jIIm/DIAGNDSTIC NUCLEAR MEDICINE

Scintigraphic Criteria for Hepatic Metastases from Cancer of the Colon and Breast

David E. Drum and Judith M. Beard

Harvard Medical Sc/zoo! and the Joint Program in Nuclear Medicine of Peter Bent Brigham Hospital, Children's Hospital Medical Center, and Sidney Farber Cancer Center, Boston, Massachusetts

Scintigraphic criteria Jor hepatic metastases were studied by examination of 333 scintigrams performed on 275 patients with primary cancers of the colon or breast. Focal defects in radiocolloid distribution correctly signaled the presence of metastatic colon carcinoma in 88% of the patients with that disease and incorrectly pointed to only 6% of the patients with out such metastases. In contrast, the same criterion detected only 67% of hepatic metastases from breast carcinoma. This lower sensitivity could be improved to 87% by adding heterogeneity or to the criteria for abnormality when patients with breast cancer are examined. Scinti graphic indicators of metastatic disease may vary according to the site of primary cancer. JNuclMed 17: 677-480, 1976

Scintigraphy of the liver plays a major role both aminations of I 67 patients with primary adenocar in detecting hepatic metastases and in evaluating cinoma of the colon or and from 125 exam their response to cancer . Despite evi inations of I08 patients with primary carcinoma of dence that the features of biologic growth of various the breast formed the basis for the analyses. No pa primary tumors differ widely ( 1) , no effort has been tient had more than one primary tumor. All repeat made to discern whether such variations are reflected examinations were made more than I month apart. in the scintigraphic properties of their hepatic me Each liver was examined either by closed needle tastases. , surgical observation, surgical biopsy, or post Recently, hepatic scintigraphy was reported to be mortem examination within 4 weeks of scintigraphy. 77% accurate in the detection of tumor metastases Any liver proven by these methods to contain meta when the primary disease was breast carcinoma and static deposits was assumed to have contained them 85 % accurate in the detection of metastatic colon at the time of scintigraphy. Further, any liver proven cancer (2). These data raise the question of whether free of metastases at postmortem examination was the growth features of metastases from these two considered to have been free of metastases at the common primary sources are sufficiently different to time of imaging. generate intrinsically different scintigraphic images, Hepatic scintigraphy. Hepatic scintiscans were be with the implication that the criteria for abnormality gun 10—30 mm after intravenous injection of 3—5 may depend upon the nature of the primary tumor. mCi of 9omTcsulfur colloid. An Anger scintillation In order to investigate this hypothesis, we studied the camera with 19 photomultiplier tubes, a 25-cm-diam medical charts of 275 patients whose had been directly examined shortly after hepatic scintigraphy. Received Dec. 3, 1975; revision accepted Feb. 23, 1976. MATERIALS AND METHODS For reprints contact: David E. Drum, Dept. of Radiology, Harvard Medical School, 25 Shattuck St., Boston, MA Data base.The scintigraphic findings from 208 cx 02115.

Volume I 7, Number 8 677 DRUM ANDBEARD

crystal, and a I40-keV diverging collimator was em scintigraphic criteria resulted in a similar true-posi ployed in all cases. Anterior (650,000 counts), p05- tive ratio (0.89) , but the false-positive ratio was tenor (650,000 counts), and right lateral (540,000 considerably larger (0.27). Thus, either strict or counts) views were obtained for every patient. For liberal criteria detected metastatic colon carcinoma all examinations two point sources, 10 cm apart, with equal sensitivity. The liberal criteria were less were photographed at the collimator face to aid esti specific, however, and would have indicated incor mation of organ size. rectly the presence of metastases in some livers when The scintigraphic images of the liver were evalu they were in fact absent. The general accuracy of ated for the following features: scintigraphy, expressed as the ratio of correctly iden Focal defect(s): One or more clearly delimited tified normal and abnormal livers to the total num areas of absent or minimal radiocolloid ac ber examined, was 9 1% with the strict criterion and cumulation surrounded by areas of normal 79% with the liberal criteria. radiocolloid uptake. Carcinoma of the breast exhibited distinctly dif Heterogeneity: Irregular distribution of radio ferent accuracy parameters (Table 2) . The strict colloid within the liver. criterion for scintigraphic abnormality led to a true Hepatomegaly: Any liver image with a vertical positive ratio of only 0.67 and a false-positive ratio @ height greater than 17 cm and a horizontal of 0.09. In contrast, the liberal criteria generated a width greater than 18 cm (3). true-positive ratio of 0.87 and a false-positive ratio In order to judge which scintigraphic features most of 0.29. Thus, for breast carcinoma the liberal cri accurately reflected the presence or absence of he teria were more sensitive to the presence of hepatic patic metastases, two kinds of criteria were em metastases than were focal defects alone. This en ployed: hanced detection was accomplished at the expense Strict: The scintigram was considered positive for metastases only when one or more focal defects could be seen (2—8). Liberal: The scintigram was considered positive TABLE 1. PARAMETERSFOR EVALUATION for metastases when focal defects, heteroge OF THE LIVER SCINTIGRAM OF PATIENTS neity, hepatomegaly, or any combination of WITH COLON CARCINOMA* these was apparent. Critera formetostasesFocal

Data analyses. The scintigraphic data were ana defects,heterogeneity, lyzed for accuracy, sensitivity, and specificity accord orFocal ing to the following definitions: onlyhepatomegalyAnalysisdefects (strictcrterion)(liberal crteria) True-positive ratio: The fraction of livers that generated abnormal scintigraphic images and True.positive ratio 0.88 0.88 False.positiveratio 0.06 0.27 were proven to have metastatic disease (a False.negative ratio 0.12 0.11 measure of sensitivity). True.negative ratio 0.94 0.73 False-positive ratio: The fraction of livers with *Atotalof208liverswere imaged: of these, 88 were out metastatic disease that generated abnor found to have metastases. mal scintigraphic images. True-negative ratio: The fraction of livers with out metastatic disease that generated normal scintigraphic images (a measure of speci ficity). TABLE 2. PARAMETERSFOR EVALUATION OF THE LIVER SCINTIGRAM OF PATIENTS False-negative ratio: The fraction of livers that WITH BREASTCARCINOMA* generated normal scintigraphic images in spite of proven metastatic disease. Criteria formetastasesFocal Each ratio was calculated twice : once with the strict defects,heterogenety, criterion for scintigraphic evidence of metastases and or- onlyhepotomegalyAnalysis Focal defects once with the liberal criteria. (strictcriterion)(liberal criteria)

True-positive ratio 0.67 0.87 RESULTS False.positive ratio 0.09 0.29 False.negative ratio 0.33 0.13 Among the 208 scintigrams of patients with pri True.negative ratio 0.91 0.71 mary carcinoma of the colon or rectum, the strict S A total of 1 25 livers were imaged: of these, 60 were criterion yielded a true-positive ratio of 0.88 and a found to have metastases. false-positive ratio of 0.06 (Table I ). The liberal

678 JOURNAL OF NUCLEAR MEDICINE DIAGNOSTIC NUCLEAR MEDICINE of a considerable increase in false-positive interpre (9). In particular, when the primary tations. Consequently, the general accuracy was simi tumor was carcinoma of the colon, the overall accu lar for scintigrams interpreted by either strict or racy of the liver scan was 90% . This figure is unsur liberal criteria (Table 2) : 80% and 79%, respec passed by any other noninvasive technique. Meta tively. static breast carcinoma appeared to cause focal Livers without metastases generated similar scm.. defects in this series less frequently than colon car tigraphic features regardless of the nature of the cinoma. Hence, its detection accuracy was lower primary tumor. Judged by the strict criterion, the (i.e., 80%). false-positive ratio was 0.06 for patients with colon When the strict criterion of focal defects was em primaries and 0.09 for patients with breast primaries. ployed as an indication of metastases, the false According to the liberal criteria, the corresponding positive ratios were identical for patients with colon ratios were 0.27 for colon carcinoma and 0.29 for or breast primaries. Livers not involved with meta breast carcinoma. Among the patients with colon static deposits were so defined by the liver scan primaries whose scintigrams were judged by the strict independently of the nature of the primary cancer criterion, three of seven false-positive interpretations and of the regimen used in managing these patients. arose from fatty infiltration, two from portal inflam Application of more liberal scintigraphic criteria matory infiltrates, one from macroscopic cirrhosis, for detecting metastatic colon carcinoma diminished and one from severe systemic and abdominal sepsis. overall accuracy by increasing the false-positive ra Three of these images exhibited significant extra tio. Thus, heterogeneity or hepatomegaly are seldom hepatic radiocolloid localization. Two of the seven unique manifestations of metastatic colon carcinoma, false-positive images arose from single focal defects a deduction consistent with the absence of these fea in the portal and left lobe areas, respectively. tures from our false-negative patterns. From Table I Six instances of focal hepatic radiocolloid defects the following may be deduced in regard to colon were observed in the absence of metastatic cancer cancer: If there are focal scintigraphic defects, there among patients with breast primaries. Two of these is a 92% chance that the patient has metastases; if livers were examined histologically and showed fatty there is only heterogeneity or hepatomegaly, there infiltration and abnormal hepatocyte regeneration. is only one chance in 26 that metastases are present. The three livers recorded as normal by surgical in According to either set of criteria, a negative scinti spection had single peripheral defects and increased gram means that the chances are 91 % that the liver splenic sequestration of radiocolloid. is free of metastases. For most of the total of 13 false-positive scans In contrast to the situation for colon carcinoma, (strict criterion) , plausible and potentially detect adoption of the more liberal criteria for metastatic able reasons for the misleading scintigram findings breast carcinoma had little effect on overall accu were evident. None were found to be due to benign racy. Both the true-positive and the false-positive space-occupying lesions or to impinging adjacent ratios increased, the former because certain instances disease. of metastatic breast carcinoma were manifest only as Among the 11 false-negative interpretations that heterogeneity or hepatomegaly. arose from application of the strict criterion to liver The probability that a patient has metastases from scintigrams of patients with colon primaries, none the colon when focal defects are observed is (0.88 X exhibited discernible focal defects and only one 88)/[(0.88x 88) + (0.06x 120)1= 0.92.Ap showed any hepatic image abnormality (hepato plying the liberal criteria, one additional case of megaly) . Only two of the livers with undetected metastases was detected, but the false-positive ratio colon metastases were so diagnosed histologically; increased from 0.06 to 0.27, indicating that 0.21 X the others were identified by surgical examination. I20 = 25 instances of heterogeneity or hepato Observation at laparotomy was also the basis for megaly did not represent metastases. establishing the presence of metastases in seven of For patients with breast primaries, a focal scinti the eight patients with carcinoma of the breast whose graphic defect gives an 87% probability of metastatic hepatic scintigrams were normal by the liberal cr1- cancer (Table 2) . However, if only heterogeneity or teria and in six of 12 cases with hepatomegaly or hepatomegaly are observed, the chance of hepatic heterogeneous and/or extrahepatic abnormalities of metastases is 48% , hardly a negligible figure. Thus, radiocolloid distribution. the liberal criteria would more frequently detect metastatic breast disease but at the price of a higher DISCUSSION false-positive ratio. In the context of modern breast These data support the confidence currently felt cancer management, this error would appear less in the capability of the liver scan to detect metastatic serious than false-negative reports. Moreover, closed

Volume 17, Number 8 679 DRUM ANDBEARD needle liver biopsy or laparoscopy with directed liver ACKNOWLEDGMENTS biopsy should identify the nonmalignant parenchy The author is indebted to S. J. Adeistein and B. J. McNeil mal diseases that cause hepatic enlargement and for helpful discussions. The quality of scintigrams permit heterogeneous radiocolloid distribution in the ab ting this study is a tribute to the technologists of the Joint sence of cancer (JO). Program in Nuclear Medicine. This study was supported Should the criteria for abnormality be deliberately by NIH Grant GM-18674. varied according to the nature of the primary tumor? In part, the answer depends upon the implications REFERENCES of false-positive and false-negative interpretations for 1. ANDERSONWAD, ed: Pathology, 6th ed. St. Louis, patient management. The requirement of focal de Mosby, 1971, Chaps. 16 and 28 fects or space-occupying lesions for the scintigraphic 2. LUNIA S, PARTHASARATHYKL, BAKSHI S, et al. : An evaluation of @mTc@sulfurcolloid liver scintiscans and their diagnosis of metastatic liver disease is the criterion usefulness in metastatic workup: A review of 1,424 studies. most commonly employed in published studies for INuciMed 16: 62—65,1975 any form of cancer (2—8). Although little is gained 3. MCAFEEJO, AUSERG, WAGNERHN: Diagnosticvalue by employing any but such strict criteria when the of scintillation scanning of the liver. Arc/i Intern Med I 16: 95—110,1965 primary tumor originates in the colon or rectum, 4. FEE HJ, PROKOP EK, CAMERON JL, et al. : Liver scan further studies may be required to establish the Va ning in patients with suspected abdominal tumor. JAMA lidity of this approach for the detection of all forms 230: 1675—1677,1974 and stages of metastatic cancer. 5. POULOSE KP, REBA RC, CAMERON JL, et al. : The The data also raise an important technical ques value and limitations of liver scanning for the detection of tion : for small metastatic deposits, is the in vivo hepatic metastases in patients with cancer. I Indian Med Assoc61: 199—202,1973 resolution of the diverging collimator (used with a 6. GENNAROAR, BACONHE : Re-evaluation of the merit I 9-photomultiplier scintillation camera at the infor of the liver scan in the management of patients with cancer mation densities employed here) inferior to that of of the colon and rectum. Dis Colon Rectum 14: 43—45, alternative methods of imaging? A definitive answer 1971 to this question would be important. 7. COVINGTONEE: The accuracy of liver photoscans. A m I Roentgenol Radium Ther Nuc! Med 109 : 742—744, Direct examination of the livers in this series re 1970 sulted in considerable dependence upon correct diag 8. DRUM DE, CHRISTACOPOULOSJS : Hepatic scintigraphy nosis by the surgeon, a potentially fallible measure in clinical decision making. I Nuc! Med 13: 908—915,1972 of histologic changes (1 1 ) . Although the limitations 9. SCuFF L: The liver: An overview. Semin Roentgenol 10:175—176,1975 of both inspection and needle biopsy (12) are well 10. SUGARBAKERPH, WILsoN RE: Using celioscopy to known, these means of diagnosing hepatic metastases determine stages of intra-abdominal malignant neoplasms. nevertheless play major roles in determining cancer ArchSurg lii: 41—44,1976 management throughout the world. In view of the ii. OZARDA A, PICKREN J: The topographic distribution disagreements between the scintigraphic and gross of liver metastases : Its relation to surgical and isotopic diag surgical diagnoses of liver disease, operating surgeons nosis.JNuc!Med3: 149—152,1962 12. CONN HO, YESNERR: A re-evaluation of needle should be encouraged to employ biopsy methods biopsy in the diagnosis of metastatic cancer of the liver. that afford maximal histologic accuracy. AnnlnternMed59: 53—61,1963

680 JOURNAL OF NUCLEAR MEDICINE