Patients were obtained from the TME trial, a large multicenter trial in the Netherlands, in which 1530 patients were included from January 1996 till December 1999 [23
]. In this randomized trial the additional value of preoperative radiotherapy (5 × 5 Gy) is studied when TME (Total Mesorectal Excision) surgery is applied. Radiotherapeutical, surgical and pathological procedures are standardized and quality controlled [24
For the analysis of the prognostic value of the amount of inflammatory infiltrate as well as eosinophilic granulocytes, we analyzed the data from all eligible patients in this trial (n = 1416). We selected 160 patients for the analysis of the cellular composition of the inflammatory infiltrate in relation to prognosis. Since local recurrence rates are very low when TME surgery is applied, an artifical selection was made to be able to study the role of inflammatory cells in both local and distant control. 40 stage II patients and 40 stage III patients without distant metastases or local recurrence (after a median follow-up of 21 months), 40 with distant metastases without local recurrence and 40 patients with local recurrence without metastases were selected. This selection of patients implies that survival percentages in this study cannot be extrapolated towards the general patient population, however, we can compare the effects of the different cell types on patients' prognosis. Patients were selected from both randomization groups. Analyses were performed using stratification for the randomization arms to exclude a possible therapy related effect.
Follow up of all patients in the trial for the endpoints was conducted according to the trial protocol for at least 36 months. Median follow up at the moment of writing was 19.6 months. Median follow up of the selection of 160 patients is 35.4 months. Distant and local recurrences were checked by a radiation oncologist (CAMM) by radiological and/or histological confirmation.
Tissue samples of the primary tumors were fixed in 4% (v/v) phosphate buffered formalin, dehydrated and embedded in paraffin. Tissue sections of 4 μm were cut, mounted onto glass slides pretreated with 2% 3-aminopropyltriethoxysilane (Sigma) and dried overnight. Serial sections were stained with hematoxylin and eosin or processed for immunohistochemistry.
Evaluation of peritumoral inflammatory infiltrate and eosinophilic granulocytes
Routine hematoxylin and eosin-stained histopathological sections were used to determine the infiltration of lymphocytes and eosinophils at the advancing border of the tumors for all patients by pathologists of the Pathology Review Committee (PRC). Tumor slides of all patients in the trial were send to the PRC, who examined them using the following categories for inflammatory infiltrate: none/few and extensive, as described by Jass et al. [20
]. Eosinophilic granulocytes were scored as none/few, moderate and extensive.
Determination of the various inflammatory cells was assessed by immunohistochemical investigation with the following antibodies: anti-CD3 (Anti Human T Cell, 1: 1600, Dako A/S Glostrup, Denmark), anti-CD4 (1: 400, Novo Castra, Newcastle, UK), anti-CD8 (1: 3200, Novo Castra, Newcastle, UK), anti-CD56 (Anti-N-CAM-16, 1: 3000, Becton Dickinson, San Jose CA, USA), anti-CD68 (Anti Human Macrophage, 1: 6400,Dako A/S, Glostrup, Denmark), AA-1 (Anti Human Mast Cell Tryptase, 1: 3200, Dako A/S, Glostrup, Denmark), elastase (Anti-Human Neutrophil Elastase, 1: 800, Dako A/S, Glostrup, Denmark), EG-2 (Anti-Human ECP/EPX, 1: 1000, Pharmacia Upjohn, Uppsala, Sweden). In brief, sections were deparaffinized in xylene, and rehydrated. Endogenous peroxidase was blocked with 0.3% hydrogen peroxide in methanol for 20 minutes. Antigen retrieval was performed using 0.01 M citrate buffer (anti-CD3, Ki-67), 1 mM EDTA (anti-CD8, anti-CD56, anti-CD4), and 0.1% Trypsine (anti-CD68), respectively. Subsequently, sections were preincubated with 1% bovine serum albumin. After overnight incubation with the primary antibody, the secondary biotin-conjugated antibody and a tertiary complex of streptavidin-avidin-biotin conjugated to 3-amino-9-ethyl-carbazole were applied. Finally, the sections were counterstained with Mayer's hematoxylin. Incubation with phosphate-buffered saline instead of the primary antibody served as a negative control.
Scoring methods and categories
The 'running mean' method was used to determine the minimum number of high power fields (HPFs) to be scored for the result to be reliable. Two independent investigators (IDN and AAMS) performed this determination on three different tumors. This was done by calculating stepwise the mean number of inflammatory cells for the total of HPFs until the difference between the successive means became negligibly small. The number of HPFs varied between 11 and 15 for the various cell types; therefor, it was decided to count 15 HPFs (2.1 mm2) for every cell type.
The scoring was performed within the tumor (intratumoral infiltrating cells, present in the stroma of the tumor), as well as separately along the invasive border of the tumor in the surrounding tissue (peritumoral infiltrating cells), by one investigator. Necrotic areas were avoided.
After analysis of the distribution of the numbers of inflammatory infiltrating cells, the counts were divided into two or three categories: none/few, (moderate) and many, based on the distribution of the numbers of cells in the study population (see table ). Because of the high numbers of peritumoral macrophages, the peritumoral macrophage infiltrate (CD68) was scored semiquantitatively by three independent investigators (IDN, AAMS, and JHJMvK), using three categories: none/few, moderate and extensive. In general, the agreement of the three investigators was good (82.5% agreement). In a few cases with poor agreement, the cases were reviewed using a multi-headed microscope and a score agreed after discussion.
Categorical arrangements for the different cell types in 160 rectal cancer patients.
Effects of radiotherapy
Since we observed differences in cell counts between both randomization groups [Nagtegaal ID, Marijnen CAM, Klein Kranenbarg E, Mulder-Stapel AA, Hermans J, van de Velde CJH, van Krieken JHJM, et al., manuscript submitted], we did a stratified analysis of all data to minimize possible therapy related effects. Neutrophils and T cells were decreased in the radiotherapy group. There were no differences between both groups in numbers of mast cell, eosinophils, NK cells and macrophages.
Differences between groups were obtained using the unpaired Student's t-test. Relations between various parameters were analyzed using Χ2 -methods, Mann Whitney tests and bivariate Pearsons' correlation analysis. Univariate survival analyses of time to local recurrence, distant metastasis or death were performed using the Kaplan Meier method, with the time of surgery as the entry date. Differences in observed survival between groups were tested for statistical significance using log ranks tests. The variables with a significant impact in the univariate analysis (p < 0.05) were further examined using the Cox proportional hazards regression model; the forward step-wise elimination method was used for this multivariate analysis. Next, the prognostic value of the selected cell types was analyzed additional to the TNM classification and the presence of inflammatory infiltrate using the enter method. Data were analyzed using the SPSS software package (SPSS 9.0 for Windows, SPSS Inc., Chicago, Illinois, USA).
All statistical analyses were stratified for randomization arm. At p < 0.05, differences were considered statistically significant.