Epidemiology quiz

Study Designs: Cohort Studies

David Celentano, ScD, MHS Johns Hopkins University

Exposure and Outcome

2

Main Types of Epidemiologic Study Designs

3

Study type Characteristics

Experimental • Studies prevention and treatment of disease • Investigator actively manipulates which groups receive the study agent

Observational • Studies causes, prevention and treatment for diseases • Investigator watches as natures takes its course

Cohort • Examines multiple health effects of an exposure • Subjects defined by exposure levels and follow for disease occurrence

Case-control • Typically examines multiple exposures in relation to a disease • Subjects are defined as cases and controls and exposure histories compared

Cross-sectional • Examine relationship between exposure and disease prevalence in a defined population at one point in time

Ecological • Examines relationship between exposure and disease with population-level data rather than individual data

The material in this video is subject to the copyright of the owners of the material and is being provided for educational purposes under rules of fair use for registered students in this course only. No additional copies of the copyrighted work may be made or distributed.

Design of a Cohort Study

Design of a Cohort Study

5

Design of a Cohort Study

6

Design of a Cohort Study

7

 Prospective cohort study of cannabis use, predisposition for psychosis, and psychotic symptoms in young people

 Mean four-year follow-up, cumulative lifetime incidence: 7.1%

An Example of Cannabis Use and Psychosis

8Source: Henquet C et al. (2004). BMJ;330:11.

Associations between Frequency of Cannabis Use at Baseline and Any Psychotic Symptoms; Figures Are Odds Ratios (95% Confidence Intervals)

9

Cumulative frequency of cannabis use

Unadjusted Adjusted *

None 1 1

<1/month 1.01 (0.55 to 1.86) 0.99 (0.53 to 1.84)

3–4 times/month 1.56 (0.91 to 2.68) 1.50 (0.86 to 2.62)

1–2 times/week 2.28 (1.28 to 4.09) 1.95 (1.07 to 3.55)

3–4 times/week 3.07 (1.49 to 6.31) 2.44 (1.16 to 5.13)

Almost daily 2.57 (1.52 to 4.34) 2.23 (1.30 to 3.84)

Linear trend 1.24 (1.15 to 1.35) 1.20 (1.10 to 1.31)

* Adjusted for age, sex, socioeconomic, urbanicity, childhood trauma, and predisposition for psychosis at baseline. † Reference category. ‡ Increase in risk with one unit in cannabis frequency.

Experimental vs. Observational Approach

10

Cohort Study

11

Disease develops

Disease does not develop

Totals Incidence

Exposed a

Not exposed

c

First, select:

Cohort Study

12

Disease develops

Disease does not develop

Totals

Exposed a b a + b

Not exposed

c d c + d

Then, follow to see whether:

First, select:

Cohort Study

13

Disease develops

Disease does not develop

Totals Incidence

Exposed a b a + b a

a +b

Not exposed

c d c + d c

c + d

Then, follow to see whether:

First, select:

a a + b = Incidence in exposed

c c + d = Incidence in non-exposed

Cohort Study: A Sample of Exposed and Non-Exposed

14

Total number of

participants Develop

CHD

Do not develop

CHD

CHD incidence per 1,000

people per year

Smoke cigarettes

3,000 84 2,916 28.0

Do not smoke cigarettes

5,000 87 4,913 17.4

Cohort Study: A Sample of Exposed and Non-Exposed

15

Total number of

participants Develop

CHD

Do not develop

CHD

CHD incidence per 1,000

people per year

Smoke cigarettes

3,000 84 2,916 28.0

Do not smoke cigarettes

5,000 87 4,913 17.4

84 3,000

= 0.0280 x 1,000 = 28.0 87

5,000 = 0.0174 x 1,000 = 17.4

 Berkey CS, Willett WC, Tamimi RM, Rosner B, Frazier AL, Colditz GA, Sep 2013, Breast Cancer Res Treat, 141(2):299–306

 doi: 10.1007/s10549-013-2686-8

Vegetable Protein and Vegetable Fat Intakes in Pre-Adolescent and Adolescent Girls, and Risk for Benign Breast Disease in Young Women

16

Girls Who Eat Peanut Butter May Face Lower Breast Cancer Risk Later in Life

17

 Relative risk =

 Atributable risk =

How Do We Estimate the Effect of a Factor on Disease Risk? Compare Incidence in Exposed and Non-Exposed

18

Incidence among exposed Incidence among non-exposed

Incidence among exposed — Incidence among non-exposed

 Relative risk = =

 Attributable risk =

=

How Do We Estimate the Effect of a Factor on Disease Risk? Compare Incidence in Exposed and Non-Exposed

19

Incidence among exposed Incidence among non-exposed

Incidence among exposed — Incidence among non-exposed

28.0/17.4 = 1.61

28.0 —17.4 = 10.6

 Relative risk: the ratio of the risk of disease or death among the exposed to the risk among the unexposed  RR = 1.0 no association  RR > 1.0 positive association, possibly causal  RR < 1.0 negative association, possibly protective

 Attributable risk: the risk of a disease or other outcome in exposed individuals that can be attributed to the exposure  Can be expressed as a risk (e.g., incidence rate) (10.6) -OR- as a proportion

10.6/28=0.38

How Do We Interpret Effect Measures?

20

 If start with a defined population and have a firm hypothesis NO  Johns Hopkins “precursors study” of medical students and health/disease outcomes

 If you have an exposure that you know you wish to examine YES  Machine-paced work vs. self-paced work on elevated blood pressure  Major advantage: assess exposures for disease incidence!

Do We Need to Know Exposure before Start?

21

The material in this video is subject to the copyright of the owners of the material and is being provided for educational purposes under rules of fair use for registered students in this course only. No additional copies of the copyrighted work may be made or distributed.

Cohort Studies and Time

2

What Do We Have Here?

Photo by Iris. (2013). “Cup of Green Tea.” Creative Commons BY-ND.

3

Is There a Protective Association between Incidence of Breast Cancer and Green Tea Consumption?

4

Cohort Study of Green Tea and Breast Cancer

If there was an association, we could expect to find…

5

Cohort Studies: Prospective vs. Retrospective

6

Timing of Follow-up

► Prospective (concurrent) ► Cohort is assembled now and followed into the future ► Obtain exposure information prior to determining endpoint status

► Retrospective (nonconcurrent/historical) ► Cohort was assembled in the past and followed forward in time up to the present ► Obtain endpoint status prior to or at the same time as exposure information

► Ambidirectional ► Combines prospective and retrospective

7

Timing of Follow-up: Summary

Source: Grimes and Schulz. (2002). Cohort studies: Marching towards outcomes. Lancet, 359(9303), 341–345. http://dx.doi.org/10.1016/S0140-6736(02)07500-1

8

Observational Study Designs

Study type Characteristics

Observational Studies causes, prevention, and treatment for diseases Investigator watches as nature takes its course

1. Cohort: Two types: 1) prospective; 2) retrospective

Examines multiple health effects of an exposure Subjects defined by exposure levels and followed for disease occurrence

2. Case-control Typically examines multiple exposures in relation to a disease Subjects are defined as cases and controls and exposure histories compared

3. Cross-sectional Examines relationship between exposure and disease prevalence in a defined population at one point in time

4. Ecological Examines relationships between exposure and disease with population-level data rather than individual data

9

Design of a Prospective Cohort Study— 1

10

Design of a Prospective Cohort Study— 2

11

Design of a Retrospective Cohort Study— 1

12

Design of a Retrospective Cohort Study— 2

13

Retrospective vs. Prospective Cohort

The material in this video is subject to the copyright of the owners of the material and is being provided for educational purposes under rules of fair use for registered students in this course only. No additional copies of the copyrighted work may be made or distributed.

Features of Cohort Studies

2

Types of Cohort Study Populations

► Single-site cohort

► Multicenter cohort

► “Collaborative cohorts”

► Population-based sample ► For an identified source population, a probability

sample may be taken

3

When Is a Cohort Study Warranted?

► Good evidence of an association between the disease and a defined exposure

► When exposure is rare, but incidence of disease among those exposed is high

► When the time between exposure and disease is short

► When attrition of study population can be minimized

► Remember, the results of a cohort study alone cannot tell us if a factor causes a disease, but it can give us important clues

4

Advantages of Cohort Studies

► Can assess several outcomes simultaneously

► Control of time and outcome measurements

► Less potential for bias than case-control studies but same potential for confounding

► Matching and masking possible

5

In a Cohort Study Starting with Exposed and Non-exposed, We Can Study Multiple Outcomes

6

In a Cohort Study Starting with a Defined Population, We Can Study Multiple Exposures and Multiple Outcomes

7

Disadvantages of Cohort Studies

► Requires large samples

► Requires long follow-up

► Not efficient for rare outcomes if exposure is low

► Financially costly

► Cohort members “choose” exposures themselves (not under our control)

8

Type and Interval of Data

► Classical (interval) cohort ► Timing: standard interval for collection

of study data (e.g., annually, semiannually)

► Data collection: follows standardized protocol

► Clinical cohort ► Timing: data collected at health care

visits ► Data collection: medical records and

laboratory tests taken as part of health care

The material in this video is subject to the copyright of the owners of the material and is being provided for educational purposes under rules of fair use for registered students in this course only. No additional copies of the copyrighted work may be made or distributed.

The Framingham Study and NA-ACCORD

Framingham Study Is a Classical Cohort Study

NA-ACCORD Is a Clinical Cohort Study

2

Framingham, Massachusetts

3

 Objectives  To study the impact of several factors on incidence of cardiovascular diseases

 Several exposures  Blood pressure  Smoking  Diet, etc.

 Multiple outcomes  Coronary heart disease (CHD)  Stroke  Congestive heart failure  Peripheral arterial disease

The Framingham Study

4

The Framingham Study: Main Hypotheses Tested

1. CHD increases with age; it occurs earlier and more frequently in males

2. Persons with hypertension develop CHD at a greater rate than those who are normotensive

3. Elevated blood cholesterol level is associated with increased risk of CHD

4. Tobacco smoking and habitual use of alcohol are associated with an increased incidence of CHD

5. Increased physical activity is associated with a decrease in the development of CHD

6. An increase in body weight predisposes to CHD

7. There is an increased rate of development of CHD in patients with diabetes mellitus

5

Derivation of the Framingham Study Population

6

Number of men

Number of women Total

Random sample 3,074 3,433 6,507

Respondents 2,024 2,445 4,469

Volunteers 312 428 740

Respondents free of CHD* 1,975 2,418 4,393

Volunteers free of CHD 307 427 734

Total free of CHD: the Framingham Study Group

2,282 2,845 5,127

Average Annual Incidence of CHD, by Systolic Blood Pressure

7

Average Annual Incidence of Coronary Heart Disease, by Weight Category and Age

8

CHD Risk Assessment based on HDL and LDL Cholesterol, Men 50–70 Years

Source: from Kannel, WB. (1990). Hosp. Pract., 25. 9

Average Annual Risk of MI per 10,000 by Age and Presence of Isolated Systolic Hypertension (ISH), Men 45–84 Years

Source: from Kannel, WB. (1990). Hosp. Pract., 25. 10

CHD Rate in Men >55 Years Old by Combination of Risk Factors

Source: Kannel WB. (1995). Circulation;92(11):3350–60. 11

1. A collaboration of North American HIV/AIDS cohorts and a data center for compilation of data to address HIV/AIDS research questions that cannot be accomplished through smaller cohorts

2. Focus on the failure of highly-active ART, with a special focus on multi- drug resistant virus and its consequences and management

3. Address additional scientific aims related to events that cannot be as well studied in smaller cohorts (for example, those that require large sample sizes, such as rare events from new HIV therapies, or those that require long-term follow-up, such as malignancy), and emerging issues in HIV clinical care such as the impact of aging on HIV treatment response

NA-Accord: a Clinical Collaboration

12

NA-ACCORD Sites in North America

13

NA-ACCORD: Virologic Failure

Source: Deeks et al. (2009). Clin Infect Dis. 14

Incidence of Second Failure: NA-ACCORD

15

Viral Suppression in NA-Accord 2000–2008

Source: Althoff KN, et al. (2012). Ann Intern Med. 16

1. The basic components and structure of cohort studies

2. Prospective (concurrent) and retrospective designs

3. Examples from the Framingham Study (“classical”) and NA-ACCORD (“clinical”), with a focus on “incident” (new) cases of outcomes

Summary

17

Main Types of Epidemiologic Study Designs

18

Study type Characteristics

Experimental • Studies prevention and treatment of disease • Investigator actively manipulates which groups receive the study agent

Observational • Studies causes, prevention and treatment for diseases • Investigator watches as natures takes its course

Cohort • Examines multiple health effects of an exposure • Subjects defined by exposure levels and follow for disease occurrence

Case-control • Typically examines multiple exposures in relation to a disease • Subjects are defined as cases and controls and exposure histories compared

Cross-sectional • Examine relationship between exposure and disease prevalence in a defined population at one point in time

Ecological • Examines relationship between exposure and disease with population-level data rather than individual data