News Media Coverage of Screening Mammography for Women in Their 40s and Tamoxifen for Primary Prevention of Breast Cancer FREE

Many people become acquainted with important health care issues through the news media. Several recent studies, however, raise questions about how well the press covers medical issues. Moynihan et al,¹ for example, found that news stories reporting on the benefits and harms of 3 popular medications were often inadequate or incomplete: less than two thirds of stories quantified treatment benefit, and only about half mentioned potential harms.

Three related events provide the opportunity to examine news reporting in the context of breast cancer prevention. In January 1997, a National Institutes of Health (NIH) consensus panel considered whether the available evidence supported a recommendation for routine screening mammography for women aged 40 to 49 years.² The panel concluded that routine screening was not warranted and, based on their evaluation of the benefits and harms, recommended that women in their 40s decide for themselves. Under political pressure,³ the National Cancer Institute (NCI) reversed this decision and subsequently recommended screening mammography for women in their 40s in March 1997. The following year, the NCI ended the Breast Cancer Control and Detection Program randomized trial of tamoxifen vs placebo for the primary prevention of breast cancer early when the data demonstrated a significant reduction in breast cancer incidence in the treatment group.⁴

The NIH consensus panel's decision not to recommend routine screening for women in their 40s, the NCI's subsequent reversal, and the early termination of the tamoxifen trial received substantial attention. We examined the nature of the news coverage following each event: for example, whether the stories encouraged women to have mammograms or take tamoxifen, whether they acknowledged uncertainty about what women should do, and how benefit and harm information was expressed.

We analyzed the 2 weeks of news coverage following 3 events: the release of the NIH consensus panel report not recommending screening mammography for women in their 40s (January 23-February 5, 1997), the NCI's announcement reversing the consensus panel's report and advocating mammography (March 27-April 9, 1997), and the dissemination of the results of the Breast Cancer Control and Detection Program regarding prophylactic use of tamoxifen (April 4-17, 1998). To analyze high-visibility reporting, we focused our study on the 10 US newspapers with the largest daily circulation⁵ and on 3 major US television networks (Table 1). We used 2 approaches to analyze the content of each report: 2 readers independently coded each report according to an explicit coding instrument, and we electronically indexed all articles by using software designed for qualitative analysis.

Newspaper Articles. We identified articles by searching Lexis-Nexis and directly contacting the newspaper not indexed (Newsday). To identify mammography stories, our search used the key words breast cancer, mammogr*, or breast cancer and screening. We found 260 citations within both 2-week periods; 80 appeared in a top-10-circulation newspaper. We excluded 44 citations because they were letters to the editor, obituaries, 2-sentence national briefs, or articles not primarily concerned with breast cancer screening for women in their 40s (eg, "Local Hospital Opens Free Mammography Clinic"). Thus, we had a total of 36 articles for analysis (19 discussing the NIH consensus panel and 17 discussing the NCI reversal).

To identify articles about tamoxifen for the primary prevention of breast cancer, we used the key words tamoxifen or Nolvadex. Twenty-seven of the 107 articles identified were published in a top-10 newspaper. After exclusion of articles that were not relevant (eg, articles about tamoxifen for preventing breast cancer recurrence) and letters to the editor, 25 articles were eligible for analysis.

All the articles analyzed were unique (ie, we found only 1 wire service article).

Television Reports. We searched the Vanderbilt Television Archives and Burrelle's Transcripts and directly contacted major television networks that were able to provide us with a list of all reports appearing in the periods studied (NBC, CBS, and CNN). We purchased transcripts of all relevant news reports (8 about the NIH consensus panel, 7 about the NCI reversal, and 9 about tamoxifen).

Manual Coding of Stories. We coded copies of each article as it appeared in the newspaper. To produce more reliable coding, we created a simple coding instrument that asked the reader to indicate the presence or absence of specific elements in each story. The coding instrument included 2 questions intended to capture overall impressions of each article. For the mammography stories, the questions were (1) Does this news story leave you with the overall impression that women in their 40s should . . . definitely get mammograms, probably get mammograms, decide for themselves (ie, the story was neutral), probably not get mammograms, or definitely not get mammograms?; and (2) Does this news story leave you with the impression that there is a lot, a little, or no uncertainty about the benefit of mammography for women in their 40s? The overall codes for tamoxifen stories were the same but substituted tamoxifen for high risk women for mammography for women in their 40s. We also coded how each story expressed the harms and potential benefits of each preventive strategy and whether the benefits and harms were quantified. We pilot tested the instrument by having multiple readers each code 1 article from each event and comparing interpretations. Based on this testing, minor revisions were made to the coding scheme.

The study authors coded each article. Agreement was generally high: raw agreement was 82% (κ= 0.72) for the overall impression codes and 93% (κ= 0.80) for the uncertainty codes; κ values ranged from 0.68 to 1.00 for the benefits and harms. To create the final set of codes for analysis, we had to resolve coding disagreements. Whenever we disagreed, we had a third person, blinded to the study's purpose, independently code each article. We resolved disagreements through discussion, factoring in the third coder's answers; when we could not agree, we used the median response.

Electronic Coding of Stories. We also electronically indexed the full text of each article and television story into a software program designed for qualitative analysis (QSR NUD*IST, version 4; Sage Publications Software, Thousand Oaks, Calif). Using this program, we were able to perform word and phrase searches on each set of stories. To assess the consistency between these 2 analytic approaches, we performed additional searches for many of the elements of our coding scheme. The 2 approaches yielded similar results. For example, searches for false positives (false positive, false alarm, false) and our coding system both found that 67% of articles mentioned false-positive mammogram results. We also used this approach to quantify the number of times harm was mentioned (search terms used were harm, downside, potential risk, and risks or side effects) and whether the story stated that women faced a choice (search terms used were caution, difficult decision, decision, and choice or weighing risks). Because the words decision and choice were used in multiple contexts, we counted only occurrences that were clearly about deciding whether to have a mammogram or take tamoxifen.

The unit of analysis was each news story. We present 2 kinds of statistics, the proportion of mammography or tamoxifen stories with a specific characteristic (eg, the proportion of stories suggesting that women in their 40s should definitely have mammograms), and, in examining how information about harm was expressed, the median number of times each story mentioned harms. We tested differences between proportions of mammography and tamoxifen stories by using χ² tests. We used the Kruskal-Wallis test to assess differences in the median of the number of times harm was mentioned. All P values were 2-tailed (P<.05). All analyses were done with STATA, version 7 (Stata Corp, College Station, Tex).

Twenty-seven NIH consensus panel, 24 NCI reversal, and 34 tamoxifen stories appeared in high-profile media within 2 weeks of each event, including a total of 18 front-page newspaper stories and 7 lead television stories (Table 1).

NIH Consensus Panel. As summarized in Figure 1, the majority of NIH consensus panel stories (67% [18/27]) left the impression that there was a lot of uncertainty about whether women aged 40 to 49 years should undergo screening. Nonetheless, 59% (16/27) suggested that women should probably or definitely be screened. Only 4 consensus panel stories suggested that undergoing screening mammography was a genuine decision for women.

Table 2 shows that the most frequently quoted sources were consensus panel members (31 quotes), NCI or NIH officials (14 quotes), and radiologists (17 quotes). The US Preventive Services Task Force recommendations,⁶ which at that time did not recommend mammography for women in their 40s, was quoted only once, in contrast to 19 quotes from the American Cancer Society, a consistently strong advocate for mammography. The newspaper headlines for the NIH consensus panel stories generally had 3 themes: expression of support for mammography (eg, "It's Basic Health Care," "Needed: Mammograms for Women in Their 40s," and "Why Do We Play Russian Roulette With Our Lives?"), frustration about the lack of a recommendation (eg, "New Mammogram Report Leaves Women Adrift"), and criticism of the consensus panel (eg, "Stand on Mammograms Greeted by Outrage" and "Mammogram Panel Defense").

NCI Reversal. The level of uncertainty reported about whether women in their 40s should undergo mammography dropped substantially following the NCI reversal: 21% (5/24) suggested a lot of uncertainty and 33% (8/24) suggested a little uncertainty (Figure 1). Yet 96% (23/24) of stories suggested that women be screened. Newspaper headlines tended to be factual (eg, "Routine Mammograms Urged at 40") and gave support to the new recommendations (eg, "A Step in the Right Direction"). Politicians were a more frequent source of quotation in this coverage (Table 2), generating a total of 21 quotes. Notably, statements by politicians supporting the NCI recommendations were invariably coupled with promises to enact legislation mandating insurance coverage for mammography. Seventy-four percent (20/27) of NIH consensus panel stories and 54% (13/24) of NCI reversal stories explicitly mentioned the cost of mammograms and the importance of insurance coverage.

Tamoxifen. In contrast, tamoxifen stories highlighted uncertainty about what women at high risk should do (62% [21/34] suggested there was a lot of uncertainty), but none left the impression that women should definitely take the drug (24% [8/34] suggested they probably should) (Figure 1). Instead, these stories suggested that women faced a genuine choice and would have to weigh the risks and benefits themselves. In fact, 65% (22/34) specifically used the words caution, decision, choice, difficult decision, or weighing risks (all but 3 of the other articles indirectly suggested that taking tamoxifen was a difficult decision by juxtaposing the benefits and harms). Typical tamoxifen headlines were "Breast Cancer Study Offers Hope but No Easy Answers," or "Caution Prescribed." The most frequently quoted sources for these stories were the NCI study investigators (n = 24), academics and researchers (n = 34), and women participating in the tamoxifen trial (n = 16) (Table 2). We found no quotations from politicians and only 3 quotes from breast cancer advocacy groups, all acknowledging uncertainty about the role of tamoxifen for primary prevention. The single quote from the American Cancer Society simply described the prevalence of breast cancer and did not state an official position about tamoxifen. Twenty-one percent (7/34) of stories mentioned cost or insurance coverage for tamoxifen. All of these stories expressed confidence that the drug would be covered once it received Food and Drug Administration approval for primary prevention.

Reporting on Benefit. Sixty-seven percent of the stories (57/85) on all 3 events quantified the benefit of the relevant intervention: 63% (17/27) of the NIH consensus panel stories, 50% (12/24) of the NCI reversal stories, and 82% (28/34) of the tamoxifen stories, respectively (Table 3). Sixty percent (34/57) of the stories quantifying benefit provided this information as a relative risk (RR) reduction without the base rate—a presentation known to exaggerate perceived benefit.^7- 9 For example, in many articles, readers were told that the tamoxifen group had 49% fewer cases of breast cancer compared with the placebo group, but they were not told the actual rate of breast cancer in the control group (eg, annual rate of 67 per 10 000).⁴Table 3 shows how data were presented during each event.

Reporting on Harm. Harms were mentioned in nearly all NIH consensus panel (93% [25/27]) and tamoxifen (94% [32/34]) stories but in less than half (42% [10/24]) of the stories about the NCI reversal (Figure 2). Tamoxifen stories were most likely to emphasize harms. We searched the stories for the words harm, downside, potential risk, and risks or side effects and found a median of 3 mentions (25th-75th percentile: 2-4) for each tamoxifen story compared with 1 mention (25th-75th percentile: 1-3) for each consensus panel story and 0 mentions (25th-75th percentile: 0-1) for each NCI recommendation story (P = .005).

The most common harms reported in mammography stories were false-positive (abnormal mammogram results where further testing reveals no cancer) and false-negative results. Overdiagnosis (the detection of early cancers that may be nonprogressive or progress so slowly that they would never become symptomatic), arguably the most important and lasting harm, was mentioned substantially less often (33% of stories after the consensus panel and 8% after the NCI reversal). The most common harm of tamoxifen noted was that of increased uterine cancer (91% of stories). All of the deaths observed in the tamoxifen trial were from pulmonary embolism,⁴ although it was not mentioned nearly as often as uterine cancer (44% of stories).

We reviewed news coverage about screening mammography for women in their 40s and tamoxifen for the primary prevention of breast cancer. We found differences in how the news media reported on these issues. Stories about the NIH consensus panel were generally enthusiastic about mammography; most left the reader with the sense that women in their 40s should undergo screening. In many cases there was a sense of anger (10 of the 27 stories included the words anger, angry, furor, outrage, or upset) directed at the consensus panel for failing to recommend screening and concern with the uncertainty surrounding the idea that women should decide for themselves.

Stories following the NCI reversal in favor of screening were almost uniformly supportive; many expressed a sense of relief that the NIH consensus panel had been refuted and their "error" corrected. These stories were also remarkable for the extent to which politicians and advocacy groups were represented and for a new focus: ensuring that mammograms were covered by insurance.

In contrast, stories about the use of tamoxifen for primary prevention of breast cancer were almost all cautious. The stories quoted scientists, few advocates, and no politicians. Rather than rejecting uncertainty, the stories accepted it, and nearly all suggested that women would have to weigh the risks and benefits of taking tamoxifen and decide for themselves about its use. In addition, almost all tamoxifen stories mentioned the increased chance of developing a second cancer, a particularly compelling harm. Several stories were critical of the decision to end the tamoxifen trial early (ie, once a decrease in breast cancer incidence was detected) and argued that the trial should have been continued until it could provide data on mortality.

In addition to characterizing the overall coverage of mammography and tamoxifen, we also focused on the ways in which news stories reported the harms and potential benefits of these interventions. Here, we modeled our study on that of Moynihan et al,¹ who examined how the news media covered the benefits and risks of popular medications. Like these researchers, we found that the majority of stories on either mammography or tamoxifen presenting data on benefit did so by reporting an RR reduction without noting the underlying event rate, a format known to exaggerate the perceived effect of an intervention.^7- 9 Curiously, although Moynihan and colleagues found that only half of news stories about medications acknowledged any harms, most of the stories we examined did so. This difference may be explained by the focus on controversy in the mammography stories, especially those about the NIH consensus panel, and the newness of the indication for tamoxifen; the drugs Moynihan et al studied, aspirin, pravastatin, and alendronate, had all been available for some time.

Although the difference in the coverage of these distinct approaches to breast cancer prevention may not be surprising, we think 2 interesting issues are raised. First, do the differences in press coverage reflect different attitudes about deemphasizing an established medical practice as opposed to publicizing a new one? The tamoxifen story was about adding a new drug therapy. Because the drug represents the first instance of cancer chemoprevention, news stories may have been appropriately cautious about the idea of healthy women (although at higher risk) taking a medication. The mammography controversy followed a recommendation to reverse support for an established practice. National survey data show that nearly 80% of US women in their 40s have had a screening mammogram,¹⁰ and 96% of women believe that screening should begin before age 50 years.¹¹

Second, news stories may reflect differences in what people think about screening tests compared with drug therapy. The negative aspects of screening may be less obvious than those of medications and may not be well appreciated by reporters and the public.¹² Two limitations of mammography—false-positive and false-negative results—are well known to women.¹³ The more troubling adverse effect, overdiagnosis (ie, the discovery of cancers that would never have become clinically important) is not well known. For younger women, a substantial proportion of screening-detected breast cancers are ductal carcinoma in situ,¹⁴ the majority of which will never become clinically important^14- 17; nonetheless, almost all will be treated surgically.¹⁵ Few women are aware of the issue,¹³ and few of the news stories discussed it.

Several study limitations should be noted. First, the quality of the news media reports we examined may not have been representative. We examined only 2-week periods after each event and considered only 10 newspapers and 3 television networks. It is possible, for example, that other newspapers or networks might have described benefit better (eg, including the relevant base rate when giving relative benefits). The 2-week windows used to select reports for analysis may seem too short. In our search of Lexis-Nexis, however, we found that few new stories appeared about the relevant topics in the subsequent 3 months. Finally, as with any qualitative analysis, our coding system was inherently subjective. We tried to increase objectivity by specifying a coding scheme, using 2 independent coders who demonstrated high interrater reliability, and resolving coding differences with a third, blinded individual. In addition, analysis of headlines, source of quotations, and an electronic search for specific word phrases yielded results consistent with our manual coding. This electronic word searching, however, was not a formal qualitative analysis.

Women concerned about breast cancer may consider a range of risk-reduction approaches. The intense news coverage about mammography for women in their 40s and the use of tamoxifen for the primary prevention of breast cancer undoubtedly made many women aware of these options. We found that stories about mammography generally supported its use, and few seemed willing to question basic assumptions about screening. Reporters seemed to allow the scientific content of the debate to be eclipsed by its political and emotional aspects. In contrast, stories about tamoxifen maintained a more dispassionate, scientific stance; reporters approached tamoxifen cautiously and were willing to question assumptions of benefit. We think reporters should scrutinize screening tests as critically as they examine drugs. By doing so, they may help the public better appreciate the benefits and risks of medical interventions.

Interestingly, another screening method, physical examination, received little attention in the news stories we reviewed (only 5 of 85 mentioned either breast self-examination or clinical breast examination). Although there are no randomized trials supporting breast self-examination screening,^18- 19 long-term data from one large trial suggests that clinical breast examination may be as effective as mammography.²⁰ As the role of tamoxifen and mammography is sorted out, the potential role of a simple low-technology screening strategy, clinical breast examination, should not be ignored.

The scientific debate has been rekindled by the publication of a new meta-analysis questioning the benefit of mammography screening for women of any age.^21- 22 This provocative analysis has recently received substantial attention in the US press. Balanced media coverage about the benefits, harms, and uncertainties of mammography—or any medical intervention—would help focus the new debate on science rather than emotions or politics.