Reply

Sir, We appreciate the kind attention and interest in our recently published paper (1), by Dr A. D. Tsakok, who presented some contradictory opinions regarding the statistical analysis methods used in our manuscript. According to the content of Dr Tsakok's letter, it was claimed that we failed to detect some significant differences in the peritoneal fluid (PF) concentrations of cytokines and ovarian steroids as presented in Tables II and IV, and some incorrect significance in the PF concentrations of these parameters as presented in Table III were pointed out by applying the Tsakok technique. Finally Dr Tsakok suggested that our data need to be reanalyzed and the conclusions reexamined. First, we would like to emphasize that the results described of our paper are the representation of our initial pilot study, which comprised a limited number of patients in the study and control groups. We are now in the process of collecting a large number of samples based on a multicenter study to confirm our initial data. Second, after reviewing an immense amount of literature from Medline, we realized that simple statistical analysis is the best way to come to a conclusion about the significance of any result between groups or among groups. Accordingly, we followed the best and most widely acceptable application of science to obtain a statistical significance by the nonparametric method between groups, such as the Mann–Whitney U-test or the Student t-test, and the Kruskal–Wallis test to normalize the distribution of data among three or more than three groups. In fact, we did not consider only the means of our data; rather we input all raw data of the variable PF concentrations of our parameters into the computer and compared their significance between groups or among groups using Statview software. It is quite reasonable to speculate that two distributions are unequal if their means are the same but not their variances. It is also true that the Mann–Whitney test and the Kruskal–Wallis test compare the entire distributions associated with the clinical groups and not just their means. As Dr Tsakok mentioned, we did not compare our results and did not elaborate the significance by means. We also agree with Dr Tsakok about the problems of comparing the means of normal populations with unknown variances. We appreciate his kindness in informing us of his new statistical method, which can easily resolve this problem (2,3). We also thank him for clarifying some undetected significance in the PF concentrations as shown in Tables II and IV. Generally speaking, the Kruskal–Wallis test is the ideal method to determine the difference among three or more than three groups and the Steel–Dwass multiple comparison method is suitable for further comparison between the study and control groups and also between the study groups. Although not described in our paper (1), we applied this technique to justify the significance of our results as presented in Table III. Interestingly, the significance of the results of Table III was unchanged as mentioned. Although currently we are unaware of the proposed Tsakok technique, we will be glad to apply that method in the future. Again, we thank Dr Tsakok for critically commenting on our results and for his valuable advice. But we also believe that the results of any clinical or basic paper should be analyzed by the method that is generally applicable, acceptable and widely appreciated.