- Email: [email protected]
Cea tissue staining in colorectal cancer patients - correlation with plasma cea, histology and staging
Pathology
ISSN: 0031-3025 (Print) 1465-3931 (Online) Journal homepage: http://www.tandfonline.com/loi/ipat20
CEA Tissue Staining in Colorectal Cancer Patients — Correlation with Plasma CEA, Histology and Staging Irene O. L. Ng, Joana Ho, C. J. Pritchett, Eric Y. T. Chan & Faith C. S. Ho To cite this article: Irene O. L. Ng, Joana Ho, C. J. Pritchett, Eric Y. T. Chan & Faith C. S. Ho (1993) CEA Tissue Staining in Colorectal Cancer Patients — Correlation with Plasma CEA, Histology and Staging, Pathology, 25:3, 219-222 To link to this article: http://dx.doi.org/10.3109/00313029309066575
Published online: 06 Jul 2009.
Submit your article to this journal
Article views: 6
View related articles
Citing articles: 1 View citing articles
Full Terms & Conditions of access and use can be found at http://www.tandfonline.com/action/journalInformation?journalCode=ipat20 Download by: [Universität Osnabrueck]
Date: 21 January 2016, At: 06:13
Pathology (1993), 25, pp. 219-222
CEA TISSUE STAINING IN COLORECTAL CANCER PATIENTS CORRELATION WITH PLASMA CEA, HISTOLOGY AND STAGING IRENE0 . L. NG, JOANA Ho, C. J . PRITCHETT,* ERICY. T. CHANAND FAITHC . S . Ho Departments of Pathology and *Surgery, University of Hong Kong, Queen Mary Hospital, Hong Kong
Downloaded by [Universität Osnabrueck] at 06:13 21 January 2016
Summary To study the expression of tissue carcinoembryonic antigen (CEA) by immunoperoxidase staining in colorectal adenocarcinomas and its relation with preoperative serum CEA and clinicopathological parameters, we studied 85 unselected patients who underwent resective surgery for colorectal adenocarcinomas and were followed up for a mean of 66 mths. All tumors except 2 showed positive staining for CEA. The staining pattern was classifiable into 4 types: apical, cytoplasmic, basolateral and stromal, according to the predominant sites of staining. We found a significant positive correlation between tissue CEA staining pattern and preoperative plasma CEA. Plasma CEA levels were higher when tumor cell staining extended to basolateral regions of the cells and into stromal tissues rather than restricted to apical and cytoplasmic regions (p = 0.012). Furthermore,tissue CEA staining also correlated positively with vascular invasion by tumor cells (p =0.046),with basolateral and stromal types associated with more frequent vascular invasion than apical and cytoplasmic types. This was in contrast to the preoperative CEA which did not correlate with vascular permeation. We speculate that tissue CEA staining is useful in indicating possible vascular invasion even at early stage whereas vascular invasion by a larger tumor bulk or even tumor metastases may be necessary to produce an increased plasma CEA level that is detectable. On the other hand, preoperative plasma CEA had a strong, positive relationship with tumor stage and mortality (p <0.001 for both). Preoperative CEA was higher in tumors showing moderate and poor differentiation, although whereas tissue it did not reach statistical significance (p = 0.068), CEA staining had no correlation with tumor differention. Although tissue CEA staining patterns appear useful in explaining some pathophysiological events, it offers little additional information to plasma CEA in management of patients with colorectal carcinoma. Key words: Colorectal, adenocarcinoma, tissue and plasma CEA
Accepted 22 December, 1992
INTRODUCTION Evaluation of serial plasma carcinoembryonic antigen (CEA) levels in patients with colorectal cancers is one of the important parameters, when coupled with other clinical and pathological parameters, used to predict the outcome of surgically treated patients. 1 - 5 Serial CEA monitoring has been advocated as an indicator of colonic
cancer recurrence and second-look operation.' However, false negatives have been reported in poorly differentiated colorectal cancers6 whereas high plasma CEA levels can be seen in several other diseases unrelated to the primary colorectal tumor, including inflammatory diseases.' The purpose of this study is to characterize the expression of tissue CEA with immunohistochemical staining in colorectal tumors and to investigate whether it contributes additional information with respect to plasma CEA. MATERIALS AND METHODS Patients A cohort of 85 unselected patients with primary colorectal adenocarcinoma who underwent resective surgery in the University Department of Surgery during the period from 1983 t o 1984 were studied. They had been followed up for a minimum of 55 mths, a mean of 66 mths, or until death. They all had a preoperative plasma CEA measured. Eight patients were lost t o follow up. Thus the analysis used 85 patients for the preoperative data and 77 patients for the follow-up data. All the plasma CEA levels were done by radioimmunoassay. Histological examination The staging of tumors was determined according to Dukes' classification.' Stages A, B and C corresponded to Dukes' classification. Stage D was defined as tumor with distant metastasis. Tumor differentiation was divided into good, moderate and poor. Presence of blood vessel and lymphatic invasion was recorded. Immunoperoxidase staining procedure lmmunoperoxidase technique was performed on formalin-fixed paraffinembedded tissue using a slightly modified indirect method of Sternberger.' All incubation steps of the staining procedure were performed at 37°C. Sections were incubated with rabbit anti-human CEA in 1:250 dilution (from Dakopatts, Copenhagen) as the primary antibody for 30 mins. IgG fraction of anti-rabbit IgG antiserum produced in swine was used as the second antibody (in 1:50 dilution for 30 mins) to link the rabbit horseradish peroxidase anti-peroxidase complex (1 :100 dilution for 30 mins) t o the primary antibody. For the chromogen, solution containing 0.05% (weight/volume) diaminobenzidine tetrahydrochloride with 0.01 To (weight/volume) hydrogen peroxide was used and the incubation last for 10 mins. For controls, normal rabbit serum was applied instead of anti-CEA antiserum. Immunoperoxidase staining pattern based on CEA localization In every specimen, 2 most representative blocks of the tumor were examined for immunoperoxidase localization of CEA. The staining pattern was classified according to a slight modification of the system used by Hamada et a1.I' and was as follows: I . apical type: CEA was
Downloaded by [Universität Osnabrueck] at 06:13 21 January 2016
220
NG el ai,
Pathology (1993). 25, July
Fig. 1 Patterns of CEA tissue staining in colorectal adenocarcinomas. A. apical ( x 500), B. cytoplasmic ( x IOOO), C. basolateral ( x lOOO), and D. stromal ( x 125).
restricted predominantly to the apical border of the malignant glands (Fig. 1A). 11. cytoplasmic type: CEA was demonstrated in the cytoplasm as well as in the apical surfaces (Fig. 1B). 111. basolateral type: CEA was found on the basolateral surfaces with o r without staining in the cytoplasm (Fig. 1C). IV. stromal type: CEA was found over the entire surfaces and cytoplasm of the malignant cells, as well as diffusely in the stroma adjacent to the malignant glands (Fig. 1D). For the purpose of assessment, intact malignant glands without necrosis were chosen. The staining pattern is tabulated in Table 1. The extent of CEA staining was also examined and divided into focal (0-30% of tumor glands sampled), moderate (30-70%) and extensive ( > 70%). Statistics
Kendall’s Tau was used for all statistical analyses except for plasma CEA with mortality where Mann-Whitney test was applied.
TABLE1 Predominant CEA localization with immunoperoxidase method in colorectal carcinoma
Type of tissue
Distribution of CEA* staining in tumor glands
Tissue CEA Normal colonic glands showed staining mainly along the free borders and the stroma did not stain up except in cases with stromal (type IV) pattern. Of all the 85 cases of adenocarcinoma examined, all but 2 were positive for CEA. The 2 tumors which showed no tissue CEA staining were well and moderately differentiated adenocarcinomas respectively and both tumors were Dukes’ class C . The most common staining pattern was apical (n = 33), followed by cytoplasmic (n = 29), and basolateral (n = 19). Only 2 cases showed a stromal pattern of staining. As for the extent, moderate extent of CEA staining was the commonest (n = 521, with focal and extensive ones seen in 7 and 26 cases respectively.
Tissue CEA and plasma CEA
I . Apical
+
+
I V . Stromal ~
RESULTS Of the 77 patients followed up, 42 are alive and 35 are dead (including all causes), giving a crude survival of 55%.
~~
~~~
*CEA = carcinoembryonic antigen.
The preoperative plasma CEA had a median of 4.0 ng/mL (range = 1.0-180 ng/mL). Figure 2 shows the relationship between the immunoperoxidase pattern of tissue CEA and preoperative plasma CEA levels. There
CEA TISSUE STAINING I N COLORECTAL CANCER
221
TABLE 2 Tissue CEA staining and vascular permeation log Plasma
"1 3 i
No staining
0
-
0
.. ..
ttt
m
Apical
Cytoplasmic
___ None __
Present Absent Total
Tissue CEA staining (No. of cases)
1 1
2
29
33
19
p=0.046.
Y
Staining Pattern Basolateral
Stromal
Fig. 2 Correlation of preoperative plasma CEA with histological CEA staining patterns. (CEA values expressed as logs as distribution is highly skewed).
Downloaded by [Universität Osnabrueck] at 06:13 21 January 2016
Vascular permeation
0
0
m 0
Kendall's Tau Correlation n.85 1.0.192 P.0.012
+median
CEA Pre-op
was significant positive correlation between the staining pattern and the plasma CEA levels (p = 0.012). Higher preoperative plasma CEA levels were seen in the cases with stromal and basolateral patterns than in those with cytoplasmic and apical ones. The extent of tissue CEA staining appeared to increase with increased preoperative CEA levels, although the relationship did not reach statistical significance (p = 0.056).
Tissue CEA and vascular permeation Table 2 shows the relationship between the tissue CEA staining and vascular permeation. There was positive correlation between vascular permeation and the patterns of staining (p=O.O46). Most of cases with apical and cytoplasmic patterns (types I and 11) of tissue CEA staining did not have either vascular or lymphatic permeation whereas over half of those with basolateral and stromal patterns (types I11 and IV) had vascular or lymphatic permeation. In the 2 cases with stromal pattern of staining, both tumors showed vascular permeation. Tissue CEA, Dukes' classification, tumor differen tiation and mortality Of the 85 cases, 9 were of Dukes' stage A, 46 stage B, 17 stage C and 13 stage D. The 2 cases of stromal type of tissue CEA staining belonged to Dukes' stage D. As a whole, there was no significant correlation between the pattern and extent of tissue CEA staining and Dukes' stage (p=O.133 and p=0.101 respectively). Tumor differentiation and mortality bore no significant association with both pattern and extent of tissue CEA staining (Table 3). The 3 poorly differentiated tumors in this study showed apical (type I) CEA staining, whereas the 2 cases with stromal (type IV) staining were moderately differentiated. On the other hand, Dukes' staging was significantly related t o mortality (p <0.001) and all Dukes stage D patients died within the period of follow-up. Plasma CEA, stage of tumor, mortality, vascular permeation and tumor differentiation Preoperative plasma CEA levels showed a significant positive correlation with the stage of the tumor
(p < 0.001). Moreover, the preoperative plasma CEA was strongly and positively correlated with mortality (p < = 0.001). If Dukes' D cases were removed from the analysis, the preoperative plasma CEA was still correlated with mortality (p = 0.013). Preoperative plasma CEA, however, did not show a statistically significant correlation with vascular permeation. Plasma CEA tended to be higher in tumors showing moderate and poor differentiation, but it did not reach statistical significance (p = 0.068).
DISCUSSION Our study showed that tissue CEA staining in colorectal adenocarcinomas could be classified into separate patterns using immunohistochemical methods. We distinguished between a cytoplasmic and a basolateral type of staining in the tumor glands and they had significantly different correlation with some of the parameters studied. Cytoplasmic staining pattern was associated with a lower preoperative plasma CEA and a lower incidence of vascular invasion, whereas the reverse applied to basolateral pattern. We confirmed that the pattern of tissue CEA staining had a good correlation with plasma CEA in colorectal cancer patients. lo Higher levels of preoperative plasma CEA were present in basolateral and stromal distributions of tissue CEA, whereas lower levels of plasma CEA were seen in apical and cytoplasmic distributions of tissue CEA. It has been speculated that a change in the surface properties of the tumor cells leads to failure to establish or maintain the polar expression of normal cell surface glycoprotein CEA, and therefore appearance of CEA in the stroma." This may play a significant role in the elevation of plasma CEA levels. In the study by Tabuchi et al.," they observed that elevation of CEA levels in draining and peripheral blood was most highly correlated with venous invasion and postulated that CEA, produced by the cancer cells in the invaded veins, might be drained mainly by the hematogenous portal system. TABLE3 Tissue CEA staining and tumor differentiation Tumor differentiation Type of tissue CEA staining No staining Apical Cytoplasmic Basolateral Stromal ~
p = 0.47
Moderate (No. of cases) -
7 8
2 -
2 23
-
21
-
17 2
-
3
Downloaded by [Universität Osnabrueck] at 06:13 21 January 2016
222
NG et ai.
From these results, we speculate that a change in the cell property of the tumor cells might on one hand lead to the release of CEA in the stroma and on the other hand contribute to the increased tendency to invade blood vessels. We showed that patterns of tissue CEA staining were correlated with vascular invasion by tumor. Thus, higher incidence of vascular invasion was seen in basolateral and stromal distributions of tissue CEA than in apical and cytoplasmic ones. However, we could not demonstrate a significant relation between plasma CEA and vascular invasion. Such a discrepancy between plasma CEA and tissue CEA staining in relation to vascular invasion led us to speculate that a larger tumor bulk might be necessary to produce a plasma CEA level to be detectable, whereas tissue CEA staining, more related to pathophysiological changes leading to vascular invasion in cellular and tissue levels, might be able to indicate possible vascular invasion even at early stage of vascular invasion, where the tumor cells in the blood are not plentiful enough to produce a detectable plasma CEA. Stromal distribution in our study was not due'to necrotic cell debris because the assessment of tissue CEA staining was done on intact malignant glands without necrosis. It was also not due to nonspecific background staining of the collagenous stroma because the stroma in the non-neoplastic colonic tissue was completely negative for CEA. We could not demonstrate any correlation between tumor differentiation and tissue CEA staining, and between tumor differentiation and plasma CEA, this being at variance with the findings of other studies.",'* Tissue CEA staining was also not associated with Dukes' staging and mortality. On the other hand, preoperative plasma CEA strongly correlated with staging of the tumors and also with mortality. This was due to the powerful influence of disseminated disease causing raised plasma CEA levels and death. However, even when Dukes' D cases were removed from the analysis, the preoperative plasma CEA still correlated with mortality. 'Thus the findings support the view that, although plasma CEA is not sensitive enough for the detection of colorectal cancers in asymptomatic individuals as a screening test, l 3 it is useful in monitoring and detection of recurrence. We conclude that although tissue CEA staining patterns appears useful in explaining pathophysiological features, it offers little additional information than plasma CEA in management of patients with colorectal carcinoma. Recently, the usefulness of other glycoproteins or antigens/tumor markers has been studied. They include epithelial membrane antigen, carbohydrate antigen 19-9,
Pathology (1993), 25, July
and newer tumor markers such as ornithine decarboxylase and polyamines. l 3 . l 4 However, studies have shown that the former 2 are either less sensitive or less specific than CEA, whereas the latter 2, although potential future tumor markers, need further studies to establish their sensitivity and specificity in colorectal cancer^."^'^ Address f o r correspondence: I.N., Department of Pathology, University of Hong Kong, Queen Mary Hospital, Hong Kong.
References 1. Minton J, Martin EW. The use of serial CEA determinations to predict recurrence of colonic cancer and when to do a second-look operation. Cancer 1978; 42: 1422-7. 2. Evans JT, Mittelman A, Chu M, Holyoke ED. Pre- and postoperative uses of CEA. Cancer 1978; 42: 1419-21. 3. Midiri G, Amanti C, Consorti F et al. Usefulness of preoperative CEA levels in the assessment of colorectal cancer patients' stage. J Surg Oncol 1983; 22: 257-60. 4. Goligher JC. What can be done to improve the results in colorectal cancer? Scand J Gastroenterol Suppl 1988; 149: 190-9. 5 . Roberts PJ. Tumour markers in colorectal cancer. Scand J Gastroenterol Suppl 1988; 149: 50-8.
6. G o s h R, O'Brien MJ, Steele G et al. Correlation of plasma CEA and CEA tissue staining in poorly differentiated colorectal cancer. Am J Med 1981; 71: 246-53. 7. Midiri G, Amanti C, Benedetti M et al. CEA tissue staining in colorectal cancer patients. A way to improve the usefulness of serial serum CEA evaluation. Cancer 1985; 55: 2624-9.
8. Dukes CE. The classification of cancer of the rectum. J Pathol Bacteriol 1932; 35: 323-32. 9. Sternberger LA, Hardy PH, Cuculis JJ, Meyer HG. The unlabelled antibody enzyme method of immunohistochemistry: Preparation and properties of soluble antigen-antibody complex (horseradish peroxidase-antihorseradish peroxidase) and its use in identification of spirochetes. J Histochem Cytochem 1970; 18: 315-33. 10. Hamada Y, Yamamura M, Hioki K, Yamamoto M, Nagura H, Watanabe K . Immunohistochemical study of carcinoembryonic antigen in patients with colorectal cancer. Correlation with plasma carcinoembryonic antigen levels. Cancer 1985; 55: 136-41. 11. Tabuchi Y, Deguchi H, Saitoh Y . Carcinoembryonic antigen and carbohydrate antigen 19-9 levels of peripheral and draining venous blood in colorectal cancer patients. Cancer 1988; 62: 1605-13. 12. Zamcheck N, Doos WG, Prudente R, Lurie BB, Gottlieb LS. Prognostic factors in colon carcinoma: correlation of serum carcinoembryonic antigen level and tumor histopathology. Hum Pathol 1975; 6: 31-45. 13. Luk GD, Desai TK, Conteas CN, Moshier JA, Silverman AL. Biochemical markers in colorectal cancer: diagnostic and therapeutic implications. Gastroenterol Clin North Am 1988; 17(4): 931-40. 14. Davidson BR, Yiu CY, Styles J, Ormerod M, Clark CG, Dean C . A comparison of carcinoembryonic antigen (CEA) and epithelial membrane antigen (EMA) in human colorectal cancer. Int J Cancer SUPPI 1988; 3: 56-60.
ISSN: 0031-3025 (Print) 1465-3931 (Online) Journal homepage: http://www.tandfonline.com/loi/ipat20
CEA Tissue Staining in Colorectal Cancer Patients — Correlation with Plasma CEA, Histology and Staging Irene O. L. Ng, Joana Ho, C. J. Pritchett, Eric Y. T. Chan & Faith C. S. Ho To cite this article: Irene O. L. Ng, Joana Ho, C. J. Pritchett, Eric Y. T. Chan & Faith C. S. Ho (1993) CEA Tissue Staining in Colorectal Cancer Patients — Correlation with Plasma CEA, Histology and Staging, Pathology, 25:3, 219-222 To link to this article: http://dx.doi.org/10.3109/00313029309066575
Published online: 06 Jul 2009.
Submit your article to this journal
Article views: 6
View related articles
Citing articles: 1 View citing articles
Full Terms & Conditions of access and use can be found at http://www.tandfonline.com/action/journalInformation?journalCode=ipat20 Download by: [Universität Osnabrueck]
Date: 21 January 2016, At: 06:13
Pathology (1993), 25, pp. 219-222
CEA TISSUE STAINING IN COLORECTAL CANCER PATIENTS CORRELATION WITH PLASMA CEA, HISTOLOGY AND STAGING IRENE0 . L. NG, JOANA Ho, C. J . PRITCHETT,* ERICY. T. CHANAND FAITHC . S . Ho Departments of Pathology and *Surgery, University of Hong Kong, Queen Mary Hospital, Hong Kong
Downloaded by [Universität Osnabrueck] at 06:13 21 January 2016
Summary To study the expression of tissue carcinoembryonic antigen (CEA) by immunoperoxidase staining in colorectal adenocarcinomas and its relation with preoperative serum CEA and clinicopathological parameters, we studied 85 unselected patients who underwent resective surgery for colorectal adenocarcinomas and were followed up for a mean of 66 mths. All tumors except 2 showed positive staining for CEA. The staining pattern was classifiable into 4 types: apical, cytoplasmic, basolateral and stromal, according to the predominant sites of staining. We found a significant positive correlation between tissue CEA staining pattern and preoperative plasma CEA. Plasma CEA levels were higher when tumor cell staining extended to basolateral regions of the cells and into stromal tissues rather than restricted to apical and cytoplasmic regions (p = 0.012). Furthermore,tissue CEA staining also correlated positively with vascular invasion by tumor cells (p =0.046),with basolateral and stromal types associated with more frequent vascular invasion than apical and cytoplasmic types. This was in contrast to the preoperative CEA which did not correlate with vascular permeation. We speculate that tissue CEA staining is useful in indicating possible vascular invasion even at early stage whereas vascular invasion by a larger tumor bulk or even tumor metastases may be necessary to produce an increased plasma CEA level that is detectable. On the other hand, preoperative plasma CEA had a strong, positive relationship with tumor stage and mortality (p <0.001 for both). Preoperative CEA was higher in tumors showing moderate and poor differentiation, although whereas tissue it did not reach statistical significance (p = 0.068), CEA staining had no correlation with tumor differention. Although tissue CEA staining patterns appear useful in explaining some pathophysiological events, it offers little additional information to plasma CEA in management of patients with colorectal carcinoma. Key words: Colorectal, adenocarcinoma, tissue and plasma CEA
Accepted 22 December, 1992
INTRODUCTION Evaluation of serial plasma carcinoembryonic antigen (CEA) levels in patients with colorectal cancers is one of the important parameters, when coupled with other clinical and pathological parameters, used to predict the outcome of surgically treated patients. 1 - 5 Serial CEA monitoring has been advocated as an indicator of colonic
cancer recurrence and second-look operation.' However, false negatives have been reported in poorly differentiated colorectal cancers6 whereas high plasma CEA levels can be seen in several other diseases unrelated to the primary colorectal tumor, including inflammatory diseases.' The purpose of this study is to characterize the expression of tissue CEA with immunohistochemical staining in colorectal tumors and to investigate whether it contributes additional information with respect to plasma CEA. MATERIALS AND METHODS Patients A cohort of 85 unselected patients with primary colorectal adenocarcinoma who underwent resective surgery in the University Department of Surgery during the period from 1983 t o 1984 were studied. They had been followed up for a minimum of 55 mths, a mean of 66 mths, or until death. They all had a preoperative plasma CEA measured. Eight patients were lost t o follow up. Thus the analysis used 85 patients for the preoperative data and 77 patients for the follow-up data. All the plasma CEA levels were done by radioimmunoassay. Histological examination The staging of tumors was determined according to Dukes' classification.' Stages A, B and C corresponded to Dukes' classification. Stage D was defined as tumor with distant metastasis. Tumor differentiation was divided into good, moderate and poor. Presence of blood vessel and lymphatic invasion was recorded. Immunoperoxidase staining procedure lmmunoperoxidase technique was performed on formalin-fixed paraffinembedded tissue using a slightly modified indirect method of Sternberger.' All incubation steps of the staining procedure were performed at 37°C. Sections were incubated with rabbit anti-human CEA in 1:250 dilution (from Dakopatts, Copenhagen) as the primary antibody for 30 mins. IgG fraction of anti-rabbit IgG antiserum produced in swine was used as the second antibody (in 1:50 dilution for 30 mins) to link the rabbit horseradish peroxidase anti-peroxidase complex (1 :100 dilution for 30 mins) t o the primary antibody. For the chromogen, solution containing 0.05% (weight/volume) diaminobenzidine tetrahydrochloride with 0.01 To (weight/volume) hydrogen peroxide was used and the incubation last for 10 mins. For controls, normal rabbit serum was applied instead of anti-CEA antiserum. Immunoperoxidase staining pattern based on CEA localization In every specimen, 2 most representative blocks of the tumor were examined for immunoperoxidase localization of CEA. The staining pattern was classified according to a slight modification of the system used by Hamada et a1.I' and was as follows: I . apical type: CEA was
Downloaded by [Universität Osnabrueck] at 06:13 21 January 2016
220
NG el ai,
Pathology (1993). 25, July
Fig. 1 Patterns of CEA tissue staining in colorectal adenocarcinomas. A. apical ( x 500), B. cytoplasmic ( x IOOO), C. basolateral ( x lOOO), and D. stromal ( x 125).
restricted predominantly to the apical border of the malignant glands (Fig. 1A). 11. cytoplasmic type: CEA was demonstrated in the cytoplasm as well as in the apical surfaces (Fig. 1B). 111. basolateral type: CEA was found on the basolateral surfaces with o r without staining in the cytoplasm (Fig. 1C). IV. stromal type: CEA was found over the entire surfaces and cytoplasm of the malignant cells, as well as diffusely in the stroma adjacent to the malignant glands (Fig. 1D). For the purpose of assessment, intact malignant glands without necrosis were chosen. The staining pattern is tabulated in Table 1. The extent of CEA staining was also examined and divided into focal (0-30% of tumor glands sampled), moderate (30-70%) and extensive ( > 70%). Statistics
Kendall’s Tau was used for all statistical analyses except for plasma CEA with mortality where Mann-Whitney test was applied.
TABLE1 Predominant CEA localization with immunoperoxidase method in colorectal carcinoma
Type of tissue
Distribution of CEA* staining in tumor glands
Tissue CEA Normal colonic glands showed staining mainly along the free borders and the stroma did not stain up except in cases with stromal (type IV) pattern. Of all the 85 cases of adenocarcinoma examined, all but 2 were positive for CEA. The 2 tumors which showed no tissue CEA staining were well and moderately differentiated adenocarcinomas respectively and both tumors were Dukes’ class C . The most common staining pattern was apical (n = 33), followed by cytoplasmic (n = 29), and basolateral (n = 19). Only 2 cases showed a stromal pattern of staining. As for the extent, moderate extent of CEA staining was the commonest (n = 521, with focal and extensive ones seen in 7 and 26 cases respectively.
Tissue CEA and plasma CEA
I . Apical
+
+
I V . Stromal ~
RESULTS Of the 77 patients followed up, 42 are alive and 35 are dead (including all causes), giving a crude survival of 55%.
~~
~~~
*CEA = carcinoembryonic antigen.
The preoperative plasma CEA had a median of 4.0 ng/mL (range = 1.0-180 ng/mL). Figure 2 shows the relationship between the immunoperoxidase pattern of tissue CEA and preoperative plasma CEA levels. There
CEA TISSUE STAINING I N COLORECTAL CANCER
221
TABLE 2 Tissue CEA staining and vascular permeation log Plasma
"1 3 i
No staining
0
-
0
.. ..
ttt
m
Apical
Cytoplasmic
___ None __
Present Absent Total
Tissue CEA staining (No. of cases)
1 1
2
29
33
19
p=0.046.
Y
Staining Pattern Basolateral
Stromal
Fig. 2 Correlation of preoperative plasma CEA with histological CEA staining patterns. (CEA values expressed as logs as distribution is highly skewed).
Downloaded by [Universität Osnabrueck] at 06:13 21 January 2016
Vascular permeation
0
0
m 0
Kendall's Tau Correlation n.85 1.0.192 P.0.012
+median
CEA Pre-op
was significant positive correlation between the staining pattern and the plasma CEA levels (p = 0.012). Higher preoperative plasma CEA levels were seen in the cases with stromal and basolateral patterns than in those with cytoplasmic and apical ones. The extent of tissue CEA staining appeared to increase with increased preoperative CEA levels, although the relationship did not reach statistical significance (p = 0.056).
Tissue CEA and vascular permeation Table 2 shows the relationship between the tissue CEA staining and vascular permeation. There was positive correlation between vascular permeation and the patterns of staining (p=O.O46). Most of cases with apical and cytoplasmic patterns (types I and 11) of tissue CEA staining did not have either vascular or lymphatic permeation whereas over half of those with basolateral and stromal patterns (types I11 and IV) had vascular or lymphatic permeation. In the 2 cases with stromal pattern of staining, both tumors showed vascular permeation. Tissue CEA, Dukes' classification, tumor differen tiation and mortality Of the 85 cases, 9 were of Dukes' stage A, 46 stage B, 17 stage C and 13 stage D. The 2 cases of stromal type of tissue CEA staining belonged to Dukes' stage D. As a whole, there was no significant correlation between the pattern and extent of tissue CEA staining and Dukes' stage (p=O.133 and p=0.101 respectively). Tumor differentiation and mortality bore no significant association with both pattern and extent of tissue CEA staining (Table 3). The 3 poorly differentiated tumors in this study showed apical (type I) CEA staining, whereas the 2 cases with stromal (type IV) staining were moderately differentiated. On the other hand, Dukes' staging was significantly related t o mortality (p <0.001) and all Dukes stage D patients died within the period of follow-up. Plasma CEA, stage of tumor, mortality, vascular permeation and tumor differentiation Preoperative plasma CEA levels showed a significant positive correlation with the stage of the tumor
(p < 0.001). Moreover, the preoperative plasma CEA was strongly and positively correlated with mortality (p < = 0.001). If Dukes' D cases were removed from the analysis, the preoperative plasma CEA was still correlated with mortality (p = 0.013). Preoperative plasma CEA, however, did not show a statistically significant correlation with vascular permeation. Plasma CEA tended to be higher in tumors showing moderate and poor differentiation, but it did not reach statistical significance (p = 0.068).
DISCUSSION Our study showed that tissue CEA staining in colorectal adenocarcinomas could be classified into separate patterns using immunohistochemical methods. We distinguished between a cytoplasmic and a basolateral type of staining in the tumor glands and they had significantly different correlation with some of the parameters studied. Cytoplasmic staining pattern was associated with a lower preoperative plasma CEA and a lower incidence of vascular invasion, whereas the reverse applied to basolateral pattern. We confirmed that the pattern of tissue CEA staining had a good correlation with plasma CEA in colorectal cancer patients. lo Higher levels of preoperative plasma CEA were present in basolateral and stromal distributions of tissue CEA, whereas lower levels of plasma CEA were seen in apical and cytoplasmic distributions of tissue CEA. It has been speculated that a change in the surface properties of the tumor cells leads to failure to establish or maintain the polar expression of normal cell surface glycoprotein CEA, and therefore appearance of CEA in the stroma." This may play a significant role in the elevation of plasma CEA levels. In the study by Tabuchi et al.," they observed that elevation of CEA levels in draining and peripheral blood was most highly correlated with venous invasion and postulated that CEA, produced by the cancer cells in the invaded veins, might be drained mainly by the hematogenous portal system. TABLE3 Tissue CEA staining and tumor differentiation Tumor differentiation Type of tissue CEA staining No staining Apical Cytoplasmic Basolateral Stromal ~
p = 0.47
Moderate (No. of cases) -
7 8
2 -
2 23
-
21
-
17 2
-
3
Downloaded by [Universität Osnabrueck] at 06:13 21 January 2016
222
NG et ai.
From these results, we speculate that a change in the cell property of the tumor cells might on one hand lead to the release of CEA in the stroma and on the other hand contribute to the increased tendency to invade blood vessels. We showed that patterns of tissue CEA staining were correlated with vascular invasion by tumor. Thus, higher incidence of vascular invasion was seen in basolateral and stromal distributions of tissue CEA than in apical and cytoplasmic ones. However, we could not demonstrate a significant relation between plasma CEA and vascular invasion. Such a discrepancy between plasma CEA and tissue CEA staining in relation to vascular invasion led us to speculate that a larger tumor bulk might be necessary to produce a plasma CEA level to be detectable, whereas tissue CEA staining, more related to pathophysiological changes leading to vascular invasion in cellular and tissue levels, might be able to indicate possible vascular invasion even at early stage of vascular invasion, where the tumor cells in the blood are not plentiful enough to produce a detectable plasma CEA. Stromal distribution in our study was not due'to necrotic cell debris because the assessment of tissue CEA staining was done on intact malignant glands without necrosis. It was also not due to nonspecific background staining of the collagenous stroma because the stroma in the non-neoplastic colonic tissue was completely negative for CEA. We could not demonstrate any correlation between tumor differentiation and tissue CEA staining, and between tumor differentiation and plasma CEA, this being at variance with the findings of other studies.",'* Tissue CEA staining was also not associated with Dukes' staging and mortality. On the other hand, preoperative plasma CEA strongly correlated with staging of the tumors and also with mortality. This was due to the powerful influence of disseminated disease causing raised plasma CEA levels and death. However, even when Dukes' D cases were removed from the analysis, the preoperative plasma CEA still correlated with mortality. 'Thus the findings support the view that, although plasma CEA is not sensitive enough for the detection of colorectal cancers in asymptomatic individuals as a screening test, l 3 it is useful in monitoring and detection of recurrence. We conclude that although tissue CEA staining patterns appears useful in explaining pathophysiological features, it offers little additional information than plasma CEA in management of patients with colorectal carcinoma. Recently, the usefulness of other glycoproteins or antigens/tumor markers has been studied. They include epithelial membrane antigen, carbohydrate antigen 19-9,
Pathology (1993), 25, July
and newer tumor markers such as ornithine decarboxylase and polyamines. l 3 . l 4 However, studies have shown that the former 2 are either less sensitive or less specific than CEA, whereas the latter 2, although potential future tumor markers, need further studies to establish their sensitivity and specificity in colorectal cancer^."^'^ Address f o r correspondence: I.N., Department of Pathology, University of Hong Kong, Queen Mary Hospital, Hong Kong.
References 1. Minton J, Martin EW. The use of serial CEA determinations to predict recurrence of colonic cancer and when to do a second-look operation. Cancer 1978; 42: 1422-7. 2. Evans JT, Mittelman A, Chu M, Holyoke ED. Pre- and postoperative uses of CEA. Cancer 1978; 42: 1419-21. 3. Midiri G, Amanti C, Consorti F et al. Usefulness of preoperative CEA levels in the assessment of colorectal cancer patients' stage. J Surg Oncol 1983; 22: 257-60. 4. Goligher JC. What can be done to improve the results in colorectal cancer? Scand J Gastroenterol Suppl 1988; 149: 190-9. 5 . Roberts PJ. Tumour markers in colorectal cancer. Scand J Gastroenterol Suppl 1988; 149: 50-8.
6. G o s h R, O'Brien MJ, Steele G et al. Correlation of plasma CEA and CEA tissue staining in poorly differentiated colorectal cancer. Am J Med 1981; 71: 246-53. 7. Midiri G, Amanti C, Benedetti M et al. CEA tissue staining in colorectal cancer patients. A way to improve the usefulness of serial serum CEA evaluation. Cancer 1985; 55: 2624-9.
8. Dukes CE. The classification of cancer of the rectum. J Pathol Bacteriol 1932; 35: 323-32. 9. Sternberger LA, Hardy PH, Cuculis JJ, Meyer HG. The unlabelled antibody enzyme method of immunohistochemistry: Preparation and properties of soluble antigen-antibody complex (horseradish peroxidase-antihorseradish peroxidase) and its use in identification of spirochetes. J Histochem Cytochem 1970; 18: 315-33. 10. Hamada Y, Yamamura M, Hioki K, Yamamoto M, Nagura H, Watanabe K . Immunohistochemical study of carcinoembryonic antigen in patients with colorectal cancer. Correlation with plasma carcinoembryonic antigen levels. Cancer 1985; 55: 136-41. 11. Tabuchi Y, Deguchi H, Saitoh Y . Carcinoembryonic antigen and carbohydrate antigen 19-9 levels of peripheral and draining venous blood in colorectal cancer patients. Cancer 1988; 62: 1605-13. 12. Zamcheck N, Doos WG, Prudente R, Lurie BB, Gottlieb LS. Prognostic factors in colon carcinoma: correlation of serum carcinoembryonic antigen level and tumor histopathology. Hum Pathol 1975; 6: 31-45. 13. Luk GD, Desai TK, Conteas CN, Moshier JA, Silverman AL. Biochemical markers in colorectal cancer: diagnostic and therapeutic implications. Gastroenterol Clin North Am 1988; 17(4): 931-40. 14. Davidson BR, Yiu CY, Styles J, Ormerod M, Clark CG, Dean C . A comparison of carcinoembryonic antigen (CEA) and epithelial membrane antigen (EMA) in human colorectal cancer. Int J Cancer SUPPI 1988; 3: 56-60.