| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Gastrointestinal Imaging |
1 From the Department of Radiology, University of California San Francisco, Box 0628, M-372, 505 Parnassus Ave, San Francisco, CA 94143-0628. Received September 11, 2003; revision requested November 24; final revision received March 18, 2004; accepted April 8. Address correspondence to F.V.C (e-mail: fergus.coakley@radiology.ucsf.edu).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
MATERIALS AND METHODS: This retrospective study was approved by the committee on human research. Written informed consent was not required. The authors retrospectively identified 153 patients with breast cancer who underwent serial abdominal CT and who did not have definite liver metastases present at initial CT. The mean age was 56 years (age range, 27–93 years). Two readers independently recorded the presence, size, and number of small (15 mm or less in diameter) hypoattenuating hepatic lesions at initial CT. Another reader independently recorded the presence or absence of definite hepatic metastases at final CT. The association between the presence, size, and number of small hypoattenuating hepatic lesions at initial CT and the subsequent development of metastases was analyzed by using Kaplan-Meier analysis.
RESULTS: One or more small hypoattenuating hepatic lesions were seen at initial CT in 54 of 153 patients (35%). After a median follow-up of 584 days (range, 16–1827 days), definite hepatic metastases developed in 43 of 153 patients (28%), including 15 of 54 patients (28%) with hypoattenuating lesions at initial CT and 28 of 99 patients (28%) without hypoattenuating lesions at initial CT. Findings from the Kaplan-Meier analysis showed no association between the presence (P = .56), size (P = .55), or number (P = .30) of small hypoattenuating hepatic lesions at initial CT and the subsequent development of hepatic metastases.
CONCLUSION: In patients with breast cancer who do not have definite hepatic metastases at initial examination, there is no evidence that small hypoattenuating hepatic lesions seen at initial CT contribute to an increased risk of subsequently developing hepatic metastases.
© RSNA, 2004
Index terms: Breast neoplasms, metastases, 00.32 • Liver neoplasms, CT, 761.1211 • Liver neoplasms, secondary, 761.332
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
The first inclusion criterion was that the reports include histopathologic findings to confirm the diagnosis of breast cancer. The second inclusion criterion was that patients must have undergone two or more contrast-enhanced abdominal CT examinations at our institution between 1997 and 2002. (Studies performed before 1997 were not included because our departmental picture archiving and communication system was installed during 1996.) For our study, the first CT examination during this period was considered the initial examination, and the most recent CT examination was considered the final examination. The third criterion was that there were no definite hepatic metastases present at initial CT. A definite hepatic metastasis was defined as a hypoattenuating or heterogeneous mass that was located in the liver and measured more than 15 mm in diameter, with no features suggestive of a cyst, hemangioma, focal fatty infiltration, or abscess. A cyst was defined as a well-circumscribed lesion with an attenuation value of –20 to +20 HU. A hemangioma was defined as a hypoattenuating lesion with peripheral nodular globular enhancement and centripetal fill-in. Focal fatty infiltration was defined as a non-masslike or geographic area of mildly reduced attenuation that was located adjacent to the porta hepatis, gallbladder fossa, or fissure for the ligamentum teres hepatis. An abscess was defined as a fluid-filled lesion with a thick enhancing wall.
One hundred fifty-three patients met these inclusion criteria. The study population consisted of 153 women with a mean age of 56 years (age range, 27–93 years). All relevant histopathologic and medical records were reviewed for these patients. Histologic findings, tumor stage, and the administration of chemotherapy during the period between the initial and final CT examinations were noted.
CT Technique and Image Interpretation
The initial and final CT examinations of the 153 patients in the study group (total of 306 examinations) were performed by using single–detector row (HiSpeed; General Electric, Milwaukee, Wis) (194 of 306 examinations) or four–detector row (Lightspeed; General Electric) (112 of 306 examinations) helical CT scanners. All patients received 150 mL of intravenous iohexol (Omnipaque 350; Nycomed Amersham, Princeton, NJ) and 800 mL of oral diatrizoate meglumine (Hypaque; Nycomed Amersham, Princeton, NJ). Portal venous images were acquired after a scanning delay of 70 seconds. A section collimation of 7 mm was used on the single–detector row scanner, and a section collimation of 5 mm was used on the multi–detector row scanner. All images were contiguous.
Two attending radiologists (F.V.C., A.Q.), who had 7 and 5 years of subspecialty experience in abdominal imaging, respectively, independently reviewed images obtained from the initial CT examinations of all 153 patients by using a picture archiving and communication system workstation (Impax; Agfa, Mortsel, Belgium). Readers were aware that the patients had breast cancer and that the patients were considered not to have metastases at initial CT; readers were unaware of all other clinical and radiologic information. Readers recorded the presence or absence of hypoattenuating (visually of lower attenuation than the adjacent hepatic parenchyma on portal venous phase images) hepatic lesions of 15 mm or less in maximum axial diameter. Readers were instructed to exclude areas of hypoattenuation that they considered to be focal fatty infiltration, which was defined as a non-masslike or geographic area of mildly reduced attenuation that was located adjacent to the porta hepatis, gallbladder fossa, or fissure for the ligamentum teres hepatis. If hypoattenuating hepatic lesions were present, readers classified the number of lesions into one of three size categories: 0–5 mm, 6–10 mm, and 11–15 mm in diameter. Our threshold of 15 mm was chosen in accordance with prior studies of small hypoattenuating lesions (1,6).
A third radiologist (G.A.K.), who was undertaking a fellowship in abdominal imaging, independently reviewed images from all the final CT examinations and recorded the presence or absence of definite hepatic metastases. This determination of hepatic masses was made by a third radiologist to avoid reader bias in the interpretation of the images from the initial CT examination; our concern was that readers might be more likely to record small hypoattenuating lesions at initial CT if they knew that images from the final CT examination showed metastases and vice versa. This third radiologist also reviewed images from the initial CT examination, as necessary, to assist in the evaluation of images from the final CT examination. A definite hepatic metastasis was defined as a hypoattenuating or heterogeneous mass in the liver measuring more than 15 mm in diameter, with no features to suggest a cyst, hemangioma, focal fatty infiltration, focal nodular hyperplasia, abscess, or adenoma. All radiologists reviewed images with our standard departmental CT window settings for the abdomen (window width, 400 HU; window level, 40 HU) and liver (window width, 150 HU; window level, 88 HU).
Statistical Analysis
Calculation of descriptive statistics and outcome analysis were performed by using data that were obtained during the identification of small hypoattenuating hepatic lesions at initial CT by the senior reader (F.V.C.). To investigate interobserver agreement in the identification of small hypoattenuating hepatic lesions, data from both readers (F.V.C., A.Q.) were used to calculate 
values with respect to the presence, size, and number of small hypoattenuating hepatic lesions. The degree of observer agreement, as indicated by values, was interpreted as follows: 0–0.20, slight agreement; 0.21–0.40, fair agreement; 0.41–0.60, moderate agreement; 0.61–0.80, substantial agreement; and 0.81–1, almost perfect agreement (7).
Survival and/or failure time analysis was used to examine the association between the presence, size, or number of small hypoattenuating hepatic lesions that were identified at initial CT, as well as to monitor the subsequent development of definite hepatic metastases. The Kaplan-Meier product-limit method (rather than the life-table method) was used to analyze "failure" (ie, development of metastases) data. The product-limit method calculates the survival function directly from the continuous failure times; this method is more robust because the resulting estimates do not depend on the grouping of data into a certain number of time intervals. If, at the time of follow-up, there are no metastases, then these results represent censored observations because metastases may have developed after this time. All survival analysis methods, including the life-table, survival-distribution, and Kaplan-Meier methods, allow for such censored data. Censoring allows data from all patients to contribute to the analysis, but only to the time of follow-up; a patient with a short follow-up interval will therefore contribute little to the overall outcome analysis and vice versa. Of note, only 11 patients in our study had a follow-up interval of less than 3 months. The Student t test was used to compare the mean follow-up interval of patients who developed metastases with the mean follow-up interval of patients who did not develop metastases.
The relationship between the presence, size, and number of small hypoattenuating lesions and the outcome was examined. For analysis of lesion size, patients were divided into three groups that were composed of patients with no lesions (n = 99), patients with lesions of 0–10 mm only (n = 35), and patients with one or more lesions of 11–15 mm (n = 19). These thresholds were chosen because prior researchers have variably defined the upper size limit of small hypoattenuating lesions as 10 or 15 mm (1,5). For analysis of lesion number, patients were divided into three groups that were composed of patients with no lesions (n = 99), patients with one to three lesions (n = 39), and patients with four or more lesions (n = 15). The threshold of four or more lesions was chosen on the basis of the authors’ experience of what would be reported as "multiple" in routine clinical practice. The log-rank test was used to compare the groups. Statistical calculations were performed by using a statistical software package (SigmaStat 3.0 and SigmaPlot 8.02; SPSS, Chicago, Ill).
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
values between the readers for the identification of small hypoattenuating hepatic lesions with respect to presence ( = 0.86), size ( = 0.72), and number ( = 0.80).
|
|
|
|
|
|
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
First, the percentage of hepatic lesions that are metastases may be so low that the relative risk of a patient developing metastases is only slightly elevated. In a study of patients with gastric and colorectal cancer, six of 226 patients (2.7%) with small hypoattenuating hepatic lesions, which were the only hepatic abnormality at CT, were subsequently found to have hepatic metastases; findings were confirmed with serial imaging or biopsy (6). The negative results of the Kaplan-Meier analyses in our study do not exclude the possibility of a small increased risk.
Second, even with thin collimation, small metastases may be less detectable than small cysts at CT because small metastases may be isoattenuating. This concept is supported by findings from a study of 31 patients with hepatic metastases, including 30 patients with colorectal carcinoma and one patient with gastrointestinal stromal tumor. Preoperative CT showed 88 small hypoattenuating hepatic lesions (2). Of these lesions, 25 were metastases. Thinner collimation (2.5 vs 5 mm) improved overall lesion detection but did not increase detection of these small metastases. If small metastases were isoattenuating, then thinner collimation would not improve lesion detection. Partial volume averaging would also tend to obscure small metastases that have an attenuation that is only slightly different from the attenuation of the adjacent hepatic parenchyma. Small cysts are more likely to be seen because of the greater attenuation difference between fluid and hepatic parenchyma. An analogy can be made with the traditional teaching that a small nodule that is visible on a chest radiograph is probably calcified and benign because a malignant soft-tissue nodule of the same size would be invisible (8). The fact that a small hepatic lesion is visible at CT may inherently suggest the lesion is cystic.
Third, our finding that 28 of 99 patients without small hypoattenuating lesions at baseline CT subsequently developed hepatic metastases suggests that these patients are at high risk, irrespective of the finding of one or more small hypoattenuating lesions at baseline CT. While these factors may explain our findings, we are not suggesting that small hypoattenuating hepatic lesions, particularly lesions in the borderline size range of 10–15 mm, can always be ignored. All radiologists who interpret abdominal CT scans will have seen lesions of this size that can be characterized as cystic, as well as irregular or heterogeneous lesions that are suggestive of metastases. Other factors, such as tumor grade and stage, may influence the interpretation of such lesions. A larger study is required to establish whether borderline hepatic lesions of 10–15 mm that are visually suspicious and that manifest in patients with advanced or aggressive tumors might be early metastases. In practice, such findings may prompt further evaluation with ultrasonography (US), magnetic resonance imaging, or even biopsy, especially if the demonstration of hepatic metastases is linked to the choice of therapy. Nonetheless, in the absence of such ancillary features, we believe it might be reasonable to omit a description of small hypoattenuating lesions from the body of the radiologic report, and particularly to omit phrases such as "small hypoattenuating hepatic lesions, metastases cannot be excluded" from the impression. Our findings can also be used to reassure patients and referring physicians that isolated small hypoattenuating hepatic lesions do not appear to be associated with an increased risk of subsequent development of hepatic metastases.
It is interesting to compare our results with those of previous similar studies. In our study, small hypoattenuating hepatic lesions that were the only abnormality in the liver at baseline CT were found in 35% of patients with breast cancer. Jones et al (1) found hypoattenuating hepatic lesions of 15 mm or less in diameter in 254 of 1454 outpatients (17%) who underwent CT, including 1192 patients with a known extrahepatic primary malignancy. Schwartz et al (5) found hypoattenuating hepatic lesions of 10 mm or less in 63 of 369 patients (17.1%) with breast cancer. The higher frequency of small hypoattenuating hepatic lesions in our study may be attributed to patient selection bias, the lower size threshold used by Schwartz et al, or to technical differences. Both of the prior studies used a collimation of 10 mm, and 17.2% of patients in the series by Schwartz et al did not receive intravenous contrast material. In a more recent study that used 5- and 7-mm collimation, researchers found small hypoattenuating lesions of 15 mm or less in diameter in 268 of 1133 patients (23.6%) with gastric or colorectal cancer (6).
Even though findings from several prior studies have suggested that such lesions are often metastases, the findings from our study did not demonstrate an increased risk of developing metastases in patients with small hypoattenuating hepatic lesions. Jones et al (1) reported that 55 of 209 patients (26%) with small hepatic lesions and a history of extrahepatic malignancy had hepatic metastases. Schwartz et al (5) reported that 14 of 63 patients (22%) with breast cancer and small hypoattenuating hepatic lesions later developed hepatic metastases. In a study that used US for characterization, Eberhardt et al (9) found that 18 of 124 (14.5%) indeterminate hypoattenuating lesions that were seen at CT were metastases. The discrepancy between these studies and our results may be explained by three factors.
First, as previously noted, even a sizable minority of small hypoattenuating lesions that are metastases may not be sufficient to cause a detectable increase in the relative risk of subsequently developing metastases when analyzed on a per-patient basis.
Second, these prior studies have generally lacked a control group and did not compare the risk of metastases developing in patients who had small hypoattenuating lesions with the risk of metastases developing in patients who did not have such lesions. The development of metastases in patients with small hypoattenuating lesions is not proof that the small lesions were metastases, and this finding could be merely coincidental. Even development of a metastasis at the site of a small hypoattenuating lesion is still is not conclusive proof that the small lesion was malignant, because a metastasis might have arisen next to and obliterated an adjacent small cyst. Such considerations suggest that the inclusion of a control group is a crucial component in the study of small hypoattenuating lesions.
Third, the histopathologic basis of small hypoattenuating lesions may be substantially different between patients who have no other CT finding in the liver and patients who have CT findings of hepatic metastases in addition to hypoattenuating lesions. In some studies, both groups were pooled together. The probability that a small lesion is a metastasis may be much higher in a patient with large, definite metastases than in a patient without large, definite metastases.
Our study has a number of limitations. First, we included only patients with breast cancer because findings from a prior study had suggested that these are the patients in whom small hypoattenuating hepatic lesions are most likely to be malignant (5). As a result, our findings cannot be directly extrapolated to the interpretation of small hypoattenuating lesions in patients with other primary malignancies.
Second, we performed our analysis on a per-patient, rather than a per-lesion, basis. It is possible that a per-lesion analysis might have demonstrated lesion progression in some small hypoattenuating lesions. As noted above, however, this would not constitute irrefutable proof that the original lesion was malignant, nor would it have addressed the clinically relevant question of the overall relative risk in a given patient.
Third, our study lacked any histopathologic standard of reference for either the nature of small hypoattenuating lesions or the diagnosis of hepatic metastases at follow-up CT. Also, the determination of metastases was made by a single radiologist who was undertaking a fellowship in abdominal imaging. That said, however, the diagnosis of hepatic metastases by using CT is widely accepted in oncologic practice and is generally straightforward. Also, it would have been unrealistic to demand histopathologic proof in these patients.
Fourth, there is undoubtedly a selection bias in our study design in that patients with breast cancer who are at high risk of metastatic spread are more likely to undergo serial CT examinations than are patients with limited early-stage disease. This bias may have inflated the number of patients who developed hepatic metastases in our study population when compared with the number of patients who develop hepatic metastases in the general population of patients with breast cancer. This bias, however, should have operated equally within the groups with and the groups without small hypoattenuating hepatic lesions at baseline CT, given that the characteristics of these subpopulations were similar (Table).
The fifth limitation of our study is that we used the results of only a single reader in the outcome analysis. The high values for interobserver agreement between the two readers, however, suggests that the evaluation of small hypoattenuating lesions is relatively objective and straightforward. Therefore, the inclusion of repeated, highly dependent data from multiple readers in the outcome analysis would have been unhelpful and would have resulted in the same conclusions.
Sixth, we included patients who underwent serial CT examinations irrespective of the interval between the two studies. As a result, the period between examinations was short and presumably insufficient in some patients to demonstrate the progression of metastases. However, data censoring by Kaplan-Meier analysis adjusts for this factor as a potential confounding variable. In addition, the mean duration of follow-up was more than 1 year, and this was not statistically different (P = .16) between the groups without small hypoattenuating lesions (mean follow-up interval, 710 days) and the groups with small hypoattenuating lesions (mean follow-up interval, 598 days).
A seventh limitation of our study is that we did not include a control group of patients without malignancy. Therefore, we cannot compare the percentage of breast cancer patients with small hypoattenuating lesions at baseline CT in our study (35%) with the percentage of patients without malignancy who have such lesions. If these small hypoattenuating lesions are truly incidental cysts, one would expect these percentages to be similar. However, while this information would be of academic interest, it would not address the clinically important question of the likelihood of subsequently developing metastases.
Finally, since it is possible that the word breast might not have appeared anywhere in the CT report of some patients, our search process may not have identified all eligible patients with breast cancer.
In conclusion, in patients with breast cancer who do not have definite hepatic metastases at presentation, there is no evidence that small hypoattenuating hepatic lesions seen at initial CT contribute to an increased risk of subsequently developing hepatic metastases.
|
FOOTNOTES |
|---|
Author contributions: Guarantor of integrity of entire study, G.A.K.; study concepts, F.V.C.; study design, G.A.K., F.V.C., A.Q.; literature research, G.A.K.; clinical studies, F.V.C., G.A.K.; data acquisition, G.A.K., F.V.C., A.Q.; data analysis/interpretation, F.V.C., G.W.; statistical analysis, G.W.; manuscript preparation, definition of intellectual content, editing, revision/review, and final version approval, all authors
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
]) and patients without hypoattenuating lesions (n = 99; 71 censored, 28 developing metastases [
]). (b) Probability of remaining free of hepatic metastases based on the size of hypoattenuating lesions at CT. Kaplan-Meier analysis showed no significant difference (P = .55) between patients with no hypoattenuating hepatic lesions (n = 99; 71 censored, 28 developing metastases [
]). (c) Probability of remaining free of hepatic metastases based on the number of hypoattenuating lesions at CT. Kaplan-Meier analysis showed no significant difference (P = .30) between patients with no hepatic lesions (n = 99; 71 censored, 28 developing metastases [
