| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Commentary |
1 Department of Radiology, Magnetic Resonance Science Center, University of California, San Francisco, AC-109, 1 Irving St., San Francisco, CA 94143-1290.
Received April 18, 2003;
accepted after revision April 18, 2003.
Address correspondence to N. M. Hylton.
The performance of MRI of the breast will depend on the clinical question and will also be influenced by characteristics of the study population. The comparison of diagnostic performance between MRI and conventional breast imaging methods, or between different breast MRI methods, is complicated by differences in study population, both in terms of cancer prevalence and disease distribution.
The article by Knopp et al. [1] describes the findings of a phase 2 double-blind randomized trial to evaluate the clinical efficacy and dose response of gadobenate dimeglumine for contrast-enhanced MRI of the breast. The study involved 14 centers performing MRI of the breast, with image interpretation conducted independently by two outside interpreters. Three doses of gadobenate dimeglumine were compared with a single dose of gadopentetate dimeglumine, a contrast agent widely used for MRI of the breast. The study was well designed and used a protocol for image acquisition and contrast administration that was reasonably standardized among the sites and could be easily adopted in clinical practice. One hundred eighty-nine women were enrolled in the study and randomly assigned to one of four treatment groups receiving either gadobenate dimeglumine at a dose of 0.05, 0.1, or 0.2 mmol/kg body weight, or gadopentetate dimeglumine at a dose of 0.1 mmol/kg body weight. All women had had a mammographic examination within 30 days before their MRI examination that revealed an abnormality that was highly suspicious for breast cancer and warranted biopsy.
The image acquisition protocol specified imaging in the coronal plane using a three-dimensional, T1-weighted spoiled gradient-echo technique with acquisitions timed at 0, 2, 4, 6, and 8 min after bolus injection of contrast agent. No fat suppression was used. Other specifications constrained the TR, flip angle, slice thickness, field of view, in-plane resolution, and total scanning time, but allowed some flexibility among the participating centers. Ten of the 14 sites used 1.5-T MRI systems, three sites used 1.0-T systems, and one site used a 0.5-T system. Image interpretation was performed independently by two interpreters, neither of whom participated in data collection or was affiliated with any of the centers. Interpreters were blinded to all patient data and to the dose and identity of the contrast agent.
A careful analysis is presented comparing a number of performance measures in the treatment groups. A lesion detection scoring system was used to assign a 0, 1, or 2 to lesions that the observers designated as uncertain, possibly or probably present, or definitely present, respectively. For the analysis of lesion detection, on-site investigators completed "final diagnosis reference maps" of the breast showing the location of all lesions diagnosed on the basis of mammography, sonography, fine-needle aspiration, biopsy, surgery, or pathology. The final diagnosis reference map was used as the gold standard for lesion detection. A score of -1 was retrospectively assigned to lesions on the final diagnosis reference map that were not detected on MRI. Diagnostic accuracy was evaluated by assessing the sensitivity and specificity for both lesion detection and malignant classification, based on histopathologic results. The analyses were made on both per lesion and per breast bases. The authors also considered the influence of image interpretation method. In the first interpretation session, unenhanced images and enhanced images were randomized and reviewed independently. In the second interpretation session, a combined data set consisting of unenhanced, enhanced, subtracted, and maximum-intensity-projection images, as well as region-of-interest measurements of signal intensity, was used to assess lesion presence and other characteristics. Two observers participated in each interpreter study, and their performances were compared.
The many layers of comparison allow a complete evaluation of the data and include many of the variables that can influence performance. The comprehensive analysis also shows the difficulty that arises in identifying the most appropriate measure for comparison. The primary findings of this study were a dose-related increase in lesion detection score for patients receiving gadobenate dimeglumine and an optimal combination of sensitivity and specificity for the 0.1 mmol/kg gadobenate dimeglumine group. In the comparison between equal-dose groups of gadobenate dimeglumine and gadopentetate dimeglumine, higher detection scores and sensitivity values were found for gadobenate dimeglumine; however, lower specificity values were measured.
Despite the prospective and randomized design of the trial, the disease distribution among the four treatment arms of the trial was nonuniform. The significance of the findings of this study is therefore uncertain. The four populations were comparable in size and demographics; however, they differed in a number of aspects that are important to the evaluation of diagnostic accuracy. Of the four groups, the proportion of malignant lesions was lowest (62%) in the 0.1 mmol/kg gadobenate dimeglumine group. It was 81%, 75%, and 79% in the 0.05 mmol/kg gadobenate dimeglumine, 0.2 mmol/kg gadobenate dimeglumine, and 0.1 mmol/kg gadopentetate dimeglumine groups, respectively. The 0.1 mmol/kg gadobenate dimeglumine group also contained a disproportionately high number of benign fibroadenomas (15% in that group vs 4%, 4%, and 0% in the 0.05 mmol/kg gadobenate dimeglumine, 0.2 mmol/kg gadobenate dimeglumine, and 0.1 mmol/kg gadopentetate dimeglumine groups, respectively). Fibroadenomas are a common false-positive finding for MRI of the breast, and the relatively high number found in the 0.1 mmol/kg gadobenate dimeglumine group may have contributed to both the higher sensitivity and lower specificity found for this dose and formulation. On the final diagnosis reference maps, a significantly lower number of lesions were identified in the 0.1 mmol/kg gadobenate dimeglumine group compared with the number identified in the others (55 lesions vs 85, 79, and 82 lesions in the 0.05 mmol/kg gadobenate dimeglumine, 0.2 mmol/kg gadobenate dimeglumine, and 0.1 mmol/kg gadopentetate dimeglumine groups, respectively). The remaining three groups were more comparable in their disease distribution.
Although dose-related improvements in lesion detection are clear from the analysis, the comparison between equal-dose groups of gadobenate dimeglumine and gadopentetate dimeglumine is less conclusive. A higher baseline sensitivity value was measured in the 0.1 mmol/kg gadobenate dimeglumine group, that is, a greater percentage of lesions were identified on unenhanced images. This may have been a consequence of the higher proportion of fibroadenomas in the 0.1 mmol/kg gadobenate dimeglumine group (15% vs 0% in the 0.1 mmol/kg gadopentetate dimeglumine group), which are often visible on unenhanced images. The sensitivity for lesion detection on unenhanced images was in the range of 50%, higher than in any other group, as seen in table 2 [1]. Thus, while the 0.1 mmol/kg gadobenate dimeglumine group showed the highest overall sensitivity values for lesion detection, the increases from baseline were actually less than those measured in the 0.2 mmol/kg gadobenate dimeglumine group and comparable to those measured in the 0.1 mmol/kg gadopentetate dimeglumine group.
Differences in disease distribution may have also contributed to the lower false-positive rate for lesion detection found in the 0.1 mmol/kg gadobenate dimeglumine group, shown in figure 3 [1]. The 0.1 mmol/kg gadobenate dimeglumine group had considerably fewer lesions on final diagnosis and a significantly lower proportion of histopathologically confirmed malignancies than all other groups.
The quantitative data in figure 9 [1] show a smaller average signal intensity increase on enhanced images for malignant lesions in the 0.1 mmol/kg gadopentetate dimeglumine group than in either the 0.05 or the 0.1 mmol/kg gadobenate dimeglumine group. Gadobenate dimeglumine molecules can bind weakly and transiently to serum albumin resulting in in vivo T1 relaxivity that is roughly twice as high as that of gadopentetate dimeglumine. It is therefore expected that similar interactions in tumors with high albumin concentrations could lead to higher relaxivities and thus greater signal enhancement at similar doses. Alternatively, it may mean that lower doses of gadobenate dimeglumine can be used. This possibility appears to be supported by the average signal intensity measurements, which were determined for the malignant lesions only. Although the overall disease distribution in the four dose groups was not uniform, good uniformity was seen in the distribution of malignant subtypes listed in table 2 [1].
Other lesion characteristics, including lesion margins, conspicuity, degree, and morphology of enhancement, were assessed qualitatively, but were less informative. In general, the study had insufficient numbers to determine if differences in these other characteristics among treatment groups, between observers, or among histopathologic subtypes were of significance. A slightly higher overall incidence of adverse events occurred in the gadopentetate dimeglumine-treated group; however, the rate was lower if events unrelated to treatment were excluded.
The many degrees of freedom are both the power and limitation of MRI in general, and they present a major challenge to defining specifications for MRI of the breast. Yet in comparison to other disease applications for MRI, the requirement for standardization of MRI of the breast may be greater because of the existence of defined standards for mammography. This study was designed and executed well to evaluate the performance of gadobenate dimeglumine-enhanced MRI. The imaging protocol was appropriately standardized to produce robust and generalizable results. The inability to draw firm conclusions from this study is unfortunate and highlights the difficulties in controlling for the many factors that can influence breast MRI performance.
 Top References |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
