VACCINE ACTIVE SURVEILLANCE I Caitlin N Dodd, MS Biostatistician, Global Child Health Cincinnati Children s Hospital Medical Center
CONTENTS M1 Compare epidemiological study methods Discuss the benefits of self controlled case series (SCCS) study methodology Discuss the application of SCCS
Slide 2 M1 These are really placeholders. Please change the Contents to suit your purpose. We will recommend that you use max of 30slides for this 60mins talk. MSH, 03/23/2012
SECTION ONE: COMPARISON OF STUDY METHODS Outline: Case Control Studies Cohort Studies
CASE CONTROL STUDIES Subjects who have experienced an event of interest or have been diagnosed with the condition of interest are matched to non diagnosed controls Popular because they are relatively inexpensive and can be completed quickly Prohibitive in that controls should ideally be matched across multiple characteristics
CASE CONTROL STUDIES Latent confounders cannot be controlled for by matching Case control studies can demonstrate association but do not demonstrate causation Vaccinated Unvaccinated Adverse a b event No AE c d Total a+c b+d
CASE CONTROL STUDIES Advantages Permit the study of rare diseases. Permit the study of diseases with long latency between exposure and manifestation. Can be launched and conducted over relatively short time periods. Relatively inexpensive as compared to cohort studies. Can study multiple potential causes of disease.
CASE CONTROL STUDIES Disadvantages Information on exposure and past history is primarily based on interview and may be subject to recall bias. Validation of information on exposure is difficult, or incomplete, or even impossible. By definition, concerned with one disease only. Cannot usually provide information on incidence rates of disease. Generally incomplete control of extraneous variables. Choice of appropriate control group may be difficult. Methodology may be hard to comprehend for non epidemiologists and correct interpretation of results may be difficult.
CASE CONTROL STUDIES VS. SCCS Self controlled case series can be thought of as a case control study in which subjects act as their own controls. Because of the underlying time scale (whether age or calendar time based), the self controlled case series is a more flexible methodology. Relative incidence can be estimated by season, exposure, or subject age
CASE CONTROL STUDIES VS. SCCS Exposed populations may be self selecting which may lead to inherent bias in a case control study The self controlled case series method is more computationally intensive and less intensive in terms of data collection.
CASE CONTROL STUDIES VS. SCCS Case control study provides relative risk = (a/a+c) / (b/b+d) SCCS provides an estimate of relative incidence associated with exposure using only a. Relative incidence associated with time varying covariates (season, age, etc.) are estimated using a and b.
COHORT STUDIES A cohort study follows a group of healthy individuals over time and then compares those subjects who experience the event of interest with those who do not experience the event of interest Analyzed by calculating incidence rate ratios for exposed and non exposed subjects, then using these to determine relative risk.
COHORT STUDIES Advantages: Allow complete information on the subject s exposure, including quality control of data, and experience thereafter. Provide a clear temporal sequence of exposure and disease. Give an opportunity to study multiple outcomes related to a specific exposure. Permit calculation of incidence rates (absolute risk) as well as relative risk. Methodology and results are easily understood by nonepidemiologists. Enable the study of relatively rare exposures.
COHORT STUDIES Disadvantages: Not suited for the study of extremely rare diseases because a large number of subjects is required. Not suitable when the time between exposure and disease manifestation is very long, although this can be overcome in historical cohort studies. Exposure patterns, for example the composition of oral contraceptives, may change during the course of the study and make the results irrelevant. Maintaining high rates of follow up can be difficult. Expensive to carry out because a large number of subjects is usually required. Baseline data may be sparse because the large number of subjects does not allow for long interviews.
COHORT STUDIES VS. SCCS In a SCCS study, exposure times are considered to be fixed and event times are random. The opposite is the case in a cohort study. Cohort studies estimate relative risk while SCCS studies estimate relative incidence.
COHORT STUDIES VS. SCCS Both allow for age and time effects Efficiency is lower in the SCCS because, by conditioning on the number of events, marginal information is lost for each individual Self controlled case series studies require far less time in patient follow up than cohort studies.
ADVANTAGES OF THE SELF-CONTROLLED CASE SERIES METHOD Based on cases only Provides consistent estimates of relative incidence Controls for fixed confounders Age and temporal variation in baseline can be allowed for Can have high efficiency relative to the retrospective cohort method.
LIMITATIONS OF THE SELF-CONTROLLED CASE SERIES METHOD It requires that the probability of exposure is not affected by the occurrence of an outcome event. Risk must be small for non recurrent events Only produces estimates of relative (not absolute) incidence Requires variability in the time or age of the event
APPLICATIONS Has typically been applied to vaccine safety questions Also: SSRIs and falls Long haul flights and Thromboembolism Beta blockers and heart failure
SECTION TWO: EXAMPLE OF STUDY USING SCCS Does oral polio vaccine cause intussusception in infants? Evidence from a sequence of three self controlled cases series studies in The United Kingdom Nick Andrews, et al.
NICK ANDREWS, ORAL POLIO VACCINE AND INTUSSUSCEPTION, 2001 Three analyses on 3 sets of linked hospital admissions data Analysis 1: exploratory, risk periods defined as days 0 13 after each of three doses Analysis 2: exploratory, risk periods 14 41 days after each dose Analysis 3: formal hypothesis testing of increased risk in days 14 27 following each dose
NICK ANDREWS, ORAL POLIO VACCINE AND INTUSSUSCEPTION, 2001 Methods Collected database data on all children age 28 to 365 days with an ICD code of intussusception For each child, date of birth, sex, postal code, hospital, hospital number, immunizations, dates of immunizations, and age in days at time of admission were collected. Validation of diagnosis was conducted on a subset of subjects
NICK ANDREWS, ORAL POLIO VACCINE AND INTUSSUSCEPTION, 2001 Statistical Analysis Variables used in the analysis: DOB, dates of vaccinations, dates of intussusceptions, region, and sex. Subjects with less than 3 recorded doses were excluded as were subjects with less than 21 days between doses Analyzed using the self controlled case series method, stratified by age for three data sets (HE1: January 1991 March 1997, HE2: April 1997 June1999, and GRPD: from the General Practice Research Database).
NICK ANDREWS, ORAL POLIO VACCINE AND INTUSSUSCEPTION, 2001 Descriptive statistics on 3 data sets
NICK ANDREWS, ORAL POLIO VACCINE AND INTUSSUSCEPTION, 2001 Results No evidence of increased relative incidence in the 0 13 days following vaccination Increased risk of intussusception in the 14 27 days following the third dose of OPV (RI = 2.15, p = 0.0082) No increased risk when 14 days pre vaccination and periods after dose 1 and 2 included in the background risk in HE2 and GRPD data (RI = 1.03, 95% CI = 0.64 1.67)
NICK ANDREWS, ORAL POLIO VACCINE AND INTUSSUSCEPTION, 2001 Results Total cases in the risk periods with relative incidence (95% CI) by dose for the three data sets
NICK ANDREWS, ORAL POLIO VACCINE AND INTUSSUSCEPTION, 2001 Results Total cases in the risk periods with relative incidence (95% CI) for dose 3 with a pre vaccination low period:
NICK ANDREWS, ORAL POLIO VACCINE AND INTUSSUSCEPTION, 2001 Interpretation and Conclusions
SECTION THREE: APPLYING THE SCCS Outline: Events and exposures appropriate to the selfcontrolled case series Determining potential exposure, seasonal, and ageassociated risk windows Power and sample size
EVENTS AND EXPOSURES PAST STUDIES MMR vaccine and autism MMR vaccine and meningitis Influenza vaccine and asthma MMR vaccine and Idiopathic thrombocytopenic purpura OPV and intussusception Rotavirus vaccine and intussusception Infant vaccines and lower respiratory disease MMR vaccine and bacterial infections SRI antidepressants and hip fracture Air travel and venous thromboembolism Influenza vaccine and Bell s Palsy MMR vaccine and gait disturbance SRI antidepressants and myocardial infarction Nicotine replacement therapy and myocardial infarction, stroke, or death Bupropian and sudden death Acute infection and pulmonary embolism Dialysis for diabetes treatment and foot ulceration Influenza vaccine and Guillain Barre Syndrome MMR vaccination and convulsions Hepatitis B vaccine and demyelinating events Oral Bisphosphonates and Atrial fibrillation Thiazolidinediones and Fractures Prescription Medications and Motor vehicle crashes
EVENTS AND EXPOSURES Questions to consider What is a recurrence and what is a new event? How many doses are given? Is this a transient or chronic exposure? What determines the date of exposure? Of an event? Can cases be reliably verified? Is exposure history available? In what form? What is the risk period following exposure?
EVENTS AND EXPOSURES Risk periods: Chosen based on prior hypotheses, previous studies, and presumed biological mechanisms. If it is too long, too short or is placed so that it does not cover the true risk period then the relative incidence estimate may be biased toward the null
EXPOSURE, SEASON, AND AGE ASSOCIATED RISK Questions to consider Are exposure, event, or both seasonal? Does data exist to answer these questions? Do subjects represent an age group or specific range of ages? Is age data available? How many exposures? Will all subjects have the same number of exposures? Is risk after each exposure expected to be the same?
POWER AND SAMPLE SIZE An Example:
MODELING WITH THE CASE SERIES METHOD Example using the global GBS/H1N1 study Assessing risk of Guillain Barre Syndrome following vaccination with pandemic influenza vaccine Includes data from 15 countries Based upon increased risk seen in the 1976 1977 swine flu pandemic
MODELING WITH THE CASE SERIES METHOD Multiple risk periods Multiple risk periods can be defined and the relative incidence in these periods can be estimated using the case series method For the Global GBS/H1N1 project, these periods are days 1 6, 7 21, and 22 42 following vaccination Each of these periods is compared to the control period, meaning days during observation before exposure and days during observation following day 42 post vaccine.
MODELING WITH THE CASE SERIES METHOD Multiple events Requires a within individual independence assumption Total number of events within each time period are analyzed For the GBS study, multiple events are not analyzed. Additional GBS diagnoses within 6 months of the initial diagnosis are considered a relapse and only the first diagnosis is analyzed.
MODELING WITH THE CASE SERIES METHOD Unique and non independent events Poisson assumption fails when events are unique (as in the case of GBS). When the baseline rate for the event approaches 0, the case series method is valid Background rate of GBS is 10 to 20 cases per million population Poisson assumption fails when occurrence of one event increases the probability of subsequent events (recurrent GBS events) If the initiating event is rare, only the first event should be used Did not see recurrent GBS in the global data set
MODELING WITH THE CASE SERIES METHOD Event dependent exposures Most restrictive assumption of the case series method is that the occurrence of the event does not alter subsequent exposures Violated in the case of GBS and influenza vaccine Three approaches: Redefine the observation period to begin at the time of exposure and only include exposed individuals. Removing a pre exposure time period from the baseline The pseudo likelihood method
MODELING WITH THE CASE SERIES METHOD Covariates and interactions Although fixed covariates are controlled, covariates may be effect modifiers. GBS is more common in males and in older subjects Interactions with sex and age group will be included in the analysis of global data A likelihood ratio test can be conducted to determine if the interaction should be included in the model.
MODELING WITH THE CASE SERIES METHOD Multiple exposures (of the same vaccine) Multiple exposures (such as doses of childhood vaccines) can be modeled, each with a risk period following exposure. The estimates are relatively unchanged if the time between exposures is longer than the risk period for each exposure. This is allowed for in the analysis of GBS and H1N1 vaccine, although few subjects received 2 doses of pandemic vaccine. Risk windows do not overlap
MODELING WITH THE CASE SERIES METHOD Multiple exposures (of different vaccines) Multiple exposures (such as a seasonal and pandemic vaccine) can be modeled, each with a risk period following exposure. Risk windows in this situation may overlap In the analysis of GBS and H1N1, we plan to assess relative incidence associated with seasonal and pandemic vaccines Challenge in that these were not always received in the same order Necessary to write new code to accommodate this situation
MODELING WITH THE CASE SERIES METHOD Confounding between exposure and age Unexposed cases help to disentangle the age effect. Not applicable to the GBS study because GBS is not hypothesized to be age dependent on the weekly or monthly scale. Long and indefinite risk periods Confounding between age and exposure effects is more likely with long risk periods. Not applicable to the GBS study risk period is hypothesized to be 6 weeks at the most.
MODELING WITH THE CASE SERIES METHOD Temporal effects Using calendar time as the underlying time line Appropriate for influenza vaccination studies due to the seasonality of influenza and influenza vaccination Determining global seasonality cut points has been one of the main challenges of the global GBS/H1N1 study
CONCLUSION The self controlled case series method is a powerful tool for vaccine safety studies Vaccine safety researchers continue to improve and modify the method The global study has been completed. Next phase to include developing countries