COLORECTAL CARCINOMA STAGING SYSTEMS

Cuthbert Dukes, a pathologist at St Mark’s Hospital, London, published his famous prognostic classification for rectal carcinomas 70 years ago.¹ He used the letters A, B, and C to classify and stage the disease, basing his observations on the results of 215 rectal cancers that exhibited a worsening prognosis in parallel with a more advanced stage. Later, it was found that this system was also suitable for the staging of colon carcinomas.^2,3 This classification was later modified slightly by Dukes and his colleagues: they divided group C into two subcategories, assigning stage C1 to those carcinomas with nodal metastases, but no metastasis in the apical nodes, and stage C2 to those tumours that displayed apical node involvement (table 1). With time, many other classifications and staging systems involving the use of the same letters were introduced (table 1). The original Dukes’s classification and its modifications with the letters A, B, C, and sometimes even D assigned to different stages and prognostic groups of CRCs remain popular and are still widely used. It seems that “Dukes’s C CRC” will long continue as the most common expression for node positive CRCs. The same seems true for “Dukes’s B CRC”, referring to node negative T3 tumours.

However, the fact that the classifications listed in table 1 involve the same letters, but with different meanings, and often appear with the eponymous form of the modified Dukes’s classification has given rise to confusion and misuse.¹⁸ It is also somewhat surprising that even the name Dukes is often misspelled (many published works mention a Duke’s classification).

“Patients with metastatic lymph nodes have a shorter survival and may require adjuvant systemic chemotherapy”

An alternative way to stage CRCs is the system based on the UICC TNM classification (table 2), which was last revised in 2002. From 2003, all tumours are expected to be staged on the basis of this most recent version. Although the A, B, C, and D based staging systems are said to be confusing,¹⁸ it is also believed that modifications and revisions of the TNM system may likewise give rise to some confusion. Whatever the staging system applied, a common language that pathologists, surgeons, and oncologists use and understand properly is a neccesity. The TNM system is a rather general system that is periodically updated in accord with new insight into the biology of malignant disease. It allows the inclusion of several prognostic variables (for example, the R classification for residual disease, or the L classification for lymphatic invasion), and therefore seems the staging system most suitable for universal usage. As suggested by a recent editorial,²¹ it is our language for cancer care. Although the definitions for the TNM categories are unchanged in the sixth edition of the TNM classification,¹⁷ there has been a minor change in the definition of LNs. Both in the TNM system¹⁹ and in the UK minimum dataset for reporting CRCs,²² tumour nodules in the pericolic or perirectal fat with no evidence of residual LNs were recorded as LNs if they were greater than 3 mm, whereas in the revised TNM classification¹⁷ they are classified as LNs only if their form and smooth contour is consistent with an LN origin; otherwise, they are classified as venous invasion (category V in the TNM system) based on likelihood. This revision is substantiated by a study demonstrating the intravascular, perivascular, or perineural origin of these tumour nodules, and the worse prognosis associated with them, independently of their size.²³

It should be borne in mind that staging systems are merely a convenient way to attempt to report uniformly a set of recognised prognostic features on which further treatment may be based, and the results of different treatment modes may be compared. Many factors influence the assessment of these features, nodal status being one of the most important.^24–26 It is of primary importance in decisions relating to the administration of adjuvant chemotherapy, and provides an objective confirmation of adequate regional LN dissection.²⁷

THE NUMBER OF LYMPH NODES ASSESSED

The number of LNs assessed depends on the number of LNs present in the patient, on the number of LNs removed by the surgeon, and on the number of LNs identified by the pathologist. It is also dependent on what are identified microscopically as LNs (for example, subserosal tumour nodules or nodes with adipose changes).

LNs are usually located in the vicinity of vascular channels, and are not evenly distributed throughout the pericolic fat or mesorectal tissues along the large bowel.²⁸ Some studies have suggested that there are fewer LNs around the rectum than around other segments of the colon,^29,30 but other studies have reported similar numbers of LNs for the rectum and other parts of the large bowel.^31,32 The transverse colon has also been claimed to have fewer LNs than other segments in some studies.^30,31 The number of LNs present in the mesocolic or perirectal fat varies from individual to individual and, under certain conditions, the number of LNs may increase even in a given individual. This increase may be related to the reactive enlargement of the LNs that are present but too small to be detected, but de novo genesis of LNs may also contribute.³³ As evidence in favour of a reactive increase in the number of LNs identifiable by palpation, we have documented that the number of LNs in each centimetre of large bowel segment is greatest in close proximity to tumours.³⁴ Anatomical and physiological factors may influence the number of LNs assessed, and probably play a major role in differences of LNs assessed in different patients undergoing the same type of surgery for the same stage of disease, but these factors cannot be influenced.

The extent of surgery has an obvious bearing on the number of LNs identifiable. It seems evident that the length of the resected bowel segment also influences the number of assessable LNs, because a longer segment has more associated LNs than a shorter one (for example, a sigmoid resection versus a left hemicolectomy). The more radical the colon resection, the more mesocolic fat is removed, and the more LNs present in the specimen. For rectal cancers, total mesorectal excision (TME)³⁵ specimens contain more LNs than anterior resection specimens performed before the era of TME. Some of these nodes are involved and are not expected to be removed by conventional resection.³⁶ Surgical factors influencing LN numbers may vary from institution to institution or from surgeon to surgeon, but these variations are expected to be relatively slight if the procedure is standardised.

It seems that the pathological work up has a major role in the establishment of the nodal status from an adequate number of LNs; with no major changes in the surgical procedures, the number of LNs assessed could be increased by means of a more thorough dissection.^37–39

“Overall, the quantitative requirement for the nodal staging of colorectal carcinomas is to recover as many lymph nodes as possible”

Guidelines on CRC reporting suggest that all LNs should be carefully dissected out and examined histologically,^22,24,40 but there is nevertheless a wide variation in the numbers of LNs assessed from study to study. It has been demonstrated that fat clearing or LN revealing substances can help in increasing the number of LNs recovered,^15,31,41,42 and that this can lead to the upstaging of some tumours reported as LN negative on the basis of palpation and manual dissection. A survival disadvantage has also been documented for these upstaged CRCs.⁴³ However, the sensitivity of manual dissection varies widely between institutions: the mean of 6.2 LNs without fat clearance in a study indicating a disadvantage for patients with small metastatic LNs discovered only after fat clearance³¹ appears inadequate, and a mean of 18.5 LNs after fat clearance is still below the mean reported in some studies where no special clearing agents were used.^32,37 Adequate fixation alone may also help in the retrieval of more LNs³² or in the identification of more LN positive CRCs.⁴⁴

It seems logical that the more LNs a pathologist assesses, the more likely it is that a metastatic LN will be picked up. Assessment of only a few LNs is therefore associated with a smaller chance of finding metastatic ones. Likewise, in series of patients with CRC, the percentage of LN positive cases reported is lower if fewer LNs are assessed.^29,37 Accordingly, there have been suggested cut off points for a minimum number of LNs that should be assessed to achieve reliable nodal staging. The recommended mean number of LNs to be assessed also varies considerably, ranging from six to 17 (table 3). Such minimum values have been disputed.^52,53 It is generally understood that pathologists should try to recover as many LNs as possible, but that if they fail to identify the suggested numbers, the staging may still be accurate.^54,55

In some series it has been found that tumours that infiltrate less deeply, and therefore have a lower T category in the TNM system, are associated with fewer LNs recovered, probably as a consequence of smaller stimuli acting in favour of nodal hyperplasia or neogenesis.^29,41 It may also be hypothesised that metastatic CRCs have more LNs that attain a detectable size, which is why there are more LN positive cases in specimens with more LNs assessed; stated otherwise, the larger number of LNs assessed is not the reason why more LN positive CRC cases are found, but is rather a consequence of the development of nodal metastases and the more advanced stage. This is certainly true to some extent, and would militate against the painstaking work of recovering more and more LNs from CRC resection specimens, but two recent publications stress the quantitative recommendations of nodal assessment. Two independent studies, a large single institutional analysis of 2427 T3 CRCs including 1305 T3N0 tumours,³⁹ and a survival epidemiology and end results database analysis of 8574 T3N0M0 CRCs,⁵¹ have documented a better survival rate for LN negative CRC cases with more LNs assessed, a fact that can be explained by the more accurate staging arising from the evaluation of more LNs. A further study has also documented an improved survival for LN negative patients with 18 LNs or more assessed, although this study correlated larger nodal numbers with better performed lymphadenectomy.⁵⁶

Overall, the quantitative requirement for the nodal staging of CRCs is to recover as many LNs as possible. Because of the extra costs, routine use of fat clearing substances is not recommended, although this technique may well be of value. The advice on fat clearance or on the use of LN revealing solutions is usually restricted to cases where the number of LNs is considered suboptimum, although the optimum number cannot be adequately defined. The number of LNs assessed and the number containing metastases should be clearly stated in the reports.

QUALITATIVE FEATURES OF LYMPH NODES THAT MAY INFLUENCE STAGING

Adequate staging is dependent not only on how many LNs are assessed histologically, but also on which LNs are assessed. If an adequate selection between metastatic and non-metastatic LNs could be performed, the quantitative requirements summarised above might also be altered.

In general, metastatic LNs are larger than non-metastatic LNs. The size of positive and negative LNs from CRC resection specimens was found to be significantly different in two studies, but this difference was too small for the practical purpose of differentiating between the LN populations by palpation, and both studies found a significant overlap between the sizes of involved and uninvolved LNs.^48,57 Large LNs may be negative and small LNs may still harbour metastases. Several studies have prompted the view that “small LNs” (usually defined as ≤ 5 mm) may be the only sites of metastases in CRC specimens.^41–60 We found that most of the LNs recovered are < 5 mm,^48,57 and therefore not above the limit assessable by imaging techniques (ultrasonography and computed tomography). This stresses the need for histopathology in the assessment of nodal status. It can be concluded that size alone is not a reliable discriminative feature between involved and uninvolved LNs. However, our series also suggested that the pathological study of the seven largest LNs could lead to the same staging as that of a mean of 12 LNs in 98% of cases.⁴⁸

The distance from the tumour may also be a feature that could be of help in differentiating negative and positive LNs. Previously, we established that most metastatic LNs are located in the vicinity of the tumour involved bowel segment. When the bowel was divided perpendicularly to its longitudinal (luminal) axis, 99 of 100 CRCs could be adequately staged as LN positive or LN negative on the basis of LNs found in association with the tumour involved bowel segment and the 1 cm segments proximal and distal to the edges of the tumour. The nodal status of all 100 cases could be adequately defined in the pTNM system on the basis of the LNs beneath the tumour and the 3 cm segments proximal and distal to it.³⁴ Incision of the mesocolic and pararectal fat helps in identifying more LN positive cases, probably by allowing better fixation and subsequently a better rate of identification of the subtumoral LNs,⁴⁴ and this also favours the hypothesis that involved LNs are likely to be located in the vicinity of the tumour involved bowel segment. One of the first studies on lymphatic mapping in CRC, using patent blue dye, also revealed that most of the blue LNs identified were located within 3 cm of the tumour margin.⁶¹ However, other lymphatic mapping studies have afforded counter examples by demonstrating unexpected direct drainage from a tumour to an anatomically distant LN.^62,63

“Size alone is not a reliable discriminative feature between involved and uninvolved lymph nodes”

The most promising qualitative feature of LNs that could help in a selection between them seems to be the occurrence of direct drainage from the tumour site. Despite previous findings on sentinal lymph nodes (SNs) in connection with parotid⁶⁴ and penile⁶⁵ tumours, the SN concept was first well established in connection with cutaneous melanomas⁶⁶ and breast carcinomas.^67,68

DIRECT DRAINAGE OF LYMPH FROM THE PRIMARY TUMOUR SITE: LESSONS LEARNT FROM OTHER CANCER SITES

Surgeons are able to localise SNs in several types of tumour with vital dyes such as isosulfane blue or patent blue, with radiolabelled colloids and intraoperative γ probes, or with a combination of these two methods. This has led to the formulation of the SN concept as a general rule for the lymphogenic spread of solid tumours,⁶⁹ a concept that has not yet been well founded in neoplasms other than cutaneous malignancies and breast carcinoma.⁷⁰

From the two sites for which most experimental data have been achieved so far, we have learnt that SN biopsy is a very promising technique for the surgical and pathological staging of breast carcinomas, where it permits (1) the selective obviation of axillary dissection in SN negative and therefore presumably axillary LN negative patients,^71,72 (2) an oriented internal mammary node biopsy if the tumour drains to this region,⁷³ and (3) a more detailed histopathological work up of a limited number of LNs^74–76; it may also have a therapeutic role.⁷⁷ In cutaneous melanomas, SN biopsy allows the mapping of nodal regions, selective nodal dissection for SN positive patients,^70,78 and the enhanced pathological assessment of a few selected LNs; the SN status is becoming a major prognostic factor.⁷⁹ Draining patterns considered uncommon on an anatomical basis have been recorded for both tumour types: for example breast cancers have been reported to drain directly to level III axillary LNs^80,81 or even supraclavicular LNs,⁸² the last of which were previously regarded as distant metastases.^19,20 Melanomas, especially those of the trunk, may often drain outside the recognised lymphatic fields, including retroperitoneal LNs.⁸³ Such experience also reflects the role that SN biopsy may play in the staging and treatment of CRCs.

THE ROLE OF LYMPHATIC MAPPING IN COLORECTAL CARCINOMA

Because the extent of surgery for CRCs is primarily defined by the location and size of the primary tumour, the identification of SNs did not initially influence the extent of surgery,⁸⁴ but was considered a tool allowing a more detailed histological approach to those LNs that are the most likely sites of metastases.⁸⁵ Missing these metastases was regarded as a possible cause of the treatment failures experienced in a rather high proportion of LN negative CRCs.^85,86 Later, lymphatic mapping studies provided evidence that the purely anatomical concept of nodal spread was not true, and that SNs may be located at unexpected sites, as seen in breast carcinoma and melanoma. The anatomical concept of lymphogenic spread suggests that LNs around the bowel would be the first to be reached by metastases, followed by intermediate and apical LNs, and then more distant (for example, para-aortic) LNs. Metastases to proximal (apical) LNs—that is, LNs at the ligature of the main blood vessel supplying the resected large bowel segment involving a tumour—were incorporated in the first modification of the Dukes’s staging system⁴ (table 1), and were also included in an earlier version of the TNM system (table 2). This was found to be of prognostic importance.^12,87,88 A failure of the anatomical order of spread is not exceptional and has long been known. The involvement of higher level LNs without metastases in LNs closer to the bowel is referred to as skip metastasis, the incidence of which may be as high as 10%.⁸⁹ Lymphatic mapping studies have demonstrated that direct drainage may occur from a tumour site to apical or even paraaortic LNs,⁹⁰ and this sheds new light on the definition of skip metastases. Even more surprisingly, the SNs may be located at sites relatively distant from the primary tumour. Surgeons from the John Wayne Cancer Institute have documented a case of CRC located in the ascendent colon that drained to an SN at the splenic flexure; this LN proved to be the only positive LN resected with the colon, but was outside the margins of a standard right hemicolectomy, the operation usually performed for a primary carcinoma at the given location.^62,63 Sterk et al reported on the detection of extramesorectal (iliac) LNs by rectal lymphoscintigraphy in four of 16 patients investigated, although they expressed doubts about the adequacy of their submucosal injections, and stated that deeper injections of the radiocolloid could be the cause for this high frequency.⁹¹ Larger series published to date suggested that unexpected lymphatic drainage occurs in 4–8% of cases,^86,92 either as deep mesenteric SNs or as SNs “placed to the left” in cases of right sided colon tumours. The pinpointing of these LNs may lead to altered surgery if they are outside the margins of the standard resection. Therefore, lymphatic mapping and SN biopsy offer the promise of a more detailed pathological staging centred on a few LNs and a patient tailored planning of the extent of resection. Despite this promise, there are many confusing factors relating to the identification of the SNs in CRCs.

TECHNICAL ISSUES OF SN BIOPSY IN COLORECTAL CANCER

The technique of lymphatic mapping (as in the case of breast carcinoma) has many unstandardised variables, which include the nature of the tracer (type of dye, type of radiocolloid, use of a tracer alone or in combination), the method of administration (submucosal or subserosal), the nature of the procedure (in vivo mapping versus ex vivo mapping; open versus laparoscopic surgery), the definition of the SNs (any blue LNs or the first few blue LNs), etc. These technical differences may be responsible for discrepant results. Table 4 summarises the details of published lymphatic mapping series. Both the extremely large variation in the false negative rate and the relevant upper extreme of this parameter suggest that SN identification in CRCs is far from perfect. Alternately, the data could also suggest that the SN concept does not hold for CRCs, but there are a few larger series that militate against this last explanation.

“Lymphatic mapping studies provided evidence that the purely anatomical concept of nodal spread was not true, and that sentinel lymph nodes may be located at unexpected sites, as seen in connection with breast carcinoma and melanoma”

On the whole, table 4 indicates that the best results in terms of reliability (low false negative rates) were attained with dynamic studies (either in vivo injection of the tracer and immediate in vivo identification of the SNs or ex vivo injection of the tracer and ex vivo identification of the SNs), because greater time scales may leave too much time for dye overflow; However, even with this method false negative results can occur. Small particle radiocolloids may also label further echelon LNs.¹⁰⁵ Overflow of the tracers can also partly explain the rather high discrepancy in the labelling of LNs despite a single combined tracer administration (81% of radioactive LNs were blue, and 76% of blue LNs were radioactive).⁹⁰ Tumour size (and the depth of invasion) is a further possible factor increasing the false negative rate in some series,^38,61,106 and massive nodal involvement may also play a part in it.^86,94,107

DETAILED ANALYSIS OF LYMPH NODES

It is not a new finding that a more detailed microscopic assessment of LNs (serial or step sectioning) or a more sensitive method of detection of microscopic involvement (immunohistochemistry) of LNs can reveal tumour cells undetected by standard haematoxylin and eosin examination.^108–111

One of the main advantages of SN biopsy is the concentration of detailed histopathology or ancillary techniques on a few LNs. The evidence to date is scarce, but serial sectioning and immunohistochemistry have been reported to upstage CRCs in 7–33% of cases (table 4). None of the articles listed in table 4 distinguishes between isolated tumour cells, micrometastases, and larger metastases,¹⁷ and the importance of detecting very small metastatic foci by enhanced histopathology or ancillary techniques is at present unclear, with some studies suggesting a survival disadvantage,^112–114 and others suggesting the opposite,^{109–111,115–117} If the technique of SN biopsy can be improved so as to reach a reasonable false negative rate in the range of 0–10%, then a more detailed work up of SNs could be advised outside research protocols, but the time has not yet come for this shift.

Immunohistochemistry can certainly (artificially) decrease the false negative rate of SNs to predict the LN status of CRCs, as shown by the study of Wong and colleagues⁹⁸ (table 4). Most of the investigations listed in table 4 used enhanced histopathology only for the SNs, and the validation of the SN theory in CRCs with an enhanced histology of both SNs and non-SNs is based on only 25 cases.⁸⁵ This study demonstrated an upstaging rate of 12% for SNs and of 1.2% for non-SNs, therefore suggesting that the SNs are really the most likely sites of metastases if the method used to identify them is adequate.

Because CRCs are generally positive for cytokeratin 20 (CK-20), this offers an ideal target for the reverse transcription polymerase chain reaction (RT-PCR) based study of regional nodal status.¹¹⁸ However, even this sensitive and specific method for both SNs and non-SNs failed to improve the false negative rate in one study,⁹⁰ again questioning either the validity of the SN theory for CRCs or the adequacy of the method used to identify SNs (table 4; in vivo tracer administration and ex vivo search for SNs). It was recently demonstrated that CK-20 is downregulated in CRC tissue samples, and the background expression may also be substantial, so that care must be taken when this single marker is used.¹¹⁹ Another study, using multiple marker RT-PCR (with primers for β chain human choriogonadotrophin, hepatocyte growth factor receptor, and universal melanoma associated antigen, all of which are frequently positive in CRCs), resulted in a high rate of upstaging of SN negative CRCs. The markers were found to be specific, and upstaging occurred more frequently with tumours that invaded deeper through the anatomical layers of the bowel.⁸⁶

It is still not clear whether or not SN biopsy with a targeted intensive pathology assessment can lead to greater insight into the biological relevance of micrometastases and isolated tumour cells.

CONCLUSIONS

Despite many attempts to define a minimum number of LNs that should be evaluated histologically to provide an adequate nodal staging of CRCs, it is advisable to assess as many LNs as possible. Although there may be some qualitative features of LNs that can be of help in the selection between them and the picking up of those that are more likely to harbour metastases, neither the size of the LNs nor their distance from the primary tumour allows a reliable selection. Lymphatic mapping may be a good adjunct to the surgical and pathological staging of CRCs by demonstrating unexpected lymphatic drainage. This review of the literature and our personal experience suggests that radioguided or dynamic dye guided studies with immediate identification of the SNs are the optimal methods for this, and the procedure is more adequate in early stage CRCs. Currently, the identification of SNs in CRCs is made uncertain by many reports of unacceptably high false negative rates, despite some optimistic results in other studies; outside research protocols, therefore, a detailed histopathology of these LNs is not justified.