To the Editor:
In their recent report, Oudejans et al1 studied 80 lymph node biopsies from patients with Hodgkin's disease (HD). In 10 biopsies, greater than 15% of the T cells contained granzyme-B (GrB). Using the Cox-proportional hazards model for multivariate analysis, the investigators found the percentage of GrB+ lymphocytes to be the strongest prognostic marker in patients with stage II disease (n = 44). Against what one may have expected, the investigators concluded that the presence of a high percentage of activated cytotoxic T cells in biopsy material of HD patients is a strong indicator for an unfavorable clinical outcome.
Many issues in this study raise the possibility that this conclusion is premature, if not inaccurate. (1) The study was performed on selected specimens and was not performed prospectively. No information was provided on what kind of therapy those patients received. (2) Relapsed patients were known before the immunohistochemical staining studies. Was the interpretation performed blindly? (3) How many slides from each lymph node were studied and were the results concordant? (4) Was there any difference between GrB+ cells in biopsies at diagnosis and at the time of relapse? (5) Finally, and most importantly, the investigators used 9 variables in their multivariable assessment to analyze 80 patients, of whom 12 died of disease (only 1.3 events per variable). The rule of thumb in the Cox model is to have at least 15 to 20 events for every additional variable in the model.2 It is more worrisome that patients with stage II (44 patients with only 5 disease-specific deaths) were also analyzed for 9 variables and the prognostic significance of GrB+ cells was made based on this inappropriate analysis.
Based on the data presented in this study,1 the only conclusion that can be made is that the majority of HD biopsies (70/80 [88%]) did not contain a high percentage of activated T cells. This low incidence of infiltrating activated T cells may explain how Reed-Sternberg cells can survive within the high number of T cells. This finding is important, especially now that we have shown that Fas ligand expression is also decreased in T cells of HD biopsies.3
To the Editor:
In response to Dr Younes' comments concerning the validity of the conclusions drawn by us in our recent report,1-1 we would like to add the following.
Our conclusion that a high percentage of activated CTLs is strongly related to a poor clinical outcome is suggested to be inaccurate by Younes on the basis of an inappropriate statistical procedure. However, in contrast to his suggestion, we came to our conclusion primarily on the basis of univariate analysis, using the log-rank test. This test clearly showed that, irrespective of stage, the presence of greater than 15% activated CTLs is strongly correlated with a poor clinical outcome (P < .0001). The Cox-proportional hazards model was used to show that, in stage II patients, the presence of many activated CTLs was a prognostic marker independent from clinical parameters as age or the presence of B symptoms. It was not mentioned clearly in our article that, when only the three variables that were significant in univariate analysis (age, presence of B symptoms, and percentage of activated CTLs) were included in the Cox model, identical results were obtained. There is therefore no “violation” of any rule of thumb.
Our study was performed retrospectively on a group of patients receiving different therapy regimens due to the fact that patients participating in different clinical trials were entered. Therefore, our study should be repeated prospectively in a group of patients, all participating in a single clinical trial.
Furthermore, the percentage of activated CTLs in primary tumors could be compared with biopsy specimens of relapsed tumors in only 7 cases. These numbers are too small to draw any (preliminary) conclusions. In addition, the immunohistochemical studies and the quantification of the number of activated CTLs were performed before analysis of clinical data. We thought that this was so obvious that it was not necessary to mention it.
Identical results were obtained when different slides from one lymph node were investigated. Moreover, the computer-assisted quantification of numbers of activated CTLs included a random selection of fields, excluding the possibility that only selected areas of the biopsy specimens were analyzed.
Finally, Younes suggests that our data merely strengthen their notion that the absence of many activated CTLs in the majority of cases explains why Reed-Sternberg cells can survive within the presence of a high number of T cells. This hypothesis would imply that cases with many activated CTLs would have a relatively favorable clinical outcome. However, our data strongly indicate that rather the opposite is true.
In our opinion, our results are best explained by the hypothesis that, in cases with many activated CTLs, only those Reed-Sternberg cells can survive that are resistant to apoptosis induced by activated CTLs, but also to apoptosis induced by either radiotherapy or chemotherapy. Functional studies should be performed to investigate whether in these cases the Reed-Sternberg cells are indeed more resistant to apoptosis than Reed-Sternberg cells in cases harboring few activated CTLs.