Rpid criticl pprisl GA: Grphic Apprisl ool for pidemiology using GA Rod Jckson University of Aucklnd, NZ Ferury 2011 1 1 picture, 2 formuls & 3 cronyms 2 ne picture: the GA frme British doctors smoking sttus mesured smokers non-smokers Lung cncer yes no 10 yers the shpe of every epidemiologicl study 3 Longitudinl (cohort) study 4 British doctors British doctors smoking sttus mesured Rndomised to spirin or plceo smokers non-smokers spirin plceo norml Lung function norml Myocrdil infrction yes no 5 yers ross-sectionl study 5 R 6 1
Middle-ged US women Middle-ged US women est pplied Brest cncer no Brest cncer Mmmogrm positive Mmmogrm negtive positive mmmogrm negtive Brest cncer yes no Dignostic test ccurcy study 7 linicl use of dignostic test 8 ne picture: the GA frme he 1 st cronym = : the 5 prts of every epidemiologicl study rticipnts xposure Group omprison Group every epidemiologicl study hngs on the GA frme 9 ime utcomes 10 1 st criticl pprisl tsk: descrie study s design y hnging on GA frme using cronym rticipnts Study Setting ligile rticipnts rticipnts 11 12 2
DV in long-hul flights: Lncet 2001;357:1485-9 rticipnts Study Setting: London, UK,? 1990s xposure & omprison Groups ligile rticipnts: no previous DV, > 50 yrs, plnned economy ir trvel 2 sectors > 8 hours xposure or Intervention Group (G) G G omprison or ontrol Group (G) rticipnts: 231, men ge 61-62 yers 13 14 xposure & omprison Groups xposure & omprison Groups 231 231 xposure or Intervention Group (G): Below knee compression stockings 115 116 omprison or ontrol Group (G): no stockings xposure or Intervention Group (G): Below knee compression stockings 115 116 100 100 omprison or ontrol Group (G): no stockings 15 16 utcomes () utcomes () 100 100 Dis-ese yes no c d utcomes () DV yes no = 0 c = 12 d rimry utcome () 17 18 3
ime () ime () utcome: e.g. deth incidence 0 0 incidence prevlence 19 = time from tkeoff to 48 hours post finl flight 20 ime () Study design: GA frme & rticipnts utcome: numer with DVs 0 12 = t post-flight ssessment (<48 hrs) prevlence 21 xposure Group omprison Group utcomes ime very epidemiologicl study hngs on the GA frme 22 DV in long-hul flights: Lncet 2001;357:1485-9 Setting: UK volunteers 479 considered ligile: > 50 yrs, no DV, flying > 8 hrs rticipnts xposure Group: stockings 231 115 116 100 100 omprison Group: no stockings Questions? ime: t post flight ssess 0 12 utcomes: DV 23 24 4
he 1 st formul: study nlyses All epidemiologicl studies involve mesuring the URRN of outcomes ccurrence (risk) of disese = Numertor Denomintor D Denomintor (rticipnts) D N 25 N Numertor (utcomes) cc = N D 26 All epidemiologicl studies involve mesuring the URRN of outcomes All epidemiologicl studies involve mesuring the URRN of outcomes D N Denomintor (rticipnts) During wht period of time () ws N mesured? (incidence) Numertor (utcomes) cc = N D (?) 27 D N Denomintor (rticipnts) At wht point in time () ws N mesured? (prevlence) Numertor (utcomes) cc = N D (?) 28 he 1 st formul: ccurrence (risk) = Numertor Denomintor 2 nd pprisl tsk: descrie nlyses y hnging numers on the GA frme nd clculting occurrences in exposure & comprison groups xposed Group D D omprison Group Denomintor 1: Denomintor 2: xposure Group omprison Group (G) (G) N N 29 Numertor 1: c d Numertor 2: 30 5
ccurrence = N D ccurrence = N D Denomintor 1: xposure Group G Denomintor 2: omprison Group G Denomintor 1: xposure Group G Denomintor 2: omprison Group G Numertor 1: c d Numertor 2: Numertor 1: c d Numertor 2: xposure Group ccurrence: G = G omprison Group ccurrence: G = G 31 xposure Group ccurrence: G = G omprison Group ccurrence: G = G 32 lculte G & G for the outcome DV post long-hul flight lculte G & G for the outcome DV post long-hul flight Denomintor 1: xposure Group G = 115 Denomintor 2: omprison Group G = 116 Denomintor 1: xposure Group G = 100 Denomintor 2: omprison Group G = 100 Numertor 1: = 0 c d Numertor 2: = 12 Numertor 1: = 0 c d Numertor 2: = 12 G = 0/115 = 0% t post flight ssessment I (intention to tret) nlysis G = 12/116= 10% t post flight ssessment 33 G = 0/100 = 0% t post flight ssessment G = 12/100= 12% t post flight ssessment (on tretment) or per-protocol nlysis 34 lculte G & G for the outcome verge pltelet count pre-long-hul flight Denomintor 1: xposure Group G = 115 Numertor 1: G = verge pltelet count = 242 (x 10 9 /L) Denomintor 2: omprison Group G = 116 Numertor 2: G = verge pltelet count = 240 (x 10 9 /L) 35 Descriing differences etween occurrences Reltive difference or Reltive Risk = G G Asolute Difference or Risk Difference = G - G Numer Needed o ret (NN) = 1 RD 36 6
Descriing differences etween occurrences Reltive difference or Reltive Risk = G G = 0/115 (0%) 12/116 (10%) = 0%/10% = 0 Asolute Difference or Risk Difference = G - G = 0/115 (0%) - 12/116 (10%) = - 10% (-10/100) Numer Needed o ret (NN) = 1 RD = 1 (- 10 /100) = - 100/10 = 10 if 10 people wore compression stockings 1 DV would e prevented 37 Descriing differences etween occurrences (verges/mens) Reltive difference or Reltive Men= G G Asolute Difference or Men Difference = G - G Numer Needed o ret (NN) = 1 RD 38 Study nlyses it s ll out G & G Questions? 39 40 he 2 nd cronym = RAMB* : ssessing is strength of study he 2 nd cronym = RAMB* : ssessing systemtic error (i.e. is) Recruitment Recruitment Alloction Mintennce Alloction Mintennce Blind or jective ssessment of outcomes 41 * ul Glsziou Blind or jective ssessment of outcomes 42 * ul Glsziou 7
3 rd pprisl tsk: ssess the degree of is y pplying the RAMB cronym Recruitment Alloction Mintennce Blind or jective ssessment of outcomes 43 Study setting ligile people RAMB were Recruitment processes pproprite to study gols? Study setting & eligiility criteri well descried? e.g. Recruit rndom/representtive smple R consecutive eligiles R volunteers from dvertisements rticipnts representtive of eligiles? rognostic/risk profile pproprite to study question? 44 Study setting RAMB ligile people were Recruitment processes pproprite to study gols? Study setting & eligiility criteri well descried? Recruited volunteers from dvertisements; eligiility criteri well descried rticipnts representtive of eligiles? 479 considered, 248 excluded - proly rognostic/risk profile pproprite to study question? yes 45 RAMB: A is for Alloction R: Allocte rndomly y investigtors (e.g drugs) Ws Alloction to G & G successful? were G & G similr t seline? ohort: Allocte y mesurement (e.g. smoking) 46 Successful rndomistion depends on 2 inter-relted processes Generting rndom numer sequence to llocte prticipnts onceling those involved (prticipnts & investigtors) from knowing to which group prticipnt will e llocted conceled lloction doule lind Rs 47 oncelment hppens during rndomistion, linding hppens fter 8
RAMB: A is for Alloction R: Allocted rndomly y investigtors using seled envelopes G & G similr t seline except more women in stocking group? Alloction is 49 50 Indequte lloction concelment & estimtes of tretment effects Study of 250 rndomised control trils of pregnncy nd childirth (Schulz et l. JAMA 1995;273:408) Results: studies with indequte or uncler lloction concelment overestimte Rx effects y 30-40% Moher et l 1998-ther trils on hert disese, stroke, mentl helth. Results: n verge indequte concelment yields 37% lrger estimtes of effect (Moher et l. Lncet 1998;352:609) RAMB were rticipnts Mintined s llocted? did most prticipnts remin in llocted groups (G & G) rticipnts &/or investigtors lind to exposure (nd comprison exposure)? omplince high & similr in G & G? ontmintion low & similr in G & G? o-interventions low & similr in G & G? ompleteness of follow-up high & similr in G & G? 52 115 116 100 100 RAMB good Mintennce? did most prticipnts remin in llocted groups (G & G) rticipnts &/or investigtors lind to exposure (nd comprison exposure)? no RAMB Were outcomes ssessed Blind or jectively? omplince high & similr in G & G? dk* ontmintion low & similr in G & G? dk o-interventions low & similr in G & G? ok ompleteness of follow-up high & similr in G & G? ok, lost < 15% 53 * dk = don t know If outcome ssessments not jective (eg. utomted or definitive) were investigtors Blind to exposure (nd comprison exposure) 54 9
RAMB Were outcomes ssessed Blind or jectively? he 4 (GA) study ises Recruitment is Alloction is Mintennce is utcome mesurements resonly ojective (ultr-sound) echnicin ws lind to exposure (lthough some prticipnts my hve hd stocking line) 55 utcomes ssessment is 56 DV in long-hul flights: Lncet 2001;357:1485-9 Recruitment ok Alloction ok Mintennce ok Questions? utcomes ssessment ok 57 58 he 2 nd formul: ssessing rndom error 4 th pprisl tsk: ssess degree of rndom error in study findings using the 2 nd formul Rndom error = 95% onfidence Intervl (1.96 x Stndrd rror) Rndom error = 95% onfidence Intervl G = 0%; 95% I = 0% - 3% G = 10%; 95% I = 4.8% - 16% 59 60 10
xcel As & pper Gte-lites Finl pprisl tsk: serch for & pprise SRs / met-nlyses using 3 rd cronym (FAIH) Find pproprite studies? Apprise selected studies? Include only vlid studies? otl-up (synthesise) ppropritely? here is GA for every study design www.epiq.co.nz 61 Heterogeneity dequted ddressed? 62 Muhmmd Shfique Sjid et l. Knee-length grduted compression stockings for thromoprophylxis in ir trvellers: A met-nlysis. Int J Angiol. 2008; 17: 119 123. Nine trils studying prticipnts using KL stockings were nlyzed. Forty-six of 1261 prticipnts rndomly ssigned to the control group developed deep vein thromosis (DV), compred with two of 1237 prticipnts (0.16%) in the KL stockings group...here ws n solute difference of 3.4% in the incidence of DV, in fvour of KL stockings. he numer needed to tret with KL stockings to void one cse of DV ws 29.4. However, there ws significnt heterogeneity mong trils. he RR for DV ws 0.08 in high-risk prticipnts nd 0.14 in low- to medium-risk prticipnts. 63 11