Correlation Versus Causation, Exercise 3
Read the excerpt below from the Journal of the American Medical and answer the following questions:
- The main question being investigated here is, “Do __________ __________ cause __________ ______________ ___________?”
- Do the results represent an anecdote, a correlation, or causation?
- Did the study described prove that implants cause anything? If so what?
- What percentage of the women with breast implants got connective tissue disease?
- What percentage of the women without breast implants got connective tissues disease?
- Based on this study, a headline in Women’s Health Weekly, 3/25/96, p14, 2p. said: “Women With Breast Implants 24% More Likely To Develop Connective Tissue Disease.” Is this headline misleading? If so why?
- Guess what might be different about the women who choose to get implants compared to those who chose not to. Which of these might be causing the connective tissue disease? (In other words, what are the possible third variables or plausible competing hypotheses?)
- What do the researchers say may have caused a flaw in their data?
- What do the researchers say is the major contribution of their study?
- Describe in detail the experiment that would have to be done to show that breast implants caused connective tissue disease. (It’s very hard to isolate the variables when experimenting on humans.)
Excerpts from: “Self-Reported Breast Implants and Connective-Tissue Diseases in Female Health Professionals: A Retrospective Cohort Study.” JAMA. 1996 Feb 28;275(8):616-21. Hennekens, C.H. et al
OBJECTIVE: To evaluate the association of breast implants with connective-tissue diseases.
DESIGN AND PARTICIPANTS: Retrospective cohort study of 395,543 female health professionals who completed mailed questionnaires for potential participation in the Women’s Health Study. A total of 10,830 women reported breast implants and 11,805 reported connective-tissue diseases between 1962 and 1991…
MAIN OUTCOME MEASURE: Self-reported connective-tissue diseases.
RESULTS: Compared with women who did not report breast implants, the relative risk…among those who reported breast implants was 1.24 …With respect to the individual diseases, the finding for other connective-tissue diseases (including mixed) was statistically significant (P=.017), the findings for rheumatoid arthritis, Sjogren’s syndrome, dermatomyositis or polymyositis, or scleroderma were of borderline statistical significance (.05<P<.10), and the finding for systemic lupus erythematosus was not statistically significant (P=.44). There were no clear trends in RR with increasing duration of breast implants.
CONCLUSION: These self-reported data from female health professionals are compatible with prior reports from other cohort studies that exclude a large hazard, but do suggest small increased risks of connective-tissue diseases among women with breast implants. The very large sample size makes chance an unlikely explanation for the results, but bias due to differential over-reporting of connective-tissue diseases or selective participation by affected women with breast implants remains a plausible alternative explanation. The major contribution of this and other observational analytic studies has been to exclude large risks of connective-tissue diseases following breast implants.”
|without implants||with implants|
|Total in group||384,713||10,830|
|Number with connective tissue disease||11,407||398|
|Percentage of group with connective tissue disease||2.97%||3.67%|
|Approximate number with disease out of every thousand||30||37|
Numbers in this chart calculated and compiled by John Norland