21st October, 2009
Many thanks for your suggestions. Enclosed is our point by point reply.
Reviewer number: 1
The Authors explored the hypothesis that an elevated pre-operative LNR and
related parameters could represents a prognostic markers of both recurrence
and survival rates after radical resection for oesophageal cancer. With this
purpose they analysed their respectful experience on 294 patients who
underwent oesophagectomy from June 1997 and May 2008.At this regard there is a possible mistake in the “Materials and Methods” section when the Authors reported that last follow up for survival analysis was set at October 2007.
We apologize for the mistake and have amended the manuscript accordingly. The patients who underwent oesophagectomy from June 1997 to September 2007 were included rather than May 2008. However, the survival was calculated at October 2007 as previously said.
But the most important concern in the reviewers’ opinion is on the type of analysis, or, at least, in the way of results presentation. It would be expected that the Authors presented their data according to the classical form of univariate and multivariate analyses where all possible prognostic parameters were inserted in to the model (age, gender, TNMG and related stages, LNR ….) resulting clear which were significantly associate with either recurrence or survival. On the contrary they preferred to report these features in regard to LNR separately.
Multivariate analysis on recurrence and survival has been done including only the clinically important variables which were related to our aims of the current study
Moreover in table 1 data related to recurrence rate and median follow up are missed, as well as it is not clear whether overall survival is reported as a mean and not in the usual way as a median.
The table 1 has been revised taking into account all your comments. The over all survival has now been reported as median.
In addition, in the paragraph of NLR and preoperative chemotherapy many parameters are reported as different but statistical significance was never quoted as well as p value was never reported.
All these comments have been taken into account. Please see the revised manuscript under the subheading of NLR and preoperative chemotherapy
Reviewer number: 2
Major compulsory revisions
The authors present a retrospective analysis of patients with oesophageal cancer who underwent surgery over a 10-year period. They aim to assess the prognostic value of pre-operative neutrophil-lymphocyte ratio in this group of patients. They report that NLR has no prognostic value in this cohort. Looking at the Kaplan-Meier curve in Figure 4, the curves are diverging at 40 months and I wonder if the lack of significance may reflect lack of statistical power rather than a true absence of effect. The authors do not provide enough data to allow a reader to judge this.
It is a retrospective analysis and the limitation was lack of statistical power with regards to effect NLR on survival. For observing significant effect on NLR on survival it will require 1600 deaths in total of 2600 patients which is rather a very large number for oesophageal cancer and in future we intend to collaborate with other units and perform the study with big numbers.
We have incorporated in the revised manuscript that this study does carries limitations of a retrospective analysis.
The statistical analysis is very confusing for the reader, as it jumps between
Kaplan-Meier, Cox-regression and univariate chi-square analyses of survival,
and recurrence. Moreover, the censored patients in the Kaplan-Meier analyses represent patients still alive. Conventionally, patients are censored when the event of interest occurs (death or recurrence) or when the patient is lost-to follow-up. The primary outcome of the study is either death or recurrence so it would be more straightforward to use univariate Kaplan-Meier analysis of overall and recurrence free survival. This would also account for the differences in duration of follow-up that a chi-square analysis cannot correct for. Significant predictors of survival should then be entered into a Cox-regression model to test for independence. It is not clear from the manuscript whether a statistician has been involved. If not, I think the authors should seek some statistical input. Detailed results of the univariate and multivariate analyses should be presented in a revised manuscript. Halazun et al (Ann Surg 2009;250:141-51) provide an excellent example of the detailed results that need to be presented.
The proposed statistical analysis has been used.The significant predictors of survival have already been mentioned into a cox regression model.
As far as the next comment is concerned the confusion arises on the survival graphs (Figures 4, 5, 6 and 7), which were copied from SPSS output and the alive cases were mentioned as censored. This has now been corrected on the graphs and the censored cases represent deathand the crosses are the alive cases.
Regarding the patient cohort, group 3 (the miscellaneous cancer sand GISTs) is so small (6 patients) that the results for this group are meaningless. These
patients should be excluded.
This group of patients has been excluded from the analysis (please see the revised table 3 and NLR in cancer subsets). We have included all the patients who underwent surgery for oesophageal cancer and hence this group was included in our cohort to represent the complete cohort of patients.
The use of pre-operative chemo may have confounded the results, especially as they had a lower NLR. The pre-treatment NLR for this group may have yielded quite different results. It may be worth running survival analyses with the chemo group excluded and using the pre-treatment NLR in this subgroup to see whether there is still no effect.
The pre-chemotherapy NLR of the patients is not available for the analysis. Hence the above mentioned analysis was not performed, .However, after exclusion of patients who had chemotherapy the NLR did not have an effect on survival in patients who did not receive any chemotherapy with p value of p=0.437
“NLR has been identified as a predictor of outcome in patients undergoing
potentially curative resection for other gastrointestinal cancers, including
hepatocellular and colorectal carcinoma” – supporting references are needed
Supporting references have been provided. (Please see the last paragraph in the introduction section on page 5 and also highlighted in red)
Under ‘Calculation of neutrophil-lymphocyte ratio’ delete either ‘routinely’ or
‘routine’ from the sentence “Routine full blood count (FBC) results were collected routinely as part of standard diagnostic and pre-operative protocols”
The suggested changes have been made in the revised manuscript.
(Please see in the material and method section under the subheading of calculation of neutrophil lymphocyte ratio)
Why would NLR correlate with either total lymph node yield or ratio of involved to total node yield?
The result section on NLR and nodal status has been rephrased to clarify the objective to correlate the NLR with known prognostic marker in the form of lymph node yield and involved nodes. Please see the result section under the subheading of NLR and nodal status on pages 8 and 9 and also highlighted in red.
“In addition, adjusting for NLR, there was no difference in survival for patients
who had received preoperative chemotherapy as compared to those without
neoadjuvant chemotherapy (p=0.0280, Cox regression analysis with interaction term)” This statement is contradictory – the p-value suggests a significant difference
There has been a typing error where I have written p value as 0.0280, while it should have been asa non-significant value of p=.280. Hence the correction has been made. Please see the result section under the subheading of NLR and preoperative chemotherapy on page 10and also highlighted in red.
“In colorectal cancer, in comparison with oesophageal cancer, the colonic lumen has a higher bacterial load which may induces local inflammation. Colorectal tumours have an efficient absorptive microvascular network, which may become more permeable as a result of neoplastic change”
.“Patients with perforated tumours have a poor prognosis not only because of the mechanical dissemination of tumour into the peritoneal cavity, but also because theinflammatory response produced may lead to a depression of anti-tumour immune responses”
These were hypothetical suggestions and will require laboratory based research to prove it and hence were not referenced, so to avoid the confusion these have been deleted from the revised manuscript.
“Immunosurveillance for cancer fails as humans age, and this may also explain changes in neutrophil and lymphocyte counts in oesophageal cancer,
predominantly a disease of older patients”
These statements need to be supported by appropriate references.
The above mentioned phrase has been referenced with the following references
1.Finch CE, Crimmins EM. Inflammatory exposure and historical changes in human life-spans. Science. 2004 Sep 17;305(5691):1736-9.
2.Krabbe KS, Pedersen M, Bruunsgaard H. Inflammatory mediators in the elderly. Exp Gerontol. 2004 May;39(5):687-99.
Table 1 categorises NLR using 2.14 as a cut-off. This value is not mentioned
anywhere else in the paper. Why is it used here instead of 3.5?
The table 1 has been revised and the cut-off of 2.14 has been deleted to avoid the confusion. (Please see the revised table1)
We will appreciate if the revised manuscript be considered for publication in
Mr Farhan Rashid,
Division of GI Surgery
Clinical Sciences Wing
Uttoxeter Road, Derby
Mob no: 07817168102