Real-World Effectiveness of Nirsevimab Against Respiratory Syncytial Virus: A Test-Negative Case-Control Study

IMPORTANCE: Nirsevimab, a long-acting monoclonal antibody, has demonstrated efficacy against RSV-related lower respiratory tract infections (LRTIs) in clinical trials. Post-licensure monitoring is essential to confirm these benefits in real-world settings. OBJECTIVE: To evaluate the real-world effectiveness of nirsevimab against medically attended RSV infections in infants and to assess how effectiveness varies by disease severity, dosage, and time since immunization. DESIGN, SETTING, AND PARTICIPANTS: This test-negative case-control study used inpatient, outpatient, and emergency room data from the Yale New Haven Health System. Nirsevimab-eligible infants who were tested for RSV using polymerase chain reaction between October 1, 2023 and May 9, 2024 were included. Cases were infants with confirmed RSV infections; controls were those who tested negative. EXPOSURE: Nirsevimab immunization, verified through state immunization registries. MAIN OUTCOMES AND MEASURES: Effectiveness was estimated using multivariable logistic regression, adjusting for age, calendar month, and individual risk factors. Separate models examined effectiveness by clinical setting, disease severity, dose, and time since immunization. Broader outcomes, including all-cause LRTI and LRTI-related hospitalization, were also analyzed, with stratification by early and late respiratory seasons. RESULTS: The analytic sample included 3,090 infants (median age 6.7 months, IQR 3.6–9.7), with 680 (22.0%) RSV-positive and 2,410 (78.0%) RSV-negative. 21 (3.1%) RSV-positive and 309 (12.8%) RSV-negative infants received nirsevimab. Effectiveness against RSV infection was 68.4% (95% CI, 50.3%–80.8%). Effectiveness was 61.6% (95% CI, 35.6%–78.6%) for outpatient visits and 80.5% (95% CI, 52.0%–93.5%) for hospitalizations. The highest effectiveness, 84.6% (95% CI, 58.7%–95.6%), was observed against severe RSV outcomes requiring ICU admission or high-flow oxygen. Although effectiveness against RSV infections declined over time, it remained significant at 55% (95% credible interval, 16%–75%) at 14 weeks post-immunization. Protective effectiveness was also observed against all-cause LRTI and LRTI-related hospitalizations during peak RSV season (49.4% [95% CI, 10.7%–72.9%] and 79.1% [95% CI, 27.6%–94.9%], respectively). However, from February to May, when RSV positivity was low, effectiveness against these broader outcomes was negligible. CONCLUSIONS AND RELEVANCE: Nirsevimab provided substantial protection against RSV-related outcomes for at least three months. These findings support the continued use of nirsevimab and provide evidence that may help build public confidence in the immunization program.


Table of contents SECTION 1. SUPPLEMENTAL FIGURES REFERENCED IN THE MAIN MANUSCRIPT
Figure S1 Correlation between potential confounders in the test-negative case-control analysis.Figure S2 RSV tests and nirsevimab doses during the study period.Figure S3 Overview of nirsevimab effectiveness: Current study estimates in context with prior research.Figure S4 Effectiveness of nirsevimab against RSV infections by dose, clinical setting, and disease severity.Figure S5 Effectiveness of nirsevimab against RSV-associated LRTI by time since immunization.Figure S6 Trace plots for the coefficients of waning effectiveness.

SECTION 2. SUPPLEMENTAL TABLES REFERENCED IN THE MAIN MANUSCRIPT
Table S1.Definition of key clinical outcomes and risk factors.Table S2.Variable selection for the multivariable logistic regression models.Table S3.Comparison of immunized and unimmunized patients.Table S4.Clinical characteristics of RSV-positive cases.Table S5.Sensitivity Analysis.S1), and insurance type (private, public, uninsured).The numbers show the correlation coefficients between any of the two variables, and a value larger than 0.5 was defined as a moderate or strong correlation.Due to collinearity between low birth weight and prematurity, and high rates of missing data (~25%), only the "at least one risk factor" variable was retained (panel B).Iterations of the burn-in period (iterations 0-50,000) are excluded, and only the sampled iterations (50,000-60,000) are presented.Data are presented as median (IQR) for continuous measures and n/total (%) for categorical measures.

SECTION 2. SUPPLEMENTAL TABLES REFERENCED IN THE MAIN MANUSCRIPT
1 Standardized mean difference: the difference in means between case and control participants in units of the pooled SD.Covariates with an absolute standardized mean difference greater than 0.2 were considered to have important imbalances.
2 Inclusing Asian, Pacific Islander, Middle Eastern or Northern American, American Indian, or Native American by self-reporting.
3 Have at least one of the following conditions recorded in the infant's medical history or diagnosis records: 1) Anemia; 2) Immunodeficiency (e.g.transplantation history, leukemia, etc.); 3) Cardiac diseases (including congenital heart diseases diagnosed at birth or any reporting of heart conditions); 4) Pulmonary diseases; 5) Down syndrome; 6) Small for gestational age (birth weight < 2,500 grams); 7) Prematurity (gestational age less than 37 weeks). 1 Data are presented as median (IQR) for continuous measures and n/total (%) for categorical measures.

Estimating the effectiveness of nirsevimab over time
We evaluated the waning of nirsevimab's protective effect over time using a logistic regression model within a Bayesian framework.For the "th test record in our dataset, the observed case status (i.e., whether the patient tested positive or negative for RSV) followed a Bernoulli distribution, such that #$%&_()$)*% " ~ -&./0*112(4 " ).
Due to the waning nature of passive immunity, we assumed that nirsevimab's effectiveness had a non-increasing trend over time.To reflect this in the model, we imposed a monotonic structure on the regression coefficients !!'s, such that M(0, ) represents truncation at 0, allowing J !)# to take only non-negative values.For ! # (coefficient for the effectiveness 0-2 weeks after vaccination), we used a weakly informative prior distribution: We examined effectiveness over time against various clinical endpoints.The model for each endpoint was fitted separately in the rjags package in R version 4.3.1, in which we collected 10,000 samples from the posterior distribution after discarding the first 50,000 samples in the burn-in period.Convergence was evaluated using trace plots (Figure S6).The estimated effectiveness of nirsevimab after a given period of time (for time interval n) since immunization NP ! was calculated as Medians and 95% quantile-based credible intervals were calculated from the collected posterior samples.

SECTION 3 . 1 SECTION 1 .
Figure S1.Correlation between potential confounders in the test-negative case-control analysis.Multivariate logistic regression was used to estimate nirsevimab effectiveness against various clinical outcomes.Potential confounders were selected using backward selection from variables in the initial model (panel A), including age at testing (<3, 3-5, 6-8, 9-11, ≥12 months), calendar month of testing, race/ethnicity, birth weight, prematurity (gestational age <37 weeks), presence of at least one risk factor (see TableS1), and insurance type (private, public, uninsured).The numbers show the correlation coefficients between any of the two variables, and a value larger than 0.5 was defined as a moderate or strong correlation.Due to collinearity between low birth weight and prematurity, and high rates of missing data (~25%), only the "at least one risk factor" variable was retained (panel B).

Figure S2 .
Figure S2.RSV tests and nirsevimab doses during the study period.Panel A shows the number of RSV tests, with orange bars for RSV-positive cases and green bars for RSV-negative controls.Panel B displays the number of nirsevimab doses administered.The vertical dashed line marks the start of nirsevimab administration in Connecticut (October 1, 2023).

Figure S3 .
Figure S3.Overview of nirsevimab effectiveness: current study estimates in context with prior research.This figure contrasts adjusted effectiveness (post-licensure) and efficacy (prelicensure) estimates from prior studies with those from the current study.The right panel shows means (dots) and uncertainty intervals (bars).Gold squares represent Phase IIb/III trial data, black squares represent observational studies, and blue circles represent the current study (highlighted in gray).Ernst et al. 2024[1] did not report uncertainty intervals.

Figure S4 .
Figure S4.Effectiveness of nirsevimab against RSV infections by dose, clinical setting, and disease severity.Square dots indicate mean effectiveness estimates, with horizontal lines representing 95% confidence intervals.All models adjusted for age and calendar month.Models for RSV-associated hospitalization and severe disease also accounted for the presence of underlying risk factors.

Figure S5 .
Figure S5.Effectiveness of nirsevimab against RSV-associated LRTI by time since immunization.The green curve and shaded area represent the median and 95% credible interval of the estimated efficacy of nirsevimab reported by Hodgson et al.[2], where efficacy over time was estimated using data from Phase IIb and Phase III trials in a survival model.The black dots denote the median estimates of nirsevimab's effectiveness in preventing RSVassociated LRTI from our current study, using the same endpoint as in Hodgson et al. for comparison[2].The error bars show the 95% credible intervals for the estimates, and the labels provide the exact values.

Figure S6 .
Figure S6.Trace plots for the coefficients of waning effectiveness.Trace plots for !!(effectiveness coefficient for each biweek interval after nirsevimab immunization, n = 1,2,3,...9) are displayed by the examined outcome.All parameters demonstrate good convergence.Iterations of the burn-in period (iterations 0-50,000) are excluded, and only the sampled iterations (50,000-60,000) are presented.

Table S2 . Variable selection for the multivariable logistic regression models.
The first row for each outcome presents the full model with all potential confounders.In each subsequent row, one variable is removed per step, with the final row showing the confounders included in the final model.Immunization status was included a priori in all models, and the final model was selected based on the lowest Akaike information criterion (AIC) score.
1Lower respiratory tract infection (LRTI): see definition in Table Give unadjusted estimates and, if applicable, confounder-adjusted estimates and their precision (eg, 95% confidence interval).Make clear which confounders were adjusted for and why they were included 7,11 (b) Report category boundaries when continuous variables were categorized NA (c) If relevant, consider translating estimates of relative risk into absolute risk for a meaningful time period Limitations 19 Discuss limitations of the study, taking into account sources of potential bias or imprecision.Discuss both direction and magnitude of any potential bias Interpretatio n 20 Give a cautious overall interpretation of results considering objectives, limitations, multiplicity of analyses, results from similar studies, and other relevant evidence Funding 22 Give the source of funding and the role of the funders for the present study and, if applicable, for the original study on which the present article is based An Explanation and Elaboration article discusses each checklist item and gives methodological background and published examples of transparent reporting.The STROBE checklist is best used in conjunction with this article (freely available on the Web sites of PLoS Medicine at http://www.plosmedicine.org/,Annals of Internal Medicine at http://www.annals.org/,and Epidemiology at http://www.epidem.com/).Information on the STROBE Initiative is available at http://www.strobe-statement.org.
9Other information 1 *Give information separately for cases and controls.Note: