Pharmacogenetics

1
Cohort Studies, Nested Case-Control Studies,
Relative Risks Brian F. Gage, MD,
MSc Department of General Medical Sciences
Associate Professor of Medicine September 30,
2009
2
Case-Control Studies
Study Onset
Time
Exposed
Cases
Unexposed
Exposed
Controls
Unexposed
Direction of Inquiry
3
Cases

• Typically have a rare disease
• Identified and selected from a defined source
  population (e.g., all patients from a clinic,
  hospital, HMO, state, or nation)
• Likelihood of a case being included in study must
  not depend on exposure to risk factor
• Criteria for defining cases should be sensitive
  and specific

4
Controls

• Identified and selected from the same source
  population as the cases
• Chosen at random from source population
• An exposed member of the population should have
  the same chance of becoming a control as an
  unexposed member

5
Assessing Exposure

• Interviews and questionnaires often used
• Objective measures limit bias
• e.g. Genotype, biological markers
• e.g. Biological markers
• Many exposures dont have biological markers
• Biological marker may not be present when test is
  taken (e.g., alcohol is quickly metabolized)
• e.g. Pharmacy records of drug exposure
• Adding an addl risk factor may allow one to
  quantify potential bias
• E.g. esophageal squamous cell CA in GERD study

6
Case-Control Studies

• Pros
• Efficient (esp. for rare outcomes)
• Faster
• Cheaper
• Allow for multiple risk factors to be assessed
• Provide for an OR (ad/bc), which estimates RR.
• Cons
• Bias (systemic error)
• Sampling bias (e.g. referral bias from cases
  referred to BJH)
• Ascertainment bias (e.g. looking harder at cases)
• Recall bias (e.g. new mom recalls pregnancy
  habits)
• Differential measurement bias (blinding can
  minimize)
• Temporal association is unclear
• Only 1 outcome can be studied
• Does not yield incident rates

7
Study Designs

• Cohort study
• Prospective
• Retrospective
• Fancy Case-control studies
• Nested case-control study
• Case-cohort study
• Case-crossover study
• Miscellaneous studies
• Clinical trial
• Intervention study
• Other (e.g., ecological study, case study, and
  case series)
• 10 slides adapted from UNC Epi 101
• www.sph.unc.edu/epid/course/Webpage/epid168/study_
  designs.ppt

8
Cohort Studies
Onset of prospective study
Onset of retrospective study
Time
Eligible subjects
Disease
Exposed
No Disease
Disease
Unexposed
No Disease
Direction of inquiry in a prospective cohort study
8 slides adapted from Ty Borders, Ph.D course on
managerial epidemiology lthttp//www.ttuhsc.edu/hsr
m/courses/GHSR5301/managerialepi_part2.pptgt
9
Cohort Studies Provide Rates

• Incidence Rate, IR, can be calculated by dividing
  the number of adverse events by the number of
  patientyears
• IR dz onsets / ?pt-yrs
• Examples
• Nurses Health Study of 121,000 nurses
• Multiple endpoints

10
Prospective Cohort Study--Example

• Source Palareti G et al (2003) Predictive value
  of D-dimer test for recurrent venous
  thromboembolism after anticoagulation withdrawal
  in subjects with a previous idiopathic event and
  in carriers of congenital thrombophilia.
  Circulation 108313-8.
• Cohorts Patients w/ and w/out an elevated
  D-Dimer after standard Rx for a venous
  thromboembolism (VTE).
• Stratified results based on presence or absence
  of genetic thrombophilia, over 1-3 years
• Outcome VTE recurrence
• How would you analyze results?

11
Is D-Dimer a Valuable Test?

• D-Dimer is known to be greater in older patients
• Older patients have higher rates of VTE
  recurrence

D-Dimer
New VTE
12
Is D-Dimer Confounded by Age?

• D-Dimer is known to be greater in older patients
• Older patients have higher rates of VTE
  recurrence

Age
D-Dimer
New VTE
13
Subset with Unprovoked VTE
14
Thrombophilia Cohort
15
Confounding variable is associated with exposure
and a cause of the outcome e.g., cigarette
smoking, coffee drinking and MI Confuses
interpretation of relationship between exposure
and outcome.
Smokes
Coffee
MI
Thanks to Dr. Jane Garbutt, Sept 10, 2008
16
Retrospective Cohort Study--Cohort

• Study Question Does warfarin increase the rate
  of osteoporotic fractures
• Primary outcome Osteoporotic fracture during up
  to 1-year follow-up
• Subjects Medicare beneficiaries with Afib
• Outcome definition ICD-9 code for Fx
• Statistical approach What would you use?

17
Results in men (minimal effect in women)
18
Results of Cox Proportional Hazards
19
Odds Ratio

• Measure strength of association of RF (exposure)
  and outcome with Odds Ratio (OR)
• Disease
• Yes No
• Risk Yes a b
• Factor No c d
• OR odds of exposure in cases a/c
• odds of exposure in controls b/d

Thanks so Dr. Jane Garbutt, Sept 10, 2008
20
Relative Risk

• Disease
• Yes No
• Outcome Risk factor a b ab
• No risk factor c d cd
• ac bd
• RR a/ (ab)
• c/ (cd)
• As formulated here, assumes that the 2 cohorts
  are followed for equal lengths of time.

21
Relative and Attributable Risk

• Relative risk (risk ratio) ratio of IR in
  exposed to IR in unexposed
• RR incidence rate among exposed
  (avoids assumption re follow up)
• incidence rate among unexposed
• NB exposure could be a risk factor
• Attributable risk (risk difference) additional
  risk of disease following exposure (i.e. the
  incremental risk above that of the unexposed).
• AR IR among exposed - IR among unexposed

22
Risk Reduction

• Relative risk reduction (RRR) is 1 RR
• Absolute risk reduction, ARR
• IRtherapy IRcontrol
• Number needed to treat (NNT)
• number of patients that have to receive the
  therapy to prevent one adverse event
• 1/ARR

23
Pros/Cons of Cohort Studies

• Advantages Disadvantages
• Calculation of RR Time consuming
• May yield info. on incidence Require large sample
  sizes
• Clear temporal relationship Expensive
• Can yield info. on multiple Not efficient for
  study of
• outcomes rare events
• Minimizes bias Losses to follow-up
• Strongest observational design
• for establishing cause-effect
• from Greenberg et al.

24
Retrospective vs. Prospective

• Attribute Retrospective Prospective
• Information Less complete More complete
• DCd exposures Useful Not useful
• Emerging
• exposures Not useful Useful
• Expense Less costly More costly
• Completion
• time Shorter Longer
• from Greenberg et al.

25
Cross-sectional Studies

• Measures prevalence of health outcomes or
  determinants of health (e.g., biomarkers) at a
  point in time or over a short period

26
Cross-sectional Studies

• Pros
• Fast (no follow up!)
• Cheap
• Allow for multiple risk factors and outcomes to
  be assessed
• Provide prevalence
• Cons
• Not feasible for rare outcomes
• Temporal association is unclear
• Biases of both exposure and outcome are common
• Does not yield incident rates
• Can identify a cohort that can be followed later
• E.g. Prevalence of non-susceptible S. pneumonia
  to amoxicillin was 3-7 in our community

27
Nested Case-Control Study
Source (2 slides) Stanford CLIO
http//clio.stanford.edu7080/cocoon/cliomods/trai
lmaps/design/design/nestedCase-Control/
28
Nested Case-Control Study

• Requires a cohort that has been followed over
  time.
• Has exposure information and/or biospecimens
  collected of interest to the investigator.
• Cases incident disease
• Controls Same cohort, but no incident disease
• Compare the exposure frequencies in cases and
  controls as in a non-nested case-control study

29
Nested Case-Control Study (2)
Consider the following cohort
X3
X case O loss to follow-up
X2
O
O
X1
t1
t2
t3
Time
30
Nested Case-Control Study (3)

• At time t1 the first case occurs for which 8
  eligible controls are identified
• Similarly, there are 5 eligible controls for the
  case at time t2, and 4 eligible controls for the
  case occurring at time t3
• A control can become a case at a later time
  (e.g., cases at t2 and t3 could serve as a
  control for case at t1)
• Controls can be selected randomly from all
  eligible controls (i.e., 1 or more controls for
  each case)
• Number of eligible controls decreases with
  increasing number of matching factors

31
Case-Cohort Study (1)
Consider the following hypothetical cohort
X
X case O loss to follow-up
X
O
O
X
t0
Time
32
Case-Cohort Study

• In a closed cohort everybody enters cohort at
  start of follow-up t0
• In open cohort the time of entry into the cohort
  is dynamic
• Case-cohort study population consists of all
  cases of any disease of interest, and members of
  the cohort
• A subject might be both case and members of the
  cohort

33
Case-Crossover Study

• Study of triggers within an individual
• Case" and "control" component, but information
  of both components will come from the same
  individual
• Case component" hazard period which is the
  time period right before the disease or event
  onset
• Control component" control period which is a
  specified time interval other than the hazard
  period
• Exposure must vary from time to time within
  person
• Disease must have abrupt onset, and effect of
  exposure must be brief
• Examples risk of MVA when using cell telephone

34
Case-Crossover Study
Event



Hazard Period
Control Period 1
Control Period 2
Measure odds of exposure here.
http//www.pitt.edu/super1/lecture/lec0821/001.ht
m
35
Redelmeier DA, Tibshirani RJ Association between
cellular-telephone calls motor vehicle
collisions. N Engl J Med 1997 336453-8.
36
Final Remarks on Cohort and Case-control Studies

• Cohort and case-control studies can be either
  prospective or retrospective, or both
• Prospective Exposure information is collected
  during follow-up, and period of time at risk
  occurs during conduct of the study
• Retrospective Cohorts (cohort study) or cases
  (case-control study) are identified from recorded
  information, and period of time at risk has
  already occurred before the study has begun
• Measures of effect (ratios)
• Cohort study Incidence rate ratio (open cohort)
  or risk ratio (closed cohort)
• Case-control study Exposure odds ratio as
  estimate of incidence rate ratio or risk ratio

37
Other Study Designs

• Clinical trials
• randomization of individuals to a treatment or
  reference group
• Intervention studies using groups
• intervention usually on community level rather
  than individual level
• Analytic plan has to take clustering into account

38
Calculate the RR of the WHI
J. Mason Estrogen plus Progestin and the Risk of
Coronary Heart Disease NEJM 349523-534 Aug 2003
39
(No Transcript)
40
Which Type of Study Design Would You Use

• To evaluate risks and benefits of a new therapy
• To evaluate how well genotype (e.g.
  apolipoprotein E2 or E4 allele) predicts
  intracerebral hemorrhage?
• Whether use of the pneumococcal polysaccharide
  vaccine alters the overall risk of
  community-acquired pneumonia
• L Jackson et al. NEJM 3481747-1755 May 2003

41
Next Week

• Diagnostic Testing (Gage)
• Read
• Hulley chapters 11 12
• Abstracts
• -Association between cellular-telephone calls and
  motor vehicle collisions
• -Maisel AS et al. Rapid Measurement of B-Type
  Natriuretic Peptide in the Emergency Diagnosis of
  Heart Failure N Engl J Med 2002.