An Analysis of Imaging Studies and Liver Function Tests to Detect Hepatic Neoplasia

Transcription

1 Digestive Diseases and Sciences, Vol, 32, No. 10 (October 1987), pp II17 An Analysis of Imaging Studies and Liver Function Tests to Detect Hepatic Neoplasia THOMAS J. McGARRiTY, MD, TODD SAMUELS, MD, and FREDERICK A. WILSON, MD A retrospective study was conducted to determine the sensitivity, specificity, and accuracy of Scintigraphy, ultrasound, and CT scanning in conjunction with biochemical tests in the detection of liver neoplasia. Sixty-three patients with metastatic liver disease and 45 patients with nonmalignant liver disease received a total of 46 liver~spleen scans, 61 ultrasounds, and 49 CT scans. The sensitivities of liver~spleen scan, ultrasound, and CT scan were 72, 73, and 81%, respectively; the specificities were 86, 94, and 83%, respectively; and, the accuracies were 78, 84, and 82%, respectively. No statistically significant difference in specificity, sensitivity, or accuracy was seen between the three imaging studies. The serum alkaline phosphatase was significantiy more accurate than total bilirubin and SGOT in detecting liver metastasis. However, 19 and t3% of the malignant and nonmalignant groups, respectively, had normal biochemical tests. The sensitivity, specificity, and accuracy of the imaging studies in the presence of an abnormal alkaline phosphatase, SGOT, or total bilirubin were not significantly different than in the presence of normal biochemical tests. These results suggest that other factors such as cost, reproducibility, and availability of skilled interpreters should be considered in the selection of imaging studies for the detection of neoplastic liver disease. KEY WORDS: scintigraphy; ultrasound; CT scan; hepatic neoplasia. The determination of liver metastasis is a critical variable in the management of patients with neoplastic disease. Therapeutic decisions such as curative surgery for localized extrahepatic cancer, partial hepatectomy for Solitary hepatic metastasis, and the implementation of chemotherapy are based on the presence or absence of liver involvement as determined by liver-imaging techniques. The ques- Manuscript received August 8, i985; revised manuscript received May 12, 1986; accepted October 10, From the Department of Medicine, The Milton S. Hershey Medical Center, The Pennsylvania State University, Hershey, Pennsylvania Address for reprint requests: Dr. Thomas J. McGarrity, Division of Gastroenterology, The Milton S. Hershey Medical Center, PO Box 850, Hershey, Pennsylvania tion of which imaging modality (CT scanning, liver scintigraphy, or liver ultrasonography) is best has been the subject of controversy for many years (1-12). Furthermore, the presence or absence of an elevated alkaline phosphatase and other biochemical parameters have been counted as sensitive and specific indicators of metastatic disease (6-14). This study is a retrospective analysis to determine the accuracy of CT scanning, liver scintigraphy, or liver ultrasound in conjunction with liver biochemical tests in the detection of liver metastasis. MATERIALS AND METHODS Records were reviewed of consecutive patients who had a histologic specimen of the liver and at least one of Digestive Diseases and Sciences, Vol. 32, No. 10 (October 1987) /87/ / Plenum Publishing Corporation 1113

2 McGARRITY ET AL the three imaging studies. Histologic specimens were from needle biopsy, surgical wedge biopsy, or autopsy material. Histologic specimens were collected from January 1982 to May Patients were excluded if imaging studies were done more than four weeks before the histologic diagnosis was made. All imaging studies and histologic examinations were conducted, respectively, by members of the departments of radiology and pathology at the Milton S. Hershey Medical Center. The patients' biochemical liver tests Were reviewed, and the last values before the histologic diagnosis was made were recorded. The alkaline phosphatase (alk phos), serum glutamic oxaloacetic transaminase (SGOT), and total bilirubin (T bili) were available in all patients, and a value greater than 10% above the upper limits of normal was considered abnormal. After meeting selection criteria, the patients were divided into two groups according to the presence of malignant or nonmalignant histology, and their radiologic and biochemical data were atialyzed. Techniques. Liver/spleen scans were observed 15 min after an injection of 3 mci technetium-99m sulfur colloid (99mTc SC). Large-field-of-view cameras (Searle) were used to perform multiple-imaging studies of the liver and spleen. Images were collected on blue-base nuclear medicine film (NMB,I, Eastman Kodak Company), 8 i0 inches in size with six views on each side. The views included anterioj: supine with rib markers, right lateral, posterior, and left lateral positions. Organ/lesion size was gauged by placing on the abdomen a standard marker consisting of 2 x 2-cm lead squares, each square separated by 2 cm. Lesion edges were estimated and measured to the nearest centimeter in both horizontal and vertical axes. Ultrasound examinations were performed with the Roche B scanner or Picker B scanner employing 2.25, 3.5, 5 short, 5 medium, or 7.5 mhz transducers. An ATL real time scanner also was employed with 3 mhz, 5 long mhz,, or multiple-frequency transducers. The liver was examined in the prone, supine, and left lateral decubittls positions. CT scans of the liver were performed with the Siemens somatome body scanner with a 5-sec scanning time. The slice thickness of the scans was 8 mm performed at 8- or 16-mm intervals. The number of scans necessary to obtain visualization of the entire liver varied according to the size of the organ. Renographin 60% contrast material was given as a 50=cc bolus unless metastatic lesions were obvious on unenhanced scans. Data Analysis. For each imaging technique, the sensitivity was defined as the proportion of patients with malignant histology whose imaging study was read as showing evidence of metastatic disease, ie, true positives divided by the sum of the true positives and false negatives [TP/(Tp + FN)]. Specificity was defined as the proportion of patients with nonmalignant histology whose imaging studies showed no evidence of ne0plasia, ie, true negatives divided by true negatives and false positives [TN/(TN + FP)]. Accuracy was defined as the number of tests that provided correct classifications divided by the total number of tests that were scored [TP + TN/(TP + FN + TN + FP)]. Similar calculations were made from biochemical liver tests alone as well as in combination TABLE 1. PRIMARY SiTE OF HEPATIC METASTATIC TUMORS Primary site Number of patients Colon 16 Unknown 11 Pancreas 9 Lung 8 Breast 5 Melanoma 5 Esophagus 4 Endometfium 2 Kidney 2 Prostate 1 Total 63 with imaging studies. Data were analyzed according to the X 2 analysis, using the fourfold table (15). RESULTS Of 108 patients who met selection criteria for review, 63 patients had malignant histologic findings and 45 patients had nonmalignant findings. Table 1 lists the primary site of the malignant tumors studied. There were nine known primary sites of hepatic metastatic tumor with the colon being the most frequent site. The primary site was unknown in 11 patients. Nonmalignant histology is listed in Table 2. The results of a total of 46 liver/spleen scans, 61 ultrasounds, and 49 CT scans were obtained. Table 3 shows the sensitivity, specificity, and accuracy for each of the imaging modalities in detection of malignant hepatic lesions. The sensitivities of liver/ spleen Scan, ultrasound, and CT scan were 72, 73, and 81%, respectively. There was no statistically significant difference between the sensitivities of the three imaging tests (P > 0.05). The specificities of liver/spleen scan, ultrasound, and CT scan were 86, 94, and 83%, respectively. There was no statistically significant difference between the specificities of the three imaging modalities (P > 0.05). TABLE 2. NONMALIGNANT HISTOLOGY Hepatitis 9 Nonspecific inflammation 9 Cholestasis 5 Cirrhosis 4 Fibrosis 4 Other* 14 Total 45 *Includes steatosis (3), normal (2), granuloma (2), adenoma (1), abscess (1), extramedullary hematopoiesis (1), centfilobular necrosis (1), bile duct hamartoma (I), hemosiderosis (1), centrilobular congestion (1) Digestive Diseases and Sciences, Vol. 32, No. 10 (October 1987)

3 LIVER IMAGING, FUNCTION AND METASTASIS TABLE 3. SENSITIVITY, SPECIFICITY, AND ACCURACY OF IMAGING STUDIES IN DETECTION OF MALIGNANT HEPATIC LESIONS Sensitivity Specificity Accuracy (%) (%) (%) Liver/spleen scan Ultrasound CT scan Finally, there was no statistically significant difference (P > 0.05) between the three imaging techniques with regard to accuracy. The accuracies of the imaging techniques were liver/spleen, 78%; ultrasound, 84%; and CT scan, 82%. Table 4 compares the sensitivity, specificity, and accuracy of three biochemical tests used in the detection of malignant hepati c lesions. Both alk phos and SGOT were significantly more sensitive than T bili in detecting malignant hepatic disease (alk phos P < 0.001; SGOTP < 0.02). There was no statistically significant difference between the sensitivity of alk phos and SGOT (P > 0.05). The alk phos was more accurate than both T bili (P < 0.05) and SGOT (P < 0.05). There was no significant difference in the mean values of alk phos, SGOT, or T bili in the malignant vs nonmalignant histology group. The percentage of patients in the two histologic groups who had 0, 1, 2, or 3 elevated biochemical liver tests is shown in Table 5. Nineteen and thirteen percent of the malignant and nonmalignant groups, respectively, had normal biochemical liver tests. A greater percentage of patients with malignant histology (35%) had one abnormal biochemical liver test than patients with nonmalignant histology (24%), whereas the reverse was true with patients with three abnormal biochemical tests (malignant; 17%; nonmalignant, 30%). Table 6 compares the accuracy of imaging studies when done in the presence of 0, 1, 2, or 3 elevated biochemical liver tests. There was no statistically significant difference (P > 0.05) in the accuracy of TABLE 4. SENSITIVITY, SPECIFICITY, AND ACCURACY OF ELEVATED BIOCHEMICAL TESTS IN DETECTION OF "MALIGNANT LIVER LESIONS Sensitivity Specificity Accuracy (%) (%) (%) Alk phos SGOT T bill 3 l TABLE 5. PERCENTAGE OF PATIENTS WITH 0-3 ABNORMAL BIOCHEMICAL LIVER TESTS IN MALIGNANT AND NONMALIGNANT GROUPS Number of abnormal liver function tests Malignant histology (%) Nonmalignant histology (%) any of the imaging modalities done in the presence of normal or elevated biochemical liver tests. Figure 1 compares the sensitivity, specificity, and accuracy of the three imaging modalities when the studies were done in the presence of a normal or an elevated alk phos. The presence of an abnormal a!k phos did not change significantly the sensitivity, specificity, or accuracy of the imaging tests in the detection of malignant hepatic lesions. The sensitivity, specificity, and accuracy of the imaging studies also were not changed significantly by the presence or absence of an abnormal SGOT and T bili. DISCUSSION Questions regarding the use of imaging studies to detect liver metastasis have been the subject of much investigation over the years. This study found no significant difference in the sensitivity, specificity, and accuracy between CT scanning, ultrasound, or liver/spleen scan (Table 3). These results resemble those of Smith et al (2), who found in 80 patients no statistical difference in the sensitivity, specificity, or overall accuracy between any of the imaging modalities. Moreover, to determine the potential benefit of using two or more tests, Smith et al (2) performed a composite analysis of two or more tests. The overall accuracy, sensitivity, and specificity were not improved by combining the results of two or three imaging studies. Another question addressed in this study is whether imaging studies are useful in the detection TABLE 6, ACCURACY OF IMAGING STUDIES IN PRESENCE OF NORMAL OR ELEVATED BIOCHEMICAL LIVER TESTS Number of abnormal Accuracy (%) liver tests Liver/spleen Ultrasound CT scan ,2, Digestive Diseases and Sciences, VoL 32, No. 10 (October 1987) I 1 15

4 McGARR!TY ET AL Sensitivity ~ % 73% 67o/0 v///~ M 82% J IN L ALK PHOS ;~t ALK PHOS Specificity /0 88% 90% 100% 88% Liver/ Ultra- CT Spleen sound Scan Accuracy 75%-71% 97%-80% 69%-83% Fig 1. Sensitivity, specificity, and accuracy of imaging studies in association with normal and elevated serum alkaline phosphatase. of metastatic liver disease when biochemical liver tests are normal. Previous studies have concluded that imaging of the liver should not be employed when biochemical liver tests are within normal limits (6-14). Indeed, Mettler et al (14) concluded that if liver/spleen scanning had been restricted to those patients with abnormal serum enzymes, at least half the scans couldhave been eliminated with no loss of accuracy. However, in the present study, the accuracy of the imaging modalities was not diminished significantly when performed in the presence of normal biochemical liver tests (Table 6). The importance of this fact is illustrated further when it is considered that 12 patients (19%) with malignant histology had normal biochemical liver tests. Our findings also contrast with those of Castagna et al (13), who concluded that patients with normal alk phos have about a 90% assurance that their liver is free of tumor. Of our 63 patients with malignant histology, 31% had a normal alk phos, ie, the sensitivity is 69% (Table 4). Thus our findings suggest that the presence of a normal alk phos should not dissuade the diagnosis of metastatic liver disease. From the results of our study, one cannot easily recommend which imaging modality to employ first in the evaluation of metastatic liver disease. Factors other than sensitivity, specificity, and accuracy should be considered. Ultrasound is the least expensive of the imaging techniques and lacks ioniz- ing radiation. For example, at our institution ultrasound is one third the cost of a CT scan with enhancement. Bernardino et al (1) recommended the initial use of ultrasound over liver/spleen scanning. They based their conclusion on the increased sensitivity of real-time sonography and the ability to detect metastatic disease outside the liver and spleen. However, ultrasound examination can be limited by obe,sity and overlying intestinal gas. In Snow's study of 94 patients, sonographic examinations were ruled inadequate for interpretation due to intestinal gas in 23% (4). Other potential limitations of ultrasound include the quality of the equip, ment, the skill of the operator, and the experienc e of the interpreter. CT scans and liver scintigraphy are less affected by these parameters. In the study by Snow et al, all three imaging studies were done on 94 patients with suspected neoplastic liver disease. CT scanning was superior in sensitivity, specificity, and accuracy of diagnosis. However, their ultrasound examinations did not employ real-time scanners, which improves accuracy. Indeed, because of its lower cost, speed, and ease of performance, these authors recommend liver/spleen scans as the initial diagnostic choice. Clouse (5) and Elyaderani and Gabriele (6) also recommended the liver/spleen scan as the initial diagnostic test for the diagnosis of liver neoplasms because of its low cost, short exam, ination time, and reproducibility. Thus, factors such as cost and physical condition of the patient must be 1116 Digestive Diseases and Sciences, Vol. 32, No. 10 (October 1987)

5 LIVER IMAGING, FUNCTION AND METASTASIS weighed in the decision as to which imaging studies are to be used to detect hepatic metastases. Our findings support the following approach to patients suspected of metastatic liver disease. Although an elevated serum alk phos level is expected, a normal alk phos level should not prevent further evaluation. Depending on local expertise and availability, ultrasonography or scintigraphy should be the initial diagnostic test. Because of its greater cost and limited patient capacity, CT scan should be reserved for technically unsatisfactory ultrasounds or liver spleen scans. In our study, the combination of a normal alk phos and a negative ultrasound was 100% specific for the absence of malignant disease. Since this combination was present in only 11 patients, malignant disease might be found if a larger number of such patients were studied. In a patient with suspected liver metastasis, normal alk phos and a normal CT scan or liver/spleen scan, the specificity is 82% and 78%, respectively (Figure 1). An ultrasound may be of benefit in this situation. Not included in this study is the correlation of biochemical liver tests, liver imaging studies, and findings on physical examination. The presence of hepatomegaly, despite negative imaging studies, might persuade the clinician to pursue further the possibility of neoplastic liver disease. Laparoscopy is superior to a single blind biopsy in the diagnosis of hepatic malignancy and would be useful in detecting hepatic metastasis in patients with hepatic enlargement and negative imaging studies (16). In a field where the technology changes rapidly, one needs to critically reassess periodically the approach to the patient with suspected liver metastasis. REFERENCES 1. Bernardino M, Thomas J, Barnes P, Lewis E: Diagnostic approaches to liver and spleen metastases. Radiol Clin North Am 20: , Smith T J, Kemeny MM, Sugarbaker PH, Jones AE, Vermess M, Shawker TH: A prospective study of hepatic imaging in the detection of metastatic disease. Ann Surg 195: , Yeh H: Ultrasonography and computed tomography of the liver. Prog Liver Dis 6: , Snow JH, Goldstein HM, Wallace S: Comparison of scintigraphy, sonography, and computed tomography in the evaluation of hepatic neoplasms. Am J Radiol 132: , Clouse ME: Roentgenographic techniques for the diagnosis and management of liver tumors. Semin Oncol 10: , Elyaderani MK, Gabriele OF: Comparison of radionuclide imaging and ultrasonography of the liver. South Med J 76(1):37-44, Petasnick JP, Ram P, Turner DA, Fordham EW: The relationship of computed tomography, gray-scale ultrasonography, and radionuclide imaging in the evaluation of hepatic masses. Semin Nucl Med 9:8-21, Rosenthal S, Kaufman S: The liver scan in metastatic disease. Arch Surg 106: , Smalley RV, Malmud LS, Ritchie WGM: Preoperative scanning: Evaluation for metastatic disease in carcinoma of the breast, lung, colon, bladder and prostate. Semin Oncol 7: , Tempero MA, Peterson RJ, Zetterman RK, Lemon HM, Gurney J: Detection of metastatic liver disease. JAMA 248: , Wiener SN, Sachs SH: An assessment of routine liver scanning in patients with breast cancer. Arch Surg 113: , Wittes RE, Yeh SDJ: Indications for liver and brain scans. JAMA 238: , Castagna J, Benfield JR, Yamada H, Johnson DE: The reliability of liver scans and function tests in detecting hepatic metastases from ovarian carcinoma. Cancer 50: , Mettler FA, Christie JH, Crow NE, Garcia JF, Wicks JD, Bartow SA: Utility of liver/spleen scanning and serum enzyme level in detecting hepatic metastases from ovarian carcinoma. Cancer 50: , Spatz C, Johnsten J: Basic statistics. Belmont, California, Wadsworth Publishing Company, Jori GP, Peschle C: Combined peritoneoscopy and liver biopsy in the diagnosis of hepatic neoplasm. Gastroenterol 63: , 1972 Digestive Diseases and Sciences, Vol. 32, No. 10 (October 1987) I 117