Pharmacoepidemiology M2 CARE 2024 PDF

Pharmacoepidemiology Justine Bénévent M2 CARE 2024  Pharmacoepidemiology What are we talking about? Epidemiology Science focused on studying the health of populations Initially: the study of infec...

Pharmacoepidemiology Justine Bénévent M2 CARE 2024  Pharmacoepidemiology What are we talking about? Epidemiology Science focused on studying the health of populations Initially: the study of infectious epidemics Pharmakos: the drug Pharmacoepidemiology Science focused on studying the effect of medications on the health of populations  Pharmacoepidemiology What are we talking about? Fundamental Pharmacology and Observational Clinical Pharmacology Mechanism of action of the drug Effect of the drug on one or more clinical parameters Comparative Clinical Pharmacology Clinical trials: scientific evidence of the difference in effects between two drugs Pharmacoepidemiology Proof that clinical trials are not or misleading  Proof, more proof! Fundamental pharmacology, logical reasoning the presence of ventricular extrasystoles (rhythm disorder) is associated with increased post-myocardial infarction mortality class 1 anti-arrhythmics can suppress this cardiac rhythm disorder Reasoning sufficient to justify the use of these drugs in subjects with extrasystoles after MI No proof of efficacy  The CAST study (NEJM 1991 ; 324 : 781-8) Flecainide and ecainide were evaluated in a randomized clinical trial of mortality  Proof, more proof! 1) Fundamental pharmacology, logical reasoning the presence of ventricular extrasystoles (rhythm disorder) is associated with increased post-myocardial infarction mortality class 1 anti-arrhythmics can suppress this cardiac rhythm disorder 1) Comparative clinical pharmacology CAST study: randomized clinical trials Clear increase in the risk of mortality in subjects treated with class 1 antiarrhythmics  Why is it different Fundamental Pharmacology Drug effect: Mechanism of action Comparative Clinical Pharmacology Drug effect: Efficacy Demonstrated benefit in an experimental setting on a selected population Pharmacoepidemiology Drug effect: Effectiveness Measured benefit-risk ratio in an uncontrolled environment on a heterogeneous population  Why is it different Comparative Clinical Pharmacology – Clinical Trials Controlled environment Patient selection: Target population Drug intake: Controlled Treatment monitoring: Controlled Pharmacoepidemiology Uncontrolled environment No patient selection: Real-world population Drug intake: Natural Treatment monitoring: Natural  Populations Not Studied in Clinical Trials Children Pregnant or breastfeeding women Elderly Patients with severe comorbidities Patients undergoing treatment with multiple drugs Sub-populations with genetic polymorphisms Others... Essais cliniques Vie réelle  The 5 « too much » of clinical trials Clinical Trials At-risk population? Too simple Too limited to a specific age group Low risks? Too few subjects Misuse, Off-label use? Too narrow Long-term effects? Too short Objectives Describe Estimate  Objectives Describe Characteristics of treated subjects (real-world population) Patterns of drug use: o Consumption volume, number of treated patients, dosage, treatment duration, indication, therapeutic combinations Estimate Drug safety Drug effectiveness in real life Impact of drugs  Concept of risk Definition Probability that an event will occur within a given time period Estimated via an incidence rate Number of new cases occurring among N at-risk individuals during a specified time period An omnipresent phenomenon for any individual Increased/decreased when exposed to a specific factor Varies depending on certain characteristics Time Place Individuals  Risk and medication Exposure factor Protective against many diseases May lead to unintended events Various types of risks Predictable o Expected effects / pharmacological properties Unpredictable o Unexpected effects Avoidable o Knowledge of risk factors Acceptable o Relative to significant benefit o In the absence of alternatives  Effectiveness the ability to be successful and produce the intended results  Studying the Effectiveness of Medications Necessary/Ideal Information for Informed Decision-Making by Health Authorities Comparative effectiveness of active medications Data generalizable to a population of current users Data collected on a large enough scale to address major safety concerns Requested by: Authorities (e.g., US, UK) Health insurers (e.g., US) Hospitals (e.g., US) Industry: an opportunity to promote adherence  Studying the Effectiveness Objective: Study the association between medication use and patient health outcomes Focus: Association between drug exposure and major events such as: o Survival o Functional improvement o Recovery Comparative effectiveness: Compare the effectiveness of two or more medications for the same indication  Impact Market withdrawal of a medication New indications/contraindications Reimbursement removals  Conclusion Pharmacoepidemiology Description and estimation of: o Drug utilization o Risk o Effectiveness o Impact Helps to understand whether fundamental pharmacology and clinical pharmacology are applicable to the real-life experience of patients Bias in Observational Studies  Bias A major issue in pharmacoepidemiology Challenges causal analysis Incorrect estimation Missing an existing association Finding a non-existent association Distortion of the exposure-event association Overestimation of association strength Underestimation of association strength  Types of bias 1) Selection Bias Related to the selection of study participants 2) Information Bias Related to the quality of collected data 3) Interpretation Bias Confounding Bias o Indication bias o Protopathic bias  Selection Bias: Definition At the time of subject inclusion: Case-control studies: Inclusion of cases/controls related to the exposure under study Cohort studies: Inclusion of exposed/non-exposed subjects related to disease occurrence Frequency of exposure (case-control) or event (cohort) differs between included subjects  Example: Non-Response Bias Study on vaccination and central nervous system demyelinating disorders (CNSDD) Controls: Children selected via telephone 5000 families contacted, 2000 agreed to participate Families who agreed may be more favorable or compliant with vaccination, leading to a higher likelihood of vaccination among controls compared to the general population  Example: Admission Bias ("Berkson's Bias") Case-control study in hospitals: NSAID use and abdominal pain Doctors may suspect gastric ulcer more strongly in NSAID users with abdominal pain Higher likelihood of hospitalization for further testing  Example: Admission Bias ("Berkson's Bias") Case-control study in hospitals: NSAID use and abdominal pain Doctors may suspect gastric ulcer more strongly in NSAID users with abdominal pain Higher likelihood of hospitalization for further testing More exposed cases registered Overestimation of the association  Preventing Selection Bias Cannot be controlled during analysis Must be anticipated during protocol design: Clearly define the "universe" of exposed individuals (cohort) the "universe" of cases (case-control) Select unexposed individuals and controls from the same universe  Information Bias: Definition Subjects are classified by: Their exposure Event occurrence Information Bias = Misclassification of exposure or event data due to systematic differences in data collection across groups  Types of Information Bias Recall Bias Subjects in different groups remember exposure or events differently Interviewer Bias Interviewers question subjects differently based on group membership  Example: Recall Bias Drug exposure during early pregnancy and birth defects Mothers of malformed children (cases) and mothers of healthy children (controls) interviewed about medications used in the first trimester Case mothers are more likely to recall drug exposure due to reflection on potential causes  Example: Interviewer Bias Medication altering alertness and road accidents Subjects hospitalized after accidents (cases) vs. for other reasons (controls) Interviewers aware of the hypothesis may probe cases more thoroughly than controls  Preventing Information Bias Standardize measurement instruments Use non-subjective data Set measurement methods at study outset and avoid modifications Blind interviewers to study hypotheses  Confounding Bias: Definition Distortion in estimates due to a confounding factor Confounder is related to: Drug exposure Event occurrence Does not lie on the causal pathway between exposure and event  Example: Oral Contraceptives (OC) and Melanoma Risk Strong association found Confounder: Sun exposure Independent risk factor for skin cancer Women using OCs are more likely to have high sun exposure Sun exposure, not OC use, explains the association  Indication Bias: Definition A special case of confounding bias One medication preferentially prescribed to patients already at higher risk of the event Confounding factor: Prescription rationale Common and challenging to control in pharmacoepidemiology Example NSAIDs (coxibs vs. traditional NSAIDs) and gastrointestinal bleeding  Coxibs prescribed more often to patients with gastrointestinal history  History itself is an independent risk factor for bleeding  Protopathic Bias: Definition Treatment for early disease symptoms appears to cause the disease itself Drug wrongly accused of causing the event  Preventing Confounding Bias A priori Restrict study to a homogeneous subject group Balance groups for potential confounders (e.g., matching) A posteriori Identify confounders beforehand and collect data Use multivariable analyses, propensity scores, etc.  Conclusion No pharmacoepidemiological study is entirely free from bias Researchers must: Anticipate potential biases during protocol design Systematically discuss biases when presenting results When reading an article: Look at the descriptive table of the population to check if the two compared groups are comparable, or if not, whether this has been accounted for in the analyses. Consider the potential biases (beyond those discussed in the article's discussion section).