This blog post is the first in a series aimed at helping the public, patients, physicians, health care providers, policy makers, and media professionals evaluate medical research findings presented in the media. I believe that a discerning public is an essential element of improved health care. All of us need to be on the same page as we evaluate, report and make decisions about medical care and policy. This series of essays is not aimed at presenting opposing opinions but intended to painlessly teach readers some of the concepts of clinical epidemiology1 in order to empower readers to better judge and utilize medical information in the news and in advertisements. Interesting media reports and advertisements will be critiqued to instruct readers.
IS THERE AN ASSOCIATION BETWEEN MODERATE ALCOHOL INTAKE AND BREAST CANCER?
ABC Evening News with Diane Sawyer reported on a study published in the November 2, 2011 issue of the Journal of the American Medical Association2 which addressed moderate alcohol intake and breast cancer. Click the link below to view the 4 minute news segment.
The news segment reported that in a group of 100,000 participants, those who drank alcohol 3 to 6 times per week had a 15% increase risk of breast cancer. Now let’s consider a different type of report.
A textbook based report3 of this journal article would have reported the study design along with the strengths and limitations of the study design for determining a cause-effect relationship and for generalizing the results to the general public. In addition a commentary on the reported statistics would have been provided. The following are a few of the details such a report would have mentioned.
The newscast reported a 15% increase risk of breast cancer associated with moderate alcohol intake (defined as 3 to 6 glasses per week). The newscast did not specify whether this figure was based on a calculation utilizing the Relative Risk statistic4 or the Risk Difference statistic (also called the Attributable Risk, Absolute Risk Reduction, and Absolute Risk ).4 The calculations for these statistics are different. The calculation using the Relative Risk 4 results in a larger number than does the calculation of the Risk Difference.4 The newscast reported the Relative Risk based statistic.2 In terms of the everyday practice of medicine and public health policy the Risk Difference statistic4 is the more important statistic according to textbooks of epidemiology and evidence based medicine.5,6,7,8 As noted above the reported Relative Risk based statistic was 15% (which is 15 in 100), however the Risk Difference statistic cannot be greater than 3% (which is 3 or less in 100)9,10,11.
The study design is known as a prospective cohort study.12 This particular study also relied heavily on survey methodology.2,13 The design of this study can suggest cause-effect associations but cannot determine the presence of cause-effect associations which can be done by a randomized controlled trial design. This is a very important distinction. The newscast did not mention the study design.
Keep in mind that the interpretation and application of any study is going to depend on one’s set of perspectives. Patients, physicians, policy makers and media professionals can have both opposing and common elements in their sets of perspectives. My opinion from the perspective set of a physician is that this study points to the possibility of a small group of women (no greater than 3% ) who may be vulnerable to moderate alcohol intake as a causative factor for cancer. Further study is needed to confirm the presence of such a group. If such a group is confirmed more study needs to be done to determine the identifying characteristics (e.g. genetic) of this group of vulnerable women so that they can be advised of their risk. Currently no written in stone recommendations can be made. However, patients should be informed of the study and the results discussed with them.
The take home lesson from this first essay is that when you hear or read the words risk and percent (%) related to health care comparisons ask the following question: is this a Relative Risk4 based statistic or a Risk Difference statistic?4 My impression is that the majority of the time the Relative Risk based statistic4 is quoted and not identified as such. An informed public can change this situation.
In case you wondered, I did not write this to pick on ABC Evening News. I have watched ABC Evening News since 1997 when Peter Jennings reported on some research in which I was involved. The breast cancer alcohol story serves as a good example for educating the public, patients, physicians, health care providers, policy makers and media professionals about concepts that help one to make fully informed health care decisions.
The next blog installment will report on an advertisement for a relatively new pharmaceutical agent. We will examine the advertized research statistics within the framework of Relative Risk and Risk Difference reports. Also, the concept of number needed to treat (NNT) will be introduced.
For more information about Dr. LaRiccia visit: http://www.drplariccia.yourmd.com
Email address for Dr. LaRiccia: firstname.lastname@example.org