CD8+ T-Cell Content in Diagnostic Lymph Nodes Measured by Flow Cytometry Is a Predictor of Survival in Follicular Lymphoma

Follicular lymphoma is the most common indolent non-Hodgkin's lymphoma, accounting for 20% to 30% of all adult cases of non-Hodgkin's lymphoma in the western world (1). In contrast to the male predominance seen in other forms of lymphoma, follicular lymphoma is equally distributed between the sexes. Follicular lymphoma arises from B cells in the germinal center of the lymphoid tissue. The t(14;18)(q32;q21) translocation is detected in over 90% of the cases (2). According to the WHO criteria, follicular lymphoma is morphologically characterized by a mixture of centrocytes and centroblasts. Follicular lymphoma is classified into grades 1, 2, and 3, depending on the number of centroblasts, and grade 3 is further subdivided into 3a and 3b, according to the presence of solid sheets of centroblasts (3). There is consensus that follicular lymphoma grades 1, 2, and 3a share clinical features as indolent lymphomas, whereas it is debated whether follicular lymphoma grade 3b should be regarded as an aggressive lymphoma (4–6).

The clinical course of indolent follicular lymphoma is highly individual: some patients have a stable disease for decades, with no symptoms and no need for treatment, whereas other patients present with symptomatic disease, necessitating treatment at the time of diagnosis. Median survival is about 10 years, but the range is wide: from <1 to >20 years (7, 8). There is no consensus regarding the best treatment for follicular lymphoma. Most patients will attain good remissions after treatment with cytostatic drugs or monoclonal antibodies. However, the disease will relentlessly relapse, and for most patients, there is no cure, even with intensive therapy, so a wait-and-watch policy is usually applied. Transformation to diffuse large B-cell lymphoma occurs in 30% to 70% of the patients, worsening the prognosis (9, 10).

Many clinical characteristics are associated with poor prognosis: age, male sex and poor performance status, high erythrocyte sedimentation rate, anemia, elevated lactate dehydrogenase, many involved lymph node stations or extranodal sites, bulky disease, hepatosplenomegaly, bone marrow involvement, and high stage (11–13). Several prognostic indices have been applied, but today, the Follicular Lymphoma International Prognostic Index (FLIPI) is mostly used (14). However, prognostic models based on clinical variables have not been successful in determining the best initial treatment, and they cannot identify the biological basis of the clinical heterogeneity.

In the microenvironement around the follicular lymphoma cells, nonmalignant T cells, macrophages, and dendritic cells are present. These host cells probably interact with the tumor cells. Recently, gene expression profiles of both tumor cells (15) and immunologic host cells (16) have been reported to be of prognostic value, but these techniques are not widely available.

The objective of this study was to analyze and clarify the importance of the T cells that coexist with the follicular lymphoma cells in the lymph nodes. Our hypothesis was that T cells, or a subset of T cells, may affect the outcome of the disease. Using flow cytometry, we estimated the numbers of different T-cell subsets and related these to conventional prognostic factors, need of treatment, time to treatment failure, and survival.

Patient selection. All cases diagnosed de novo with follicular lymphoma between January 1, 1994, and January 1, 2004, at the Department of Pathology of the Karolinska University Hospital, Huddinge, were identified and retrieved. The reason for starting inclusion from 1994 was that by then, flow cytometry analysis had become a routine procedure in the lymphoma diagnostic workup at the department. Out of the identified 219 follicular lymphoma patients, 168 cases had had flow cytometry assays done on pretreatment diagnostic lymph nodes. Cases in which flow cytometries were only from extranodal sites were excluded, because T-cell numbers may vary between different tissues. The cases in which flow cytometry had not been done were due to an insufficient amount or quality of tissue or to inappropriate handling (prior fixation). Those cases did not clinically differ from our study population. All available histologic slides from each of the 168 patients were reviewed by two experienced hematopathologists (B. Sander and B. Christensson). The grade and the diffuse component were reassessed according to the current WHO criteria. Immunohistochemical stainings for CD3, CD4, CD8, CD10, CD20, Bcl-2, Bcl-6, p53, and Ki-67 were available in most cases. Ten cases were found not to be follicular lymphoma (five cases of diffuse large B-cell lymphoma, two of Burkitt's lymphoma, one each of Hodgkin's lymphoma, uncategorized lymphoma, and follicular hyperplasia). Thirteen follicular lymphoma patients were excluded due to concomitant diffuse large B-cell lymphoma, rendering 145 true follicular lymphoma patients without evidence of transformed disease at the time of diagnosis. Finally, we excluded the grade 3b cases, resulting in a cohort of 139 consecutive patients with indolent follicular lymphoma and flow cytometry data from diagnostic lymph nodes.

The Department of Pathology at Karolinska Huddinge serves all outpatient clinics and hospitals in the region of South Stockholm County (population 895,089 as of December 31, 2003), and this cohort is representative for an entire, well-defined region, where lymphoma patients are treated at two hospitals, Karolinska Huddinge (75 patients in the study) and Stockholm Söder Hospital (54 patients). Ten patients had been treated outside the region (three patients at Karolinska Solna, three at St. Göran Hospital, and four at Visby Hospital). Baseline clinical information and prospective updates were gathered from patient charts. Survival status was available for all patients, save a nonresident who returned to South America after a successful first-line cyclophosphamide-Adriamycin-vincristine-prednisone (CHOP) regimen, making this case the only one with short follow-up time (0.6 years). Excluding this patient, the median follow-up time for the surviving patients was 6.8 years (range, 2.8-12.6 years). The last evaluation of all patients was done in September 2006.

This study was approved by the Ethics Committee of Clinical Research, Stockholm.

Treatment regimens. The patients had received a variety of treatment algorithms: wait-and-watch, local irradiation, single alkylators, fludarabine (sometimes in combination with alkylators), anthracycline-containing regimens, autologous and allogeneic stem cell transplantations, and biological therapies (rituximab and IFN).

Clinical characteristics and the FLIPI. Baseline clinical data were gathered from the time of diagnosis, before the start of treatment. The FLIPI risk group was calculated from the baseline Ann Arbor stage, hemoglobin, number of nodal areas involved, lactate dehydrogenase, and age, as previously reported (14). The follicular lymphoma grade and proportion of the diffuse component of the diagnostic lymph node specimens were evaluated in all but 13 biopsies that were too small to allow grading, but large enough to ensure the follicular lymphoma diagnosis. Other baseline clinical characteristics considered to be important for patient outcome were also recorded and evaluated.

Flow cytometry analysis. The phenotypes of the cells in the lymph nodes were examined by flow cytometry using three-color fluorescence according to standard procedures. Briefly, suspended cells from biopsies were washed before mixing with appropriate concentrations of fluorochrome-conjugated monoclonal antibodies to B-cell antigens CD19, CD20, CD22, CD23, CD10, and immunoglobulin κ and λ chains. T-cell markers analyzed were CD2, CD3, CD4, CD5, CD7, and CD8. In addition, cells were analyzed for CD45, CD14, CD52, HLA-DR, CD56, CD16, and CD25. All antibodies were obtained from Becton Dickinson (Mountain View, CA). After incubation for 30 min at room temperature, cells were washed with PBS. For data acquisition and analysis, a FACSCalibur (Becton Dickinson) was used with Cell Quest software (Becton Dickinson). All samples were analyzed by setting appropriate side and forward scatter gates around the mononuclear cell population using CD45/CD14 for gate setting. Consistency of analysis parameters was ascertained by calibrating the flow cytometer with calibrating beads and FacsComp software, both from Becton Dickinson. The results are reported as percents of gated cells positive for each antibody. To ensure that the specimen used for flow cytometry was representative of the tumor biopsy, the sum of the percentage points of CD4+ cells, CD8+ cells, and CD19+ cells had to equal 100 ± 20% to allow the specimen to be kept in the study, which was true in all cases. All flow cytometry results obtained from the pretreatment follicular lymphoma nodes were reviewed.

Statistical analysis. Overall survival (OS) and disease-specific survival (DSS) were used as the major end points. Both OS and DSS were calculated from the date of diagnosis to the date of death or last follow-up. For DSS, data were censored at the time of death if lymphoma was not the primary or underlying cause of death; deaths during lymphoma treatment were always considered lymphoma related. Effect on OS and DSS was analyzed by proportional hazards (Cox) regression (17), and for ordinal predictors, the Mantel-Haenszel log-rank test (18) and the Kaplan-Meier method (19) were also applied. We decided to fit a multivariable Cox model where the lymphocyte subsets found to be marginally significant or better (nominal P < 0.200) after bivariable analysis (adjusted to the FLIPI) would compete against each other, again adjusted to the FLIPI. This we called the “FLIPI model” that was used to identify the T-cell subset(s) with the greatest effect on OS and DSS. Because of the skewed distributions of the lymphocyte subsets, competition in the FLIPI model was repeated with rank-transformed lymphocyte variables. To reduce the numbers of factors competing in the larger multivariable Cox “Total model”, a prior identification of predictors was done with bivariable (FLIPI-adjusted) OS and DSS analysis (except the factors used for calculating the FLIPI; these were univariable tested). All factors with an apparent effect on survival (nominal P < 0.200 in Table 1) then competed with the identified T-cell subset(s). In the Total model, the impact of the identified T-cell subset(s) would thus be adjusted to all important clinical factors. This model includes more variables but fewer patients, because of missing data (cases with missing information were listwise deleted). The proportionality hazard assumption was verified using tests and plots based on Schoenfeld and scaled Schoenfeld residuals.

For a descriptive analysis, the time to transformation was defined as the time between the date of diagnosis and the date of histologically verified transformation (failure), censoring all patients without evidence of transformed disease (patients still alive were censored at the date of last follow-up, and deceased patients were censored at the date of death).

Also, for a descriptive analysis, the time to treatment failure was used for relations between predictors and treatment efficacy. The time to treatment failure was defined as the time between the date of the start of the first-line regimen and the date of the start of the second-line regimen (failure) or the date of lymphoma-related death (also failure), censoring patients alive at the date of last follow-up and patients dead from causes unrelated to lymphoma.

Clinical characteristics. Baseline clinical characteristics of the 139 patients in the cohort are shown in Table 1. The median age was 60 years, and the female/male ratio was 1.2. Seventeen percent of the patients had low hemoglobin, and 38% had elevated lactate dehydrogenase, whereas 73% had stages III to IV disease. The patients were equally distributed in the three FLIPI strata. In total, 54 patients out of 139 had died at the time of last follow-up. Three deceased patients, for whom sufficient clinical data were not available, were not included in the DSS analysis. Out of the 136 remaining patients, 37 had died of a lymphoma-related cause. The causes of the 14 deaths unrelated to lymphoma were dementia (n = 4), cardiac disease (n = 3), and other malignancies (n = 7; two from lung cancer, one each from acute myeloid leukemia, kidney cancer, breast cancer, colon cancer, and esophagus cancer). Estimated median OS for all patients was 10.4 years. The predicted 5-year OS was 70%. Out of the 139 patients, 130 could be stratified into the FLIPI risk groups. The prognostic strength of the FLIPI was confirmed, showing a strong association with both OS and DSS (Table 1; Fig. 1). The nine patients with insufficient data for FLIPI risk stratification had an increased risk of death, similar to patients in the high-risk FLIPI group. In univariable analysis, the clinical characteristics that constitute the FLIPI were all important for both OS and DSS. Some other clinical characteristics seemed also associated with OS or DSS (bivariable P < 0.200 in Table 1) and competed with the FLIPI-constituting factors to fit the Total models for OS and DSS. The follicular lymphoma grade and the proportion of diffuse component had no effect on survival.

Flow cytometry results. The flow cytometry results from the diagnostic lymph nodes, where each subtype was reported as the percent of all gated cells, revealed large variations in the levels of the lymphocyte subsets (Table 2). The T-cell populations were easily distinguishable for all subset markers. In all cases, the phenotypes of the T cells were normal (CD2+, CD3+, CD5+, and CD7+). The median ratio of CD4+/CD8+ was 3.8. Interestingly, upon reevaluation of the flow cytometry assays, a small (generally <1% of the lymphocytes) population of double-positive CD4+CD8+ was observed. As expected, the levels of CD3+, CD4+, and CD8+ cells showed positive linear correlations with one another, and the linear correlations between the T-cell subsets and CD19+ B cells were negative. All investigated lymphocyte subsets seemed associated with OS and DSS after bivariable analysis, but levels of clonal (highest of κ or λ) or nonclonal (lowest of κ or λ) B cells did not have an association with survival (Table 2).

Survival analysis. The levels of CD19+, CD3+, CD4+, CD8+, the CD4+/CD3+ ratios, and the CD8+/CD3+ ratios had bivariable P < 0.200 for either OS or DSS. These factors, together with the FLIPI as a multiple dichotomized variable, were put into competition in Cox forward stepwise regression, where only the levels of CD8+ T cells and the FLIPI remained independently associated with OS. The risk of death decreased with higher levels of CD8+ cells: an increase of one percentage point of CD8+ had a hazard ratio (HR) of 0.90 (P = 0.017; Table 3). To minimize outlier effect, this Cox forward stepwise regression was repeated with rank-transformed values of the lymphocyte variables: ranked CD8+ remained the sole lymphocyte variable significantly associated with survival (P = 0.003). To ascertain minimal missing data bias, these two analyses were also done with the nine patients with missing FLIPI included as a risk group of its own, again showing only CD8+ cells (P = 0.037) and ranked CD8+ (P = 0.008) as the lymphocyte variables associated with OS (n = 137). The clinical characteristics that were marginally significant or better in univariable or bivariable analysis were put with CD8+ in stepwise forward regression to fit the final Total model for OS. The CD8+ level was still significant: HR, 0.86; P = 0.001 (Table 3). Evaluation of DSS was also done. The lymphocyte variables competed in Cox forward stepwise regression to achieve the final FLIPI model for DSS, in which only the FLIPI strata, together with the CD8+ cells, were significant (Table 3). In addition, with rank-transformed values, ranked CD8+ cells were the sole lymphocyte variable associated with DSS (P = 0.002). Including the patients with missing FLIPI, the CD8+ cells remained the only subset associated with DSS (P = 0.020; rank-transformed P = 0.002; n = 134). The final multivariable Total model for DSS contained the same factors as for OS (Table 3). CD8+ cells were also significantly correlated with better 5-year OS and 5-year DSS, independently of the factors in the FLIPI model and in the Total model (data not shown).

Thus, higher CD8+ T-cell levels in the diagnostic pretreatment lymph nodes strongly correlated with better patient survival in both OS and DSS. In the Total model, the hemoglobin value was the only factor, beside the CD8+ cell number, where higher values were associated with better OS and DSS. The variables that correlated with inferior OS and DSS were older age (in years), higher lactate dehydrogenase level (in multiples of upper normal limit), male sex, stages III to IV disease, ≥2 extranodal sites, and dementia. There was no effect modification on survival between the CD8+ cells and the other predictors in the FLIPI model or in the Total model. The equal distribution between dichotomous clinical characteristics and the CD8+ T-cell numbers were investigated with the Wilcoxon-Mann-Whitney rank sum test, and no significant difference could be seen (Table 1). Correlations between ordinal or continuous variables and the CD8+ numbers were analyzed with Spearman's correlation. For one characteristic, there was a correlation: higher levels of CD8+ correlated with a larger proportion of diffuse component (P = 0.026). However, this variable did not have an effect on survival (Table 1).

Risk stratification by the CD8+ cell levels. By dividing the patients in groups according to the number of CD8+ T cells in the 25th and 75th percentiles (4.2% and 8.6%, respectively), three CD8+ strata were created. Thirty-four patients (25%) had <4.2% CD8+ cells (lower quarter), 69 patients (50%) had ≥4.2% and ≤8.6% CD8+ cells (middle half), and 34 patients (25%) had >8.6% CD8+ cells (upper quarter). This stratification correlated significantly with survival [P = 0.045 (OS) and P = 0.029 (DSS)]. The importance of the CD8+ cell strata was more accentuated in the intermediate and high-risk FLIPI patients (Fig. 2). Among patients in the upper quarter of CD8+ cells, there were few deaths, and no survival difference could be seen between the FLIPI risk groups (P = 0.953), and likewise, no survival difference between the three CD8+ strata could be seen among low-risk FLIPI patients (P = 0.957). After adjustment in the FLIPI and Total models, the CD8+ stratification remained strongly correlated with both OS and DSS. Patients with >8.6% CD8+ cells had five times lower risk of death, and patients with 4.2% to 8.6% CD8+ cells had twice lower risk, compared with patients with <4.2% CD8+ cells (Table 4). When including the patients with missing FLIPI risk group into the FLIPI model, the CD8+ stratification remained significantly associated with OS (P = 0.012) and DSS (P = 0.007). Among the 34 patients in the upper quarter of CD8+ cell levels, there were five (15%) lymphoma-related deaths. Two patients died at older age (92 and 80 years), both 7 years after diagnosis. The other three patients had experienced early (<2 years after the follicular lymphoma diagnosis) lymphoma-related deaths, none of which was from indolent follicular lymphoma but all were from transformation to diffuse large B-cell lymphoma. Among the patients in the middle half of CD8+, there were 17 lymphoma-related deaths out of 67 evaluable patients (25%), and eight were early, of which four could be attributed to indolent follicular lymphoma, whereas verified transformed disease was the cause of death in two cases and clinically strongly suspected in two. Among patients in the lower quarter of CD8+, 14 out of 33 evaluable patients had died from lymphoma (42%), and 6 had died early. Of these six patients, four died from the indolent disease, one died from a verified transformation, and one died from a strongly suspected transformation.

CD8+ levels and transformation. Transformation to diffuse large B-cell lymphoma occurring after the diagnosis of follicular lymphoma had been histologically verified in 25 of 137 evaluable patients (18%). There was no association between the time to transformation and CD8+, either with the continuous CD8+ cell levels or with the CD8+ strata (data not shown); nor was there an association between CD8+ cells and risk of early transformation (<2 years after diagnosis).

CD8+ levels and treatment. Ninety patients required early treatment (within 6 months from diagnosis), and these had lower levels of CD8+ cells in their diagnostic follicular lymphoma lymph nodes, compared with the 45 patients who were asymptomatic in the first 6 months. After multivariable logistic regression analysis, the only factors that significantly correlated with early treatment were the CD8+ cell percentage (OR, 0.89; P = 0.011) and the lactate dehydrogenase levels as multiples of the upper normal limit (OR, 15.1; P = 0.006; n = 126). The FLIPI, the grade, and all other factors were not significant. Using the CD8+ stratification, the fraction of patients not requiring early treatment was 27% in the lower quarter, 27% in the middle half, and 55% in the upper quarter of CD8+ cells (χ² P = 0.015). In total, 22 out of 136 evaluable patients remained untreated throughout the study: 7 until death and 15 until last follow-up. The surviving untreated patients had a median follow-up time of 6.5 years (range, 2.8-9.0 years). In the entire cohort, the median length of treatment-free survival was 12 weeks.

A total of 114 of 136 evaluable patients had received treatment for follicular lymphoma. In first line, 17 had been solely treated with irradiation, 13 with single alkylators, and 11 with rituximab monotherapy. Seventy-three patients had combination therapies in the first line; of these, 14 received fludarabine-based regimens, and 59 received anthracycline-based regimens. The Kruskal-Wallis test revealed no differences in the levels of CD8+ cells between patients these five first-line algorithms. In the anthracycline group, nine patients received treatment in combination with rituximab, and seven had additional irradiation. Including secondary and later therapies, a total of 78 out of 136 patients (57%) had received anthracyclines, 37 patients (27%) had received fludarabine, and 52 patients (38%) received rituximab until the date of last follow-up, and the CD8+ cells were equally distributed in these categories. High and intermediate FLIPI were strongly associated with shorter time to treatment failure (P = 0.0001; n = 108), but the CD8+ levels were not.

In this cohort of patients with nodal indolent follicular lymphoma, we have shown that the level of the CD8+ T cells in the diagnostic lymph nodes, estimated by flow cytometry, predicted OS and DSS, both as a continuous and as a stratified variable. Higher levels of CD8+ cells seemed to protect patients from death, independently of all other identified predictive factors.

It has been shown that tumor-infiltrating T cells are part of a highly potent and specific antitumor immune response against certain tumors, e.g., melanoma (20, 21). In addition, studies in non-Hodgkin's lymphoma have shown that the presence of a T-cell response may affect the outcome (22–29). It is well known that follicular lymphoma may spontaneously regress, suggesting immunologic interactions with the tumor cells (7). Tumor vaccinations against the clonal immunoglobulin expressed by follicular lymphoma have also highlighted the potential of the immune system to stall disease progression (30). A large study using gene expression analysis on follicular lymphoma lymph nodes has shown a good prognosis-immune signature with high expressions of genes encoding T-cell proteins such as CD7 and CD8β1, but not CD4 or CD2, and a poor prognosis immune signature with high expressions of genes encoding proteins in macrophages and dendritic cells (16). Moreover, high macrophage content measured by immunohistochemistry has been shown to predict short survival in follicular lymphoma (31). There are thus suggestions of both antitumor immune response mechanisms and tumor-promoting immune response mechanisms. These mechanisms are probably not mutually exclusive. There is also evidence that signals from malignant B cells can induce defects in otherwise normal T cells (32). Recently, it has been shown that T_reg CD4+CD25+ cells with high levels of CTLA-4 and Foxp3 are overrepresented in non-Hodgkin's lymphoma lymph nodes, and that these T_reg cells are recruited to the microenvironment by the lymphoma cells, where they suppress the proliferation and cytokine production of the other T cells (33).

We have investigated the levels of T-cell subsets with flow cytometry assays on lymph nodes in a well-defined and well-characterized large cohort of consecutive follicular lymphoma patients. The median OS times and the distributions of the clinical characteristics in our material are similar to data from other large studies on follicular lymphoma. We could reproduce the predictive strength of the FLIPI, both for survival and for treatment response duration (34), although the latter should be interpreted with caution because the patients had received different treatment regimens in the first line. In multivariable analysis, CD8+ T cells was the only lymphocyte subset important for survival. The levels of B cells (total, clonal, or nonclonal) did not have an association with survival. Thus, the importance of the CD8+ cell level is not a secondary effect due to the numbers of the malignant cells in the specimen. Higher CD8+ T-cell levels in the diagnostic pretreatment biopsy specimens correlated with better OS and better DSS, independently of the FLIPI and independently of all other predictors. We found no evidence that the CD8+ cells had an association with the transformation to diffuse large B-cell lymphoma or to treatment response duration.

On the review of the histologic and immunohistochemical slides, it was observed that the CD8+ cells were mostly perifollicular, whereas the CD4+ cells were seen inside as well as outside the follicles. It could be speculated that these mostly perifollicular CD8+ cells perform their protective action as watchmen of the perimeter or as captors of the disease, halting its progression. There are at least two functional subsets of cytotoxic CD8+ T cells: Tc1, producing high amounts of IFN-γ, and Tc2, producing interleukin-4, interleukin-5, interleukin-10, and interleukin-13 and low levels of IFN-γ (35). Because these subsets express different chemokine receptors, they may have different capabilities of migrating into tumors. Once in the tumor, each subset may perform different effector functions dependent on the cytokines it produces. Furthermore, different kinds of regulatory CD8+ T cells exist (36). We do not know if one of these subsets of CD8+ T cells is of greater importance than the others. We plan to study the relation between outcome and the different subsets of CD8+ T cells with immunohistochemistry.

In our material, male sex and older age were strong risk factors for lymphoma-related death. It could be speculated that the immunomodulatory effects of androgens (37) might be negative in the follicular lymphoma setting. In old patients, both the well-known decreased ability to mediate effective immune responses (38) as well as weakened tolerability to therapy might explain the poor prognosis. However, the CD8+ cell numbers were equally distributed between the sexes, and there was no correlation between age and CD8+ levels. Likewise, CD8+ cell levels did not correlate with any other predictors important for survival.

Higher levels of CD8+ T cells increased the chance of treatment-free survival at 6 months, but it could be that the main effect of the CD8+ cells is in conjunction with treatment. We do not know whether the CD8+ cells could modify the antibody-dependent cell-mediated cytotoxicity involved in the effect of rituximab. Conversely, we cannot say whether treatments such as fludarabine, a well-known T-cell suppressor (39), could disrupt the numbers or actions of the nodal CD8+ cells.

We divided the patients into three groups by the 25th and 75th percentile levels of CD8+ cells: lower quarter (<4.6%), middle half (4.2-8.6%), and upper quarter (>8.6%). After multivariable adjustment, patients with >8.6% CD8+ cells had five times lower risk of death, and patients with 4.2-8.6% CD8+ had twice lower risk, compared with patients with <4.2% CD8+. With respect to lymphoma-related death, the risk differences were further accentuated; there were only two deaths from indolent follicular lymphoma in the upper quarter. It should be emphasized that we do not believe the cut points at 4.2% and 8.6% to be biologically significant, rather, that the degree of CD8+ T-cell presence is a continuum with increased numbers of CD8+ cells correlating with improved survival. On the other hand, this stratification does provide a simple but powerful tool for assessing the prognosis already at diagnosis. The cutoffs at 4.2% and 8.6% are instantly applicable in the clinical setting. Standard flow cytometry analysis done at diagnosis could be used to identify the patients that are at high risk (<4.2% CD8+) for an early death, with median survival of about 6 years, the patients with an intermediate risk (4.2-8.6% CD8+), with median survival of about 10 years, and the patients that are likely to have a docile disease (>8.6% CD8+), where median survival is not yet reached. Even among the high-risk FLIPI patients, the three CD8+ strata could predict differences in median DSS (2.7 years, 4.1 years, and not reached, respectively). In conclusion, CD8+ cell content evaluated by flow cytometry was a predictor of survival, independent of all available clinical prognostic variables. A better understanding of the functional properties of these immune cells may suggest new ways of therapy for patients with follicular lymphoma.

We thank Bo Nilsson, M.Sci., biostatistician, for statistical analysis.