Senior Scientist University of British Columbia, Canada
Background: Previous studies have reported treatment with nirmatrelvir-ritonavir is associated with reduced risk of Covid-19 hospitalization or death in high risk individuals.
Objectives: To identify medical conditions most strongly associated with nirmatrelvir-ritonavir treatment effectiveness.
Methods: We analyzed individuals from a previous matched cohort study in British Columbia administrative claims data between February 2022 and February 2023, which found nirmatrelvir-ritonavir treatment (the exposure) was associated with reduced risk of Covid-19 hospitalization or death (the outcome). The study included severely or moderately immunocompromised individuals, and individuals who were not immunocompromised but with medical conditions thought to be associated with complications from Covid-19 infection. To determine which medical conditions were most strongly associated with treatment effectiveness, we designed a bootstrap algorithm which, in each of 10,000 iterations, selected a random 50% sample of matched pairs from the original cohort, estimated a matched odds ratio for the association between nirmatrelvir-ritonavir and Covid-19 hospitalization or death, and calculated Pearson correlations between the natural logarithm of the matched odds ratio and each of 63 potential risk factors. The original cohort was then incrementally reassembled by including matched pairs of individuals who possessed factors correlated with treatment effectiveness, in ascending order of Pearson correlation. In each increment, the cumulative proportion of total outcome events captured was estimated, and a matched odds ratio was estimated to determine the association between treatment and the outcome.
Results: The study included 6,866 individuals with Covid-19 between February 1, 2022, and February 3, 2023, who were previously matched on high-dimensional propensity score for treatment with nirmatrelvir-ritonavir. After iterative sampling, the factors most correlated with nirmatrelvir-ritonavir treatment effectiveness were use of immunosuppressants, treatment for cancer, moderate renal disease, and acquired or congenital heart disease. The incremental analysis, restricted to matched pairs where some or all of these factors were present, captured 90.7% of all events which occurred in the original cohort, while preserving the magnitude of the matched odds ratio of 0.74 (95% confidence interval 0.43 to 1.27).
Conclusions: This study supports a hypothesis that population effectiveness of nirmatrelvir-ritonavir treatment may be largely attributable to use in individuals possessing one or more of a limited number of commorbidities.