Week 3: revisist Flashcards
Pharmacoepidemiology
-study of use, risks, and benefits of drugs in populations (not individuals)
-pharma + epidemiology
-studies to estimate beneficial or adverse effects in population
experimental vs quasi/nonexperimental (observational)
-experiment: RCTs
-nonexperimental (observational): case-control, cohort
pharmacoepidemiologic and pharmacovigilance studies are primarily ___
-observational (nonexperimental)
Applications of pharmepi
-new info from premarketing studies
-better info on ADRs (more ppl w more conditions)
-patterns of use
-economic impact
-drug safety
-ethical and legal obligation
Data sources for pharmepi
-ADR reports (many go unreported tho)
-medical claims data (private, gov, insurance, some sold by companies): (diagnostic, procedure, lab, rx codes); (not very granular)
-EMRs (granular, but tmi)
-Indiana Network for Patient Care (INPC)
Indiana Network for pt care (INPC)
->100 separate healthcare entities providing data
-hospital, health networks, insurance providers
->18 million pt
-rx data
Limitations of observational studies in pharmepi
-confounding (independently related to BOTH exposure and outcome)
-information bias
-detection bias
-selection bias
-referral bias (encounter due to drug tx)
-protopathic bias (drug initiated before diagnosis)
-prevalence bias
-lag time
-immortal time bias (pt will not survive to measure outcome)
Information bias
-related to info regarding exposure/outcome
-includes measurement and/or classification error, or patient reporting/recall
-hawthorne effect: knowing they are being studied
Detection bias
-specific outcome dx preferentially in subjects exposed to agent
-more likely to look for an AE in someone exposed to a drug
-more pt on amiodarone may report more pulmonary toxicity but that is bc they are also being more routinely screened
-investigator detection bias in unblinded studies
Confounding by indication
-indication for drug or severity of disease predicts use of drug
-ex: ACE preventing MI in pt w HTN (HTN pt w comorbidities like DM are more likely to get ACE than other pt)
-ex: COXIBs and GI bleeds (coxibs reserved for people with higher GI bleed risk, so they might not cause GI bleed, pt just might have ulcer or smth already)
-occurs when risk of event is related to INDICATION for med use but not the use of the med itself
-appears when the REASON of rx is associated w outcome of interest
Selection bias
-bias related to procedures used to select subjects/influence study participation
-due to systematic diff in pt selected for study vs pt not selected
Referral bias
-reason for encounter related to drug treatment
-ex: use of drug contributes to diagnostic process
healthy user effect
-access to health care resources have higher level of education
-those who are adherent are healthier
Protopathic bias
-using drug to treat manifestation of undx disease, but drug may actually cause disease
-ex: antipsychotic started to tx delirium but anticholinergic effects contribute to delirium, NSAID for GI pain that turns out to be ulcer
-“reverse causality”
-occurs if tx was stared, then stopped or changed bc baseline manifestation caused by disease or other outcome event
-occurs when the drug is initiated in response to first sx of disease while still technically undiagnosed
-different than confounding by indication which is when risk is related to indication, not medication
Misclassification effect
-classify patient wrong
-missing data also a prob
Prevalence bias
-prevalent cases rather than new (incident) cases are selected
-need to make sure we’re selecting pt that don’t have abnormal baseline for what side effect we’re looking for
time related biases
-lag time
-immortal time bias
Lag time bias
-ex: PPI in fracture risk
-delayed time to see drug to start working
immortal time bias
-period of follow-up when outcome being studied could never occur
-if pt died before receiving heart transplant, they were defaulted to non-transplant
-so theres a time between waiting for transplant where transplant group is “immortal” until transplant
-pt in late initiation group had to be inhospital for at least 7 days to be in that group
Pharmacovigilance
-CONTINUAL monitoring for unwanted effects and other SAFETY aspects of marketed drugs
-detect, evaluate, understand ADR postmarketing
-wider use of data
Pharmacovigilance programs
-FDA adverse event reporting system (FAERS): receives postmarketing ADR reports
-FDA Sentinel System: monitor safety of FDA products
-FDA vax adverse event reporting system (FDA VARES)
-FDA started v-safe after covid for ppl to self-report vax adrs
Pharmacovigilance data use
-post-marketing surveillance
-signal detection
-data mining (social media)
-often regulatory agencies and industry (FDA)
-get data on pt excluded from premarketing studies, ADRs, similar to phxepi
phxepi vs pharmavigilance
-epi: more hypothesis driven, discrete studies
-vigilance: more ongoing, continuous processes
Comparative effectiveness research (CER)
-studying interventions in real world settings
-determine what therapeutic intervention (not just drug products) works best for a given disorder in patients likely to be seen in clinical practice
-conduct of research comparing interventions
-multiple study designs rct and observational
-efficacy (ideal) vs effectiveness (real)
-focus on effectiveness
Goals of CER
-overcome external validity
-compare tx
Efficacy vs effectiveness of CER
-efficacy: ideally will drug work
-effectiveness: in real world, does drug work
-focus on effectiveness
Pragmatic research
-studies that often test small practical changes in real world settings that could have impact on health outcomes
Pragmatic RCTs
-RCT w one or more pragmatic element
-intend to overcome limitations of traditional RCTs in order to answer CER questions
-hybrid RCT and routine care design
-intervention is the only thing controlled, everything else happens w clinicians in office
Pragmatic trial vs RCT
-effectiveness
-normal practive
-little selection
-more flexible
-no placebo, usually standard of care is control
-normal adherence
-higher external validity
-real world outcomes
-direct relevance to practice
pharmacoepidemiology opioid example?
-aim: estimate incidence and risk factors of pt that abuse opioids
pharmacovigilance review of fluoroquinolones example
-muscle, tendon, neurological effects
-outcome reported as disability
-varying durations
-lead to box warning and indication changes
What are primary limitations of RCTs related to applying them to clinical care?
-external validity
-comparison to placebo
Epidemiology
-study of distribution and determinants of health related events in POPULATION
-application of info to control health problems
-distribution: focus on freq and patterns
-frequency: # of events, rate or risk of disease
-patterns: person, place, time
Determinants
-causes and factors that cause disease
-why/how
-demographics, genetics, immunologic patterns, behaviors, comorbidities, environement, etc
-exposures and outcomes
Epidemiology data
-used to inform public health efforts
-basic science of pop health
-life expectancy, mortality, etx
-explain disease etiology/cause
-predict disease occurence
-control spread of disease
-assess efficacy of public health efforts
-inform decisions at individual level
-complete clinical picture
John snow
-cholera in london 1800s
-identified source of outbreak with spot map to find contaminated water
-established steps to investigate disease
core epidemiologic functions
-public health surveillance (collect mortality reports, identify new diseases or changes in patterns)
-field investigators (environement, food-borne, contact tracing)
-analytic studies llinked to biostatistics
-programmatic evaluation (vax efforts)
-policy development
social determinants of health
-education access/quality
-economic stability
-health access/quality
-neighborhood/environment
-social/community context
time in epidemiology
-change in occurence over time
-on graph as rate of disease or number of cases vs time
-rapid changes, seasonal trends, long-term trends
-epidemic period (not limited to infectious disease): time course and epidemic curve
Endemic
-baseline level of disease found in community for disease that is usually present in community
-expected level of disease over time
-can be high or low
Hyperendemic
-persistent, high levels of disease
Sporadic
-disease occurs infrequently and irregularly
Epidemic
-inc (maybe sudden) in cases above expected
-relative to usual freq of disease
-can be single case of long absent communicable disease
-can be first invasion of disease
Pandemic
-global epidemic
Epidemics
-occur when agent and host present in adequate numbers to spread
epidemics can result from
-inc in amt/virulence of agent
-reintroduction
-change in transmission, susceptibility, exposure
Common source outbreak
-exposure originates from same source
-point: all exposed at once, sudden, one incubation period, stops unless 2nd spread, steep upslope w gradual downslope (food-borne, nuclear disaster)
-continuous: over time from common source
-intermittent: exposure reemerges over time
Propogated outbreak
-transmission from person to person
-typical of community outbreaks (can be vehicle,syringe, born or vectorborne,mosquito)
-incubation period
-generation period
incubation period
-amt of time between initial contact w agent and onset of disease
-may create multiple peaks
-seondary cases appear once incubation period after peak of first wave due to secondary spread
-wanes after a few generations bc number of susceptible people falls below critical number or intervention measures
generation period
-amt of time between peaks in spread
-(estimate of incubation period)
mixed epidemic
-common-source outbreak followed by propagated spread
COVID-19
-incubation period 4-5 days
-provide basis for quarantine recommendations
Serial interval
-time between successive cases primary to secondary
-interval between clinical onset of disease
-4-7 days for COVID
if serial interval < incubation period
-can indicate that disease may be transmitted prior to onset of sx
incidence
-# of new cases of disease that occur over time period
-rate of development of disease
=(# of new cases over time period/total population at risk during same time period): incidence proportion or rate
Incidence Rate (IR)
-can incorporate person-time
-usually for longer-term follow up
-IR = (# of cases/time each person was observed totalled for all persons)
-decribes how quickly disease occurs
-assumes probability of disease is constant through time period
-report results as cases per person years
IR incorporating peron-time example
-adult opioid naive pt who received rx between 2012 and 2017
-1.3 mil rx for 341k pt
= incidence of death was 3.52 per 1,000 person-years
Prevalence
-NOT a RATE
-proportion of ppl w disease at given time or over time period
-total cases in population
=all new and pre-existing cases/population
-number, percentage, per unit size
-proportional to incidence rate and disease duration
-if duration short, prevalence similar to incidence
Prevalence is proportional to
-incidence rate and disease duration
-if duration is short, prevalence is similar to incidence
Attack rate
-alt form of incidence rate in outbreak settings
-used for diseases for short times
-often specific exposure (food-borne)
-not a true rate bc time may be unknown
-measure of risk
AR (attack rate) formula
(# of new cases) / (total population)
Secondary attack rate
-rate of disease in group among those exposed to initial case
-document transmission in defined/closed population
-index of spread in defined group
-measure contagiousness
-useful in evaluating control measures
-denominator restricted to susceptible contacts
= (# of new cases) / (# of exposed susceptible individuals)
secondary attack rate (SAR) formula
= (# of new cases) / (# of exposed susceptible individuals)
Seoncdary attack rate of covid
-16.6%
-harder to measure bc underreporting
Basic reproductive number (Ro)
-avg number of secondary cases produced by one infected individual introduced into a population of susceptible individuals
-estimates epidemic potential
-<1 disease likely dying out
->1 likely to spread
-depends on location and population density and other factors
Basic reproductive number (Ro) in covid
-initially around 2
-now less than 1 in Indiana
Mortality (or morbidity) rate
-freq of death
= (deaths/population) x 10^n
-denominator can vary (vital stats, may use size of population in middle of time period
-can report number per 1,000 or 100,000
mortality rate variations
-crude mortality rate (all causes)
-cause-specific
-age-specific
-infant mortality rate (<1 year of age/# of live births)
-maternal mortality rates
-race specific
-age adjusted
Case fatality rate
-proportion of people with a given condition who dies from condition
-proportion not a tre rate
-indication of VIRULENCE in population
-how fatal is disease
-compare fatality to other diseases
= [(# of cause specific deaths among incident cases) / (# of incidence cases)] x 10^n
CFR for COVID-19
-around 1%
-need better data tho
-but flu CFR was 0.1%
Risk ratio (relative risk)
-measures of association in cohort study
-risk of outcome (disease) among one group with risk or exposure, among another group w/o risk or exposure
=risk of disease in group of interest / risk of disease of comparator
=indicdence of disease in exposed / incidence of disease in unexposed
-dont need to calculate
-outcome could be good tho
Rate ratio
-compares INCIDENCE RATES between 2 groups (cohort study)
-may include person-time
=rate for group A/ rate for group B
statistical significance indicated by
-95% Cl
-not sig if includes 1
->1 is higher risk
-<1 is lower risk
Odds ratio (OR)
-used for CASE CONTROL studies
-estimates relative risk
-don’t use RR in case control since we don’t know total population
= odds of exposed person being a case / odds of unexposed person being a case
-95% Cl
