GYNECOLOGIC
ONCOLOGY
37, 279-283(19%)
Correlation of Abdominal Ultrasound and Computed Tomography Scans with Second- or Third-Look Laparotomy in Patients with Ovarian Carcinoma BIRTHE LUND,’
KRISTIAN JACOBSEN, LONE RASCH, FLEMMING JENSEN, KNUD OLESEN, AND KNUD FELDT-RASMUSSEN
Departments of Oncology, Diagnostic Radiology, and Clinical Physiology and Nuclear Medicine, The Finsen Institute, Rigshospitalet, and the Departments of Internal Medicine C and Diagnostic Radiology, Bispebjerg Hospital, University Hospitals of Copenhagen, Copenhagen, Denmark Received August 2, 1989
To determine the best noninvasive means of evaluating response in patients with advanced ovarian carcinoma, 50 abdominal ultrasound (US) and computed tomography (CT) scans were performed in clinically disease-free ovarian cancer patients. The scans were correlated with the results obained at a subsequent second- or third-look laparotomy. CT and US were not complementary, and only mebstases larger than 2 cm were detected. The overall positive predictive value of nonconcordant scans was 57% compared with 100% for concordant CT and US (95% confidence limits: 18.4-90.1 and 29.2-N%, respectively). The corresponding negative predictive values were 45 and 47% (30.259.9 and 30.4-61.2%, respectively), if undetected microscopic disease was classified as a false-negative result. The negative predictive value of US and CT increased only to 60% in both cases,if undetected microscopic diseasewas registered as a truenegative result. Compared with the pelvic examination CT and US added positive information for 4 of 22 (18%) patients with macroscopic residual disease. In this study neither CT nor US was sensitive enough to preclude second-look laparotomy. 8 1990 Academk Pms,Inc.
(CT) scans in the evaluation of tumor response the proportion of clinically tumor-free patients who actually have obtained a pathological complete response has remained stationary at about 40-50% [2]. It is essential to detect progressive disease as early as possible so that treatment can be stopped or changed in nonresponding patients and unnecessary toxicity avoided. No prospective trials comparing both US and CT with the results obtained at a second- or third-look laparotomy have yet been published. Two questionswhich of these imaging modalities is better and are US and CT complementary in patients with residual disease not detected by a pelvic examination-have not been solved. To establish a better noninvasive means of response evaluation in clinically disease-free ovarian cancer patients a prospective trial was performed correlating the results obtained by both US and CT with the results found at a subsequent second-/third-look laparotomy.
INTRODUCTION
Chemotherapy in ovarian cancer patients has become more aggressive during the last decade, resulting in higher clinical response rates [I]. Combined with a more aggressive primary debulking surgery this has resulted in a greater number of patients who at some point during their treatment course will be classified as being clinically disease free. Disregarding the increasing use of either abdominal ultrasound (US) or computed tomography ’ To whom reprint requests should be addressed at the Department of Oncology, The Finsen Institute, Rigshospitalet, 9 Blegdamsve, DK2100 Copenhagen $3, Denmark.
MATERIAL
AND METHODS
Included in the trial were only patients with epithelial ovarian carcinoma who according to a clinical and pelvic examination were eligible for second- or third-look laparotomy. This group consisted of patients who either were clinically nonevaluable from the start of combination chemotherapy and in whom progressive disease had not been detected, or patients who had obtained a clinical complete response during treatment. Informed consent was obtained from all patients. To be eligible US and CT scans had to be performed within the 21 days preceding the laparotomy, and the gynecologists were not to be given the results of the scans. 279 ao!30-82.58/90 $1.50 Copyright 0 1990by AcademicPress,Inc. All rights of reproductionin any form reserved.
280
LUNDETAL.
The scans were performed at two hospitals by four experienced radiologists, who evaluated either US or CT scans without knowledge of the results of the complementary scanning modality. At the Finsen Institute, US was performed with a real-time scanner type 1846 (Bruel and Kjaer) with a 3.5/5 MHz sector-transducer, and CT with a Somatom DR 2 scanner (Siemens) with I.O-cm intervals and an g-mm “slice thickness.” At Bispebjerg Hospital, US was performed with a Siemens Sonoline 2, 3.5/5 MHz sector scanner, and CT with a Somatom DRG scanner (Siemens) with 1.O-cm intervals and a 1.Ocm “slice thickness.” Prone longitudinal and transverse US projections were obtained, and the patients were scanned from the diaphragm to the symphysis pubis with a full bladder. The patients received oral contrast 1 hr before the CT scan which was performed without a vaginal tampon. Local recurrences, metastases especially in the liver, adrenals, lymph nodes, omentum, and mesenterium, peritoneal carcinomatosis, ascites, and hydronephrosis were looked for. Subsequently, second-/third-look laparotomy was performed consisting of inspection and biopsy of suspected lymph nodes, peritoneal surfaces, and liver; diaphragm scrapings; and cytologic evaluation of ascitic fluid if present, otherwise peritoneal washings. Tumor debulking was performed if possible. The findings obtained at the laparotomy were the key to the scanning results. Nosographic sensitivity and specificity indicate the frequency rate of either true-positive or true-negative scanning results defined as either the probability that the scans will be positive or negative, respectively, if residual disease is present or not [3]. Positive and negative predictive values, though, are of a greater value in clinical decision making. The positive and negative predictive values of both CT and/or US scans were defined as the probability that residual disease was either present or not present if the scans were either positive or negative, respectively [3]. The following definitions were applied: Positive predictive value = the true positive ratio (TP)/total (true and false) positive ratio. Negative predictive value = the true negative ratio (TN)/total (true and false) negative ratio. Diagnostic accuracy = (TP + TN)/total true and false ratios.
The significance of differences was assessed by means of the x2 test of contingency tables [4]. RESULTS
Between January 1985 and June 1988, sixty-two patients entered the trial. Fifteen patients were ineligible. Four of these patients developed histologically verified progressive disease (PD) while awaiting second-look laparotomy, which as a consequence was not performed. In three of these patients the CT or US scans also revealed residual disease, whereas both were negative in the last patient. No statistical analysis can be performed in the present small group of this patient category, as patients with clinical residual disease were not otherwise included in the trial. In another 7 patients second-look laparotomy was postponed more than 21 days after the scans had been performed due to patient-related personal reasons. Reasons for ineligibility in two other patients were insufficient US scan of the retroperitoneum and the patient’s refusal of US scan, respectively. In the remaining two patients none of the scans were evaluated by any of the participating radiologists. Fifty CT and US scans were performed in 47 eligible patients. Three patients participated twice in the trial with a second- and a third-look laparotomy. They were considered as new cases as the time intervals between the two laparotomies were 7, 9, and 11 months, respectively. Primary FIG0 stages in the 47 eligible patients were IIB (4), IIIA (9), IIIB (4), IIIC (23), and IV (7) [5]. Pelvic examination and/or US at the start of chemotherapy classified 29 of the patients as clinically evaluable. Pretreatment CT scans were not performed routinely. The interval between the second-/third-look laparotomy and US was a median of 9 days (range l-21 days), and for CT, a median of 8 days (range 1-21 days), with 72% of the scans performed less than 14 days before the laparotomy. Forty (80%) of the patients were scanned and operated at the Finsen Institute and ten at Bispebjerg or Hvidovre Hospital. The results obtained at the laparotomies are registered in Table 1. Twenty-nine patients (58%) had pathologically verified residual disease; 11 of these (22%) had residual tumor >2 cm. In Table 2 are listed the overall numbers of true-posAs the purpose of the study was to disclose the ability of CT and US to detect “residual disease” versus “no itive, true-negative, false-positive, and false-negative reresidual disease,” scanning results were classified as true sults obtained with either CT, US, or CT and US compositive if all detected metastases were in the correct bined. In four patients (8%) the results were equivocal. location, even if some metastases were overlooked. In US was true positive/negative in two patients, and false case the detected metastases were in the wrong location positive/negative in the remaining two patients. CT was the scanning results were classified as false positive. Mi- true negative in one, false positive in two, and false croscopic residual disease not detected by CT or US was negative in one patient. Six of seven patients with microscopic residual disease registered as a false-negative result.
IMAGING MODALITIES
TABLE 1 Residual Tumor Size at 50 Second- or Third-Look Laparotomies in 47 Ovarian Cancer Patients Tumor size
n
0
21 I 11 5
Microscopic 5 cm
6
were claimed to be disease free by both CT and US. One patient, who only had microscopic disease in the pouch of Douglas, was diagnosed by US as having a 0.5 x 4-cm tumor in the bladder wall. This US result was registered as false positive, the CT result as false negative. Twenty-two patients had macroscopically visible residual disease. A thick coherent layer of intestinal carcinomatosis, which caused multiple adhesions, made one patient inoperable. This patient was claimed to have obtained a complete response by CT and US. In addition, peritoneal carcinomatosis was registered in 37 other locations and distinct residual metastases 2 cm were with microscopic residual disease, whereas Stehman et detected by either CT or US. Inter- and intraobserver al. [12] and Calkins et al. [13] have classified their pa- variability was not assessed in the present trial, but actients as having no residual disease if not detected by cording to Warde et al. [16] the intraobserver reliability of CT scans is good, whereas the interobserver reliability CT. The setup and definition of positive and negative re- is somewhat poorer. The possibility of performing a CT sults in two of the studies are comparable with the results review may thus be of educational value for the radiolof the present trial [6,14]. One might object that it is ogists, especially if performed together with the surgeon. illogical to assume that CT or US should be able to detect Because of the high false-negative rate a CT review has, microscopic disease. As microscopic residual disease, however, from the clinician’s point of view only resmall peritoneal implants, and tumors
ONCOLOGY
37, 279-283(19%)
Correlation of Abdominal Ultrasound and Computed Tomography Scans with Second- or Third-Look Laparotomy in Patients with Ovarian Carcinoma BIRTHE LUND,’
KRISTIAN JACOBSEN, LONE RASCH, FLEMMING JENSEN, KNUD OLESEN, AND KNUD FELDT-RASMUSSEN
Departments of Oncology, Diagnostic Radiology, and Clinical Physiology and Nuclear Medicine, The Finsen Institute, Rigshospitalet, and the Departments of Internal Medicine C and Diagnostic Radiology, Bispebjerg Hospital, University Hospitals of Copenhagen, Copenhagen, Denmark Received August 2, 1989
To determine the best noninvasive means of evaluating response in patients with advanced ovarian carcinoma, 50 abdominal ultrasound (US) and computed tomography (CT) scans were performed in clinically disease-free ovarian cancer patients. The scans were correlated with the results obained at a subsequent second- or third-look laparotomy. CT and US were not complementary, and only mebstases larger than 2 cm were detected. The overall positive predictive value of nonconcordant scans was 57% compared with 100% for concordant CT and US (95% confidence limits: 18.4-90.1 and 29.2-N%, respectively). The corresponding negative predictive values were 45 and 47% (30.259.9 and 30.4-61.2%, respectively), if undetected microscopic disease was classified as a false-negative result. The negative predictive value of US and CT increased only to 60% in both cases,if undetected microscopic diseasewas registered as a truenegative result. Compared with the pelvic examination CT and US added positive information for 4 of 22 (18%) patients with macroscopic residual disease. In this study neither CT nor US was sensitive enough to preclude second-look laparotomy. 8 1990 Academk Pms,Inc.
(CT) scans in the evaluation of tumor response the proportion of clinically tumor-free patients who actually have obtained a pathological complete response has remained stationary at about 40-50% [2]. It is essential to detect progressive disease as early as possible so that treatment can be stopped or changed in nonresponding patients and unnecessary toxicity avoided. No prospective trials comparing both US and CT with the results obtained at a second- or third-look laparotomy have yet been published. Two questionswhich of these imaging modalities is better and are US and CT complementary in patients with residual disease not detected by a pelvic examination-have not been solved. To establish a better noninvasive means of response evaluation in clinically disease-free ovarian cancer patients a prospective trial was performed correlating the results obtained by both US and CT with the results found at a subsequent second-/third-look laparotomy.
INTRODUCTION
Chemotherapy in ovarian cancer patients has become more aggressive during the last decade, resulting in higher clinical response rates [I]. Combined with a more aggressive primary debulking surgery this has resulted in a greater number of patients who at some point during their treatment course will be classified as being clinically disease free. Disregarding the increasing use of either abdominal ultrasound (US) or computed tomography ’ To whom reprint requests should be addressed at the Department of Oncology, The Finsen Institute, Rigshospitalet, 9 Blegdamsve, DK2100 Copenhagen $3, Denmark.
MATERIAL
AND METHODS
Included in the trial were only patients with epithelial ovarian carcinoma who according to a clinical and pelvic examination were eligible for second- or third-look laparotomy. This group consisted of patients who either were clinically nonevaluable from the start of combination chemotherapy and in whom progressive disease had not been detected, or patients who had obtained a clinical complete response during treatment. Informed consent was obtained from all patients. To be eligible US and CT scans had to be performed within the 21 days preceding the laparotomy, and the gynecologists were not to be given the results of the scans. 279 ao!30-82.58/90 $1.50 Copyright 0 1990by AcademicPress,Inc. All rights of reproductionin any form reserved.
280
LUNDETAL.
The scans were performed at two hospitals by four experienced radiologists, who evaluated either US or CT scans without knowledge of the results of the complementary scanning modality. At the Finsen Institute, US was performed with a real-time scanner type 1846 (Bruel and Kjaer) with a 3.5/5 MHz sector-transducer, and CT with a Somatom DR 2 scanner (Siemens) with I.O-cm intervals and an g-mm “slice thickness.” At Bispebjerg Hospital, US was performed with a Siemens Sonoline 2, 3.5/5 MHz sector scanner, and CT with a Somatom DRG scanner (Siemens) with 1.O-cm intervals and a 1.Ocm “slice thickness.” Prone longitudinal and transverse US projections were obtained, and the patients were scanned from the diaphragm to the symphysis pubis with a full bladder. The patients received oral contrast 1 hr before the CT scan which was performed without a vaginal tampon. Local recurrences, metastases especially in the liver, adrenals, lymph nodes, omentum, and mesenterium, peritoneal carcinomatosis, ascites, and hydronephrosis were looked for. Subsequently, second-/third-look laparotomy was performed consisting of inspection and biopsy of suspected lymph nodes, peritoneal surfaces, and liver; diaphragm scrapings; and cytologic evaluation of ascitic fluid if present, otherwise peritoneal washings. Tumor debulking was performed if possible. The findings obtained at the laparotomy were the key to the scanning results. Nosographic sensitivity and specificity indicate the frequency rate of either true-positive or true-negative scanning results defined as either the probability that the scans will be positive or negative, respectively, if residual disease is present or not [3]. Positive and negative predictive values, though, are of a greater value in clinical decision making. The positive and negative predictive values of both CT and/or US scans were defined as the probability that residual disease was either present or not present if the scans were either positive or negative, respectively [3]. The following definitions were applied: Positive predictive value = the true positive ratio (TP)/total (true and false) positive ratio. Negative predictive value = the true negative ratio (TN)/total (true and false) negative ratio. Diagnostic accuracy = (TP + TN)/total true and false ratios.
The significance of differences was assessed by means of the x2 test of contingency tables [4]. RESULTS
Between January 1985 and June 1988, sixty-two patients entered the trial. Fifteen patients were ineligible. Four of these patients developed histologically verified progressive disease (PD) while awaiting second-look laparotomy, which as a consequence was not performed. In three of these patients the CT or US scans also revealed residual disease, whereas both were negative in the last patient. No statistical analysis can be performed in the present small group of this patient category, as patients with clinical residual disease were not otherwise included in the trial. In another 7 patients second-look laparotomy was postponed more than 21 days after the scans had been performed due to patient-related personal reasons. Reasons for ineligibility in two other patients were insufficient US scan of the retroperitoneum and the patient’s refusal of US scan, respectively. In the remaining two patients none of the scans were evaluated by any of the participating radiologists. Fifty CT and US scans were performed in 47 eligible patients. Three patients participated twice in the trial with a second- and a third-look laparotomy. They were considered as new cases as the time intervals between the two laparotomies were 7, 9, and 11 months, respectively. Primary FIG0 stages in the 47 eligible patients were IIB (4), IIIA (9), IIIB (4), IIIC (23), and IV (7) [5]. Pelvic examination and/or US at the start of chemotherapy classified 29 of the patients as clinically evaluable. Pretreatment CT scans were not performed routinely. The interval between the second-/third-look laparotomy and US was a median of 9 days (range l-21 days), and for CT, a median of 8 days (range 1-21 days), with 72% of the scans performed less than 14 days before the laparotomy. Forty (80%) of the patients were scanned and operated at the Finsen Institute and ten at Bispebjerg or Hvidovre Hospital. The results obtained at the laparotomies are registered in Table 1. Twenty-nine patients (58%) had pathologically verified residual disease; 11 of these (22%) had residual tumor >2 cm. In Table 2 are listed the overall numbers of true-posAs the purpose of the study was to disclose the ability of CT and US to detect “residual disease” versus “no itive, true-negative, false-positive, and false-negative reresidual disease,” scanning results were classified as true sults obtained with either CT, US, or CT and US compositive if all detected metastases were in the correct bined. In four patients (8%) the results were equivocal. location, even if some metastases were overlooked. In US was true positive/negative in two patients, and false case the detected metastases were in the wrong location positive/negative in the remaining two patients. CT was the scanning results were classified as false positive. Mi- true negative in one, false positive in two, and false croscopic residual disease not detected by CT or US was negative in one patient. Six of seven patients with microscopic residual disease registered as a false-negative result.
IMAGING MODALITIES
TABLE 1 Residual Tumor Size at 50 Second- or Third-Look Laparotomies in 47 Ovarian Cancer Patients Tumor size
n
0
21 I 11 5
Microscopic 5 cm
6
were claimed to be disease free by both CT and US. One patient, who only had microscopic disease in the pouch of Douglas, was diagnosed by US as having a 0.5 x 4-cm tumor in the bladder wall. This US result was registered as false positive, the CT result as false negative. Twenty-two patients had macroscopically visible residual disease. A thick coherent layer of intestinal carcinomatosis, which caused multiple adhesions, made one patient inoperable. This patient was claimed to have obtained a complete response by CT and US. In addition, peritoneal carcinomatosis was registered in 37 other locations and distinct residual metastases 2 cm were with microscopic residual disease, whereas Stehman et detected by either CT or US. Inter- and intraobserver al. [12] and Calkins et al. [13] have classified their pa- variability was not assessed in the present trial, but actients as having no residual disease if not detected by cording to Warde et al. [16] the intraobserver reliability of CT scans is good, whereas the interobserver reliability CT. The setup and definition of positive and negative re- is somewhat poorer. The possibility of performing a CT sults in two of the studies are comparable with the results review may thus be of educational value for the radiolof the present trial [6,14]. One might object that it is ogists, especially if performed together with the surgeon. illogical to assume that CT or US should be able to detect Because of the high false-negative rate a CT review has, microscopic disease. As microscopic residual disease, however, from the clinician’s point of view only resmall peritoneal implants, and tumors
