Influenza Vaccine Effectiveness National Immunization Conference 2009 Dallas, TX April 2, 2009 Anthony Fiore, MD, MPH CAPT, USPHS Influenza Division, CDC

Overview Cursory review of influenza epidemiology and vaccinology Key elements of vaccine effectiveness (VE) studies Recent VE studies and controversies Looking ahead and summary

Influenza Highly transmissible respiratory illness caused by influenza viruses Currently circulating strains Influenza A (H1N1) Influenza A (H3N2) Influenza B (two lineages, Victoria and Yamagata) Yearly winter epidemics in temperate climates (seasonal influenza) Illness cannot be distinguished from other acute respiratory illnesses (“influenza-like illness”) without laboratory testing Antigenic drift – point mutations and recombinations Diminished immunity among previously infected or immunized persons to “drifted” strains Requires updating trivalent vaccine yearly to match predicted strains that will be circulating Sporadic, unpredictable pandemics

Composition of Vaccines against Seasonal Influenza Three strains selected to make a trivalent vaccine based on global viral surveillance and availability of suitable vaccine candidate 1 representative of each influenza A subtype Influenza A(H1N1) Influenza A(H3N2) 1 of the two predominant influenza B lineages Selection decision precedes typical peak influenza season by 10-12 months Northern Hemisphere strains selected in February Southern hemisphere strains selected in September “New vaccine” (one or more new strains) nearly every season Three strains are selected to make a trivalent vaccine against seasonal influenza. Strains selected are based on global viral surveillance. The selection decision precedes typical peak influenza season by 10-12 months. Northern Hemisphere strains are selected in February for the upcoming influenza season, and Southern hemisphere strains are selected in September. The result is that there is usually at least one new strain in each vaccine each season.

Determinants of Immune Response to Influenza Vaccines Age and immune competence Elderly, infants and chronically ill generally lower antibody response Immunosenescence – deterioration in overall immune capacity with older age especially beyond 75 years Prior exposure to virus strains similar to those in vaccine (infection or vaccination) Virus – strains vary as to how robust immune responses will be How effective vaccination is going to be in preventing influenza is often predicted by using the antibody response to vaccination as an indicator. For example, the very young and the elderly generally have lower antibody responses. Persons with previous exposure to virus strains similar to those in the vaccine (either because of previous infection or previous vaccination) will usually have a better antibody response. The immune competence of the person being vaccinated will affect response. A larger amount of antigen in the vaccine can produce a better response, although a point of diminishing returns is often reached as antigen amounts are increased. Finally, the type of vaccine affects the immune response. Live attenuated vaccines might provide a better immune response, but this response is more difficult to measure because it requires measuring types of immune responses beyond those that can be measured using serologic samples. Whole virus inactivated vaccines can be more immunogenic, but are also more likely to cause adverse reactions to the vaccine.

Measuring Seasonal Influenza Vaccine Effectiveness: What to look for in VE studies Virus and illness variables Antigenic match (circulating viruses vs. vaccine) Intensity of season Types of viruses circulating (virulence:H3N2 > B > H1N1) Multiple vs. single season Study design Outcomes measured Influenza-like illness (ILI) Laboratory-confirmed influenza Modeling studies of Influenza-related hospitalizations and deaths based medical databases and viral surveillance Randomized controlled trial (placebo-controlled ideal) vs. observational study Setting and participants Age and risk group Controls Clearly effectiveness, by any measure, varies according to the age group, influenza risk group, and the degree that the vaccine strains match the circulating strains. One of the challenges of determining the effectiveness of seasonal influenza vaccination is how to choose an outcome to measure effectiveness. The variety of different outcomes used in the influenza literature to assess effectiveness has made comparison across studies and time very difficult. Some outcomes often used to assess effectiveness is quite non-specific, such as influenza-like illness or ILI, which is usually defined as acute onset of fever and either cough or sore throat. This syndrome can be caused by a variety of respiratory pathogens or even non-infectious etiologies. However, influenza vaccine is only going to prevent ILI caused by influenza virus. Therefore effectiveness estimates that use ILI as an outcome typically find the vaccine to have a relatively low effectiveness in preventing ILI. Effectiveness of influenza vaccine in preventing laboratory-confirmed influenza is typically much higher, but requires more expensive and carefully designed studies. Effectiveness estimates using severe outcomes such as hospitalization or death are usually done retrospectively, and typically require modeling techniques. These studies can substantially overestimate the effectiveness of vaccination in preventing these outcomes if generally healthier persons tend to be preferentially vaccinated.

Vaccine match issues: Example of influenza B virus lineages Two influenza B lineages have co-circulated over the past 10 years Yamagata-like Victoria-like Protection after vaccination with one lineage against other lineage is limited Co-circulation of both lineages over the past several years means some degree of mismatch between B virus represented in vaccine and circulating strains

Influenza B by lineage and season, United States, 1998-2008 Virologic surveillance data Season %B % B not represented in vaccine '98-99 23% 0% '99-00 1% '00-01 46% '01-02 13% 77% '02-03 43% '03-04 93% '04-05 25% 26% '05-06 19% 78% '06-07 21% 24% '07-08 29% 98%

Estimated mortality by influenza virus type/subtype, 1990-1999: Impact of variability in circulating strains Season A(H1N1) A(H3N2) B 1990-91 1,386 4435 11,235 1991-92 4,594 33,927 357 1992-93 822 14,465 12,067 1993-94 188 35,763 253 1994-95 389 24,475 4,473 1995-96 10,080 16,865 4,639 1996-97 0 40,131 7,803 1997-98 47 50,858 394 1998-99 203 39,514 6,076 Thompson WW, et al. Mortality associated with influenza and respiratory syncytial virus in the United States. JAMA 2003;289:179-86

Influenza Strains by type/subtype over time, USA, 2007-08 1000 H1N1 B H3N2 800 600 400 200 40 42 44 46 48 50 52 02 04 06 08 10 12 14 16 18 20 Oct Nov Dec Jan Feb Mar Apr 2007 2008

Virus Type/Subtype by Region 2007-08 %A/H1 %A/H3 %B

Ideal influenza VE studies: Take virus and season variability into consideration Multiple seasons Multiple sites Good virus surveillance Typing/subtyping Antigenic characterization Reality: Ideal rarely achieved, but important to keep principles in mind when thinking about generalizibility of any single study

Measuring Seasonal Influenza Vaccine Effectiveness: What to look for in VE studies Virus and illness variables Antigenic match (circulating viruses vs. vaccine) Intensity of season Types of viruses circulating (virulence:H3N2 > B > H1N1) Multiple vs. single season Study design Outcomes measured Influenza-like illness (ILI) Laboratory-confirmed influenza Modeling studies of Influenza-related hospitalizations and deaths based medical databases and viral surveillance Randomized controlled trial (placebo-controlled ideal) vs. observational study Setting and participants Age and risk group Controls Clearly effectiveness, by any measure, varies according to the age group, influenza risk group, and the degree that the vaccine strains match the circulating strains. One of the challenges of determining the effectiveness of seasonal influenza vaccination is how to choose an outcome to measure effectiveness. The variety of different outcomes used in the influenza literature to assess effectiveness has made comparison across studies and time very difficult. Some outcomes often used to assess effectiveness is quite non-specific, such as influenza-like illness or ILI, which is usually defined as acute onset of fever and either cough or sore throat. This syndrome can be caused by a variety of respiratory pathogens or even non-infectious etiologies. However, influenza vaccine is only going to prevent ILI caused by influenza virus. Therefore effectiveness estimates that use ILI as an outcome typically find the vaccine to have a relatively low effectiveness in preventing ILI. Effectiveness of influenza vaccine in preventing laboratory-confirmed influenza is typically much higher, but requires more expensive and carefully designed studies. Effectiveness estimates using severe outcomes such as hospitalization or death are usually done retrospectively, and typically require modeling techniques. These studies can substantially overestimate the effectiveness of vaccination in preventing these outcomes if generally healthier persons tend to be preferentially vaccinated.

Selection of Influenza Vaccine Effectiveness Outcomes Outcomes based on symptoms or syndrome without laboratory confirmation Less specific outcome than laboratory confirmed Lowers vaccine effectiveness estimate Affected by level co-circulating respiratory pathogens, i.e. what proportion of influenza-like illness cases are actually influenza? Examples of outcomes: Influenza-like illness (e.g. fever + [cough or sore throat]) Influenza-related hospitalization or death based on ICD-9 codes and virus surveillance

Effect Of Co-circulation Of Non-influenza Pathogens/Outcome Specificity On VE Estimate Assuming 100 vaccinated and 100 unvaccinated in each set (A and B): VE against influenza infection (red) = 75% for both sets A and B, VE against all respiratory illness (yellow+ red) = 30% in set A and 15% in set B.

Influenza-Associated Hospitalizations By Age Group, 1979-2001 Data for this slide comes from Dr. Bill Thompson at CDC and illustrate the substantial number of influenza-related hospitalizations each year. The highest rates occur among persons aged 65 years and older followed by children <5 years of age and then adults 50-64 years. Among young children, risk of severe outcomes drops steeply after age 1, and the risk is reduced 5 to tenfold by age 5 0-4 Yrs 5-49 Yrs 50-64 Yrs >65 Yrs Thompson, et al. JAMA 2004. Hospitalizations modeled using viral surveillance and National Hospital Discharge Survey data

Selection of Influenza Vaccine Effectiveness Outcomes Outcomes based on symptoms or syndrome without laboratory confirmation Less specific outcome than laboratory confirmed Lowers vaccine effectiveness estimate Affected by level co-circulating respiratory pathogens, i.e. what proportion of influenza-like illness cases are actually influenza? Examples of outcomes: Influenza-like illness (e.g. fever + [cough or sore throat]) Influenza-related hospitalization or death based on ICD-9 codes Laboratory confirmed influenza Most specific because ILI can caused by many pathogens Potential pitfall: Bias in who may be chosen for testing

Issues in Vaccine Effectiveness among Older Persons

Controlling for differences between vaccinees and unvaccinated persons Major issue with observational VE studies among elderly that use medical diagnostic databases “Confounding by indication” Greater representation in vaccinated group of persons with chronic medical conditions “Healthy vaccinee effect” Greater representation in vaccinated group of persons who might less likely to have outcome studied “Frailty bias” Greater representation in unvaccinated group of persons with diminished functionality for activities of daily living who are more likely to die or be hospitalized during the observation period Analysis methods to control for confounding Statistical control using stratified or multivariable model analyses Subclassifying or matching patients on propensity to seek care Frailty bias is difficult to control for when using medical databases Suggested reading: Nichol N Engl J Med 2007; Simonsen Lancet Infectious Diseases 2007

Recent re-analysis of only published randomized, controlled trial of influenza vaccine in persons 60 years and older* *Thijs et al. Mortality benefits of influenza vaccination in elderly people (letter). Lancet 2008. Reanalysis of Govaert et al. JAMA 1994 (Note: 27% of participants with chronic illness).

ACIP’s Influenza Working Group Statement Supports prospective, population-based studies with laboratory-confirmed endpoints to monitor influenza vaccine effectiveness Encourages efforts to increase vaccine immunogenicity in frail elderly population Plans to review studies of new adjuvanted vaccines, novel delivery methods and alternative doses and schedules *Slide presented by K Neuzil, Chair, ACIP Influenza Vaccine Working Group ACIP meeting, October 23, 2008

Vaccine Effectiveness Studies among Younger Persons

Randomized placebo-controlled trials among young adults: Efficacy of TIV against lab-confirmed* influenza, MI Season Placebo (%+) TIV (%+) VE (95%CI) 2004-051 7.8 1.9 75% (42-90) 2005-062 1.8 1.5 16% (10 to -171) 2007-083 10.8 3.4 68% (46 to 81) *Culture and/or PCR 1Ohmit et al. N Engl J Med 2006 2Ohmit et al. Clin Infect Dis 2008 3Ohmit et al. IDSA: Seasonal and pandemic influenza 2009. Available at: http://www.idsociety.org/WorkArea/showcontent.aspx?id=13422

Interim Within-Season Estimate of VE: Marshfield, Wisconsin, 2007-08 Influenza Season Objective: Estimate VE for preventing medically attended acute respiratory illness (MAARI) which is lab-confirmed as influenza Patients living in a 14 postal-code area surrounding the Marshfield Clinic eligible for the study Enrolled subjects: Cases: MAARI with influenza infection diagnosed by RT-PCR Controls: MAARI negative by RT-PCR

Interim Within-Season Estimate of VE: Marshfield, Wisconsin, 2007-08 Influenza Season: Conclusions Suboptimal match (based on antigenic characterization) between 2 of 3 vaccine strains and viruses isolated from study participants Patient group /influenza type VE (95% CI) All enrollees 44% (11 to 65) ACIP recommended 34% (-34 to 67) Health 5-49 year olds 54% (12 to 76) Influenza A, all enrollees 58% (28-76) Influenza B, all enrollees -35% (-172 to 33) MMWR 2008;57:393-8

Effectiveness of Maternal Influenza Immunization in Mothers and Infants* Study participants and design Bangladesh, 2004-05 Randomized controlled trial 340 pregnant women received either influenza vaccine or pneumococcal polysaccharide vaccine (control) during 3rd trimester Follow-up through pregnancy and first 6 months after birth Outcomes Febrile respiratory illness among infants Lab-confirmed influenza among infants Febrile respiratory illness among mothers The next 3 slides summarizes results form a study conducted in Bangladesh in 2004-2005, with the results published last month in the New England Journal of Medicine. * K Zaman et al. N Engl J Med 2008

Cumulative Cases of Lab-Confirmed Influenza among Infants, by Receipt of Influenza vaccine, Zaman et al, Bangladesh, 2004-05 We don’t know how generalizable these results are for the United States, and additional studies need to be done to better understand the benefits of influenza vaccination for the infants born to vaccinated women. But these are powerful results, and should be of interest to clinicians who provide care for pregnant women.

Summary of common VE study methodologies Randomized controlled studies Gold standard for controlling for confounding Might not be possible in many age/risk groups already recommended for routine vaccination Persons who volunteer for RCT are not normal Observational studies – administrative databases Might better capture a representative population Confounding issues unresolved, particularly for studies among the elderly Observational studies – using test negative controls with acute respiratory illness Not certain that test negative controls are representative of source populations

CDC Plans for Vaccine Assessments 4 U.S. sites funded for 3 years to determine VE for lab-confirmed medically attended outcomes Include all groups recommended for vaccination Controls: Persons with acute respiratory illness who are influenza test-negative Powered to assess VE for hospitalization Potential for within-season VE estimates Emerging Infections Program study of VE against hospitalization with lab-confirmed influenza Controls – persons who were not hospitalized, matched on age and county Two international sites (locations in Senegal and India) funded to measure VE and indirect effects

Issues on the Near Horizon Replication of intriguing findings Intradermal vaccination VE among infants born to women vaccinated while pregnant Better methods to reduce confounding in observational studies among elderly? Ongoing discussions among persons on both sides of debate

Issues on the Near Horizon More effective vaccines? Large RCTs of adjuvanted vs. standard vaccine in elderly are in progress FDA and manufacturers discussing a quadrivalent (2 A and 2 B strains) vaccine for United States Better ability to predict circulating strains? Increased, more representative virus surveillance globally Molecular epidemiology and geographic spread of drifted strains coming into focus Better understanding of critical antigenic properties