Estimated Vaccine Effectiveness for Respiratory Syncytial Virus–Related Lower Respiratory Tract Disease

Respiratory syncytial virus (RSV) infection is a leading cause of severe acute respiratory disease in older adults and adults with underlying comorbidities¹ and may also result in long-term sequelae and trigger cardiac events.^2-4 Among US adults aged 65 years and older, approximately 160 000 RSV-related acute respiratory infection (ARI) hospitalizations and 120 000 emergency department (ED) visits are estimated to occur annually.¹

On June 21, 2023, the US Advisory Committee on Immunization Practices recommended persons aged 60 years and older may receive a single dose of a vaccine licensed to prevent RSV lower respiratory tract disease (LRTD) using shared clinical decision-making,⁵ citing in part lack of vaccine effectiveness (VE) data from the phase 3 pivotal trials regarding RSV-related hospitalizations and among high risk groups.⁶ We evaluated, in a large US health care network, Respiratory Syncytial Virus Prefusion F (RSVpreF) effectiveness against RSV-related LRTD inpatient or ED visit.

We performed a test-negative case-control study comparing the odds of RSVpreF (Abrysvo) receipt between RSV cases and test-negative controls at Kaiser Permanente Southern California (KPSC) (providing care to approximately 4.8 million members, including 1.1 million aged ≥60 years). We included patients aged 60 years or older with an International Statistical Classification of Diseases and Related Health Problems, Tenth Revision (ICD-10) code consistent with LRTD in any position for a hospitalization or ED encounter who underwent molecular testing for RSV via respiratory swab from November 24, 2023, to April 9, 2024 (eTable 1 in Supplement 1). Participants were required to have 1 year or more of health plan membership to determine comorbidities. Non-RSVpreF vaccine recipients (eg, recipients of Arexvy) and those who received RSV vaccination less than 21 days before the LRTD encounter were excluded. KPSC’s institutional review board approved this study and waived the informed consent requirement because the research involved minimal risk to participants. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.

To reduce bias associated with infrequent standard of care RSV testing, we enhanced specimen testing by salvaging respiratory swabs collected for standard-of-care molecular testing of SARS-CoV-2 and influenza. Swabs that were not SARS-CoV-2 or influenza positive and not already tested for RSV using the Roche Diagnostics Cobas eplex respiratory pathogen panel 2 (RP2) as part of routine care were defined as salvaged specimens and assayed using the same RP2 panel, which tests for 16 respiratory viral targets, and Chlamydia pneumoniae and Mycoplasma pneumoniae.

Outcomes were defined based on the results of Cobas RP2 testing 14 or fewer days before through 3 or fewer days after the LRTD encounter. Cases were RSV-positive LRTD events. Two sets of prespecified controls were used: (1) Strict (primary analysis): LRTD events that were negative for RSV, human metapneumovirus (HMPV), influenza, and SARS-CoV-2, and were positive for a nonvaccine preventable disease (VPD); adenovirus, coronavirus (229E, HKU1, NL63, or OC43), human rhinovirus/enterovirus, parainfluenza 1-4, Chlamydia pneumoniae, Mycoplasma pneumoniae; and (2) Broad (sensitivity analysis): RSV-negative LRTD events, regardless of other identified causes (and including those who tested negative for all pathogens). The strict approach accounted for potential bias associated with VPD controls⁷ and for potential RSVpreF VE against HMPV.⁸ Additional outcomes using only the strict control definition included RSV-related LRTD hospitalizations, RSV-related LRTD hospitalizations among patients with high-risk chronic medical conditions, RSV-related LRTD ED events, RSV-related severe (ie, supplemental use of oxygen) LTRD hospitalizations and ED events. Patients could contribute 1 or more LRTD event to the study if events were more than 30 days apart; deduplicated sensitivity analyses to estimate VE against LRTD hospitalization or ED events on only first event per patient were conducted.

KPSC members aged 60 years or older were eligible for RSV vaccines at no cost based on FDA-authorized indications. KPSC electronic health records captured all vaccinations administered within the health system. Records were supplemented with vaccine administration data from California’s Immunization Registry, to which clinicians are required by law to report RSV vaccinations, rendering misclassification of vaccination status unlikely. RSVpreF was administered at 7 study hospitals in the KPSC network. Patients who received RSVpreF 21 or more days before the LRTD encounter were considered vaccinated.

Odds ratios (ORs) and 95% CIs were calculated from multivariable logistic regression including prespecified and model-specified covariates: month of encounter, age, sex, self-reported race and ethnicity (Hispanic, non-Hispanic Asian or Pacific Islander, non-Hispanic Black, non-Hispanic White, or non-Hispanic multiple/other/unknown), modified Charlson score,^9-11 and health care utilization in the year before encounter. The only variable with missing values was race and ethnicity, which was combined as a non-Hispanic multiple/other/unknown category. Race and ethnicity were assessed in this study due to association with vaccination and association with the outcomes of interest. A 2-sided α of .05 was used for logistic regression modeling. Corresponding 95% CIs were calculated using the Wald method. VE was calculated as 1 − OR × 100%. Analyses were performed using SAS version 9.4 (SAS Institute). Data were analyzed from May to September 2024.

A total of 8253 patients aged 60 years and older had LRTD hospitalizations or ED encounters at study sites. A total of 6122 (74.2%) had a nasal or nasopharyngeal swab, and of these, 4084 (66.7%) were tested for RSV (3232 salvage specimens and 852 standard-of-care specimens). After applying eligibility criteria, 5714 LRTD events with tested specimens were included in the analysis (ie, met age, membership, and vaccination criteria) (Figure). The mean (SD) age was 77.4 (9.5) years, 3092 (54.1%) were female, 658 (11.5%) were non-Hispanic Asian or Pacific Islander, 1788 (31.3%) were Hispanic, 936 (16.4%) were non-Hispanic Black, and 2266 (39.7%) were non-Hispanic White. A total of 860 (15.1%) were immunocompromised. The distribution of LRTD-defining ICD-10 codes was similar between RSV-positive and RSV-negative LRTD events, with pneumonia being the most common diagnosis in both groups (eTable 2 in Supplement 1). The study population included 2722 specimens (47.6%) tested for RSV as part of standard of care, and 2992 salvaged specimens (52.4%). Overall, patient characteristics were similar between patients with salvaged vs standard of care specimens; however, salvaged specimens were more likely to be from non-White patients, those presenting in the ED, and those who received RSVpreF; and less likely to be from those who are immunocompromised (eTable 3 in Supplement 1). Compared with those in the study population and tested for RSV, LRTD inpatient or ED encounters with no swab collected or not tested for RSV were more likely to be ED visits, earlier in the RSV season, among non-White patients, among patients with lower health care utilization, among those with lower Charlson Index scores, and among those with generally fewer comorbidities (eTable 3 in Supplement 1). The majority of the study population had comorbidities (5415 [94.8%] had ≥1 Charlson comorbidity) and were aged 75 years or older (3414 [59.7%]) (Table 1). Overall, 8.8% (502 of 5714) tested positive for RSV and 3.3% (189 of 5714) received RSVpreF. Among all those who tested RSV-negative (5212 broad controls), 187 (3.6%) were vaccinated. A total of 26 RSV cases were coinfected with another VPD or non-VPD pathogen. Significant differences between vaccinated and unvaccinated persons in the broad analysis included a higher proportion vaccinated among older, non-Hispanic Asian or Pacific Islander and non-Hispanic White patients, those with 10 or more outpatient encounters in the year before LRTD encounter, those with peripheral vascular disease, and those with chronic obstructive pulmonary disease (COPD). Those with diabetes and those who were hospitalized were less likely to have received the vaccine (eTable 4 in Supplement 1). To construct the strict control group (661 participants), VPD-positive specimens (164 hMPV, 182 SARS-CoV-2, and 162 influenza; nonmutually exclusive) and those that were not positive for a non-VPD pathogen (4054 participants) were removed from the broad control group (5212 participants).

In the analysis comparing cases with strictly defined controls (502 cases and 661 controls), adjusted VE was 89% (95% CI, 52%-97%) with vaccinees receiving RSVpreF a median (IQR) of 51 (36.5-87) days before LRTD encounter (Table 2). In the analysis comparing cases with any RSV-negative controls (502 cases and 5212 controls), adjusted VE was 87% (95% CI, 48%-97%) with vaccinees receiving RSVpreF a median (IQR) of 62 (41-83) days before LRTD encounter.

Additional VE estimates against other LRTD end points are provided in Table 3. Notably, estimated VE against severe LRTD hospitalizations and ED events was 91% (95% CI, 28%-99%).

Overall, 200 patients contributed more than 1 LRTD event more than 30 days apart. Of those, 195 (97.5%) were RSV-negative, 5 (2.5%) were RSV-positive; 191 (95.5%) were unvaccinated, and 9 were vaccinated (4.5%). Sensitivity VE analyses including only the first LRTD event for each patient demonstrated similar results to the main analyses, (strict: adjusted VE [aVE], 89%; 95% CI, 52%-97%; broad: aVE, 87%; 95% CI, 48%-97%).

Since licensure in 2023, there are few reports of RSVpreF effectiveness in adults in a clinical practice setting. In our primary analysis, RSVpreF effectiveness was 89% against RSV-related LRTD requiring hospitalization or ED visit among adults aged 60 years or older in the first 5 months of vaccine use.

RSVpreF’s pivotal clinical trial, RENOIR,¹² reported similar VE of 85.7% (95% CI, 32.0%-98.7%) at the end of season 1 for RSV-related LRTD with 3 or more signs or symptoms overall, with a similar result of 84.6% (95% CI, 32.0%-98.3%) for medically attended events.⁵ Medically attended LRTD in RENOIR included a mix of health care encounters (ie, outpatient and inpatient), but consistent with other phase 3 trials, VE estimates for hospitalization could not be estimated due to few RSV-positive hospitalizations among enrolled participants. In RENOIR, among those with 1 or more high risk conditions, VE against RSV-LRTD with 3 or more symptoms was 81.8% (95% CI, 16.7%-98.0%) at the end of season 1.¹³ While this latter finding is promising, our study included a substantially larger proportion of persons at high risk for severe disease than RENOIR (eg, median age 77 vs 68 years, 56% vs 6% with COPD, and 15% vs 0% immunocompromised) and assesses VE against a severe clinical outcome of ED and hospitalization visits. Moreover, in our study, VE was maintained when evaluating severe LRTD, as defined by oxygen supplementation.

Results from this general population evaluation, therefore, expand on clinical trial data by documenting VE against more severe outcomes (hospitalization and ED encounters or need for oxygen supplementation) among an older patient population with a higher proportion of high risk conditions compared with those in the clinical trial.¹⁴ Indeed, in light of the 2024 Advisory Committee on Immunization Practices update to use of RSV vaccines,⁵ which now recommends their use for all adults aged 75 years and older, and for adults aged 60 to 74 years at increased risk of severe RSV, our results are particularly relevant as they address data gaps for groups called out in this guidance.

Our analytic approach allowed for minimization of bias from several sources. First, we salvaged remnant swabs collected for standard of care testing for influenza and SARS-CoV-2 to overcome low routine clinical RSV testing rates. Next, our strict control definition, which required a positive result for a non-VPD, was used to ensure effective specimen collection technique and mitigate underascertainment associated with low sensitivity of nasopharyngeal/nasal swab reverse transcription polymerase chain reaction testing in adults (ie, approximately 50% of RSV events not detected),^5,15 leading to RSV cases being misclassified as controls and potentially biasing VE estimates toward a null value. Next, removal of VPD infections from the control group addressed the correlation of respiratory vaccines that can confound test-negative design VE estimates.⁷ Lastly, we designated hMPV as a possible VPD, based on literature describing the isolation of multiple cross-reactive neutralizing antibodies between RSV and hMPV from human donors^8,16 While it is currently unknown if this may result in RSVpreF vaccination providing cross-protection against hMPV, removing this pathogen from the control group allowed us to minimize this potential bias in our VE estimates against RSV end points. Regardless of the control selection approach used, VE was similar (89% vs 87%).

A limitation of this initial analysis is that less than 5% of those with LRTD events were vaccinated in this first season. This may impact generalizability of estimates and prevent stable VE estimates for some subgroups. This vaccination rate is lower than what is reported in national estimates,¹⁷ which may be due to a later rollout of RSVpreF at KPSC, absence of an automated alert for clinicians to prompt RSV vaccination, or other issues related to implementation of shared clinical decision-making for RSV vaccination at KPSC in this first year of administration. These early first-season estimates miss the first part of RSV season, resulting in median follow-up times of 51 and 62 days for strict and broad control group analyses, respectively. If our vaccinated population is older and with more comorbidities than recipients nationwide, our VE estimates may be biased toward the null, or vice versa. Further case accrual and vaccine uptake after the 2024 to 2025 RSV season will continue and will power future stratified analyses. Multiseason future analyses will be better suited to capture potential waning of protection with time since vaccination.

Not all LRTD events in the ED or hospital setting were tested for RSV. In general, LRTD events without a test were in the ED and among a younger, healthier population with events later in the RSV season. While the inclusion of higher-risk patients in the study population may bias VE lower, the intended target population for RSV vaccines consists of those at highest risk for severe outcomes. Important strengths of our enhanced testing approach include substantial increase of our study size and more robust analyses in the ED setting.

This demonstration of the general population effectiveness of RSV vaccination among older adults provides insight into the potential public health impact of this strategy. Based on our study results and RSV incidence in older adults,¹ for approximately every 240 persons vaccinated, 1 RSV-related ED or hospitalization encounter could be prevented in the first season after vaccination. These data suggest use of RSVpreF in older adults, providing an opportunity to reduce severe medically attended RSV disease burden.