Radiology
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

DOI: 10.1148/radiol.2342040048
This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF)
Right arrow Submit a response
Right arrow Alert me when this article is cited
Right arrow Alert me when eLetters are posted
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Yankaskas, B. C.
Right arrow Articles by Barlow, W. E.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Yankaskas, B. C.
Right arrow Articles by Barlow, W. E.
(Radiology 2005;234:363-373.)
© RSNA, 2005

Breast Imaging

Association between Mammography Timing and Measures of Screening Performance in the United States1

Bonnie C. Yankaskas, PhD, Stephen H. Taplin, MD, MPH, Laura Ichikawa, MS, Berta M. Geller, EdD, Robert D. Rosenberg, MD, Patricia A. Carney, PhD, Karla Kerlikowske, MD, Rachel Ballard-Barbash, MD, MPH, Gary R. Cutter, PhD and William E. Barlow, PhD

1 From the Department of Radiology, University of North Carolina at Chapel Hill, CB 7515, 106 Mason Farm Rd, Chapel Hill, NC 27599-7515 (B.C.Y.). Affiliations for all other authors are listed at the end of this article. Received January 9, 2004; revision requested March 12; revision received April 20; accepted May 24. Supported by cooperative agreements UO1CA63731, UO1CA63736, UO1CA63740, UO1CA69976, UO1CA70013, UO1CA70040, UO1CA86076, and UO1CA86082 from the National Cancer Institute as part of the Breast Cancer Surveillance Consortium of the National Cancer Institute. S.H.T. was the principal investigator on grant CA69976. Address correspondence to B.C.Y. (e-mail: bcy@med.unc.edu).


   ABSTRACT
TOP
 ABSTRACT
INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
DISCUSSION
 REFERENCES
 
PURPOSE: To evaluate whether there is an association between the number of months since previous mammography (MSPM) and performance measures (sensitivity, specificity, recall rate, cancer detection rate, and positive predictive value) in women who underwent U.S. community-based screening mammography.

MATERIALS AND METHODS: Data from seven registries (Breast Cancer Surveillance Consortium) and mammographic data and cancer outcome in regard to 1 213 754 screening mammographic examinations performed in 680 641 women who were 40–89 years old for the years 1996–2000 were used in this study. These data are submitted annually in a standard format to a central statistical coordinating center that is subject to institutional review board approval, quality control, and confidentiality standards. Performance measures were calculated for first and subsequent screening mammography. For subsequent mammography, performance measures were calculated according to categories of MSPM (9–15, 16–20, 21–27, and ≥28 months). Receiver operating characteristic and multivariable logistic regression analyses were conducted to test the association between the number of MSPM and performance measures.

RESULTS: With increasing MSPM in each category from 9–15 to 28 months or more and for first mammographic examinations, respectively, there was increased sensitivity (70.9%, 75.7%, 85.4%, 82.5%, and 88.6%), decreased specificity (93.3%, 92.7%, 91.6%, 91.0%, and 85.9%), increased recall rate (7.0%, 7.6%, 8.8%, 9.4%, and 14.7%), and increased cancer detection rates (3.2, 3.5, 4.5, 4.6, and 6.1 per 1000 mammographic examinations). When the category of 9–15 MSPM was compared with that of 21–27 MSPM, there was a slight increase in positive predictive value from 4.6% to 5.1%. Confidence intervals were narrow and did not overlap. Age affected these associations for all performance measures except sensitivity.

CONCLUSION: Performance measures increased as MSPM increased, except for specificity, which decreased. Time between mammograms is an important factor to consider when audits are reviewed or screening performance measures are compared.

© RSNA, 2005


   INTRODUCTION
TOP
 ABSTRACT
 INTRODUCTION
MATERIALS AND METHODS
 RESULTS
DISCUSSION
 REFERENCES
 
Researchers in some studies (1–3) have documented a marked increase in the use of mammography throughout the United States from the middle 1980s to the present. Although some results remain controversial, those in most clinical trials of mammography indicate reduced breast cancer mortality among screened women. In the United States, two major organizations recommend screening every 1–2 years for women 40 years and older (4,5). In the United Kingdom and most of Europe, screening is performed every 2–3 years. To our knowledge, in no U.S. study has an evaluation been conducted about whetherdifferences in the time since previous mammography affect performance parameters, and very few investigators include sensitivity in their comparison of measures of screening performance. Thus, the purpose of our study was to evaluate whether there is an association between the number of months since previous mammography (MSPM) and performance measures (sensitivity, specificity, recall rate, cancer detection rate, and positive predictive value [PPV]) in women who underwent U.S. community-based screening mammography.


   MATERIALS AND METHODS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS
DISCUSSION
 REFERENCES
 
All authors in consensus were involved with the study design and data analysis.

Data Sources
Data were available from registry sites of the Breast Cancer Surveillance Consortium. The sites draw from geographically, racially, and economically diverse counties of the United States with characteristics that closely replicate the national unemployment rate (Breast Cancer Surveillance Consortium, 3.4%; United States, 4.1%), national median income (Breast Cancer Surveillance Consortium, $55 189; United States, $50 984), national racial mix (African American group: Breast Cancer Surveillance Consortium, 9%; United States, 11%), and national Hispanic ethnicity (Breast Cancer Surveillance Consortium, 6.9%; United States, 7.3%) (6,7). The National Cancer Institute created the Breast Cancer Surveillance Consortium in 1994 for the purpose of evaluation of the performance of mammography in the community setting (6). Seven community-based mammography registries located in Vermont, New Hampshire, North Carolina, Colorado, New Mexico, California, and the state of Washington have created mammography databases that are linked with population-based cancer databases. Data are submitted annually in a standard format to a central statistical coordinating center that is subject to institutional review board approval, quality control, and confidentiality standards (8).

Mammographic Data
All data related to screening mammography were collected at the facility at the time mammography was performed. Data collected from the patient and used for this study included date of birth, history, history and date of previous mammography, and reported presence of signs and symptoms (lump, nipple discharge, or others, but not breast pain). Mammographic data were recorded directly by the radiologist or technologist and included indication for the examination, additional imaging studies performed, date of previous mammography, and date of comparison mammography. In addition, breast density and mammographic assessment were recorded by using the recommended categories of the Breast Imaging Reporting and Data System of the American College of Radiology (9). Breast density was categorized as extremely dense, heterogeneously dense, scattered fibroglandular densities, or almost entirely fat. Mammographic assessments were coded as follows: 0, additional imaging evaluation needed; 1, negative; 2, benign; 3, probably benign; 4, suspicious abnormality; and 5, highly suggestive of malignancy.

Breast Cancer Data
The seven registries collected breast cancer information from one or more of three sources: regional Surveillance, Epidemiology, and End Results (known as SEER) programs; state cancer registries; and pathology databases. Cancers were categorized as either ductal carcinoma in situ or invasive disease (lobular carcinoma in situ was considered benign for this analysis). Cancers were also classified into two size categories, smaller than 1 cm or 1 cm or larger.

Study Population
A computerized search of the pooled data identified 2 235 409 mammographic examinations performed in women 40–89 years old for the years 1996–2000 that were indicated by the radiology facility as performed for routine screening. To ensure that diagnostic mammography was not included, mammographic examinations associated with any of the following were excluded: only unilateral views, a patient who had a history of breast cancer, a patient who had undergone imaging within the previous 9 months, and a patient who had breast implants. After all exclusions were applied, 91% (2 026 606) of all mammographic examinations remained that were indicated as screening mammographic examinations by the radiology facilities. Further exclusions from 91% of all mammographic examinations included 11% (230 213) of mammographic examinations in which breast density was not recorded, 23% (474 881) of those in which density was not reported by using the Breast Imaging Reporting and Data System, and another 5% (107 758) of those in which the time since previous screening for subsequent mammography was missing. The final study population, after the initial and further exclusions, included 54% (1 213 754) of the total initial population of mammographic examinations. Most exclusions (23%) occurred because the Breast Imaging Reporting and Data System four-point scale for density classification was not used. These exclusions generally involved entire mammography facilities and were therefore not expected to affect the relationship between the number of MSPM and performance measures. When the characteristics of the women whose mammographic examinations were excluded were compared with those of women whose mammographic examinations were included, no differences were found.

Our final study population had a total of 680 641 women 40–89 years of age in whom a combined total of 1 213 754 screening mammographic examinations were performed between 1996 and 2000. Because our unit of analysis was the bilateral mammographic examination (and not the woman), for one woman, one or multiple mammographic examinations could be included in the data. Our use of the term mammographic examination refers to the bilateral mammographic examination.

Of the 680 641 women whose mammographic examinations were included in the analysis, 20% (134 277) were 40–44 years old, 18% (121 065) were 45–49 years old, 16% (113 729) were 50–54 years old, 12% (80 076) were 55–59 years old, 18% (123 239) were 60–69 years old, and 16% (108 255) were 70–89 years old at the time of screening mammography. At each visit, women separately recorded their racial group and whether they were of Hispanic origin ("yes" or "no" response). Racial information was missing for 19% (130 547) of all women, and data about Hispanic origin were missing for 22% (147 690) of all women. Among the women for whom there was racial information (550 094), the distribution was as follows: 88% (483 349) were white, 8% (43 500) were black, 2% (12 898) were Asian, 1% (4021) were American Indian or Alaskan native, and 1% (6326) were of other races or of mixed races. Among the women who responded to the question about Hispanic origin (532 951), 6% (31 212) were of Hispanic origin.

Time Since Performance of Previous Mammography
The number of MSPM was our primary variable of interest. The number of MSPM was calculated by using the most recent of the following date sources: date of previous mammography in the database, date of comparison mammography, date of previous mammography reported by the radiologist, and self-reported date given by the patient. Self-reported time that previous mammography was performed was used in 24% (272 141 of 1 143 558) of women who underwent subsequent mammography. Women who underwent first mammography were included, though there is not a time measurement. We give results for this group separately from the remaining patients in whom previous mammography was performed.

We categorized MSPM into five groups as follows: 9–15, 16–20, 21–27, and 28 months or more, and no previous mammography, which we refer to as first mammography. We chose these time periods on the basis of the frequency distribution of calculated months (Fig 1). The 9–15- and 21–27-month categories captured most women who were returning for annual or biannual screening. We made the 16–20-month category distinct, as it contained a sufficiently large number of mammographic examinations and they did not naturally belong in either the 9–15- or the 21–27-month group.



View larger version (31K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 1. Graph shows frequency distribution of MSPM.

 
Performance Measures
Definitions that we used for positive and negative screening mammographic assessment, cancer outcomes, and performance measures are presented in Table 1. Performance indexes were based on the assessment of screening mammography (two views) only prior to consideration of any additional imaging (10). Although a short-term follow-up is recommended with Breast Imaging Reporting and Data System category 3 assessments, 37% of these assessments were associated with a recommendation for additional imaging, as reported in the Breast Cancer Surveillance Consortium data (11). Since this was more consistent with the Breast Imaging Reporting and Data System category 0 management recommendation, we considered this subset of category 3 cases as positive assessments for analysis. In cases in which the screening visit included both screening mammography and additional imaging to arrive at an assessment, the screening mammographic assessment was considered Breast Imaging Reporting and Data System category 0 for the analyses. We used a single overall assessment per mammographic examination. When a woman had different assessments for the left and right breasts, the higher category was used according to the following hierarchy: 1, 2, 3, 0, 4, 5, from lowest to highest suspicion for cancer.


View this table:
[in this window]
[in a new window]

  		TABLE 1. Definitions Used for Classification and Analysis

 
Covariates of Interest
Previous research indicates that women’s characteristics, including age, breast symptoms, hormone replacement therapy use, and breast density, affect performance estimates. It was important to consider these in the analyses (12–14). Although we evaluated distribution of hormone replacement therapy use in our descriptive data, we did not include hormone replacement therapy use in our models because we would have had to exclude a large number of mammographic examinations from analyses on the basis of missing hormone replacement therapy use data. There are published reports showing that hormone replacement therapy use affects breast density (12–17) and that breast density is the most important predictor in a model with age, breast density, and hormone replacement therapy use (12). Thus, exclusion of hormone replacement therapy should not have had a substantive effect on our results. Family history was not included, as it does not appear to influence sensitivity (18), although it may influence other performance estimates.

Statistical Methods
We examined sensitivity, specificity, recall rates, PPV, and cancer detection rates according to 5-year age groups within each MSPM category. When we compared our data with published data, we combined the categories of 9–15, 16–20, 21–27, and 28 months or more into a total "subsequent" mammography category and considered patients in whom no previous mammography was performed as those who underwent "first" mammography.

In general, our numbers were sufficiently large that statistical testing for differences in the descriptive data were not very informative, as even very small absolute differences were statistically significant.

We plotted receiver operating characteristic curves to allow a simultaneous evaluation of sensitivity and specificity at different cut points for assessing the mammographic examination as positive. The cut points used were the Breast Imaging Reporting and Data System assessment categories ordered by increasing likelihood of cancer by using the hierarchy described previously. Receiver operating characteristic analysis corresponds to ordinal regression analysis of the ordered categories (19). We estimated the area under the receiver operating characteristic curve as our measure of overall accuracy. The area under the receiver operating characteristic curve can range from 0.5 (purely random performance) to 1.0 (perfect performance). The NLMIXED procedure in the statistical software (SAS/STAT, version 8; SAS Institute, Cary, NC) was used to fit an ordinal regression model. We included a scale parameter to allow the receiver operating characteristic curves to be nonsymmetrical. Also, the scale parameter was allowed to differ between the cases with cancer and cases without cancer, since the underlying latent (normal) distribution for cases may have a different standard deviation than that for cases without cancer. We produced the receiver operating characteristic curves for all ages combined and separately for mammography performed in women who were 40–49 years old.

We used unconditional logistic regression analysis to examine the effect of the number of MSPM on all performance measures while controlling for age, breast density, registry site, and reported presence of signs and symptoms. For the recall rate model, we used all mammographic examinations and estimated the odds of a positive mammographic assessment versus a negative mammographic assessment. For the PPV model, we included mammographic examinations with a positive assessment and estimated the odds ratio for a true-positive versus a false-positive mammographic examination. For sensitivity, we used all mammographic examinations associated with a cancer occurrence within the follow-up interval and estimated the odds for a true-positive mammographic examination versus a false-negative mammographic examination. For specificity, we used all mammographic examinations that were not associated with a cancer within the follow-up period and estimated the odds of a true-negative mammographic examination versus a false-positive mammographic examination. For the cancer detection rate model, all screening mammographic examinations were included, and we estimated the odds ratio for a true-positive mammographic examination versus all other mammographic examination assessments. An odds ratio greater than 1.00 indicates better performance. Since women may have had more than one screening mammographic examination included in the data, a repeated-measures analysis was also performed. The repeated-measures results did not vary from those presented. All statistical analyses were performed with software (SAS/STAT, version 8.2; SAS Institute).


   RESULTS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
DISCUSSION
 REFERENCES
 
Study Population Characteristics according to MSPM
Among 1 213 754 screening mammographic examinations, 70 196 (5.8%) were first (no previous) mammographic examinations, and almost half of them were performed in women in the 40–44-year age group; 1 143 558 (94.2%) were subsequent mammographic examinations. Of the subsequent mammographic examinations, 50.1% (572 639 of 1 143 558) were in the 9–15-MSPM group, with 13.0%, 18.5%, and 18.5%, respectively, in the other three MSPM groups (16–20, 21–27, and ≥28 months).

In 680 641 women, one screening mammographic examination was included in the data for 50.9% (346 768). Eighty-five percent (294 587) of these 346 768 women had undergone a previous examination at some time. In 27.6% (187 634) of all the women, two mammographic examinations were performed, and in 21.5% (146 239), three or more screening mammographic examinations were performed. In women in whom two or more mammographic examinations were performed, they were either both first and subsequent mammographic examinations or subsequent mammographic examinations only.

With the exception of age, the distribution of women’s characteristics associated with mammography across MSPM intervals varied only slightly (Table 2). Sixty-four percent of first mammographic examinations were performed in women in the two youngest age groups (47.3% of those 40–44 years old and 16.5% of those 45–49 years old), while subsequent mammographic examinations in these younger women were performed only 21% and 45% of the time in the 9–15 MSPM and the 28 or more MSPM time periods, respectively.


View this table:
[in this window]
[in a new window]

  		TABLE 2. Patient and Mammographic Characteristics according to MSPM

 
Hormone replacement therapy use was lowest among women who underwent first mammography (9.2%) compared with that among women who underwent subsequent mammography, which was 34% in the 9–15-MSPM category and which decreased to 20.2% in the 28-month or more MSPM interval. The percentage of women with a positive family history of breast cancer according to MSPM followed the same pattern as that of women who used hormone replacement therapy. There was little difference in the distribution of mammographic breast density across MSPM intervals in women who underwent subsequent mammography.

There were 6090 cancers diagnosed within 1 year of mammography; 483 were linked to first mammography, and 5607 were linked to subsequent mammography. Among women who underwent first mammography, 82.8% (400 of 483) of cancers were invasive and 17.2% (83 of 483) were cases of ductal carcinoma in situ; among women who underwent subsequent mammography, 80.3% (4504 of 5607) of cancers were invasive and 19.7% (1103 of 5607) were cases of ductal carcinoma in situ.

Performance Characteristics of Mammography
Overall performance.—For the entire population of mammographic examinations, performance results were as follows: recall rate, 8.2%; sensitivity, 77.8% (4735 of 6090); specificity, 92.1%, cancer detection rate, 3.9 per 1000; and PPV, 4.7%. All confidence intervals were narrow (≤ ±0.2% or 0.2 per 1000).

Performance results in women who underwent first mammography included the following: recall rate, 14.7% (95% confidence interval [CI]: 14.4%, 14.9%); sensitivity, 88.6% (428 of 483) (95% CI: 85.8%, 91.4%); specificity, 85.9% (95% CI: 85.6%, 86.1%); cancer detection rate, 6.1 per 1000 (95% CI: 5.5, 6.7); and PPV, 4.2% (95% CI: 3.8%, 4.5%). The percentage of invasive cancers detected at mammographic screening that were smaller than 1 cm was 25.3% (79 of 312) (95% CI: 20.5%, 30.1%).

Compared with women who underwent first mammography, all women who underwent subsequent mammography as a group had a significantly lower recall rate of 7.8% (95% CI: 7.8%, 7.9%), sensitivity of 76.8% (4307 of 5607) (95% CI: 75.7%, 77.9%), and cancer detection rate of 3.8 per 1000 (95% CI: 3.7, 3.9), and they had significantly higher specificity of 92.5% (95% CI: 92.5%, 92.5%) and PPV of 4.8% (95% CI: 4.7%, 4.9%) (P < .01). Among the women who underwent subsequent mammography, the percentage of invasive cancers detected at mammographic screening that were smaller than 1 cm was 39.5% (1194 of 3019) (95% CI: 37.8%, 41.3%), and this percentage was greater than that for women who underwent first mammography.

Performance characteristics according to MSPM.—Sensitivity increased with increasing number of MSPM, as observed with the following values: 70.9% (1859 of 2621) (95% CI: 69.2%, 72.7%) for 9–15 months, 75.7% (519 of 686) (95% CI: 72.4%, 78.9%) for 16–20 months, and 85.4% (945 of 1107) (95% CI: 83.3%, 87.4%) for 21–27 months. Sensitivity decreased slightly to 82.5% (984 of 1193) (95% CI: 80.3%, 84.6%) for the 28-month or more interval. Specificity decreased slightly with the number of MSPM from 93.3% (95% CI: 93.2%, 93.4%) for 9–15 months to 91.0% (95% CI: 90.9%, 91.2%) for the 28-month or more interval (Fig 2a).



View larger version (18K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 2a. Graphs show performance measures according to MSPM. (a) Sensitivity (black bars) and specificity (gray bars). (b) Recall rate (black bars), cancer detection rate per 1000 (gray bars), and PPV (white bars).

 


View larger version (12K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 2b. Graphs show performance measures according to MSPM. (a) Sensitivity (black bars) and specificity (gray bars). (b) Recall rate (black bars), cancer detection rate per 1000 (gray bars), and PPV (white bars).

 
Recall rate increased with the number of MSPM from 7.0% (95% CI: 6.9%, 7.1%) for 9–15 months to 9.4% (95% CI: 9.3%, 9.5%) for 28 months or more. Cancer detection rates per 1000 mammographic examinations increased slightly with the increase in MSPM: 3.2 per 1000 (95% CI: 3.1, 3.4) for 9–15 MSPM, 3.5 per 1000 (95% CI: 3.2, 3.8) for 16–20 MSPM, 4.5 per 1000 (95% CI: 4.2, 4.8) for 21–27 MSPM, 4.6 per 1000 (95% CI: 4.4, 4.9) for 28 or more MSPM, and 6.1 per 1000 (95% CI: 5.5, 6.7) for first mammographic examination. PPV showed a slight increase from 9–15 months to 21–27 months, with values of 4.6% (95% CI: 4.4%, 4.8%) and 5.1% (95% CI: 4.8%, 5.4%), respectively, with no further increase at 28 months or more (Fig 2b).

Performance according to age within time.—In regard to the distribution of sensitivity, recall rate, cancer detection rate, and PPV stratified by 5-year age groups stratified by the number of MSPM (Fig 3), there were general patterns stratified by age for all the measures. For sensitivity, cancer detection rate, and PPV, there was an increase in these measures with increasing age at all MSPM intervals. As MSPM increased, there was a larger increase in these performance measures (with the exception of sensitivity) among older women compared with younger women. Recall rates decreased with age but increased slightly with the number of MSPM. Specificity decreased with more MSPM, but specificity increased with age for each MSPM group. (Specificity was not shown in graphs.)



View larger version (61K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 3a. Performance measures according to MSPM and age in years, represented by numbers with A-H. (a) Sensitivity. (b) Recall rate. (c) Cancer detection rate (per 1000). (d) PPV.

 


View larger version (56K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 3b. Performance measures according to MSPM and age in years, represented by numbers with A-H. (a) Sensitivity. (b) Recall rate. (c) Cancer detection rate (per 1000). (d) PPV.

 


View larger version (38K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 3c. Performance measures according to MSPM and age in years, represented by numbers with A-H. (a) Sensitivity. (b) Recall rate. (c) Cancer detection rate (per 1000). (d) PPV.

 


View larger version (31K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 3d. Performance measures according to MSPM and age in years, represented by numbers with A-H. (a) Sensitivity. (b) Recall rate. (c) Cancer detection rate (per 1000). (d) PPV.

 
The receiver operating characteristic curves, controlling for 5-year age groups, registry site, breast density, and presence of signs and symptoms, are displayed in Figure 4. There was a significant increase in the area under the receiver operating characteristic curve as the number of MSPM increased for all ages combined (P < .001) and also among the curves for women 40–49 years old (P < .001), despite effectively no increase in cancer detection rates according to MSPM (Fig 3c). However, receiver operating characteristic curves for women 40–49 years old were lower compared with the curves for women of all ages.



View larger version (38K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 4a. Graphs show receiver operating characteristic curves according to MSPM for women of all ages and for the youngest women. (a) Women 40-89 years old. (b) Women 40-49 years old.

 


View larger version (35K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 4b. Graphs show receiver operating characteristic curves according to MSPM for women of all ages and for the youngest women. (a) Women 40-89 years old. (b) Women 40-49 years old.

 
In a logistic regression model, with control for 5-year age group, registry, breast density, and reported signs and symptoms, there was a significant increase in recall rate, PPV, and cancer detection rate (P < .001) with an increase in MSPM. There was a significant interaction between MSPM and age for all three of these measures, although this interaction of MSPM and age was weaker for PPV. The logistic regression analysis results show a significant association between MSPM and sensitivity and specificity. For specificity, the effect of MSPM was greater among older women compared with younger women (P < .001, test for interaction with age for specificity), but no difference was shown across ages for sensitivity (P = .15, test for interaction of MSPM with age for sensitivity).


   DISCUSSION
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
DISCUSSION
REFERENCES
 
In our study, we evaluated the effect of the number of MSPM on screening performance (sensitivity, specificity, recall rate, cancer detection rate, and PPV) in the United States. Women presented for screening mammography at varying intervals, with the majority returning at approximately 12 months, a significant proportion returning at nearly 24 months, and yet another sizable proportion returning between these two time intervals. Thus, it seems relevant to evaluate the effect of the MSPM on performance measures. We demonstrated that performance of screening mammography is influenced by elapsed time, and the effect of MSPM is stronger among older compared with younger (40–44 and 45–49 years old) women for all the performance measures except sensitivity. As breast density did not differ in distribution across MSPM intervals, breast density cannot be used to explain the difference in measures with time. With one exception, the highest sensitivity and lowest specificity are among women who underwent first mammography, and all performance measures increase with increasing MSPM between 9–15 months and 28 months or more, except for specificity, which decreases with increasing MSPM.

Of clinical importance is our finding that specificity differed 2%–6% with varying MSPM for women who underwent subsequent mammography. This is a small percentage, but it translates into a very large number of false-positive mammographic examination assessments, considering that most screening mammographic examinations are performed in women without cancer. For every 1000 women screened, only four to six will have cancer, so a decrease in specificity of 2%–6% means that 20–60 more women per 1000 screening mammographic examinations will have false-positive assessments. Though the explanation is not clear, specificity may be higher with shorter MSPM intervals between the times that mammographic examinations are performed. With longer MSPM, benign solid tumors, as well as malignant tumors, will increase in size. This may result in benign tumors having mammographic assessments interpreted as abnormal, adding to higher false-positive rates (20). Another explanation may be that women with mammographic examination assessments that are interpreted in the shorter time period have a higher percentage of lesions that were determined to be benign at a previous work-up, resulting in a lower probability that a follow-up work-up will be performed. It is also possible that women waiting longer periods to schedule mammography often self-identify problems, many of which are not cancer, and prompt physicians to do further work-ups.

We also determined that the probability of finding cancer was extremely low in women who were 40–44 years old with 9–15 MSPM. According to our results, there was no significant improvement in sensitivity with increased MSPM in these younger women. Although there was a 6% decreased specificity, PPV and cancer detection rates did not change with number of MSPM in the younger women. At all MSPM intervals, women in their 40s had lower cancer detection rates than did women who were 50–54 years old. Our receiver operating characteristic analysis demonstrated lower overall accuracy in these younger women. Women 40–49 years old should understand that the incidence of breast cancer is lower at their age, compared with that in older women, and there is a very low probability that a positive work-up will lead to a cancer diagnosis. They should also be aware of the importance of evaluation of a breast symptom even if a recent mammographic examination assessment is negative.

The results of our work add another factor to consider when results of audits of performance of screening mammography are interpreted. There is a sufficient effect of the number of MSPM within the subgroup of subsequent mammographic examinations to warrant considering the MSPM in the evaluation of performance, and certainly when comparisons of performance data are made.

Screening results (recall rates, cancer detection rates, and stage and size distributions), to date, have been reported at first and subsequent mammography, and the comparisons between these two major subgroups of mammographic examinations among countries are routinely made (21–29). In general, screening programs in Europe and the United Kingdom have screening intervals of 2 and 3 years. Researchers in two recently published articles compared screening mammography in the United States with that in the United Kingdom and other countries. In both, researchers concluded that when comparing screening mammography in the United States with screening in other European countries, recall rates are higher in the United States, with no gain in cancer detection rates (30,31). Just as in previous comparative studies, in neither of these recent studies was the issue of variation in time since previous mammography addressed when making comparisons for subsequent mammography. Given the policy in the United Kingdom of a longer screening interval than that in the United States, the high cancer detection rate does not seem surprising. Our results clearly show that in women 65 years and older, the longer the interval in MSPM, the higher the cancer detection rate.

We make comparisons between our performance measures and those of other countries with hesitation, with the understanding that rates are not necessarily comparable and not always available. Our recall rate for first mammography (14.7%) is higher than all other reported recall rates in other countries, and these rates range from 1.3% in the Netherlands (22), to 7.4% or 8.3% in the United Kingdom (30,32), and to 13.8% in Canada (24). Only Spain has higher reported rates (18.4%) (32). Because the screening population of women who underwent first mammography reported here comprised more than 50% of those who are younger than 50 years, and many European and other programs start screening at that age, the young age of our population may at least partially explain the high recall rate, as recall rate decreases with increasing age. The recall rates among the women who underwent subsequent mammography are also high in the United States (7.8%), compared with a range of 0.7%–8.6% elsewhere (21,22,24–26,33,34). In the majority of women who underwent subsequent mammography in our study, the MSPM was 9–15 months, compared with 24–36 months for other countries, and 28% of subsequent mammographic examinations in our study were performed in women 65 years or older. Having both a shorter interval of MSPM and a population of older women should contribute to lower recall rates. Therefore, it is difficult to assume that the higher U.S. recall rates among women who underwent subsequent mammography can be attributed to the number of MSPM or age alone. Reasons previously suggested, including legal concerns, volume of reading, lack of double reading, and different quality standards, are worth further study (30).

If we compare our cancer detection rates (6.1 per 1000) with those from other countries (range, 1.8–10.1 per 1000) that have been published (21,22,24–28,30,33–36), our rate in women who underwent first mammography is in the middle range. For women who underwent subsequent mammography, our rate of 3.8 per 1000 is also in the middle of the range reported from other countries (1.5–5.4 per 1000). In the longer MSPM intervals (21–27 and ≥28 months), our cancer detection rates were 4.5 per 1000 and 4.6 per 1000, respectively. Comparisons of the performance measures at 21–27 months and 28 months or more in the United States with the European data may be more accurate. Our PPV rates were low for women who underwent first and subsequent screening examinations compared with reported rates from the studies noted previously.

When we compared performance measures in women in the United States who underwent first mammography with those in women in Europe, we expected sensitivity to be lower, as women who undergo first mammography are in a younger population. Sensitivity reported here for women who underwent first mammography was 88.6% compared with a range of 74%–90% internationally (24,27,36). The country with the highest reported sensitivity, Canada, had a recall rate similar to ours (24). In Canada, the definition of cancer included lobular carcinoma in situ, and many women underwent a clinical breast examination prior to screening; both may have contributed to higher values than those reported here. In another report, the sensitivity was equal to that in our study, yet the recall rate (5%) and cancer detection rate (4.3 per 1000) (27) were lower. The higher sensitivity in these women may be explained by the inclusion of all probably benign assessments as positive mammographic examination assessments, which would increase the reported sensitivity. We found two publications in which sensitivity was higher for women who underwent subsequent mammography (81.4% and 91%) than that reported in our study (76.8%). In our study, however, sensitivity values for the longer MSPM intervals (21–27 and ≥28 months) were 85.4% and 82.5%, which were similar to these rates. In another U.S. study, a rate similar to the sensitivity in our study for women who underwent subsequent mammography was reported (37).

One possible explanation for the higher sensitivity values is that larger and more easily detected cancers were observed in women who underwent subsequent mammography because the longer time intervals between mammographic examinations would allow the tumors to grow. Researchers in previous studies have found that the proportions of small (<1 cm) invasive cancers are lower for women who undergo first mammography compared with women who undergo subsequent mammography: 23%–30% for women who undergo first mammography and 33%–37% for those who undergo subsequent mammography (21,22,24,25,27,33,38,39). Investigators in one U.S. study of a series of 120 breast cancers in women older than 64 years found significantly smaller and less advanced cancers with 1- versus 2-year screening intervals (40). We were unable to fully explore the results with respect to tumor size in our analysis, since data about size were missing in more than 10% of cancers diagnosed earlier in the observation period. The proportion of data that were missing decreased with calendar time, and this is an area for further study in the future.

There are some limitations to consider in this study. The performance measures depend on completeness of cancer reporting to tumor registries and pathology laboratories at the mammography registries. This reporting has been estimated to be more than 94.3% complete for the Breast Cancer Surveillance Consortium (41). Incomplete follow-up information could potentially have increased the sensitivity in our study, but this lack of information should not have affected the comparison across our MSPM intervals. We controlled for registry in our models to account for any possible difference in the practice of mammography or data collection across the seven registries. Last, it is important to note that regardless of the number of MSPM, all women were followed up after each mammographic examination for only 12 months for evidence of a cancer diagnosis. Our objective was to measure the effect of the elapsed time at screening. We did not design this study to examine the question of appropriate intervals for screening. In our data, the MSPM may differ in a series of mammographic examinations for one woman. Since for each mammographic examination a woman was followed up for 12 months regardless of the time that a woman may have waited to undergo subsequent mammography, we underestimated the number of cancers that may have developed subsequent to the mammographic examination, if the woman was not regularly screened at the same interval as that which we used. This has the potential effect of causing overestimation of the sensitivity gain with a longer MSPM interval and underestimation of the cancer detection rate and PPV for longer time intervals between screenings. Another limitation may result from our inferences concerning MSPM, which are based on the behavior of women who determine the timing of their mammographic examinations. The associations we see may be confounded by their behaviors. While certainly some of the differences seen are due to the number of MSPM, others may be due to differences among the health-seeking behaviors of the women.

A strength of our study is the very large number of women from counties with population characteristics that closely replicate those of the U.S. population for whom data were collected prospectively in the routine practice of mammography, with the same definitions and guidelines for quality control. Other researchers who compare their results with ours should use comparable definitions, since these will influence their conclusions. Similar to results in other studies, our results revealed higher recall rates in the United States, yet sensitivity was as good as, if not better than, sensitivity reported in other literature. One published study in the United States found that a recall rate of 5%–8% would maximize sensitivity without a loss of PPV (42). An effort to reduce recall rates in the United States needs more attention. In addition, U.S. mortality rates are lower than those reported in the United Kingdom and in many other countries in Europe (43–45). It has been suggested that the improved mortality rates may be explained as much by the increased proportion of small cancers that are detected outside of screening combined with adjuvant therapy as by improved screening performance (43). This is an area worthy of further study.

In conclusion, sensitivity, PPVs, cancer detection rates, and recall rates will increase with longer MSPM intervals, and these effects are stronger in older women for all measures except sensitivity. These factors should be considered when radiologists audit their performance. When researchers compare performance measures with those in studies in which data from screening programs with differing screening intervals are reported, it is important to understand the effect of MSPM on outcomes. U.S. recall rates remain higher than those in most other countries, even when the comparison is restricted to similar time intervals and accounts for age. However, when similar time intervals are used, the remaining U.S. performance measures are more similar to reported international rates. Time between mammographic examinations is an important factor to consider in the evaluation of screening performance.

Author affiliations: Division of Cancer Control and Population Sciences, Applied Research Program, National Cancer Institute, Bethesda, Md (S.H.T., R.B.B.); Center for Health Studies, Group Health Cooperative, Seattle, Wash (S.H.T., L.I.); Health Promotion Research, University of Vermont, Burlington, Vt (B.M.G.); Department of Radiology, University of New Mexico, Albuquerque, NM (R.D.R.); Department of Community & Family Medicine, Dartmouth Medical School, Norris Cotton Cancer Center, Lebanon, NH (P.A.C.); General Internal Medicine Section, Veterans Affairs Medical Center and Department of Medicine, UCSF, San Francisco, Calif (K.K.); University of Alabama Birmingham School of Public Health, Birmingham, Ala, and Cooper Institute, Denver, Colo (G.R.C.); and Cancer Research and Biostatistics and Department of Biostatistics, University of Washington, Seattle, Wash (W.E.B.).


   ACKNOWLEDGMENTS
 
The authors acknowledge Mark Dignan, PhD, Charles R. Key, MD, PhD, Charles R. Lynch, MD, and Nicole D. Urban, ScD, for their contribution to the development of the definitions and plans for this work. We are indebted to all the members of practices who contribute data about all their screening patients for the purpose of research. We also thank Kathleen C. Barry, who has been instrumental in helping us communicate throughout the entire process. Last, we thank all the members of the Breast Cancer Surveillance Consortium who have participated in data collection, data quality control, and development of the research plan and who have provided constructive criticism every step along the way.


   FOOTNOTES
 
Abbreviations: CI = confidence interval, MSPM = months since previous mammography, PPV = positive predictive value

Authors stated no financial relationship to disclose.

All opinions are those of the authors and cannot be construed to imply the opinion or endorsement of the Federal Government or the National Cancer Institute.

Author contributions: Guarantors of integrity of entire study, B.C.Y., S.H.T., B.M.G., P.A.C., K.K., R.B.B., W.E.B.; study concepts, B.C.Y., S.H.T., B.M.G., R.D.R., P.A.C., K.K., R.B.B., G.R.C., W.E.B.; study design, all authors; literature research, B.C.Y., S.H.T., B.M.G., R.D.R., P.A.C., K.K., R.B.B.; data acquisition and data analysis/interpretation, all authors; statistical analysis, B.C.Y., L.I., R.B.B., G.R.C.; manuscript preparation, B.C.Y., S.H.T., L.I., B.M.G., R.D.R., P.A.C., K.K., R.B.B., W.E.B.; manuscript definition of intellectual content, editing, and final version approval, all authors; manuscript revision/review, B.C.Y., S.H.T., L.I., B.M.G., P.A.C., K.K., R.B.B., G.R.C., W.E.B.


   REFERENCES
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
DISCUSSION
 REFERENCES
 

  1. Frazier EL, Jiles RB, Mayberry R. Use of screening mammography and clinical breast examinations among black, Hispanic, and white women. Prev Med 1996; 25:118-125.[CrossRef][Medline]
  2. Blackman DK, Bennett EM, Miller DS. Trends in self-reported use of mammograms (1989–1997) and Papanicolaou tests (1991–1997): behavioral risk factor surveillance system. MMWR CDC Surveill Summ 1999; 48:1-22.[Medline]
  3. Breen N, Wagener DK, Brown ML, Davis WW, Ballard-Barbash R. Progress in cancer screening over a decade: results of cancer screening from the 1987, 1992, and 1998 National Health Interview Surveys. J Natl Cancer Inst 2001; 93:1704-1713.[Abstract/Free Full Text]
  4. Smith RA, Cokkinides V, Eyre HJ, American Cancer Society. American Cancer Society guidelines for the early detection of cancer, 2003. CA Cancer J Clin 2003; 53:27-43.[Abstract/Free Full Text]
  5. U.S. Preventive Services Task Force. Screening for breast cancer: recommendations and rationale. Ann Intern Med 2002; 137(5 pt 1):344-346.
  6. Ballard-Barbash R, Taplin SH, Yankaskas BC, et al. Breast Cancer Surveillance Consortium: a national mammography screening and outcomes database. AJR Am J Roentgenol 1997; 169:1001-1008.[Free Full Text]
  7. National Cancer Institute. Available at: http://breastscreening.cancer.gov/. Accessed 2003.
  8. Carney PA, Geller BM, Moffett H, et al. Current medicolegal and confidentiality issues in large, multicenter research programs. Am J Epidemiol 2000; 152:371-378.[Abstract/Free Full Text]
  9. D’Orsi C, Bassett LW, Feig SA, et al. Breast Imaging Reporting and Data System 3rd ed. Reston, Va: American College of Radiology, 1998; 165-181.
  10. Rosenberg RD, Yankaskas BC, Hunt WC, et al. Effect of variations in operational definitions on performance estimates for screening mammography. Acad Radiol 2000; 7:1058-1068.[CrossRef][Medline]
  11. Taplin SH, Ichikawa LE, Kerlikowske K, et al. Concordance of breast imaging reporting and data system assessments and management recommendations in screening mammography. Radiology 2002; 222:529-535.[Abstract/Free Full Text]
  12. Carney PA, Miglioretti DL, Yankaskas BC, et al. Individual and combined effects of age, breast density, and hormone replacement therapy use on the accuracy of screening mammography. Ann Intern Med 2003; 138:168-175.[Abstract/Free Full Text]
  13. Litherland JC, Evans AJ, Wilson AR. The effect of hormone replacement therapy on recall rate in the National Health Service Breast Screening Programme. Clin Radiol 1997; 52:276-279.[CrossRef][Medline]
  14. Stomper PC, Van Voorhis BJ, Ravnikar VA, Meyer JE. Mammographic changes associated with postmenopausal hormone replacement therapy: a longitudinal study. Radiology 1990; 174:487-490.[Abstract/Free Full Text]
  15. Berkowitz JE, Gatewood OM, Goldblum LE, Gayler BW. Hormone replacement therapy: mammographic manifestations. Radiology 1990; 174:199-201.[Abstract/Free Full Text]
  16. Cyrlak D, Wong CH. Mammographic changes in postmenopausal women undergoing hormonal replacement therapy. AJR Am J Roentgenol 1993; 161:1177-1183.[Abstract/Free Full Text]
  17. Laya MB, Larson EB, Taplin SH, White E. Effect of estrogen replacement therapy on the specificity and sensitivity of screening mammography. J Natl Cancer Inst 1996; 88:643-649.[Abstract/Free Full Text]
  18. Kerlikowske K, Carney PA, Geller B, et al. Performance of screening mammography among women with and without a first-degree relative with breast cancer. Ann Intern Med 2000; 133:855-863.[Abstract/Free Full Text]
  19. Tosteson AN, Begg CB. A general regression methodology for ROC curve estimation. Med Decis Making 1988; 8:204-215.
  20. Williams RS, Brook D, Monypenny IJ, Gower-Thomas K. The relevance of reported symptoms in a breast screening programme. Clin Radiol 2002; 57:725-729.[CrossRef][Medline]
  21. Doyle TC, Elwood JM, Smale P, et al. Clinical outcomes of the Otago-Southland Breast Cancer Screening Programme 1991–1996. N Z Med J 1998; 111:380-383.[Medline]
  22. Fracheboud J, de Koning HJ, Beemsterboer PM, et al. Nation-wide breast cancer screening in the Netherlands: results of initial and subsequent screening 1990–1995. National Evaluation Team for Breast Cancer Screening. Int J Cancer 1998; 75:694-698.
  23. Kavanagh AM, Mitchell H, Giles GG. Hormone replacement therapy and accuracy of mammographic screening. Lancet 2000; 355:270-274.[CrossRef][Medline]
  24. Libstug AR, Moravan V, Aitken SE. Results from the Ontario breast screening program, 1990–1995. J Med Screen 1998; 5:73-80.[Abstract/Free Full Text]
  25. Paquette D, Snider J, Bouchard F, et al. Performance of screening mammography in organized programs in Canada in 1996. The Database Management Subcommittee to the National Committee for the Canadian Breast Cancer Screening Initiative. Can Med Assoc J 2000; 163:1133-1138.
  26. Thurfjell E. Population-based mammography screening in clinical practice: results from the prevalence round in Uppsala county. Acta Radiol 1994; 35:487-491.[Medline]
  27. Vizcaino I, Salas D, Vilar JS, Ruiz-Perales F, Herranz C, Ibanez J. Breast cancer screening: first round in the population-based program in Valencia, Spain. Collaborative Group of Readers of the Breast Cancer Screening Program of the Valencia Community. Radiology 1998; 206:253-260.
  28. Dean PB, Pamilo M. Screening mammography in Finland: 1.5 million examinations with 97 percent specificity. Mammography Working Group, Radiological Society of Finland. Acta Oncol 1999; 38(suppl 13):47-54.
  29. Blanks RG, Moss SM, Patnick J. Results from the UK NHS breast screening programme, 1994–1999. J Med Screen 2000; 7:195-198.[Abstract/Free Full Text]
  30. Smith-Bindman R, Chu PW, Miglioretti DL, et al. Comparison of screening mammography in the United States and the United Kingdom. JAMA 2003; 290:2129-2137.[Abstract/Free Full Text]
  31. Elmore JG, Nakano CY, Koepsell TD, Desnick LM, D’Orsi CJ, Ransohoff DF. International variation in screening mammography interpretations in community-based programs. J Natl Cancer Inst 2003; 95:1384-1393.[Abstract/Free Full Text]
  32. Lynge E, Olsen AH, Fracheboud J, Patnick J. Reporting of performance indicators of mammography screening in Europe. Eur J Cancer Prev 2003; 12:213-222.[CrossRef][Medline]
  33. Robinson JI, Crane CE, King JM, Scarce DI, Hoffmann CE. The South Australian Breast X-Ray Service: results from a statewide mammographic screening programme. Br J Cancer 1996; 73:837-842.[Medline]
  34. Olivotto IA, Kan L, d’Yachkova Y, et al. Ten years of breast screening in the Screening Mammography Program of British Columbia, 1988–97. J Med Screen 2000; 7:152-159.[Abstract/Free Full Text]
  35. Doyle GJ, McLean L. Unilateral increase in mammographic density with hormone replacement therapy. Clin Radiol 1994; 49:50-51.[CrossRef][Medline]
  36. Kavanagh AM, Giles GG, Mitchell H, Cawson JN. The sensitivity, specificity, and positive predictive value of screening mammography and symptomatic status. J Med Screen 2000; 7:105-110.[Abstract/Free Full Text]
  37. Kerlikowske K, Grady D, Barclay J, Sickles EA, Ernster V. Likelihood ratios for modern screening mammography: risk of breast cancer based on age and mammographic interpretation. JAMA 1996; 276:39-43.[Abstract]
  38. Rickard M, Donnellan M. Diagnosis of small sized invasive breast cancer by an Australian mammography screening service: surrogate end-points for mortality reduction. Aust N Z J Surg 1998; 68:415-418.[Medline]
  39. Frankel SD, Sickles EA, Curpen BN, Sollitto RA, Ominsky SH, Galvin HB. Initial versus subsequent screening mammography: comparison of findings and their prognostic significance. AJR Am J Roentgenol 1995; 164:1107-1109.[Abstract/Free Full Text]
  40. Field LR, Wilson TE, Strawderman M, Gabriel H, Helvie MA. Mammographic screening in women more than 64 years old: a comparison of 1- and 2-year intervals. AJR Am J Roentgenol 1998; 170:961-965.[Abstract/Free Full Text]
  41. Ernster VL, Ballard-Barbash R, Barlow WE, et al. Detection of ductal carcinoma in situ in women undergoing screening mammography. J Natl Cancer Inst 2002; 94:1546-1554.[Abstract/Free Full Text]
  42. Yankaskas BC, Cleveland RJ, Schell MJ, Kozar R. Association of recall rates with sensitivity and positive predictive values of screening mammography. AJR Am J Roentgenol 2001; 177:543-549.[Abstract/Free Full Text]
  43. Jatoi I, Miller AB. Why is breast-cancer mortality declining? Lancet Oncol 2003; 4:251-254.[CrossRef][Medline]
  44. Quinn MJ. Cancer trends in the United States: a view from Europe. J Natl Cancer Inst 2003; 95:1258-1261.[Free Full Text]
  45. Sant M, Allemani C, Berrino F, et al. Breast carcinoma survival in Europe and the United States. Cancer 2004; 100:715-722.[CrossRef][Medline]



This article has been cited by other articles:


Home page
 JNCI J Natl Cancer InstHome page
S. Hofvind, P. M. Vacek, J. Skelly, D. L. Weaver, and B. M. Geller
Comparing Screening Mammography for Early Breast Cancer Detection in Vermont and Norway
J Natl Cancer Inst, August 6, 2008; 100(15): 1082 - 1091.
[Abstract] [Full Text] [PDF]

Home page
 JNCI J Natl Cancer InstHome page
S. Taplin, L. Abraham, W. E. Barlow, J. J. Fenton, E. A. Berns, P. A. Carney, G. R. Cutter, E. A. Sickles, D. Carl, and J. G. Elmore
Mammography Facility Characteristics Associated With Interpretive Accuracy of Screening Mammography
J Natl Cancer Inst, June 18, 2008; 100(12): 876 - 887.
[Abstract] [Full Text] [PDF]

Home page
 Am. J. Roentgenol.Home page
M. Gromet
Comparison of Computer-Aided Detection to Double Reading of Screening Mammograms: Review of 231,221 Mammograms
Am. J. Roentgenol., April 1, 2008; 190(4): 854 - 859.
[Abstract] [Full Text] [PDF]

Home page
 JCOHome page
T. L. Lash, M. P. Fox, D. S.M. Buist, F. Wei, T. S. Field, F. J. Frost, A. M. Geiger, V. P. Quinn, M. U. Yood, and R. A. Silliman
Mammography Surveillance and Mortality in Older Breast Cancer Survivors
J. Clin. Oncol., July 20, 2007; 25(21): 3001 - 3006.
[Abstract] [Full Text] [PDF]

Home page
 NEJMHome page
S. Ciatto, N. Houssami, D. Gur, R. M. Nishikawa, R. A. Schmidt, C. E. Metz, J. F. Ruiz, S. A. Feig, R. L. Birdwell, M. N. Linver, et al.
Computer-aided screening mammography.
N. Engl. J. Med., July 5, 2007; 357(1): 83 - 84.
[Full Text] [PDF]

Home page
 NEJMHome page
J. J. Fenton, S. H. Taplin, P. A. Carney, L. Abraham, E. A. Sickles, C. D'Orsi, E. A. Berns, G. Cutter, R. E. Hendrick, W. E. Barlow, et al.
Influence of Computer-Aided Detection on Performance of Screening Mammography
N. Engl. J. Med., April 5, 2007; 356(14): 1399 - 1409.
[Abstract] [Full Text] [PDF]

Home page
 RadiologyHome page
J. D. Keen, J. E. Keen, M. A. Ganott, J. H. Sumkin, and D. Gur
Does the Marginal Benefit Exceed the Marginal Cost?
Radiology, April 1, 2007; 243(1): 299 - 300.
[Full Text] [PDF]

Home page
 Am J EpidemiolHome page
D. L. Miglioretti and P. J. Heagerty
Marginal Modeling of Nonnested Multilevel Data using Standard Software
Am. J. Epidemiol., February 15, 2007; 165(4): 453 - 463.
[Abstract] [Full Text] [PDF]

Home page
 Am. J. Roentgenol.Home page
P. A. Carney, L. A. Abraham, D. L. Miglioretti, K. R. Yabroff, E. A. Sickles, D. S. M. Buist, C. J. Kasales, B. M. Geller, R. D. Rosenberg, M. B. Dignan,