Causal Inference Causal Inference or or Truth in the Universe Truth in the Universe Importance of clinical Importance of clinical trials trials Major pitfalls in Major pitfalls in clinical trials clinical trials • Low power Low power • Not randomized Not randomized • Unblinded Unblinded • Incomplete follow-up Incomplete follow-up
32
Embed
Causal Inference or Truth in the Universe Importance of clinical trials Importance of clinical trials Major pitfalls in clinical trials Major pitfalls.
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Causal InferenceCausal Inferenceoror
Truth in the UniverseTruth in the Universe
Causal InferenceCausal Inferenceoror
Truth in the UniverseTruth in the Universe Importance of clinical trialsImportance of clinical trials Major pitfalls in clinical trialsMajor pitfalls in clinical trials
Untruth - spurious associationsUntruth - spurious associations• chance (small sample size)chance (small sample size)• bias (selection bias and other biases)bias (selection bias and other biases)
Truth - real associations, not always causalTruth - real associations, not always causal• effect - causeeffect - cause• effect - effect (confounding)effect - effect (confounding)• cause - effect (truth in the universe)cause - effect (truth in the universe)
Untruth - spurious associationsUntruth - spurious associations• chance (small sample size)chance (small sample size)• bias (selection bias and other biases)bias (selection bias and other biases)
Truth - real associations, not always causalTruth - real associations, not always causal• effect - causeeffect - cause• effect - effect (confounding)effect - effect (confounding)• cause - effect (truth in the universe)cause - effect (truth in the universe)
Estrogen and CHD in WomenEstrogen and CHD in WomenEstrogen and CHD in WomenEstrogen and CHD in Women
RQRQ: Does postmenopausal estrogen therapy : Does postmenopausal estrogen therapy reduce CHD risk in women?reduce CHD risk in women?
DesignDesign: Cross-sectional: Cross-sectional
SubjectsSubjects: 20 postmenopausal women - entire : 20 postmenopausal women - entire population of my Tuesday clinicpopulation of my Tuesday clinic
Estrogen and CHD in WomenEstrogen and CHD in WomenCross-Sectional StudyCross-Sectional Study
Estrogen and CHD in WomenEstrogen and CHD in WomenCross-Sectional StudyCross-Sectional Study
CHDCHD No CHD No CHD
EE
No ENo E
CHDCHD No CHD No CHD
EE
No ENo E
11 44
66 99
55
1515
131377 2020
RR = 0.5RR = 0.5
Estrogen and CHD in WomenEstrogen and CHD in WomenEstrogen and CHD in WomenEstrogen and CHD in Women
RQRQ: Does estrogen therapy reduce CHD risk?: Does estrogen therapy reduce CHD risk?
DesignDesign: Case-control: Case-controlCases:Cases: 1000 women admitted to SFGH over a 5-year 1000 women admitted to SFGH over a 5-year
period with discharge diagnosis of CHD (ICD-9 codes)period with discharge diagnosis of CHD (ICD-9 codes)
Controls:Controls: 1000 women identified by random digit dialing 1000 women identified by random digit dialing in SF who report no CHDin SF who report no CHD
MeasurementsMeasurements: CHD based on discharge : CHD based on discharge diagnosis; estrogen therapy based on self-reportdiagnosis; estrogen therapy based on self-report
RQRQ: Does estrogen therapy reduce CHD risk?: Does estrogen therapy reduce CHD risk?
DesignDesign: Case-control: Case-controlCases:Cases: 1000 women admitted to SFGH over a 5-year 1000 women admitted to SFGH over a 5-year
period with discharge diagnosis of CHD (ICD-9 codes)period with discharge diagnosis of CHD (ICD-9 codes)
Controls:Controls: 1000 women identified by random digit dialing 1000 women identified by random digit dialing in SF who report no CHDin SF who report no CHD
MeasurementsMeasurements: CHD based on discharge : CHD based on discharge diagnosis; estrogen therapy based on self-reportdiagnosis; estrogen therapy based on self-report
Estrogen and CHD in WomenEstrogen and CHD in WomenCase-Control StudyCase-Control Study
Estrogen and CHD in WomenEstrogen and CHD in WomenCase-Control StudyCase-Control Study
CHDCHD No CHD No CHD
EE
No ENo E
CHDCHD No CHD No CHD
EE
No ENo E
200200 300300
800800 700700
500500
15001500
1000100010001000 20002000
OR = .6; p = .01OR = .6; p = .01
Estrogen and CHD in WomenEstrogen and CHD in WomenEstrogen and CHD in WomenEstrogen and CHD in Women
RQRQ: Does estrogen therapy reduce CHD risk?: Does estrogen therapy reduce CHD risk?
DesignDesign: Case-control: Case-controlCases:Cases: 1000 women admitted to Kaiser over a 5-year 1000 women admitted to Kaiser over a 5-year
period with discharge diagnosis of CHDperiod with discharge diagnosis of CHD
Controls:Controls: 1000 women admitted to Kaiser over the same 1000 women admitted to Kaiser over the same 5-year period with no discharge diagnosis of CHD5-year period with no discharge diagnosis of CHD
MeasurementsMeasurements: CHD based on discharge : CHD based on discharge diagnosis; estrogen therapy based on diagnosis; estrogen therapy based on computerized pharmacy recordscomputerized pharmacy records
RQRQ: Does estrogen therapy reduce CHD risk?: Does estrogen therapy reduce CHD risk?
DesignDesign: Case-control: Case-controlCases:Cases: 1000 women admitted to Kaiser over a 5-year 1000 women admitted to Kaiser over a 5-year
period with discharge diagnosis of CHDperiod with discharge diagnosis of CHD
Controls:Controls: 1000 women admitted to Kaiser over the same 1000 women admitted to Kaiser over the same 5-year period with no discharge diagnosis of CHD5-year period with no discharge diagnosis of CHD
MeasurementsMeasurements: CHD based on discharge : CHD based on discharge diagnosis; estrogen therapy based on diagnosis; estrogen therapy based on computerized pharmacy recordscomputerized pharmacy records
Estrogen and CHD in WomenEstrogen and CHD in WomenCase-Control StudyCase-Control Study
Estrogen and CHD in WomenEstrogen and CHD in WomenCase-Control StudyCase-Control Study
Analysis:Analysis: Multivariate logistic regression - age, Multivariate logistic regression - age, ethnicity, education, blood pressure, diabetes, ethnicity, education, blood pressure, diabetes, smoking, alcohol, family history of CHD and smoking, alcohol, family history of CHD and hypercholesterolemiahypercholesterolemia
RQRQ: Does estrogen therapy reduce CHD risk?: Does estrogen therapy reduce CHD risk?
Analysis:Analysis: Multivariate logistic regression - age, Multivariate logistic regression - age, ethnicity, education, blood pressure, diabetes, ethnicity, education, blood pressure, diabetes, smoking, alcohol, family history of CHD and smoking, alcohol, family history of CHD and hypercholesterolemiahypercholesterolemia
Nurses’ Health StudyNurses’ Health StudyNurses’ Health StudyNurses’ Health Study
HormonesHormones NN PYARPYAR CHDCHD RRRR P-valueP-value
Complete Complete Follow-upFollow-up Comparability Comparability at the end of the trialat the end of the trial
Power of the PlaceboPower of the PlaceboPower of the PlaceboPower of the Placebo
Internal Mammary Artery Ligation for Angina Internal Mammary Artery Ligation for Angina In unblinded trials, IM ligationIn unblinded trials, IM ligation
• reduced angina 60%reduced angina 60% In blinded trials, reduced angina 65% inIn blinded trials, reduced angina 65% in
Internal Mammary Artery Ligation for Angina Internal Mammary Artery Ligation for Angina In unblinded trials, IM ligationIn unblinded trials, IM ligation
• reduced angina 60%reduced angina 60% In blinded trials, reduced angina 65% inIn blinded trials, reduced angina 65% in
• subjects who underwent IM ligation subjects who underwent IM ligation • subjects who underwent IM ligation subjects who underwent IM ligation
• subjects who underwent sham IM ligation subjects who underwent sham IM ligation • subjects who underwent sham IM ligation subjects who underwent sham IM ligation
Outcome = structured neurologic exam by Outcome = structured neurologic exam by blinded and unblinded neurologistsblinded and unblinded neurologists
More improvement with plasma exchange by More improvement with plasma exchange by unblinded, but not blinded assessmentunblinded, but not blinded assessment
Canadian Cooperative MS Trial Canadian Cooperative MS Trial 165 patients with multiple sclerosis165 patients with multiple sclerosis
Outcome = structured neurologic exam by Outcome = structured neurologic exam by blinded and unblinded neurologistsblinded and unblinded neurologists
More improvement with plasma exchange by More improvement with plasma exchange by unblinded, but not blinded assessmentunblinded, but not blinded assessment
Unintended effective interventionsUnintended effective interventions• participantsparticipants use other therapy or change behavior use other therapy or change behavior• study staff, medical providers, family or friendsstudy staff, medical providers, family or friends
Unintended effective interventionsUnintended effective interventions• participantsparticipants use other therapy or change behavior use other therapy or change behavior• study staff, medical providers, family or friendsstudy staff, medical providers, family or friends
HHeart and eart and EEstrogen-progestin strogen-progestin RReplacement eplacement SStudy tudy (HERS)(HERS)
HHeart and eart and EEstrogen-progestin strogen-progestin RReplacement eplacement SStudy tudy (HERS)(HERS)
2763 postmenopausal women < 80 years old 2763 postmenopausal women < 80 years old with documented CHD and a uteruswith documented CHD and a uterus
Randomized to CEE 0.625 mg plus MPA 2.5 Randomized to CEE 0.625 mg plus MPA 2.5 mg or identical placebomg or identical placebo
Followed every 4 months for 4.2 yearsFollowed every 4 months for 4.2 years Separate gynecology group managed bleedingSeparate gynecology group managed bleeding Outcome = nonfatal MI and CHD deathOutcome = nonfatal MI and CHD death
2763 postmenopausal women < 80 years old 2763 postmenopausal women < 80 years old with documented CHD and a uteruswith documented CHD and a uterus
Randomized to CEE 0.625 mg plus MPA 2.5 Randomized to CEE 0.625 mg plus MPA 2.5 mg or identical placebomg or identical placebo
Followed every 4 months for 4.2 yearsFollowed every 4 months for 4.2 years Separate gynecology group managed bleedingSeparate gynecology group managed bleeding Outcome = nonfatal MI and CHD deathOutcome = nonfatal MI and CHD death
NONO• generate important hypothesesgenerate important hypotheses• provide only answer if trial not feasibleprovide only answer if trial not feasible• generally produce correct answergenerally produce correct answer
But bias and confounding always worrisomeBut bias and confounding always worrisome Particularly problematic for interventions Particularly problematic for interventions
that require selection and adherencethat require selection and adherence
NONO• generate important hypothesesgenerate important hypotheses• provide only answer if trial not feasibleprovide only answer if trial not feasible• generally produce correct answergenerally produce correct answer
But bias and confounding always worrisomeBut bias and confounding always worrisome Particularly problematic for interventions Particularly problematic for interventions
that require selection and adherencethat require selection and adherence