Review Br. J. Surg. 1992, Vol. 79, July. 608-61 3
G. T. Deans, T. G. Parks, B. J. Rowlands and R . A. J. Spence Queen's University Department of Surgery, Belfast, UK Correspondence to: Mr R. A. J. Spence, University
Department of Surgery, Belfast City Hospital, Belfast BT9 7AB, UK
Prognostic factors in colorectal cancer The prognostic power of the extent of tumour invasion is indisputable; Dukes' classijication has repeatedly been proven to be strongly correlated with patient suruiual. Modijications have led only to confusion, resulting in caution being required in the classiJication of patients with Dukes' A tumours. In the UK, the American tumour node metastasis and Australian clinicopathological systems are frequently considered too complex for routine clinical use. Meanwhile, Jass 's classiJication may be complicated by observer variation between pathologists, and recent evidence suggests that it ofers no advantage over that of Dukes. All the conventional staging systems also fail to take the skill of the surgeon into account when determining outcome. Attempts at quantifying tumour structure have not heralded the expected major advance. For instance, the expense and uncertain prognostic value of tumour D N A content assessed by flow cytometry are likely to restrict widespread use of this technique. It may soon be possible, however, to prouide optimum treatment for patients based on individual tumour doubling times. Classifications using knowledge of how a small number of cells in the tumour have the ability to invade locally, enter blood vessels and metastasize would also provide important prognostic information on which treatment could be based. Until then, the ease of use and high prognostic power of Dukes' classiJication ensure that, after 60 years, it is still the 'gold standard' against which all other prognostic classiJications in colorectal cancer should be assessed.
Dukes' classification In 1927 Lockhart-Mummery3, noting a relationship between prognosis and extension of rectal tumours, suggested a clinical classification that was later modified to include pathological data4. Dukes subsequently changed the criteria, developing a system for the pathological classification of rectal carcinoma'. Three stages were described: A, growth limited to the rectal wall; B, extension of growth to extrarectal tissues but no metastases in regional lymph nodes; and C, metastases in regional lymph nodes. Class C was later subdivided into C,, only regional lymph nodes involved, and C,, nodal spread to the level of the point of ligature of the blood vessels6. A class ' D has crept into popular usage to indicate tumours with
distant metastases. The classification was soon applied to colonic as well as to rectal turnours'. Two modifications have complicated this classification. In 1949, Kirklin et al.' subdivided Dukes' system into: A, lesion limited to the mucosa; B,, lesion extends to, but does not penetrate, the muscularis propria; B,, lesion penetrates through the muscularis propria; and C, lesion of either B, or B, with involvement of lymph nodes. Class C of this system was later modified by Astler and Collerg into C1, lesion limited to bowel wall with positive nodes, and C,, lesion through all layers of bowel with positive nodes. These changes have created more confusion than clarity, making it difficult to compare reported series". A modified B, lesion corresponds to Dukes' A. Tumours classified as A, as designated by Kirklin et al., and by Astler and Coller, are rare; of the 981 tumours studied by the authors of the two classifications, only nine (fewer than 1 per cent) met their criteria for stage A. In contrast, 15 per cent of tumours in the series reported by Dukes and Bussey' were classified as Dukes' A. The reported frequency of Dukes' A tumours varies from 6 per to 25 per and even 35 per cent". Although differing referral patterns or detection of earlier lesions by colonoscopy'g may be contributory, the various interpretations of what constitutes a Dukes' A lesion account for many of these differences. The Cl and C, categories proposed by Astler and Collerg are different from those described by Dukes; another source of confusion results from the use of the term 'muscularis propria'. In the colon and rectum, the muscularis propria consists of the longitudinal and circular layers of smooth muscle in the intestinal wall. Although Astler and Coller used this in their staging system, the term has subsequently been misinterpreted to mean muscularis mucosa. This has resulted in both B, and
608
0007-1323/92/07060846
Colorectal carcinoma is the second commonest cause of death from cancer in western countries, accounting for 14 and 16 per cent of cancer deaths in men and women, respectively'. Approximately one person in five in England and Wales now dies from cancer, and one in eight of these will die from tumours of the large bowel2. A clinician encountering a patient with colorectal cancer needs to know the expected outcome to plan appropriate therapy. Over the past 60 years, claims have been made of numerous variables being related to survival. However, the application of many of these in a clinical context has serious drawbacks, in that they are complex, or poorly reproducible, or have not been subjected to adequate statistical analysis. This review discusses the relationship to survival of clinical and pathological variables, and highlights the disadvantages of the principal staging systems that are currently in clinical use.
Tumour staging
'
0 1992 Butterworth-Heinemann
Ltd
Prognostic factors in colorectal cancer: G. T. Deans et al.
B, of the modified Dukes’ systems corresponding to class A of Dukes’ original scheme. Indeed, Fisher et al. found that subdividing patients according to Astler and Coller A and B, did not improve on the predictability of survival indicated by Dukes’ A alone2’. Irrespective of these problems, there are deficiencies in all the Dukes classifications.Analysis based on a few tissue samples may miss breakthrough of the muscularis propria, such that some patients who die quite soon are misclassified into the good prognosis Dukes’ A group. In a series of 2037 rectal cancers, Dukes reported a crude 5-year survival rate of 81.2 per cent for patients with class A tumours, instead of the near 100 per cent expected if the classification were perfect”. The histological grade of the tumour and the number of involved nodes, which are both known to be related to s u r v i ~ a I ~are ~ - not ~ ~ ,considered in Dukes’ classification. The thoroughness of the pathologist in examining for lymph nodes is also important. Blenkinsopp et aLZ4,reviewing pathological assessment of the depth of tumour penetration in 2046 patients from several institutions, noted that variation between observers ranged from 5 to 30 per cent ( P < 0.001).Differences were also noted between centres in the mean number of lymph nodes harvested per specimen ( P < 0.001) and the proportion of cases in which no nodes were identified. Because microscopic appearances are rarely difficult to interpret, these differences are most likely to be caused by errors in the macroscopic assessment of the specimen. Such results suggest that caution is necessary when comparing the proportions of cases in the various Dukes’ classes between institutions. Despite these problems, Dukes’ classification remains an extremely powerful prognostic tool. Dukes and Bussey reported 5-year survival rates of 81.2, 64 and 27.4 per cent for patients with A, B and C tumours, respectively, figures comparable to those of most modern series. Studies employing univariate analysis’2~zs-29have repeatedly emphasized the prognostic importance of tumour staging in colorectal cancer. Similarly, major series employing multivariate analysis have identified tumour stage as the primary determinant of prognosis in colorectal cancer15,17,2z,30-33.The only improvement required on Dukes’ assessment of the degree of tumour penetration may be the use of Astler and Coller stages C, and C,. Multivariate analysis has revealed this to be superior to a single Dukes’ C class, and to be independent of the number and site of nodal metastases and the degree of tumour differentiation,’. In summary, the deficiency of Dukes’ classification is that it fails to take clinical factors and the number of involved nodes into account. Modifications have resulted only in confusion over the interpretation of the various classes, particularly A. In spite of these problems, the prognostic significance and comparative ease of Dukes’ classification ensure that it remains the ‘gold standard’ against which all other prognostic variables in colorectal cancer should be tested.
’
Clinicopathological staging The turnour node metastasis system In an attempt to overcome the problems of Dukes’ classification a clinicopathological system was proposed34. This consisted of grading the extent of the primary tumour (T), the condition of regional lymph nodes ( N ) and the presence or absence of distant metastases (M). The Union Internacional Contra la Cancrum, however, agreed that the extent of colonic carcinoma could not be fully assessed clinically at the time of surgery. The American Joint Committee on Cancer therefore developed a series of prefixes for the TNM classification to relate to the extent of the disease at given sites and time^^'-^'. These are: cTNM, clinical diagnostic staging; sTNM, surgical evaluative staging; pTNM, postsurgical pathological staging; rTNM, retreatment staging (e.g. at second laparotomy ); and aTNM, autopsy staging. This American TNM classification was considered by many
Br. J. Surg., Vol. 79, No. 7, July 1992
to be too complex for routine use. In Australia a separate clinicopathological system was developed. This employed an A, B, C, D grading but combined clinical information, curative or palliative operation status, and a pathological subdivision of Dukes’ classes before a final grade was ascribed40A6. In Spain a further system, based on a subclassification of Dukes’ ‘D’ tumours, was developed4’. This international variation in the surgical, clinical and pathological classifications was addressed in 1987 with the adoption of identical criteria between c ~ u n t r i e s;~guidelines ~ . ~ ~ for an internationally agreed staging system were p r ~ p o s e d ~ ’ , ~Despite ’. attempts to produce a standardized clinicopathological classifications2, the TNM system is complex relative to Dukes’ classification. For instance, subdividing patients into T,NoMo and T,NoMo does not significantly improve the prediction of survival over a simple Dukes’ A classification2’. Consequently, the TNM system has not become widely accepted in the UK. Other clinicopathological variables In an attempt to improve on both Dukes’ and the TNM classifications, several studies have assessed the prognostic significance of numerous clinical and pathological variables. These include tumour size53, lymphatic i n ~ a s i o n ’ ~ - ~ ~ , mucinous tumourss7, laminin”, involvement of the lateral resection margin30~59.60, silver-binding nucleolar organizing region^^'-^^, lymphocytic and vascular invasion in the individual Dukes’ classes64, biochemical markers6’, bloodgroup antigens66p69,carcinoembryonic antigen70-72 and CA 19-973-77.Despite suggestions that some of these variables may be related to survival on univariate or multivariate analysis, none consistently improves on the prognostic value of Dukes’ classification in a regression analysis model. Three clinicopathological variables, however, are often related to survival and warrant further discussion. These are patient age, histological grade and venous invasion. Patient age. The most important clinical variable appears to be patient age. Whether older or younger patients have the poorer prognosis is, however, unclear. Several studies have reported that increasing age is associated with a poorer prognosisl 5,17,78-80 . Th’is appears to be independent of the fact that older patients are more likely to die from their operation” or concomitant diseases2than are younger patients. Nevertheless, recent reports have suggested that age itself should not be a determinant when considering operation in patients with colorectal ~ancer’~-’~. In contrast, some authors have noted a particularly poor prognosis in those aged
G. T. Deans, T. G. Parks, B. J. Rowlands and R . A. J. Spence Queen's University Department of Surgery, Belfast, UK Correspondence to: Mr R. A. J. Spence, University
Department of Surgery, Belfast City Hospital, Belfast BT9 7AB, UK
Prognostic factors in colorectal cancer The prognostic power of the extent of tumour invasion is indisputable; Dukes' classijication has repeatedly been proven to be strongly correlated with patient suruiual. Modijications have led only to confusion, resulting in caution being required in the classiJication of patients with Dukes' A tumours. In the UK, the American tumour node metastasis and Australian clinicopathological systems are frequently considered too complex for routine clinical use. Meanwhile, Jass 's classiJication may be complicated by observer variation between pathologists, and recent evidence suggests that it ofers no advantage over that of Dukes. All the conventional staging systems also fail to take the skill of the surgeon into account when determining outcome. Attempts at quantifying tumour structure have not heralded the expected major advance. For instance, the expense and uncertain prognostic value of tumour D N A content assessed by flow cytometry are likely to restrict widespread use of this technique. It may soon be possible, however, to prouide optimum treatment for patients based on individual tumour doubling times. Classifications using knowledge of how a small number of cells in the tumour have the ability to invade locally, enter blood vessels and metastasize would also provide important prognostic information on which treatment could be based. Until then, the ease of use and high prognostic power of Dukes' classiJication ensure that, after 60 years, it is still the 'gold standard' against which all other prognostic classiJications in colorectal cancer should be assessed.
Dukes' classification In 1927 Lockhart-Mummery3, noting a relationship between prognosis and extension of rectal tumours, suggested a clinical classification that was later modified to include pathological data4. Dukes subsequently changed the criteria, developing a system for the pathological classification of rectal carcinoma'. Three stages were described: A, growth limited to the rectal wall; B, extension of growth to extrarectal tissues but no metastases in regional lymph nodes; and C, metastases in regional lymph nodes. Class C was later subdivided into C,, only regional lymph nodes involved, and C,, nodal spread to the level of the point of ligature of the blood vessels6. A class ' D has crept into popular usage to indicate tumours with
distant metastases. The classification was soon applied to colonic as well as to rectal turnours'. Two modifications have complicated this classification. In 1949, Kirklin et al.' subdivided Dukes' system into: A, lesion limited to the mucosa; B,, lesion extends to, but does not penetrate, the muscularis propria; B,, lesion penetrates through the muscularis propria; and C, lesion of either B, or B, with involvement of lymph nodes. Class C of this system was later modified by Astler and Collerg into C1, lesion limited to bowel wall with positive nodes, and C,, lesion through all layers of bowel with positive nodes. These changes have created more confusion than clarity, making it difficult to compare reported series". A modified B, lesion corresponds to Dukes' A. Tumours classified as A, as designated by Kirklin et al., and by Astler and Coller, are rare; of the 981 tumours studied by the authors of the two classifications, only nine (fewer than 1 per cent) met their criteria for stage A. In contrast, 15 per cent of tumours in the series reported by Dukes and Bussey' were classified as Dukes' A. The reported frequency of Dukes' A tumours varies from 6 per to 25 per and even 35 per cent". Although differing referral patterns or detection of earlier lesions by colonoscopy'g may be contributory, the various interpretations of what constitutes a Dukes' A lesion account for many of these differences. The Cl and C, categories proposed by Astler and Collerg are different from those described by Dukes; another source of confusion results from the use of the term 'muscularis propria'. In the colon and rectum, the muscularis propria consists of the longitudinal and circular layers of smooth muscle in the intestinal wall. Although Astler and Coller used this in their staging system, the term has subsequently been misinterpreted to mean muscularis mucosa. This has resulted in both B, and
608
0007-1323/92/07060846
Colorectal carcinoma is the second commonest cause of death from cancer in western countries, accounting for 14 and 16 per cent of cancer deaths in men and women, respectively'. Approximately one person in five in England and Wales now dies from cancer, and one in eight of these will die from tumours of the large bowel2. A clinician encountering a patient with colorectal cancer needs to know the expected outcome to plan appropriate therapy. Over the past 60 years, claims have been made of numerous variables being related to survival. However, the application of many of these in a clinical context has serious drawbacks, in that they are complex, or poorly reproducible, or have not been subjected to adequate statistical analysis. This review discusses the relationship to survival of clinical and pathological variables, and highlights the disadvantages of the principal staging systems that are currently in clinical use.
Tumour staging
'
0 1992 Butterworth-Heinemann
Ltd
Prognostic factors in colorectal cancer: G. T. Deans et al.
B, of the modified Dukes’ systems corresponding to class A of Dukes’ original scheme. Indeed, Fisher et al. found that subdividing patients according to Astler and Coller A and B, did not improve on the predictability of survival indicated by Dukes’ A alone2’. Irrespective of these problems, there are deficiencies in all the Dukes classifications.Analysis based on a few tissue samples may miss breakthrough of the muscularis propria, such that some patients who die quite soon are misclassified into the good prognosis Dukes’ A group. In a series of 2037 rectal cancers, Dukes reported a crude 5-year survival rate of 81.2 per cent for patients with class A tumours, instead of the near 100 per cent expected if the classification were perfect”. The histological grade of the tumour and the number of involved nodes, which are both known to be related to s u r v i ~ a I ~are ~ - not ~ ~ ,considered in Dukes’ classification. The thoroughness of the pathologist in examining for lymph nodes is also important. Blenkinsopp et aLZ4,reviewing pathological assessment of the depth of tumour penetration in 2046 patients from several institutions, noted that variation between observers ranged from 5 to 30 per cent ( P < 0.001).Differences were also noted between centres in the mean number of lymph nodes harvested per specimen ( P < 0.001) and the proportion of cases in which no nodes were identified. Because microscopic appearances are rarely difficult to interpret, these differences are most likely to be caused by errors in the macroscopic assessment of the specimen. Such results suggest that caution is necessary when comparing the proportions of cases in the various Dukes’ classes between institutions. Despite these problems, Dukes’ classification remains an extremely powerful prognostic tool. Dukes and Bussey reported 5-year survival rates of 81.2, 64 and 27.4 per cent for patients with A, B and C tumours, respectively, figures comparable to those of most modern series. Studies employing univariate analysis’2~zs-29have repeatedly emphasized the prognostic importance of tumour staging in colorectal cancer. Similarly, major series employing multivariate analysis have identified tumour stage as the primary determinant of prognosis in colorectal cancer15,17,2z,30-33.The only improvement required on Dukes’ assessment of the degree of tumour penetration may be the use of Astler and Coller stages C, and C,. Multivariate analysis has revealed this to be superior to a single Dukes’ C class, and to be independent of the number and site of nodal metastases and the degree of tumour differentiation,’. In summary, the deficiency of Dukes’ classification is that it fails to take clinical factors and the number of involved nodes into account. Modifications have resulted only in confusion over the interpretation of the various classes, particularly A. In spite of these problems, the prognostic significance and comparative ease of Dukes’ classification ensure that it remains the ‘gold standard’ against which all other prognostic variables in colorectal cancer should be tested.
’
Clinicopathological staging The turnour node metastasis system In an attempt to overcome the problems of Dukes’ classification a clinicopathological system was proposed34. This consisted of grading the extent of the primary tumour (T), the condition of regional lymph nodes ( N ) and the presence or absence of distant metastases (M). The Union Internacional Contra la Cancrum, however, agreed that the extent of colonic carcinoma could not be fully assessed clinically at the time of surgery. The American Joint Committee on Cancer therefore developed a series of prefixes for the TNM classification to relate to the extent of the disease at given sites and time^^'-^'. These are: cTNM, clinical diagnostic staging; sTNM, surgical evaluative staging; pTNM, postsurgical pathological staging; rTNM, retreatment staging (e.g. at second laparotomy ); and aTNM, autopsy staging. This American TNM classification was considered by many
Br. J. Surg., Vol. 79, No. 7, July 1992
to be too complex for routine use. In Australia a separate clinicopathological system was developed. This employed an A, B, C, D grading but combined clinical information, curative or palliative operation status, and a pathological subdivision of Dukes’ classes before a final grade was ascribed40A6. In Spain a further system, based on a subclassification of Dukes’ ‘D’ tumours, was developed4’. This international variation in the surgical, clinical and pathological classifications was addressed in 1987 with the adoption of identical criteria between c ~ u n t r i e s;~guidelines ~ . ~ ~ for an internationally agreed staging system were p r ~ p o s e d ~ ’ , ~Despite ’. attempts to produce a standardized clinicopathological classifications2, the TNM system is complex relative to Dukes’ classification. For instance, subdividing patients into T,NoMo and T,NoMo does not significantly improve the prediction of survival over a simple Dukes’ A classification2’. Consequently, the TNM system has not become widely accepted in the UK. Other clinicopathological variables In an attempt to improve on both Dukes’ and the TNM classifications, several studies have assessed the prognostic significance of numerous clinical and pathological variables. These include tumour size53, lymphatic i n ~ a s i o n ’ ~ - ~ ~ , mucinous tumourss7, laminin”, involvement of the lateral resection margin30~59.60, silver-binding nucleolar organizing region^^'-^^, lymphocytic and vascular invasion in the individual Dukes’ classes64, biochemical markers6’, bloodgroup antigens66p69,carcinoembryonic antigen70-72 and CA 19-973-77.Despite suggestions that some of these variables may be related to survival on univariate or multivariate analysis, none consistently improves on the prognostic value of Dukes’ classification in a regression analysis model. Three clinicopathological variables, however, are often related to survival and warrant further discussion. These are patient age, histological grade and venous invasion. Patient age. The most important clinical variable appears to be patient age. Whether older or younger patients have the poorer prognosis is, however, unclear. Several studies have reported that increasing age is associated with a poorer prognosisl 5,17,78-80 . Th’is appears to be independent of the fact that older patients are more likely to die from their operation” or concomitant diseases2than are younger patients. Nevertheless, recent reports have suggested that age itself should not be a determinant when considering operation in patients with colorectal ~ancer’~-’~. In contrast, some authors have noted a particularly poor prognosis in those aged
