				Using the search form

IN THE NEWS

Commentary on advances in medicine

Does antibiotic use cause breast cancer?

Theresa Hegmann, MPAS, PA-C

Ms. Hegmann is Clinical Assistant Professor and Director of Clinical Education, Physician Assistant Program, Roy J. and Lucille A. Carver College of Medicine, University of Iowa, Iowa City. The author has indicated no relationships to disclose relating to the content of this article.

In mid February 2004, news services all over the world reported some startling information: a researcher at the University of Washington in Seattle had documented a link between antibiotic use and breast cancer. In fact, many of the headlines used the word link, probably taken from a February 17 press release from the National Institutes of Health entitled, "Study Shows Link Between Antibiotic Use and Increased Risk of Breast Cancer."¹

Without a doubt, the findings of this study were newsworthy. Women in the United States have a 1 in 7 lifetime probability of developing breast cancer, which is the second most common cause of cancer-related death in women.² Breast cancer touches countless lives and triggers anxiety and fear for both patients and their care providers. Predictably, many of the news reports on the study advised women who frequently take antibiotics to talk with their physicians, so it is important for clinicians to understand what this research "link" between breast cancer and antibiotic use really means.

Links, coincidences, and associations

In the context of medical research, what is a link? The occurrence of breast cancer in a single patient treated with antibiotics marks a case where cancer and antibiotic treatment coincide. If many women treated with antibiotics are observed to develop breast cancer, the coincidence between antibiotic treatment and cancer occurrence becomes high. High coincidence is often referred to as a link between an environmental factor and a disease state. Such links might mean that the environmental factor has a causal role in the disease state—but links can also be sample specific, spurious, and misleading. The challenge in researching a link of this sort is to determine the likelihood that the high rate of coincidence results from a true causal relationship.

The gold standard experimental design to answer this question is a randomized, double-blind, placebo-controlled trial. Groups of patients would be randomized to receive either antibiotics or placebo; then they would be evaluated over time for differences in cancer rates. Although testing causal ties this way with human subjects is neither ethical nor feasible, it is, unfortunately, exactly what we would have to do to determine with any degree of certainty whether antibiotic use causes breast cancer.

Ethically, researchers can use only observational design—taking people as they find them and observing, but not manipulating, the factors being studied. Even with careful choice of control groups and sophisticated data analysis to adjust for as many confounding variables as possible, observational studies still can establish only a high level of coincidence—frequently called an association—between two factors.

Antibiotics and breast cancer

The study in question is an excellent example of a specific type of observational study called a case-control study.³ Case-control studies are generally "retrospective"—they start with the identification of the disease or condition of interest and work backward to identify possible risk factors associated with that condition.

Author Christine M. Velicer, PhD, and colleagues compared 2,266 women who had a new diagnosis of invasive breast cancer with 7,953 control patients. Subjects were selected from a population enrolled in a large health plan that has kept pharmacy data in an electronic database since 1977. The authors found that, as cumulative days of antibiotic use increased, so did rates of incident breast cancer: Women who took antibiotics for more than 500 days, or who received more than 25 antibiotic prescriptions over an average of 21 years of enrollment, had an approximate doubling in risk. Women with lower rates of antibiotic use, such as 1 to 10 prescriptions filled over an average of 17 years, still had an odds ratio of 1.49, indicating that they were about 50% more likely to receive a diagnosis of breast cancer than women who had not used any antibiotics. These associations were true across all common classes of antibiotics, and P values were generally less than .001.

One drawback to a case-control design is "confounding"—when an observed association between two variables is actually caused totally or in part by an underlying variable or variables. For example, an association between the rate of drowning deaths and the rate of ice-cream-cone consumption over a year's time would almost certainly exist if we were to look for it. Here, the confounding variable in the system is obvious—hot summer weather. Often, however, confounding variables are not so obvious, or the system is so complex that many such confounding variables exist. Just as forbidding kids to eat ice cream is unlikely to make them safer swimmers, we would be unwise to conclude on the basis of the JAMA study that avoiding antibiotic use would prevent breast cancer.

Efforts by researchers to control for potential confounding variables can never be perfect. Although a good case-control study that does not find an association between two factors can provide proof of the absence of a relationship, even the best-designed case-control study can never prove a causal connection between two factors. To their credit, the authors of the JAMA study made no attempt to claim that their results supported a causal link between antibiotic use and breast cancer, and they were careful to caution the reader against making that assumption.

Moving from link to causation

And so we are left with a large, relatively well-designed observational study that has found a significant "link" between antibiotic usage and breast cancer. The big question in the minds of patients and clinicians is surely this: Does antibiotic use cause breast cancer?

The classic criteria for establishing causality were developed by Sir Austen Bradford Hill, and in a modified form, they include affirmative answers to the following questions:⁴

• Is there evidence from true experiments in humans?

• Is the association strong?

• Is the association consistent from study to study?

• Does the postulated cause clearly precede the postulated effect?

• Is there a dose-response gradient?

• Does the association make both epidemiologic and biological sense?

• Is the association specific?

• Is the association similar to a previously proven causal relationship?

In this case, we can answer only a few of these questions in the affirmative, and then only tentatively. Dr. Velicer's is the second observational study to suggest an association between antibiotic use and breast cancer, and her results offer some support for a dose-response gradient. The relationship indicated in the JAMA study is not particularly strong (odds ratios between 0.5 and 2.0 require cautious interpretation), though the P values are impressive. The temporal relationship is problematic, as we could argue that breast cancer might have been developing for years before diagnosis in many of these women. We do not have proof that the association is specific; the reference group of "never users" of antibiotics might have routinely used antibiotics before they joined the health plan.

As for whether the association makes biological sense, reasonable arguments can be made on both sides. The authors mention that antibiotics might increase the risk of breast cancer by decreasing the levels of cancer-fighting phytochemicals metabolized by the intestinal microflora. They also note that some classes of antibiotics have an effect on the production of prostaglandin E2, which might lead to overexpression of cyclooxygenase 2, which has in turn been associated with mammary carcinogenesis. On the other hand, a weakened immune system could reasonably be associated both with increased antibiotic use and with an increased rate of cancer. Or, as mentioned by Roberta Ness, MD, and Jane Cauley, DrPH, in an editorial commenting on the JAMA study, both antibiotic use and breast cancer incidence may be related to a state of chronic infection and/or inflammation.⁵

Unfortunately, based on this study, we do not know whether antibiotics cause breast cancer. This is a common predicament in clinical medicine, since most studies reported in the news that describe risk factors for cancer have an observational design and cannot establish causation. A part of the evolving role of PAs may be to help our patients negotiate this maze of information overload. Avoiding every factor that has ever been linked with cancer is virtually impossible—and it might not make a whit of difference anyway.

Right now, we cannot apply the results of this research study to clinical practice. Certainly, the study reemphasizes that antibiotics should be used only when they are clearly indicated and only for as long as necessary. For most thoughtful clinicians, this probably does not represent a change in practice. The main value of this study may well be to open the question for further research. Drs. Ness and Cauley put it this way: "As is often true for reports of new associations, this study provides . . . more . . . questions than answers."⁵

Acknowledgment

The author is grateful to Joseph P. Hegmann and Rick Dehn for their suggestions and editorial assistance.

REFERENCES

1. Study shows link between antibiotic use and increased risk of breast cancer. National Institutes of Health. NIH News. February 17, 2004. Available at: http://www.nih.gov/news/pr/feb2004/nci-17.htm . Accessed April 12, 2004.

2. American Cancer Society. Cancer Facts & Figures 2004. Available at: http://www.cancer.org/downloads/STT/CAFF_finalPWSecured.pdf . Accessed April 12, 2004.

3. Velicer CM, Heckbert SR, Lampe JW, et al. Antibiotic use in relation to the risk of breast cancer. JAMA. 2004;291:827-835.

4. Greenhalgh T. How to Read a Paper: The Basics of Evidence Based Medicine. 2nd ed. London, UK: British Medical Assoc; 2001.

5. Ness RB, Cauley JA. Antibiotics and breast cancer—what's the meaning of this? JAMA. 2004;291:880,881.