Many studies have already demonstrated the prognostic value of lymphoid infiltration (LI) in CRC
[3–5, 11–13]. However, LI evaluation is still not used in daily practice to determine whether a patient should receive adjuvant treatment. Automated counts on virtual immune-stained slides from patients with CRC have already been investigated in other studies
[12, 13]. We developed an automated method of LI evaluation suitable for use in routine clinical practice.
Our automated method has several specific features accounting for its robustness and high level of reproducibility. The size of area analyzed (more than 14 times the area of a TMA core) limits the impact of count variability due to the heterogeneity of LI. Furthermore, taking into account the demonstration, by Gallon et al., of the prognostic value of lymphocyte density in a region adjacent to the tumor, the method developed here includes both tumor tissue and the surrounding non tumor tissue
. Finally, we decided to focus on the variation of lymphocyte density across the invasive front, from the non tumor region to the centre of the tumor, rather than obtaining a mean value for one or two areas. Interestingly, we were able, with a sample of 117 patients with stage II or III CRC, to identify three different curve patterns and to associate these patterns with long-term outcome. One of the original findings of this study was the presence of a peak of lymphocyte density around the invasive front of the tumor in the curves or more than half (61%) the study population. This finding is reminiscent of the high lymphocyte densities found at the invasive margin of colorectal liver metastases, which are also associated with a favorable outcome
Lymphocyte density has been assessed by several methods, including the counting of lymphocytes on H&E-stained sections and immunostaining. It has been evaluated on TMA and on whole slides
[5, 11–13, 21]. Immunostained slides have been evaluated by a pathologist
 or by dedicated software. A recent study confirmed that the results of automated counting of tumor-infiltrating lymphocytes based on image analysis are closely correlated with the semi-quantitative results obtained by a pathologist
. Software can be used to evaluate the percentage of the study area that is immunoreactive
 or to identify lymphocytes by segmentation
. Such segmentation-based analysis may prove difficult if the lymphocytes are located close together
. We obtained identical patterns for analyses of the percentage of the area displaying immunoreactivity and for analyses based on the identification of positive cells.
In most studies, receiver operating characteristic (ROC) curve analysis, with survival as the end-point, is used to determine the cutoff values. This method is problematic in that the mean values obtained cannot be used for patients of other series or in routine practice. This drawback does not apply to LQLI patterns, which are not dependent on the definition of a single threshold value. The LQLI method had a better prognostic value than counts based on the use of TMA technology, and the results obtained were independent of the analysis area. The LQLI method was also feasible for all tumor subtypes, provided that the invasive front was included in the paraffin-embedded block. The only processes requiring direct human intervention in the LQLI method are the positioning of the rectangle for analysis on the virtual slides and determination of the curve pattern. Interobserver reproducibility was excellent (kappa = 0.93).
Microsatellite instability has been shown to be predictive of a favorable outcome, even in the absence of adjuvant therapy, in colorectal cancer. Tumors displaying microsatellite instability are known to have rich lymphoid infiltrates
[24–26]. It is therefore important to consider microsatellite status when assessing LI and prognosis. In our study, only one of the 10 patients with MSI-positive tumors had a low density of lymphocytes both in and around the tumor (pattern 3). CD3+ cell density was higher in the MSI+ group than in tumors without MSI, and the other nine patients displayed LI patterns 1 and 2, but these differences were not statistically significant. No relapses were recorded in the MSI-positive group, but MSI status was not identified as significantly predictive of disease-free survival, possibly due to the small number of patients with MSI-positive tumors.
Another important feature of LI is the location of lymphocytes within the tumor. Indeed, the infiltration of cancer cell networks by CD8+ T cells has been reported to be of prognostic value in colorectal cancer
. Kayser et al. have shown that stromal T cells affect survival in patients with non-small cell lung cancer
. We did not investigate the prognostic significance of the location of the lymphocytes within the tumor (stroma versus epithelial tumor cells) with our method. Little is currently known about the precise role of lymphocyte location and its impact on prognosis, and further investigations and confirmation in a different subset of solid tumors are required before this method can be adopted as a standard approach.
By contrast, a number of studies have investigated the nature of the lymphocytes infiltrating tumors. Galon et al. analyzed four lymphoid markers (CD3, CD8, CD45R0, granzyme B) and found that combined analysis of CD3+ and CD45RO+ lymphocyte densities in the center of the tumor and the invasive margin had the best prognostic value
. In an independent series, FoxP3 was the only prognostic marker identified, with CD45RO and CD8 significant only in univariate analysis
. Both these studies were performed on tissue arrays. The innate immune response may also play an important role in tumor control, as shown by the prognostic value of the density of tumor-infiltrating NK cells and macrophages
. Moreover, the respective proportions and the location of the various leukocyte subpopulations with respect to the tumor may be important, as already demonstrated for breast carcinomas