Marketing Medicine to the Public: A Reader's Guide | JAMA | JAMA Network
[Skip to Navigation]
Sign In
February 13, 2002

Marketing Medicine to the Public: A Reader's Guide

Author Affiliations

Not Available

Not Available

JAMA. 2002;287(6):774-775. doi:10.1001/jama.287.6.774-JMS0213-4-1

Efforts to market medicine to the public have greatly expanded in the last several years. In addition to seemingly ubiquitous direct-to-consumer drug advertisements, medical centers increasingly advertise services such as cancer care or surgical procedures. Recently, a number of companies have begun soliciting patients to order their own tests, from simple blood tests to advanced imaging studies. As medicine becomes increasingly commercialized, the prominence of direct-to-consumer marketing efforts is likely to grow.

Ideally, medical advertisements would promote informed decision making by educating consumers about medical conditions, tests, and treatment options. Unfortunately, such ads often present medical information in a way that exaggerates disease risk and thus the value of the marketed products in reducing that risk. The purpose of this essay is to help physicians critically read medical advertising so they are better prepared to respond to patients with misconceptions about advertised claims. We analyze 3 actual advertisements to illustrate how to approach messages about disease risk, screening, and medication.

How Many People Get or Die From a Disease?

"This year 180 000 Americans will be found to have a brain tumor, 21 000 of which originate in the brain. The others are of "metastatic origin". . . . Brain tumors are the second-leading cause of cancer death in children under the age of 15."1

This example, from a company marketing brain imaging to the public, uses a common strategy to exaggerate risk. The message begins with attention grabbing large numbers, but it provides little context to make sense of them. To understand a message about disease risk, readers need to know the population at risk and how this risk compares to other risks.

In the example, the population at risk appears to be the entire US population. The 180 000 brain tumors (and 21 000 primary brain tumors) occur among approximately 275 million Americans.2,3 By highlighting the numerator (ie, number of tumors) without mentioning the denominator (ie, total number at risk), the reader's attention is focused on a large number instead of a small proportion. Explicitly providing this denominator would probably change how readers perceive their risk of having a brain tumor. A clearer expression of the reader's risk of having a primary brain tumor might say: "This year 0.07% of Americans (7 in 10 000) will be found to have a brain tumor."

The original message also notes that brain tumors are "the second leading cause of death among children under the age of 15." This alarming phrase further exaggerates risk by suggesting that death from brain cancer is common during childhood. To realistically assess the magnitude of the threat, readers need information not provided in the ad: How often do children die before the age of 15, and what proportion of these deaths are attributable to brain cancer? Despite what the ad suggests, on average, a child's chance of dying of any cause by age 15 is small (about 1%). Injuries are the single largest cause of deaths in this age range and account for 40% of all deaths in childhood. Compared with a risk of 1 in 10 000 of dying before age 15 of a brain tumor4, the risk of death from injury is about 16 times that of death from brain tumor.

What Is the Benefit of Early Detection?

"CT [computed tomographic] scans—when used in conjunction with an analytical model developed here—can detect lung cancer five to seven years before an x-ray, when the tumor is no bigger than a grain of rice. This early detection breakthrough could raise the lung cancer survival rate from a discouraging 12% to as high as 80%".5

This advertisement uses the 5-year survival metric (ie, the proportion of cases alive 5 years after diagnosis) to suggest that screening can dramatically improve prognosis in lung cancer. Specifically, it suggests that the chance of surviving 5 years after diagnosis is 12% for people who are not screened, and 80% for those who are.

This advertisement is misleading because the comparison is inherently biased in favor of screening.6 Consider the effect of lead time bias. Imagine a person with an advanced lung cancer causing cough, hemoptysis, and weight loss; these symptoms prompt medical attention and a diagnosis of lung cancer is made in 1997. Despite treatment he dies in 1999. Now imagine that CT screening detected the tumor at an earlier stage in 1995, two years before the development of symptoms. Because treatment could not affect the progression of his tumor, despite the fact that it was diagnosed earlier, he still dies of lung cancer in 1999. In the first case, survival from time of diagnosis was 2 years; in the second, 4 years. Although measured survival improved by 2 years, he died at exactly the same time. Thus, the only thing the screening accomplished was informing the patient earlier that he had an incurable disease. Two other biases—length bias (ie, the tendency of screening to discover relatively less aggressive cancers) and overdiagnosis bias (ie, the fact that some screen-detected cancers would never become clinically significant even without treatment)—also distort comparisons of 5-year survival in favor of screening.6 Whether or not screening really works, screening tests that identify presymptomatic disease will always improve 5-year survival even if treatment is completely ineffective.6 While there is good evidence that CT screening can detect lung cancers before they become symptomatic,7 there are no data demonstrating improved outcomes.8 While it seems intuitive that early detection must improve outcomes, this intuition was wrong for lung cancer screening using chest radiographs.6 Lower mortality rates from a randomized trial—not improved 5-year survival—would constitute proof of the benefit of screening.8

The advertisement also misleads consumers by failing to mention 2 potentially harmful consequences of testing: the risks associated with further diagnostic testing, and overdiagnosis. Some people undergoing CT screening have suspicious lesions detected that are not cancer (ie, false positives). In general, cancer is ruled out through a biopsy, which carries significant risks in itself. Patients requiring thorascopy or open lung biopsy need general anesthesia and often require chest tubes. These procedures can be complicated on rare occasions by serious infection or death. A second way that screening can result in harm is through overdiagnosis, which is the detection of lung cancers (sometimes called pseudodisease) that either would not have progressed or would have progressed so slowly that the patient would have died of other causes before ever experiencing symptoms.6,9 Treatment of pseudodisease is unnecessary and can only cause harm. Pseudodisease can only be found by screening, and there is no way to know which lesions represent pseudodisease during a patient's lifetime. Given the current state of knowledge about screening for lung cancer, there is no way to know how many patients who are screened will be injured by unnecessary treatment.

How Well Does Treatment Work?

A frequently published direct-to-consumer advertisement states: "42% fewer deaths from heart attack among those taking ZOCOR".10

This advertisement is a good example of the kind of the numbers readers are likely to see. The 42% reduction in heart attack deaths sounds impressive. But the key question to ask is "42% fewer than what?" Without knowing what number is being lowered by 42%, it is impossible to know the absolute magnitude of the change.

This information is provided in small print at the bottom of the advertisement: "42% reduction based on 111/2221 (ZOCOR) vs. 189/2223 (placebo)." Therefore, ZOCOR lowered the risk of heart attack death in the next 5 years from 8.5% (placebo) to 5% (ZOCOR)—a 42% relative risk reduction (1-5%/8.5%) and a 3.5% absolute risk reduction (8.5%-5%). The impressive figure—42%—reflects the difference between 8.5% and 5%. To understand how well a treatment works, a comparison should be made between the risk of an outcome for people who do not receive treatment (ie, the base rate) with the risk for those who do.

One way to increase demand for medical services is to promote exaggerated beliefs about disease risk and intervention benefit. A few simple questions may help readers critically evaluate such claims (Box).

Box Section Ref ID

Box. What to Look for in Messages That Market Medicine

How many people get or die from disease? Denominator (calculate the chance of the outcome in the target population) Time frame (learn over what period of time the risk refers to) Context (compare how this chance compares to the chance of other events)

What is the benefit of early detection? Evidence for delayed death—not just earlier diagnosis (lower mortality rates from randomized trials–not improved 5-year survival for cancers detected by screening–constitute proof of the benefit of screening) Potential harms (false-positive test results and the follow-up testing needed, and the detection of pseudodisease)

How well does treatment work? Absolute event rates with and without treatment (beware of reports that present a relative risk reduction without the base rate) Outcomes that matter to patients (eg, reduced mortality rather than reduced cholesterol)

Author contributions: Drs Schwartz and Woloshin contributed equally to this manuscript, and the order of authorship is entirely arbitrary.
Funding/Support: Drs. Schwartz and Woloshin are supported by Veterans Affairs Career Development Awards in Health Services Research and Development and a grant from the National Cancer Institute (CA91052-01). The views expressed herein do not necessarily represent the views of the Department of Veterans Affairs or the United States government.
Acknowledgment: The authors wish to thank Elliott Fisher, MD, MPH, and H. Gilbert Welch, MD, MPH, for helpful comments.
This article was corrected on February 13, 2002.
Not Available, home page. Accessed October 12, 2001.
US Bureau of the Census, National population estimates for the 1990's. Available at: Accessed June 19, 2001.
Greenlee  RTHill-Harmon  MBMurray  TThun  M Cancer statistics, 2001.  C A Cancer J Clin. 2001;5115- 36Google ScholarCrossref
National Center for Health Statistics, Death rates for 282 selected causes by five-year age groups, race, and sex: United States, 1978-1998. Available at: Accessed October 15, 2001.
Not Available, Columbia Presbyterian and New York Weill Cornell have come together to fight cancer. We have no intention of losing.  New York Times. June18 2000;34- 35Google Scholar
Kramer  BBrawley  O Cancer screening.  Hematol Oncol Clin North Am. 2000;14831- 848Google ScholarCrossref
Henschke  CIMcCauley  DIYankelevitz  DF  et al.  Early Lung Cancer Action Project: overall design and findings from baseline screening.  Lancet. 1999;35499- 105Google ScholarCrossref
Frame  P Routine screening for lung cancer? maybe someday, but not yet.  JAMA. 2000;2841980- 1983Google ScholarCrossref
Black  W Overdiagnosis: an underrecognized cause of confusion and harm in cancer screening.  J Natl Cancer Inst. 2000;921280- 1282Google ScholarCrossref
Not Available, Zocor Direct to Consumer Advertisement.  New York Times. September23 2001;47- 48Google Scholar