Comparison of Screening Mammography in the United States and the United Kingdom FREE

The provision of screening mammography differs greatly between the United States and the United Kingdom. In the United States, screening is provided in diverse settings, such as private practice, health maintenance organizations, and academic medical centers¹; whereas in the United Kingdom, a single organized screening program run by the National Health Service provides virtually all mammographic screening for women aged 50 years or older.^2- 3 There are also differences between the ages of women screened; the recommended interval between mammographic examinations; the proportion of women recalled for additional imaging examinations, such as diagnostic mammography or ultrasound; and the methods used to further evaluate findings considered suspicious for cancer.^4- 6 However, it is not clear if there are actual differences in the performance and outcomes of screening mammography between the 2 countries. Comparing the performance of screening mammography between the 2 countries may suggest methods to improve mammography practice.

We compared recall (the percentage of mammograms in which there is a recommendation for prompt additional testing, clinical evaluation, or percutaneous biopsy), surgical biopsy, and cancer detection rates for screening mammography among similarly aged women between the United States and the United Kingdom.

Data on US screening mammography was obtained from the Breast Cancer Surveillance Consortium (BCSC)⁷ and the National Breast and Cervical Cancer Early Detection Program (NBCCEDP).^8- 9 In the United Kingdom, data were obtained from the National Health Service Breast Screening Program (NHSBSP).^3,10 Results of all screening mammograms in women aged 50 years or older conducted within each of these settings between January 1, 1996, and December 31, 1999, were included. More than 1 screening examination was included if the examinations occurred more than 9 months apart. We excluded mammograms obtained to further evaluate a mass detected by clinical breast examination, because of a breast symptom, or to follow up an abnormal mammogram. The study was approved by the institutional review board at the University of California, San Francisco.

The BCSC is a National Cancer Institute–funded consortium of mammography registries in San Francisco, Calif; Colorado; New Hampshire; New Mexico; North Carolina; Seattle, Wash; and Vermont.⁷ The primary purpose of the consortium is to collect data pertaining to mammography performance in a uniform fashion across diverse settings and populations.¹¹ Women are included if they self-refer or are referred by a physician for a mammogram to 1 of 202 contributing facilities. Data are obtained for individual women from self-administered questionnaires¹² and radiologist reports (medical records). Mammography results are reported using the categories of the American College of Radiology's Breast Imaging Reporting and Data Systems.¹³ Cases of cancer are ascertained through active case follow-up and through linkages with state tumor registries, Surveillance, Epidemiology and End Result programs, or pathology databases, and cancer ascertainment has been found to be 94% complete.¹⁴ Although all US facilities and radiologists must follow the Mammography Quality Standards Act/Mammography Quality Standards Reauthorization Act regulations,¹⁵ the BCSC offers no specific guidelines for, or has authority in advising, how mammograms should be interpreted.

The NBCCEDP, which is funded by the Centers for Disease Control and Prevention, provides breast and cervical cancer screening to poor uninsured women throughout the United States.^8- 9 Although funding limitations have allowed only 15% to 20% of eligible women to be served, screening mammography for women aged 40 years or older have been provided in all 50 states, tribes, and territories since 1996. The Centers for Disease Control and Prevention funds each state, which in general contracts for the mammographic screening through diverse settings. Data are collected for individual women from self-administered questionnaires and medical records from primary physicians and radiologists. Mammography results are reported to the programs using the categories of Breast Imaging Reporting and Data Systems.¹³ Cancer occurrences are ascertained primarily through active follow-up of mammograms with abnormal findings and review of pathology reports but some programs also link to state tumor registries or Surveillance, Epidemiology and End Result programs. The NBCCEDP offers no specific guidelines on how mammograms should be interpreted, but it works with all of the state programs to improve program performance, including mammography.

The government-funded NHSBSP provides free breast cancer screening in the United Kingdom for women 50 years or older.^3,10 Women aged 50 to 64 years are invited by postcard to attend breast screenings every 3 years through a system that relies on centralized computer databases. From age 65 years onward, women are encouraged to self-refer. By 1995, the NHSBSP achieved national coverage so that screening mammography was available to all eligible women. The program is currently organized into 95 separate breast-screening programs that coordinate the provision of screening services and cancer ascertainment. Data are collected and analyzed locally as well as centrally in the Department of Health and the Cancer Screening Evaluation Unit, University of London, London, England. Women specifically concerned about breast problems are referred to hospital breast clinics for diagnostic mammography and the results of such testing are not included in this study.

For the BCSC and NBCCEDP, a mammogram was classified as positive (recall) if the assessment was abnormal or incomplete (Breast Imaging Reporting and Data Systems¹³ categories 0, 3, 4, and 5) and a recommendation for prompt diagnostic imaging, clinical evaluation, or biopsy (including fine-needle aspiration, core biopsy, and open surgical biopsy) was given. Clinical evaluation was the reason for a positive examination in a small minority of cases (<2% of the recalls) but was included to be consistent with the NHSBSP. For the NHSBSP, a mammogram was classified as positive (recall) if there was a recommendation for further work-up, including diagnostic imaging, clinical examination, or pathological evaluation. Any additional views that were recommended contributed to the recall rate. Mammograms that were recommended for short-interval follow-up only were not considered positive.

Because recall and cancer detection rates vary by whether women have undergone previous mammography,^9,16- 17 all analyses were subdivided into first or subsequent screening examination (screening cycle). For the BCSC and NBCCEDP data, a mammogram was classified as first if the woman had no prior mammogram in the database and self-reported no prior mammogram within 5 years. For the NHSBSP, the first mammogram that a woman underwent in the program was considered first. Information on race, ethnicity, socioeconomic status, and cancer risk factors, such as menopausal status and use of hormone therapy, are not collected by the NHSBSP or NBCCEDP and were not included.

Women were considered to have breast cancer detected if active-case follow-up or reports from a pathology database, Surveillance, Epidemiology and End Result program, or state tumor registry showed invasive carcinoma or ductal carcinoma in situ within 12 months of a positive screening mammogram. Cancers that occurred after a mammogram with negative findings (false-negative examinations) were not included in this analysis.

Recall, noninvasive work-up, and biopsy rates were calculated per 100 screening mammograms and stratified by first or subsequent examinations and by 5-year age groups, or age-adjusted to a standard age distribution. The standard age distribution was the mean of the age distributions of the 3 data sources, in which each data source was weighted equally. The recall rate was calculated as the number of mammograms with positive findings per 100 screening mammograms. The noninvasive work-up rate was calculated as the number of recommendations for only noninvasive work-up, including ultrasound, diagnostic imaging, other noninvasive tests, or breast examination per 100 mammograms. Each mammogram was counted 1 time when calculating the noninvasive work-up rate, even if more than 1 test was recommended. The biopsy rate (any type of biopsy) was calculated as the number of mammograms with a recommendation for fine-needle aspiration, core biopsy, biopsy where the type was not specified, or open surgical biopsy per 100 mammograms. Each mammogram was counted 1 time when calculating the overall biopsy rate, even if more than 1 biopsy was recommended. The percutaneous biopsy rate was calculated as the number of fine-needle aspirations or core biopsies per 100 mammograms. The open surgical biopsy rate was calculated as the number of open surgical biopsies per 100 mammograms. Women could have contributed to both the percutaneous biopsy rate and the open surgical biopsy rate, hence these numbers do not sum to the overall biopsy rate. The open surgical biopsy rate was subdivided into 2 groups: those that resulted in a diagnosis of cancer (positive open surgical biopsy rate) and those that did not (negative open surgical biopsy rate). The specific method of biopsy could not be determined for the NBCCEDP data and for 3 of the 7 BCSC sites and thus the percutaneous and open surgical biopsy rates could not be calculated for these sites. The cancer detection rate was calculated as the number of breast cancers detected per 1000 examinations. The rate of invasive cancer by tumor size (<10 mm, 10-20 mm, or >20 mm) was calculated per 1000 examinations using the standard age distribution.

Because mammographic screening is recommended^3,18- 21 and performed^3,22 more frequently in the United States than the United Kingdom, one would expect fewer cancers to be diagnosed per subsequent screening examination in the United States. To compare cancer detection rates for a similar period of screening, we used 4 years of actual data to estimate the number of cancers detected and women recalled per 1000 women undergoing screening mammography during a 10-year period. For these estimates, we assumed that screening started at age 50 years (or 60 years) and continued for 10 years using an estimated screening interval for each setting. For the BCSC and NBCCEDP, the estimated screening interval was 18 and 19 months, respectively, and was based on the mean time between mammograms that women obtained between 1998 and 1999. These estimates are similar to those reported by others.^22- 23

For the NHSBSP, screening occurred about every 3 years²⁴ and correspondingly, the interval was estimated at 36 months. To calculate 10-year estimates of cancer detection and recall for each program, a 50-year-old woman was assumed to have undergone a single first mammogram and several subsequent examinations, and the age-specific recall rate and cancer rate of these first and subsequent examinations were those reported herein. We assumed that women aged 60 years or older underwent only subsequent examinations so only age-specific recall and cancer rates for subsequent screens were used to calculate 10-year estimates. We also assumed that the likelihood of recall and cancer detection were independent from one examination to the next, and that a woman could be recalled or have cancer detected only once. We estimated the chance of at least 1 recalled examination or cancer diagnosis during a 10-year period for a 50-year-old woman who underwent routine screening and a 60-year-old woman who underwent routine screening in each setting. To estimate the variability of these 10-year estimates, we used the 95% confidence interval for the recall rates and cancer detection rates, and varied the screening interval from 16 to 20 months (BCSC), 17 to 21 months (NBCCEDP), and 33 to 39 months (NHSBSP). The lower estimate for the range in the cancer rate was calculated by assuming the lower bound of the 95% confidence interval for cancer detection and screening interval. SAS version 8.2 (SAS Institute Inc, Cary, NC) was used for all statistical analyses.

This analysis included 5.5 million mammograms: 978 591 from the BCSC, 613 388 from the NBCCEDP, and 3.94 million from the NHSBSP, which led to the diagnosis of 27 612 cases of breast cancer among women aged 50 years or older (Table 1).

Recall rates were similar between the BCSC and the NBCCEDP for both first and subsequent examinations (Table 2). Recall rates in these 2 US settings were approximately twice as high as those in the United Kingdom for all age groups, for first as well as subsequent examinations. Among first screening mammograms for women aged 50 to 54 years, 14.4% of women in the BCSC and 12.5% in the NBCCEDP vs only 7.6% in the NHSBSP were recalled for further evaluation, including diagnostic imaging, ultrasound, clinical examination, or biopsy. Biopsy rates were similar across all settings: 2.3% to 3.4% of first screening mammograms and 0.84% to 1.7% of subsequent screening examinations were followed up with a recommendation for biopsy. The higher US recall rate was primarily because of a higher rate of diagnostic imaging, ultrasound, and clinical evaluation.

Although the biopsy rates were similar between the 2 countries, biopsies were more likely to be open surgical biopsies in the United States (Table 3). For 100 first screening mammograms, 1.1% in the United States compared with 2.4% in the United Kingdom resulted in a recommendation for percutaneous biopsy; for 100 subsequent screening mammograms, 0.4% in the United States compared with 0.8% in the United Kingdom resulted in a recommendation for percutaneous biopsy (age-adjusted data). Conversely, for 100 first screening mammograms, 1.15% in the United States compared with 0.72% in the United Kingdom resulted in a recommendation for open surgical biopsy (age-adjusted data). Most of the difference in open surgical biopsy rates was attributed to procedures among women who did not have breast cancer, with negative open surgical biopsy rates 2 to 3 times as high in the United States vs the United Kingdom. For 100 first screening examinations, 0.82% resulted in negative open surgical biopsy in the United States compared with 0.36%. Positive surgical biopsy rates were more similar between the 2 countries but tended to be higher in the United Kingdom for subsequent examinations.

The cancer detection rates increased with age and were 2 to 3 times as high for first vs subsequent mammograms in both countries (Table 4). Despite substantially higher recall rates in the United States, cancer detection rates were similar across settings, particularly for first screening examinations. For 1000 first examinations among women aged 50 to 54 years, 5.8, 5.9, and 6.3 cancers were diagnosed in the BCSC, NBCCEDP, and NHSBSP, respectively. Differences in cancer detection rates between the 2 countries were higher for subsequent examinations, likely reflecting more frequent US screenings.

The estimated number of cancers detected per 1000 women screened during 10 years was also similar between both countries (Table 5). If 1000 women aged 50 to 59 years underwent regular mammographic screening during 10 years, approximately 24.5 cancers would be detected in the BCSC, 23.8 in the NBCCEDP, and 19.4 in the NHSBSP. If 1000 women aged 60 to 69 years underwent regular mammographic screening during 10 years, approximately 31.5 cancers would be detected in the BCSC, 26.6 in the NBCCEDP, and 27.9 in the NHSBSP. Although invasive cancer detection rates are more similar between the 2 countries, the in situ cancer rates are higher in the United States. Among women aged 50 to 59 years, approximately 5.8, 7.4, and 3.8 in situ cancers would be detected, respectively. The higher frequency of screening in the United States magnifies the difference in the estimated recall rates between the countries when projected over 10 years. After 10 years of screening 1000 women aged 50 to 59 years, 477 women in the BCSC and 433 in the NBCCEDP vs 175 in the NHSBSP would have been recalled for additional work-up. After 10 years of screening women aged 60 to 69 years, 396 women in the BCSC and 334 in the NBCCEDP vs 133 in the NHSBSP would have been recalled for additional work-up.

For first screening mammograms, there were slightly fewer US invasive cancers diagnosed per 1000 examinations in most size categories (Figure 1). For subsequent examinations, there were lower rates of invasive cancer in all size categories in the United States vs the United Kingdom. The absolute difference in cancer rates between the United States and United Kingdom was highest for invasive tumors 10 to 20 mm.

The recall and negative open surgical biopsy rates associated with screening mammograms were twice as high in US settings than in the United Kingdom; however, cancer detection rates were similar in the 2 countries. In the United Kingdom, half as many women are recalled for diagnostic examinations and half as many women without breast cancer undergo open surgical biopsies as in the United States. These results observed in large numbers of women are similar to recent findings from a series of 60 test cases evaluated by physicians in both countries in whom false-positive rates were higher among US physicians but cancer detection rates were not.²⁵ The goal of any cancer screening effort is to obtain high cancer detection rates while avoiding unnecessary diagnostic evaluation following false-positive results, which are costly and associated with ongoing psychological morbidity.²⁶

There are several possible explanations for the differences in recall rates between the 2 countries. Much higher rates of US malpractice lawsuits that focus on missed breast cancer diagnoses²⁷provide a strong incentive to increase sensitivity at the expense of specificity, possibly leading US radiologists to recall women when they identify a finding with a low likelihood of cancer.^25,28 In addition, US physicians must read only 480 mammograms annually to fulfill Mammography Quality Standards Act requirements,¹⁵ while radiologists in the United Kingdom are required to read at least 5000 mammograms annually²⁹ and on average, radiologists interpret 5 to 7 times their US counterparts. Furthermore, although more than 90% of programs in the United Kingdom use double reading, this practice is much less common in the United States. Although the exact impact of double reading remains uncertain, some evidence shows that double reading by consensus or arbitration, as used in the United Kingdom, raises detection rates and decreases recall rates.^30- 31 Another consideration is the higher centralization of mammogram reading in the United Kingdom, as well as a less mobile population, which might make prior mammograms more readily available for comparison when interpreting results. Although the latter difference might reduce recall rates for subsequent mammograms,^32- 33 it does not account for higher recall rates for first screening mammograms.

Most importantly, quality assurance standards for the NHSBSP programs are set nationally and are regularly monitored through a quality assurance network. Ranges of acceptable data for recall, biopsy, and cancer detection rates have been established and an organized program operates at the local and national levels to monitor and achieve these targets.^5,29 All screening programs in the United Kingdom receive data that enable a comparison of their recall and cancer detection rates with other programs. Both programs and individual radiologists below a minimum standard are subject to quality assurance scrutiny. In contrast, the United States has only voluntary guidelines³⁴ and there is no national organization to collect or monitor data to promote high levels of performance. Finally, an organized program of professional development in the United Kingdom specifically provides instruction and individual feedback regarding recall and cancer detection rates by using a set of test mammography cases called PERFORMS.³⁵ Although not compulsory, 85% of mammographers from the United Kingdom participated in this test in 2001. Continuing medical education is a requirement for US radiologists but the actual content is not uniformly organized and almost never targets specific recall or cancer detection rates.

The NHSBSP has integrated quality assurance into all clinical aspects of its program^5,29,36 and as a result, the United Kingdom has observed dramatic improvements in the performance of screening mammography since the program began in 1988. For example, cancer detection rates have increased dramatically for both first and subsequent screening examinations, as has the positive predictive value of mammography.³⁶ The United Kingdom observed a rapid 50% decline in the open surgical biopsy rate between 1996 and 1999, as a result of a coordinated effort to increase the use of percutaneous biopsy and to decrease the percentage of women without breast cancer who underwent open surgical biopsy.^3,37 The well-documented improvements in the United Kingdom⁵ demonstrate that implementation of quality control can be efficient and feedback mechanisms effective. Despite the differences between the 2 countries in the provision and funding of screening mammography, mammography technology is very similar between the 2 countries and similar targets for mammography outcomes, including specific recall and biopsy rates, could be established in the United States. Success in reaching technical targets set by the Mammography Quality Standards Act/Mammography Quality Standards Reauthorization Act demonstrates how a coordinated quality assurance program can work in the United States.³⁸

Screening mammography is performed more frequently in the United States than in the United Kingdom. During a 10-year period, women aged 50 years or older will undergo approximately 7 mammograms in the United States vs only 3 in the United Kingdom. More frequent screening likely translates into smaller average cancer size at diagnosis, as evidenced by the slightly lower rates of invasive cancer for 10 mm or more and the higher rates of in situ cancer diagnosed in the United States. Additionally, US screening tends to begin at an earlier age than in the United Kingdom. From our results, it cannot be determined whether these differences affect breast cancer mortality.

We compared the cancer detection rates, which are widely used as a measure of mammography performance,^36,39- 40 as they approximate the total cancer rates and can be readily measured for quality assurance purposes. We found the breast cancer detection rates in both countries to be similar. Given that the overall age-adjusted breast cancer incidence rates are slightly higher in the United States,⁴¹ one would expect that the United States would have similar or higher cancer detection rates than in the United Kingdom. Thus, it is unlikely that the United Kingdom is missing cancers despite a much lower recall rate. It has been shown that at high recall rates, cancer detection rates levels off.⁴² Thus, despite recalling more women, more cancers are not detected in the United States.

The main limitation of our study is that we cannot be certain that our definition of screening mammography was the same across all 3 settings. Specifically, we do not know if there was a higher proportion of diagnostic examinations among the US women, which might account for a higher recall rate. However, more diagnostic mammograms should produce a substantially higher cancer rate^43- 44 in the 2 US settings, which we did not find. We should also note that our estimation of the total cancers detected during 10 years was based on only 4 years of screening data and the assumptions of the model were simplistic. When we used different values for these assumptions, our results did not appreciably change. Additionally, our estimated recall rates are similar to those results found by others.^32,45 There is likely a small degree of overlap between the 2 US data sources but this is estimated to be less than 3% of the mammograms described. Additionally, by pooling data within each program, we have ignored variations by region, physician, and other variables in each program.^36,46 Lastly, although the data from the United Kingdom includes virtually all mammographic screening performed, the US data reflects only a small percentage of mammography performed. Because mammograms from all 50 US states were included and the results from the BCSC and NBCCEDP were broadly similar, these results probably provide the best current evidence of the US performance of mammography screening.

We did not focus on differences between the BCSC and NBCCEDP (such as the slightly higher diagnostic imaging rate and slightly lower biopsy rate in the BCSC) because the differences between the 2 US data sources were small compared with the differences between the 2 countries and these programs describe different populations, in which breast cancer rates, as well as tumor characteristics, might be different.

Women undergoing screening mammography should consider going to facilities where physicians read a large number of mammography examinations,^28,40 radiologists devote a large percentage of their practice to mammography,⁴⁰ and comprehensive auditing of outcomes is undertaken on a routine basis.¹³ Additionally, women should return to the same facility for repeat screening or ensure that comparison films are available to radiologists at the time of imaging interpretation, if they change facilities.^32- 33 Lastly, if they do have an examination with abnormal findings and an open surgical biopsy is recommended, they should discuss all options with a radiologist or surgeon, and consider getting a second opinion.

In the United Kingdom, the NHSBSP has set and reached targets that emphasize high rates of cancer detection and low recall. Recall rates in the United Kingdom are now substantially lower than in the United States with no substantial reduction in cancer detection. We believe this success stems primarily from a centralized program of continuous quality improvement. A large portion of the costs associated with mammographic screening comes from frequent screening⁴⁷ and the relatively high percentage of women who undergo additional testing.⁴⁸ Screening women aged 50 to 69 years biennially and reducing recall rates could substantially decrease the cost of mammography, as well as associated anxiety caused by false-positive diagnoses.⁴⁹ Efforts to improve US mammographic screening should be targeted to lowering the recall rate without substantially lowering the cancer detection rate.