L8 Case Contr Studies

Question 1

LO2: outl faces sugg case contr design suitab for partic epidemic Q.

Answer 1

-need case contr because cohort long time (excep historic). Cohort expensive esp if want detailed info. Cohort bad for rare events as would need huge samp. Prospec cohort meas absol risk (IR/MR) and comp int to ext.

Question 2

LO3: anal case contr. Estim str of assoc from res of simp case contr study. Calc and interp ef and 95% CI.

Answer 2

-rare dis assump- in epidemic case contr studies where hyp tests assoc btw expos and dis. Assumes if dis prev low, then expos OR approaches the rel risk.
OR= (exposed/unexp cases)/(exposed contr/unexp contr). Odds of expos if case/ odds of expos if not.
OR= exposed cases(a) X unexp contr(d)/ unexp cases(b) X exp contr(C). Under rare dis assump this equates to IRR. OR is val meas of excess risk in cases comp to contr.
Ef= exp (2x sqroot 1/a+1/b+1/c+1/d)
-ratio of odds of being exposed= ratio of odds of being a case. Estim of risk from OR v simil to IRR. CI not incl 1=sig.
-how many controls? Unlike IRR precis of ORR is aff by numb of healthy as well as cases. Ef incr if any grp too small. So typ 4-6x more contr than cases=narrower CI.

Question 3

LO4: desc diffs and similes btw case contr and other epidem design.

Answer 3

-retrospect= relies on data on expos and/or outc already collec. Decr reliab as recall bias.
-prospect= asks specif study Q, recruits approp pp’s, looks at expos and outc in foll yrs.
-anal forw/backs. And data collec as outc occ/historic review.
Cohort anal looks forw- look at expos status, IRR over time.
Case contr always looks back- given case status now, OR of prev expos.
-concur cohort- prospect data collec, wait for events.
- historic coh- retrospect data collec from records.
-convent case contr- retrospect data from recall.
-nested case contr= cases drawn from same cohort as controls. Collec data on cases and controls is defined bef study. Nested= decr recall bias and temporal a ambig, cheaper and less time. Incid and prev can be estim somet unlike norm case contr where tot exposed and foll up time not us known. Id cases in a cohort, match no case controls, calc CI like for CC.
-advs over norm CC- IR calc (samp fractions known-can find tot dis/not, known samp frame). Pop for sampling of controls already defined. Already have def meas.
-adv over norm cohort- can collec more def info for minority of pp’s.

Question 4

CC vs cohort

Answer 4

Cohort:
- selec based on expos status, comp outc. 
-large, time cons and expens. 
-bad for rare dis/outc. 
-can study range outc for each oxphos. 
-lim bias in ascert of expos but prone to loss dur foll up. 
-can set expos precede out, prev Rev causality. (Nested also). 
-can direc meas incid (nested also). 
-IRR
CC:
-selec based on outc, comp expos. 
-quicker and cheaper. 
-bad for rare expos. 
-can study range expos for 1 dis/outc. 
-prone to slice and info bias/recall bias. 
-up to 6 contr per case. 
-OR

Question 5

LO5: desc lims and assumps inherent in CC.

Answer 5

-selec bias= distorts ev/data from way collected. Predom contr selec eg contr from resp ward likely smoke. Study would underestimate OR=bias, would sugg not linked. Direc of bias imp!
Cases should be repres eg not just spec centres. Controls should be repres of pop from which cases came. How to define?
-info bias=
Non diff misclassif ie rand inacc meas. CCwrong in eq meas and direc not matter, bal out.
Systematic misclassif- eg recall bias=recall of susp RF infl by fact they have dis. Assessor bias, data collec methods btw diff grps.
Even 10% non diff misclassif can= BIG underestimate OR. Obs OR tends tow 1 ie shrinkage to null.
Systemat eff as 10% unexposed cases wrongly classic as exposed= sys misclassif=OR furth away or tow null val ef 1. Away from 1 eg want blame someth for dis=overestim. Tow 1 eg under rep expos.
-confounding= distortion of true rel btw 2 chars. Can=case conclus if not acc for. Min by matching using imp confounder said. Adj for by anal with logistic regress. Ask multip confound immed, keeps acc.

Question 6

LO1: outl princips underlying case contr studies.

Answer 6

-is type of analytical study which is type of obs study. Obs study= researchers no contr over expos, obs what happens to grps people.
Cohort study- exp and unexposed grps, foll up outcomes and p-y. obs outcomes.
-case contr have outc of int, comp with grp that don’t. Id on basis of dis NOT expos. NOT observed, are investig. Gen retrospect.
Case contr study= epidemic study often used to ID RFs. Comps pts with cond to w/o. Looks back to see char/expos diffs. Sample of cases and defined controls. Sample at 1 time.