Tumour marker CA 50 levels compared to signs and symptoms in the diagnosis of pancreatic cancer

European Journal of Surgical Oncology 1997; 23:151-156

Tumour marker CA 50 levels compared to signs and symptoms in the diagnosis of pancreatic cancer Birger P~lsson*, Parvesh Massont and ~,ke Andr~n-Sandberg* *Department of Surgery, University Hospital, Lund, Sweden

The diagnostic merits of CA 50 and of symptoms indicating pancreatic cancer (pain, jaundice, weight loss, malabsorption) were compared prospectively in 512 consecutive patients. Among the final diagnoses were: exocrine pancreatic cancer, 175; periampullary cancer, 44; other gastrointestinal cancer, 45; and chronic pancreatitis, 64 cases. The suspected diagnoses based on symptoms and signs were correct in 80% of the patients with exocrine pancreatic cancer, in 78% with periampullary, in 76% with other gastrointestinal cancer and in 90% with chronic pancreatitis. CA 50 was pathological in 96% of the cases with exocrine pancreatic cancer, in 70% with periampullary, in 78% with other gastrointestinal malignancies and in 36% with chronic pancreatitis. The sensitivity was 96%, specificity 48%, positive prediction 49% and negative prediction 96%, depending on cut-off level. The single CA 50 value was comparable to symptoms and signs regarding sensitivity and negative prediction. In 28 of 42 cases incorrectly clinically classified, CA 50 alone indicated a benign or malignant diagnosis. If both the modalities 'signs and symptoms' and CA 50 were combined, the sensitivity was 91%, the specificity 92%, the positive prediction 86% and the negative prediction 95%. The initial CA 50 value can help to indicate in which patients a pancreatic malignancy should be suspected.

Key words: CA 50; tumour marker; tumour-associated antigens; exocrine pancreatic cancer.

Introduction The initial symptoms of exocrine pancreatic cancer are often vague or similar to those of different benign diseases, especially chronic pancreatitis and biliary disorders. Common clinical signs are upper abdominal pain, obstructive jaundice and decreased appetite with weight loss. t~ The most valuable diagnostic tools have been ultrasonography (U.S.), computerized tomography (CT), endoscopic retrograde cholangiopancreaticography (ERCP), percutaneous fine needle biopsy (guided by U.S. or CT) and exploratory laparotomy.~'2 However, differential diagnostic problems remain, especially between pancreatic cancer and chronic pancreatitis, sometimes even at surgery. Monoclonal antibodies (mabs) developed by the hybridoma technique have offered new diagnostic aids by identifying tumour-associated antigens in blood and tissues and have become recommended clinical routine in some situations? In pancreatic cancer, the antigen CA 50 has shown high sensitivity ranging between 69-96% and specificity varying between 71-98%, depending on differences in published patient materials. 6-~4The antigen is identified by the mab C 50 against a colonic carcinoma cell line in mice and is recognized by the epitopes sialosyllactoN-fucopentaose II (corresponding to sialylated blood group Correspondence to: B. Phlsson, Dept. of Surgery, University Hospital, S-221 85 Lund, Sweden. "tin memoriam. Earlier at Department of Bioanalytical Chemistry, Astra-Draco AB, Lund, Sweden.

substance Lewis~) and the analogue sialyllacto-N-tetraose, lacking fucose. 1517 Hence, it differs from the closely related CA 19-9, derived from the same procedure, which only identifies the former structure, t8"~9 Despite several retrospective reports on the diagnostic accuracy of these mabs in patients with previously defined pancreatic diseases, little is known about the diagnostic properties in relation to the symptoms and signs of the patients. The aim of the present investigation was to compare prospectively the diagnostic accuracy of symptoms and signs as assessed by the physician, in a large series of consecutive, symptomatic patients with suspected pancreatic diseases, to the discriminatory capability of the tumour marker CA 50 at the initial stage of a clinical investigation.

Patients and methods

Patients Five hundred and twelve consecutive patients were included. They were admitted to the Department of Surgery, Lund University Hospital, Sweden, due to defined symptoms and signs, which could indicate pancreatic disease: abdominal pain (alone or in combination with high lumbar pain) for more than 4 weeks without any other explanation, visible jaundice or bilirubin in serum >25 pmol/l, weight loss > 10% of the body weight during the last 3 months or signs of malabsorption (mainly diarrhoea or steatorrhoea). No further selections were made among the patients showing

B. Pdlsson et al.

152

Table I. Characteristics of the patient population Diagnoses n

Age mean (range)

Sex M:F

Pain (%)

Jaundice (%)

Weight loss n*

Malabsorption n*

Exocrine pancreatic cancer Chronic pancreatitist Endocrine pancreatic tumours:l: Periampullary cancer§ Gastrointestinal cance~ Benign bile duct diseasesll Benign hepatic diseases** Benign gastrointestinal diseasett Other cancerSJ;

175 64 15 44 45 52 14 96 7

66.6 (38-93) 51.3 (16-75) 60.4 (34-84) 69.2 (49-91 ) 71.7 (53-90) 66.5 (39-93) 57.2 (38-80) 63.2 (19-90) 60.4 (21-78)

99:76 48:16 6:9 26:18, 18:27 19:33 9:5 35:61 5:2

58 77 40 45 62 62 36 19 100

67 14 -61 53 58 71 15 29

42/88 1/34 5/9 8/16 13/30 1/43 2/10 2/52 0/3

18/97 1/37 I/12 2•20 4/30 1/44 1/10 1/51 I/4

Total

512

64.2 (16-93)

265:247

63

45

74/285

301305

* Fewer registrations; positive numbers of made observations. i Including eight with recurrent acute pancreatitis and five with pancreatic cysts. $ Insulinoma six, 'mixed' five, PP-om two, gastrinoma one and glucagonoma one. § Bile duct cancer 19, papilla Vateri cancer 18 and cancer duodeni juxtapapillary part seven. ¶ Colonic from right upper quadrant of abdomen 14, gall bladder 12, primary hepatic 10, hepatic metastases from colonic adenocarcinomas five, gastric three and oesophagus one. IIStones dc. choledochus 31, benign stenosis dc. choledochus or papilla Vateri 18, cholangitis three. ** Cirrhosis seven and chronic hepatitis seven. t t Gallbladder stones 73, ulcus duodeni/gastritis five, no diagnoses 18 (see Patients). ~:~Lymphoma four, gynecological one, pulmonary with hepatic metastases one and renal one.

one or more of these symptoms and signs. Further characteristics of the patient population and final diagnoses are shown in Table 1.

Diagnostic procedures The patients were examined according to a standardized protocol, where the symptoms and signs mentioned earlier were registered by the surgeon in charge. All patients were examined by one of two responsible surgeons. The surgeon's estimation of each patient was used in a scoring system, where the surgeon had to score the grade of suspicion of exocrine pancreatic cancer as: very high, high, not unlikely or unlikely. This was done irrespective of what the surgeon finally stated as the most probable diagnosis. Blood tests were taken at the initial examination, including plasma for CA 50 analyses, which were performed without any knowledge about the individual patient. The results were for scientific use only and blinded until the final diagnoses were established. The main diagnostic methods used, alone or in combinations, were U.S., CT, ERCP, percutaneous transhepatic cholangiography (PTC) and explorative laparotomy. These diagnostic procedures were requested as normal routine investigations, not as special tests performed for the purpose of this study, and were the same throughout the investigation period. All cases of malignancy were histologically or cytologically confirmed. All patients were followed up for at least 1 year or until death; the benign cases were followed up for a median of 3.5 years.

CA 50 assay Plasma samples were stored at - 2 0 ° C until analysed. CA 50 was analysed with an immunoradiometric method (CA 50 I R M A , Pharmacia CanAg, Gothenburg, Sweden) according to the manufacturer's instructions. The dynamic range of

the method extended to 160 U/ml. The upper reference value defined as the cut-off level was 20 U/ml, representing the 95th percentile of a population of blood donors. Every sample was analysed in duplicate and dilutions were made when appropriate to keep within the analytical range, according to our previous findings when comparing different CA 50 assays. -'°

Statistics Sensitivity is defined as the frequency of pathological tests in patients with the disease, and specificity as the frequency of values lower than the reference limit in patients without the disease. The predictive value of a positive result is the frequency of pathological tests in patients with the disease compared to all pathological tests and the predictive value of a negative result is the frequency of normal tests in patients without the disease compared to all normal tests. Statistical comparisons were made using the z-test or M a n n - W h i t n e y U-test where appropriate.

Results

The 512 patients were divided in nine diagnostic groups, described in Table I. In total, 29 different diagnoses were established, of which 18 were malignant, among them 34% exocrine pancreatic cancer. A conclusive diagnosis could not be obtained in 18 cases with abdominal pain without any other sign or symptom, despite extensive investigations and follow-up for at least 2 years. These cases were classified as benign gastrointestinal diseases. The frequencies of symptoms and signs are also shown in Table 1, where the most common symptom in the patients suffering from exocrine pancreatic cancer was jaundice in 67%. It was also frequent in cases with other diseases related to the hepato-biliary system. Pain was mostly recorded in

CA 50, symptoms and signs in pancreatic cancer

153

Table 2. Number of patients (n) with CA 50 >20 U/ml and CA 50 values in different diagnoses Diagnoses

CA 50

Exocrine pancreatic cancer* Chronic pancreatitis Endocrine pancreatic tumour Periampullary cancer Gastrointestinal cancert Benign bile duct diseases:~ Benign hepatic diseases Benign gastrointestinal disease Other cancer

>20 U/ml n

Mean U/ml

Median U/ml

Range U/ml

SEM U/ml

Lower quartile U/ml

Upper quartile U/ml

168/175 23/64 5/15 31/44 35/45 36/52 I0/14 31/96 4/7

905 37 18 195 541 130 65 40 35

175 14 10 50 83 43 41 13 21

10-15800 0-593 2-48 3-2126 3-4043 8-984 8-181 0-661 3-104

167 12 3 62 147 31 15 9 14

109 9 8 14 23 18 18 I0 7

73 I 25 28 138 711 120 116 27 65

* Six extreme values ranging from 30,000 to 483,000 U/ml excluded (if included mean 6373 U/ml, median 178 Ulrnl, SEM 3044 U/ml, range: 10--483,000U/ml, lower quartile I10 and upper quartile 780 U/ml). t One extreme value 11,840 U/ml excluded (if included mean 797 U/ml, median 90 U/ml, SEM 294 U/ml, range: 3-11 840 U/ml, lower quartile 30 and upper quartile 754 U/ml). 3~One extreme value 2960 U/ml excluded (if included mean 185 U/ml, median 44 U/ml, SEM 62 U/ml, range: 8-2960 U/ml, lower quartile 18 and upper quartile 136 U/ml).

Table 3. Statistical comparisons (P-values) between mean levels of CA 50 in different diagnoses Diagnoses Diagnoses Chronic pancreatitis (CP) Endocrine pancreatic tumour (EPT) Periampullary cancer (PaC) Gastrointestinal cancer (GIC) Benign bile duct diseases (BBD) Benign hepatic diseases (BHD) Benign gastrointestinal diseases

PC

CP

EPT

PaC

G1C

BBD

BHD

<0.000 <0.000 <0.000 0.003 <0.000 <0.000 <0.000

NS <0.000 <0.000 <0.000 0.004 NS

0.002 <0.000 0.001 0.008 NS

NS NS NS <0.000

0.03 NS <0.000

NS <0.000

0.009

PC, exocrine pancreatic cancer; NS, not significant.

those with chronic pancreatitis with 77%. Weight loss and malabsorption were most frequently observed in patients with pancreatic cancer. The n u m b e r of patients with plasma-CA 50 >20 U/ml and statistical comparisons are given in Tables 2 and 3. In total, eight extreme values (outliers) of CA 50 were excluded in the statistical comparison, in order to avoid high statistical skewness. CA 50 exceeded the cut-off level 20 U/ml in 343 of the 512 cases, and 234 (68%) of these had some kind of malignancy (except the 15 endocrine pancreatic neoplasias and the seven extra-gastrointestinal cancers). In the patients suffering from exocrine pancreatic cancer, the CA 50 values were raised in 96%. Thirty-two per cent of the patients with benign diseases had elevated CA 50 values, dominated by 71% of those with hepatic diseases and 69% of the patients with benign bile duct diseases. CA 50 showed significantly higher plasma levels in the 175 patients with exocrine pancreatic cancer than in all the other cases; it was <20 U/ml in only seven patients (Table 2) and none had a value below 10 U/ml. These seven patients showed no uniform characteristics. Also, the highest values overall were found among the patients suffering from pancreatic cancer, up to 483,000 U/ml, resulting in the broadest range (10-483,000 U/ml). Even when the six most extreme values were excluded, the median value was more

than twice that of the next group (175 U/ml vs 83 U/ml in other gastrointestinal cancer). To compare, the highest CA 50 value in patients with chronic pancreatitis was 593 U/ ml, exceeded by 30% of the cases with pancreatic cancer. In the groups with periampullary cancer and other gastrointestinal cancer CA 50 was >20 U/ml in 70% and 78%, respectively (Table 2). Although the plasma levels of CA 50 were lower in patients with benign bile duct diseases and benign hepatic disorders, there was no or only a slight statistical difference in relation to the two cancer groups mentioned. Patients with endocrine pancreatic tumours, chronic pancreatitis and benign gastrointestinal diseases all showed similarly low CA 50 levels (medians 10-14U/ml), without statistical differences between them. The operative parameters for CA 50 are shown in Table 4 at five different cut-off levels and at three different comparisons (further explained in Discussion). The accuracy of the most probable diagnosis, based on the initial clinical examination and symptoms at the same time as the blood samples were drawn, illustrated in Table 5, was at least 73% (benign bile duct diseases) and 80% concerning suspected cases of exocrine pancreatic cancer. The distribution of the scores of suspicion of pancreatic cancer in relation to the final diagnoses and compared to the CA 50 levels is shown in Table 6.

B. Pdlsson et al.

154

Table 4. Assay parameters in percentage for CA 50 at different cut-

with initial benign diagnosis but where the correct diagnosis was a malignancy; twelve (92%) showed pathological C A 50 values. A slight tendency towards higher C A 50 levels in the malignant group was noticed (mean: 200+278 (SD)) with the extreme value of 1736 U/ml excluded vs 143 + 2 0 0 (SD) U/ml in the benign). A probable malignant diagnosis finally showing another type of malignancy is not regarded as inrorrect, neither is a probable benign diagnosis turning out to be another benign disease. In Table 7 an overall comparison is made between the diagnostic reliability of the clinical diagnosis and the results of C A 50.

off levels with respect to exoerine pancreatic cancer vs pancreatic diseases, exocrine pancreatic cancer vs all diseases and 'relevant cancer' (exocrine pancreatic cancer, periampullary cancer and gastrointestinal cancer) vs all diseases, n=number of patients included Cut-off level (U/ml):

20

27

37

70

220

Pancreatic cancer vs pancreatic diseases (n = 175 vs 79) Sensitivity 96 94 93 85 Specificity 64 77 85 95 Positive predictive value 86 90 93 97 Negative predictive value 88 86 79 90 Pancreatic cancer vs all diseases (n = 175 vs 337) Sensitivity 96 94 93 85 Specificity 48 57 63 78 Positive predictive value 49 '53 56 67 Negative predictive value 96 96 95 91 'Relevant' cancer vs all diseases (n = 264 vs 248) Sensitivity 88 85 82 72 Specificity 55 65 71 86 Positive predictive value 68 72 75 85 Negative predictive value 82 80 80 74

47 97 98 45 47 88 67 76

Discussion

40 94 87 60

The differential diagnostic problems concerning pancreatic cancer could be large, as illustrated by the 29 different diagnoses in our patients. Therefore, our study was designed to evaluate the contribution of C A 50 to the surgeon's initial diagnosis in consecutive, symptomatic patients, rather than C A 50 with respect to certain diagnoses. Even though our patients were consecutive, they were selected with respect to the symptoms and signs and only two surgeons were responsible for the investigation. This contributed to a

In 29 cases an initial malignant diagnosis was wrong and the final outcome was benign; in 16 of these patients (55%) C A 50 was <20 U/ml. The opposite occurred in 13 patients

Table 5. Most probable initial* diagnoses vs final diagnoses. Figures in brackets denominate percentage of correct most probable initial

diagnoses Probable diagnoses Final diagnoses Exocrine pancreatic cancer (PC) Chronic pancreatitis (CP) Endocrine pancreatic tumour (EPT) Periampullary cancer (PaC) Gastrointestinal cancer (GIC) Benign bile duct diseases (BBD) Benign hepatic diseases (BHD) Benign gastrointestinal disease (BGID) Other cancer Total

PC

CP

EPT

PaC

GIC

BBD

BHD

BGID

Total

166 (80) 8 2 10 7 6 3 4 1

I 55 (90) 0 0 0 0 1 4 0

0 0 13 (93) 0 0 0 0 0 1

2 1 0 32 (78) 3 3 0 0 0

2 0 0 0 29 (76) I I I 4

1 0 0 2 2 32 (73) 0 7 0

0 0 0 0

3 0 0 0

I

3

0 9 (90) 0 0

10 0 80 (83) 0

175 64 15 44 45 52 14 96 6*

61

14

41

38

44

10

96

207

* Seven 'other cancer' only shown with respect to final diagnoses; only one case initially classified as 'other cancer'.

Table 6. Grade of suspicion of exocrine pancreatic cancer at admission in relation to final diagnoses. Figures in brackets percentage of

final diagnoses with CA 50 values >20 U/ml Grade of suspicion % of final diagnoses (% with CA 50 >20 U/ml) Final diagnoses Exocrine pancreatic cancer Chronic pancreatitis Endocrine pancreatic tumour Periampullary cancer Gastrointestinal cancer Benign bile duet diseases Benign hepatic diseases Benign gastrointestinal diseases Other cancer

n 175 64 15 44 45 [ 52 14 96 7

Very high

High

Not unlikely

Unlikely

84 (81) 8 (3) 13 (7) 20 (16) 16 (11) 5 (6) 14 (7) 2 (I) 29 (14)

14 (14) 20 (5) 7 (7) 41 (30) 16 (18) 21 (13) 14 (14) 11 (1) 14 (14)

2 (0.5) 42 (17) 47 (13) 34 (23) 50 (38) 40 (31) 58 (43) 67 (23) 43 (28)

0 (0) 30 (I 1) 33 (7) 5 (2) 18 (I 1) 34 (21) 14 (7) 20 (7) 14 (0)

CA 50, symptoms and signs in pancreatic cancer Table 7. Comparison between the diagnostic accuracy in percentage of symptoms and signs vs CA 50 with cut-off level 20 U/ml in exocrine pancreatic cancer vs other diseases and 'relevant' cancer (exocrine pancreatic cancer, periampullary cancer and gastrointestinal cancer) vs other diseases CA 50 Pancreatic cancer vs all diseases (n= 175 vs 337) Sensitivity 96 Specificity 48 Positive predictive value 49 Negative predictive value 96 'Relevant' cancer vs all diseases (n=264 vs 248) Sensitivity 88 Specificity 55 Positive predictive value 68 Negative predictive value 82

Symptoms and signs 95 86 80 91 86 76 79 84

high prevalence of malignancy with 56% overall and 34% pancreatic cancer (Table 1). The frequencies of the symptoms and signs were in concordance with the literature, with pain reported in 72%, jaundice in 30-68% and weight loss in 7 I% of the pancreatic cancer patients. 2-4 Weight loss and malabsorption were only registered in part of our cases, but when the patient records were reexamined, it was clear that patients without these symptoms actually were more numerous than indicated in Table I. Based primarily on the symptoms and signs, the patient history and the clinical examination, 80% of the patients with probable exocrine pancreatic cancer were correctly classified (Table 5). The frequencies were of the same order for the other malignancies. On the other hand, this means that every fifth patient thought to have an exocrine pancreatic cancer was classified incorrectly. The majority of patients with benign conditions, 73-90%, was also assigned a correct diagnosis initially. This could be expected, since some dominating benign disorders, in particular biliary stones (70% of the benign bile duct and other benign gastrointestinal diseases), often show a typical patient history and clinical picture. However, about 20% overall were not judged correctly and 13 cases (6%) actually had malignancies. The scoring system was used to judge the probability of pancreatic cancer in each patient, despite the preliminary diagnosis. Ninety-eight per cent of the pancreatic cancer cases were scored as suspicion 'very high' or 'high', followed by peri-ampullary cancer with 61%, confirming the difficulty in differentiating these malignancies (Table 6). The majority (72-87%) of the patients with benign conditions was scored as 'not unlikely' or 'unlikely' to have pancreatic cancer. Thus, the systematic scoring gave a quite good indication whether a pancreatic cancer should be suspected, even in disorders where jaundice and pain were frequent. The comparison with the raised CA 50 values (in brackets in Table 6) implies what the contribution of these could have been at the initial examination, if they had been known to the surgeon. For instance 28% of the patients with chronic pancreatitis were denominated the two high-risk scorings of having pancreatic cancer, but only 8% had raised CA 50 values, i.e. the CA 50 test indicated a benign disease.

155

In Table 7, we have attempted to illustrate the diagnostic accuracy of symptoms and signs in relation to CA 50. This comparison needs to be interpreted with caution, since the quantity 'symptoms and signs' is multifactorial, assessed by the clinician and more diverse than CA 50, which is just a single laboratory parameter on a one-dimensional scale without any knowledge of the patient by the examiner. Despite this the sensitivity and negative predictive value were comparable for CA 50 and 'symptoms and signs' concerning both pancreatic cancer vs all other diseases and 'relevant' cancer vs all other diseases. As a comparison, both the modalities CA50 and 'symptoms and signs' could be combined. With the requirement that at least one should be 'positive', the following assay parameters were yielded for pancreatic cancer (n = 175) vs all diseases (n = 337): sensitivity, 100%; specificity, 43%; positive predictive value, 48%; and negative predictive value, 100%. If both modalities were required 'positive', the sensitivity would be 91%; specificity, 92%; positive predictive value, 86%; and negative predictive value, 95%. CA 50 had a sensitivity of 96% regarding pancreatic cancer (Table 4), which is among the highest reported. 6-14 Moreover, most studies are retrospective and difficult to compare with a prospective series. As the CA 50 antigen was derived from a colorectal adenocarcinoma cell line, it is not expected to be specific for pancreatic cancer, and it has consequently been identified in other gastrointestinal adenocarcinomas. 2t'z~ Therefore Table 4 contains three different calculations: exocrine pancreatic cancer compared to other pancreatic diseases, compared to all other diseases and 'relevant' cancer vs all diseases. 'Relevant' cancer should be interpreted clinically and refers to other adenocatcinomas of gastrointestinal origin, where CA 50 has been identified. Despite this broader use, the sensitivity was 88%. The specificity varied between 48-64% in the three different groups. This was partly due to the study design, since no further selections were made among the patients once they were referred, i.e. CA 50 was measured even if other factors indicated a benign disease. However, raising the cut-off level improved the specificity without appreciably reducing the sensitivity, e.g. the specificity improved from 48% at 20 U/ml to 63% at 37 U/ml when pancreatic cancer was compared to all diseases, while the sensitivity was lowered only from 96% to 93% (Table 4). This reflects that the 'false' positive cases generally show low plasma levels of CA 50, also illustrated in Table 2. A contributing factor to 'false' positive CA50 values might be cholestasis.6'7'23 The frequencies of jaundice were about the same in patients with exocrine pancreatic cancer (67%), periampullary cancer (61%), benign bile duct diseases (58%) and benign hepatic diseases (71%) (Table 1). The plasma levels of CA 50 in these two benign groups showed no statistical difference compared with periampullary cancer and only slightly to other gastrointestinal cancer, while a strong difference occurred between these benign groups and pancreatic cancer (Table 3). Thus, it seems probable that jaundice to some extent might influence the plasma levels of CA 50, at least in benign disorders, but only slightly in pancreatic cancer. In the clinical situation, the prediction rates are probably

156

B. Pdlsson et al.

the most important. The positive predictive rates were 49% comparing pancreatic cancer to all other diseases, 68% in 'relevant' cancer vs other disorders and 86% comparing pancreatic cancer vs other pancreatic diseases (Table 4). This means that two of three patients with a pathological C A 50 value actually had some kind of gastrointestinal malignancy. The prediction of a negative test result ranged from 82 to 96% in the three groups, implying that a normal plasma C A 5 0 level strongly indicated no malignancy, especially not pancreatic cancer. The choice of cut-off level is debatable. The recommended 2 0 U / m l is derived from a population of blood donors. This is more relevant when screening a mainly healthy population. The Working G r o u p on Tumour Marker Criteria ( W G T M C ) has therefore recommended that the reference population should consist of patients with relevant benign diseases. '4 To illustrate this, we have calculated five different cut-off levels in Table 4; the mean represents 20 U/ ml (13U/ml) + 2 S D (I S D = 3 . 5 U/ml) derived from the mentioned blood donors and 27 U/ml is the same mean + 4 SD. The mean is 37 U/ml for our patients with chronic pancreatitis, i.e. the group with a benign disease in the organ of interest. Taking into account the SD for this group (92 U/ml), the mean + 2 SD would be 220 U/ml. Another way to establish a cut-off level could be to use the 95th percentile. In the pancreatic cancer patients this was 27 U/ ml, whereas for patients with chronic pancreatitis 95% were excluded at 70 U/ml. As seen (Table 4), the balance between the assay parameters was better if a higher cut-off level was used, e.g. 27, 37 or even 70 U/ml, depending on which diseases that were compared. Therefore, it seems reasonable to change the cut-offvalue according to the purpose of using a certain tumour marker, based upon earlier investigations. In conclusion, in symptomatic patients the single plasma CA 50 value alone showed about the same sensitivity and negative predictive value as symptoms and signs and patient history interpreted by the clinician. A pathological CA 50 value strongly indicated a pancreatic cancer, which should be investigated by ordinary methods, while a normal C A 50 level was highly indicative of the absence of cancer.

5. 6.

7. 8. 9. 10.

I 1. 12. 13.

14.

15.

16.

17.

18. 19.

Acknowledgement Excellent technical assistance was provided by Mrs AnneCathrine L6fstr6m. This work was supported financially in part by grants from The Faculty of Medicine, University of Lund, and by The County Hospital of V~.xj6, Sweden.

20.

References

22.

1. Go VLW, Taylor WF, DiMagno EE Efforts at early diagnosis of pancreatic cancer: the Mayo Experience. Cancer 1981; 47: 1698-703. 2. Moossa AR. Pancreatic cancer. Approach to diagnosis, selection for surgery and choice of operation. Cancer 1982; 50: 2689-98. 3. Kalser MH, Barkin JS, Maclntyre IM. Pancreatic cancer. Assessment of prognosis by clinical presentation. Cancer 1985; 56: 397-402. 4. Cubilla AL, Fitzgerald PJ. Tumors of the exocrine pancreas.

21.

23.

24.

In: Atlas o f Tumor Patholog); 2nd series, fasc. 19. Armed Forces Institute of Pathology: Washington, DC, 1984: 1-4, 111-54. International Union Against Cancer Report. Workshop on immunodiagnosis. Cancer Res 1986; 46: 3744-6. Kuusela P, Haglund C, Roberts PJ, Jalanko H. Comparison of CA-50, a new tumour marker, with carcinoembryonic antigen (CEA) and alpha-fetoprotein (AFP) in patients with gastrointestinal diseases. Br J Cancer 1987; 55: 673-6. Paganuzzi M, Onetto M, Marroni P, Barone D, Conio M, Aste ~, Pugliese V. CA 19-9 and CA 50 in benign and malignant pancreatic and biliary diseases. Cancer 1988; 61: 2100-8. Haglund C, Kuusela P, Jalanko H, Roberts P. Serum CA 50 as a tumor marker in pancreatic cancer: a comparison with CA 19-9. lnt J Cancer 1987; 39: 477-81. Masson P, P~lsson B, Andr~n-Sandberg ,/k. Evaluation of CEA, CA 19-9, CA-50, CA-195 and TATI with special reference to pancreatic disorders. Int J Pancreatol 1991; 8: 333-44. Benini L, Cavallini G, Zordan D, et al. A clinical evaluation of monoclonal (CA 19-9, CA 50, CA 12-5) and polyclonal (CEA, TPA) antibody-defined antigens for the diagnosis of pancreatic cancer. Pancreas 1988; 3: 61-6. Blind PJ, Dahlgren ST. Serum levels ofthe carbohydrate antigen CA-50 in pancreatic disease. Acta Chh" Scand. 1986; 153: 45-9. Haglund C, Roberts PJ, Jalanko H, Kuusela P. Tumour markers CA 19-9 and CA 50 in digestive tract malignancies. Scand J Gastroenterol 1992; 27: 169-74. Habib NA, Hershman MJ, Haberland F, Papp L, Wood CB, Williamson RCN. The use of CA-50 radioimmunoassay in differentiating benign and malignant pancreatic disease. Br J Cancer 1986; 53: 697-9. Kobayashi T, Kawa S, Tokoo M, et al. Comparative study of CA-50 (Time-Resolved Fluoroimmunoassay), Span-l, and CA 19-9 in the diagnosis of pancreatic cancer. Scand J Gastroenterol 1991; 26: 787-97. Lindholm L, Holmgren J, Svennerholm L, et al. Monoclonal antibodies against gastrointestinal tumour-associated antigens isolated as monosialogangliosides. Int Archs Allergy Appl bnmun 1983; 71: 178-81. M~msson JE, Fredman P, Nilsson O, Lindholm L, Holmgren J, Svennerholm L. Chemical structure ofcarcinoma ganglioside antigens defined by monoclonal antibody C-50 and some allied gangliosides of human pancreatic adenocarcinoma. Biochhn Biophys Acta 1985; 834: 110-7. Nilsson O, M~nsson JE, Lindholm L, Holmgren J, Svennerholm L. Sialosyllacto-tetraosyl ceramide, a novel ganglioside antigen detected in human carcinomas by monoclonal antibody. FEBS Lett 1985; 182: 398-402. Koprowski H, Herlyn M, Steplewski Z, et al. Colorectal carcinoma antigens detected by hybridoma antibodies. Somat Cell Genet 1979; 5: 957-72. Magnani JL, Nilsson B, Brockhaus M, et al. A monoclonal antibody-defined antigen associated with gastrointestinal cancer is a ganglioside containing sialylated lacto-Nfucopentaose II. J Biol Chem 1982; 257: 14365-9. Masson P, P~lsson B, Andr6n-Sandberg A. CA-50 in patients with different pancreatic diseases--an evaluation of three different laboratory techniques. Scand J Clin Lab Invest 1988; 48: 751-5. Kuusela P, Haglund C, Roberts PJ. Comparison of a new tumour marker CA 242 with CA 19-9, CA 50 and carcinoembryonic antigen (CEA) in digestive tract diseases. Br J Cancer 1991; 63: 636-40. Nilsson O, Johansson C, GIimelius B, Persson B, N6rgaardPedersen B, Andr6n-Sandberg ,/k, Lindholm L. Sensitivity and specificity ~f CA 242 in gastrointestinal cancer. A comparison with CEA, CA 50 and CA 19-9. Br J Cancer 1991; 65: 215-21. Phlsson B, Masson P, Andr6n-Sandberg ,~,. The influence of cholestasis on CA 50 and CA 242 in pancreatic cancer and benign biliopancreatic diseases. Scand J Gastroentero11993; 28: 981-7. Bonfrer JMG. Working group on tumour marker criteria (WGTMC). Tumor Biol 1990; 11: 287-8.

Accepted for publication 5 February 1997