The Genomic Grade Index predicts postoperative clinical outcome in patients with soft-tissue sarcoma

The Genomic Grade Index predicts postoperative clinical outcome in patients with soft-tissue sarcoma Abstract Background Soft-tissue sarcomas (STSs) are a group of rare, heterogeneous, and aggressive tumors, with high metastatic risk and relatively few efficient systemic therapies. We hypothesized that the Genomic Grade Index (GGI), a 108-gene signature previously developed in early-stage breast cancer, might improve the prognostic assessment of patients with early-stage STS. Patients and methods We collected gene expression and clinicopathological data of 678 operated STS, and searched for correlations between the GGI-based classification and clinicopathological variables, including the metastasis-free survival (MFS). Results Based on GGI, 275 samples (41%) were classified as ‘GGI-low’ and 403 (59%) as ‘GGI-high’. The ‘GGI-high’ class was more associated with poor-prognosis features than the ‘GGI-low’ class: pathological grade 3 (P = 9.50E–11), undifferentiated sarcomas and leiomyosarcomas (P < 1.00E–06), location in extremities (P < 1.00E–06), and complex genetic profile (P = 2.1E–20). The 5-year MFS was 53% (95%CI 47–59) in the ‘GGI-high’ class versus 78% (95%CI 72–85) in the ‘GGI-low’ class (P = 3.02E–11), with a corresponding hazard ratio for metastatic relapse equal to 2.92 (95%CI 2.10–4.07; P = 2.23E–10). In multivariate analysis, the GGI-based classification remained significant, whereas the pathological grade did not. In fact, the GGI-based classification stratified the patients with pathological grades 1 and 2 and those with pathological grade 3 in two classes with different 5-year MFS. Comparison of the GGI and CINSARC multigene signatures revealed similar correlations with clinicopathological variables, which were, however, stronger with GGI than with CINSARC, a strong concordance (71%) in terms of low-risk or high-risk classifications, and independent prognostic value for MFS in multivariate analysis, suggesting complementary prognostic information. Conclusion GGI refines the prediction of MFS in operated STS and might improve the tailoring of adjuvant chemotherapy. Further clinical validation is warranted in larger retrospective, then prospective series, as well as the functional validation of relevant genes that could provide new therapeutic targets. cell cycle, Genomic Grade Index, gene expression signatures, soft-tissue sarcoma, survival Key Message The Genomic Grade Index (GGI) is a 108-gene signature previously developed in early-stage breast cancer. We analyzed gene expression and clinicopathological data of 678 localized STS and showed that GGI is an independent prognostic factor for metastatic relapse, which outperforms the pathological grading. Introduction Soft-tissue sarcomas (STSs) are rare tumors that represent a heterogeneous group with at least 50 different pathological subtypes [1]. Surgery is the standard treatment of adult patients with an early-stage STS, but up to 50% of them experience postoperative metastatic recurrence, lethal in most of cases [2]. Results of adjuvant chemotherapy are conflicting, with negative results from the largest randomized study [3], but positive results in terms of relapses in meta-analyses [4, 5]. Therefore, adjuvant chemotherapy is not a standard treatment in adult-type STS, and can be given to high-risk patients (high-grade, deep-seated tumor, tumor size >5 cm) after a shared decision-making with the patient [2]. Pathological grading, most commonly based on the Federation Francaise des Centres de Lutte Contre le Cancer (FNCLCC) grading system, is a good predictor of metastatic relapse in early-stage STS [6], and the most influential for the challenging decision of adjuvant chemotherapy [7]. It distinguishes three increasing-malignancy grades from 1 to 3, based on evaluation of three pathological parameters—tumor differentiation, necrosis, and mitotic index. However, it displays limitations: questionable reproducibility between pathologists, poorly informative prognostic value for grade 2, which represents ∼40% of the cases, nonapplicability in all pathological subtypes, and difficulty of assessment on small tumor biopsies [8, 9]. The same limitations exist for the frequently used three-grade system developed by the National Cancer Institute [10]. Thus, tumor grading remains imperfect to solve STS clinical heterogeneity in clinical practice. In this context, gene expression [11] may improve our capacity to predict clinical outcome. In 2010, the French Sarcoma Group identified and validated a 67-gene expression signature (CINSARC) associated with metastasis-free survival (MFS) in a cohort of 310 STS patients [12]. The signature had been identified using a hypothesis-driven approach, by comparing STS samples with high versus low number of genomic alterations and high versus low pathological grade. To our knowledge, this signature, which defines ‘high-risk’ and ‘low-risk’ patients, represents the most studied prognostic multigene signature in STS, although other prognostic expression signatures have been reported in adults’ STS [13, 14], or more specifically in liposarcomas [15], and leiomyosarcomas [16, 17]. In breast cancer, pathological grade is based on cell differentiation, nuclear pleomorphism, and mitotic index: it includes three classes from 1 to 3, and displays the same weaknesses as in STS. To improve performances, a molecular grading, the Genomic Grade Index (GGI), was constructed, by comparing the gene expression profiles of grade 3 versus grade 1 early-stage tumors [18]. This 108-gene mRNA signature reclassified patients with histologic grade 2 tumors into two groups (‘GGI-high’ and ‘GGI-low’) with high risk versus low risk of recurrence, improved the accuracy of tumor pathological grade, and outperformed its prognostic value in multivariate analysis [18]. It was also associated with pathological response to neoadjuvant anthracycline-based chemotherapy in breast cancer [19], as did CINSARC [20]. Because the respective pathological grades of STS and breast cancer are based on similar major morphological criteria (differentiation, mitotic index), we tested the hypothesis that GGI could improve the prognostic classification of STS, as reported in breast cancer [18]. We analyzed expression data of 678 STS profiled using DNA microarrays and RNA-sequencing, and searched for correlation between GGI-based classification and clinicopathological variables. Patients and methods More details are available in the supplementary Methods file, available at Annals of Oncology online. Soft-tissue sarcoma samples We collected clinicopathological and mRNA expression data of 678 clinical STS samples from four datasets (supplementary Table S1, available at Annals of Oncology online). All samples were from operative specimen of patients operated for nonmetastatic STS and corresponded to untreated primary tumors. The selection of datasets was based according to the following criteria: availability of data in the GEO, Array-Express, or TCGA databases, and the presence of at least 50 informative samples. Gene expression data analysis Data analysis required preanalytic processing. We then applied the GGI signature and the CINSARC signature [12], to each dataset separately, by using strictly the same method as that reported in the original publications: each sample was classified as ‘low’ associated with low risk of relapse, or ‘high’ associated with high risk of relapse. Statistical analysis Our primary end point was the MFS calculated from the date of diagnosis until the date of distant relapse or death from STS. Survival was calculated using the Kaplan–Meier method and curves were compared with the log-rank test. Correlations between the GGI-based groups (low versus high) and the clinicopathological features were calculated with the Student’s t-test or the Fisher’s exact test when appropriate. Uni- and multivariate prognostic analyses were done using Cox regression analysis (Wald test). Results Patients’ characteristics Gene expression profiles of 678 clinically annotated STS samples were available. As presented in Table 1, the median patients’ age was 63 years. The sex ratio was balanced, with 171 females (46%) and 200 males (54%). The most frequent anatomical sites were the extremities (n = 159; 42%), followed by the internal trunk (n = 151; 40%), and the superficial trunk (n = 64; 17%); 160 cases (82%) were deeply seated, below or through the superficial fascia. The most frequent pathological types included liposarcomas (n = 256; 38%), undifferentiated sarcomas (n = 202; 30%), and leiomyosarcomas (n = 149; 22%). Based on the pathological type, 433 samples (65%) were STS with complex genetic profile. The median pathological tumor size at the time of surgery was 9 cm. Fifty-six percent of the samples (n = 171) were FNCLCC grade 3. With a median follow-up of 32 months, 209 of 678 (31%) patients experienced metastatic relapse and the 5-year MFS was 63% (95%CI 59–68; Figure 1A). Based on the GGI signature, 275 samples (41%) were classified as ‘GGI-low’ and 403 (59%) as ‘GGI-high’. Table 1. Clinicopathological characteristics and correlations with the GGI-based classification Characteristics  All, n = 678  GGI classes   P-value  Low, n = 275 (41%)  High, n = 403 (59%)  Age, median  Years (range)  63 (17–93)  62 (25–93)  63 (17–90)  0.955  Sex          0.910    Female  171 (46%)  52 (45%)  119 (46%)      Male  200 (54%)  63 (55%)  137 (54%)    Tumor site          <1.00E–06    Extremity  159 (42%)  26 (23%)  133 (49%)      Internal trunk  151 (40%)  69 (62%)  82 (30%)      Superficial trunk  64 (17%)  15 (13%)  49 (18%)      Head and neck  8 (2%)  2 (2%)  6 (2%)    Depth          0.340    Deep  160 (82%)  56 (78%)  104 (84%)      Superficial  36 (18%)  16 (22%)  20 (16%)    Pathological type          <1.00E–06    Liposarcoma  256 (38%)  151 (55%)  105 (26%)      Undifferentiated sarcoma  202 (30%)  51 (19%)  151 (37%)      Leiomyosarcoma  149 (22%)  43 (16%)  106 (26%)      Myxofibrosarcoma  39 (6%)  14 (5%)  25 (6%)      Others  32 (5%)  16 (6%)  16 (4%)    Genetic profile          2.10E–20    Simple  233 (35%)  149 (56%)  84 (21%)      Complex  433 (65%)  117 (44%)  316 (79%)    Pathological size, median  cm (range)  9 (2–40)  8.75 (2–40)  9 (2–40)  0.909  Pathological FNCLCC grade        9.50E–11    1 and 2  136 (44%)  84 (66%)  52 (29%)      3  171 (56%)  43 (34%)  128 (71%)    Metastatic events  Number of patients  209 (31%)  45 (16%)  164 (41%)  7.58E–12  5-year MFS  Months [95% CI]  63% [59-68]  78% [72-85]  53% [47-59]  3.02E–11  Characteristics  All, n = 678  GGI classes   P-value  Low, n = 275 (41%)  High, n = 403 (59%)  Age, median  Years (range)  63 (17–93)  62 (25–93)  63 (17–90)  0.955  Sex          0.910    Female  171 (46%)  52 (45%)  119 (46%)      Male  200 (54%)  63 (55%)  137 (54%)    Tumor site          <1.00E–06    Extremity  159 (42%)  26 (23%)  133 (49%)      Internal trunk  151 (40%)  69 (62%)  82 (30%)      Superficial trunk  64 (17%)  15 (13%)  49 (18%)      Head and neck  8 (2%)  2 (2%)  6 (2%)    Depth          0.340    Deep  160 (82%)  56 (78%)  104 (84%)      Superficial  36 (18%)  16 (22%)  20 (16%)    Pathological type          <1.00E–06    Liposarcoma  256 (38%)  151 (55%)  105 (26%)      Undifferentiated sarcoma  202 (30%)  51 (19%)  151 (37%)      Leiomyosarcoma  149 (22%)  43 (16%)  106 (26%)      Myxofibrosarcoma  39 (6%)  14 (5%)  25 (6%)      Others  32 (5%)  16 (6%)  16 (4%)    Genetic profile          2.10E–20    Simple  233 (35%)  149 (56%)  84 (21%)      Complex  433 (65%)  117 (44%)  316 (79%)    Pathological size, median  cm (range)  9 (2–40)  8.75 (2–40)  9 (2–40)  0.909  Pathological FNCLCC grade        9.50E–11    1 and 2  136 (44%)  84 (66%)  52 (29%)      3  171 (56%)  43 (34%)  128 (71%)    Metastatic events  Number of patients  209 (31%)  45 (16%)  164 (41%)  7.58E–12  5-year MFS  Months [95% CI]  63% [59-68]  78% [72-85]  53% [47-59]  3.02E–11  Bold entries correspond to significant P-values ≤ 0.05. GGI, Genome Grade Index; MFS, metastasis-free survival. Figure 1. View largeDownload slide Metastasis-free survival (MFS) in patients with STS and according to Genomic Grade Index (GGI). (A) Kaplan–Meier MFS curves in all patients with STS. (B) Similar to (A), but according to the GGI-based classification (‘GGI-low’ and ‘GGI-high’ classes). (C) Similar to (B), but in patients with grades 1 and 2 STS. (D) Similar to (C), but in patients with grade 3 STS. The P-values of the log-rank test are indicated. Figure 1. View largeDownload slide Metastasis-free survival (MFS) in patients with STS and according to Genomic Grade Index (GGI). (A) Kaplan–Meier MFS curves in all patients with STS. (B) Similar to (A), but according to the GGI-based classification (‘GGI-low’ and ‘GGI-high’ classes). (C) Similar to (B), but in patients with grades 1 and 2 STS. (D) Similar to (C), but in patients with grade 3 STS. The P-values of the log-rank test are indicated. GGI-based classification and clinicopathological characteristics There was no correlation between the two GGI classes and patients’ age and sex, tumor depth and size (Table 1). By contrast, correlations were found with tumor site, pathological type, genetic profile, and pathological grade: compared with the ‘GGI-low’ class, the ‘GGI-high’ class included more STS located in extremities and less on internal trunk (P < 1.00E–06), more undifferentiated sarcomas and leiomyosarcomas and less liposarcomas (P < 1.00E–06), more samples with complex genetic profile (P = 2.1E–20), and more samples with pathological grade 3 (P = 9.50E–11). Thus, the ‘GGI-high’ class was more associated with poor-prognosis features than the ‘GGI-low’ class. GGI-based classification and metastatic relapse The 5-year MFS was shorter in the ‘GGI-high’ class (53%, 95%CI 47–59) than in the ‘GGI-low’ class (78%, 95%CI 72–85; P = 3.02E–11, log-rank test; Figure 1B). In univariate analysis (Table 2), the hazard ratio (HR) for metastatic relapse was 2.92 (95%CI 2.10–4.07) in the ‘GGI-high’ class when compared with the ‘GGI-low’ class (P = 2.23E–10, Wald test). Other variables associated with MFS included the pathological type (P = 1.35E–06), with a trend for significance for the pathological grade (P = 8.80E–02). Patients’ age and sex, tumor site and depth, and pathological size were not associated with MFS. Table 2. Univariate and multivariate prognostic analyses for MFS Characteristics    Univariate      Multivariate    n  HR [95%CI]  P-value  n  HR [95%CI]  P-value  Age    371  1.00 [0.99–1.01]  0.902        Sex  Male versus female  371  1.02 [0.70–1.49]  0.909        Tumor site  Head and neck versus extremities  382  0.00 [0.00–Inf.]  0.660          Internal trunk versus extremities    0.77 [0.50–1.18]            Superficial trunk versus extremities    0.81 [0.47–1.40]          Depth  Superficial versus deep  196  0.78 [0.38–1.61]  0.495        Pathological type  Liposarcoma versus leiomyosarcoma  678  0.48 [0.35–0.67]  1.35E–06  307  0.46 [0.23–0.95]  3.60E–02    Myxofibrosarcoma versus leiomyosarcoma    0.45 [0.24–0.86]    307  0.49 [0.24–1.00]  5.03E–02    Undifferentiated sarcoma versus leiomyosarcoma    0.43 [0.30–0.61]    307  0.36 [0.22–0.59]  4.40E–05    Others versus leiomyosarcoma    0.21 [0.08–0.57]    307  0.28 [0.07–1.16]  7.95E–02  Pathological size (cm)    210  1.00 [0.96–1.04]  0.898        Pathological FNCLCC grade  1-2 versus 3  307  1.43 [0.95–2.17]  0.088  307  1.39 [0.87–2.20]  0.166  GGI classes  High-risk versus low-risk  678  2.92 [2.10–4.07]  2.23E–10  307  2.24 [1.34–3.74]  2.05E–03  Characteristics    Univariate      Multivariate    n  HR [95%CI]  P-value  n  HR [95%CI]  P-value  Age    371  1.00 [0.99–1.01]  0.902        Sex  Male versus female  371  1.02 [0.70–1.49]  0.909        Tumor site  Head and neck versus extremities  382  0.00 [0.00–Inf.]  0.660          Internal trunk versus extremities    0.77 [0.50–1.18]            Superficial trunk versus extremities    0.81 [0.47–1.40]          Depth  Superficial versus deep  196  0.78 [0.38–1.61]  0.495        Pathological type  Liposarcoma versus leiomyosarcoma  678  0.48 [0.35–0.67]  1.35E–06  307  0.46 [0.23–0.95]  3.60E–02    Myxofibrosarcoma versus leiomyosarcoma    0.45 [0.24–0.86]    307  0.49 [0.24–1.00]  5.03E–02    Undifferentiated sarcoma versus leiomyosarcoma    0.43 [0.30–0.61]    307  0.36 [0.22–0.59]  4.40E–05    Others versus leiomyosarcoma    0.21 [0.08–0.57]    307  0.28 [0.07–1.16]  7.95E–02  Pathological size (cm)    210  1.00 [0.96–1.04]  0.898        Pathological FNCLCC grade  1-2 versus 3  307  1.43 [0.95–2.17]  0.088  307  1.39 [0.87–2.20]  0.166  GGI classes  High-risk versus low-risk  678  2.92 [2.10–4.07]  2.23E–10  307  2.24 [1.34–3.74]  2.05E–03  Bold entries correspond to significant P-values ≤ 0.05. GGI, Genome Grade Index; MFS, metastasis-free survival. In multivariate analysis (Table 2) including pathological type and grade, and GGI-based classification, two factors remained significant, including the GGI-based classification (P = 2.05E–03). The pathological grade lost its prognostic value. In fact, the GGI-based classification stratified the patients with pathological grades 1 and 2 in two classes with different 5-year MFS (Figure 1C): 74% (95%CI 62–87) in the ‘GGI-low’ class versus 59% (95%CI 46–76) in the ‘GGI-high’ class (P = 3.17E–02, log-rank test). Similarly, the patients with pathological grade 3 were separated in two classes with 82% 5-year MFS (95%CI 70–95) in the ‘GGI-low’ class versus 54% (95%CI 45–65) in the ‘GGI-high’ class (P = 5.20E–03, log-rank test; Figure 1D). Even if the follow-up was relatively limited, GGI better discriminated the relapses during the first 5 years in the whole population (Figure 1B) and in the grade 3 patients (Figure 1D). Comparison with the CINSARC signature Given the potential value of CINSARC in STS [12, 13], we assessed its prognostic value and compared it with that of the GGI. There were 39 genes common to the two signatures, representing 36% of GGI genes and 58% of CINSARC genes. Based on CINSARC, 387 samples (57%) were classified as ‘CINSARC-low’ and 291 (43%) as ‘CINSARC-high’. The CINSARC classes showed the same clinicopathological correlations as the GGI-based classes, with correlations with the tumor site, pathological type and grade, and genetic profile (supplementary Table S2, available at Annals of Oncology online). However, all these correlations were more significant with GGI than with CINSARC. There was also a strong correlation between the GGI and CINSARC classes (P = 3.11E–35; Fisher’s exact test), with 482 of 678 (71%) samples classified similarly as low-risk or high-risk; the 196 discordant samples were more often CINSARC-low/GGI-high than CINSARC-high/GGI-low. CINSARC was strongly associated with MFS with 50% 5-year MFS (95%CI 44–57) in the ‘CINSARC-high’ class and 74% 5-year MFS (95%CI 68–79) in the ‘CINSARC-low’ class (P = 4.77E–11, log-rank test; Figure 2A). In univariate analysis (supplementary Table S3, available at Annals of Oncology online), the HR for metastatic relapse was 2.48 (95%CI 1.87–3.28) in the ‘CINSARC-high’ class when compared with the ‘CINSARC-low’ class (P = 2.03E–10, Wald test). In multivariate analysis (supplementary Table S3, available at Annals of Oncology online) including the CINSARC and the GGI classifications, both classifications remained significant (P = 1.44E–05 for GGI, P = 1.51E–04 for CINSARC). The GGI classification affected the clinical outcome of the CINSARC classes (Figure 2B): the 5-year MFS was 79% (95%CI 72–86) in the ‘CINSARC-low’/‘GGI-low’ class and 65% (95%CI 56–75) in the ‘CINSARC-low’/‘GGI-high’, and 73% (95%CI 60–89) in the ‘CINSARC-high’/‘GGI-low’ class and 46% (95%CI 39–53) in the ‘CINSARC-high’/‘GGI-high’ class (P = 1.1E–13, log-rank test). Figure 2. View largeDownload slide Metastasis-free survival (MFS) in patients with STS according to CINSARC and Genomic Grade Index (GGI). (A) Kaplan–Meier MFS curves in all patients with STS according to the CINSARC-based classification (‘CINSARC-low’ and ‘CINSARC-high’ classes). (B) Similar to (A), but according to the four-class classification based on both GGI and CINSARC. The P-values of the log-rank test are indicated. Figure 2. View largeDownload slide Metastasis-free survival (MFS) in patients with STS according to CINSARC and Genomic Grade Index (GGI). (A) Kaplan–Meier MFS curves in all patients with STS according to the CINSARC-based classification (‘CINSARC-low’ and ‘CINSARC-high’ classes). (B) Similar to (A), but according to the four-class classification based on both GGI and CINSARC. The P-values of the log-rank test are indicated. Discussion The absence of accurate prognostic features, such as pathological grade, and of predictors of response to anthracycline/ifosfamide-based chemotherapy in patients with STS, combined with the scarcity and heterogeneity of the disease, explain in part the difficulty to prove the benefit, if any, of adjuvant chemotherapy. Because pathological grade is considered as the best prognostic factor of STS and given the performance of GGI in breast cancer, we tested the prognostic value of this latter in a large series of 678 operated STS samples. We showed that GGI is an independent prognostic factor for metastatic relapse. To our knowledge, this is by far the largest prognostic study of gene expression profiles in STS. During the last decades, high-throughput molecular analyses, notably gene expression profiling, have provided insights into the extensive heterogeneity of cancers. For example, several multigene signatures such as Oncotype™ or Prosigna™ are marketed in early-stage breast cancer. No similar signature is currently available in clinical routine for STS patients. The scarcity of STS and the paucity of tumor specimens available for analysis explain the relatively small number of samples profiled in previous prognostic studies, 310 in the largest one [12]. We overcame the problem by pooling four public sets including one multicentric series prospectively collected (TCGA) and three unicentric or multicentric retrospective series, representing a total of 678 operated primary cancers. The whole series displayed the expected clinicopathological characteristics and poor prognosis with ∼60% 5-year MFS. Its size allowed multivariate analysis, and the transcriptional nature of data allowed the comparison of the GGI signature with CINSARC, a promising expression signature. When applied to our series of 678 STS, GGI classified 41% of samples as ‘GGI-low’ and 59% as ‘GGI-high’. Of course, none of the STS samples had been used to generate the GGI signature. The HR for metastatic relapse was close to 3 in the ‘GGI-high’ class when compared with the ‘GGI-low’ class. In breast cancer, GGI is strongly associated with pathological grade [18]. Here too, and as expected given the similarity between STS and breast cancer regarding the definition of grade, there was strong correlation with pathological tumor grade, and with other grade-associated clinical variables. Despite the association of the ‘GGI-high’ class with poor-prognosis features, the GGI classification remained an independent prognostic feature in multivariate analysis, and provided additional information to pathological grade by discriminating within the patients with grades 1 and 2 tumor and with grade 3 tumor those with good-prognosis (‘GGI-low’) from those with poor-prognosis (‘GGI-high’). This prognostic superiority of GGI suggests that it is biologically more coherent than pathological grade to capture the proliferation level of tumors, likely thanks to its more quantitative and objective character. Like CINSARC, GGI provides the advantage over the pathological grading of stratifying patients into two groups instead of three, thus facilitating the clinical management. Thus, as in breast cancer [18], GGI improved the prognostic value of pathological grade in STS. We have reported a similar result in patients with GIST [21]. Such transversality of prognostic value in different cancers has been recently reported for CINSARC [22], which, like GGI, likely reflects a fundamental biological property of tumors able to metastasize. Several genes included in the GGI signature and/or overexpressed in the ‘GGI-high’ samples encode potential therapeutic targets involved in cell cycle regulation that could be, if functionally validated, targeted by new drugs: kinases (AURKA/B, BUB1, CDC2, CDK4, CHEK1, NEK2, PLK1/4) and phosphatase (CDC25). Among the prognostic gene signatures previously established in STS [12–17], only two [12, 15]—corresponding to those based on the largest series of cases—combined both validation in an independent tumor set and multivariate prognostic analysis, and only one, the CINSARC signature, concerned all STS pathological types. We thus compared the GGI and CINSARC signatures in our present series. Several similarities were observed. First, there was a relatively strong gene overlap, representing 36% of GGI genes and 58% of CINSARC genes. Second, we found similar correlations with clinicopathological variables (notably pathological grade), which were, however, stronger with GGI than with CINSARC, suggesting more homogeneity within classes and more differences between classes with GGI. Third, there was a strong correlation between the GGI classes and the CINSARC classes with 71% of samples classified similarly as low-risk or high-risk. Fourth, both signatures showed prognostic value in uni- and multivariate analyses, suggesting independent information. The two signatures showed complementarity since each signature was able to stratify each class of the other signature in two subclasses with different MFS; however, the 5-year MFS difference was 14% between the GGI subclasses defined in the ‘CINSARC-low’ class and 27% between those defined in the ‘CINSARC-high’ class, whereas it was lesser between the CINSARC subclasses defined in the ‘GGI-low’ class (6%) and in the ‘GGI-high’ class (19%). Conclusion In conclusion, we have shown that a GGI-based classification of operated STS outperforms the prognostic performances of the FNCLCC grading system in terms of MFS and provides independent information complementary to the CINSARC-based classification. The strength of our results lies in several aspects: the number of 678 samples that, to our knowledge, makes our series the largest prognostic gene expression study reported so far in STS; the biological relevance of involved genes; and the independent prognostic value through three different technological platforms. Limitations include the retrospective nature of the series and associated statistical biases, such as missing data for some variables for some patients or different lengths of follow-up between the different datasets, and the relatively large number of genes included (108 genes) for developing a signature applicable in clinical routine. Efforts are ongoing in breast cancer to improve the clinical applicability of GGI though a qRT-PCR test carried out on formalin-fixed paraffin-embedded samples [23]. But yet, by refining the prediction of MFS, GGI might improve our ability to better tailor adjuvant chemotherapy: patients defined as ‘GGI-low’ could be spared adjuvant chemotherapy, whereas those defined as ‘GGI-high’ should be, all the more so because ‘GGI-high’ tumors, might be more sensitive to anthracycline-based chemotherapy than ‘GGI-low’ tumors, as demonstrated in breast cancer [20]. Additionally, the predictive value of GGI for the pathological response to neoadjuvant chemotherapy deserves to be tested in STS patients. But yet, by helping stratify STS, GGI could revive the interest of adjuvant chemotherapy in STS, and improve patients’ survival. Because of the exploratory nature of our study, further clinical validation is warranted in larger retrospective series, then prospective trials of adjuvant chemotherapy in STS. Funding Institut Paoli-Calmettes (no grant number is applicable). Disclosure The authors have declared no conflicts of interest. Author ORCIDs FB: http://orcid.org/0000-0002-0157-0959. References 1 Wunder JS, Nielsen TO, Maki RG et al.   Opportunities for improving the therapeutic ratio for patients with sarcoma. Lancet Oncol  2007; 8( 6): 513– 524. Google Scholar CrossRef Search ADS PubMed  2 ESMO/European Sarcoma Network Working Group. Soft tissue and visceral sarcomas: ESMO Clinical Practice Guidelines for diagnosis, treatment and follow-up. Ann Oncol  2014; 25(Suppl 3): iii102– iii112. PubMed  3 Woll PJ, Reichardt P, Le Cesne A et al.   Adjuvant chemotherapy with doxorubicin, ifosfamide, and lenograstim for resected soft-tissue sarcoma (EORTC 62931): a multicentre randomised controlled trial. Lancet Oncol . 2012; 13( 10): 1045– 1054. Google Scholar CrossRef Search ADS PubMed  4 Adjuvant chemotherapy for localised resectable soft-tissue sarcoma of adults: meta-analysis of individual data. Sarcoma Meta-analysis Collaboration. Lancet  1997; 350: 1647– 1654. CrossRef Search ADS PubMed  5 Pervaiz N, Colterjohn N, Farrokhyar F et al.   A systematic meta-analysis of randomized controlled trials of adjuvant chemotherapy for localized resectable soft-tissue sarcoma. Cancer  2008; 113( 3): 573– 581. Google Scholar CrossRef Search ADS PubMed  6 Guillou L, Coindre JM, Bonichon F et al.   Comparative study of the National Cancer Institute and French Federation of Cancer Centers Sarcoma Group grading systems in a population of 410 adult patients with soft tissue sarcoma. JCO  1997; 15: 350– 362. Google Scholar CrossRef Search ADS   7 Italiano A, Delva F, Mathoulin-Pelissier S et al.   Effect of adjuvant chemotherapy on survival in FNCLCC grade 3 soft tissue sarcomas: a multivariate analysis of the French Sarcoma Group Database. Ann Oncol  2010; 21( 12): 2436– 2441. Google Scholar CrossRef Search ADS PubMed  8 Coindre JM. Grading of soft tissue sarcomas: review and update. Arch Pathol Lab Med  2006; 130( 10): 1448– 1453. Google Scholar PubMed  9 Coindre JM, Terrier P, Guillou L et al.   Predictive value of grade for metastasis development in the main histologic types of adult soft tissue sarcomas: a study of 1240 patients from the French Federation of Cancer Centers Sarcoma Group. Cancer  2001; 91( 10): 1914– 1926. Google Scholar CrossRef Search ADS PubMed  10 Costa J, Wesley RA, Glatstein E, Rosenberg SA. The grading of soft tissue sarcomas. Results of a clinicohistopathologic correlation in a series of 163 cases. Cancer  1984; 53( 3): 530– 541. Google Scholar CrossRef Search ADS PubMed  11 Bertucci F, Houlgatte R, Nguyen C et al.   Gene expression profiling of cancer by use of DNA arrays: how far from the clinic? Lancet Oncol  2001; 2( 11): 674– 682. Google Scholar CrossRef Search ADS PubMed  12 Chibon F, Lagarde P, Salas S et al.   Validated prediction of clinical outcome in sarcomas and multiple types of cancer on the basis of a gene expression signature related to genome complexity. Nat Med  2010; 16( 7): 781– 787. Google Scholar CrossRef Search ADS PubMed  13 Francis P, Namlos HM, Muller C et al.   Diagnostic and prognostic gene expression signatures in 177 soft tissue sarcomas: hypoxia-induced transcription profile signifies metastatic potential. BMC Genomics  2007; 8( 1): 73. Google Scholar CrossRef Search ADS PubMed  14 Hoffmann AC, Danenberg KD, Taubert H et al.   A three-gene signature for outcome in soft tissue sarcoma. Clin Cancer Res  2009; 15( 16): 5191– 5198. Google Scholar CrossRef Search ADS PubMed  15 Gobble RM, Qin LX, Brill ER et al.   Expression profiling of liposarcoma yields a multigene predictor of patient outcome and identifies genes that contribute to liposarcomagenesis. Cancer Res  2011; 71( 7): 2697– 2705. Google Scholar CrossRef Search ADS PubMed  16 Beck AH, Lee CH, Witten DM et al.   Discovery of molecular subtypes in leiomyosarcoma through integrative molecular profiling. Oncogene  2010; 29( 6): 845– 854. Google Scholar CrossRef Search ADS PubMed  17 Lee YF, John M, Falconer A et al.   A gene expression signature associated with metastatic outcome in human leiomyosarcomas. Cancer Res  2004; 64( 20): 7201– 7204. Google Scholar CrossRef Search ADS PubMed  18 Sotiriou C, Wirapati P, Loi S et al.   Gene expression profiling in breast cancer: understanding the molecular basis of histologic grade to improve prognosis. J Natl Cancer Inst  2006; 98( 4): 262– 272. Google Scholar CrossRef Search ADS PubMed  19 Liedtke C, Hatzis C, Symmans WF et al.   Genomic grade index is associated with response to chemotherapy in patients with breast cancer. J Clin Oncol  2009; 27( 19): 3185– 3191. Google Scholar CrossRef Search ADS PubMed  20 Bertucci F, Finetti P, Sabatier R, Birnbaum D. The CINSARC signature: prognostic and predictive of response to chemotherapy? Cell Cycle  2010; 9( 19): 4025– 4027. Google Scholar CrossRef Search ADS PubMed  21 Bertucci F, Finetti P, Ostrowski J et al.   Genomic Grade Index predicts postoperative clinical outcome of GIST. Br J Cancer  2012; 107( 8): 1433– 1441. Google Scholar CrossRef Search ADS PubMed  22 Lesluyes T, Delespaul L, Coindre JM, Chibon F. The CINSARC signature as a prognostic marker for clinical outcome in multiple neoplasms. Sci Rep  2017; 7( 1): 5480. Google Scholar CrossRef Search ADS PubMed  23 Toussaint J, Sieuwerts AM, Haibe-Kains B et al.   Improvement of the clinical applicability of the Genomic Grade Index through a qRT-PCR test performed on frozen and formalin-fixed paraffin-embedded tissues. BMC Genomics  2009; 10( 1): 424. Google Scholar CrossRef Search ADS PubMed  © The Author 2017. Published by Oxford University Press on behalf of the European Society for Medical Oncology. All rights reserved. For Permissions, please email: journals.permissions@oup.com. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Annals of Oncology Oxford University Press

The Genomic Grade Index predicts postoperative clinical outcome in patients with soft-tissue sarcoma

Loading next page...
 
/lp/ou_press/the-genomic-grade-index-predicts-postoperative-clinical-outcome-in-S0hLqyddGz
Publisher
Oxford University Press
Copyright
© The Author 2017. Published by Oxford University Press on behalf of the European Society for Medical Oncology. All rights reserved. For Permissions, please email: journals.permissions@oup.com.
ISSN
0923-7534
eISSN
1569-8041
D.O.I.
10.1093/annonc/mdx699
Publisher site
See Article on Publisher Site

Abstract

Abstract Background Soft-tissue sarcomas (STSs) are a group of rare, heterogeneous, and aggressive tumors, with high metastatic risk and relatively few efficient systemic therapies. We hypothesized that the Genomic Grade Index (GGI), a 108-gene signature previously developed in early-stage breast cancer, might improve the prognostic assessment of patients with early-stage STS. Patients and methods We collected gene expression and clinicopathological data of 678 operated STS, and searched for correlations between the GGI-based classification and clinicopathological variables, including the metastasis-free survival (MFS). Results Based on GGI, 275 samples (41%) were classified as ‘GGI-low’ and 403 (59%) as ‘GGI-high’. The ‘GGI-high’ class was more associated with poor-prognosis features than the ‘GGI-low’ class: pathological grade 3 (P = 9.50E–11), undifferentiated sarcomas and leiomyosarcomas (P < 1.00E–06), location in extremities (P < 1.00E–06), and complex genetic profile (P = 2.1E–20). The 5-year MFS was 53% (95%CI 47–59) in the ‘GGI-high’ class versus 78% (95%CI 72–85) in the ‘GGI-low’ class (P = 3.02E–11), with a corresponding hazard ratio for metastatic relapse equal to 2.92 (95%CI 2.10–4.07; P = 2.23E–10). In multivariate analysis, the GGI-based classification remained significant, whereas the pathological grade did not. In fact, the GGI-based classification stratified the patients with pathological grades 1 and 2 and those with pathological grade 3 in two classes with different 5-year MFS. Comparison of the GGI and CINSARC multigene signatures revealed similar correlations with clinicopathological variables, which were, however, stronger with GGI than with CINSARC, a strong concordance (71%) in terms of low-risk or high-risk classifications, and independent prognostic value for MFS in multivariate analysis, suggesting complementary prognostic information. Conclusion GGI refines the prediction of MFS in operated STS and might improve the tailoring of adjuvant chemotherapy. Further clinical validation is warranted in larger retrospective, then prospective series, as well as the functional validation of relevant genes that could provide new therapeutic targets. cell cycle, Genomic Grade Index, gene expression signatures, soft-tissue sarcoma, survival Key Message The Genomic Grade Index (GGI) is a 108-gene signature previously developed in early-stage breast cancer. We analyzed gene expression and clinicopathological data of 678 localized STS and showed that GGI is an independent prognostic factor for metastatic relapse, which outperforms the pathological grading. Introduction Soft-tissue sarcomas (STSs) are rare tumors that represent a heterogeneous group with at least 50 different pathological subtypes [1]. Surgery is the standard treatment of adult patients with an early-stage STS, but up to 50% of them experience postoperative metastatic recurrence, lethal in most of cases [2]. Results of adjuvant chemotherapy are conflicting, with negative results from the largest randomized study [3], but positive results in terms of relapses in meta-analyses [4, 5]. Therefore, adjuvant chemotherapy is not a standard treatment in adult-type STS, and can be given to high-risk patients (high-grade, deep-seated tumor, tumor size >5 cm) after a shared decision-making with the patient [2]. Pathological grading, most commonly based on the Federation Francaise des Centres de Lutte Contre le Cancer (FNCLCC) grading system, is a good predictor of metastatic relapse in early-stage STS [6], and the most influential for the challenging decision of adjuvant chemotherapy [7]. It distinguishes three increasing-malignancy grades from 1 to 3, based on evaluation of three pathological parameters—tumor differentiation, necrosis, and mitotic index. However, it displays limitations: questionable reproducibility between pathologists, poorly informative prognostic value for grade 2, which represents ∼40% of the cases, nonapplicability in all pathological subtypes, and difficulty of assessment on small tumor biopsies [8, 9]. The same limitations exist for the frequently used three-grade system developed by the National Cancer Institute [10]. Thus, tumor grading remains imperfect to solve STS clinical heterogeneity in clinical practice. In this context, gene expression [11] may improve our capacity to predict clinical outcome. In 2010, the French Sarcoma Group identified and validated a 67-gene expression signature (CINSARC) associated with metastasis-free survival (MFS) in a cohort of 310 STS patients [12]. The signature had been identified using a hypothesis-driven approach, by comparing STS samples with high versus low number of genomic alterations and high versus low pathological grade. To our knowledge, this signature, which defines ‘high-risk’ and ‘low-risk’ patients, represents the most studied prognostic multigene signature in STS, although other prognostic expression signatures have been reported in adults’ STS [13, 14], or more specifically in liposarcomas [15], and leiomyosarcomas [16, 17]. In breast cancer, pathological grade is based on cell differentiation, nuclear pleomorphism, and mitotic index: it includes three classes from 1 to 3, and displays the same weaknesses as in STS. To improve performances, a molecular grading, the Genomic Grade Index (GGI), was constructed, by comparing the gene expression profiles of grade 3 versus grade 1 early-stage tumors [18]. This 108-gene mRNA signature reclassified patients with histologic grade 2 tumors into two groups (‘GGI-high’ and ‘GGI-low’) with high risk versus low risk of recurrence, improved the accuracy of tumor pathological grade, and outperformed its prognostic value in multivariate analysis [18]. It was also associated with pathological response to neoadjuvant anthracycline-based chemotherapy in breast cancer [19], as did CINSARC [20]. Because the respective pathological grades of STS and breast cancer are based on similar major morphological criteria (differentiation, mitotic index), we tested the hypothesis that GGI could improve the prognostic classification of STS, as reported in breast cancer [18]. We analyzed expression data of 678 STS profiled using DNA microarrays and RNA-sequencing, and searched for correlation between GGI-based classification and clinicopathological variables. Patients and methods More details are available in the supplementary Methods file, available at Annals of Oncology online. Soft-tissue sarcoma samples We collected clinicopathological and mRNA expression data of 678 clinical STS samples from four datasets (supplementary Table S1, available at Annals of Oncology online). All samples were from operative specimen of patients operated for nonmetastatic STS and corresponded to untreated primary tumors. The selection of datasets was based according to the following criteria: availability of data in the GEO, Array-Express, or TCGA databases, and the presence of at least 50 informative samples. Gene expression data analysis Data analysis required preanalytic processing. We then applied the GGI signature and the CINSARC signature [12], to each dataset separately, by using strictly the same method as that reported in the original publications: each sample was classified as ‘low’ associated with low risk of relapse, or ‘high’ associated with high risk of relapse. Statistical analysis Our primary end point was the MFS calculated from the date of diagnosis until the date of distant relapse or death from STS. Survival was calculated using the Kaplan–Meier method and curves were compared with the log-rank test. Correlations between the GGI-based groups (low versus high) and the clinicopathological features were calculated with the Student’s t-test or the Fisher’s exact test when appropriate. Uni- and multivariate prognostic analyses were done using Cox regression analysis (Wald test). Results Patients’ characteristics Gene expression profiles of 678 clinically annotated STS samples were available. As presented in Table 1, the median patients’ age was 63 years. The sex ratio was balanced, with 171 females (46%) and 200 males (54%). The most frequent anatomical sites were the extremities (n = 159; 42%), followed by the internal trunk (n = 151; 40%), and the superficial trunk (n = 64; 17%); 160 cases (82%) were deeply seated, below or through the superficial fascia. The most frequent pathological types included liposarcomas (n = 256; 38%), undifferentiated sarcomas (n = 202; 30%), and leiomyosarcomas (n = 149; 22%). Based on the pathological type, 433 samples (65%) were STS with complex genetic profile. The median pathological tumor size at the time of surgery was 9 cm. Fifty-six percent of the samples (n = 171) were FNCLCC grade 3. With a median follow-up of 32 months, 209 of 678 (31%) patients experienced metastatic relapse and the 5-year MFS was 63% (95%CI 59–68; Figure 1A). Based on the GGI signature, 275 samples (41%) were classified as ‘GGI-low’ and 403 (59%) as ‘GGI-high’. Table 1. Clinicopathological characteristics and correlations with the GGI-based classification Characteristics  All, n = 678  GGI classes   P-value  Low, n = 275 (41%)  High, n = 403 (59%)  Age, median  Years (range)  63 (17–93)  62 (25–93)  63 (17–90)  0.955  Sex          0.910    Female  171 (46%)  52 (45%)  119 (46%)      Male  200 (54%)  63 (55%)  137 (54%)    Tumor site          <1.00E–06    Extremity  159 (42%)  26 (23%)  133 (49%)      Internal trunk  151 (40%)  69 (62%)  82 (30%)      Superficial trunk  64 (17%)  15 (13%)  49 (18%)      Head and neck  8 (2%)  2 (2%)  6 (2%)    Depth          0.340    Deep  160 (82%)  56 (78%)  104 (84%)      Superficial  36 (18%)  16 (22%)  20 (16%)    Pathological type          <1.00E–06    Liposarcoma  256 (38%)  151 (55%)  105 (26%)      Undifferentiated sarcoma  202 (30%)  51 (19%)  151 (37%)      Leiomyosarcoma  149 (22%)  43 (16%)  106 (26%)      Myxofibrosarcoma  39 (6%)  14 (5%)  25 (6%)      Others  32 (5%)  16 (6%)  16 (4%)    Genetic profile          2.10E–20    Simple  233 (35%)  149 (56%)  84 (21%)      Complex  433 (65%)  117 (44%)  316 (79%)    Pathological size, median  cm (range)  9 (2–40)  8.75 (2–40)  9 (2–40)  0.909  Pathological FNCLCC grade        9.50E–11    1 and 2  136 (44%)  84 (66%)  52 (29%)      3  171 (56%)  43 (34%)  128 (71%)    Metastatic events  Number of patients  209 (31%)  45 (16%)  164 (41%)  7.58E–12  5-year MFS  Months [95% CI]  63% [59-68]  78% [72-85]  53% [47-59]  3.02E–11  Characteristics  All, n = 678  GGI classes   P-value  Low, n = 275 (41%)  High, n = 403 (59%)  Age, median  Years (range)  63 (17–93)  62 (25–93)  63 (17–90)  0.955  Sex          0.910    Female  171 (46%)  52 (45%)  119 (46%)      Male  200 (54%)  63 (55%)  137 (54%)    Tumor site          <1.00E–06    Extremity  159 (42%)  26 (23%)  133 (49%)      Internal trunk  151 (40%)  69 (62%)  82 (30%)      Superficial trunk  64 (17%)  15 (13%)  49 (18%)      Head and neck  8 (2%)  2 (2%)  6 (2%)    Depth          0.340    Deep  160 (82%)  56 (78%)  104 (84%)      Superficial  36 (18%)  16 (22%)  20 (16%)    Pathological type          <1.00E–06    Liposarcoma  256 (38%)  151 (55%)  105 (26%)      Undifferentiated sarcoma  202 (30%)  51 (19%)  151 (37%)      Leiomyosarcoma  149 (22%)  43 (16%)  106 (26%)      Myxofibrosarcoma  39 (6%)  14 (5%)  25 (6%)      Others  32 (5%)  16 (6%)  16 (4%)    Genetic profile          2.10E–20    Simple  233 (35%)  149 (56%)  84 (21%)      Complex  433 (65%)  117 (44%)  316 (79%)    Pathological size, median  cm (range)  9 (2–40)  8.75 (2–40)  9 (2–40)  0.909  Pathological FNCLCC grade        9.50E–11    1 and 2  136 (44%)  84 (66%)  52 (29%)      3  171 (56%)  43 (34%)  128 (71%)    Metastatic events  Number of patients  209 (31%)  45 (16%)  164 (41%)  7.58E–12  5-year MFS  Months [95% CI]  63% [59-68]  78% [72-85]  53% [47-59]  3.02E–11  Bold entries correspond to significant P-values ≤ 0.05. GGI, Genome Grade Index; MFS, metastasis-free survival. Figure 1. View largeDownload slide Metastasis-free survival (MFS) in patients with STS and according to Genomic Grade Index (GGI). (A) Kaplan–Meier MFS curves in all patients with STS. (B) Similar to (A), but according to the GGI-based classification (‘GGI-low’ and ‘GGI-high’ classes). (C) Similar to (B), but in patients with grades 1 and 2 STS. (D) Similar to (C), but in patients with grade 3 STS. The P-values of the log-rank test are indicated. Figure 1. View largeDownload slide Metastasis-free survival (MFS) in patients with STS and according to Genomic Grade Index (GGI). (A) Kaplan–Meier MFS curves in all patients with STS. (B) Similar to (A), but according to the GGI-based classification (‘GGI-low’ and ‘GGI-high’ classes). (C) Similar to (B), but in patients with grades 1 and 2 STS. (D) Similar to (C), but in patients with grade 3 STS. The P-values of the log-rank test are indicated. GGI-based classification and clinicopathological characteristics There was no correlation between the two GGI classes and patients’ age and sex, tumor depth and size (Table 1). By contrast, correlations were found with tumor site, pathological type, genetic profile, and pathological grade: compared with the ‘GGI-low’ class, the ‘GGI-high’ class included more STS located in extremities and less on internal trunk (P < 1.00E–06), more undifferentiated sarcomas and leiomyosarcomas and less liposarcomas (P < 1.00E–06), more samples with complex genetic profile (P = 2.1E–20), and more samples with pathological grade 3 (P = 9.50E–11). Thus, the ‘GGI-high’ class was more associated with poor-prognosis features than the ‘GGI-low’ class. GGI-based classification and metastatic relapse The 5-year MFS was shorter in the ‘GGI-high’ class (53%, 95%CI 47–59) than in the ‘GGI-low’ class (78%, 95%CI 72–85; P = 3.02E–11, log-rank test; Figure 1B). In univariate analysis (Table 2), the hazard ratio (HR) for metastatic relapse was 2.92 (95%CI 2.10–4.07) in the ‘GGI-high’ class when compared with the ‘GGI-low’ class (P = 2.23E–10, Wald test). Other variables associated with MFS included the pathological type (P = 1.35E–06), with a trend for significance for the pathological grade (P = 8.80E–02). Patients’ age and sex, tumor site and depth, and pathological size were not associated with MFS. Table 2. Univariate and multivariate prognostic analyses for MFS Characteristics    Univariate      Multivariate    n  HR [95%CI]  P-value  n  HR [95%CI]  P-value  Age    371  1.00 [0.99–1.01]  0.902        Sex  Male versus female  371  1.02 [0.70–1.49]  0.909        Tumor site  Head and neck versus extremities  382  0.00 [0.00–Inf.]  0.660          Internal trunk versus extremities    0.77 [0.50–1.18]            Superficial trunk versus extremities    0.81 [0.47–1.40]          Depth  Superficial versus deep  196  0.78 [0.38–1.61]  0.495        Pathological type  Liposarcoma versus leiomyosarcoma  678  0.48 [0.35–0.67]  1.35E–06  307  0.46 [0.23–0.95]  3.60E–02    Myxofibrosarcoma versus leiomyosarcoma    0.45 [0.24–0.86]    307  0.49 [0.24–1.00]  5.03E–02    Undifferentiated sarcoma versus leiomyosarcoma    0.43 [0.30–0.61]    307  0.36 [0.22–0.59]  4.40E–05    Others versus leiomyosarcoma    0.21 [0.08–0.57]    307  0.28 [0.07–1.16]  7.95E–02  Pathological size (cm)    210  1.00 [0.96–1.04]  0.898        Pathological FNCLCC grade  1-2 versus 3  307  1.43 [0.95–2.17]  0.088  307  1.39 [0.87–2.20]  0.166  GGI classes  High-risk versus low-risk  678  2.92 [2.10–4.07]  2.23E–10  307  2.24 [1.34–3.74]  2.05E–03  Characteristics    Univariate      Multivariate    n  HR [95%CI]  P-value  n  HR [95%CI]  P-value  Age    371  1.00 [0.99–1.01]  0.902        Sex  Male versus female  371  1.02 [0.70–1.49]  0.909        Tumor site  Head and neck versus extremities  382  0.00 [0.00–Inf.]  0.660          Internal trunk versus extremities    0.77 [0.50–1.18]            Superficial trunk versus extremities    0.81 [0.47–1.40]          Depth  Superficial versus deep  196  0.78 [0.38–1.61]  0.495        Pathological type  Liposarcoma versus leiomyosarcoma  678  0.48 [0.35–0.67]  1.35E–06  307  0.46 [0.23–0.95]  3.60E–02    Myxofibrosarcoma versus leiomyosarcoma    0.45 [0.24–0.86]    307  0.49 [0.24–1.00]  5.03E–02    Undifferentiated sarcoma versus leiomyosarcoma    0.43 [0.30–0.61]    307  0.36 [0.22–0.59]  4.40E–05    Others versus leiomyosarcoma    0.21 [0.08–0.57]    307  0.28 [0.07–1.16]  7.95E–02  Pathological size (cm)    210  1.00 [0.96–1.04]  0.898        Pathological FNCLCC grade  1-2 versus 3  307  1.43 [0.95–2.17]  0.088  307  1.39 [0.87–2.20]  0.166  GGI classes  High-risk versus low-risk  678  2.92 [2.10–4.07]  2.23E–10  307  2.24 [1.34–3.74]  2.05E–03  Bold entries correspond to significant P-values ≤ 0.05. GGI, Genome Grade Index; MFS, metastasis-free survival. In multivariate analysis (Table 2) including pathological type and grade, and GGI-based classification, two factors remained significant, including the GGI-based classification (P = 2.05E–03). The pathological grade lost its prognostic value. In fact, the GGI-based classification stratified the patients with pathological grades 1 and 2 in two classes with different 5-year MFS (Figure 1C): 74% (95%CI 62–87) in the ‘GGI-low’ class versus 59% (95%CI 46–76) in the ‘GGI-high’ class (P = 3.17E–02, log-rank test). Similarly, the patients with pathological grade 3 were separated in two classes with 82% 5-year MFS (95%CI 70–95) in the ‘GGI-low’ class versus 54% (95%CI 45–65) in the ‘GGI-high’ class (P = 5.20E–03, log-rank test; Figure 1D). Even if the follow-up was relatively limited, GGI better discriminated the relapses during the first 5 years in the whole population (Figure 1B) and in the grade 3 patients (Figure 1D). Comparison with the CINSARC signature Given the potential value of CINSARC in STS [12, 13], we assessed its prognostic value and compared it with that of the GGI. There were 39 genes common to the two signatures, representing 36% of GGI genes and 58% of CINSARC genes. Based on CINSARC, 387 samples (57%) were classified as ‘CINSARC-low’ and 291 (43%) as ‘CINSARC-high’. The CINSARC classes showed the same clinicopathological correlations as the GGI-based classes, with correlations with the tumor site, pathological type and grade, and genetic profile (supplementary Table S2, available at Annals of Oncology online). However, all these correlations were more significant with GGI than with CINSARC. There was also a strong correlation between the GGI and CINSARC classes (P = 3.11E–35; Fisher’s exact test), with 482 of 678 (71%) samples classified similarly as low-risk or high-risk; the 196 discordant samples were more often CINSARC-low/GGI-high than CINSARC-high/GGI-low. CINSARC was strongly associated with MFS with 50% 5-year MFS (95%CI 44–57) in the ‘CINSARC-high’ class and 74% 5-year MFS (95%CI 68–79) in the ‘CINSARC-low’ class (P = 4.77E–11, log-rank test; Figure 2A). In univariate analysis (supplementary Table S3, available at Annals of Oncology online), the HR for metastatic relapse was 2.48 (95%CI 1.87–3.28) in the ‘CINSARC-high’ class when compared with the ‘CINSARC-low’ class (P = 2.03E–10, Wald test). In multivariate analysis (supplementary Table S3, available at Annals of Oncology online) including the CINSARC and the GGI classifications, both classifications remained significant (P = 1.44E–05 for GGI, P = 1.51E–04 for CINSARC). The GGI classification affected the clinical outcome of the CINSARC classes (Figure 2B): the 5-year MFS was 79% (95%CI 72–86) in the ‘CINSARC-low’/‘GGI-low’ class and 65% (95%CI 56–75) in the ‘CINSARC-low’/‘GGI-high’, and 73% (95%CI 60–89) in the ‘CINSARC-high’/‘GGI-low’ class and 46% (95%CI 39–53) in the ‘CINSARC-high’/‘GGI-high’ class (P = 1.1E–13, log-rank test). Figure 2. View largeDownload slide Metastasis-free survival (MFS) in patients with STS according to CINSARC and Genomic Grade Index (GGI). (A) Kaplan–Meier MFS curves in all patients with STS according to the CINSARC-based classification (‘CINSARC-low’ and ‘CINSARC-high’ classes). (B) Similar to (A), but according to the four-class classification based on both GGI and CINSARC. The P-values of the log-rank test are indicated. Figure 2. View largeDownload slide Metastasis-free survival (MFS) in patients with STS according to CINSARC and Genomic Grade Index (GGI). (A) Kaplan–Meier MFS curves in all patients with STS according to the CINSARC-based classification (‘CINSARC-low’ and ‘CINSARC-high’ classes). (B) Similar to (A), but according to the four-class classification based on both GGI and CINSARC. The P-values of the log-rank test are indicated. Discussion The absence of accurate prognostic features, such as pathological grade, and of predictors of response to anthracycline/ifosfamide-based chemotherapy in patients with STS, combined with the scarcity and heterogeneity of the disease, explain in part the difficulty to prove the benefit, if any, of adjuvant chemotherapy. Because pathological grade is considered as the best prognostic factor of STS and given the performance of GGI in breast cancer, we tested the prognostic value of this latter in a large series of 678 operated STS samples. We showed that GGI is an independent prognostic factor for metastatic relapse. To our knowledge, this is by far the largest prognostic study of gene expression profiles in STS. During the last decades, high-throughput molecular analyses, notably gene expression profiling, have provided insights into the extensive heterogeneity of cancers. For example, several multigene signatures such as Oncotype™ or Prosigna™ are marketed in early-stage breast cancer. No similar signature is currently available in clinical routine for STS patients. The scarcity of STS and the paucity of tumor specimens available for analysis explain the relatively small number of samples profiled in previous prognostic studies, 310 in the largest one [12]. We overcame the problem by pooling four public sets including one multicentric series prospectively collected (TCGA) and three unicentric or multicentric retrospective series, representing a total of 678 operated primary cancers. The whole series displayed the expected clinicopathological characteristics and poor prognosis with ∼60% 5-year MFS. Its size allowed multivariate analysis, and the transcriptional nature of data allowed the comparison of the GGI signature with CINSARC, a promising expression signature. When applied to our series of 678 STS, GGI classified 41% of samples as ‘GGI-low’ and 59% as ‘GGI-high’. Of course, none of the STS samples had been used to generate the GGI signature. The HR for metastatic relapse was close to 3 in the ‘GGI-high’ class when compared with the ‘GGI-low’ class. In breast cancer, GGI is strongly associated with pathological grade [18]. Here too, and as expected given the similarity between STS and breast cancer regarding the definition of grade, there was strong correlation with pathological tumor grade, and with other grade-associated clinical variables. Despite the association of the ‘GGI-high’ class with poor-prognosis features, the GGI classification remained an independent prognostic feature in multivariate analysis, and provided additional information to pathological grade by discriminating within the patients with grades 1 and 2 tumor and with grade 3 tumor those with good-prognosis (‘GGI-low’) from those with poor-prognosis (‘GGI-high’). This prognostic superiority of GGI suggests that it is biologically more coherent than pathological grade to capture the proliferation level of tumors, likely thanks to its more quantitative and objective character. Like CINSARC, GGI provides the advantage over the pathological grading of stratifying patients into two groups instead of three, thus facilitating the clinical management. Thus, as in breast cancer [18], GGI improved the prognostic value of pathological grade in STS. We have reported a similar result in patients with GIST [21]. Such transversality of prognostic value in different cancers has been recently reported for CINSARC [22], which, like GGI, likely reflects a fundamental biological property of tumors able to metastasize. Several genes included in the GGI signature and/or overexpressed in the ‘GGI-high’ samples encode potential therapeutic targets involved in cell cycle regulation that could be, if functionally validated, targeted by new drugs: kinases (AURKA/B, BUB1, CDC2, CDK4, CHEK1, NEK2, PLK1/4) and phosphatase (CDC25). Among the prognostic gene signatures previously established in STS [12–17], only two [12, 15]—corresponding to those based on the largest series of cases—combined both validation in an independent tumor set and multivariate prognostic analysis, and only one, the CINSARC signature, concerned all STS pathological types. We thus compared the GGI and CINSARC signatures in our present series. Several similarities were observed. First, there was a relatively strong gene overlap, representing 36% of GGI genes and 58% of CINSARC genes. Second, we found similar correlations with clinicopathological variables (notably pathological grade), which were, however, stronger with GGI than with CINSARC, suggesting more homogeneity within classes and more differences between classes with GGI. Third, there was a strong correlation between the GGI classes and the CINSARC classes with 71% of samples classified similarly as low-risk or high-risk. Fourth, both signatures showed prognostic value in uni- and multivariate analyses, suggesting independent information. The two signatures showed complementarity since each signature was able to stratify each class of the other signature in two subclasses with different MFS; however, the 5-year MFS difference was 14% between the GGI subclasses defined in the ‘CINSARC-low’ class and 27% between those defined in the ‘CINSARC-high’ class, whereas it was lesser between the CINSARC subclasses defined in the ‘GGI-low’ class (6%) and in the ‘GGI-high’ class (19%). Conclusion In conclusion, we have shown that a GGI-based classification of operated STS outperforms the prognostic performances of the FNCLCC grading system in terms of MFS and provides independent information complementary to the CINSARC-based classification. The strength of our results lies in several aspects: the number of 678 samples that, to our knowledge, makes our series the largest prognostic gene expression study reported so far in STS; the biological relevance of involved genes; and the independent prognostic value through three different technological platforms. Limitations include the retrospective nature of the series and associated statistical biases, such as missing data for some variables for some patients or different lengths of follow-up between the different datasets, and the relatively large number of genes included (108 genes) for developing a signature applicable in clinical routine. Efforts are ongoing in breast cancer to improve the clinical applicability of GGI though a qRT-PCR test carried out on formalin-fixed paraffin-embedded samples [23]. But yet, by refining the prediction of MFS, GGI might improve our ability to better tailor adjuvant chemotherapy: patients defined as ‘GGI-low’ could be spared adjuvant chemotherapy, whereas those defined as ‘GGI-high’ should be, all the more so because ‘GGI-high’ tumors, might be more sensitive to anthracycline-based chemotherapy than ‘GGI-low’ tumors, as demonstrated in breast cancer [20]. Additionally, the predictive value of GGI for the pathological response to neoadjuvant chemotherapy deserves to be tested in STS patients. But yet, by helping stratify STS, GGI could revive the interest of adjuvant chemotherapy in STS, and improve patients’ survival. Because of the exploratory nature of our study, further clinical validation is warranted in larger retrospective series, then prospective trials of adjuvant chemotherapy in STS. Funding Institut Paoli-Calmettes (no grant number is applicable). Disclosure The authors have declared no conflicts of interest. Author ORCIDs FB: http://orcid.org/0000-0002-0157-0959. References 1 Wunder JS, Nielsen TO, Maki RG et al.   Opportunities for improving the therapeutic ratio for patients with sarcoma. Lancet Oncol  2007; 8( 6): 513– 524. Google Scholar CrossRef Search ADS PubMed  2 ESMO/European Sarcoma Network Working Group. Soft tissue and visceral sarcomas: ESMO Clinical Practice Guidelines for diagnosis, treatment and follow-up. Ann Oncol  2014; 25(Suppl 3): iii102– iii112. PubMed  3 Woll PJ, Reichardt P, Le Cesne A et al.   Adjuvant chemotherapy with doxorubicin, ifosfamide, and lenograstim for resected soft-tissue sarcoma (EORTC 62931): a multicentre randomised controlled trial. Lancet Oncol . 2012; 13( 10): 1045– 1054. Google Scholar CrossRef Search ADS PubMed  4 Adjuvant chemotherapy for localised resectable soft-tissue sarcoma of adults: meta-analysis of individual data. Sarcoma Meta-analysis Collaboration. Lancet  1997; 350: 1647– 1654. CrossRef Search ADS PubMed  5 Pervaiz N, Colterjohn N, Farrokhyar F et al.   A systematic meta-analysis of randomized controlled trials of adjuvant chemotherapy for localized resectable soft-tissue sarcoma. Cancer  2008; 113( 3): 573– 581. Google Scholar CrossRef Search ADS PubMed  6 Guillou L, Coindre JM, Bonichon F et al.   Comparative study of the National Cancer Institute and French Federation of Cancer Centers Sarcoma Group grading systems in a population of 410 adult patients with soft tissue sarcoma. JCO  1997; 15: 350– 362. Google Scholar CrossRef Search ADS   7 Italiano A, Delva F, Mathoulin-Pelissier S et al.   Effect of adjuvant chemotherapy on survival in FNCLCC grade 3 soft tissue sarcomas: a multivariate analysis of the French Sarcoma Group Database. Ann Oncol  2010; 21( 12): 2436– 2441. Google Scholar CrossRef Search ADS PubMed  8 Coindre JM. Grading of soft tissue sarcomas: review and update. Arch Pathol Lab Med  2006; 130( 10): 1448– 1453. Google Scholar PubMed  9 Coindre JM, Terrier P, Guillou L et al.   Predictive value of grade for metastasis development in the main histologic types of adult soft tissue sarcomas: a study of 1240 patients from the French Federation of Cancer Centers Sarcoma Group. Cancer  2001; 91( 10): 1914– 1926. Google Scholar CrossRef Search ADS PubMed  10 Costa J, Wesley RA, Glatstein E, Rosenberg SA. The grading of soft tissue sarcomas. Results of a clinicohistopathologic correlation in a series of 163 cases. Cancer  1984; 53( 3): 530– 541. Google Scholar CrossRef Search ADS PubMed  11 Bertucci F, Houlgatte R, Nguyen C et al.   Gene expression profiling of cancer by use of DNA arrays: how far from the clinic? Lancet Oncol  2001; 2( 11): 674– 682. Google Scholar CrossRef Search ADS PubMed  12 Chibon F, Lagarde P, Salas S et al.   Validated prediction of clinical outcome in sarcomas and multiple types of cancer on the basis of a gene expression signature related to genome complexity. Nat Med  2010; 16( 7): 781– 787. Google Scholar CrossRef Search ADS PubMed  13 Francis P, Namlos HM, Muller C et al.   Diagnostic and prognostic gene expression signatures in 177 soft tissue sarcomas: hypoxia-induced transcription profile signifies metastatic potential. BMC Genomics  2007; 8( 1): 73. Google Scholar CrossRef Search ADS PubMed  14 Hoffmann AC, Danenberg KD, Taubert H et al.   A three-gene signature for outcome in soft tissue sarcoma. Clin Cancer Res  2009; 15( 16): 5191– 5198. Google Scholar CrossRef Search ADS PubMed  15 Gobble RM, Qin LX, Brill ER et al.   Expression profiling of liposarcoma yields a multigene predictor of patient outcome and identifies genes that contribute to liposarcomagenesis. Cancer Res  2011; 71( 7): 2697– 2705. Google Scholar CrossRef Search ADS PubMed  16 Beck AH, Lee CH, Witten DM et al.   Discovery of molecular subtypes in leiomyosarcoma through integrative molecular profiling. Oncogene  2010; 29( 6): 845– 854. Google Scholar CrossRef Search ADS PubMed  17 Lee YF, John M, Falconer A et al.   A gene expression signature associated with metastatic outcome in human leiomyosarcomas. Cancer Res  2004; 64( 20): 7201– 7204. Google Scholar CrossRef Search ADS PubMed  18 Sotiriou C, Wirapati P, Loi S et al.   Gene expression profiling in breast cancer: understanding the molecular basis of histologic grade to improve prognosis. J Natl Cancer Inst  2006; 98( 4): 262– 272. Google Scholar CrossRef Search ADS PubMed  19 Liedtke C, Hatzis C, Symmans WF et al.   Genomic grade index is associated with response to chemotherapy in patients with breast cancer. J Clin Oncol  2009; 27( 19): 3185– 3191. Google Scholar CrossRef Search ADS PubMed  20 Bertucci F, Finetti P, Sabatier R, Birnbaum D. The CINSARC signature: prognostic and predictive of response to chemotherapy? Cell Cycle  2010; 9( 19): 4025– 4027. Google Scholar CrossRef Search ADS PubMed  21 Bertucci F, Finetti P, Ostrowski J et al.   Genomic Grade Index predicts postoperative clinical outcome of GIST. Br J Cancer  2012; 107( 8): 1433– 1441. Google Scholar CrossRef Search ADS PubMed  22 Lesluyes T, Delespaul L, Coindre JM, Chibon F. The CINSARC signature as a prognostic marker for clinical outcome in multiple neoplasms. Sci Rep  2017; 7( 1): 5480. Google Scholar CrossRef Search ADS PubMed  23 Toussaint J, Sieuwerts AM, Haibe-Kains B et al.   Improvement of the clinical applicability of the Genomic Grade Index through a qRT-PCR test performed on frozen and formalin-fixed paraffin-embedded tissues. BMC Genomics  2009; 10( 1): 424. Google Scholar CrossRef Search ADS PubMed  © The Author 2017. Published by Oxford University Press on behalf of the European Society for Medical Oncology. All rights reserved. For Permissions, please email: journals.permissions@oup.com.

Journal

Annals of OncologyOxford University Press

Published: Feb 1, 2018

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off