Presentation is loading. Please wait.

Presentation is loading. Please wait.

Measuring covariate data_Presentation (November 14, 2007) 1 Measuring covariate data in subsets of study populations: Design options Jean-François Boivin,

Similar presentations


Presentation on theme: "Measuring covariate data_Presentation (November 14, 2007) 1 Measuring covariate data in subsets of study populations: Design options Jean-François Boivin,"— Presentation transcript:

1 Measuring covariate data_Presentation (November 14, 2007) 1 Measuring covariate data in subsets of study populations: Design options Jean-François Boivin, MD, ScD McGill University 19 August 2007

2 2

3 3 16 th International Conference on Pharmacoepidemiology Barcelona 2000

4 4 What about missing covariate data?

5 5 Do not research that topic Option #1

6 6 Conduct study without covariates Scientifically reasonable for certain questions Example: Sharpe et al. 2000 Option #2

7 7 British Journal of Cancer 2002 The effects of tricyclic antidepressants on breast cancer risk Genotoxicity in Drosophila Comparison of antidepressants: –6 genotoxic vs 4 nongenotoxic Confounding unlikely

8 8 Option #3 “Confounding by other determinants was studied in analyses with data obtained by interviewing samples of subjects…”

9 9 List 4 - 6 different sampling strategies: “Confounding by other determinants was studied in analyses with data obtained by interviewing samples of subjects…” a) ? b) ? c) ? d) ?

10 10

11 11

12 12 Two-stage sampling

13 13 Entire population (=truth) OR=0.5 OR=2.5 Obese Not obese All E+E- D+ D- 12,000140 10,20010,400 22,20010,54032,740 2,000 40 10,000 100 200400 10,000 2,200440 20,00010,100

14 14 Obese Not obese All E+E- D+ D- 22,20010,540 not available computerized databases 2,200440 20,00010,100 D+ D-

15 15 Two-stage sampling

16 16 Obese Not obese All E+E- 250/ 2,200 440 20,000 10,100 32,740 227 23 125 2 23227 125 248 D+ D- D+ D- D+ D- Two-stage sampling OR 1  biased OR 2  biased 250 x 250 = 1

17 17 White. AJE 1982 Walker. Biometrics 1982 Cain, Breslow. AJE 1988 Weinberg, Wacholder. Biometrics 1990 Weinberg, Sandler. AJE 1991 Statistical analysis; further design issues

18 18

19 19 Option 1: Option 2: Option 3: Option 4: No study No covariate measurement 2-stage sampling Case only measurement

20 20 Ray et al. Archives of Internal Medicine 1991

21 21 Cyclic antidepressants and the risk of hip fracture

22 22 E+E- All RR=0.5 RR= D+ D- D+ D- D+ D- All Not obese Obese RR=0.5 N 1 =? N 2 =? RR=0.5 N 3 =?N 4 =? RR= RR=0.5 N 1 =1,000 N 2 =1,000 RR=0.5 N 3 =1,000N 4 =1,000 RR=0.5 N 1 =1,000 N 2 =1,000 cross-product ratio =1 RR=0.5 N 3 =1,000N 4 =1,000 RR= RR=0.5 N 1 =1,000 N 2 =1,000 RR=0.5 N 3 =1,000N 4 =1,000 RR= Confounding: Quick review

23 23 Obese Not obese All D+ D- OR=0.5 OR= E+E- OR=0.5 5001,500 OR=0.5 1,0003,000 OR= OR=0.5 cross-product ratio =1 OR=0.5 OR= Case-control study

24 24 Cyclic antidepressants and the risk of hip fracture

25 25 E+E- D+ Obese Not obese All D- D+ D- D+ D- 2,200 440 computerized database 20,000 10,100 22,200 10,540 medical record review 2,200 440 computerized database 20,000 10,100 22,200 10,540 2,000400 ?? 20040 ?? 2,200 440 20,000 10,100 22,200 10,540 Covariate data on cases only

26 26 E+E- D+ Obese Not obese All D- D+ D- D+ D- 2,000400 ?? 20040 ?? 2,200 440 20,000 10,100 22,200 10,540 OR 1 OR 2 assume OR 1 = OR 2 then: cross-product ratio =1 implies no confounding Covariate data on cases only

27 27 What if confounding seems to be present? Extensions

28 28

29 29 Option 1: No study Option 2: No covariate measurement Option 3: 2-stage sampling Option 4: Case only measurements Suissa, Edwardes. 1997

30 30 Confounder data on cases only Obese Not obese E+E- D+ D- 2,000 220 ? ? 200220 ?? Cross-product ratio =10 Confounding plausible D+ D-

31 31 Epidemiology 1997 Extensions of Ray’s method to presence of confounding Requires additional data from external sources

32 32 Smoker Nonsmoker All E+E- D+ D- Theophylline 171330 9563,1544,080 14519 3811 14519 24% of 4,080 3811 76% of 4,080 14519 24% of 4,080 obtained from population survey 3811 76% of 4,080 Confounding; no interaction

33 33 Extensions of Ray’s method to presence of interaction Requires further additional data from external sources Suissa, Edwardes. 1997

34 34 No interaction OR=0.5 Obese Not obese E+E- D+ D- 12,000140 10,20010,400 2,000 40 10,000 100 200400 10,000

35 35 Option 1: No study Option 2: No covariate measurement Option 3: 2-stage sampling Option 4: Case only measurements Suissa, Edwardes. 1997 Multi-stage sampling Partial questionnaires Propensity score adjustments Others:

36 36

37 37

38 38 Monotone missingness

39 39 Wacholder S, et al.

40 40 Cov 12345678 Subject 1 2 3 4 5 6 7 8 9 10 … n Cov 12345678 Subject 1 2 3 4 5 6 7 8 9 10 … n Cov 12345678 Subject 1 2 3 4 5 6 7 8 9 10 … n Cov 12345678 Subject 1 2 3 4 5 6 7 8 9 10 … n Cov 12345678 Subject 1 2 3 4 5 6 7 8 9 10 … n Cov 12345678 Subject 1 2 3 4 5 6 7 8 9 10 … n Cov 12345678 Subject 1 2 3 4 5 6 7 8 9 10 … n Cov 12345678 Subject 1 2 3 4 5 6 7 8 9 10 … n Cov 12345678 Subject 1 2 3 4 5 6 7 8 9 10 … n

41 41 Wacholder S, et al. Restricted to a small number of discrete covariates

42 42 Methodologic research Stürmer et al. AJE 2005, 2007 Propensity score calibration

43 43 Summarizes information about several covariates into a single number Used for matching, stratification, regression Propensity score

44 44 Main cohort: selected covariates -“error-prone” scores estimated -regression coefficients estimated Sample: additional covariates -gold standard scores -regression calibration Advantage: multivariable technique Stürmer et al. 2005

45 45 “Until the validity and limitation of… [propensity score calibration] have been assessed in different settings, the method should be seen as a sensitivity analysis.” Stürmer et al. 2005

46 46

47 47

48 48 Stage 1: 278 cases in 4561 pregnancies Stage 2: 244 cases + 728 non cases

49 49

50 50 “Relatively few examples of two-and three- phase sampling designs for case-control studies have appeared to date in the epidemiologic literature.This is unfortunate, because the stratified designs are easy to implement and can result in substantial savings.” NE Breslow (2000)

51 51 Consent for second-stage interviews: Cases: 49% Controls: 39%

52 52 jean-f.boivin@mcgill.ca


Download ppt "Measuring covariate data_Presentation (November 14, 2007) 1 Measuring covariate data in subsets of study populations: Design options Jean-François Boivin,"

Similar presentations


Ads by Google