Speaking Notes

PADM 5324

September 22, 2009

Dr. Neubauer

WHERE WE ARE:

• Are there "In the News" presentations this evening?
• I do not have a Homework One assignment for you at this time. There is not a Homework One assignment due next week.
• I will distribute a Midterm Examination soon.
• We can spend some time in class this evening practicing with the collaborative system of Google Docs pages.

CHAPTER 7 -- ASSESSING THE EFFICACY OF PREVENTIVE AND THERAPEUTIC MEASURES: RANDOMIZED TRIALS

In medical research, life and death really are often in the balance, at least for some people.

There are implication for the participants in research but for a far larger "audience" of others.

National Institutes of Health fund massive amounts of medical research.

http://www.nih.gov/

Until recently, their reviewers have favored INCREMENTAL approaches to the addition of new human knowledge to what is already known. Now there is a growing emphasis on funding VERY INNOVATIVE research that is HIGH RISK (in terms of findings) but with the change of MAJOR BREAKTHOUGHS.

http://www.fastercures.org/index.cfm/AboutFasterCures/Overview

Given that we have EXISTING TREATMENTS for almost every human medical condition, we must not assume that existing treatments are necessarily the best possible treatments.

We are interested not only in treatments but in PREVENTIONS and CURES.

Some people have an inherent FAITH IN INCREMENTALISM and a huge capacity to be PATIENT. There are at least TWO PROBLEMS. . .

1) sometimes you simply can't get there from here (Memphis)

2) patience is a luxury that dying people don't have

THIS IS NOT THE KIND OF RESEARCH I TEACH IN PADM 5502 INVOLVING "INFERENTIAL STATISTICS."

We are not studying the attributes, attitudes, or opinions of people.

Generally, we want to know if a new or alternative drug/treatment is more effective than other treatments, and, if so, for what kinds of patients. We are also interested in understanding SIDE EFFECTS.

I don't think a STUDY POPULATION refers to the entire population of people for whom the research may have implications. I believe it refers to the SAMPLE of people who will participate in the study in one way or another.

The idea of a CONTROL GROUP is important. You cannot both TREAT and NOT TREAT the same person at the same time. You generally cannot give the same patient two different treatments at the same time and figure out which one helped (or possibly killed) the patient. You need a BASIS OF COMPARISON. If you have two similar patients and one gets a new treatment and the other gets nothing (or an old treatment) you have some basis for saying what the affect of the new treatment was.

Without a comparison group you may "leap to conclusions" of the kinds described on page 133 of our textbook. Just because you did something and something happened does not mean that what you did produced the observed result.

Figure 7.1 from textbook -- Design of a randomized trial.

HOW ABOUT USING "HISTORICAL CONTROLS?"

• This may provide useful insights.
• But you have no control over what data was historically collected. Research data is often not exactly the same as clinical data.
• Other things may have changed over time which may compromise your ability to draw conclusions based upon comparisons of "historic" patients and newly treated participants in your research.

HOW ABOUT KNOWINGLY SELECTING PEOPLE TO BE IN THE TREATMENT GROUP

• Selection bias becomes an issue.
• Person making the selection may not be aware of his/her bias.
• Truly random selection "should" remove all kinds of possible selection bias.
• The researcher's first priority is the objectivity of the findings.
• A physician's first priority is his/her perception of the best interests of particular patients.

It is not ethical to deny patients treatment because they are agreed to participate in research.

It may not be ethical to give patients a placebo if it is known that there is a drug that can help them.

You cannot force patients to do what you tell them to do, even if they signed an informed consent.

Sick people are by definition a VULNERABLE population. Prisoners, for example, are protected by IRB rules. I do not know what the current standards are regarding medical research involving soldiers.

Note about random methods of selection: If there are few people in the "study population" there is no reason to believe that random selection will produce two groups that are comparable "in every possible way." THE LARGER THE "STUDY POPULATION" the more likely it is that RANDOM SELECTION will produce comparable (similar) groups that are similar, "in every possible way."

If you don't want to use a random method for some reason you can INTENTIONALLY PRODUCE groups that have similar percentages of composition regarding specific attributes. This is called STRATIFIED RANDOMIZATION. If you decide that AGE and GENDER are important variables you can FORCE the sample to have "THE RIGHT PERCENTAGES" by age and my gender. This is sometimes referred to in research methods as a QUOTA SAMPLE. The resulting groups LOOK AS IF THEY WERE RANDOMLY selected but they were not.

THE IMPORTANCE OF MASKING (BLINDING)

We don't want the participants to know which group they are in. Their knowing my affect the outcome. Medicine is full of mind-body issues.

We don't want the people taking the measurements of outcome to know either. Their preferred outcomes (of the research) may influence how they interpret their measurements.

CROSSOVER

Crossover means that all participants start with one kind of treatment and then switch to the other treatment. This only works if each treatment has only an immediate and not-lasting effect.

FACTORAL DESIGN

Apparently, you can sometimes use the same two groups (treatment and control) to conduct two different drugs (or other treatments) at the same time, so long as their two MODES OF ACTION are different. Given that there may be complex INTERACTION AFFECTS that cannot be anticipated, this seems to me to be an effort to gain ECONOMY and may produce some risks and possibly some misleading findings.

PARTICIPANTS SOMETIMES "DROP OUT" AND/OR BECOME NONCOMPLANT. The researcher cannot prevent this.

Sometimes a study is modified in the MIDDLE OF EXECTION because the pattern of outcomes is so apparent that it become UNETHICAL to continue.

CONCLUSION

• Randomized clinical trials ate the "gold standard" of medical study designs.
• It appears to me that the research is essentially MEDICAL and that the possible applications are both in the practice of MEDICINE and in PUBLIC HEALTH.
• Some additional issues are address in Chapter 8.