Evidence that Screening Is Beneficial

A full understanding of the controversy must start with a review of the scientific evidence, principally that from eight population-based, randomized trials of screening efficacy. Results from all but one of these trials consistently show that deaths from breast cancer are reduced in groups of women who are screened compared with control groups; the mortality reduction was statistically significant in several of the trials [7-12]. These findings apply to all those in the populations studied; the lower limit of age ranged from 40 to 45 years and the upper limit from 64 to 74 years. When the data are analyzed in the ways anticipated in the trial designs, the evidence indicates that screening is beneficial from age 40 to age 74 [2, 13, 14].

The controversy began with the retrospective analysis of results by age subgroups, a method of evaluation unanticipated in the design of all but one of the trials [12]; these trials were planned with marginal power that was sufficient only for whole-group examination of the data. When used to interpret the end results of the trials, retrospective subgroup analysis reduces the power of the results and thus substantially increases the likelihood that an observed benefit will not achieve statistical significance [15, 16]. This was discussed eloquently in the editorial that recently inaugurated the "In the Balance" series in Annals [17]: "... post hoc data-dredging for subgroups of patients or subsets of events, however qualitatively dramatic their results, should only generate hypotheses for the next trial, not conclusions from this one."

The scientific limitations of retrospective subgroup analysis notwithstanding, results show consistent point estimates of mortality reduction in women who are screened compared with controls for women aged 50 and older (average mortality reduction of approximately 30%; statistically significant mortality reduction in several individual trials), probably because of the much larger numbers of older women in the trials [2, 7-11]. For women aged 40 to 49 years, most point estimates show mortality reductions ranging from 22% to 49%, but three trials show no benefit and none of the trials individually shows statistically significant results [2].

Analysts have combined the randomized trial data using meta-analysis. When 7-year results are evaluated for all trials, there is no evidence for mortality reduction in women aged 40 to 49 years, and the point estimate indicates a relative risk close to unity [2, 18]. However, meta-analysis is most meaningful when it evaluates trials that are similar in design. Seven of the eight trials had similar designs: Women from defined populations were randomly selected, individually or by community, and then were offered screening. The Canadian National Breast Screening Study (CNBSS), on the other hand, involved self-selected participants (volunteers) who were randomly selected only after clinical breast examination, and all controls were screened with a clinical examination [12]. Although staunchly defended by its principal investigators [19-22], this trial has been widely criticized, both within and outside Canada, for deficiencies in design and execution [3, 23-31].

A recent meta-analysis of data on women aged 40 to 49 years in the five Swedish trials shows the emergence of a 13% mortality reduction in screened women compared with controls, beginning after 8 years of follow-up [32]. A subsequent meta-analysis [33] was completed using updated results and shows that results from trials mature enough to report follow-up of 10 years or more (therefore excluding data from the CNBSS) indicate a 17% mortality reduction in screened women compared with controls. The most recent meta-analysis [34], which is currently in press, provides results that exclude data from the CNBSS but incorporate the recently presented 10-year follow-up data from the Goteborg trial [35]. This meta-analysis shows a 21% mortality reduction in screened women compared with controls, a statistically significant result (upper bound of 95% CI, 0.98) [34]. Demonstration of significance is remarkable indeed, given the numerous deficiencies in design and execution, not only in the CNBSS but in all the trials [3]. These deficiencies have probably lessened the observed benefit, perhaps substantially [3, 14, 34].

"Evidence-based" guidelines, by definition, should take into account all available information. Because each of the randomized trials was planned more than 10 years ago and because subsequent major advances in mammographic imaging have been made [36], some investigators have supplemented evidence from the trials with reports of intermediate screening outcomes in large demonstration projects that use more modern mammographic techniques. These results, although limited by the absence of matched populations of unscreened women and the unavailability of mortality data, show that the major prognostic factors (median tumor size, axillary lymph node status, and tumor stage) of screening-detected breast cancers indicate that screening is at least as beneficial for women aged 40 to 49 years as for women aged 50 to 64 years, the age range for which screening is widely accepted [3, 37-39].

In summary, for women aged 40 to 49 years, data from randomized trials (which are limited by retrospective subgroup analysis, low statistical power, and use of older mammographic techniques) show a statistically significant mortality reduction that appears only 8 to 10 years after entry into screening programs. It also can be inferred from data from modern mammography demonstration projects (which are limited by a lack of control group comparisons and the unavailability of mortality results) that screening is as or more beneficial for women aged 40 to 49 years than it is for older women, for whom the benefit has already been proved.

How to Interpret the Evidence That Screening Is Beneficial

Those interpreting the evidence summarized previously have come to strikingly different conclusions about the routine use of mammographic screening, depending on the level of proof they use to establish efficacy. Those who require the highest degree of proof will deny that screening is beneficial for women in their 40s because published studies, including meta-analyses, do not show a statistically significant mortality reduction in this group. It should be noted that the use of such rigorous interpretive criteria will not only reject the efficacy of mammographic screening among women aged 40 to 49 years but also that of screening using clinical breast examination, because results from randomized trials that include screening with clinical breast examination do not show a statistically significant benefit [7, 8, 12]. Those who do not require such strict criteria will accept the combination of data from the recent meta-analysis that shows statistical significance after 8 to 10 years of follow-up [34] and inferences that can be made from data on the efficacy of more modern mammographic techniques as convincing evidence that screening is beneficial for women aged 40 to 49 years. Both interpretations of the evidence are defensible, given the assumptions on which each is based. Indeed, a multinational group of eminent clinicians and epidemiologists (including representatives from the NCI) who attended a workshop sponsored by the Union Internationale Contre le Cancer (UICC) came to this conclusion using the then available data; they stated that the evidence can support clinical guidelines that recommend breast cancer screening beginning at either age 40 or age 50 [31, 40].

The NCI has suggested a compromise as an alternative to either endorsing or not endorsing screening for all women aged 40 to 49 years of age, but this approach lacks internal consistency: The revised NCI guidelines recommend that screening be limited to women at high risk for developing breast cancer. In fact, little if any scientific evidence supports such an approach. The same randomized trial results used by the NCI to justify not endorsing screening for average-risk women also do not show that screening is beneficial for high-risk women because none of the randomized trials was designed to evaluate such women and because no retrospective subgroup analyses have specifically assessed results for such women. The only evidence that favors selective screening involves data from demonstration projects that indicate a higher positive predictive value for detecting cancer in high-risk women [41], a finding explained by the increased prior probability of malignancy in these women rather than by any intrinsic difference in the efficacy of screening for women with different risk profiles.

Another compromise is embodied in the candid statement of Samuel Broder, MD, Director of the NCI: "A physician's advice may be different from a public health official's. What I do as an individual is recommend annual mammograms, but I can't recommend it to the public because I don't have the facts" [42]. We suspect that many practitioners understand Dr. Broder's quandary, despite his obvious inconsistency in logic. An even more troubling ethical dilemma faces practitioners who achieve compromise by recommending screening to their own families and friends but not to their patients.

Other Factors Affecting the Endorsement of Mammographic Screening

Factors other than scientific analysis affect the decision of whether to endorse the routine mammographic screening of women aged 40 to 49 years. One such factor is economic: specifically, the weighing of cost against benefit. Opponents of mammographic screening often criticize the costs induced by false-positive interpretations, especially the cost of surgical biopsies for otherwise undetected lesions that prove to be benign. A balanced discussion of this issue must include a comparison of false-positive cases that result from mammography with those that result from clinical breast examination. Viewed in this context, mammographic screening appears to be the more favorable intervention. Although the positive predictive value (number of cancers detected divided by number of biopsies recommended) ranges from 15% to 35% for both palpable and nonpalpable lesions (43-45; Brown ML. Personal communication), mammographically detected cancers are smaller and are less often associated with axillary nodal and distant metastases [45].

Furthermore, recent advances in mammographic imaging now permit substantial reductions in the number of benign surgical biopsies for screening-detected lesions. The efficacy of periodic mammographic follow-up for "probably benign" lesions (those having a very low likelihood of malignancy) has been shown by large-scale prospective studies [46-48]; this approach increases the positive predictive value for mammographically detected lesions to approximately 40% [46, 49]. Percutaneous imaging-guided tissue sampling also can serve as an alternative to surgical biopsy. This approach is usually used for nonpalpable lesions that are more likely to be malignant [50-52]; it has been reported to increase the positive predictive value to between 60% and 75% [9, 38, 50, 53]. Increasing acceptance and use of these reduced-cost approaches will occur as we enter an era of managed care and capitation payment; primary care practitioners can facilitate the process by encouraging mammography facilities to adopt such new approaches.

Already, estimates of the cost-effectiveness of mammographic screening are favorable. Because of differences in cost estimates and methods of measuring benefits, it is difficult to compare the wide range of results from various published studies [54]. However, the few reports that describe age-related differences consistently show that mammographic screening is less cost-effective in women aged 40 to 49 years than in women aged 50 to 69 years [55-57] but that the marginal cost per year of life saved for those aged 40 to 49 years is similar to that for those who participate in other widely accepted medical prevention and treatment programs, including the screening for and treatment of hypertension, coronary artery bypass surgery, and hemodialysis for renal disease [57, 58].

Another economic factor serves as a disincentive for primary care practitioners to recommend mammographic screening. It often takes considerable time for a practitioner to convince a woman to have routine mammography, but there is no mechanism to bill for such a consultation. Indeed, none of the reimbursement for mammographic screening goes to the primary care practitioner who recommends the procedure and who takes responsibility for follow-up of any screening-detected abnormalities. Correction of this inequity might go a long way toward achieving better compliance with mammographic screening guidelines among women of all ages.

Political factors also affect the decision of whether to endorse routine mammographic screening. These are important primarily to the politicians who control government agencies and who must make the difficult decisions of how to allocate limited health care funds but who also are accountable to an electorate that often expects more in government-supported services than it is willing to pay for in taxes. In our opinion, the recent decision of the NCI to abandon its previous recommendation for routine mammographic screening of women aged 40 to 49 years was based primarily on such economic and political considerations. This opinion is sustained by a recent report of the NCI's Congressional Oversight Committee, which severely criticizes the NCI for ignoring much of the scientific evidence and for its biased review process [59]. We believe that the NCI changed its guidelines on the basis of a perceived need to limit expenditures while maintaining a position of active support for women's health issues. Why else would the NCI tell women and their physicians that scientific studies do not show sufficient evidence of benefit to endorse mammographic screening in women aged 40 to 49 years and yet simultaneously recommend as "prudent practice" the annual screening of these women using clinical breast examination [1]? Such a position is scientifically inconsistent but politically expedient. Consistency would dictate either rejecting the use of both mammography and clinical breast examination (if one applies rigorous criteria for accepting evidence that screening is beneficial) or accepting both examinations as prudent practice but indicating the lack of political will to fund the mammography component.

What Should the Primary Care Practitioner Do?

Ultimately, this will be a personal decision made on the basis of the level of proof one requires to accept screening as beneficial, and it might also be influenced by economic and political considerations. Various choices are available, including the scientifically consistent extremes and those in the middle, which lack scientific support. In making your decision, please consider that the weight of evidence suggests that screening is beneficial, whether or not you agree with the validity of retrospective subgroup analysis; this is especially true because meta-analysis of the data from randomized controlled trials is beginning to show the emergence of a statistically significant benefit.

We support the routine use of mammographic screening for all women aged 40 to 49 years because we interpret the evidence as indicating a high probability of benefit. Currently, more than 10 000 deaths occur per year in the United States among women who developed breast cancer between the ages of 40 and 49, and 30% of the years of life lost to breast cancer are among these women. Because the probable benefit of screening involves at least a 20% reduction in deaths from breast cancer, thousands of lives are probably lost with each year of indecision and inaction. It seems much more prudent to endorse screening now and risk the unlikely subsequent determination that the effort was ineffective, than to withhold screening until it is determined whether "proof" will be obtained and risk the loss of so many women in the prime of life.

However, if you choose not to recommend routine screening because you believe that efficacy has not been adequately proved, then be consistent in your practice and also omit clinical breast examination for asymptomatic women aged 40 to 49 years. If you choose not to recommend routine screening only because of population-based cost considerations, then tell your patients that you believe screening is beneficial and also tell them what you recommend for friends and family members who can afford screening.

The decision to screen represents an opportunity for—not a guarantee of—early detection and diagnosis, but the decision not to screen represents the loss of this opportunity.