1. 1954 Salk vaccine field trials
Biggest public health experiment ever
Polio epidemics hit U.S. in 20th century
Struck hardest at children
Responsible for 6% of deaths among 5-9 year olds

3. Salk vaccine field trial
Polio is rare but virus itself is common
Most adults experienced polio infection without being aware of it.
Children from higher-income families more vulnerable to polio! Paradoxical, but
Children in less hygienic surroundings contract mild polio early in childhood while still protected from mother’s antibodies. Develop immunity early.
Children from more hygienic surroundings don’t develop such antibodies.

4. Salk vaccine field trial
By 1954, Salk poliomyelitis vaccine was promising
Public Health Service and National Foundation for Infantile Paralysis (NFIP) ready to try the vaccine in population
Vaccine could not be distributed without controls
A yearly drop might mean the drug was effective, or that that year was not an epidemic year.
Some children would get vaccine, some would not
Raises question of medical ethics

5. Salk vaccine trial Polio rate of occurrence about 50 per 100,000
Clinical trials needed on massive scale
Suppose vaccine was 50% effective and 10,000 subjects in control group, 10,000 subjects in treatment group
Would expect 5 polio cases in control group and 2-3 in treatment group
Difference could be attributed to random variation
Ultimate experiment involved over 1 million

6. How to design the experiment
Treatment and control groups should be as similar as possible
Any difference in response should be due to the treatment rather than something else
Taking volunteers biases the experiment
Fact: volunteers tend to be better educated and more well-to-do than those who don’t participate
Relying on volunteers biases the results because subjects are not representative of the population
Definition: A study is biased if it systematically favors certain outcomes

7. NFIP study: Observed Control approach
Offer vaccination (treatment) to 2nd graders
Control group: 1st and 3rd graders
Three grades drawn from same geographical location
Advantage: Not much variability between grades
Objections:
Uncertainty of the diagnostic process
Selective use of volunteers

8. NFIP Observed Control study
In making diagnosis physicians would naturally ask whether child was vaccinated
Many forms of polio hard to diagnose
Borderline cases could be affected by knowledge of whether child was vaccinated
Volunteers would result in more children from higher income families in treatment group
Treatment group is more vulnerable to disease than control group
Biases the experiment against the vaccine

9. Randomized control approach
Subjects randomly assigned to treatment and control groups
Control group given placebo
Placebo material prepared to look exactly like vaccine
Each vial identified only by code number so no one involved in vaccination or diagnostic evaluation could know who got vaccine
Experiment was double-blind, neither subjects nor those doing the evaluation knew which treatment any subject received

10. Results of vaccine trials
The randomized, controlled experiment
Size
Rate (per 100,000)
Treatment
200,000 28
Control 71
No consent
350,000 46
The NFIP/Observed Control study
Size
Rate (per 100,000)
Grade 2 (vaccine)
225,000 25
Grade 1, 3 (control)
725,000 54
Grade 2 (no consent)
125,000 44
Source: Thomas Francis, J r., “An evaluation of the 1954
Poliomyelitis vaccine trials---summary report,” American Journal
of Public Health vol 45 (1955) pp

11. Are the results significant?
Results show NFIP study biased against vaccine
Chance enters the study in a haphazard way: what families will volunteer, which children are in grade 2, etc.
For randomized controlled experiment chance enters the study in a planned and simple way: each child has chance to be in treatment or control
Allows for use of probability to determine if the results are significant

12. Are the results significant?
Two competing positions
1: The vaccine is effective.
2: (Devil’s Advocate) The vaccine has no effect. The differences between the two groups is due to chance.
Probability to the rescue: Suppose vaccine has no effect. What are the chances of seeing such a large difference in the two groups?
We’ll do the calculations in a few weeks. But they are a billion to one against!
Definition: An outcome is statistically significant if the effect is so large that it would rarely occur by chance

13. Basic principles of statistical design of experiments
Randomization
Uses chance to assign subjects to treatments
Tends to prevent bias, or systematic favoritism, in experiments
Replication
Repeating the treatment on many subjects reduces role of chance variation
Comparison and Control
Compare treatments to prevent confounding effect of treatment with other influences
Also tends to prevent bias

14. More complex designs
Blocking allows for greater control of influential variables
Perhaps vaccine works differently on men and women
Set up separate “blocks” of men and women
Carry out randomization separately within
each block
Allows for separate conclusions about each block

15. Matched pairs design Special case of blocking
Pair up individuals or apply two treatments to same individual
Often used for before-and-after studies
Example: Effectiveness of a diet using weights of subjects measured before and after the diet treatment

16. Randomization How to assign subjects at random
Pick from a hat, computer generator, tables of random numbers
Table B:
Line .
Simulates random digits
Every position has “the same probability” of being 0, 1, , 9
The digits in any position have “no influence” over the digits in any other position (e.g., they are “independent” of each other)
Doesn’t matter whether you pick a row, a column or a block as long as you do so consistently and without peeking

17. Using Table B Assign numerical labels to population
00 Alex Baum
01 Tim Blaha
02 Rachel Br.
03 Kari Chr.
04 Danielle D.
05 Alex Eis.
06 Patricia Glab
07 Danny Gr.
08 Vince Huang
09 Nita Jain
10 Peter Juul
11 Mou Khan
12 Asad Khawar
13 Lyndsey Kl.
14 Jamie Ly.
15 Dan Masello
16 Alicia Mazzara
17 Martha Mont.
18 Jesse Moreno
19 Oyeyinka Oy.
20 Ginger Price
21 Claire Rich.
22 Leon Schneider
23 Richard South.
24 Hikaru Ter.
25 Jacob Titus
26 Jesse Trent.
27 Andrew Tulloch
28 Bryce Turner
29 Jonathan White
30 Matthew Yelle
Assign numerical labels to population
Start anywhere in Table B and read off groups of numbers
Example: Pick random sample of 5 students
Label students from 00 to 30
We used 2 digits to label so pick digits in pairs.
E.g., from Line 116 pick . . .