The Malignant Polyp: Polypectomy or Surgical Resection?

Endoscopic resection of colorectal polyps is a well-recognised therapy for the prevention of colorectal carcinoma. Roughly 10% of resected polyps contain foci of carcinoma and are of‐ ten termed malignant polyps or polyp cancers. Their incidence is increasing in line with the increasing use of colonoscopy.[1] A proportion of these will have progressed to nodal dis‐ ease before presentation and a further oncological resection should be considered for high risk patients.[2]


Introduction
Endoscopic resection of colorectal polyps is a well-recognised therapy for the prevention of colorectal carcinoma. Roughly 10% of resected polyps contain foci of carcinoma and are often termed malignant polyps or polyp cancers. Their incidence is increasing in line with the increasing use of colonoscopy. [1] A proportion of these will have progressed to nodal disease before presentation and a further oncological resection should be considered for high risk patients. [2] The risk of nodal disease at presentation can be stratified by histology but definitive staging information can currently only be obtained by oncological resection, a procedure which can cause significant morbidity and mortality, especially in the elderly. This is of particular relevance as the majority of patients do not have nodal disease, even for the most dangerous categories of polyp. [3] There is a real risk of causing excess morbidity by over treating the majority in order to adequately treat the minority.

Malignant polyps
Not all polyps are created equal. The adenoma carcinoma sequence has long been recognised as the natural history of colorectal carcinoma and it is therefore logical that some adenomas will be discovered with foci of malignancy within them.
For those confined to the mucosa, polyps showing foci of potentially malignant cells are often termed carcinoma in situ. The lack of lymphatics in the mucosa prevents distant spread and, as these lesions are neither regarded as malignant or treated as malignancies, the term high grade mucosal neoplasm is now preferred. [4] The definition of colorectal carcinoma is dysplasia crossing the muscularis mucosa, so when high grade dysplasia in these polyps crosses this barrier the lesion is termed a malignant polyp. A malignant polyp is essentially a macroscopically benign lesion that contains malignant foci on further examination. When the totality of the polyp is comprised of malignancy the term polypoid carcinoma is often used.
T1 lesions are therapeutically significant as they are the first lesions where nodal and distant metastases must be considered. The management of these polyps is based on the belief that the risk of spread can be stratified according to the histology of the resected polyp. [5] In the past authors used various criteria to define favourable or unfavourable histology and guide management. [1,2,6,7] For a large part, this has involved dividing patients into two groups. A "low risk" group, who are safe without further treatment and a "high risk group", for whom surgery should be considered. [8,9] Unfortunately published studies disagree about the factors that are most significant. [3,[10][11][12] 3. Factors affecting risk of nodal disease 3

.1. Morphology
The Paris classification [13] of gastro-intestinal tumours recognises that adenomas may be polypoid or non polypoid. Non polypoid (0-II, 0-III ) lesions are not usually removed endoscopically as they are more challenging to remove and are recognised to have high malignant potential.  Polypoid lesions can be pedunculated (Type 0-1p) or sessile (Type 0-1s). Due to their shape malignant sessile polyps are harder to remove with clear margins and have more ready access to the deeper portions of the submucosa. They are therefore more likely to be classified as high risk. Seitz et al [9] presented a series of 114 endoscopically removed malignant polyps. Overall 46% of these polyps were sessile, but 65% of "high risk" (ie. requiring surgical removal) polyps were sessile. Conversely only 23% of "low risk" polyps were sessile.
An earlier literature series of 741 malignant polyps reported that 58.3% of sessile polyps had "high risk features" (Grade 3-4, vascular or lymphatic invasion, positive resection margin) whereas only 10% of pedunculated polyps were similarly classified. [11] One meta-analysis reported positive resection margins in 56.8% of sessile lesions verses 18.7% in polypoid lesions (P < 0.0001).
Size and tubular or villous architecture are also well known to affect the malignant potential of polyps. However, in a similar fashion to flat or depressed areas of dysplasia, very large polyps are seldom excised endoscopically and are not relevant to the current topic.

Grading
Polyps are defined by dysplasia and the varying degree displayed by different polyps is thought to explain a large degree of their different metastatic potential. [14] Negative for Intraepithelial neoplasia.
Indefinite for Intraepithelial neoplasia.
In a meta-analysis of published series, Hassan et al. [1] reported a 3.9 (1.9-8.4) odds ratio for nodal metastasis with regard to high vs low grade malignant polyps. The odds ratio for mortality was reported as 9.2 (4.7-18.2). Determining the exact risk from high grade dysplasia is complicated by their relative rarity. One study of 80 malignant polyps found only 2 poorly differentiated polyps. [12] In a meta-analysis 7.2% of 1612 malignant polys were high grade. [1] It is interesting to note that despite poor differentiation being recognised as an important determinant of nodal disease, no universally accepted definition exists. Indeed in studies where the prevalence of highly dysplastic lesions was lower, the risk of nodal disease in these polyps was increased. (See Table 3). This suggests that poor differentiation, when a rigorous definition is used is an extremely important predictor of nodal disease. Those studies that did not find the degree of dysplasia to be significant are hampered by the very small number of highly dysplastic lesions in their sample.

Depth of invasion
Haggitt's classification is based on the greatest anatomical depth of invasion in pedunculated polyps. [5] Haggitt 0 lesions are confined to the mucosa. Haggitt grades 1-3 breach the submucosa within the polyp, and they are confined to the head, neck and stalk of the polyp respectively. Only Haggitt 4 lesions invade past the stalk into the submucosa of the wall. Most authors would agree that only Haggit 4 lesions require further treatment. If adequately excised, Haggitt 0-3 lesions have a risk of recurrence (<1%) which is lower than the predicted mortality of an oncological resection. [15,16] Conversely, for level 4 lessions, Haggitt reported nodal disease rates of almost 13%.
All sessile lesions are Haggitt 4 by definition, but other authors have treated selected sessile lesions with polypectomy alone to good effect. Kudo produced a refinement for sessile polyps by dividing the submucosa into thirds. [13] This has become known as the Kikuchi classification. [2] Lesions confined to the superficial third of the submucosa (called Sm1) demonstrated very low rates of nodal disease and many authors recommend no further treatment after polypectomy for Sm1. In the absence of other risk factors most would agree for Sm2 lesions as well. [7] The Kikuchi classification has been widely accepted for the assessment of T1 colorectal tumours but can be difficult to perform on endoscopy specimens as the muscularis propria is not usually included in the specimen. [2] Difficulty is also encountered when the muscularis muscosa cannot be identified. A large collaborative Japanese study used Haggit level 2 (i.e. the border between the head and neck of the polyp) as a baseline for pedunculated polyps. Provided that there was no lymphatic invasion, they found no nodal disease if the depth of invasion from here was <3mm. For sessile polyps the superficial aspect of the lesion was used and again no nodal disease discovered if the invasion was <1mm, regardless of other lymphatic invasion. [17] Other Japanese studies have also found good correlation between quantitative measures of submucosal invasion and risk of lymph node metastasis. [18,19] Although the study included operative specimens as well as endocsopically removed malignant polyps, Ueno [14] showed that the width of tumour invasion is also an important factor.

Incomplete or piecemeal resection
Involved resection margins have been shown to be strongly associated with poor outcomes. These patients have higher mortality, local recurrence and rates of residual disease. In one, all be it small, study 75% of incompletely resected polyps were associated with an adverse outcome. [20] It should be noted that even when incomplete resection is reported, absence of residual disease in the surgical specimen is the rule (94% in one study) rather than the exception [21]. This is likely due to diathermy electrofulguration of the remnant.
The European recommendations state that tumour cells within 1mm of the margin represents a positive margin [22], with some authors arguing then >2mm represents the true safe margin [21].
Incomplete removal is failure of primary therapy and requires further resection. Piecemeal removal of the polyp prevents proper histological assessment and surgery is mandated in all cases. For this reason endoscopic mucosal resection by the strip biopsy method is discouraged for the removal of potentially malignant lesions.

Lymphatic and vascular invasion
Lymphatic invasion has been sighted by some authors as an important predictor of nodal disease. Controversy exists however as reported cases are rare and usually associated with poorly differentiated tumours or incomplete resection. Inter-observer variability and the ease of mistaking retraction artefact for lymphatic invasion also make interpretation difficult. [6,12] Lymphatic invasion is usually associated with other high risk factors and in those cases with adverse outcomes, almost invariably so. Many authors would regard its status as an independent risk factor is unclear. [20] However a large multi-centre retrospective study from Ja-pan found lymphatic involvement to be highly significant for nodal metastases (odds ratio 4.69 P<0.0001) in a multivariate analysis of risk factors. [17] They also found that in a small number of cases adverse outcomes were seen from cases of lymphatic invasion, despite invasion being confined to the head of the polyp (Haggit 1).
Vascular invasion is also considered difficult to identify, but where present, it is strongly associated with nodal disease. Yasuda [18] studied T1 rectal tumours, including specimens from primary resections and resections after polypectomy. The odds ratio for the nodal metastasis with reference to the presence or absence of vascular invasion was 12.023 (3.751-116.751 p=0.001). Another study of sessile T1 colorectal carcinomas found that vascular invasion was a significant factor in both univariate and multivariate analysis. However they admit that the small number of cases of vascular invasion were found in lesions with deeper Sm3 invasion. [23] An odds ratio of 7 (2.6-19.2) for lymph node metastasis was reported in the only meta-analysis looking specifically at malignant polyps and the presence of vascular invasion. [1] However, the same analysis demonstrated no such increased risk in polyps that would otherwise be considered low risk. It may well be that vascular invasion carries no special significance in itself and should not be emphasised in decision making.

Tumour budding
Tumour budding is the presence of microscopic islands of tumour cells out ahead of the main front of tumour invasion. At present there is no defined agree standard to reporting the phenomenon but several authors have found it to be highly significant. Yasuda reported an odds ratio of 11.11 (3.64-146.03) [18] for predicting nodal disease but until further study occurs it is difficult to make firm recommendations.

Location
T1 rectal tumours seem to be particularly likely to cause nodal disease, especially when located in the lower third. [15,23] However at this stage there have been no studies looking at this relationship specifically in malignant polyps.

"The low risk polyp"
Several authors, starting with Morson in 1984 [24], have developed the concept of the low risk polyp. That is, a polyp which can be safely treated with polypectomy alone, as there is minimal risk of nodal disease. The concept has been incorporated into the American College of Gastroenterology guide lines [25]. They recommend no further treatment if: The polyp is considered to be completely excised by the endoscopist and is submitted in toto for pathological examination.
In the pathology laboratory, the polyp is fixed and sectioned so that it is possible to accurately determine the depth of invasion, grade of differentiation, and completeness of excision of the carcinoma.
The cancer is not poorly differentiated.
There is no vascular or lymphatic involvement.
The margin of excision is not involved. Invasion of the stalk of a pedunculated polyp, by itself, is not an unfavourable prognostic finding, as long as the cancer does not extend to the margin of stalk resection.
The European recommendations, while noting the potential of tumour budding and lymphatic and vascular invasion as prognostic factors, decline to provide a guideline as they have not been statistically significant in all cases. [22] Another perspective is given by Nicholls, [7] who instead offered an algorithmic approach. He differentiates between colonic and rectal polyps. For rectal lesions judged to be adenoma prior to resection he recommends that all poorly differentiated lesions be removed. For colonic polyps he suggested further resection solely for Haggit 4 polyps (including by definition all sessile polyps) with a depth of invasion >1000 µm.
Systematic review of studies which selected low risk polyps using methodology broadly similar to the American criteria has demonstrated very low rates of nodal recurrence. (See Table 4). Mortality from oncological resection varies greatly by age and co-morbidity but is usually quoted around 3-5%. [26][27][28] Therefore, for these lesions, the safest course of action is surveillance rather than further resection. [8,29] It should be noted that these criteria take no account of the depth of invasion and that these guidelines would encourage the removal of some lesions that have been safely treated by endoscopy. It may be that they are documenting many of the same characteristics but in a different way. It is not hard to imagine that Sm3 lesions are less likely to be excised with clear margins and are more likely to show poor differentiation. Indeed a large study of surgically resected sessile T1 colorectal tumours found Sm3 invasion in 68% of G3+4 tumours and on 33% in G1+2 (P=0.001). This study also found tumour grade not to be significant on multivariate analysis. [23] Given Japanese experience it maybe be better to refine the criteria for a low risk polyp as any polyp lacking all of these features: High grade (G3-4) lesions.
Incomplete resection or other factors preventing adequate histological assessment of the lesion.

Piecemeal resection
Depth of invasion greater than 2mm from muscularis mucosa Width of invasion greater than 4mm.
The utility of including lymphatic invasion, vascular invasion or tumour budding is unclear at this time. Further work should be done to examine the risk from polyps of the lower third of the rectum, especially as these can often require a permanent stoma if oncological resection is performed.

"The high risk polyp"
Polyps that do not meet the low risk criteria should be considered for surgical removal even if there has been total excision of the primary lesion. Indeed, it is unusual to find residual tumour in the surgical specimen, especially if the lesion had clear histological margins. [6] The justification for surgery is the desire for regional control as the risk of nodal disease is much higher in these patients and oncological resection is required to obtain regional control in a similar manner to other colorectal malignancies. The dilemma is that only a minority of these patients have nodal disease requiring control and these patients are only reliably identified after resection. Especially in elderly, the decision to resect has the possibility to cause considerable harm without producing a benefit to the patient.

"First do no harm..."
It is an old surgical adage that surgery is only indicated if the natural history of the cure is better than the natural history of the disease. In situations of uncertainly like this it is useful to examine the possible outcomes of proposed courses of action in order to see were the survival advantage lies.
The outcomes of the decision to operate will be a function of the risk of nodal disease and the risk of operative mortality and morbidity. We feel it is useful to consider these decisions with reference to a 2x2

Nodal disease and oncological resection
We see no reason to regard these patients as any different from patients who had proceded straight to oncological resection and post operatively were staged either IIIa to IIIc (TNM v5). In the SEER data from 1998 to 2000 there is a huge difference between those with regard to five year survival (73% vs. 28% respectively). Clearly this stage differentiation has huge implications for the advisability of surgery. It has been suggested that T1-2N2 tumours have a better survival that T3-4N2 tumours and the TNMv6 classification has been changed to reflect this. Newer SEER data shows five year survival of 87.7% for T1-2N1 disease and 75% for T1-2N2 disease. [30] To our knowledge there is no current method of estimating the extent of nodal disease from polypectomy histology.

Nodal disease and no further resection
For those patients with nodal disease who do not have it resected the prognosis is likely to be compromised. Intensive surveillance is likely to detect continued disease progression.
The role of chemotherapy and or radiotherapy has not been clearly defined in this group but is likely to be palliative in nature.

Absence of Nodal disease
The survival of patients after endoscopic removal of T1 lesions and no nodal disease is excellent. The mortality in these groups will be limited to the operative mortality from further resection.

Risk of Nodal disease
Various figures have been quoted in the text for the risk of nodal disease in high risk patients. This can partly be explained by the differing criteria used to define risk by various authors. Further stratification within the "high risk group" may become apparent with further study.
The St Mark's Lymph Node Positivity Model [31] can be used to predict the individual risk of nodal metastasis after local resection of rectal tumours. However it makes no distinction within T1 tumours. Such assessment of individual risk factors to produce a personalised risk is not possible based on the current evidence. Further studies using multivariate analysis will be required to tease out the importance of individual risk factors.
For our analysis we have chosen to present data based on Sm depth as this has shown to be a reproducible predictor of nodal disease. Both Kikuchi and Nascimbeni reported rates of roughly 5, 10 and 25% for Sm 1, 2 and 3 respectively. [23,32] Risk Factor Incidence of nodal disease

The risk of further resection
Oncological resection of colorectal lesions is performed via segmental resection of the affected potion of the bowel and its draining lymph node basin. Harvesting these nodes gains local control and definatively stages the disease. It is a major undertaking with considerable risks. In the case of very low rectal tumours an abdomino-perinal excision of rectum (APER) results in permanent stoma formation.
In the UK at least, operative mortality has fallen in recent years. 30 day mortality was 6.8% in 1999, falling to 3.7% in 2009/10. [33,34] Rates from Scandinavia (4.8%) [26] and the US (3.1%) [27] are broadly similar. This remains considerably higher than the rate of nodal disease in the low risk malignant polyps. The 90 day mortality rate, considered by some authors to be a more accurate measure of operative mortality is higher still, is 5.6% in the UK. [34] This baseline rate is affected by both tumour and patient factors. Patients over the age of 80 are over ten times more likely to die than those under 50 (15% vs 1.2% 30 day mortality). [28,33] Comparing ASA1 and ASA4 patients, the odds ratio for death at 30 days is 14.06. Patients with rectal tumours do better than those with colonic tumours, though this seems to be due to high mortality for patients undergoing subtotal or total colectomy. Female sex, affluence, high volume surgical centres and elective rather than emergency surgery all also have a beneficial effect. [33] On top of mortality, anastomotic leaks, wound complications, cardiovascular complications, defecatory disorders and the psychological impact of stoma formation must also be considered when deciding whether or not to resect. Morbidity rates of up to 35% have been reported in the past. [15] These seem to affect laparoscopic surgery as much as open resections, but hospital discharge and return to work occurs sooner following laparoscopic procedures. [35,36] A more accurate individualised operative risk can be estimated from risk scoring systems. CR-POSSUM uses patient and operative parameters to estimate operative risk on an individual basis and has been well validated. [37,38] Cardiopulmonary exercise testing is also useful in predicting complications and the length of hospital stay. Both these tools can be of great use to the surgeon and patient when used thoughtfully during surgical planning. [39]

The special case of rectal tumours
There has been considerable interest in recent years in local resection of early rectal tumours to avoid stoma formation. This is relevant as malignant polyps in the rectum are a variety of early rectal tumour. Transanal Endoscopic Microsurgery (TEMS) allows full thickness excision of rectal lesions below the peritoneal fold, with excellent rates of local recurrence. [15] Its ability to harvest local lymph nodes is limited, and as such it is not generally a suitable second procedure for high risk lesions. It may have a role as secondary procedure for incompletely removed polyps which otherwise show favourable features. In its guidance the ACPGBI recommended full classical resection of rectal tumours that show the high risk features described earlier. [40] The anatomical location of the draining nodes in rectal lesions has also encouraged more extensive use of imaging to predict local nodal metastases. Endoanal Ultrasound and MRI both have the ability to detect enlarged local nodes; however distinguishing between the commonly found reactive nodes and metastases can be difficult. Micrometastases have also been detected in radiologically normal nodes. The use of new contrast agents may improve accuracy but currently histological examination remains the gold standard for detecting nodal disease. [15] For locally excised T1 and T2 tumours adjuvant chemoradiotherapy has been used with success to prevent local recurrence if further surgery is not deemed appropriate. The role of adjuvant therapy in malignant polyps is unexplored at this time.

Calculating the survival advantage
As the nodal status of these patients is unknown prior to surgery, mortality is a composite of the mortality of those with and without nodal disease. The contribution from each group will be in proportion to the risk of nodal disease.

cM = R.NM + (1-R).nM
Where cM= Composite mortality R =Risk of nodal disease NM =Mortality of those with nodal disease nM =Mortality of those without nodal disease The best course of action can be discerned by calculating the difference between cM with and without surgery. Tables 7-9 contain sample composite survival figures and number needed to treat at five years for various stages of malignant polyp.
Using this method we can see that for a patient with a Sm3 lesion (25% of nodal disease) and a predicted operative mortality of 2% there will be an absolute risk reduction of mortality at 5 years of 16% (NNT 6.25) if 5 year survival of node positive patients is 75% and a 4.5% reduction (NNT 21.05) if 5 year survival of node positive patients is 27%. In stage IIIa (75% five year survival) disease the absolute risk is reduced by 10% (NNT 1), but this disappears for stage IIIc. There has been considerable debate regarding the need to resect Sm2 lesions. In this model the benefit from resection disappears once operative mortality reaches 5% for IIIa lesions and 10% for IIIc lesions. Clearly careful though needs to be given to risk when choosing to operate on these patients.
Obviously this model makes no account of operative morbidity. For patients with IIIc disease and Sm3 lesions there is a survival advantage to operating; however the decision to subject 40 patients to major surgery to save 1 life at five years needs careful consideration.

Conclusion
This problem has been known and debated for over 30 years. [41] As the role of endoscopy has grown and developed, guidelines have been formulated to help clinicians make beneficial choices. Unfortunately the small scale and heterogeneity for published work had prevented any guidelines from gaining universal acceptance. The focus on tumour grade in the American guidance has been challenged by work from Japan that emphasises the importance of quantitative measures of the depth of invasion. Japanese work has also shown lymphatic invasion, vascular invasion and tumour budding to be of high prognostic significance, but concerns about reproducibility have prevented their universal adoption. It is also unclear which observed prognostic factors are truly significant and which are cofounding. None of the prognostic factors identified are highly specific and clinicians must still make difficult decisions based on the balance of risk.
The solution to this problem will surely come from improved pre-operative staging. Endo-anal ultra sound and targeted contrast MRI have both shown promise for rectal tumours. Sentinel node mapping in the colon has also been investigated but remains experimental. [42] Until highly accurate pre-operative staging of nodal disease is possible effort must be made to refine the classification of malignant polyps to identify the truly significant prognostic factors. It is the opinion of the authors that an individualised prediction model comparing operative surgical risk and risk of progressive disease should be used to counsel patients regarding future strategies. Creation of a national or international database would facil;itate better predictive models.  Table 10. 5 year survival, survival advantage with further resection and number needed to treat if 5 year survival in node positive patients is 27% after resection.