Lymph node density as a surrogate marker for positive lymph nodes

Abstract

British Journal of Cancer (2011) 104, 221–222. doi:10.1038/sj.bjc.6606000 www.bjcancer.comPublished online 23 November 2010& 2011 Cancer Research UKSir,We read with interest the paper by Polterauer et al (2010) inwhich they evaluated the prognostic value of lymph node density(LND) in patients with lymph node-positive cervical cancer. LNDis defined as the ratio of the number of metastatic lymph nodesto the total number of lymph nodes removed. They showed thatLND410% is associated with an impaired disease-free and overallsurvival. Lymph node involvement and lymph vascular spaceinvolvement have always been noted to be a poor prognostic factorin cancer of the cervix (Pecorelli, 2006), and thus this confirms thefact that the higher the number of lymph nodes involved,the worse the prognosis of the patient becomes. The benefit ofusing LND could be that it also incorporates the extent of surgicalstaging. Indeed, it is probably true that when you perform acomplete lymph node dissection instead of a sampling you havemore chance of finding all involved lymph nodes. However, thereis more to it than positive lymph nodes and surgical skills.Although the authors gave the mean/median number ofremoved lymph nodes and positive lymph nodes for all patients,we were not able to discern this information for the LNDp10% orLND410% group separately. We believe that this information ispivotal in a study that examines the ratio of both parameters. In aratio, both the numerator and denominator have an equal role. Theauthors state that they performed a systemic lymphadenectomy inall patients. This means that the mean/median number of removedlymph nodes cannot be statistically different in both groups. As aconsequence, only the denominator will become important, andthus the number of positive lymph nodes will become theprognostic factor. If for any reason the total number of lymphnodes removed in both groups is different, then one must questionthe reasons why. Did the surgeon stop the operation prematurelybecause during the operation he/she discovered bulky involvedlymph nodes? This would make a complete systemic lymphade-nectomy redundant, as the patient is already known to be lymphnode positive based on these few lymph nodes. This is generallyaccepted as an indication for (chemo-)radiotherapy, and thecombination of a radical surgery and pelvic (chemo-)radiotherapywill increase therapy-related morbidity (Quinn et al, 2006). Maybethe pathologist stopped looking intensively for other lymph nodesbecause he/she already found several positive lymph nodes.Second, with respect to the technique used, they referred to aprevious study in which a laparoscopic pelvic lymph node stagingwas described (Polterauer et al, 2008). In this study, patients wereincluded between 1995 and 2007 and, remarkably, the mediannumber of lymph nodes was 15, being more than 3 lymph nodesless than in the present study. Does this implicate a learning curverelated to the laparoscopic technique used? If yes, once more asurgical bias is introduced.Third, we question the cutoff value and how it was established.The authors mentioned that this was based on preliminary datafrom a study by Ooki et al (2007). We wonder how welloesophageal cancer can be compared with cervical cancer. Bothorgans are located in completely different anatomic regions, with adifferent lymph vessel drainage system and lymph node distribu-tion. Did the authors try different cutoff points or was the 10%cutoff the only value examined? If the authors tried different cutoffpoints, one could argue that the authors should have used anindependent validation set to validate this cutoff point.Fourth, with a median of 18.5 lymph nodes (range 12–27),Polterauer et al demonstrated that they routinely performedcomplete lymph node dissections. As the ratio of the mediannumber of involved lymph nodes (n¼2) to the median number ofremoved lymph nodes (n¼18.5) was 10.8%, removing one or twolymph nodes would more or less determine whether the patientbelonged to the category of patients with LNDp10% or LND410%. This would implicate that the surgeon could influence towhich prognostic group the patient will belong. After all, it sufficesto remove more lymph nodes to shift the patient from theLND410% group to the LNDp10% group. It is contradictory thatthe surgeon has to remove more ‘healthy’ lymph nodes to‘improve’ the prognosis of the patient. With more healthy lymphnodes the denominator will become bigger and thus the LNDsmaller. This strikes us as ironic.In conclusion, as LND is influenced by surgical technique,anatomic circumstances and the quality and accuracy of thepathological analysis, we believe that LND is not an objectiveparameter and should not influence the decision on what kind ofadjuvant treatment should be given to a patient. We believe thatthe number of positive lymph nodes is probably the true predictor,but as the number of patients in each of the different groups (1 vs 2vs 42) was low, the authors were unable to reach significance.Using a ratio solved this problem by creating two arbitrary groupsof patients and by using denominators that enlarged the differencePublished online 23 November 2010 between the numerators.*Correspondence: Dr T Van Gorp; E-mail: toon.van.gorp@mumc.nl