The American humorist Will Rogers once joked that the move of many residents of Oklahoma to California was increasing the average IQ in both places. An analogous explanation for the apparent linkage between extent of pathological examination and cancer survival was first proposed by Feinstein in 1985, and named ‘stage migration'; the argument is that in a centre where pathologists make a greater effort to identify lymph nodes it is more likely that patients with nodal involvement will be staged accurately as being in stage C (Feinstein et al, 1985
; Poller, 2000
). This would have the effect of apparently improving survival in patients in stage B by excluding patients with a worse prognosis. At the same time, these patients might improve survival amongst Stage C patients, as they are likely to be better prognostically than those graded as Dukes' C by pathologists who do not make such an effort. It can be expected, therefore, that the number of lymph nodes identified by a pathologist following surgery for colorectal cancer predicts long-term survival in both stage B and stage C disease. In our study, however, the pattern is seen in the population as a whole, and is similar within each individual stage group, although significance is reached only in Dukes' B and Dukes' C groups. Several alternative explanations for these findings need to be considered.
First, stage migration. This is the conclusion reached by most authors who have published on the subject, and is a perfectly acceptable explanation for the relationship observed between number of nodes removed and prognosis in Dukes' B and C disease seen in this study. However, had IQ been measured in the combined population of Oklahoma and California this phenomenon could not have been observed, nor become the subject of a joke: there would simply have been no change in IQ associated with population movement. Our study shows a significant survival advantage related to number of nodes removed across the whole study population, without division into stages. This cannot be explained by stage migration alone, as any artefact brought about by incorrect staging is eliminated when all stages are considered together. Stage migration may still contribute to the relationship seen within individual stages, however, except for one: it is interesting to note that a relationship between number of nodes removed and survival, similar in both direction and magnitude to that seen in Stage B and C patients, is also seen in Stage D patients, although this does not attain statistical significance (P=0.068). Type II statistical error is the most likely explanation for this, based on the smaller group size. However, as node status misclassification cannot influence whether a patient is placed in stage D, patients being classified thus on the basis of presence of distant metastases, this finding supports, albeit softly, our contention that stage migration cannot be the sole explanation for our findings.
If the artefactual explanation offered by stage migration cannot completely explain our results, we must consider other means by which the number of nodes identified by the pathologist could impact upon survival. Several suggest themselves. First, taking out more lymph nodes may reflect superior technical ability on behalf of the surgeon. Second, the identification of lymph nodes containing tumour may change the management of the patient in some way (e.g. by the giving of adjuvant chemotherapy). Third, it may be that the number of lymph nodes present may relate to other factors affecting survival, principally geographical site of the tumour. Last, lymph node enlargement leading to easier harvesting may in itself be a prognostic factor in the patient's outcome.
The first explanation is the most intuitive: surgeons removing a specimen containing more lymph nodes are performing a more radical and technically accomplished operation than those who have a small harvest of lymph nodes, and the impact of the surgeon on outcome in patients having colorectal surgery is well reported (Meagher, 1999
). The traditional and radical approach to colon cancer is to remove the tumour with all its associated nodes up to and including the level of the principal arterial supply, and it is self evident that the more of the mesentery that is removed along with the tumour the more nodes there will be that are available for pathological harvest. However, most surgeons will aim to carry out the same operation on each similar patient they treat, whereas it is apparent from our data that the difference in numbers of nodes removed within
a surgical firm (i.e. between patients) is greater than that between
surgical firms. Mathematically, the average number of nodes taken out by a surgical firm is not a significant predictor of survival in this multi-variable model, whereas the advantage conferred by the individual number of nodes removed remains.
Another explanation concerns changes in management brought about by the finding of involved lymph nodes. However, participants in this study were recruited between 1991 and 1994. During that period only a minority of patients in the UK, even in Dukes stage C, had postoperative chemotherapy in the hospitals concerned, and receipt of chemotherapy did not contribute significantly to our regression model.
It is possible that number of nodes is associated with site of tumour, itself a prognostic variable in colorectal cancer. Right-sided tumours have worse survival, principally due to late presentation. Also, rectal tumours are thought by many to have a poorer prognosis (stage for stage) than colonic cancers, and some authors have reported that there are fewer nodes found around the rectum than around other segments of the colon (Maurel et al, 1998
). Again, however, site of tumour did not contribute significantly to our regression model.
Another possible explanation is based upon the observation that reactive lymph node enlargement is itself a prognostic factor in colorectal carcinoma (Pihl et al, 1980
). The number of nodes found within a specimen relies on two principal factors: first, the amount of tissue, particularly mesenteric, taken out by the surgeon and, second, the number of nodes that the pathologist is able to find in the specimen. The first of these we have already found not to be a significant predictor of survival in this model. The real explanation for our findings may relate to how easy it is for the pathologist to locate lymph nodes. Tiny lymph nodes in a fatty mesentery can be extraordinarily difficult to locate, whereas large nodes in a thin mesentery are comparatively easy to find. Lymph nodes enlarged because of malignancy are more easily found but, in addition, it is probable that uninvolved nodes enlarged because of reactive change are also more easily found. It is known that immune response to colorectal cancer affects survival: patients who have tumours with a pronounced lymphocytic infiltrate have a good prognosis compared with those who have no such infiltration (Jass, 1986
; Fernandez-Acenero et al, 2000
; Murphy et al, 2000
; Canna et al, 2005
). We have shown here that the number of lymph nodes found by the pathologist is increased in those with prominent lymphocytic infiltration of the primary tumour. It may be, therefore, that the number of lymph nodes found is a function of their reactive enlargement. Recent UK guidelines suggesting that a minimum of 12 nodes be examined may be easier to implement in some patients than in others, therefore (National Institute for Clinical Excellence, 2004
). It may also be time to consider whether to include an assessment of the degree of lymphocytic infiltration into the primary tumour as part of the minimum data set.
In conclusion, we have demonstrated that the number of lymph nodes harvested in patients with curable colorectal cancer correlates with a better survival, and also with immune reaction to tumour. It is likely that some of the survival advantage noted in this study is due to stage migration, and we do not doubt the value of impeccable surgical technique and pathological examination. However, stage migration alone cannot explain our results, and it may be that biological predetermination is a key factor in differences in patient survival from colorectal cancer (Macdonald, 1951
; Jass, 1999