Ovarian carcinoma is the first cause of death by gynecologic malignancy in western countries. In 2010 in USA, around 22 000 cases were diagnosed and 14 000 deaths were reported 1 . Such a poor prognosis is due to late diagnosis and relative lack of efficacy of current treatments. The therapeutic sequence used by most of clinicians is maximal cytoreductive surgery (also called debulking surgery) followed by adjuvant chemotherapy for undifferentiated or advanced tumors 2 3 4 5 6 7 . Nevertheless, 20% of patients are initially refractory to this treatment and more than 50% of patients who are initially in complete remission will relapse and ultimately succumb from disease 8 9 . Consequently, overall survival is quite reduced and has remained stable since 20 years (30-40% at five years for all stages). Early stages have a favorable prognosis (~90%), while life expectancy is only 30% after 5 years when disease is extended to peritoneal cavity and only 5-10% when there is distant metastasis 8 9 .

A combination of a platinum agent and paclitaxel is the standard therapy with benefits in terms of response, progression-free and overall survivals, leading in stages III and IV to a median survival of more than 35 months 10 11 . Several laboratory models 12 as well as retrospective analyses of clinical studies 13 14 have strongly suggested that chemotherapy dose could favorably influence ovarian cancer outcome. Major chemotherapy dose intensification using alkylating agents with autologous hematopoietic stem cell support (HSCS) has been investigated in this setting, with encouraging results in pilot studies 15 16 17 18 . However, these promising results have not been confirmed in randomized phase III trials 19 20 , and high-dose chemotherapy (HDC) is currently not recommended for advanced ovarian carcinomas (AOC).

Nevertheless, ovarian cancer is clinically heterogeneous. Patients with morphologically similar, advanced-stage tumors display a broad range of clinical outcomes. Features currently used for prognosis and chemotherapy decision are clinicopathological and include patient’s age, performance status, FIGO stage, histological tumor grade and subtype, initial surgery results and response to chemotherapy. These factors were not incorporated in the initial design of randomized studies although they might be associated with different responses to HDC.

The present study is a retrospective comparative survival analysis, including subsets analysis based on usual clinicopathological features. A survival comparison was done between 103 patients with AOC treated by surgery plus platinum/taxane-based conventional chemotherapy alone (CCA) and 60 patients who received the same treatment plus HDC and autologous HSCS.

Even though HDC plus HSCS cannot be considered as a standard of care for all AOC patients, results from this monocentric comparative retrospective study including 163 patients suggest that it may be beneficial to young patients. In women under 50 years of age, addition of HDC to platinum/taxane-based chemotherapy improves not only PFS (p=0.02), but also OS (median of 54.6 months versus 36 months with conventional therapy alone, p=0.05).

Despite advances in chemotherapy and multidisciplinary management of ovarian carcinomas, the prognosis of patients with advanced stages (FIGO III/IV) remains poor. Median PFS and OS of our cohort treated with a platinum/taxane combination alone (18.1 and 41.3 months, respectively) were similar to those of phase III pivotal studies: 18 and 38 months 10 , and 19.4 and 48.7 months 6 with cisplatin and paclitaxel; 20.7 and 57.4 months for carboplatin and paclitaxel 6 . Our population was thus similar to previously described cohorts. Prognosis is known to be dramatically influenced by cytoreductive surgery and response to adjuvant platinum/taxane-based chemotherapy. However, even good responders to initial treatment often have a poor prognosis due to secondary relapse. Such relapses are generally chemoresistant and remain the major cause of death. Thus, it may be useful to treat chemosensitive patients in order to kill residual clones and avoid the chemoresistant relapse. Different consolidation therapies have been considered: conventional maintenance chemotherapy, intraperitoneal treatment with chemotherapy and/or hyperthermia, and HDC with HSCS. The latter has been widely used in the context of poor risk hematological malignancies and sometimes in chemosensitive solid tumors such as metastatic breast cancer 21 22 23 24 25 or germ cell tumors 26 with controversial results.

The main toxicity of high-dose alkylating agents is hematological. Stem cell transplantation is needed in such treatment strategies to limit the duration and consequences of aplasia. Nevertheless, severe infection can always occur during grade 4 neutropenia and remains the major potential risk during severe aplasia. However we observed no toxic death after HDC in this study.

Several promising but preliminary studies have reported that HDS plus HSCS may improve ovarian cancer outcome in first-line therapy. These results were observed when HDC was used either as front-line treatment 19 27 , or as consolidation therapy 17 28 29 30 31 32 . However published randomized phase III trials did not confirm these results. In a single center small-sized study from Papadimitriou et al. 19 , although PFS was numerically improved by HDC (85.2 months versus 18 months), the difference was not significant (p=0.059). Moreover, no significant difference was observed in OS (not reached after 75 months of follow-up versus 75 months, p=0.38). The authors attributed PFS gain to the higher rates of stages IV (14% vs. 8.1%) and larger post-operative residue (32.6% vs. 21.6%) in the conventional therapy arm. Mobus et al. reported similar findings in their relatively large phase III trial published in 2007 20 . Median PFS was 29.5 months in the HDC arm versus 20.5 in the control arm (p=0.40). There was also no difference regarding OS (54.4 vs. 62.8 months, p=0.54). Conclusions of these studies were that HDC does not improve outcome in advanced ovarian cancer.

Nevertheless a question that could be asked is: are these conclusions relevant for all patients or is there a subset of patients who may benefit from HDC? In this retrospective study, we tried to address this issue using a subgroup analysis approach in a large population of more than 160 patients. We have explored prognostic value of the different histoclinical features used in ovarian cancer evaluation: age, performance status, FIGO stage, histological subtype, histological grade, debulking status and response to conventional chemotherapy. Age was the only parameter correlated to HDC efficacy, both in PFS and OS. Intriguingly, patients under 50 years of age had a gain in survival when HDC was performed after platinum/taxane-based chemotherapy: median OS of 54.6 months vs. 36 months with standard treatment (p=0.05). This benefit was observed independently of the response after standard treatment. A possible hypothesis is that, in young patients known to have a better prognosis than older women, HDC may be more efficient regardless of the persistence of residual disease after conventional therapy. A hypothesis to explain these results could be the higher prevalence of BRCA-related tumors in younger patients compared to sporadic forms 33 34 . Indeed, BRCA-related ovarian cancers display distinctive biological and clinical characteristics including genomic instability, dysfunction in DNA repair processes especially homologous recombination and thereby higher sensitivity to platinum-based chemotherapy and better outcome 35 36 . Of note, recent data have shown that this phenotype could be extended to a larger group of tumors without germline BRCA mutations, the so-called “BRCAness” phenotype 37 38 . Thus, the benefit of alkylating agents-based HDC in younger patients observed in this study may reflect the enrichment in BRCA-related or BRCAness-associated forms in this subgroup and therefore a higher sensitivity of ovarian cancer cells to DNA damages that can be induced by alkylating agents. As suggested by the dose-effect concept, more chemotherapy –and thus more DNA lesions- may lead to an increase in tumor cells death.

A similar exploitation of this Achilles’ heel of the BRCAness-related phenotype was recently demonstrated with the new therapeutic class of PARP1 inhibitors 39 , which also target DNA repair processes. PARP1 inhibitors are able to induce DNA single-strand breaks that will accumulate and degenerate to DNA double-strand breaks, which are not appropriately repaired if the BRCA pathway is deficient or dysfunctional, the so-called synthetic lethality concept. Olaparib has been shown to induce relevant and promising rates of response when used as single agent in AOC. Interestingly, its activity was documented not only in patients carrying BRCA mutations 40 41 , but also in patients without constitutive mutations 42 , further validating the BRCAness concept.

This phenomenon may be increased with the association of PARP inhibitor and alkylating drugs. Such an additive activity may not be necessary in case of complete remission after standard treatment, but may have a positive effect when the tumor burden has been decreased but not eliminated by the initial treatment.

Our observations show that more treatment may be more effective in young patients. Addition of HDC after platinum/taxane-based chemotherapy in this population should be compared to other ways to enhance treatment exposure. Intra-peritoneal chemotherapy may be an option to increase the doses of platinum and/or taxane administered to cancer cells, with less hematological adverse events 43 . Another issue is the lack of studies comparing consolidation (such as HDC) and maintenance therapy, which could be based on cytotoxic treatments 44 as well as angiogenesis inhibitors 45 . Nevertheless it is of note that, except angiogenesis inhibiting agents, none of the treatments cited above has shown his superiority in randomized trials versus observation alone, but without age consideration as we have done in this analysis. These new findings must be balanced with the fact that this study was retrospective, and that HDC regimens were heterogeneous. Nevertheless, despite its retrospective nature, this study, based on a large population, used a comparative design and included subgroup analyses with traditional clinical and pathological prognostic factors. Another limitation of this work is the absence of relevant information about the BRCA status of our patients. Unfortunately, this data was available only for few patients in our retrospective cohort (21 of 163), with only six BRCA1 and two BRCA2 mutations identified.

We have shown in this retrospective comparative study including more than 160 women, that, when applied to all patients, HDC does not improve advanced ovarian cancer survival. However, HDC seems to benefit to young patients (less than 50 years of age). Median overall survival in this subset presented an improvement of 18 months when HDC was performed after initial platinum/taxane-based chemotherapy versus standard chemotherapy alone. This work is the first to make the hypothesis of a differential benefit from HDC according to age. As we know that young patients have a higher frequency of BRCA alterations than older women, they may have a more important benefit from HDC. That may lead to new clinical trials to explore this hypothesis of HDC usefulness in young patients, without or with combination with drugs targeting DNA repair such as olaparib.

AOC: Advanced ovarian carcinoma; CCA: Conventional chemotherapy alone; CCR: Clinical complete response; HDC: High-dose chemotherapy; HR: Hazard ratio; HSCS: Hematopoietic stem cell support; OS: Overall survival; PFS: Progression-free survival; SLS: Second look surgery.

The authors declare that they have no competing interests.

Conception and design: RS. Acquisition of data: RS, AG, MAC, FR, EL, CC, PV, JME. Statistical analysis: RS. Manuscript writing: RS, AG, FB, JME. Final approval: all authors.