Effectiveness of BNT162b2 COVID-19 Vaccination in Children and Adolescents

VISION Network is a collaboration between the Centers for Disease Control and Prevention and healthcare systems with integrated medical, laboratory, and vaccination records (obtained from electronic health records, state and local immunization registries, and claims).²⁴  This study included pediatric data from 201 emergency departments (EDs), 105 UC clinics and 164 hospitals from 9 VISION network partners across 10 states: Baylor Scott and White Health, Columbia University Irving Medical Center, HealthPartners/Children’s Minnesota, Intermountain Healthcare, Kaiser Permanente Northern California, Kaiser Permanente Northwest, Paso del Norte Health Information Exchange, Regenstrief Institute, and University of Colorado.

We included all patients aged 5 to 17 years with eligible ED/UC encounters or hospitalizations from April 2021 through September 2022. We obtained demographic characteristics and underlying medical conditions from electronic health records.

This study was approved by partners’ institutional review boards or under a reliance agreement with Westat, Inc.’s institutional review board.

This was a case-control study with a test-negative design²⁴  that assessed BNT162b2 COVID-19 vaccine effectiveness against ED/UC encounters and hospitalizations lasting ≥24 hours among patients aged 5 to 17 years with diagnoses consistent with COVID-19-like illness (CLI).²⁴ 

We defined CLI using International Classification of Diseases, ninth and 10th Revision (ICD) diagnosis codes from ED/UC encounters and hospital discharge (Supplemental Table 4). We considered 4 CLI categories: (1) acute respiratory CLI (eg, COVID-19 pneumonia, viral pneumonia, influenza diseases, acute respiratory distress syndrome, asthma exacerbation); (2) acute nonrespiratory CLI (cause- unspecified gastroenteritis, thrombosis, acute myocarditis); (3) acute respiratory illness signs and symptoms consistent with CLI (eg, cough, dyspnea, hypoxemia); and (4) acute febrile illness signs and symptoms.

We defined cases as having ≥1 CLI code in any 1 of the 4 categories and ≥1 positive severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) reverse transcription polymerase chain reaction (RT-PCR) test result; controls had ≥1 CLI code and negative RT-PCR test results. We included RT-PCR results occurring ≤14 days before to 72 hours after the ED/UC encounter date or hospital admission. We defined the index date for medical encounters as the earlier date of either specimen collection associated with the most recent SARS-CoV-2 RT-PCR result (if before the medical encounter) or medical encounter date (if testing occurred only after the encounter). We also identified ICU admissions and inpatient deaths as measures of COVID-19 disease severity.²⁶ 

We excluded patients vaccinated before the Advisory Committee on Immunization Practice recommendation date for their age, patients who received only 1 vaccine dose, or if index date was <14 days after receipt of dose 2 or <7 days after dose 3. Additionally, we excluded patients who received a third dose before monovalent boosters were recommended for their age, a monovalent booster <5 months after dose 2, >3 doses of vaccine, or mRNA-1273 vaccine (Moderna¹¹ ). Because of differences in authorization timing and number of recommended doses, we also excluded patients who were likely immunocompromised. We considered moderate to severely immunocompromising conditions based on ICD-9 or ICD-10 codes for solid malignancy, hematologic malignancy, rheumatologic or inflammatory disorder, other intrinsic immune condition or immunodeficiency, or organ or stem cell transplant.

We included medical encounters occurring ≥14 days after age-specific COVID-19 vaccine eligibility dates to allow all patients the opportunity to be fully vaccinated before their index date. For adolescents aged 16 to 17 years, the study period began April to May 2021 (when vaccines were recommended and available at study sites). For adolescents aged 12 to 15 years, the study period began June 16, 2021. For children aged 5 to 11 years, the study period began on January 22, 2022. Therefore, we estimated VE only during the Omicron predominant period for the 5 to 11-year age group.

We determined the site-specific dates when the Delta and Omicron variants became predominant (ie, >50% of sequenced viruses) based on state and local surveillance data (Supplemental Table 5). Delta became predominant between June 1 and July 3, 2021, and Omicron between December 16 and 26, 2021.

We compared cases versus controls with respect to their demographic characteristics and comorbidities. A standardized mean difference (SMD) >0.2 was considered significant.

We used conditional logistic regression to examine case-control status in relation to vaccination status. Vaccination status was defined by the number of doses received (0, 2, or 3) with 2-dose vaccinees further categorized by whether time between the second dose and the encounter was 14 to 59, 60 to 149, or ≥150 days. The unvaccinated were the reference group. Each category of vaccinees was separately compared with the unvaccinated reference group using separate logistic regression models, each specified with a dichotomous indicator of vaccination status. Each model was conditioned on month and site such that cases were compared with controls tested during the same month at the same site. Logistic regression models were also adjusted for covariates specifying year of age, geographic region, and spline functions of calendar time, the time-and-area-specific percent of SARS-CoV-2 tests that were positive, and a propensity-to-be-vaccinated score. A separate propensity score for each level of vaccination versus none was estimated by generalized boosted regression based on demographic characteristics and underlying medical conditions.^27,28  Separate analyses were conducted in 3 time periods defined by the predominant variant (pre-Omicron [including Delta], Delta [only], and Omicron [including BA1 through BA.4/5]) and in 3 age groups (5–11, 12–15, and 16–17 years). Separate propensity scores were calculated for the analysis of each age group in each period.

VE was estimated by (1 minus the adjusted odds ratio) ×100%. VE estimates were considered too imprecise to report if there were no cases in the analysis, the total of cases plus controls was <20, or the 95% confidence interval (CI) was wider than 50 percentage points.

We conducted 4 sensitivity analyses by restricting the analytic population as follows: (1) SARS- CoV-2 testing on the same day as the hospital admission or ED/UC visit; (2) CLI definition limited only to codes listed in the primary diagnostic position; (3) CLI inclusion limited to clinical respiratory diagnoses (excluding asthma exacerbation); and (4) CLI inclusion limited to asthma exacerbation.

We used SAS (version 9.4; SAS Institute) and R (version 4.1.2; R Foundation for Statistical Computing).

There were 80 032 ED/UC CLI-associated encounters among children aged 5 to 17 years from April 9, 2021, to September 2022; 7294 (74%) of 9800 cases were unvaccinated, whereas 44 371 (63%) of 70 232 controls were unvaccinated (Table 1). Of the cases, 28% occurred during Delta and 71% during Omicron-predominant periods; 32% of the controls occurred during Delta and 67% during Omicron. Children in the analysis were predominantly non-Hispanic white (48%), Hispanic (23%), and non-Hispanic Black (13%), with similar proportions among the cases and controls.

Diagnoses most frequently recorded with CLI-associated ED/UC encounters were acute upper respiratory tract infections (56%), acute respiratory illness signs and symptoms (26%), and acute febrile illness signs and symptoms (20%), with similar proportions among cases and controls, except for influenza disease, where 3% of cases versus 11% of controls received this diagnosis (Supplemental Table 6).

Among 17 998 CLI-associated ED/UC encounters in adolescents aged 16 to 17 years, 1879 (69%) of 2728 cases were unvaccinated, whereas 7820 (51%) of 15 270 controls were unvaccinated (Table 1). During the Delta period, VE against COVID-19-associated ED/UC encounters was 93% (95% CI 86% to 97%) 14 to 59 days after dose 2 and decreased to 72% (63% to 79%) after ≥150 days (Fig 1). During Omicron, 2-dose VE estimates had confidences intervals which included the null, except for 60 to 149 days after dose 2 (VE = 31%, 10% to 47%) and ≥7 days following a monovalent booster dose (VE = 46%, 30% to 58%).

Among 30 321 CLI-associated ED/UC encounters in adolescents aged 12 to 15 years, 3319 (74%) of 4487 cases were unvaccinated, whereas 14 623 (57%) of 25 834 controls were unvaccinated (Table 1). During the Delta period, VE against COVID-19-associated ED/UC encounters was 93% (89% to 95%) 14 to 59 days after dose 2 and decreased to 77% (69% to 84%) after ≥150 days (Fig 1). During Omicron, VE 14 to 59 days after dose 2 was 64% (44% to 77%), decreased to 13% (3% to 23%) ≥150 days, and subsequently increased to 54% (40% to 65%) ≥7 days after a monovalent booster dose.

Among the 31 713 CLI-associated ED/UC encounters in children aged 5 to 11 years, 2096 (81%) of 2585 cases were unvaccinated, whereas 21 928 (75%) of 29 128 controls were unvaccinated (Table 1). VE against COVID-19-associated ED/UC encounters was 49% (33% to 61%) 14 to 59 days after 2 doses and 41% (29% to 51%) after ≥150 days (Fig 1). No boosters were administered during the study in this age group.

Among children aged 5 to 17 years, there were 2917 CLI-associated hospitalizations; 83% (n = 254) of 305 hospitalized cases were unvaccinated and 70% (n = 1822) of 2612 controls were unvaccinated (Table 2). Of cases, 48% occurred during Delta and 49% during Omicron-predominant periods, whereas of controls, 23% were during Delta and 74% during Omicron. Of hospitalized children, 1106 (38%) were non-Hispanic white, 763 (26%) were Hispanic, and 488 (17%) were non-Hispanic Black, with similar proportions among cases and controls.

Among those hospitalized, 61% of cases and 59% of controls had a chronic respiratory condition; 66% of cases and 55% of controls had a chronic nonrespiratory condition (Table 2). Diagnostic codes associated with hospitalization varied, with 47% of cases and 0.5% of controls receiving COVID-19 pneumonia codes, whereas 15% of case and 29% of control hospitalizations receiving acute asthma exacerbation codes (Supplemental Table 7).

Among 671 CLI-associated hospitalizations in adolescents aged 16 to 17 years, 97 (85%) of 114 cases were unvaccinated, whereas 326 (59%) of 557 controls were unvaccinated (Table 2). Among 989 CLI-associated hospitalizations in adolescents aged 12 to 15 years, 113 (85%) of 133 cases were unvaccinated, whereas 552 (65%) of 856 controls were unvaccinated. Among the 1257 CLI-associated hospitalizations in children aged 5 to 11 years, 44 (76%) of 58 positive cases were unvaccinated, whereas 944 (79%) of 1199 negative controls were unvaccinated.

During the Delta-predominant period, VE against COVID-19–associated hospitalizations in persons aged 16 to 17 years was 99% (90% to 100%) during 60 to 149 days after dose 2 and 98% (73% to 100%) after ≥150 days (Fig 2). Similarly, in adolescents aged 12 to 15 years VE was 98% (21% to 100%) 14 to 59 days after dose 2 and 97% (74% to 100%) 60 to 149 days after dose 2. VE estimates for other variant periods and intervals defined by time-since-vaccination or after the monovalent booster doses were too imprecise to report. VE estimates against hospitalizations among 5 to 11-year-olds were also imprecise and did not meet reporting criteria because of few hospitalized cases.

Among 2917 5 to 17-year-olds with CLI-associated hospitalization, 602 (21%) were admitted to the ICU, of whom 61 (10.1%) tested SARS-CoV-2 positive (Table 2). Of 61 cases in the ICU, 46 (75%) were unvaccinated; of 541 controls, 365 (68%) were unvaccinated (Table 2 and Supplemental Table 8).

Among 2917 5 to 17-year-olds with CLI-associated hospitalization, 25 (0.9%) died, 4 of whom tested SARS-CoV-2 positive (Table 2). All 4 cases who died were unvaccinated and 15 (71%) of 21 controls who died were unvaccinated. VE estimates against ICU admissions and deaths were too imprecise to report.

All 4 sensitivity analyses assessing ED/UC encounters among 5 to 17-year-olds yielded results similar to the primary analyses (Table 3).

In this multisite, test-negative case-control study of nonimmunocompromised children and adolescents aged 5 to 17 years, we found that BNT162b2 effectiveness against COVID-19– associated ED/UC encounters and hospitalizations varied by time since vaccination, predominant circulating SARS-CoV-2 variant, and by receipt of a monovalent booster dose.

VE against hospitalization was high (≥97%) in adolescents aged 12 to 17 years during Delta predominance. VE remained high beyond 150 days after dose 2 for 16 to 17-year-olds; we did not find evidence of waning protection against hospitalizations for this age group. Despite VISIONS’s large size, we were unable to estimate VE against hospitalizations during Omicron because of few COVID-19-associated hospitalizations among both vaccinated and unvaccinated persons. However, other studies demonstrated that 2 BNT162b2 doses protected children and adolescents against COVID-19 hospitalization during Omicron predominance, with 1 estimating 40% to 68% VE within a network of US pediatric hospitals²³  and another reporting 83% VE among children aged 5 to 11 years in Singapore.²⁹  Further, 83% of our SARS-CoV2 positive hospitalizations were among unvaccinated patients, which, together with other reports, provides evidence that BNT162b2 offers protection against severe disease in children and adolescents. However, increased vaccine uptake is essential to optimize protection in this population.

For ED/UC encounters, VE was highest in the Delta period, consistent with other studies.³⁰  During the Omicron-predominant period, VE was lower and waned more. These findings align with other pediatric studies demonstrating lower VE during Omicron,^21,22  including our prior report showing lower VE against COVID-19-associated ED/UC visits during the early Omicron period compared with Delta.²⁵  Our VE estimates were also lower than those reported by clinical trials conducted before Omicron variants emerged.^18–20  Omicron variants can escape neutralizing antibodies from BNT162b2 vaccination,³¹  suggesting that the lower VE during Omicron predominance in all age groups was because of Omicron-variant evasion of vaccine-induced immunity rather than age-specific differences in vaccine response.

Our results align with other studies showing waning vaccine protection over time in all ages, especially during Omicron predominance and against mild to moderate outcomes, including ED/UC visits.^{22,29,30,32–34}  Consistent with other studies,³²  our study found that protection against ED/UC visits increased following receipt of the monovalent booster. This finding highlights the benefit of booster doses, however studies from adults demonstrate that VE against ED/UC visits and hospitalizations also wane with increasing time since an mRNA monovalent booster dose.³³  The recently authorized bivalent boosters containing Omicron subvariants BA.4/BA.5 are expected to provide higher levels of protection during ongoing Omicron predominance.³⁵  Whether protection from bivalent vaccines will similarly wane is unknown. Continued monitoring of VE after bivalent boosters will be critical to inform need for future boosters and composition of next generation vaccines.

All vaccinated children and adolescents included in this study received BNT162b2 vaccine, although the authorized dosage differs by age (10mcg/dose for 5–11-year-olds and 30mcg per dose for ≥12 years). Such differences have raised questions regarding potential for lower VE or faster waning in children compared with adolescents and adults.³⁶  Although we could not measure VE in 5 to 11-year-olds before Omicron-predominance, VE against ED/UC encounters during Omicron was similar across age groups, suggesting that the 10mcg per dose in children provided similar protection to the 30mcg per dose in older individuals.

BNT162b2 VE shortly after a second or monovalent booster dose during Omicron was similar or lower than influenza VE during recent influenza seasons.^37–40  Even with modest VE, vaccinating children against COVID-19 could substantially reduce the burden of illness and strain on healthcare facilities, especially when multiple respiratory viruses circulate.⁴¹ 

This study had limitations. First, although this study was designed to limit cases to those with COVID-19-related illness, we likely included some cases identified by surveillance screening with asymptomatic SARS-CoV-2 infections that were not the cause of CLI symptoms. Second, we could not accurately measure prior SARS-CoV-2 infection, which likely was common during Omicron predominance.⁴²  Consequently, we could not ascertain to what degree we estimated VE in a population with high levels of immunity from prior infection, nor were VE estimates adjusted for potential confounding that would arise if the prevalence of prior infections differed between vaccinated and unvaccinated children. Third, comparisons of VE by age group must be made with care. COVID-19 vaccines became available to different age groups at varying times when different variants were predominant. Fourth, adolescents had more postvaccination follow-up during the Omicron period than earlier because of the timing of vaccine recommendations. Although we adjusted for time since dose and variant predominance, longer postvaccination follow-up may have reduced VE estimates for the Omicron period compared with earlier periods. Fifth, statistical power was limited in some analyses because of small numbers of cases. Sixth, only a modest amount of observation time was available after the monovalent booster dose. Seventh, although models were adjusted to balance the differences between unvaccinated and vaccinated persons, unmeasured and residual confounding, such as differences in mask use, might have biased VE estimates. Eighth, determination of periods of Delta and Omicron predominance was based on ecologic surveillance data; patient-level genomic data were not available. Finally, we were not able to assess VE by specific Omicron sublineage (ie, BA.1, BA.2/BA.2.12.1, and BA.4/BA.5), which may have obscured sublineage-specific differences.

In summary, this study found that BNT162b2 protected children and adolescents aged 5 to 17 against COVID-19-associated ED/UC encounters and hospitalizations. Protection against ED/UC visits was lower during Omicron predominance and waned over time but increased after receipt of a monovalent booster. Children and adolescents should remain up to date with recommended COVID-19 vaccinations to protect against severe disease. An Omicron-containing bivalent booster may optimize protection.

We would like to acknowledge the contributions of the VISION Network: Children’s Minnesota (Ernest Krause, BA); HealthPartners Institute (Inih Essien, OD, Sunita Thapa, MPH, and Sheryl Kane, BS); Baylor Scott & White Health (I-Chia Liao, MPH; Mufaddal Mamawala; MBBS; Deborah Hendricks, Jason Ettlinger, MA; Joel Blais, BTh; Elisa Priest, DrPH; Michael Smith, BS; Spencer Rose, BS; Natalie Settele, PA; Jennifer Thomas, MS; Muralidhar Jatla, MD; Madhava Beeram, MD; Javed Butler, MD and Alejandro Arroliga, MD); Westat (Akintunde Akinseye, MSPH, Elizabeth Bassett, BA, Bria Berry, MPH, Rebecca Birch, MPH, Kevin Cheng, BS, Sumathi Croos, BA, Jonathan Davis, PhD, Margaret Dunne, MSc, Rebecca Fink, MPH, Nina Hamburg, MBA, Alex Hughes, PhD, Jean Keller, MS, Salome Kiduko, MPH, Lindsey Kirshner, MPH, Magdalene Kish, BS, Victoria Lazariu, PhD, Yong Lee, BSEE, Matthew Levy, PhD, Vanessa Masick, MS, Thomas Mienk, MPA, Patrick Mitchell, ScD, Jean Opsomer, PhD, Sarah Reese, PhD, Weijia Ren, PhD, John Riddles, PhD, Anna Rukhlya, MA, Kristin Schrader, MA, Patricia Shifflett, MS, Talia Spark, PhD, Brenda Sun, MS, Duck-Hye Yang, PhD, and Yan Zhuang, PhD); School of Medicine, University of Colorado Anschutz Medical Campus, Health Data Compass (David Mayer, BS, Bryant Doyle, Briana Kille, PhD, and Catia Chavez, MPH); Regenstrief Institute (Ashley Wiensch, MPH, and Amy Hancock, MPA); and Center for Health Research, Kaiser Permanente Northwest (Padma Dandamudi, MPH).

CI

confidence interval

CLI

COVID-19-like illness

ED/UC

emergency department/urgent care

ICD

International Classification of Diseases, 9th and 10th Revision

SMD

standardized mean difference

VE

vaccine effectiveness