Speaking Notes

PADM 5324

November 3, 2009

Dr. Neubauer

 

CHAPTER 13 -- Review of Cohort and Case-control Studies

 

There are really three approaches:

 

 

In a cohort study you identify a population and then watch for exposures and incidents of disease.  The comparison is between people exposed and people not exposed.

 

In a case-control study you begin with a group of diseased persons and then define a control group.  You then try to figure out which exposures may have caused the disease.  In other words, what things were the diseased group exposed to that the control group was exposed to?  The selection of the members of the control group is a serious methodological issue. 

 

BENEFITS OF EACH APPROACH

 

 

 

Good things

Not so good things

Cohort

Less dependent upon ability of participants to recall things.

 

Requires large number of participants

Is relatively costly

Takes years to complete

Not useful if the disease is rare

Not useful is the exposure is rare

Participants may drop out

Case-Control

Relatively inexpensive

Requires relatively few participants

Useful when instance of the disease is rare

 

The findings will depend upon how the controls were selected

Data collection depends upon ability of participants to remember and report possible exposures

Nested Case-Control

Recall bias eliminated regarding anticipated exposures

Lab tests only required on those selected as cases and controls

???

Cross-sectional study design

Based on survey research in which data on exposures and diseases is collected at the same time

May fail to capture the temporal sequence of events necessary to assess possibility of causality

 

 

The other point in this review chapter is that you need to get all the "bang for your buck" possible for your investment in data collection.

 

 

 

 

 

CHAPTER 14 --From Association to Causation

 

Using OBSERVATION we can often identify apparent ASSOCIATIONS among factors and diseases.  Put patterns of association are not necessarily SPECIFIC and SIMPLE.  The nature of relationships among factors and diseases can be VERY COMPLEX and illusive.  We are not dealing with rocks and minerals.  We are dealing with LIFE ITSELF, which we cannot create and perhaps will never fully understand.

 

LIMITATIONS OF HUMAN EXPERIMENTATION

 

 

THE LIMITATION OF ANIMAL STUDIES

 

 

LIMITATION OF VITRO SYSTEMS STUDIES

 

 

 

In an ECOLOGICAL STUDY the UNIT OF ANALYSIS shifts from individuals to nations or other aggregations of people.  Our textbook says that there is evidence that incidence of breast cancer in women tend to higher in NATIONS with higher per capita supply of fat calories.  The problem is, THIS DOES NOT NECESSARILY MEAN THAT THE WOMEN WITH BREAST CANER CONSUMED MORE DIETARY FAT. 

From Wikipedia, the free encyclopedia

Jump to: navigation, search

An ecological study is an epidemiological study in which the unit of analysis is a population rather than an individual. For instance, an ecological study may look at the association between smoking and lung cancer deaths in different countries. An ecological study is normally regarded as inferior to non-ecological designs such as cohort and case-control studies because it is susceptible to the ecological fallacy.

In the United States presidential elections of 2000, 2004, and 2008, wealthier states (states with higher per capita incomes) tended to vote Democratic and poorer states tended to vote Republican. Yet wealthier voters tended to vote Republican and poorer voters tended to vote Democratic. For example, in 2004, the Republican candidate, George W. Bush, won the fifteen poorest states, and the Democratic candidate, John Kerry, won 9 of the 11 wealthiest states. Yet 62% of voters with annual incomes over $200,000 voted for Bush, but only 36% of voters with annual incomes of $15,000 or less voted for Bush.[1]

 

BOTTOM LINE: Ecological studies can be useful in the conception and design of NEW RESEARCH cast at the level of INDIVIDUALS as the UNIT OF ANALYSIS.  But conclusions that gloss the different units of analysis should be avoided.

 

 

SPURIOUS RELATIONSHIPS

 

A high "correlation" between B and C might be interpreted as being causal when in fact it is the result A being the cause of both B and C.

 

CONFOUNDING RELATIONSHIPS

 

The confounding effect.  Both A and B may seem to cause C but A and B are themselves highly correlated.  It may be the confounded effect of the two of them together that causes C and not the independent effect of either one of them alone.  I may refer to this as covariation between (among) two or more INDEPENDENT VARIABLES.   In this example, C is the DEPENDENT VARIABLE.

 

INDIRECT RELATIONSHIPS

 

If A causes B and B causes C and our intent is to prevent C, it is not necessary to address A if we can get rid of B.  I may refer to this as a MEDIATED relation between A and C.

 

NECESSARY AND SUFFICIENT

 

B only happens after A.  A alone is sufficient to cause B.  If we can prevent A there will be no incidence of B.

 

NECESSARY BUT NOT SUFFICIENT

 

A and B together can cause C but neither of them alone will cause C.  It s not necessary to prevent both A and B.  Preventing EITHER of them will prevent C.

 

 

 

 

 

SUFFICIENT BUT NOT NECESSARY

 

ANY ONE of three things (A, B, or C) is sufficient ALONE to cause D.  To prevent D we must prevent A, B and C.

 

NEITHER SUFFICIENT NOR NECESSARY

 

In this case, every known cause of the disease is A COMBINATION of multiple factors.  The textbook says that this probably most accurately represents he causal relationships that operate in most chronic diseases.

 

GUIDELINES FOR JUDGING WHETHER AN ASSOCIATION IS CAUSAL (page 236)

 

1)         Does the pattern appear to be in the "right" temporal order.  We are dealing with causation and not destiny.  http://en.wikipedia.org/wiki/Teleology  Teleology is very interesting, but it is not what we are studying now. 

 

2)         Strength of association.  The stronger the apparent association the more likely that the relationship may be causal.

 

3)         Dose-Response relationship.  If the incidence seems to reflect the dose (degree of exposure) the relationship may be causal.  There may be a threshold effect.

 

4)         If the association can be observed multiple times involving different groups and different kinds of research populations the relationship may be causal.

 

5)         Biologic plausibility.  If the relationship can be explained as causal based upon what we know about how biological systems work, the relationship may be causal.

 

6)         Consideration of alternate explanations.  If there is no other likely candidate factor that may be causing the disease, the relationship between the focal factor and the disease may be causal.

 

7)         Cessation of exposure.  If the exposure can be stopped and the disease stops getting worse (or goes away) the relationship may be causal.  However, in some diseases, the pathogenic process may have been irreversibly initiated.

 

8)         Consistency with other knowledge.  If our hypothesis makes sense in terms of what we already know the relationship may be causal.

 

9)         Specificity of the association.  The authors say this criteria has problems and perhaps should not be included on the list.  I think specificity refers to a clear "one to one" relationship between a factor and a disease.  We are in fact dealing with very complex factor-environments and very complex organic systems.  It is true that simplicity underlies complexity.