| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
1 The Sydney-Square Breast Clinic, Medical Benefits Fund of Australia, Level 12,
97-99 Bathurst St., Sydney, New South Wales 2000, Australia.
2 School of Public Health, Faculty of Medicine, A27, The University of Sydney,
New South Wales 2006, Australia.
3 NSW Breast Cancer Institute, Westmead Hospital, P. O. Box 143, Westmead, New
South Wales 2145, Australia.
4 Northern Sydney and Central Coast BreastScreen, The Cottage, Royal North Shore
Hospital, St. Leonards, New South Wales 2065, Australia.
5 Department of Diagnostic Radiology, Royal North Shore Hospital, St. Leonards,
New South Wales 2065, Australia.
Received May 13, 2002;
accepted after revision September 3, 2002.
Address correspondence to N. Houssami.
Abstract
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
MATERIALS AND METHODS. Four hundred eighty subjects were sampled from all women consecutively attending a symptomatic breast clinic between 1994 and 1996 and ranging in age from 25 to 55 years. We included all 240 women shown to have breast cancer (thus avoiding selection bias) and 240 age-matched women shown not to have cancer. Mammograms and sonograms were prospectively interpreted independently and without knowledge of age by two radiologists in a blinded manner, with a third radiologist arbitrating disagreements. Sensitivity and specificity of each imaging test in relation to age were examined using logistic regression modeling, and accuracy was compared using the chi-square test for paired proportions.
RESULTS. Sensitivity and specificity of each test were not linearly
associated with age; however, the sensitivity of mammography increased
substantially in women older than 50 years. Sonographic sensitivity of 81.7%
was not significantly greater than mammographic sensitivity of 75.8%
(
21 = 2.06, p = 0.15). However, in women
45 years old or younger, the sensitivity of sonography was 13.2% (95%
confidence interval, 2.1–24.3%) greater than that of mammography. The
specificity of both tests was approximately 88.0%.
CONCLUSION. These data show that sonography is the more accurate imaging test in women 45 years old or younger who present with breast symptoms and may be an appropriate initial imaging examination.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
A comprehensive review of the literature found little evidence about the comparative age-specific accuracy of mammography and breast sonography in symptomatic women [2]. A recent study has subsequently provided the first empiric evidence regarding the comparative sensitivity of both imaging tests in symptomatic women who underwent both examinations [3]. That study found that sonography was more sensitive than mammography in women younger than 62 years, the so-called crossover age, and mammography was more sensitive than sonography in women older than 62 years [3]. However, the study's authors acknowledged that the non-independent interpretation of the two tests and the analysis used may have led to under-estimation of the sensitivity of mammography, and that the crossover age may be as early as 48 years [3]. For a valid comparison of the accuracy of two tests, the tests need to be interpreted independently (without knowledge of each other) in the same subjects [4].
In this article, we report findings from the Sydney Breast Imaging Accuracy Study, the first study designed to establish the age-related accuracy of diagnostic mammography and sonography using independent (and blinded) interpretation of the two imaging tests. The study examines the comparative sensitivity and specificity of the two tests in symptomatic women 55 years old or younger. This study was of a clinically referred population and makes no inferences as to the accuracy or merits of screening mammography.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Cancer patients included all women in the defined inclusion criteria proven to have a breast malignancy. All women who were recommended to have surgery or who underwent surgery regardless of the clinic's recommendation, were followed up as standard monitoring procedure in the study center. Final histologic diagnosis was obtained for all patients who underwent surgical biopsy, and all cancer cases were verified by reviewing the histopathology report. Two cancer patients with typical malignant clinical findings and proven malignancy at cytology were not treated surgically but were not excluded from the study. Cancer cases included all types of invasive carcinoma as well as ductal carcinoma in situ. Eighteen eligible patients (7%) were excluded because images were not available.
To avoid verification bias [5, 6], subjects without cancer consisted of a randomly generated sample of consecutive women from the same population (irrespective of whether they had undergone biopsy) who were age-matched to cancer patients by year of birth. A subject was classified as not having breast cancer on the basis of the diagnostic assessment and on the basis of not being registered as having a breast cancer diagnosis at the New South Wales Central Cancer Registry during a 2-year interval after attending the study center. (Notification of malignant neoplasms is a statutory requirement in the state of New South Wales.) Two subjects without cancer were age-matched per each cancer patient. However, only one subject without cancer was included for each patient, using the following method to maximize the number of subjects without cancer from this randomly selected group who had undergone both tests. When both age-matched subjects without cancer had undergone only one test or both tests (27%), one was selected randomly by a coin toss. When only one of the two age-matched subjects without cancer had undergone both imaging tests (73%), that subject was chosen for inclusion. Because this method may theoretically lead to subjects without cancer having denser mammograms (and hence a higher specificity for mammography), we explored any potential effect this method may have in the analysis. The method has no potential influence on test sensitivity, however, because it applied only to subjects without cancer.
Medical records of all subjects were reviewed to verify eligibility, and
all images performed at the time of diagnostic assessment were retrieved. In
the study center, mammography is performed before sonography in all women
older than 30 years who are referred for evaluation of symptoms. Sonography is
performed in all women younger than 30 years and whenever a localized or focal
finding (such as asymmetric thickening, lump) is noted on breast clinical
examination regardless of the level of clinical suspicion. Mammography and
sonography are almost always used in combination whenever some suspicion of
malignancy exists regarding any individual component of the assessment.
Mammography consisted of baseline mediolateral oblique and craniocaudal
projections and all additional workup images obtained at the time of
evaluation (
60% of subjects had workup films). Workup films included
films taken with a lead skin marker indicating the clinical area or modified
projections that accommodate the clinical area. All sonographic images taken
at the time of diagnosis were included. Sonography was performed with
knowledge of the mammographic results by accredited sonographers with
experience in breast sonography. It is routine practice in the study center to
scan and include an image (in at least two planes) of the symptomatic area or
the area palpated on clinical examination (as indicated by the examining
clinician). The limitation of interpreting sonographic images on hard copy,
rather than in real-time, is acknowledged and is explored further in the
Discussion.
For the years included in the study, mammography was performed using one of three dedicated machines (Senograph 600T FD HF and 600T HF; General Electric Medical Systems, Milwaukee, WI), all of which used film-screen techniques. Sonographic equipment early in the study also included Aloka sonographic units (SSD-630; Aloka, Tokyo, Japan) with a 7.5-MHz transducer (mechanical sector probe). Two other units were used in 1995–1996: an Apogee 800 scanner (11.5-MHz linear array transducer; Advanced Technology Laboratories, Bothell, WA) and an SSA 250 scanner (7.5-MHz annular array transducer; Toshiba Medical Systems Division, Tokyo, Japan).
Prospective Reporting of Images in a Blinded Manner
For all subjects in the study, images (which were collected
retrospectively) were prospectively interpreted by two radiologists who had no
knowledge of the initial imaging reports or of the subjects' final diagnoses.
The radiologists were aware of the study's aim and design but were unaware of
the number and distribution of subjects and their ages. Disagreement between
the two radiologists was resolved using a third radiologist, who was asked to
arbitrate in disagreements [7]
and given the other radiologists' opinions. All three radiologists had
extensive experience in mammography and breast sonography and interpretation.
Radiologists' reports were recorded on standardized forms using a categoric
scale of 1–5 with an increasing level of suspicion (1 = no significant
abnormality, 2 = benign, 3 = indeterminate, 4 = suspicious, and 5 =
malignant). Radiologists were requested to interpret sonograms with
consideration of published criteria
[8].
To ensure independent interpretation of the two imaging tests and blinding of the radiologists performing the prospective interpretations, we used the following methods: the interpretation of each imaging test was maintained independent of knowledge of the other imaging test findings by completely separating mammographic and sonographic films, with each set of films being interpreted on separate days; films were presented to radiologists in batches of either mammography or sonography, with batches being interpreted at least 2–4 weeks apart; subjects' names and ages were deleted from images, thus eliminating any possibility of associating the mammographic films with the corresponding sonographic films; reviewers were separated (interpreting of films was performed by the radiologists on separate days to avoid any potential for discussion). Similarly, arbitration of disagreements was done on a separate day. Film interpretation throughout the phases of the study was performed in a separate room dedicated for this purpose, with restricted access while participating radiologists were interpreting.
Data Analysis
The prevalence of breast cancer in the study population was calculated. The
overall sensitivity and specificity of each test were calculated. Categories 1
and 2 were considered negative for cancer, and categories 3–5, positive
for cancer (in line with clinical practice in the study center, where imaging
findings reported as categories 3–5 are recommended for further
investigation). The accuracy of each test was also calculated in 5-year age
groups and presented in tables and graphs. Sensitivity and specificity of the
two tests were compared in all subjects and for the different age groups using
McNemar's chi-square test for paired proportions (accuracy of the two imaging
tests in subjects); the 95% confidence interval (CI) for the difference
between paired proportions was also calculated.
Logistic regression analysis and cross-tabulations were performed using statistical software (version 6.12; SAS Institute, Cary, NC). The logarithm of the odds of a positive result was regressed on age (as a continuous variable) for cancer patients for each imaging modality to assess the influence of age on test sensitivity. The logarithm odds of a negative result were regressed on age for subjects without cancer for each imaging modality to assess the influence of age on test specificity. We cross-tabulated mammographic findings by sonographic findings in patients with cancer and subjects without cancer.
A post hoc analysis was performed to compare the sensitivity of the two tests in two age groups, older and younger than 45 years, using McNe-mar's chi-square test. We also tested whether the difference in sensitivity of the two tests was significant between the two age groups by testing the interaction between the test and the dichotomized age groups using logistic regression, taking into account the paired data using a generalized estimating equations model [9].
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
|
|
|
|
|
Figure 1 shows that for
women younger than 50 years, the sensitivity of both tests in relation to age
has little variability. However, the sensitivity of mammography increases
substantially after age 50. For women 46–55 years old, the sensitivity
of the two tests becomes similar (Fig.
1) and the specificity of sonography and mammography are also
similar (Fig. 2).
Cross-tabulation of mammographic and sonographic results is shown in Tables
4 and
5. The sonographic sensitivity
of 81.7% was 5.9% (95% CI, –1.5% to 13.2%) greater than the mammographic
sensitivity of 75.8%. The difference in the sensitivities of the two imaging
tests, however, was not statistically significant
(21 = 2.06, p = 0.15).
|
|
For the 233 women without cancer who had both tests
(Table 5), the specificity of
sonography was 88.0%, which was not significantly different from the
specificity of mammography (87.6%) (21 = 0,
p = 1), with a difference in test specificity of 0.4% (95% CI,
–5.0% to 5.8%). The specificity of mammography in women chosen for
having had both tests was only 2% higher than the specificity in women
randomly included, indicating that the method used for inclusion of subjects
without cancer had little effect on estimates of specificity for
mammography.
Table 4 shows that, of all cancers, 20% were correctly identified as cancer on sonography but not correctly identified as cancer on mammography, and 14% were correctly identified on mammography but not on sonography. Combining the two imaging tests, in which case imaging is considered positive if either test is positive, gives a high sensitivity of approximately 96%. That is, about 4% of cancers will be misdiagnosed when combining mammography and sonography in the investigation of symptomatic women. Table 5 shows that, of all women who did not have breast cancer, about 9% had false-positive findings on one test but were correctly identified as not having cancer on the other test. Combining the two tests, if imaging is considered negative when both tests have negative findings, gives a specificity of 79%. If both tests are combined in the investigation of symptomatic women, approximately 3% of women who do not have breast cancer will have a false-positive result on both mammography and sonography.
In women 45 years old or younger, the sensitivity of sonography was 84.9%,
which was 13.2% (95% CI, 2.1%–24.3%) greater than the sensitivity of
mammography (71.7%), indicating a significant difference in sensitivity for
the two tests (21 = 4.45, p = 0.035). In
women 46–55 years old, the sensitivity of sonography was 79.1%, which
was the same as the sensitivity of mammography of 79.1%
(21 = 0, p = 1). Testing for interaction
between test type and dichotomized age group showed weak evidence (p
= 0.08) of a difference between the two age groups in the difference in
sensitivity of the two tests.
Regression of the probability of positive mammographic results on age in
cancer patients (sensitivity) was not significant (likelihood ratio
21 = 1.23, p = 0.27). Regression of the
probability of a positive sonographic result on age in cancers was also not
significant (likelihood ratio 21 = 0.16, p
= 0.69). For every 10 years' increase in age, the odds ratio for the
probability of a positive mammographic result is 1.28 (95% CI,
0.83–1.97), and the odds ratio for the probability of a positive
sonographic result is 0.91 (95% CI, 0.55–1.48). These results indicate
that no significant linear relationship exists between age and the logarithm
odds of a positive mammographic or sonographic result. Similarly, regression
of the probability of a negative mammographic result on age in subjects
without cancer (specificity) was not significant for mammography (likelihood
ratio 21 = 0.06, p = 0.80), and the odds
ratio for a negative result for every 10 years' increase in age is 0.93 (95%
CI, 0.53–1.64). Regression of the probability of a negative sonographic
result on age in subjects without cancer was also not significant (likelihood
ratio 21 = 0.19, p = 0.66); and for every
10 years' increase in age, the odds ratio for a negative result is 0.88 (95%
CI, 0.48–1.42).
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The study contributes several findings to our knowledge of the accuracy of mammography and sonography in relation to age in women with breast symptoms. The greater sensitivity of mammography in women 50 years or older relative to younger women has been shown in other studies that have considered mammography only [10, 11, 12]. However, unlike some of these studies that show a progressive improvement in sensitivity with age, the sensitivity of mammography in our data is not linearly associated with age. The sensitivity of mammography in the Sydney Breast Imaging Accuracy Study remains relatively constant in women 45 years and younger and increases in women older than 45 years and, to a greater extent, in women 51–55 years.
Overall, the difference in the sensitivity of the two tests in all subjects is not statistically significant; however, in women 45 years or younger, sonography has a significantly greater sensitivity than mammography and a similar specificity. Although 45 years was a post hoc choice of age dichotomy, on the basis of these data sonography may be the more appropriate initial test for investigating women with symptoms who are 45 years old or younger. It may be argued that the age of 45 years, as a means of deciding first-line breast imaging in women with symptoms, is later than expected. However, from a biologic perspective, this age as the decision basis makes sense because it correlates with the transition of hormonal (menopausal) status. In addition, despite differences in study design and analysis, this finding is in keeping with the recent study based on data from Florence that has also shown this crossover age to occur in the range of 48–62 years [3] and later than expected on the basis of expert opinion.
Our study also shows that there is little difference in the specificity of the two imaging tests, and that specificity is not influenced by age for either mammography or sonography. This fact may explain the different findings in published studies, with some reporting a greater specificity for sonography than for mammography [8, 11, 13], and others reporting a greater specificity for mammography than for sonography [14, 15]. Although seven of our subjects without cancer had not undergone sonography (< 2% of subjects), that fact is not likely to have any effect on comparative specificity because the proportion is very small and because specificity is approximately the same for both tests in these data. That fact also has no effect on test sensitivity because it does not apply to cancer patients and therefore does not invalidate our recommendations based on the differences in test sensitivity.
It may be argued that sonography is a real-time imaging test and any judgment of its accuracy should be based on real-time or dynamic interpretation, and that using hard-copy sonographic images in this study may lead to underestimation of the accuracy of sonography. Although this potential effect is acknowledged, dynamic features basically consist of compressibility and, to a lesser extent, mobility, which are not considered primary features of breast lesions seen on sonography [16]. We also point out that research undertaken to define breast sonographic criteria has similarly used sonographic hard-copy images [17]. This potential effect may be balanced by other factors inherent in sonography that may potentially overestimate its accuracy. Because sonography was performed with knowledge of the clinical opinion and the mammographic result, the operator may have been influenced by the level of suspicion or other information obtained from these tests. The hard-copy image is also likely to be the most representative image of the lesion obtained by systematic scanning. In summary, we believe that using a hard copy of sonographic films gives a realistic and valid estimate of test accuracy. Although using hard-copy sonograms prevents assessment of the dynamic features of lesions, which may potentially underestimate accuracy, this effect is likely to be balanced by a potential to slightly overestimate the performance of sonography by presenting the most representative image of the lesion.
Our study shows that combining both mammography and sonography has a greater sensitivity (96%) than either sonography alone (81.7%) or mammography alone (75.8%), although combining reduces specificity. These findings are in keeping with other studies of the accuracy of combined mammography and sonography, which have been extensively discussed in a recent paper that included an appraisal of published work on this topic [18].
Using the age of 45 years to decide the choice of first-line imaging in women with breast symptoms raises the question of whether mammography is appropriate as second-line imaging in women younger than 45 years. Apart from the individual clinical situation that guides decision making, the additional use of mammography may be decided on the basis of the sonographic result and the prevalence of breast cancer in the population being investigated (which is an indicator of the woman's probability of having the disease before test findings). If the sonographic result is positive, most clinicians would agree that mammography is indicated to define the extent of disease, in addition to contributing further diagnostic information. However, if the sonographic findings are negative, the benefit of mammography is related to the remaining breast cancer prevalence, once sonography has identified (and therefore removed) most cancers.
As an example, in a population of 1000 women younger than 45 years and with symptoms, in which the initial breast cancer prevalence is 2%, sonography will identify approximately 85% of cancers, so the remaining breast cancer prevalence is 0.3%. Mammography in this context would identify two additional cancers not identified on sonography and cause an additional falsepositive result in 85 women who do not have breast cancer. In a population of 1000 women with symptoms in whom the initial breast cancer prevalence is 10%, mammography would identify 11 additional cancers not identified on sonography and cause an additional false-positive result in 78 women who do not have breast cancer.
We have shown that breast sonography is more accurate than mammography in symptomatic women 45 years or younger and may be an appropriate initial imaging test in those women. We recognize that the evidence from this research is contrary to current thinking and clinical practice in breast imaging of symptomatic women, and we recommend further research that examines this issue using a prospective study. Future research should also examine how hormonal status (rather than age alone) influences the accuracy of both tests. We suggest that any change from current practice should be introduced initially in the context of dedicated breast services, in which systematic sonographic scanning and monitoring of accuracy are routine practice, before wider implementation.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  M. R. Del Turco, P. Mantellini, S. Ciatto, R. Bonardi, F. Martinelli, B. Lazzari, and N. Houssami Full-Field Digital Versus Screen-Film Mammography: Comparative Accuracy in Concurrent Screening Cohorts Am. J. Roentgenol., October 1, 2007; 189(4): 860 - 866. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 O. Graf, T. H. Helbich, G. Hopf, C. Graf, and E. A. Sickles Probably Benign Breast Masses at US: Is Follow-up an Acceptable Alternative to Biopsy? Radiology, July 1, 2007; 244(1): 87 - 93. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 R S Saunders and E Samei Improving mammographic decision accuracy by incorporating observer ratings with interpretation time Br. J. Radiol., December 1, 2006; 79(Special_Issue_2): S117 - S122. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
W. A. Berg, J. D. Blume, J. B. Cormack, and E. B. Mendelson Operator Dependence of Physician-performed Whole-Breast US: Lesion Detection and Characterization. Radiology, November 1, 2006; 241(2): 355 - 365. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S.-P. Cheng, C.-L. Liu, and Y.-C. Chang Extraintestinal Salmonella Infection Appearing as a Suspicious Breast Mass J. Ultrasound Med., February 1, 2006; 25(2): 277 - 279. [Full Text] [PDF]
|
|
|
|
|
|
W. A. Berg, L. Gutierrez, M. S. NessAiver, W. B. Carter, M. Bhargavan, R. S. Lewis, and O. B. Ioffe Diagnostic Accuracy of Mammography, Clinical Examination, US, and MR Imaging in Preoperative Assessment of Breast Cancer Radiology, December 1, 2004; 233(3): 830 - 849. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
D. B. Kopans, N. Houssami, L. Irwig, and S. Blome Breast Imaging and the Symptomatic Patient: Enough with the "Diagnostic" Mammography Am. J. Roentgenol., November 1, 2003; 181(5): 1423 - 1424. [Full Text] [PDF]
|
|
|
|
|
|
F. M. Hall, N. Houssami, L. Irwig, and S. Blome Mammography and Sonography in Young Symptomatic Women Am. J. Roentgenol., November 1, 2003; 181(5): 1424 - 1425. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
