Nirmatrelvir-Ritonavir and COVID-19 Mortality and Hospitalization Among Patients With Vulnerability to COVID-19 Complications

Key Points Question What is the association of nirmatrelvir and ritonavir exposure with the risk of death or COVID-19–related hospitalization when accounting for patient vulnerability to complications from COVID-19 infection? Findings In this cohort study of 6866 individuals with COVID-19, treatment with nirmatrelvir and ritonavir was associated with lower risk of death or hospitalization in the most clinically extremely vulnerable individuals but not in less vulnerable individuals. Individuals who were not extremely vulnerable to experiencing complications from COVID-19, whose median age was 79 years, had greater risk of the outcome while receiving nirmatrelvir and ritonavir, but the finding was not statistically significant. Meaning In this study, treatment with nirmatrelvir and ritonavir was not associated with reduced risk of death or hospitalization among individuals who were not extremely vulnerable to complications from COVID-19 infection, regardless of age.


Introduction
Nirmatrelvir and ritonavir (Paxlovid [Pfizer]) is an oral antiviral drug combination that targets a key SARS-CoV-2 protease enzyme.Nirmatrelvir and ritonavir was approved based on interim efficacy and safety data from the Evaluation of Inhibition for COVID-19 in High-Risk Patients (EPIC-HR) trial, conducted before the emergence of the Omicron variant. 1,2In the trial, 2246 patients who were not vaccinated against COVID-19 were randomized to receive nirmatrelvir and ritonavir or placebo within 5 days of symptom onset.Nirmatrelvir and ritonavir reduced the primary composite end point of 28-day risk of death or COVID-19-related hospitalization by 5.6% absolute (88% relative) compared with placebo.By day 28, there were 0 deaths in the nirmatrelvir and ritonavir group and 12 in the placebo group.EPIC-HR reported numerically fewer serious adverse events but more suspected drug-related adverse events with nirmatrelvir and ritonavir than placebo.
Parallel to EPIC-HR, a further 1141 lower-risk adults were studied in a second trial known as EPIC-SR (called standard-risk). 3,4The manufacturer announced the closure of the trial in a June 14, 2022, media release, "due to a very low rate of hospitalization or death observed in the standard-risk patient population." 4 Nirmatrelvir and ritonavir was not associated with reduced symptoms (the primary end point), and the media release reported a non-statistically significant reduction of 0.9% absolute (51% relative) in hospitalization or death.Peer-reviewed observational analysis has also shown COVID-19-related hospitalization or death from any cause was significantly lower with nirmatrelvir and ritonavir treatment, 5 although by approximately half the magnitude reported in the EPIC-HR trial.Observational studies have demonstrated a benefit of nirmatrelvir and ritonavir treatment in higher risk individuals. 5,6otective associations with nirmatrelvir and ritonavir therapy have been shown to be variable across the different populations and study periods, which suggests the benefit risk profile of nirmatrelvir and ritonavir depends on vulnerability to complications from COVID-19 infection.The benefit-harm profile of nirmatrelvir and ritonavir thus remains uncertain.British Columbia (BC) provides a natural experiment for examining nirmatrelvir and ritonavir according to vulnerability to complications from COVID-19.BC adopted eligibility criteria for nirmatrelvir and ritonavir that differed substantially from participants in the EPIC-HR trial and included individuals who were more comparable with those studied in the EPIC-SR trial.Furthermore, patients enrolled in EPIC-HR were unvaccinated, had no natural immunity from prior COVID-19 infection, were infected by COVID-19 variants that were different from those now circulating, and were not taking drugs with known CYP 3A4 interactions. 7As in the EPIC-HR trial, we sought to analyze the 28-day risk of death or COVID-19associated hospitalization in the 4 groups of vulnerable individuals in BC with elevated risk of complications who were given access to nirmatrelvir and ritonavir.

Study Design and Data Source
We undertook a retrospective cohort study between February 1, 2022, and February 3, 2023, of individuals who had increased vulnerability to complications from COVID-19 infection.Death from any cause and COVID-19-related hospitalization were compared between individuals who were either prescribed or not prescribed nirmatrelvir and ritonavir.The study used anonymized, individual-level, and linkable administrative health databases from the BC Ministry of Health.

Identification of Study Cohorts
Four mutually exclusive populations of individuals at higher risk for complications from COVID-19 infection were eligible for nirmatrelvir and ritonavir during the study period.Detailed definitions used to identify these populations are provided in eAppendix 1 in Supplement

Identification of Outcome Events
The primary outcome was a composite

Sensitivity and Subgroup Analyses
Several sensitivity and subgroup analyses were performed.The primary outcome was further

Statistical Analysis
After assembling nirmatrelvir and ritonavir exposure and outcome data for the matched cohorts in 2 × 2 tables, we estimated the 28-day risk difference (RD) and the 28-day relative risk (RR) for each study outcome.These parsimonious analyses allowed for the straightforward reporting of RDs and numbers-needed-to-treat but did not account for the possible influence of competing risks.The reasonability of this approach was checked by comparing the RR estimates with hazard ratios estimated using the method of Fine and Gray 12 and by estimating outcome-specific hazard functions. 13Both of those methods impose a proportional hazards assumption, which was checked Results for the association between nirmatrelvir and ritonavir and risk of death or COVID-19related emergency hospitalization, the primary end point, are shown in Table 2.There was a statistically significant RD of −2.5% in the CEV1 group (95% CI, −4.8% to −0.2%), a −1.7% RD in the CEV2 group (95% CI, −2.9% to −0.5%), a nonstatistically significant −1.3% RD in the CEV3 group (95% CI, −2.8% to 0.1%), and a nonstatistically significant RD of 1.0% in the EXEL group (95% CI, −0.9% to 2.9%).Cumulative incidence functions are shown in the Figure for the CEV2, CEV3 and EXEL groups (to protect privacy, the CEV1 group could not be displayed).The Fine and Gray subdistribution hazard ratios and outcome-specific hazard ratios accounting for competing risks were the same or similar to the 28-day RR estimates from the fixed cohort analysis (eAppendix 4 in Supplement 1).
In Table 3, the primary outcome is reported for the subgroups of individuals who were age 70 years or older and for male compared with female participants.RDs were nominally, but not statistically significantly, further from the null in the subgroup aged 70 years or older compared with the main analysis.RDs were nominally more protective in male compared with female participants, but these differences were also not statistically significant.Detailed data could not be presented for the CEV1 groups because of restrictions on reporting small numbers of events.Although event counts could not be reported, in individuals aged 70 years or older, the RD in 98 nirmatrelvir and ritonavir-exposed and 98 unexposed individuals was −8.2% (95% CI, −13.6% to −2.7%).In the CEV1 group, 179 nirmatrelvir and ritonavir-exposed women were matched to 179 unexposed women, and 101 nirmatrelvir and ritonavir-exposed men were matched to 101 unexposed men.Respectively, the RDs in these subgroups were −1.1% (95% CI, −2.7% to 0.4%), and −5.0%(95% CI, −10.6% to 0.7%).
Among vaccinated individuals, RD estimates in the 4 vulnerability groups were similar to estimates for the whole study group.Small cell restrictions prevented reporting of the association in unvaccinated individuals in the 3 CEV groups.However, in the EXEL group, there were 7 events in 210 nirmatrelvir and ritonavir-exposed unvaccinated individuals and 6 events in 170 non-nirmatrelvir and ritonavir-exposed unvaccinated individuals.The RD in this subgroup was −0.2% (95% CI, −3.9% to 3.5%).There were no statistically significant differences between nirmatrelvir and ritonavirexposed and unexposed individuals in ED visits (Table 4).
In the main analysis, nirmatrelvir and ritonavir-exposed individuals did not require a PCR test because COVID-19 was the only indication for nirmatrelvir and ritonavir.A sensitivity analysis that required all nirmatrelvir and ritonavir-exposed individuals to also have a positive PCR test resulted in a subgroup of 1362 individuals from the original 6866 study population.A small number of events prevented reporting the sensitivity analysis in the CEV1 group.In 480 individuals in the CEV2 group,

Discussion
In this observational study of 6866 individuals in BC, nirmatrelvir and ritonavir exposure was associated with lower risk of death or COVID-19-related hospitalization in individuals with extreme   4 where no difference in the primary outcome was seen, although we observed a nominally higher hospitalization and mortality rate in those exposed to nirmatrelvir and ritonavir.With one important exception, our results were generally consistent with observational studies of nirmatrelvir and ritonavir, 5,6 where the benefit of nirmatrelvir and ritonavir is driven by high-risk groups within an exposed population.The important exception in our study was that, owing to the stratified approach in our analysis, older age (Ն70 years) did not have a statistically significant beneficial association with nirmatrelvir and ritonavir after accounting for other comorbidities.Abbreviations: CEV, clinically extremely vulnerable; EXEL, expanded eligibility; NA, not applicable.
a Events for the CEV1 groups cannot be reported because of small cell restrictions.Among individuals aged 70 years or older, the risk difference was −8.2% (95% CI −13.4% to −2.7%).
b Number needed to treat is for harm, but not statistically significant.
analyzed in subgroup analyses of COVID-19 vaccination history, age 70 years or older, sex, history of diabetes, and history of kidney disease.To avoid confounding by indication, we needed to ensure that all people entering the study were infected with COVID-19.The widespread and unrecorded use of rapid antigen tests during the study period required restricting non-nirmatrelvir and ritonavirexposed control individuals to those who received positive PCR tests.Nirmatrelvir and ritonavirexposed individuals did not require a positive PCR test because COVID-19 was the only indication for the drug, and prescribing physicians were required to make sure their patients tested positive within 5 days prior to prescribing nirmatrelvir and ritonavir.A sensitivity analysis was conducted in which nirmatrelvir and ritonavir-exposed individuals were also required to have a positive COVID-19 PCR test.

Figure . No
Figure.Cumulative Incidence of Death or COVID-19-Related Emergency Hospitalization Nirmatrelvir and Ritonavir and COVID-19 Mortality and Hospitalization Prescription data for nirmatrelvir and ritonavir and other drugs were obtained from the PharmaNet database of all prescriptions filled at community pharmacies.COVID-19 vaccination records and polymerase chain reaction (PCR) test records were also obtained from the Ministry of Health.Demographic, diagnostic, and medical procedure data, used to identify clinically extremely vulnerable (CEV) cohorts and perform statistical adjustment, were obtained from the Medical Services Plan database, the hospital Discharge Abstract Database, and the National Ambulatory Care Reporting System database (NACRS).This study followed the Strengthening the Reporting of JAMA Network Open.2023;6(10):e2336678. doi:10.1001/jamanetworkopen.2023.36678(Reprinted) October 2, 2023 2/11 Downloaded From: https://jamanetwork.com/ on 10/06/2023 11rameters used for matching were age (±2 years), sex, propensity score (±0.2 multiplied by the standard deviation of the pooled standard deviation of the logit of the propensity score),11and year and calendar month (±1 month) of COVID-19-positive test.Follow-up for study outcomes in an nirmatrelvir and ritonavir-exposed individual began on their nirmatrelvir and ritonavir initiation date.For each unexposed matched control participant, to avoid immortal time bias, follow-up began on the individual's COVID-19 test collection date, plus the number of days that lapsed between the test collection date and treatment initiation in the control individual's nirmatrelvir and ritonavir-exposed counterpart.Where the nirmatrelvir and ritonavirexposed individual in a match did not have a PCR test, follow-up in the non-nirmatrelvir and ritonavir-exposed individual began 3 days after their PCR test sample collection date.
Cohorts of individuals with COVID-19, identified from either a PCR test, a prescription for nrirmatrelvir and ritonavir, or both, were assembled from the 4 vulnerable populations.Individuals were not allowed to enter a cohort more than once.For individuals who at different times qualified as nirmatrelvir and ritonavir exposed and unexposed, the nirmatrelvir and ritonavir instance was included.Prior to initiating treatment with nirmatrelvir and ritonavir or having a positive COVID-19 test, individuals required at least 730 days of continuous enrollment in the BC Medical Services Plan (gaps of Յ30 days were permitted).Individuals were excluded from the study if, at the time before their test sample collection date (or 3 days prior to initiating nirmatrelvir and ritonavir if there was no test), they were younger than 18 years of age, had a history of severe kidney or liver disease, pregnancy, use of remdesivir, or, within 30 days before their test, were in hospital for any reason.Detailed exclusion criteria are provided in eAppendix 2 in Supplement 1.Individuals prescribed nirmatrelvir and ritonavir were matched to individuals with COVID-19 who were not prescribed nirmatrelvir and ritonavir.Nirmatrelvir and ritonavir-exposed individuals were not required to have a positive PCR test, but all non-nirmatrelvir and ritonavir-exposed individuals required a positive test.Before matching, to minimize potential confounding, highdimensional propensity score (HDPS) models were estimated for each of the 4 groups.10TheHDPS algorithm empirically selected covariates from the 730-day period prior to the COVID-19 test sample JAMA Network Open | Infectious Diseases Nirmatrelvir and Ritonavir and COVID-19 Mortality and Hospitalization JAMA Network Open.2023;6(10):e2336678. doi:10.1001/jamanetworkopen.2023.36678(Reprinted) October 2, 2023 3/11 Downloaded From: https://jamanetwork.com/ on 10/06/2023 collection date.Within each cohort, nirmatrelvir and ritonavir-exposed individuals were matched 1-to-1 with non-nirmatrelvir and ritonavir-exposed individuals, without replacement, using the nearest neighbor method.
outcome of COVID-19-related emergency hospital visit or admission, or death from any cause, within 28 days of an individual's cohort entry date.A secondary outcome was emergency department visit for any reason, with or without a subsequent admission to hospital.Estimation of hospitalization outcomes relied on emergency department (ED) records after March 31, 2022.This was because complete hospital discharge abstract data for the remainder of study period will not be available until late 2023.Instead, COVID-19-related emergency admissions after March 31, 2022, were identified using ED visit records from the NACRS database, which was complete for our study period.A NACRS record was counted as a COVID-19-related emergency hospitalization if it indicated COVID-19 infection (International Statistical Classification of Diseases and Related Health Problems, Tenth Revision, codes U07.1 and U07.2) and subsequent admission to hospital.To assess the reliability of this approach, we compared complete hospital discharge records with NACRS records in the year before our study.From March 1, 2021, to February 28, 2022, a NACRS record that indicated subsequent admission to hospital was 88.5% sensitive and 92.3% specific for the presence of an emergency hospital admission record in the hospital discharge abstract database.

JAMA Network Open | Infectious Diseases
14ing cumulative sums of Martingale-based residuals.14Dataanalysis was conducted in SAS version 7.15 (SAS Institute).We used a 95% CI that excluded the null as our level of statistical significance.

Table 1 .
Nirmatrelvir and ritonavir-exposed individuals were balanced on sex and age in all

Table 1 .
Baseline Characteristics for the Matched Study Cohorts a Data on the first 3 could not be shown to protect privacy.For the remaining conditions, there were no statistically significant differences, with the exception of more nirmatrelvir and ritonavir-exposed individuals with neurological conditions in the Abbreviations: CEV, clinically extremely vulnerable; NA, not applicable.aCohortswerematched on age (±2 years), sex, propensity score, and year and month of cohort entry (±1 month).bAvalue of 3 days was imputed for exposed individuals without a recorded positive SARS-CoV-2 PCR test within 5 days prior to nirmatrelvir and ritonavir dispensing.cDenotes a number less than 5. d Definitions for baseline conditions are provided in eAppendix 1 in Supplement 1.JAMA Network Open | Infectious DiseasesNirmatrelvir and Ritonavir and COVID-19 Mortality and Hospitalization JAMA Network Open.2023;6(10):e2336678. doi:10.1001/jamanetworkopen.2023.36678(Reprinted)October 2, 2023 5/11 Downloaded From: https://jamanetwork.com/ on 10/06/2023were also compared for each matched group.

Table 2 .
Risk of Death or COVID-19-Related Emergency Hospitalization (Primary End Point) by Cohort Number needed to treat is for harm, but not statistically significant.
a Number masked to preserve privacy.b Downloaded From: https://jamanetwork.com/

JAMA Network Open | Infectious Diseases
vulnerability to complications from COVID-19.The same association was not observed in lower-risk individuals in the EXEL group, a result that was robust across sex and older vs younger age.The results from our CEV cohorts were compatible with the results of the EPIC-HR trial, 2 where a statistically significant reduction of hospitalization or death was seen.The results in the lower-risk EXEL group appeared comparable with the unpublished of the EPIC-SR trial, JAMA Network Open.2023;6(10):e2336678. doi:10.1001/jamanetworkopen.2023.36678(Reprinted) October 2, 2023 7/11 Downloaded From: https://jamanetwork.com/ on 10/06/2023

Table 3 .
Risk of Death or COVID-19-Related Emergency Hospitalization, by Sex and in Individuals Aged 70 Years or Older

Table 4 .
Risk of Any Emergency Department Visit, by Group Abbreviations: CEV, clinically extremely vulnerable; EXEL, expanded eligibility.aNumber needed to treat is for harm, but not statistically significant.