Post on 22-Dec-2015
OSLER JOURNAL CLUBCOHORT STUDY
8/12/09
Racial Differences in Incident Heart Failure among Young Adults
Bibbins-Domingo K, et al. N Engl J Med 360(12):1179-90
Presented by: Cristina Alewine , Raymond Givens, Zoe OreckiFaculty Advisor: J. Hunter Young
Cohort Study
Observational Group of subjects followed over time Non-randomized Compares differences in outcomes
between groups Types of cohort studies
Prospective Retrospective Nested case-control Household panel survey
Cohort Study Design
Defined Population
Exposed
Develop Disease
Do Not Develop Disease
Non-exposed
Develop Disease
Do Not Develop Disease
Group A Group B
Cohort Study Limitations
Expensive Time-consuming Attrition Biases
Assessment bias due to lack of blinding Information bias Bias due to attrition Analytic bias
Lack of causal inference: confounding
Cohort Study Strengths
Can define incidence and possible causes of a condition
Efficient for rare exposures Can establish timing of exposure to
outcome Allow study of outcome when
randomization to exposure is unethical or impractical
Heart Failure Epidemiology 5.7 million Americans with HF 670,000 new cases diagnosed each year U.S. mortality rate related to HF
estimated at 20.2 deaths per 100,000 HF prevalence increases with age Prevalence and etiology differ by
ethnicity and gender HF incidence twice as high among older
African-American as among older Caucasian
American Heart Association: Heart Disease and Stroke
Statistics
Bibbins-Domingo K, et al. N Engl J Med 360(12):1179-90
HF Risk FactorsNHANES I
010203040506070 61.6
17.110.1 9.2 8.9 8.9 8
3.1 2.2
Risk Factor
Popu
lati
on
att
ributa
ble
ri
sk (
%)
Modified from: He J, et al. Arch Intern Med 161:996, 2001
HF Prevalence by Age and Gender
NHANES III
20-24 25-34 35-44 45-54 55-64 65-74 75+0
1
2
3
4
5
6
7
8
9
10
0.1 0.10.7
1.8
6.26.8
9.8
0.1 0.10.5
1.3
3.4
6.6
9.7
MenWomen
Perc
en
t of
popu
lati
on
(%)
American Heart Association: Heart Disease and Stroke Statistics
HF Prevalence by Ethnicity
From: Yancy CW. Heart Failure in African Americans. Am J Cardiol 2005;96[suppl]:3i-12i
Heart Failure Epidemiology
• Limited data about HF incidence among people younger than 50
Better understanding of HF among young adults needed for improving targeting of screening and treatment
Friedman GD, et al. CARDIA: Study design, recruitment, and some characteristics of the examined subjects. J Clin Epidemiol 1988;41:1105-16.
CARDIA
Coronary Artery Risk Development in Young Adults
Prospective Cohort- initiated in 1984
“Initiated to investigate life-style and other factors that influence , favorably or unfavorably, the evolution of coronary heart disease risk factors during young adulthood.”
Friedman GD, et al. CARDIA: Study design, recruitment, and some characteristics of the examined subjects. J Clin Epidemiol 1988;41:1105-16.
CARDIA- Recruitment
Population Goal: Obtain a representative sample of underlying
population of black and white adults aged 18 to 30 years
Stratify to achieve equal numbers by race, gender, age, education
Centers: Birmingham Chicago Minneapolis OaklandFriedman GD, et al. CARDIA: Study design, recruitment, and some characteristics of the examined subjects. J Clin Epidemiol
1988;41:1105-16.
CARDIA- Eligibility
Age - 18-30 years at initial telephone recruitment interview - initial exam before 31st birthday
Race Residence Health/Medical - “free of long-term disease or disability” - excluded if pregnant or up to 3 months post-partum Other
- excluded if “unsuitable subjections” emotional instability, drug effects, or hostility
Friedman GD, et al. CARDIA: Study design, recruitment, and some characteristics of the examined subjects. J Clin Epidemiol 1988;41:1105-16.
CARDIA- Design
Brief Screening Telephone Interview 16 Questions-
Verification Demographics Medical Eligibility
CARDIA Exam Additional Questionnaires
Sociodemographics, Medical, Psychosocial Interviews
A/B Behavior Patterns, Diet Phlebotomy Blood Pressure Pulmonary Function Testing Anthropometry Treadmill Test
Friedman GD, et al. CARDIA: Study design, recruitment, and some characteristics of the examined subjects. J Clin Epidemiol 1988;41:1105-16.
CARDIA- Participants
Friedman GD, et al. CARDIA: Study design, recruitment, and some characteristics of the examined subjects. J Clin Epidemiol 1988;41:1105-16.
CARDIA- Participants
Friedman GD, et al. CARDIA: Study design, recruitment, and some characteristics of the examined subjects. J Clin Epidemiol 1988;41:1105-16.
CARDIA- Time Line
CARDIA Examination at Baseline and 2, 5, 7, 10, 15, and 20 years
Transthoracic Echo at 5 years
Hospitalizations
Deaths at 6 month intervals
0 2 5 7 10 15
20ECHO
Friedman GD, et al. CARDIA: Study design, recruitment, and some characteristics of the examined subjects. J Clin Epidemiol 1988;41:1105-16.
Bibbins-Domingo et al. (2009). Racial Differences in Incident Heart Failure among Young Adults. NEJM. 360:12- 1179-1190.
Study Cohort Retention
Retention at Year 20 Telephone Interview 87.5% Examination 71.8%
Noted- Black Men most likely to be lost to follow-up.
However statistics not supplied by authors.
Bibbins-Domingo et al. (2009). Racial Differences in Incident Heart Failure among Young Adults. NEJM. 360:12- 1179-1190.
CHF Related- End Points
questioned about overnight hospitalizations records requested in cases of suspected cv
events classified as heart failure if
physician diagnosis medical treatment (diuretic and digitalis or after-load reducing agent)
deaths reported at 6 month intervals
records requested after getting consent from next of kin
Classified as heart failure if appropriate ICD-9Bibbins-Domingo et al. (2009). Racial Differences in Incident Heart Failure among Young Adults. NEJM. 360:12- 1179-1190.
Heart Failure Incidence by Race and Gender
0.9%
1.1%
0 %
0.08%
Bibbins-Domingo et al. (2009). Racial Differences in Incident Heart Failure among Young Adults. NEJM. 360:12- 1179-1190.
20 yr Risk of Heart Failure Based on Demographic Measures
Bibbins-Domingo et al. (2009). Racial Differences in Incident Heart Failure among Young Adults. NEJM. 360:12- 1179-1190.
BP, HTN, BMI, DM, HDL and CKD Increased in Participants with Heart
FailureWhite Black participants
Blacks +HF vs.All Participants No HF ***p <0.001, ** <0.01, *<0.05Blacks +HF vs. Blacks No HF ### p <0.001, ## <0.01, 0.05
##
***###
***###***###***###
**#**##
**###
Prevalence of HTN in Participants with HF
Bibbins-Domingo et al. (2009). Racial Differences in Incident Heart Failure among Young Adults. NEJM. 360:12- 1179-1190.
Lower EF and Worse Systolic Fxn Seen in Pts with HF
*#
*#
*#
Blacks +HF vs.All Participants No HF ***p <0.001, ** <0.01, *<0.05Blacks +HF vs. Blacks No HF ### p <0.001, ## <0.01, 0.05
White Black participants
Bibbins-Domingo et al. (2009). Racial Differences in Incident Heart Failure among Young Adults. NEJM. 360:12- 1179-1190.
20 yr Risk of Heart Failure Based on Echo Measurements at Year 5
Not statistically significant in Multivariate Model Adjusted for Clinical Measures
Conclusions of the Study
Racial disparity in development of early HF Rates of HF in white pts confirmed earlier
studies Risk factors for heart failure in black pts:
Elevated blood pressure Obesity Chronic kidney disease Systolic dysfunction in early adulthood
Need aggressive screening and intervention in young patients at risk
Need studies to determine best ways to intervene
VALIDITY:
Should we believe the results?
YES ISSUES
Large study size Big Association Long observation Standardization Specific risk factors
associated Result makes sense
given prior studies
Differential drop-out Reliance on self-
report Misdiagnosis
Confounded by Chronic Kidney Disease
Missed cases The missing risk
factors: LDL Cocaine
Chronic Kidney Disease
Heart failure or kidney failure? Hospitalizations (N= 23)
n= 9 kidney dysfunction as a co-existing condition
and 3 of these are ESRD Deaths (n= 5)
n= 1 kidney dysfunction as a co-exisiting condition
and it is classified as ESRD
Missed Cases?
Unreported hospitalizations Subclinical cases
Diagnosis based on review of hospital admissions
Excludes diagnoses in clinic Why not review med lists for drugs like lasix or
digitalis that would suggest failure? Bias
Are the persons on the reviewing committee more likely to diagnose HF in black vs. white patients?
GENERALIZABILITY:
Can results apply to everybody?
YES SOME ISSUES
Multiple study centers
Men and women Black and white
subjects Varied socio-
economics Varied educational
background
Does not give info on HF cases by location
Non-black minority groups excluded
Excludes “unsuitable subjects”
What does this mean in clinic?
“Our data suggest that the number of young, black patients with hypertension that would need to be treated to prevent one case of heart failure before 50 years of age could be as low as 21.”
Evidence of causality
Temporal association Strong association Dose-response Consistency/replication Biologic plausibility No alternate explanation (confounding) Cessation of exposure Specific association
Types of Studies
Trial: Cohort assembled and exposure assigned, usually by randomization
Cohort study: Cohort assembled and followed over time. Exposures are measured.
Case-control study: Subjects selected based on presence or absence of disease
Cross-sectional study: Exposures and outcomes measured at one point in time
From Journal to Bedside
Internal validity: Is the association real and causal?
External validity (generalizability): Do the findings apply to other populations (your patient)?
Statistical significance: It’s unlikely the results occurred by chance
Clinical Significance: Findings are compeling enough to influence your practice
Internal Validity: Sources of error
Bias: Association not real due to systematic error Selection bias Information bias
Chance: Association not real due to random error Small sample size Subgroup analyses
Confounding: Real association; wrong inference Grey hair associated with heart disease
Study type: Trials
Strength: validity Trials provide the stongest evidence of causation
Key: the exposure is assigned, usually through randomization
Weaknesses May not be generalizable
Volunteers Clinically homogeneous Ideal setting (extraneous factors controlled)
Expensive Short duration Bias: Minimize by blinding participants & staff
Study type: Cohort Studies
Strengths Long duration of follow-up Temporal association of exposure with outcome Increased generalizability
Weaknesses: Validity Confounding
Factor related to exposure and outcome Exposure is often a choice (diet, exercise, drug)
Bias Assessment of outcome or exposure can be unduly
influenced by factors unrelated to disease process
Study type: Cross-Sectional Studies
Strengths: Efficient Can address prevalence
Weaknesses: Validity
Confounding Bias
Survivor bias Reverse causality
Cannot address incidence
Study type: Case-Control Studies
Strengths: Efficient
Weaknesses: Validity
Confounding Bias:
Selection bias Recall bias
Cannot address prevalence or incidence
Current Article Bibbins-Domingo et al. NEJM 2009; 360:1179-90 Study question: Association of ethnicity with heart
failure in young adults Results: Young African Americans have greater risk of
heart failure than young Americans of European descent Internal validity:
Is the association real? Yes, but with following caveats Differential drop outs: probably underestimated incidence in AA
men Authors could have assessed effect using baseline characteristics
Diagnostic bias: Ethnicity may have influenced probability of naming a clinical scenario as heart failure
Differential access to care: European-Americans may have been diagnosed in clinic more often
Subclinical heart failure was not assessed and may account for a substantial portion of heart falure cases underestimating incidence
Current Article Internal validity: (continued)
Is the association confounded? Renal disease: High prevalence in African Americans and
could both lead to and mimic heart failure (volume overload) External Validity:
Those more likely to be loss to follow-up were excluded Statistical significance: No question here. Just
lack of power to further explore predictors Clinical significance: Not sure these findings were
not unexpected. Incidence is still low complared to renal disease. Another reason to be aggressive with blood pressure control (although this is extrapolating from the data)