Should MMMT still be treated with adjuvant taxane-based combination chemotherapy?

Carcinosarcoma or malignant mixed Mullerian tumors (MMMT) are biphasic tumors that contain malignant mesodermal and epithelial components, in contrast to sarcomas, which contain exclusively malignant mesodermal elements (Kernochan and Garcia 2009). Typically, the metastatic sites mainly consist of carcinoma elements (Silverberg et al. 1990). Due to their rareness, there are no large epidemiological studies available and data are in general inconclusive. It has been proposed that the epithelial and sarcomatous elements develop from the same stem cell, with the carcinoma component being undifferentiated and of a specific histotype, whilst the sarcomatous component shows a mixed homologous appearance (McCluggage 2002; Gorai et al. 1997). Relapse-free and overall survival in MMMT-E is poor, and treatment failure at the time of recurrence appears to be at distant sites (Gadducci et al. 2002; Gonzalez Bosquet et al. 2010; Yamada et al. 2000; Manolitsas et al. 2001; Vorgias and Fotiou 2010). Surgical therapy is the initial treatment modality and necessitates hysterectomy, bilateral salpingo-oophorectomy and at least pelvic lymphadenectomy (NCCN 2001). If metastatic disease is present, there is a need for full cytoreduction, which can be achieved in 57% of patients, resulting in a significant improvement of median overall survival (OS 53.3 vs. 8.6 months) (Tanner et al. 2011).

The benefit of chemotherapy for MMMT-E has been clearly shown, with active substances being cisplatin, carboplatin, ifosfamide, anthracyclines and paclitaxel (Sutton et al. 2000, 2005). It has been proposed that FIGO Stage I patients should be treated with 4 cycles and Stage II–IV patients with 6 cycles of carboplatin and epirubicin (Manolitsas et al. 2001; Berton-Rigaud et al. 2014; Wolfson et al. 2007). A recent Cochrane review including 579 patients from two randomized studies compared adjuvant radiotherapy with combination chemotherapy (ifosfamide with paclitaxel versus ifosfamide alone) in recurrent Stage III/IV MMMT-E, finding an improved OS in the combination chemotherapy treatment arms (Sutton et al. 2000; Homesley et al. 2009; Galaal et al. 2011). The promising survival data from various studies and the reduced toxicity found in the carboplatin and paclitaxel combination treatments which are predominantly used in endometrial and ovarian adenocarcinoma, led several institutions, including our own, to change the chemotherapy regimen in MMMT-E to this more tolerable combination. However, there is no established consensus for therapeutic management in this patient group (Berton-Rigaud et al. 2014; NCCN 2001). With only a few prospective randomized controlled trials reported in MMMT-E, the optimal chemotherapy modality is yet to be determined, particularly in view of new targeted therapies.

Therefore, this study retrospectively analyzed the outcomes of a large cohort of MMMT-E to evaluate the response to various kind of adjuvant chemotherapies.

We analyzed 103 histologically confirmed MMMT-E cases against 378 controls, consisting of 172 EC, 189 OC and 17 MMMT-O. The clinicopathological characteristics showed statistically significant differences for death from disease, age, BMI, FIGO Stage, grade, histotype, type of chemotherapy, adjuvant radiotherapy, residual disease, and lymph node dissection (P ≤ 0.001, Table 1). Approximately, one-third of patients with MMMT-E (31.1%) and MMMT-O (35.5%) died due to their disease by the end of this study, in contrast to one tenth (9.9%) of EC and one-fifth (19%) of OC patients. Almost two-thirds (64.9%) in the MMMT-E study group and three quarters (76.4%) of the EC cases were early FIGO Stage I/II patients, in contrast MMMT-O (17.6%) and OC (32.4%). MMMT-O and OC patients were mainly advanced FIGO Stage III/IV cases (82.4% and 67.6%, respectively). Significantly more undifferentiated and high-grade cancers were found in MMMT-E and MMMT-O patients (82.2 and 93.3%, respectively) than in EC and OC patients (33.5% and 66.5%, respectively). Mainly, mixed and endometrioid histotypes were found in the MMMT-E study group (“other”, 71.4%) and in MMMT-O (55.6%) and EC (83.11%) patients; whereas OC patients were predominately of serous (66.7%). A total of 222 patients received adjuvant chemotherapy, hereby platinum and either different anthracyclines or ifosfamide were mainly given in MMMT-E (88.7%) and less in EC (31.6%), MMMT-O (42.7%) and OC (14.7%). In contrast, a platinum/taxane regimen was used most commonly in OC (85.3%), MMMT-O (57.1%), EC (68.7%) and only in 11.3% in MMMT-E. Adjuvant radiotherapy was applied in 65.3% to MMMT-E, 45.9% to EC, and 26.7% to OC.

We compared the long-term outcome, expressed as the cumulative risk of relapse, over a period of 20 years regardless of FIGO Stage. The cumulative relapse risk initially increased for all four cancers, strongest for MMMT-O patients and to comparable extents for OC, EC, and MMMT-E (Fig. 1). Intriguingly, however, the cumulative risk for MMMT-E patients remained stable, namely reaching a plateau after 2.5 years until the end of the observation period (20 years); whereas, it further increased for the three other cancers to differing extent over this time period. We speculated as to whether the observed divergent survival results for MMMT-E and EC was dependent on the FIGO Stage. The relapse-free survival of MMMT-E and EC were, therefore, compared for early Stage (FIGO I/II) (Fig. 2a) and late Stage (FIGO III/IV) (Fig. 2b) patients in a Kaplan–Meier presentation. Indeed, the relapse-free survival rate of early Stage MMMT-E patients decreased to a greater extent within the first 2.5 years when compared to EC patients, but then remained stable at 0.75, meaning that 75% of MMMT-E patients remained without any case of recurrence occurring for 20 years. The relapse-free survival rate of early FIGO stage EC patients was significantly different from that of the MMMT-E patients (strong intersection of the curves, test for proportional hazard P < 0.001) and decreased to a lesser extent than in MMMT-E patients within the first 2.5 years but then further decreased within the subsequent 11.5 years. The respective relapse-free survival rates for advanced FIGO Stage MMMT-E and EC patients decreased to a comparable extent (proportional hazard P = 0.118).

We also analyzed whether the choice of the chemotherapy regimen affected the observed favorable relapse-free survival of MMMT-E patients, regardless of FIGO Stage. Indeed, a significantly shorter time to relapse for patients receiving platinum and taxanes than those receiving platinum and anthracyclines or ifosfamide was observed (HR 4.69, CI 1.23–17.87, P = 0.024, Fig. 3). This indicates that MMMT-E patients who received taxane-free platinum combination chemotherapy and particularly Stage FIGO I/II MMMT-E patients who do not relapse within the first 2.5 years have excellent long-term survival outcomes.

The increased risk of relapse in MMMT-E patients receiving taxane-based chemotherapy prompted us to also examine whether a similar effect could be present in MMMT-O. We, therefore, examined the mutational load and the protein expression of various drug targets in a large independent US-American cohort of MMMT-O (N = 450). MMMT-O and HGSOC (N = 1290) displayed a similarly high load of p53 mutations (77.8% vs 80.2%, Fig. 4a). In contrast, significant differences were found for KRAS and PI3KCA mutations, both being more frequent in MMMT-O compared to HGSOC (KRAS: 5.7% vs 2.4%, P = 0.015; PI3KCA: 6.2% vs 3.3%, P = 0.018).

The majority of proteins selected showed a statistically significant difference in the expression between MMMT-O and HGSOC (Supplement Table S1). These results, therefore, suggest that crizotinib (targeting ALK), cetuximab (EGFR), topotecan (TOPO1), anthracyclines (TOP2A), and alkylating agents such as ifosfamide (MGMT loss) would have a potential treatment benefit in MMMT-O. We next calculated a presumed sensitivity for the chemotherapy combination carboplatin/taxane compared to carboplatin/anthracycline or carboplatin/alkylating agents in MMMT-O. Hereby, the prediction of sensitivity to platinum/taxane combination was based on the loss of expression of ERCC1 and TUBB3 proteins, the prediction for the platinum/anthracycline combination was based on ERCC1 loss and TOP2A expression, and the prediction of platinum/alkylating agents combination was based on ERCC1 loss and MGMT loss. The results demonstrated for MMMT-O a presumed chemotherapy sensitivity of 39.5% for a combination of carboplatin/taxane, of 73.9% to a combination of carboplatin/anthracyclines, and of 24.2% to a combination of carboplatin alkylating agents (Fig. 4b). These data suggest that not only MMMT-E but also MMMT-O patients may benefit from a taxane-free chemotherapy.

It is increasingly important to define malignant diseases in relation to their genomic similarity instead of the organ of origin. This has been nicely shown in the M-PACT-trial from the US-National Cancer Institute, where the intention was to detect the molecular signature of diseases, rather than classify them by the tissue of origin. Subsequently, patient outcomes were examined according to mutations for p53 and PIK3CA. Whilst the curves initially looked genetically driven, they diverged when the researchers looked at the tissue of origin (Schott et al. 2015).

MMMT of endometrial and ovarian origin share the same histological signature but their genetic similarity is still widely unknown. As MMMT-E and MMMT-O derive from different organs, they are not managed identical but due to the tissue of origin. One study suggests that MMMT-O and MMMT-E are different diseases in terms of their genetic landscape. In this study, 110 MMMT-O, 141 MMMT-E and 1587 OC of all histotypes where compared. TP53 was the most commonly mutated gene in all three cancers with 76.4% in MMMT-O, 68.8% in MMMT-E and 69% in OC. Genetic alterations of PI3K/AKT/mTOR and MAPK pathways were noted to be similar in MMMT-O and OC but less frequent in MMMT-E (p < 0.001). In OC, the chance of having a BRCA1/2 mutation was highest compared to MMMT-E and MMMT-O (20% and 18% vs. 9%, respectively) (Mahdi et al. 2015). These data are similar to our own NGS analysis; however, we have 4-times more numbers of MMMT-O included and compared them to the most aggressive subtype of OC, HGSOC. Our own data in conclusion with the literature demonstrate a higher rate of KRAS and PI3KCA mutations in MMMT-O compared to HGSOC.

Since MMMT are known to be metaplastic carcinoma, they are no longer considered a subtype of sarcoma or managed as such. Instead, despite the lack of specific data, the management of MMMT has been extrapolated from studies of EC and OC (Berton-Rigaud et al. 2014; Cantrell et al. 2015). In the past 13 years, 9 GOG trials were performed in MMMT-E and MMMT-O. In total, 21 studies were found in our systematic literature search. Hereby, 16/21 were performed in MMMT-E only, 4/21 in MMMT-O only and 1/21 in both types (Table 2). In total, 1214 patients were included in these heterogeneous studies. The largest GOG trial incorporated 206 patients, but most studies examined MMMT numbers below 100, which clearly limits its results. Cisplatin-based chemotherapy and adjuvant setup were most commonly studied (67%), hereby the drug combination was mainly platinum plus paclitaxel or ifosfamide. Throughout these investigations, with mostly insufficient numbers of MMMT-E and MMMT-O patients, best response rate of a combination of carboplatin and paclitaxel was 62% and 55%, respectively, and 5-year overall survival 62–88% and 30%, respectively (Table 2). These data differ from our own results and might be due to the heterogeneity of the various cohorts and the small patient numbers in the published literature.

Targeted drug trials have been scarce for MMMT with only two studies examining the role of VEGF-directed therapy, one with pazopanib (MMMT-E), the other with aflibercept (MMMT-E and MMMT-O), both demonstrating only a minimal efficacy (Campos et al. 2014; Mackay et al. 2012). Our large retrospective case–control study with over 20 years of follow-up has clearly shown that the combination of carboplatin/taxanes is less efficient in MMMT-E. Our exploratory data further suggest that the same could be the case for MMMT-O.

The concept of personalized treatment is based upon NGS, CISH and IHC data on potentially targetable biomarkers that describe an individual molecular footprint of a tumor (Janssens et al. 2017). Using precision IHC, we found proteins that were differentially expressed between MMMT-E and MMMT-O. It has been hypothesized that these markers could be associated with the likelihood of response to chemotherapy. If confirmed, this may explain the poorer performance of the current standard treatments. Obviously, this is only a hypothesis generating observation and a carefully designed prospective clinical trial would be necessary to validate these findings.

The rarity of MMMT, in particular of MMMT-O, accounts for the fact that subgroup analyses in randomized controlled chemotherapy trials can rarely be performed because of a lack of statistical power. Secondly, different histopathological definitions of MMMT make it difficult to compare studies that focus on this group. Thirdly, many reports combine MMMT-E and MMMT-O with their organ-based adenocarcinoma counterparts. Treatment options tailored to the mutational driver p53 or KRAS/PI3KCA as shown in our NGS analysis should, therefore, be considered in future studies to tailor treatment in relation to the genetic origin. For this, collaborative studies within trial networks performing whole genomic sequencing of these tumors is needed to identify potential targeted therapies for the future.