, albeit very rare, is potentially more likely than among other demographic strata [14]. The risk–benefit ratio of the COVID-19 antivirals, especially among very old fully-vaccinated hospitalised patients, remain unclear. In this study, using a territory-wide public healthcare database in Hong Kong SAR, we aim to examine the effectiveness of molnupiravir and nirmatrelvir-ritonavir among hospitalised oldest-old, i.e. defined as people aged 80 years or older, typically with high underlying morbidity burden in terms of mortality.

Methods

Data sources

This study aimed to emulate a target trial using observational data from Hong Kong to investigate the effectiveness of COVID-19 oral antiviral medications among the oldest-old. Clinical data for this research were obtained from the electronic health record database of the Hospital Authority (HA), vaccination records from the Department of Health (DH) and records of confirmed COVID-19 cases from the Centre for Health Protection (CHP) of the Government of the Hong Kong Special Administrative Region (HKSAR). These databases were integrated using anonymised unique patient identifiers as required by the data custodians due to patient confidentiality and privacy concerns. The HA manages public inpatient and outpatient services, providing real-time clinical information across all clinics and hospitals. The DH maintains a vaccination record database, while the CHP maintains a comprehensive COVID-19 database that documents all verified COVID-19 cases, encompassing reporting of positive results from both Polymerase Chain Reaction (PCR) and Rapid Antigen Test. Previous studies have utilised these population-based databases for COVID-19 pharmacovigilance research [15–23].

Study design and eligibility criteria

Participants included patients aged 80 years or older who were hospitalised with COVID-19 (indicated by PCR test) during the inclusion period. The inclusion period started from the date when the oral antiviral drugs became available (i.e. February 26, 2022, for molnupiravir, March 16, 2022, for nirmatrelvir-ritonavir) and ended on 26 December 2022. The date of admission was used as the index date. We excluded patients with a previous history of COVID-19, received the other COVID-19 oral antiviral treatment (specifically: nirmatrelvir-ritonavir for the molnupiravir trial or molnupiravir for the nirmatrelvir-ritonavir trial) treatment during the study period, those who received either molnupiravir or nirmatrelvir-ritonavir before hospital admission, and those died on the index date. Additionally, considering that oral antivirals are not a common treatment option for patients admitted to the ICU or requiring ventilatory support, individuals who were admitted to the ICU or required ventilatory support on or before the index date were also excluded. Patients with contraindications to nirmatrelvir–ritonavir, such as severe liver impairment (cirrhosis, hepatocellular carcinoma or liver transplant), severe renal impairment (eGFR <30 mL/min/1.73 m2, dialysis or kidney transplant) and those using interacting drugs (such as amiodarone, apalutamide, rifampicin, rifapentine, carbamazepine, primidone, phenobarbital or phenytoin) [3] within 90 days prior to the index date, were also excluded to minimise confounding by indication. Patients were followed from the index date until the earliest of outcome occurrence, i.e. death, 28 days after the index date or the administrative end of the study.

Treatment strategies

We aimed to compare two treatment strategies for COVID-19 patients. The first strategy involved initiating either molnupiravir or nirmatrelvir-ritonavir within 5 days of the index date. Separate emulated trials were conducted for each drug, and patients assigned to this strategy were expected to complete a full 5-day course of treatment. Concomitant treatments, such as corticosteroids, were allowed and adjusted for as post-assignment confounders. The second strategy involved no COVID-19 oral antiviral (molnupiravir or nirmatrelvir-ritonavir) treatment during follow-up. It is worth noting that in Hong Kong, all public hospitals are managed by the HA, which sets territory-wide guidelines for physicians’ use across hospitals in different districts. Therefore, both drugs are equally accessible across all public hospitals.

Outcome

The primary outcome was all-cause mortality. Death information was linked from the Hong Kong Deaths Registry, an official government registry that documents all registered deaths in Hong Kong.

Statistical analysis

We emulated a target trial of COVID-19 oral antiviral treatment initiation using observational data, estimating the per-protocol average treatment effect. The statistical analysis plan is based on a previous published large-scale study [11]. To minimise potential biases, the cloning, censoring, and weighting approach was adopted due to the five-day grace period for initiating treatment, which is commonly used interval in previous research to allow slightly delayed treatment initiation [5]. An intermediate dataset with two copies of each eligible individual was created at the index date, with each clone assigned to one of the treatment strategies. Clones assigned to the ‘oral antiviral’ strategy were censored on day four if they had not initiated either oral antiviral treatment by then. Clones assigned to the ‘no oral antiviral’ strategy were censored when they initiated oral antiviral treatment during follow-up. Stabilised inverse probability weights were estimated to account for potential bias introduced with artificial censoring due to baseline and post-assignment confounders. Charlson Comorbidity Index (CCI) and concomitant treatments received (remdesivir, tocilizumab, baricitinib, interferon beta-1b, corticosteroids) were included as post-assignment time-varying covariates in addition to other baseline covariates. The post-assignment time-varying covariates were updated daily until the end of follow-up. The full list of covariates included in the regression models and the resulting variable estimates used in the construction of inverse probability weights are shown in Supplemental Table 1.

To estimate the odds ratio of mortality, a pooled logistic regression was conducted, which was adjusted for treatment, days of follow-up, and baseline covariates, such as age, sex, CCI and number of vaccine doses received. The regression model was used to emulate randomisation of treatment assignment using observational data. The odds ratio was used to approximate the hazard ratio (HR) since the outcomes were rare at all times. The robust variance estimator was used to estimate the 95% confidence intervals (CIs). To estimate absolute risk differences and cumulative incidence of outcomes, predicted values from the pooled logistic regression were used, with interaction terms between treatment and day of follow-up included. The number needed to treat (NNT) to prevent one additional outcome was calculated based on the estimated risk differences. The 95% CIs for the absolute risk difference and cumulative incidence were obtained from a nonparametric bootstrap of 100 samples.

Prespecified subgroup analyses were conducted, which were stratified by vaccination status, sex and CCI. Interaction effects between treatment and vaccination status, sex and CCI were also tested, with P values for the interaction reported.

All statistical tests were two-sided, and a P value of less than 0.05 was considered statistically significant. R version 4.0.3 (R Foundation) was used for statistical analysis. To ensure the quality of the analysis, two investigators conducted the statistical analyses independently. The STROBE (Strengthening the Reporting of Observational Studies in Epidemiology) statement checklists were followed to ensure transparent reporting of the study.

Results

We enrolled a total of 20 751 patients who were hospitalised with COVID-19 for the molnupiravir trial and 15 913 patients for the nirmatrelvir–ritonavir trial. After excluding some participants, there were 13 642 eligible participants for the molnupiravir trial and 9553 for the nirmatrelvir–ritonavir trial. Figure 1 shows the procedures of cohort selection from the database.

The mean age of the participants in the molnupiravir trial was 88.34 [standard deviation (SD) 5.29] years, while for the nirmatrelvir–ritonavir trial it was 87.92 (SD 5.17) years. Almost half of the participants in the molnupiravir trial (47.4%) and nirmatrelvir–ritonavir trial (48.0%) were men. In the molnupiravir trial, 16.1% received one dose of COVID-19 vaccination, 17.5% received two, and 30.3% received at least three. In the nirmatrelvir–ritonavir trial, 9.0% received one, 18.8% received two, and 49.0% received at least three doses. Of note, the CCI score for participants of both trials were high, i.e. molnupiravir: 5.55 (SD: 1.67); nirmatrelvir–ritonavir: 5.22 (SD: 1.52) (Table 1).

During the follow-up period, a small percentage of participants who did not start oral antiviral treatment within 5 days eventually initiated it. Specifically, 412(3.02%) of the participants in the molnupiravir trial and 101(1.06%) of the participants in the nirmatrelvir–ritonavir trial-initiated treatment after the initial 5 days. The majority of participants, 4595 (92.36%) in the molnupiravir group and 4963 (96.09%) in the nirmatrelvir–ritonavir group, were prescribed the full treatment course of 5 days. Figure 2 shows the cumulative incidence of mortality across different arms of the two trials.

The median follow-up for both trials was 28.0 days (IQR, 0 days). In the emulated trials of molnupiravir and nirmatrelvir–ritonavir, 2956 and 1125 participants, respectively, died during follow-up. The HRs (95% CI) of mortality among initiators versus noninitiators were 0.895 (0.826–0.970) for molnupiravir and 0.804 (0.678–0.955) for nirmatrelvir–ritonavir (Table 2). The 28-day risk of mortality (95% CI) in initiators was 17.68% (16.49–18.70%); and was 18.77% (17.80–19.72%) in non-initiators (risk difference[95% CI], −1.09% [−2.29,0.11], NNT: 92) in the molnupiravir trial; and was 8.35% (7.59–9.23%) in initiators and 10.06% (9.01–11.33%) in non-initiators (risk difference[95% CI], −1.71% [−3.30,-0.16], NNT: 58) in the nirmatrelvir–ritonavir trial.

We found no significant interaction between oral antiviral treatment and vaccination status for any outcome, suggesting the effectiveness of COVID-19 oral antivirals does not vary with different vaccination status (Table 2).

Discussion

In this target trial emulation study using territory-wide public healthcare records in Hong Kong, we demonstrated that a moderate mortality risk reduction is associated with the use of molnupiravir or nirmatrelvir-ritonavir among the oldest-old people hospitalised for COVID-19. This inverse association was not significantly altered with different vaccination status prior to the hospitalisation. Likewise, sex or multimorbidity burden proxied by CCI was not found to be associated with varied strength of the observed association between the medications and reduction of mortality risks.

Our findings largely support the effectiveness of molnupiravir or nirmatrelvir-ritonavir in the prevention of mortality in oldest-old people [5]. Indeed, in comparison with previous studies using a more broadly defined population, no marked inconsistency was observed in terms of the effect size of mortality risk reduction [5]. With advanced age and possibly complex multimorbidity status, the probability of seeing drug–drug interactions or side effects are far from zero [24]. Indeed, although antivirals are generally well tolerated, they are not without risk, such as potential renal complications. The decision to use them should be based on good scientific information indicative of the obvious benefits that outweighs the risks. For our study population, the significance of investigating the effectiveness of the antivirals in reducing mortality risks is essentially twofold. First, the oldest-old are the demographic stratum with the highest risk of death from COVID-19 and, also, should represent the most frequently admitted patients in the hospital that clinicians have to look after [25, 26]. Evidence specific to this population is much needed to inform the day-to-day clinical decisions. Second, further complicating the matter is the availability of largely effective vaccines against COVID-19 which did not exist at the time the antivirals were trialled in the development process. Without further information, it is reasonable to question the additional benefits given the preexisting immunity acquired from vaccination. The same problem exists for people who already had COVID-19 and recovered. Hence, our study provides important evidence to bridge this knowledge gap. Our results confirm that regardless of vaccination status, both antiviral medications are moderately associated with a reduction of mortality risk even among hospitalised people at an advanced age.

The existing clinical and public health practice is mainly focused on prescribing the antivirals to people in the community or primary care settings, which is based on the previous strong evidence of good effectiveness against hospitalisation and mortality among people with mild COVID-19 [27]. The direct implication of the findings of this study, however, is that since both antivirals are effective against mortality risks among the oldest old people hospitalised for COVID-19, hospitals admitting patients with COVID-19 should include the medications in their respective formulary to be equipped for the management of COVID-19 admissions of people at an advanced age. This is especially valid for hospitals used to admitting very old patients in their routine practice. In the current study, the NNT to avert one death estimated for molnupiravir and nirmatrelvir-ritonavir were only 92 and 58, respectively. Although the cost may vary from country to country, recent study in Hong Kong indicated that the use of both antiviral medications was cost savings compared to standard care in inpatients settings [27]. Specifically, molnupiravir cost USD 449.57 per person and reduced 17.1% of mortality. The estimated incremental cost-effectiveness ratios were USD2629.08 per death averted. The use of nirmatrelvir-ritonavir saved USD1265.58 and reduced 23.0% of the mortality events, indicating an ICER of USD5502.53. Another implication of this study is that prior vaccination status might be of less relevance when prescribing antivirals for patients admitted for COVID-19. Even for very old people with full or booster vaccination, the antivirals will be effective in further reducing their risk of death from COVID-19. The mechanisms of risk reduction between the vaccines and antivirals are likely unrelated and independent.

There are clear key strengths to this study. First, the database used in this study covers the entire territory of Hong Kong SAR conferring very high representativeness of the findings. Second, there is a consistent and comprehensive coding system commonly used across the territory to ensure the reliable recording of events and medication practices. Third, since the setting was tertiary care, the medication adherence is likely better than studies using community-based data. Despite these strengths however, there are several limitations to the analysis that require caution. First, the data presented in this study are observational and no randomisation processes were involved. Similar with all other pharmacoepidemiologic studies, there may be unmeasured indication bias and confounders. The results should be interpreted cautiously considering the observational study design. Further studies are needed to confirm the findings of this study. Second, we were not able to retrieve clinical records from private hospitals and practices. There may be patients who used the antivirals prescribed by private clinicians such that the effectiveness observed in this study could have been underestimated. That is because people included in the non-use group had a higher risk than if private prescriptions were recorded. Lastly, the study population is predominantly ethnic Chinese and future studies of other ethnicities will be required to broaden generalisability.

In conclusion, we found a statistically significant moderate risk decrease associated with the use of molnupiravir or nirmatrelvir-ritonavir among the oldest old people who were hospitalised for COVID-19, many of whom were already fully- or booster-vaccinated. As the association is consistent across prior vaccination status, antiviral prescribing decisions could be made regardless of the vaccination status in this specific patient group.

Acknowledgements:

We gratefully acknowledge the CHP, the DH and the HA for facilitating data access.

Declaration of Conflicts of Interest:

F.T.T.L. has been supported by the RGC Postdoctoral Fellowship under the Hong Kong Research Grants Council and has received research grants from the Health Bureau of the Government of the HKSAR, outside the submitted work. C.S.L.C. has received grants from the Health Bureau of the Hong Kong Government, Hong Kong Research Grant Council, Hong Kong Innovation and Technology Commission, Pfizer, IQVIA, and Amgen; and personal fees from PrimeVigilance; outside the submitted work. X.L. has received research grants from the Health Bureau of the Government of the HKSAR; research and educational grants from Janssen and Pfizer; internal funding from the University of Hong Kong; and consultancy fees from Merck Sharp & Dohme; Dohme, unrelated to this work. C.L.C. reports grants and personal fees from Amgen outside the submitted work. I.C.K.W. has received research grants from Amgen, Janssen, GSK, Novartis, Pfizer, Bayer and Bristol-Myers Squibb and Takeda, Institute for Health Research in England, European Commission, National Health and Medical Research Council in Australia, The European Union’s Seventh Framework Programme for research, technological development, Research Grants Council Hong Kong and Health and Medical Research Fund Hong Kong; consulting fee from IQVIA and WHO; payment for expert testimony for Appeal Court in Hong Kong; serves on advisory committees for Member of Pharmacy and Poisons Board; Member of the Expert Committee on Clinical Events Assessment Following COVID-19 Immunisation; Member of the Advisory Panel on COVID-19 Vaccines of the Hong Kong Government; is a non-executive director of Jacobson Medical in Hong Kong; a director of Advance Data Analytics for Medical Science (ADAMS) Limited (HK), Asia Medicine Regulatory Affairs (AMERA) Services Limited and OCUS Innovation Limited (HK, Ireland and UK). He was a director of Therakind Limited (UK) until April 2024. E.W.Y.C. reports grants from Research Grants Council (RGC, Hong Kong), Research Fund Secretariat of the Health Bureau, National Natural Science Fund of China, Wellcome Trust, Bayer, Bristol-Myers Squibb, Pfizer, Janssen, Amgen, Takeda, and Narcotics Division of the Security Bureau of the HKSAR; honorarium from HA; outside the submitted work. E.Y.F.W. has received research grants from the Health Bureau of the Government of the HKSAR, the Hong Kong Research Grants Council of the Government of the HKSAR, Narcotics Division, Security Bureau of the Government of the HKSAR, and National Natural Science Foundation of China, outside the submitted work. All other authors declare no competing interests.

Declaration of Sources of Funding:

Health and Medical Research Fund Research on COVID-19, Government of the Hong Kong Special Administrative Region; Research Grants Council, Collaborative Research Fund; and Health Bureau, Government of the Hong Kong Special Administrative Region.

Ethics approval:

This study was approved by the Central Institutional Review Board of the HA of Hong Kong (CIRB-2021-005-4) and the DH Ethics Committee (LM171/2021).

References

Author notes