3 Causal Models Part II: Counterfactual Theory and Traditional Approaches to Confounding (Bias?) Confounding, Identifiability, Collapsibility and Causal inference Thursday reception at lunch time at SACEMA Review Yesterday Causes – definition Sufficient causes model – – – Component causes Attributes Causal complements Lessons – – – – Disease causation is poorly understood Diseases don’t have induction periods Strength of effects determined by prevalence of complements Only need to prevent one component to prevent disease This Morning Counterfactual model – – Susceptibility types Potential outcomes Confounding under the counterfactual susceptibility model of causation Stratification – – Identifying confounders Standardization versus pooling What is Confounding? Give me the definition you were taught or describe how you understand it What is an “adjusted” measure of effect? Is red wine cardio-protective? In an adjusted model to remove confounding of the E-D relationship, is it reasonable to remove variables that are not statistically significant and include those that are? Counterfactual Theory Potential Outcomes, Susceptibility types Poor Clare Doctor prescribes antibiotics 3 days later she is cured Did the antibiotic cure her? Cinema d’Counterfactual The counterfactual model: The counterfactual ideal Disease experience, given exposed Hypothetical disease experience, if unexposed The Counterfactual Ideal Counterfactual theory Only one can actually be observed – Ask, what would have happened had things been different, all other things being equal? – The other is “counterfactual” in that it is counter to what is actually observed Leads to the causal contrast Exposure must be changeable to have effect – We will come back to this The counterfactual model: The counterfactual ideal Approximation to The Counterfactual Ideal Disease experience, given exposed Substitute disease experience of truly unexposed Take home message 1: We’re often interested in what happens to index (exposed). Reference (unexposed) are useful only insofar as they tell us about index group. Must Specify a Causal Contrast Events are not causes themselves – Only causes as part of a causal contrast What is the effect of oral contraceptives on risk of death? – The question, as defined, has no meaning Compared – to condoms, increased risk Through stroke and heart attack Compared to no contraceptive, maybe decreased risk – Some places childbirth may be a greater risk Take home message 2: “Effects” of exposures only have meaning when defined in contrast to an alternative If ethics were not a concern, how would you design an RCT of smoking and lung cancer? Think about dose, duration What about obesity and MI? What about gender and cancer? Effects Must be Amenable to Action To have an effect, must be changeable – – What is effect of sex on heart disease? How would you change sex? Defining the action helps define the causal contrast well – – What is the effect of obesity on death? How would you change obesity? Each has a different effect, some good, some bad To remind us, use A for Action, not E Take Home Message 3: For etiologic observational studies, think of RCT you would do first. Develop your observational study with the RCT in mind. Think of the action, inclusion criteria, the placebo, etc. To identify a causal effect in an individual Need three things: – Call the counterfactual outcomes: – Ya=1 vs Ya=0, read: Y that would occur if A=a Note counterfactuals different from: – Outcome, actions compared, person whose 2+ counterfactual outcomes compared Y|A=1 (or just Y), read: Y given A=1 Effect can be precisely defined as: – Ya=1 ≠Ya=0 Assume infinite population with no information or selection bias, a dichotomous A and Y All examples, assume each person represents 1,000,000 people exactly the same as them so no random error problem Assume each person represents 100,000 people Person A Person B A (E) 1 1 Ya=1 1 1 Ya=0 0 1 Y 1 1 Person C Person D Person E 1 1 0 0 0 0 1 0 0 0 0 0 Person F Person G Person H 0 0 0 1 1 0 0 1 1 0 1 1 Effect : [Pr(Ya=1=1) - Pr(Ya=0=1)] = Assume each person represents 100,000 people Person A Person B A (E) 1 1 Ya=1 1 1 Ya=0 0 1 Y 1 1 Person C Person D Person E 1 1 0 0 0 0 1 0 0 0 0 0 Person F Person G Person H 0 0 0 1 1 0 0 1 1 0 1 1 Effect : [Pr(Ya=1=1) - Pr(Ya=0=1)] = [4/8 – 4/8] = 0 Association : [Pr(Y=1|A=1) - Pr(Y=1|A=0)] = Assume each person represents 100,000 people Person A Person B A (E) 1 1 Ya=1 1 1 Ya=0 0 1 Y 1 1 Person C Person D Person E 1 1 0 0 0 0 1 0 0 0 0 0 Person F Person G Person H 0 0 0 1 1 0 0 1 1 0 1 1 Effect : [Pr(Ya=1=1) - Pr(Ya=0=1)] = [4/8 – 4/8] = 0 Association : [Pr(Y=1|A=1) - Pr(Y=1|A=0)] = [2/4 – 2/4] = 0 The counterfactual model Susceptibility types Susceptibility Type Envision 4 responses to exposure, relative to unexposed – – – – Type 1 - Doomed Type 2 - E causal Type 3 - E preventive Type 4 - Immune CST: Counterfactual susceptibility type Exposed Outcome (Ya=1) Unexposed Outcome (Ya=0) Type 1 – doomed 1 1 Type 2 – E causal 1 0 Type 3 – E preventive 0 1 Type 4 – immune 0 0 The counterfactual model The index condition, relative to the reference condition, affects only susceptibility types 2 and 3 – – Types 2 get the disease, but would not get disease had they had the reference condition Types 3 do not get the disease, but would have got the disease had they had the reference condition Individual Susceptibility under the CST model Individual Susceptibility Risk Difference Risk Ratio Ya=1 – Ya=0 Ya=1 / Ya=0 Type 1 1–1=0 1/1=1 Type 2 1–0=1 1 / 0 = undef Type 3 0 – 1 = -1 0/1=0 Type 4 0–0=0 0 / 0 = undef Can type 2 and 3 co-exist? Are there exposures that can both prevent and causes disease? – – – – Vaccination and polio Exercise and heart attack Seat belts and death in a motor vehicle accident Heart transplant and mortality So what does RD = 0 or RR=1 mean? – – – Could mean no effect Could be balance of causal/preventive mechanisms We call no effect “sharp null” but it is not identifiable Take home message 4: If exposures can be causal and preventive, estimates of effect only tell us about the balance of causal and preventive effects Average causal effects Individual effects rarely identifiable because we don’t have both conditions – But average causal effects may be identifiable in populations An average causal effect of treatment A on outcome Y occurs when: – – Pr(Ya=1 = 1) ≠ Pr(Ya=0 = 1) Or more generally, E(Ya=1) ≠ E(Ya=0) Note makes no reference to relative vs. absolute Effects vs. Associations Effects measures – – – RD: Pr(Ya=1 = 1) - Pr(Ya=0 = 1) RR: Pr(Ya=1 = 1) / Pr(Ya=0 = 1) OR: Pr(Ya=1 = 1)/Pr(Ya=1 = 0)/ Pr(Ya=0 = 1)/Pr(Ya=0 = 0) Associational measures – – – RD: Pr(Y = 1|A=1) - Pr(Y = 1|A=0) RR: Pr(Y = 1|A=1) / Pr(Y = 1|A=0) OR: Pr(Y = 1|A=1) / Pr(Y = 0|A=1) / Pr(Y = 1|A=0) / Pr(Y = 0|A=0) Traditional Approaches to Confounding and Confounders Extend the CST model of causation to populations Susceptibility Type Index Outcome Reference Outcome Proportion in Index Pop Proportion in Reference Pop Type 1 – doomed 1 1 p1 q1 Type 2 – index causal 1 0 p2 q2 Type 3 – index preventive 0 1 p3 q3 Type 4 – immune 0 0 p4 q4 1 1 What is the risk of disease in exposed? Susceptibility Type Type 1 – doomed Index Outcome Observed Reference Outcome risk in exposed is p1 + p2, but we 1 cannot tell 1 how many of each Proportion in Index Pop Proportion in Reference Pop p1 q1 Type 2 – index causal 1 0 p2 q2 Type 3 – index preventive 0 1 p3 q3 Type 4 – immune 0 0 p4 q4 1 1 What would the risk of disease be in the exposed had they been unexposed? Susceptibility Type Index Outcome Reference Outcome Counterfactual risk is the risk the Type 1 – 1 1 doomed exposed would have had had they Type 2 – 1 0 index causalbeen exposed: p1+p3 Type 3 – index preventive Type 4 – immune Proportion in Index Pop Proportion in Reference Pop p1 q1 p2 q2 0 1 p3 q3 0 0 p4 q4 1 1 When can reference group stand in for the exposed had they been unexposed? To have a valid comparison, Susceptibility Index Reference Typerequire Outcome Outcome we the disease experience of reference Type 1 – be able to stand in for group 1 1 doomed the counterfactual risk. This Type 2– is partial exchangeability 1 0 index causal Proportion in Index Pop Proportion in Reference Pop p1 q1 p2 q2 Type 3 – index preventive 0 1 p3 q3 Type 4 – immune 0 0 p4 q4 1 1 Exchangeability Observed Counterfactual Full exchangeability means the two groups can stand in for each other – Risk exposed had = risk unexposed would have had if they were exposed Pr(Ya=1=1|A=1) – = Pr(Ya=1=1|A=0) Risk unexposed had = risk exposed would have had if they were unexposed Pr(Ya=0=1|A=1) = Pr(Ya=0=1|A=0) Exchangeability Observed Counterfactual Partial exchangeability means the E- can stand in for what would have happened to the E+ had they been unexposed – Risk unexposed had = risk exposed would have had if they were unexposed Pr(Ya=0=1|A=1) = Pr(Ya=0=1|A=0) Take Home Message 5: The unexposed have to be able to stand in for the exposed had they been unexposed. Not vice versa. Partial exchangeability Two possible definitions of no confounding (1) Definition One — the risk of disease due to background causes is equal in the index and reference populations p1 So p11 = q11 under this definition. The risk difference [(p1 + p2) - (q1 + q3)] equals (p2 assuming partial p2 –- qq3), 3 exchangeability. But effect should be based only on exposed Two possible definitions of no confounding (2) Definition Two -- the risk of disease in the reference population equals the risk the index population would have had, if they had been unexposed p1 + p3 p1 ++ pp3 So p1 +3qunder this definition. 3 ==qq1 1+ q The risk difference [(p1 + p2) - (q1 + q3)] equals (p2 p2 –- p3 p3 ), assuming partial exchangeability. NOTE that RD related to balance of p2 and p3 We choose the second definition First forces inclusion of effect of absence of exposure in reference group Second measures effect of exposure only in index group – – Holds under randomization However, it is counterfactual If exposure is never preventive, they are same We choose the second definition A measure of association is unconfounded if: – Experience of the reference group = the disease occurrence the index population would have had, had they been unexposed Risk difference tells about balance of causal/preventive action in index – Effect, not an estimate To put it mathematically Suppose we have two populations A and B We want to observe: IAE+ - IAEWe observe: IAE+ - IBEIf we add IAE- - IAE- to this we get: (IAE+ - IAE-) + (IAE- - IBE-) (IAE+ - IAE-) is the causal RD (IAE- - IBE-) is a bias factor (i.e. confounding) Bias is difference between counterfactual unexposed experience of exposed and experience of truly unexposed Causal RD vs. Observed Susceptibility Index Reference Type Frequency Frequency Causal RD? – Type 1 – doomed Type 2 – index causal Type 3 – index preventive Type 4 – immune 10 – 10 5 10 Observed RD? – – 10 5 75 – 100 100 (p1+p2) – (q1+q3) 15/100 – 15/100 = 0 Confounding? – 75 p2 – p3 5/100 – 10/100 = -5/100 Does (p1+p3) = (q1+q3) ? 20/100 ≠ 15/100, Yes Causal = Observed? – No Causal RD vs. Observed Susceptibility Index Reference Type Frequency Frequency Causal RD? – Type 1 – doomed Type 2 – index causal Type 3 – index preventive Type 4 – immune 10 – 5 5 10 Observed RD? – – 5 10 75 – 100 100 (p1+p2) – (q1+q3) 15/100 – 15/100 = 0 Confounding? – 80 p2 – p3 5/100 – 5/100 = 0 Does (p1+p3) = (q1+q3) ? 15/100 = 15/100, No Causal = Observed? – Yes Take Home Message 6: Lack of confounding doesn’t mean perfect balance of CST types which we would expect under randomization Take Home Message 7: If there is no confounding, the causal risk difference (i.e. the true effect) is the observed effect Assuming no other bias and random error Getting the observed contrast close to the counterfactual ideal Design – – Randomization Creating similar populations Matching Restriction Analysis – Stratification based methods – Stratification, MantelHaenszel, Regression Standardization based methods Standardization, Gestimation, IPTW, Marginal structural models Confounders Note we have defined confounding with no reference to imbalances in covariates – – Separate confounder from confounding Confounder is a factor that explains discrepancy between observed risk in reference and desired counterfactual risk Must be imbalanced in index/reference groups, a cause of disease and not on causal pathway – Use data as guide only Non-identifiability and collapsibility: Identifying confounding in practice Because we can’t identify individuals’ CST types, can’t use comparability definition in practice – – Instead a traditional approach uses the collapsibility criterion – Call this “ the non-identifiability problem” Except thoughtfully If crude measure equals adjusted for potential confounder, no confounding by that variable What adjusted measure of effect? Take Home Message 8: Confounding is when the unexposed can’t stand in for the exposed had they been unexposed. Confounders are variables that explain confounding. Stratified Analysis: Introduction One method for control of extraneous variables in the analysis – Advantages/disadvantages – – Analysis of disease-exposure association within categories of confounder / modifier prevents external influence of that variable Straight-forward, few statistical assumptions Data become thin with many categories/ variables Candidate variables – Confounders, Modifiers, Matched factors Stratified Analysis 1 Variable Stratify then ask: Are measures of effect within each stratum heterogeneous? – Yes = Interaction, stratified analysis? – No = No Interaction, assess confounding Does summary measure of effect across strata equal crude? – Yes = No confounding, collapse – No = Confounding, use summary measure Note, this is about change in estimate of effect, nothing about p-values Example (1-1) (CST balance within strata) cases non-cases total risk risk ratio All ages Men Women 150 70 850 930 1000 1000 0.15 0.07 2.1 Example (1-2) (CST balance within strata) Old cases non-cases total risk risk ratio Men 90 243 333 0.27 Women 60 607 667 0.09 3.0 Men 60 607 667 0.09 Young Women 10 323 333 0.03 3.0 Take home message 9: In practice, confounding USUALLY presents as – within levels of the confounder, uneven distribution of the exposure and different risk of outcome among unexposed But be careful, as this can be misleading as this is NECESSARY but not SUFFICIENT Example (1-3) (CST balance within strata) cases non-cases total risk risk ratio type1 type2 type3 type4 total confounding All ages Men Women 150 70 850 930 1000 1000 0.15 0.07 2.1 30 120 20 830 1000 -0.020 35 40 35 890 1000 Does p1 + p3 = q1 + q3? (30+20)/1000 <> (35+35)/1000 Example (1-4) (CST balance within strata) Old cases non-cases total risk risk ratio type1 type2 type3 type4 total confounding Men 90 243 333 0.27 Young Women Men Women 60 60 10 607 Does 607 323 667 667 333 p10.09 + p3 = q1 + q3 0.03 0.09 3.0 within strata? 3.0 20 70 10 233 333 0.000 30 30 30 577 667 10 50 10 597 667 0.000 5 10 5 313 333 Example (2-1) (choice of effect measure) cases non-cases total risk RR odds OR cases non-cases total risk RR MHRR odds OR MHOR E+ 100 900 1000 0.1 E50 950 1000 0.05 2 0.111 0.053 Collapsible? Does Collapsible? crude = adjusted? Outcome needs to be rare in levels of the Does all crude = adjusted? exposure/confounder 2.11 males E+ 90 110 200 0.45 E45 155 200 0.225 cases non-cases total females E+ 10 790 800 0.0125 2 E5 795 800 0.00625 2 2 0.818 0.290 0.013 2.82 0.006 2.01 2.68 Take Home Message 10: The odds ratio is not strictly collapsible. Change in estimate of effect after adjustment can be just an artifact of the data. Outcome must be rare in ALL strata. But this can go wrong Type 1 Type 4 total risk RR Type 1 Type 4 total risk RR MHRR E+ 100 100 200 0.5 E250 250 500 0.5 1 C+ E+ 60 40 100 0.6 E70 30 100 0.7 Counfounding? Is the exposure distribution different across strata? Does p1+p3=q1+q3? Is the risk in the unexposed different? Type I Type 4 total 0.86 CE+ 40 60 100 0.40 E180 220 400 0.45 .88 .87 Take Home Message 11: Statistical Criteria Are Not Sufficient to Determine What to Keep in a Model to Observe Causal Effects Pooled adjusted estimate Assumes uniform RR/RD across strata – Pooled estimates are weighted averages of effects in strata – – Precision enhancing Pooled estimate are between stratum estimates Weights measure information in strata (inverse variance) but can be computed differently Ex: Mantel-Haenzel, Logistic/Cox Reg – – So long as there are no interaction terms Regression models are analogous to stratification Review of weighting Pooling means we average the stratum specific estimates to get one estimate – Thus the pooled estimate must be between the two stratum specific estimates We can choose the weights however we like – Different weighting schemes have different properties and logics wR 1 weight ed RR wR 0 w w Example: MH Pooling Crude E+ p for heterogeneity 0.09 C1 E- C0 E+ E- E+ E- D+ 420 396 D+ 320 4 D+ 100 392 D- 1390 1785 D- 1200 45 D- 190 1740 Total 1810 2181 Total 1520 49 Total 290 2132 RR RR RR 1.3 2.6 1.9 420 / 1810 ˆ CrudeR R 1.3 396 / 2181 aN0 320* 49 100* 2132 N 1569 2422 ˆ MH RR 1.9 bN1 1520 290 4* N 392* 2422 1569 95% CI: 1.4, 2.7 Example: MH Pooling Crude C1 E+ ED+ 420 396 D1390 1785 Total 1810 2181 RR 1.3 E+ D+ 320 D- 1200 Total 1520 RR 2.6 C0 E4 45 49 E+ ED+ 100 392 D190 1740 Total 290 2132 RR 1.9 Weight is N1*N0/N which weights towards the strata with highest total N and most evenly distributed exposure distribution N 0 N1 a N *N 1 MH RRˆ N 0 N1 c N *N 0 N N N0 1 N 0 N1 aN1 320* 49 100* 2132 N 1569 2422 1.9 bN0 1520 290 4* N 392* 1569 2422 Mantel Haenszel Weights The weight, (N1*N0 )/ N is at its minimum if N1=1 so N0 = (N-1). Weight is then (N-1)/N which is about 1 The weight, (N1*N0 )/ N is at its maximum if N1= N0 = N/2. Weight is then (N/2)2/N which is N/4 So a larger sample size will increase the weight, as will an even distribution of exposed an unexposed subjects Summary estimates: Mantel-Haenszel A pooled summary estimate: – – Weighted average of estimates of effect from each stratum Weight is highest for stratum with most information (subjects) Precision optimizing Calculation depends on design Exposed (Index) MH estimatesafor g Cases Controls 3 Unexposed (reference) designs bg cg Case - control design ORMH dg ag d g g n g bg cg g n g Exposed (Index) MH estimatesafor g Cases Undiseased Total 3 Unexposed (reference) designs bg cg dg n1g n0g Risk design RRMH ag n0 g g n g bg n1g g n g Exposed (Index) MH estimatesafor g Cases Person-time 3 L1g Rate design IRRMH Unexposed (reference) designs bg L0g ag L0 g g L g bg L1g g L g Summary estimates: Standardized RR (SMR) Standardize the risk or rate – Choose index group because: – Weighted average of risk or rate in strata, using the index group’s experience as the weight Want reference group to reflect the rate we would have seen in the exposed had they been unexposed No assumption of homogeneity across strata 1.4, 2.7 1.7, 2.7 Example: Standardization 0.9, 0.7,6.4 2.5 Crude D+ DTotal RR C1 E+ E420 396 1390 1785 1810 2181 1.3 C0 E+ ED+ 320 4 D1200 45 Total 1520 49 RR 2.6 E+ ED+ 100 392 D190 1740 Total 290 2132 RR 1.9 420 / 1810 ˆ CrudeR R 1.3 396 / 2181 SMRˆ O N b * N 1 o 320 100 4 392 1520* 290* 49 2132 2.4 95% CI: 0.9, 6.4 Example: Standardization Crude D+ DTotal RR E+ E420 396 1390 1785 1810 2181 1.3 C1 E+ ED+ 320 4 D- 1200 45 Total 1520 49 RR 2.6 C0 E+ ED+ 100 392 D190 1740 Total 290 2132 RR 1.9 When we standardize, we can use whatever distribution we want. If we use the distribution of the exposed group, we call this an SMR. N1 * a N1 320 100 a 1520* 290* W a 1520 290 N 1 SMRˆ 2.4 b b b 4 392 N1 * N N1 * N N1 * N 1520* 49 290* 2132 0 0 0 W N1 * Example: Standardization Crude D+ DTotal RR C1 E+ E420 396 1390 1785 1810 2181 1.3 E+ D+ 320 D- 1200 Total 1520 RR 2.6 C0 E4 45 49 E+ ED+ 100 392 D190 1740 Total 290 2132 RR 1.9 We could also ask what would happen if everyone was both exposed and unexposed: corresponds to PO model W * a N1 320 100 a 1569* 2422* W 1520 290 N 1 SMRˆ 2.0 b b 4 392 W * N N * N 1569* 49 2422* 2132 0 0 W N* Exposed (Index) MH estimatesafor g Cases Controls 3 Unexposed (reference) designs bg cg dg a g Case - control design SMR g bg cg dg g Exposed (Index) MH estimatesafor g Cases Undiseased Total 3 Unexposed (reference) designs bg cg dg n1g n0g a g Risk design SMR g bg n1g n0g g Exposed (Index) MH estimatesafor g Cases Person-time 3 Unexposed (reference) designs bg L1g L0g a g Rate design SMR g bg L1g L0g g 1 1.3 1.3 Practical summary Use the RRc to measure the direction and magnitude of confounding: cRR = SMR*RRc RRc = cRR/SMR Use pooled estimates to maximize precision when effects are homogeneous within strata. Use the SMR as an unconfounded summary estimate when effects are heterogeneous 0.6 1.3 2.1 Practical summary Use the RRc to measure the direction and magnitude of confounding: cRR = SMR*RRc RRc = cRR/SMR Use pooled estimates to maximize precision when effects are homogeneous within strata. Use the SMR as an unconfounded summary estimate when effects are heterogeneous Take Home Message 12: Mantel-Haenszel is only appropriate when no interaction. Standardization can be used with interaction but isn’t precision optimizing. Conclusion Counterfactual model Causal contrast is between disease experience of exposed and counterfactual experience they would have had had they been unexposed Use unexposed group to stand in for counterfactual ideal Confounding occurs when the unexposed can’t stand in for exposed had they been unexposed