|Home | About | Journals | Submit | Contact Us | Français|
We sought to determine whether tumor-infiltrating lymphocytes and/or tumor mitotic activity could identify subgroups of patients with advanced serous epithelial ovarian cancer who would maximally benefit from aggressive surgical cytoreduction.
Snap-frozen specimens from 134 consecutive patients with stage III or IV serous or poorly differentiated ovarian adenocarcinoma undergoing primary debulking surgery from a single US institution were characterized based on CD3+, CD8+, FoxP3+ tumor infiltrating lymphocytes, and Ki67 expression. Kaplan-Meier survival curves were estimated and compared using a log-rank statistic. A multivariate Cox model was used to estimate adjusted hazard ratios (HR). Interactions were modeled using recursive partitioning based on maximal prognostic differentiation.
Brisk intraepithelial CD8+ cells (p=0.035) and low Ki67 expression (p=0.042) portended prolonged survival. T cell infiltration was more likely to occur in tumors with high proliferation index. Patients whose tumors exhibited low Ki67 expression and high intraepithelial CD8+ frequency had a 5-year survival rate of 73.3%. Patients with aggressive tumor behavior, i.e. whose tumors exhibited low frequency of intraepithelial CD8+ T cells or high Ki67 expression were more likely to draw benefit from aggressive surgical cytoreduction. Survival was similar for patients with brisk CD8+ T cells who had optimal or suboptimal debulking. Likewise, survival was similar for patients with low Ki67 expression who had optimal or suboptimal debulking.
These novel interactions of T cells, tumor proliferation index and surgical treatment reveal for the first time that biological prognosticators may be useful for surgical decision making in ovarian cancer.
Epithelial ovarian cancer (EOC) is the leading cause of death from gynecologic malignancies.1 Survival is known to be affected by tumor characteristics such as stage and histology, as well as host factors including patient age and performance status.2, 3 Apart from performance status, however, these prognosticators do not provide information that allows clinicians to tailor therapy. Currently, the standard therapeutic approach entails the administration of cytoreductive surgery and combination carboplatin/paclitaxel chemotherapy, when feasible, to all patients. Reducing tumor burden surgically to a maximal diameter of one half to two centimeters has been considered optimal cytoreduction, and has been shown in numerous studies to predict improved response to chemotherapy and prolonged survival.2-9 Patients with microscopic residual disease have the best outcomes in advanced EOC,2, 3 and this may be surgically achievable even in the presence of bulky upper or lower abdominal disease, diffuse carcinomatosis, or liver involvement, although ultra-radical procedures may result in significant morbidity.7, 10 Chemotherapy intensification and high dose approaches followed by stem cell support have not yielded any improvement in progression-free or overall survival.11-13 Thus, aggressive efforts to intensify therapy may increase morbidity, but have not been shown to significantly improve long term survival over the past two decades.14
It is conceivable that identification of biomarkers that capture intrinsic tumor properties or host factors affecting tumor behavior may help individualize therapy and result in improved therapeutic results and/or reduced toxicity. An important biomarker reflecting intrinsic tumor aggressiveness is the rate of tumor cell proliferation. In epithelial malignancies, tumor cells characterized by accelerated mitosis and proliferation are likely to result in rapid tumor growth, recurrence after therapy, and poor survival. Ki67 is a nuclear protein frequently used as a reliable marker for determining the growth fraction within tumors, as it is expressed during all active phases of the cell cycle (G1, S, G2, and mitosis), but is absent from resting cells (G0). In ovarian cancer, several studies have previously demonstrated Ki67 overexpression is associated with poor prognosis.15-19 Despite this, the clinical utility of Ki67 expression in guiding treatment decisions has not been shown.
A second source of important prognostic biomarkers derives from the patient's immune response, specifically tumor-infiltrating lymphocytes (TIL). Histologic evidence of T cells infiltrating tumor islets (named intratumoral or intraepithelial T cells) has been associated with improved survival in patients with ovarian cancer.20-22 In contrast, the presence of immunosuppressive regulatory T (Treg) cells, identified by co-expression of CD4 and CD25 or the FoxP3 transcription factor, has been associated with diminished survival.21, 23 To date, a role for this robust prognostic information in therapeutic decision-making has not been defined.
The amount of residual disease following surgery is considered the most important predictor of subsequent clinical course and overall survival in advanced stage ovarian cancer. Presently, there are no biological factors known to influence the impact of surgical cytoreduction on survival. We examined whether intrinsic tumor aggressiveness, assessed by the tumor proliferation index, or host response mechanisms, determined by T cell infiltration of the tumor, affect the impact of surgical cytoreduction on survival in EOC. Our findings indicate that tumor cell proliferation and T cell infiltration are independent prognosticators. Interestingly, T cell infiltration was more likely to occur in aggressive tumors with high proliferation index. Furthermore, our findings reveal that high proliferation index or low frequency of intraepithelial T cells identify patients with aggressive tumor behavior who are more likely to draw benefit from aggressive surgical cytoreduction. These observations suggest for the first time that biological prognosticators may be useful for clinical decision making in ovarian cancer and these biomarkers warrant investigation on a larger scale.
We identified a cohort of 134 consecutive patients with stage III/IV serous or poorly differentiated epithelial ovarian cancer with tumor available for analysis, undergoing primary debulking surgery at Memorial Sloan-Kettering Cancer Center between 1989 and 2003. Tumors were collected at the time of primary surgery following informed consent and under procedures approved by the local Institutional Review Board. Optimal surgical debulking was defined as residual tumor nodules measuring ≤1 cm. The primary study outcome was overall survival and survival times were censored when patients were lost to follow-up.
Six-micron sections of snap-frozen tumor specimens were subjected to immunohistochemical analysis. Specimens were fixed in cold acetone and stained using an automated stainer (DAKO cytomation). After hydrogen peroxide treatment, sections were incubated for 30 minutes with primary antibodies: rabbit anti-human CD3 (DAKO, 1:200 dilution); mouse anti-human CD8 (DAKO, 1:100 dilution); goat anti-human FoxP3 (Novus, 1:100 dilution); or mouse anti-human Ki67 (DAKO, 1:100 dilution). This was followed by incubation with polymer-labeled secondary antibodies, or a biotinylated anti-goat antibody in the case of FoxP3 staining, and 3,3′-diaminobenzidine peroxidase substrate (DAB+, DAKO). Slides were counterstained with hematoxylin. Normal tonsil and thymus samples were used as positive controls. Internal positive controls for CD3, CD8 and FoxP3 staining were provided by stromal T cells when present. Negative controls consisted of cases for which the primary antibody was omitted.
Specimens were evaluated blindly by two independent investigators and differences in interpretation were resolved by consensus. At least ten high-powered fields (HPF, 20X) were evaluated for each case, and the number of positively staining intraepithelial T cells per field was averaged for scoring. An ordinal score was defined based on the average T cell counts as follows: 0 = 0; 1+ = 1-5; 2+ = 6-10; 3+ = 11-19; 4+ ≥ 20 cells/HPF. Tumors with scores greater than 0 were considered positive. For intraepithelial CD3 staining, cases were considered to have CD3+ intraepithelial T cells present with any score > 0, consistent with our previous work.20 For both intraepithelial CD8 and intraepithelial FoxP3 cells, cases were considered to have a high frequency of intraepithelial T cells if they had greater than 10 positively staining cells per HPF. This value was chosen because it was just above the median value (50th percentile) of intraepithelial CD8 TIL counts used in a previous report.21 Nuclear Ki67 staining was also scored using ordinal categories that were based on the percentage of positively staining cells within tumor islets. High Ki67 expression was defined as >25% of tumor cells staining positively, as used in other solid tumors.16, 24
Pearson's chi-square test for independence was used to evaluate the associations among the biomarkers (low versus high expression), as well as between the biomarkers and other clinical characteristics. Kaplan-Meier survival curves were used to estimate five-year survival rates, and were compared using a log-rank statistic. A multivariate Cox model controlling for surgical debulking was used to estimate the adjusted hazard ratios (HR). A classification tree was developed as previously described25 using recursive partitioning that identified important prognostic factors among seven binary variables: the four study biomarkers, debulking status, stage and age. The first factor selected by the algorithm defined two groups that had the largest difference between the two survival curves. This process was then repeated in each subgroup. The five-year survival rates computed from the survival and their 95% confidence intervals are presented for the groups defined by the classification tree.
The characteristics of the 134 patients with stage III or IV serous or poorly differentiated epithelial ovarian cancer are presented in Table 1. The average age at diagnosis was 62 (26-84 years). All patients received first-line platinum-based chemotherapy. The 5-year survival rate for the entire cohort was 37.1%. Sixty-one patients (45%) were optimally debulked, and these patients had significantly improved 5-year survival (45.4%; HR=0.5, p=0.007) compared to those not optimally debulked (30.6%). The average duration of follow-up was 42 months (range 0.7-157.2 months). At the end of follow-up, 89 patients, (66% of the original cohort), had died of disease.
We first examined which intraepithelial T cell populations best predicted survival in the cohort of patients under study. The distribution of scores observed is presented in Figure 1.
These data collectively confirm previous findings and indicate that T cells have an impact on survival in the patient population under study, enabling further analyses of interactions. Intraepithelial CD8+ T cells were the most robust biomarker of host immune response, as low frequency of CD8+ T cells defined an unfavorable tumor subtype. Frequency of CD8+ cells was therefore used in subsequent analyses.
Next, we examined the impact of tumor cell proliferation on clinical outcome.15-18, 24 As with the other biomarkers, there was no association between Ki67 expression and patient age, tumor stage, grade, or histology. Forty-one patients (30.6%) exhibited high Ki67 expression (Table 1) with a 5-year survival of 24.7%; the remaining patients with low Ki67 expression had a 5-year survival of 41.4% (p=0.042, Figure 2B). Ki67 was an independent prognostic factor in multivariate Cox models including debulking, CD3, CD8, and FoxP3 (Table 2) with high Ki67 score defining an unfavorable tumor subtype.
To determine whether the increased survival associated with CD8+ T cell infiltration of ovarian cancers could be due to an association with low tumor cell proliferation rate and indolent tumor growth, we examined the association between intraepithelial CD8+ T cells and tumor Ki67 expression. There was a significant and positive association between expression of Ki67 and frequency of intraepithelial CD8+ T cell (p=0.041). In addition, the proportion of tumors exhibiting high frequency of CD8+ T cells increased as Ki67 expression increased, and patients with highly proliferative tumors were more likely to have high intraepithelial CD8+ T cell counts (odds ratio=2.4, p=0.041, Figure 3A and Table 1). Thus the presence of high numbers of intraepithelial CD8+ T cells did not associate with indolent tumors exhibiting low proliferative activity.
We examined the survival curves of four groups of patients defined by the combinations of Ki67 expression (high or low) and frequency of intraepithelial CD8+ cells (high or low). This comparison defined a group of patients with a significantly more favorable outcome: patients whose tumors exhibited low Ki67 expression and high intraepithelial CD8+ frequency (n=15, 11.2%) had a 5-year survival rate of 73.3% (Figure 3B). Interestingly, the survival curves of the other three groups, i.e. patients with high Ki67 expression (and high or low CD8+ frequency) as well as low Ki67 expression and low intraepithelial CD8+ frequency were similar.
Optimal debulking is an important prognostic factor in ovarian cancer. In agreement with our previous study20, optimal debulking was not associated with a high frequency of intraepithelial T cells (p=0.719). Similarly, high intraepithelial CD8+ cell score was associated with improved survival independently of the debulking status (HR 0.5, p=0.028) (Table 2).
We examined the impact of surgical debulking on survival separately in two subgroups of patients, defined by the presence of either a high or a low frequency of intraepithelial CD8+ cells. For the group of patients with a high frequency of intraepithelial CD8+ T cells, the survival curves for those who had optimal and suboptimal debulking were not significantly different (p=0.625, Figure 4A). Thus, optimal debulking did not impact the survival of patients whose tumors had a high frequency of intraepithelial CD8+ T cells. However, for patients with a low frequency of intraepithelial CD8+ T cells, those that had optimal debulking had a significantly better prognosis than those who had suboptimal debulking; the five-year survival rates were 41.2% and 25.1% for those with optimal and suboptimal debulking, respectively (p=0.002, Figure 4B). Importantly, the 5-year survival rate of suboptimally debulked patients with high intraepithelial CD8+ cell counts was similar to that of optimally debulked patients. Thus, optimal debulking impacts the natural course mainly of ovarian cancers with unfavorable biologic characteristics such as poor T cell infiltration.
Because Ki67 was an independent biological prognosticator, we also examined the impact of surgical debulking on survival separately in subgroups of patients whose tumors had high or low Ki67 expression. In patients with high Ki67 expression, there was a significant survival advantage associated with optimal debulking, with a 5-year survival rate of 48% for patients with optimal debulking and 16% for those with suboptimal debulking (p=0.023, Figure 5A). Conversely, in patients with low Ki67 expression, optimal debulking was not significantly associated with a survival advantage relative to suboptimal debulking (p=0.18, Figure 5B). Thus, optimal debulking was prognostic only for ovarian cancer patients whose tumors were characterized by an unfavorable biologic characteristic, high Ki67 expression.
The above results indicate that intraepithelial CD8+ T cells and Ki67 expression define subsets of patients with ovarian cancers that benefit differently from debulking. To further understand how each variable might affect the natural course of advanced stage serous ovarian cancer, we developed a prognostic tree using recursive partitioning. This algorithm identifies important prognostic factors and associated survival rates. We used seven binary variables: the four study biomarkers, Ki67, CD3, CD8 and FoxP3; as well as three well established prognostic factors, debulking status; stage; and age.2, 3 As previously described,25 with this algorithm the first factor selected has the strongest impact on separating patient groups with different survival. Debulking status was the first factor selected by the algorithm and it defined two groups that had the largest difference in survival curves (Figure 6). In suboptimally debulked patients, Ki67 expression was identified as the second strongest prognostic factor, the one that best predicted survival differences (p=0.027). Furthermore, for suboptimally debulked patients with low Ki67 expression, intraepithelial CD8+ T cell frequency was the third strongest prognostic factor (p=0.029). Among optimally debulked patients, stage (p=.059) was most predictive of survival, followed by frequency of intraepithelial CD8+ T cells (p=0.145) for stage III patients. Five-year survival rates are presented for each group of the prognostic tree in Figure 6. Based on this model, the estimated 5-year survival rate for a patient whose tumor was suboptimally debulked and had a low level of Ki67 with a high level of intraepithelial CD8+ T cells was 70% (95% CI=41.6 to 98.4) compared to a survival rate of 29.9% (95% CI=14.4 to 45.5) for patients whose tumor was suboptimally debulked and had a low level of Ki67 and a low level of intraepithelial CD8+ T cells.
The present data reveal novel interactions among T cell infiltration, tumor proliferation index and surgical treatment that are important for prognosis in advanced serous ovarian cancer, and for the first time indicate that biological factors intrinsic to the tumor or to the host response can provide useful information to individualize the surgical management of patients with this disease. Indeed, high frequency of intraepithelial CD8+ T cells and/or low proliferation (Ki67 expression) defined subsets of patients with more indolent tumors in whom the presence of residual tumors larger than one cm did not significantly alter survival. Vice versa, patients with tumors exhibiting unfavorable biologic features such as low frequency of intraepithelial CD8+ T cells and high proliferation had a markedly better prognosis when they had optimal cytoreductive surgery. Significantly, these data also provide evidence that the effects of tumor debulking, CD8+ T cells and tumor proliferation on clinical outcome are independent prognostic factors, but an association exists between increased tumor proliferation and increased frequency of intraepithelial CD8+ T cells.
The benefit of surgical cytoreduction has been established through numerous retrospective studies4-9 and was recently confirmed by analysis of more than 2,250 patients with stage III or IV EOC from four Gynecologic Oncology Group (GOG) phase III trials.2, 3 In fact, these and other smaller studies26 indicate that maximal cytoreduction as close as possible to microscopic residual is associated with the best survival. Our study is in agreement with this line of evidence, as recursive partitioning analysis identified residual tumor of less than one centimeter as the most important predictor of better survival in this cohort of patients.
Although biological factors have been shown to impact the natural course of ovarian cancer, it has remained unclear whether such factors may alter the benefit afforded by surgical cytoreduction. We used Ki67 as a reliable marker of intrinsic tumor aggressiveness and identified CD8 as the best cellular immune response biomarker in this patient cohort. Our study provides the first evidence that intrinsic tumor and host biological factors can influence the benefit of surgical cytoreduction. Although this study did not evaluate prospectively the association of surgical cytoreduction and biological factors with survival, the data still provide provocative evidence to suggest that biological biomarkers can provide useful information for selecting appropriate candidates for aggressive surgical cytoreduction. Based on analysis of four GOG studies,2, 3 Winter et al concluded that ultraradical procedures might be justified to achieve maximal cytoreduction and proposed that a diagnostic surgical procedure should be used first in an attempt to best assess resectability. Patients with disease that is considered optimally resectable could then undergo laparotomy and radical surgical cytoreduction. Our findings provide an additional tool to select patients for aggressive debulking and suggest that every attempt should be made for optimal cytoreduction in patients whose tumors exhibit unfavorable features such as high Ki67 and low CD8 score. Our study should be considered exploratory, as it is based on a relatively small cohort of patients. Although the number of patients in each subgroup are small, these results are strengthened by the fact that all patients in this cohort were treated in a single institution during a period of reasonably standardized surgical approach, and all had serous or poorly differentiated carcinoma. Additional studies are warranted to confirm our findings and test the hypothesis that selection of patients based on CD8 or Ki67 scores may reduce morbidity and maximize efficacy of surgical management. It should be noted that our study did not evaluate patients undergoing upper abdominal ultraradical procedures to achieve microscopic residual, as these were pioneered at a later date in the institution under study.10 Thus, future studies will have to address whether the observed interactions persist following more aggressive debulking procedures.
We identified CD8 as the most robust immune biomarker in this cohort. Although a significant correlation was found between the frequency of CD3+ and CD8+ intraepithelial T cells, intratumoral CD3+ T cells was not a prognostic factor. This is likely due to sample and effect size, as CD3 predicted outcome in other studies.20, 27 There is substantial evidence that T cells infiltrating ovarian cancer recognize and react to tumor antigens ex vivo.28-33 However, it remains unclear to date whether the observed association between intraepithelial T cells and improved clinical outcome can be attributed to the direct function of tumor-infiltrating T cells or merely to an association of T cells with indolent tumors with low proliferation. For example, it could be hypothesized that intraepithelial T cells are more able to accumulate in slowly growing tumors, which are associated with improved survival, while they are outpaced and excluded by aggressive tumors with elevated rates of proliferation, which are associated with poor survival.
The present data identified a positive association between high Ki67 expression and CD8+ T cell infiltrate. Mitotically active ovarian cancers are more likely to exhibit genetic instability34 and we hypothesize that these tumors express a more diverse antigenic repertoire, including neoantigens that elicit a cellular immune response. Similar observations have been made in breast carcinomas with mutations in BRCA1, a gene involved in maintaining genomic stability,35 which are higher grade but are also characterized by T cell infiltration and improved survival.36, 37 Similarly, colorectal cancers with microsatellite genetic instability are characterized by T cell infiltration and better prognosis in spite of being higher grade.38 Interestingly, a recent study found that ovarian cancers with p53 mutations were more likely to have intraepithelial T cells.39 This data, together with previous epidemiologic evidence,20-23 and molecular and functional studies20, 28-33 suggest that T cells may indeed react to tumor antigens and directly contribute to reduced tumor growth. If this were the case, it is possible that the biomarkers under study could help individualize not only surgery but also medical therapy and guide the selection of chemotherapy combinations that synergize with spontaneous antitumor mechanisms and immune or immunomodulatory therapy.40-42
This work was supported by NCI P01-CA83638 SPORE in Ovarian Cancer; NIH R01-CA116779-01A1; and a grant by the Ovarian Cancer Research Fund.
This work was presented in part at the annual meetings of the Society of Gynecologic Oncologists, San Diego, California, March, 2007, and the American Society of Clinical Oncology, Chicago, Illinois, June, 2007.