Freshman Statistics Seminar

Week 5: Case study on correlation vs causation and levels of evidence

Marta Shore

Objective: To look at the evidence for circumcision and HIV infection reduction in terms of both our understanding of correlation and levels of evidence.

Article Summary:

  • Caldwell 1996 Sci Amer 274 p62 “The African AIDS Epidemic”

The first article here, “The African AIDS Epidemic” by John C Caldwell and Pat Caldwell, is an article demonstrating correlation without actually using the word “correlation”.  They argue convincingly that circumcision is the only thing that “explains” reduced infection rates in certain parts of Africa, by showing that other potential explanations do not overlap the region of highest infection as well as lack of circumcision does.  But what they are showing is that lack of circumcision and high HIV infection rates are correlated.  They are doing an observational study.

  • Lafraniere 2006 NYT “Circumcision Studied In Africa As AIDS Preventive”

The next article: “Circumcision Studied in Africa as AIDS Preventative” from the New York Times (4/28/06) illustrates the uncertainty of the correlation.  Although there is overwhelming evidence of correlation, the UN is waiting for “randomized controlled trials” before making the recommendations.  In other words, they are worried that there may be some other underlying factor that results in the correlation.

  • McNeil 2006 NYT “HIV Risk Halved By Circumcision- US Agency Finds”

The final recommended article “H.I.V. Risk Halved by Circumcision, U.S. Agency Finds” from the New York Times (12/14/06)  shows the results from a randomized clinical trial (the additional reading article from PloS Medicine gives more detail).  It tests men in two regions in Africa, Kenya and Uganda.  It confirms the results shown in the previous article from the French study in South Africa.  We now have  randomized controlled trials in three regions in Africa.  It is important to note that now the UN is recommending circumcision to men in Africa as long as done properly.

Suggested Lesson Structure:

  • First, students can examine the evidence over time, to try to understand why the UN  took so long to recommend circumcision. This is a chance to explore concepts from the last two classes: correlation and levels of evidence. The questions in the discussion points are designed to lead this discussion.
  • Second, there are some extra issues that, if you have time, would be good to explore.  One important point is that circumcision leads to a reduction in the risk of contracting HIV.  This does not mean that men who are circumcised are immune, or that risky behaviors are OK with circumcision.  This is explored in the active learning module.
    • Another concept to explore is the concept of randomization.  In particular, why are the experiments “randomized”?  If there’s time, an exploration of randomization as a way to ensure the two groups are as alike as possible.
    • Finally, if there’s time, the active learning module is designed to help understand odds of infection.  Circumcision is not a cure, it merely reduces the odds of getting infected.

Discussion Points:

  • Based on this first article, would you recommend circumcision?
  • What else, besides what was mentioned in the article, could be linked to both lack of circumcision and high HIV infection rates?
  • (If no ideas) What if you were told that circumcision tends to occur mainly in Muslim populations Africa, and these populations have more taboos on extramarital sexual relations?  Also, what if Muslims are least common in the AIDS belt?

NOTE: the additional article from Health Transition Review: “The East African AIDS Epidemic and the absence of male circumcision: what is the link?” by James P.M. Ntozi gives some other plausible reasons that lack of male circumcision and higher HIV rates could be correlated.

  • Moving on to the second article: “Circumcision Studied in Africa as AIDS Preventative” (4/28/06), why is the UN not recommending circumcision yet?
  • What does “randomized clinical trial” mean? How is this different from what was done in the Caldwells’ article?
  • What else was listed as a ’cause’ of high HIV infection rates, according to Daniel Halpern from USAID?
  • Based on the third article: “H.I.V. Risk Halved by Circumcision, U.S. Agency Finds” (12/14/06) why are more organizations (including, after the article was published, the UN) willing to recommend circumcision now?

Active Learning Module:
*** need to design a program that randomly selects either HIV infection or non-infection at the levels for the circumcised (0.8 per 100) or non-circumcised (2.1 per 100) groups***
Divide the class into two groups.  In group 1, each student is to try the above program 10 times for the circumcised group.  If they get infected any one of those 10 times, they should raise their hand.  In group 2, they do the same with the uncircumcised odds.

Download HIV simulation data:

HIV Simulation (Week5simulation_DS_5-28-07.xls)

Additional Links:

The ANRS 1265 Trial

The East African AIDS epidemic and the absence of male circumcision: what is the link ?

The African AIDS Epidemic – John C Caldwell

Circumcision Studied in Africa as AIDS Preventive

HIV Risk Halved by Circumcision, US Agency Finds