2 tables omitted
During clinical trials, the efficacy of vaccination with inactivated influenza vaccines for the prevention of serologically confirmed influenza infection has been estimated as high as 70%-90% among healthier adults. However, the effectiveness of annual influenza vaccination typically is lower during those influenza seasons when a suboptimal match between the vaccine strains and circulating influenza strains is observed. For example, in a 4-year randomized study of influenza vaccine among healthy persons aged 1-65 years, the predominant strain was drifted from the vaccine strain in 2 of the 4 years. Inactivated vaccine effectiveness (VE) against culture-confirmed influenza ranged from 71% to 79% when the vaccine and circulating strains were suboptimally matched to 74% to 79% when the matches were well matched.1 In contrast, a 2-year study of inactivated influenza vaccine among healthy adults aged 18-64 years found no measurable VE during a year when a poorly matched strain circulated, but found VE of 86% against laboratory-confirmed influenza during the following year when the vaccine and circulating strains were well matched.2 Although laboratory data on the antigenic characteristics of circulating influenza viruses compared with vaccine strains are available during influenza seasons, estimates of VE usually have not been made until months after the conclusion of the season. This report summarizes interim results of a 2008 case-control study to estimate the effectiveness of trivalent inactivated influenza vaccine for prevention of medically attended, laboratory-confirmed influenza during the 2007-08 influenza season, when most circulating influenza A (H3N2) and B viruses were suboptimally matched to the vaccine strains. Despite the suboptimal match between two of three vaccine strains and circulating influenza strains, overall VE in the study population during January 21–February 8, 2008, was 44%. These findings demonstrate that, in any season, assessment of the clinical effectiveness of influenza vaccines cannot be determined solely by laboratory evaluation of the degree of antigenic match between vaccine and circulation strains.
Patients living in a 14 postal-code area surrounding Marshfield, Wisconsin, were eligible to participate in this study. Nearly all residents in this area receive outpatient and inpatient care from Marshfield Clinic health-care providers. Study enrollment began on January 21, 2008, based on laboratory evidence of influenza circulation from both Marshfield Clinic laboratories and the Wisconsin State Laboratory of Hygiene and continued through March 28, 2008. Patients who visited a Marshfield Clinic facility with medically attended illnesses were screened for study eligibility during outpatient or inpatient visits. Patients who reported feverishness, chills, or cough were eligible for enrollment. Those who reported symptoms for 8 or more days were not eligible for enrollment because influenza virus shedding decreases with illness duration, making detection of the virus unlikely after 8 days of symptoms. The majority of ill patients not approached during a clinical encounter were identified the next day by using electronic diagnosis codes entered by the clinician; these patients were contacted by telephone and enrolled at home if they met eligibility criteria. The Marshfield Clinic Research Foundation institutional review board approved this study.
Nasal or nasopharyngeal swabs were obtained from consenting patients and were tested for influenza A or B infection by reverse transcription–polymerase chain reaction (RT-PCR) at the Marshfield Clinic Research Foundation using CDC-recommended probes and primers. Viral culture was performed on all samples that were RT-PCR positive to provide virus isolates for antigenic characterization. Influenza vaccination status was determined through an immunization information system (Regional Early Childhood Immunization Network*) used by all public and private immunization providers for vaccines administered to adults and children. Previous validations have demonstrated that the system captures 96%-98% of influenza vaccines administered to area residents (Marshfield Clinic Research Foundation, unpublished data, 2005-2007). Trivalent inactivated influenza vaccine from Sanofi-Pasteur ([Fluzone®], Swiftwater, Pennsylvania) was the only influenza vaccine used by Marshfield Clinic during the 2007-08 influenza season.
For this case-control study, a case of medically attended influenza was defined as an acute illness in a patient with feverishness, chills, or cough and documentation of influenza infection by RT-PCR. Controls were defined as patients with the same symptoms who had a negative RT-PCR test for influenza. Using persons with acute respiratory symptoms who test negative for influenza as controls is a method that in modeling studies has compared favorably with cohort studies and traditional case-control designs for the assessment of vaccine effectiveness.3 Patients were categorized as immunized if they had received influenza vaccine 14 days or more before enrollment; in addition, children aged <9 years were categorized as immunized if they had received 2 doses of influenza vaccine. Twenty-three children were excluded because they had received only 1 of the 2 recommended doses; this subgroup was too small to permit a separate analysis of VE for partial immunization.
VE was estimated by using logistic regression to compare patients with laboratory-confirmed influenza with patients who tested negative for influenza. The likelihood of receiving influenza vaccination in this population is associated with a propensity to seek health care, and use of test-negative controls helped adjust for this source of bias by estimating VE for medically attended influenza illness. Comparisons of this study design to traditional cohort and case-control methods for assessing VE have been published recently.3 For this analysis, the enrolled patients were categorized into two groups: persons for whom influenza vaccine was recommended by the Advisory Committee on Immunization Practices (ACIP) for the 2007-08 season based on age or an existing chronic medical condition† that increased their risk for influenza-related complications (i.e., the ACIP recommended group), and healthy children and adults aged 5-49 years (i.e., the healthy group).
Logistic regression models were adjusted for age, week of enrollment, and presence of a chronic medical condition. The last variable was not included in the models restricted to healthy patients aged 5-49 years. VE was estimated as 100 × [1 – adjusted odds ratio]) and was interpreted as zero if the percentage was negative. The first 59 influenza virus isolates obtained during the study were submitted to CDC for detailed antigenic characterization.
During January 21–February 8, 2008, a total of 1,779 patients were assessed for study eligibility after a clinical encounter for acute respiratory illness or febrile illness. A total of 850 (48%) did not meet eligibility criteria; 773 (91%) of exclusions resulted from absence of feverishness, chills, or cough or an illness duration 8 days or longer. Of the 929 eligible patients, 639 (69%) consented to the study and were tested for influenza infection. Final enrollment for this interim analysis was reduced to 616 patients after exclusion of 23 partially immunized children who had received only 1 of 2 recommended vaccine doses.
Influenza was detected by RT-PCR in 191 (31%) enrollees; 75% of influenza infections were type A. Distribution by sex was similar for patients who tested positive and patients who tested negative for influenza; however, the median age was higher for patients who tested positive (21 years) than those who tested negative (10 years). Approximately 19% of patients who tested positive and 39% of those who tested negative had been vaccinated against influenza.
The overall interim estimate of VE was 44%; the estimate was higher among persons in the healthy group aged 5-49 years (54%). The overall estimate of VE for prevention of medically attended influenza A infections was 58%. No VE was observed for prevention of medically attended influenza B infections.
Subtyping by RT-PCR performed at CDC demonstrated that 40 of 41 influenza A specimens were influenza A (H3N2) viruses; the remaining specimen was an H3N2 and B virus mixture. Preliminary data on antigenic characterizations were available for nine influenza A (H3N2) viruses and 18 of 20 influenza B viruses. Two of nine influenza A (H3N2) viruses were A/Wisconsin/67/2005-like, the H3N2 component of the 2007-08 Northern Hemisphere vaccine; the other seven were A/Brisbane/10/2007-like (H3N2) viruses, a strain that is drifted from the A/Wisconsin/76/2005 strain. All 18 influenza B viruses were B/Florida/04/2006-like, belonging to the B/Yamagata/16/88 lineage of viruses. B/Yamagata-like viruses are antigenically distinct from the B/Victoria-like lineage virus that was included in the 2007-08 influenza vaccine.
E Belongia, MD, B Kieke, L Coleman, PhD, J Donahue, DVM, PhD, S Irving, J Meece, PhD, M Vandermause, Marshfield Clinic Research Foundation, Marshfield, Wisconsin. D Shay, MD, P Gargiullo, PhD, A Balish, A Foust, MA, L Guo, MD, S Lindstrom, PhD, X Xu, MD, A Klimov, PhD, J Bresee, MD, N Cox, PhD Influenza Div, National Center for Immunization and Respiratory Disease, CDC.
Influenza infections result in substantial morbidity and mortality each year in the United States.4,5 Because of the sizeable burden of influenza-associated disease, annual influenza vaccination was recommended by ACIP for the 2007-08 season for children aged 6-59 months, adults aged ≥50 years, persons with chronic medical conditions that place them at high risk for serious influenza-related complications, and close contacts of these groups and of children aged <6 months.6
Viral data reported to World Health Organization (WHO) and National Respiratory and Enteric Virus Surveillance System (NREVSS) laboratories in the United States during the 2007-08 influenza season through April 5, 2008, demonstrated that influenza A and B viruses accounted for 74% and 26%, respectively, of influenza viruses characterized in the United States.7 Of influenza A viruses subtyped, 27% were influenza A (H1N1) viruses, and 73% were influenza A (H3N2) viruses. Antigenic characterization of a subset of these viruses by CDC indicated that 69% of A (H1N1) viruses were A/Solomon Islands/3/2006-like, the A (H1N1) vaccine component, but that 71% of A (H3N2) viruses were A/Brisbane/10/2007-like, a recent antigenic variant of the A/Wisconsin/67/2005-like virus, the A (H3N2) vaccine component. In addition, 95% of antigenically characterized B viruses belonged to the B/Yamagata lineage. Viruses in this lineage are antigenically distinct from the B/Malaysia/2506/2004-like component of the 2007-08 vaccine, which is in the B/Victoria lineage. These viral surveillance data suggested that the effectiveness of the 2007-08 influenza vaccine might be reduced against circulating influenza A (H3N2) and B viruses. However, in this analysis, preliminary VE results indicated that, despite the antigenic differences between vaccine and circulating H3N2 strains, the effectiveness of vaccine in preventing medically attended respiratory illnesses from influenza A infections was 58%. In contrast, no VE could be demonstrated against influenza B.
Multiple previous studies of the effectiveness of influenza vaccines have been reported (i.e., observational studies of the clinical effects of vaccination as opposed to randomized clinical trials).8 VE varies from influenza season to season, based in part on the degree of antigenic match between vaccine and circulating influenza strains. VE previously has been assessed sporadically in different populations and by using different methods. Annual systematic assessments of VE using laboratory-confirmed outcomes have not been available within an influenza season. Furthermore, antigenic characterization data rarely have been available for influenza viruses isolated from participants of VE studies, and not previously from the population for whom annual vaccination is recommended by ACIP. Despite a mismatch between the vaccine influenza A (H3N2) strain and seven of nine influenza A (H3N2) viruses isolated from study participants, the data in this report are consistent with results obtained in seasons with a moderate antigenic mismatch between vaccine and circulating strains of H3N2 viruses.1,8
Based on preliminary analyses of A/Brisbane/10/2007-like (H3N2) viruses and the 2007-08 vaccine H3N2 strain using the method of antigenic mapping,9 an average fourfold difference was observed between the homologous titer for the vaccine strain and average titers for circulating strains. These differences were measured with hemagglutination inhibition tests by using a panel of reference postinfection ferret antisera. The degree of mismatch between the A/Wisconsin/67/2005 vaccine strain and H3N2 viruses tested at CDC thus far during the U.S. 2007-08 influenza season can be described as moderate in relation to antigenic distances seen over time for H3N2 viruses.10 By contrast, all the influenza B viruses isolated in the Marshfield Clinic study this season and antigenically characterized thus far belong to the B lineage not contained in this season's vaccine. Viruses from the B/Victoria-like lineage and B/Yamagata-like lineage are substantially more antigenically distinct from each other than A/Wisconsin/67/2005-like and A/Brisbane/10/2007-like H3N2 viruses are from each other.
The findings in this report are subject to at least four limitations. First, analyses were conducted while enrollment and laboratory testing were ongoing, and not all RT-PCR positive samples had yet been confirmed by culture. Thus, the preliminary subtype distribution and antigenic characterization results might not be representative of all patients in the study with influenza. Second, VE was estimated only for prevention of influenza among persons who sought care for acute respiratory illness, comparing patients who tested positive for influenza with patients who tested negative. Certain patients who tested negative for influenza might actually have had influenza virus infections, although RT-PCR is the most sensitive diagnostic test available. In addition, although simulation models have demonstrated that VE estimated with test-negative controls was close to the actual VE when test specificity was high, as is also the case with RT-PCR,3 this method is only beginning to be used in studies. VE was assessed against medically attended influenza and not against more severe outcomes of influenza infection, such as influenza hospitalizations; VE might vary with severity of the outcome studied. Third, if the antigenic characteristics of influenza viruses circulating in other regions of the United States differ substantially from viruses isolated from the Marshfield, Wisconsin, study participants, VE might vary by region. Finally, enrollment of patients continued in this study thorough March 28, and final analyses might differ from these interim assessments of VE.
These preliminary data based on study enrollment during January 21–February 8 suggest several conclusions. First, when assessing VE, laboratory data on antigenic characterization of circulating influenza viruses compared with vaccine strains should be interpreted together with data on the clinical effectiveness of vaccination in preventing laboratory-confirmed influenza illnesses. Although two of three vaccine strains were not optimally matched with circulating viruses this season, an interim VE estimate suggests that vaccination provided substantial protection against medically attended acute respiratory illness in this study population. In addition, intraseason estimates of VE, such as those from this analysis, might be useful to public health authorities and medical practitioners in their communications about the benefits of vaccination, especially late in the influenza season. Such data also might be helpful to practitioners when evaluating the need for antiviral treatment and prophylaxis for their patients. Therefore, creating systems that enable collection and dissemination of timely VE data during an influenza season are a priority for CDC. Finally, health-care providers should be aware of the types and subtypes of influenza circulating in their communities over the course of each influenza season. If influenza B strains predominate during the remainder of this season, providers can anticipate an increased risk for vaccine failures and should consider early use of antiviral medications for treatment and prophylaxis of persons at high risk for complications from influenza infection.
The findings in this report are based, in part, on contributions from V Allison, J Anderson, E Bergmann, C Beyer, L Bennetti, N Berger, C Becker, A Bernitt, A Brockman, K Buedding, D Cole, A Deedon, J Frahmann, D Gamble, L Gavigan, D Gentz, G Greenwald, N Hartl, J Herr, D Hilgemann, L Ivacic, D Johnson, D Kempf, T Kronenwetter-Koeppel, D Marx, C Meyer, C Reis, S Reisner, J Salzwedel, S Strey, P Siegler, P Stockwell, L Verhagen, D York, J Zygarlicke, Marshfield Clinic Research Foundation, Marshfield, Wisconsin.
*Available at http://www.recin.org/default.asp.
†Defined as existing if the patient had two or more health-care visits with relevant International Classification of Diseases, Ninth Revision, Clinical Modification diagnosis codes during 2007. Diagnosis codes were based on ACIP criteria, including cardiac, pulmonary, renal, neurological/musculoskeletal, metabolic, cerebrovascular, immunosuppressive, circulatory system, and liver disorders; diabetes mellitus; and malignancies.
Interim Within-Season Estimate of the Effectiveness of Trivalent Inactivated Influenza Vaccine—Marshfield, Wisconsin, 2007-08 Influenza Season. JAMA. 2008;299(20):2381-2384. doi:10.1001/jama.299.20.2381