It is 1950. Dr. L. Craven specializes in treatment of arthritis. Many of his patients receive aspirin and he notes that the incidence of myocardial infarction.

It is 1950. Dr. L. Craven specializes in treatment of arthritis. Many of his patients receive aspirin and he notes that the incidence of myocardial infarction (heart attack) is lower than expected in these patients.

Craven LL: Acetylsalicylic acid, possible preventive coronary thrombosis. Ann. West. Med. Surg. 4:95-99, 1950.

Craven LL: Experiences with aspirin (acetylsalicylic acid) in the nonspecific prophylaxis of coronary thrombosis. Miss. Valley Med. J. 75:38-44, 1953.

What type of study?What type of study?What type of study?What type of study?

20%20%20%20%20%1. Case series

2. Case-control

3. Retrospective cohort

4. Prospective cohort

5. Clinical trial

• 1529 patients with a history of myocardial infarction, most of them several years earlier.

• Randomly assigned (double-blind) to aspirin therapy - 324-mg tablet three times daily or placebo treatment.

• Length of follow-up ranged from 10-28 months (average 22 months).

Coronary Drug Project, 1976

CHD Mortality

35 723

49 722

758

771

Yes No

Aspirin

Placebo

How would you analyze these data?

RR = 35/758 = 0.73 (0.48-1.11) 49/771

p=0.14

Interpretation? Conclusions?

4.6%6.4%

Coronary Drug Project, 1976

Definite Non-fatal MI

28 730

32 739

758

771

Yes No

Aspirin

Placebo

RR = 28/758 = 0.89 (0.54-1.45) 32/771

p=0.65

Interpretation? Conclusions?

Coronary Drug Project, 1976

“This difference is suggestive of a beneficial effect for aspirin in the treatment of post-MI men, but not large enough to be conclusive.”

Coronary Drug Project, 1976

Followed arthritis patients on aspirin and determined MI, angina, stroke rates.

Compared to overall rates “in the community” (unpublished data).

Mayo Clinic Proceedings, 1978

Primary prevention in 473 patients with arthritis?

Overall, no difference in observed and expected rates of any outcome.

However, in men there were 30-50% reductions (but NSD).

MI in men: Expected 7.2; observed 5

(30% reduction)

A variety of case-control and cohort type studies with relatively small numbers of subjects suggested that aspirin might reduce coronary “events” by perhaps 20-30%, and it might be beneficial in primary prevention, but none of the observational studies had been conclusive.

What were the major limitations of all of these observational studies?

1. Small sample size; limited power to detect a modest difference

2. Potential biases: misclassification, selection bias, recall bias

3. Uncontrolled confounding: many other factors influence risk of heart disease

What should be done to determine whether aspirin is effective in primary prevention of cardiovascular disease?

How would you do it?

Treat (or cure) disease

Do combinations of 2 or 3 antiretroviral drugs prolong survival of AIDS patients better than single drugs?

Keep healthy people well

Does tamoxifen lower the incidence of breast cancer in women with high risk profile compared to high risk women not given tamoxifen?

Preventive RCTs

Therapeutic RCTs

New Drug TrialsNew Drug Trials

Phase I Preliminary information on safety

Phase II Efficacy; dosage Additional data on safety & side-effects

Phase III Advantage over existing treatments

Phase IV Post-marketing surveillance

Equipoise: Will the new treatment be better? Maybe; Maybe not.

Sufficient belief to justify exposing some subjects to new treatment.

Sufficient doubt to justify withholding new treatment from some subjects.

Back to Aspirin –

Is it ethical to do this?

Institutional Review Board ApprovalInstitutional Review Board Approval

Required for all research involving human subjects (not just clinical trials).

Benefits

Risks

“Human Research” means a systematic investigation involving living humans (including research development, testing and evaluation), designed to develop or contribute to generalizable knowledge.

Who should we invite to participate?

(age, gender, occupation?)

Inclusion/exclusion criteria?

Study or experimental population –

Potential participants (people considered for enrollment).

• Breast Cancer patients at BUMC• U.S. Physicians• U.S. Health Professionals

Selection of Study Population Selection of Study Population

Reference population – The general group to whom results should be applicable.• All humans• Women with breast cancer• Men at risk for MI

What groups will be compared?

Are you going to use a placebo?

Why or why not?

Is it ethical to use a placebo?

Why Use A Placebo? Why Use A Placebo?

• Minimize bias in assessing outcomes

• Makes groups as comparable regarding the perception of treatment.

• Allows “blinding” (not always possible).

Blinding (or Masking) Blinding (or Masking)

Minimizes bias in assessing outcomes

Single-blind: subjects unaware of treatment group Double-blind: subjects and investigators unaware

Exposure status is unknown.

Pain relief >20%

MinimalEffect

Total # Subjects

Glucosamine + Chondroitin

211 106 317

211/317 = .666 = 67/100

From Clegg DO, et al.:Glucosamine, chondroitin sulfate, and the two in combination for painful knee osteoarthritis. N Engl J Med 354:795, 2006.

Pain relief >20%

MinimalEffect

Total # Subjects

Placebo 188 125 313

Anti-inflam. drug 223 95 318

Glucosamine 203 114 317

Chondroitin 208 110 318

Glucosamine + Chondroitin

211 106 317

Glucosamine + Chondroitin: 211/317 = 67% Placebo: 188/313 = 60%

The “placebo effect”

From Clegg DO, et al.:Glucosamine, chondroitin sulfate, and the two in combination for painful knee osteoarthritis. N Engl J Med 354:795, 2006.

What can you conclude?How would you go about making a formal comparison?

How many should we invite?

• How many will have expected end points? Restrict study to persons at higher risk?

• Duration of Follow-up: Is there sufficient time to accumulate end points?

• Acute vs. long-term effects of treatment.

• Do you need multiple study sites?

Statistical power: the ability of a study to demonstrate a statistically significant association, if one exists.

Sample Size Estimates

What do we do after we have a group of willing subjects?

Informed consent?(What is informed consent)?

Informed about: Treatments Potential outcomes; risks/benefits Randomization (i.e., treatment is

not their choice) What is required of them

Informed ConsentInformed Consent

Obtain informed consent prior to randomization.

How should subjects be assigned to the comparison groups?

RandomizationRandomization

Method of assignment such that each individual has an equal chance of receiving each possible treatment.

Benefits/Strengths:• Unpredictability• Leads to comparability

Balance of known & unknown confounders Minimizes selection bias

Other methods have potential problems Subject self-selects Day of week, order of visit, etc.

Equal chance: Coin toss Random numbers (table or computer)

Large, carefully done randomized clinical trials are particularly useful for assessing moderate or small effects which may be clinically important.

1. Random assignment of a sufficiently large number of subjects provides exquisite control over both known and unsuspected confounding factors.

2. A large sample size increases the ability to identify modest, but significant differences.

Active Intervention Group

Reference Population

Experimental Population(potential participants)

Participants(willing and eligible)

Compliers Non-compliers

Comparison Group

Allocation

Non-participants

Population Hierarchy

Population Hierarchy

Compliers Non-compliers

Ok, so now the trial is under way and they plan to follow subjects for 8 years.

Anything else we should do while we are waiting for the results?

What about maintaining compliance (adherence to the protocol) and follow up?

Does compliance matter?

What steps can you take to maintain compliance and follow up?

How do you know if they have complied?

Maintaining Compliance & Follow-upMaintaining Compliance & Follow-up

Begin with a motivated, knowledgeable group. Identify subjects who are unlikely / unable to comply. Make the protocol as simple as possible. Be clear about what is involved. If possible, mask the treatment so all groups comply. Maintain frequent contact with subjects without interfering

with treatment. Provide incentives (free check-ups, transportation, etc.) Conduct a “run-in” or “lead-in” period before the real trial.

• Following assigned protocol for the duration of the study is crucial to ability to demonstrate a true effect.

• Noncompliance makes the groups more alike & reduces the ability to detect a difference (bias towards null).

Assessing ComplianceAssessing Compliance

Self-report

Pill counts

Biological measures (blood, urine)

Assess compliance, if possible.

Non-compliance Tends to Bias Toward the “Null”

Non-compliance Tends to Bias Toward the “Null”

0.3 0.5 1.0 2 3

“Null” means no difference

Relative Risk

What about monitoring for safety?

Data and Safety Monitoring Board

Data and Safety Monitoring Board

A scientifically & financially independent group with expertise in various disciplines charged with safeguarding participants.

Reviews progress of trial & data on outcome Interim results Adverse events

Can recommend modification or termination of all/part of trial. Treatment results so good cannot withhold from others. Treatment side effects/outcome so bad cannot continue.

Physicians' Health StudyPhysicians' Health Study

• Randomized, double-blind, placebo controlled trial.

• Primary prevention of CVD & cancer in 22,071 U.S. male physicians, aged 40-84 at baseline.

• A 2x2 factorial design to test:• 325 mg aspirin on alternate days• 50 mg beta carotene on alternate days

Physicians' Health StudyExclusion Criteria

Physicians' Health StudyExclusion Criteria

History of:• Heart attack• Stroke• Cancer• Current liver or kidney disease• Peptic ulcer or gout

Contraindication to aspirin Current use of aspirin or other drugs

affecting platelet function Current use of a vitamin A or carotene

Physicians' Health StudyPhysicians' Health Study

261,248 invitation letters sent with questionnaire

112,528 questionnaires returned

59,285 willing to participate

33,223 eligible and enrolled in 18-week run-in phase on active ASA & carotene placebo.

22,071 randomized

Randomization SchemeRandomization Scheme

22,071 U.S. Male Physicians

aged 40-84

-carotene5,517

-carotene placebo5,520

-carotene5,519

-caroteneplacebo5,515

Aspirin11,037

Aspirin Placebo11,034

Both Aspirin -carotene Neither

2x2 factorial design

How can we tell if randomization was successful?

What are some known risk factors for heart disease?

Were these confounding factors distributed equally?

Baseline Characteristics Baseline Characteristics

Aspirin Placebo(n=11,037) (n=11,034)

Age (yrs) 53.2 9.5 53.2 9.5Systolic BP (mmHg) 126.1 11.3 126.1 11.1Diastolic BP (mmHg) 78.8 7.4 78.8 7.4History of hypertension (%) 13.5 13.6History of high cholesterol (%) 17.5 17.3Cholesterol level (mg) 212.1 44.2 212.0 45.1History of diabetes (%) 2.3 2.2History of angina (%) 1.3 1.2Parental MI (%) 13.0 13.1

Baseline CharacteristicsBaseline Characteristics

Aspirin Placebo(n=11,037) (n=11,034)

Current smoking (%) 11.0 11.1Past smoking (%) 39.4 39.1Daily alcohol (%) 24.9 25.0Exercise >1/week (%) 71.7 71.2Body mass index (kg/m2) 24.9 3.1 24.9 3.0Multivitamin 19.9 19.9

Basic analysis is that of a cohort study (i.e. – compare incidence….)

How will we analyze the data?

Should we include all subjects in the analysis?

ComplianceCompliance

Placebo Group14.23% took aspirin or other

platelet-active drugs

Aspirin group85.71% took treatment

as intended

After 60.2 months

Who should we include in our analysis?

Secondary analysis: Compliers only. Not a randomized comparison.

Intention to Treat AnalysisIntention to Treat Analysis

Primary analysis: “Intention to treat” For the primary analysis all subjects should be

included in the groups to which they were randomly assigned, even if they did not complete or even receive the appropriate treatment.• Preserves baseline comparability & control of

confounding• Since compliers and non-compliers may be

systematically different it prevents bias.• It reflects efficacy in everyday practice.

Fatal Myocardial InfarctionFatal Myocardial Infarction

10 11,027

26 11,008

11,037

11,034

Yes No

Aspirin

Placebo

Use “Epi_Tools.xls” to calculate the risk ratio, the p-value, & 95% CI.

Interpret your findings in words.

Non-fatal Myocardial InfarctionNon-fatal Myocardial InfarctionNon-fatal Myocardial InfarctionNon-fatal Myocardial Infarction

126 10,911

213 10,821

11,037

11,034

Yes No

Aspirin

Placebo

Use “Epi_Tools.xls” to calculate the risk ratio, the p-value, & 95% CI.

Interpret your findings in words.

Follow-up After 60.2 Months

Morbidity follow-up 99.7%

Vital status only 0.3%

Mortality follow-up 100.0%

Compare Incidence of disease

aspirin

placebo

Stroke & Total CV DeathsStroke & Total CV Deaths

Aspirin PlaceboGroup Group RR 95% CI

Total Stroke 119 98 1.22 (0.93-1.60)

Total CV Deaths 81 83 0.96 (0.60-1.54)

More follow-up needed, but . . .

Data & Safety Monitoring Board Data & Safety Monitoring Board

On December 18, 1987, the Data & Safety Monitoring Board recommended early termination of the aspirin component of the trial.

Myocardial Infarction 125.9 216.6

Aspirin Placebo RR(/10,000) (/10,000)

0.58

What should we recommend?

Benefit vs. RisksBenefit vs. Risks

Stroke 107.8 88.8 1.2

Ischemic 82.4 74.3 1.1

Hemorrhagic 20.8 10.9 1.9

Upper GI ulcer 153.1 125.1 1.2

with hemorrhage 34.4 19.9 1.7

Bleeding 2699.1 2037.3 1.3

Transfusion need 43.5 25.4 1.7

Aspirin Placebo RR(/10,000) (/10,000)

Myocardial Infarction 125.9 216.6 0.58

What should we recommend?

Benefit vs. RisksBenefit vs. Risks

Stroke 107.8 88.8 1.2 19

Ischemic 82.4 74.3 1.1 8

Hemorrhagic 20.8 10.9 1.9 10

Upper GI ulcer 153.1 125.1 1.2 28

with hemorrhage 34.4 19.9 1.7 15

Bleeding 2699.1 2037.3 1.3 690

Transfusion need 43.5 25.4 1.7 18

Myocardial Infarction 125.9 216.6 0.58 -91

Aspirin Placebo RR RD(/10,000) (/10,000) (/10,000)

Benefit vs. Risks

Stroke 107.8 88.8 1.2 19

Ischemic 82.4 74.3 1.1 8

Hemorrhagic 20.8 10.9 1.9 10

Upper GI ulcer 153.1 125.1 1.2 28

with hemorrhage 34.4 19.9 1.7 15

Bleeding 2699.1 2037.3 1.3 690

Transfusion need 43.5 25.4 1.7 18

Myocardial Infarction 125.9 216.6 0.58 -91

Aspirin Placebo RR RD(/10,000) (/10,000) (/10,000)

Results Of The Physicians' Health StudyResults Of The Physicians' Health StudyResults Of The Physicians' Health StudyResults Of The Physicians' Health Study

Clear benefit of aspirin on risk of myocardial infarction in male physicians.

Unable to evaluate aspirin's effect on stroke or total cardiovascular mortality, due to early termination of the trial.

In primary prevention, aspirin should be prescribed on an individual basis by the health care provider as an adjunct to, not an alternative to, the management of other risk factors.

Sample size:

Only 1 out of 10 physicians over 40 years were women in 1982.

By age 60 women have only 1/3 the # of cardiovascular events as men.

Why Were Women Physicians Not Why Were Women Physicians Not Included in the Physicians' Health Study?Included in the Physicians' Health Study?

External ValidityExternal Validity

(Generalizability)(Generalizability)

Unable to say whether aspirin is effective in women or if the effect is different in women.

And…And…

The effect of aspirin did turn out to be different in men & women.

Low-dose aspirin reduced strokes in healthy women, but had no effect on heart attacks unless the women were >65 years old.

Double cheeseburger, large fries, jumbo coffee… oh, and an aspirin – gotta take care of the old ticker, ya know.

“Take an aspirin every day, but before you swallow it, take it for a five-mile walk.”

Community TrialsCommunity Trials

35 miles

Entire communities can be assigned to a treatment or intervention.

Does fluoridation of water prevent dental caries? (1944 - The Newburgh-Kingston Caries Trial)

Herd immunity in adults against influenza-related illnesses with use of the trivalent-live attenuated influenza vaccine (CAIV-T) in children. Pedro A. Piedra et al.: Vaccine 2005;23:1540–1548.

School children have the highest attack rates for influenza, and many observers believe that student proximity in classrooms promotes the spread of the influenza virus. Infected children then infect family members. Studies in Michigan and Japan suggested that influenza immunization programs in children reduced mortality and morbidity in adults, because school-to-home transmission was reduced. The study cited above tested this by conducting a childhood immunization program in two communities (Temple and Belton) and measuring subsequent rates of influenza in adults. About 20-25% of eligible children were vaccinated. The adult rates of influenza in the test communities were compared to adult rates of influenza in three communities that did not have childhood immunization programs (Waco, Bryan, and College Station).

Handwashing

“Every year, more than 3.5 million children aged less than 5 years die from diarrhea and acute lower respiratory-tract infection. These deaths are concentrated in low-income communities in developing countries. Several studies have shown that regular hand washing with soap reduces the incidence of diarrhea in children younger than 5 years in communities with a high incidence of diarrhea, although we are unaware of any reports of the effect of hand washing on acute respiratory-tract infections in settings where pneumonia is a leading cause of death. In developed countries, the promotion of hand washing has reduced respiratory-tract infections in several settings....”

What gap in the existing knowledge base on the control of serious infections of childhood is this excerpt describing?

What hypothesis needs to be tested?

“We undertook the Karachi Soap Health Study… to measure the broad health benefits brought about by improvement of handwashing and bathing with soap in settings where communicable diseases are leading causes of childhood morbidity and mortality.”

How would you design a study to accomplish this goal, specifically focusing on pneumonia?

– What epidemiologic study design would you use. – What would be the main exposure variable, and

what would be the main outcome variable? This study will be set in Karachi, the largest city in Pakistan. Briefly

describe the study population you would use for your study.

“Fieldworkers identified 42 candidate neighbourhoods separated by a street or market area. Eligible households were located in a candidate neighbourhood, had at least two children younger than 15 years (at least one of whom was less than 5 years old), and were planning to continue to reside in the same home for the duration of the study….”

“Overall, 25 neighborhoods were assigned to hand washing promotion and 11 to control.”

Within neighborhoods candidate households continued to be enrolled by random selection until there were 600 hand washing households (300 regular soap; 300 antibacterial soap) and 300 control households.

“Hand washing promotion was assigned at the neighbourhood level, because several activities promoting hand washing were undertaken at this level.”

It is notoriously difficult to get people to wash their hands thoroughly and consistently. (Disturbingly enough, physicians are among the worst offenders.) How would you consider promoting hand washing for this study?

“During the week before soap was distributed and disease surveillance was begun, fieldworkers gathered participants [in groups of 10-15 households] … for neighborhood meetings that lasted 30-45 minutes.”

They provided slide shows, showed videotapes and distributed pamphlets to illustrate health problems resulting from not washing and to provide specific hand washing instructions

For the first 2 months they held neighborhood meetings 2-3 times per week. In months 3-9 the meetings occurred weekly. Field workers also visited intervention households weekly to observe/correct hand washing procedures and to encourage adherence. Field workers encouraged all family members to wash after defecation, after cleaning an infant, before preparing food, before eating, and before feeding an infant. They also encouraged subjects to bathe once a day with soap and water.

“Field workers provided control families with a regular supply of books, notebooks, pens, and pencils to help their children’s education, but they gave no products that would be expected to affect rates of respiratory illness, diarrhea, or impetigo. Field workers neither encouraged nor discouraged hand washing in control households, and visited both control and intervention households with equal frequency.”

What data will you collect? What are the key comparisons you would make? What calculations will you do? How will you measure “association”?

Table to report pneumonia results for Karachi Soap Health Study

New cases of pneumonia Person time at risk

Plain soap # cases 20,009

Antibacterial soap # cases 20,191

No intervention (control) # cases 18,392

Incidence rates of pneumonia in children younger than 5 years

Antibacterial soap 4.21 per 100 person-weeksPlain soap 4.16 “Control 8.50 “

What do you conclude about the efficacy of hand washing in preventing pneumonia in young children?

Luby SP, Agboatwalla M, Feikin DR, Painter K, Billhimer W, Altaf A, Hoekstra RM. Effect of handwashing on child health: a randomized controlled trial. Lancet 2005;366:225-33.

Do these findings have any relevance for measures to deal with the threat of a flu pandemic?

What implications do these findings have for disease control in the US?

Hand washing is a simple, highly effective way to prevent common gastrointestinal and respiratory diseases!

Pneumonia Results for Karachi Soap Health Study

Comparison Data IRR

Antibacterial soap vs control 2.42/4.40 0.55

Plain soap vs control 2.20/4.40 0.50

Plain soap vs antibacterial soap 2.42/2.20 1.1

Compare the efficacy of handwashing in preventing pneumonia in infants under 1 year old, children 1-2 years old, and children 2-5 years old? What do you think might explain the differences between age groups?

“Provision of free soap, frequent community meetings, and weekly household handwashing promotion visits to all impoverished households worldwide is prohibitively expensive.

As public health professionals, how can we find cost-effective techniques for handwashing promotion?

Luby SP, Agboatwalla M, Feikin DR, Painter K, Billhimer W, Altaf A, Hoekstra RM. Effect of handwashing on child health: a

randomized controlled trial. Lancet 2005;366:225-33.

Compare the effectiveness of handwashing in preventing pneumonia in infants under 1 year old, children 1-2 years old, and children 2-5 years old? What do you think might explain the differences between age groups?

Summarize the data in this graph. What do you think explains what you see here?