Recurrent ovarian cancer remains a difficult disease to treat. Drugs used in the recurrent disease setting include platinum compounds, taxanes, pegylated liposomal doxorubicin (PLD), gemcitabine,topotecan, pemetrexed, and etoposide. Investigational agents with activity include bevacizumab, AZD2171, poly ADP ribose polymerase inhibitors, and epothilones.^1–5 After a trial of three to four such drugs in the recurrent disease setting, patients usually do not derive a clinically meaningful response to additional agents as a result of mechanisms of cross-resistance that are still poorly understood. Each of us can recount exceptions to this rule, especially in the setting of germlineBRCA1 or BRCA2 mutation, but they are not common.

The algorithm for chemotherapy selection in the recurrent disease setting is fairly straightforward.⁶Patients who experience disease progression through first-line, platinum-based therapy (platinum-refractory) or those who experience relapse within 6 months of receiving prior platinum therapy (platinum-resistant) are typically treated with a non–platinum-based regimen, such as single-agent PLD, gemcitabine, weekly paclitaxel, or topotecan, yielding responses in the range of 10% to 15% that last a median of approximately 4 months. Such patients represent a heterogeneous group, with patients with platinum-refractory disease being unlikely to experience an objective response tocurrently available agents and patients with platinum-resistant disease experiencing the occasional good response that provides meaningful palliation (and sometimes even experiencing anotherresponse to platinum rechallenge). Patients who experience relapse after a platinum-free interval (PFI) of greater than 6 months (platinum-sensitive) typically receive a platinum-based regimen, often in combination with a drug like paclitaxel, gemcitabine, or, more recently, PLD,⁷ yielding responses in the range of 30% to 40% or higher, especially for those with a PFI of greater than 24 months.⁸

Unlike platinum-resistant relapse, where single agents are preferred outside of a clinical trial setting, patients with platinum-sensitive disease seem to experience an overall survival (OS) benefit through the use of the combination of paclitaxel and platinum.⁹ This conclusion is based on the results of the International Collaborative Ovarian Neoplasm-4 trial (ICON4), involving 802 patients with platinum-sensitive recurrence who received either single-agent platinum or a combination of paclitaxel and platinum. Thus far, this is the only trial to demonstrate a statistically significant improvement in OS in favor of combination chemotherapy for recurrent ovarian cancer (7% absolute improvement at 2 years; P = .023). ICON4 was well conducted but has some methodologic limitations, including the fact that approximately 40% of patients randomly assigned to the control (single-agent) arm never received a taxane at any point in their disease course, including first-line therapy. In addition, the majority of patients in the control arm did not receive paclitaxel on disease progression, raising the possibility that the sequential use of platinum followed by paclitaxel on disease progression might have conferred the same survival advantage, but with less toxicity. Of note, the OS benefit observed with combination paclitaxel and platinum in ICON4 was restricted to those patients whose PFI was greater than 12 months, although many practitioners seem to extrapolate these data to any patient for whom platinum is considered appropriate, such as those with a PFI of between 6 and 12 months. Regardless of these issues, ICON4 has justifiably broadened the goal of treating platinum-sensitive relapse from one of symptom control to one of also improving survival.

For all the good that ICON4 has done, it seems that this study has unintentionally created a precedent whereby it has become fashionable to accept looser criteria as evidence of benefit for combination therapy in this disease setting. For example, a subsequent study performed by the Arbeitsgemeinschaft Gynaekologische Onkologie compared single-agent carboplatin with the combination of carboplatin and gemcitabine in 356 patients with platinum-sensitive relapse.¹⁰ This trial showed that the combination was associated with a 2.8-month improvement in PFS (P = .0031), greater toxicity than single-agent carboplatin, no improvement in quality of life (QOL), and, unlike ICON4, no improvement in OS (the fact that the study was not powered to detect an OS advantage begs the question). Nevertheless, despite the absence of an OS or QOL advantage, the combination of carboplatin and gemcitabine was ultimately approved for use by the US Food and Drug Administration (FDA). In the post-ICON4 era, is it appropriate to decrease the rigor with which we evaluate subsequent combination trials in recurrent disease, and in particular, is it appropriate to make assumptions about the value of PFS as a surrogate for OS in this setting? These issues will be explored further as we discuss the results of an interesting combination published in this issue of Journal of Clinical Oncology by Monk et al.¹¹

The report by Monk et al¹¹ describes the results of a randomized phase III trial comparing PLD with a combination of PLD and trabectedin in 672 patients with recurrent epithelial ovarian cancer. Trabectedin is derived from the marine tunicate Ecteinascidia turbinata and exerts its cytotoxic effects through binding to the minor groove of DNA, interfering with nucleotide excision repair, and causing lethal DNA strand breaks. This drug has single-agent activity in relapsed ovarian cancer in approximately the same range for platinum-sensitive patients as drugs like topotecan, PLD, and gemcitabine.¹² In addition, there is in vitro evidence to suggest that the combination of PLD and trabectedin is synergistic in its cytotoxicity against sarcoma cell lines.¹³ Finally, there are patients with platinum-sensitive relapse who cannot receive platinum-based combinations because of neuropathy or hypersensitivity. Taken together, these considerations provided the rationale for studying a non–platinum-based combination such as PLD and trabectedin in patients with recurrent ovarian cancer. For the entire patient cohort in the trial by Monk et al¹¹ (both patients with platinum-sensitive and platinum-resistant disease, excluding platinum-refractory disease), the combination of PLD and trabectedin was associated with a 1.5-month improvement in median PFS, no improvement as yet in OS, and no improvement in QOL as compared with single-agent PLD. The hazard ratio (HR) for PFS was 0.79 (95% CI, 0.65 to 0.96; P = .019). In the platinum-sensitive subset, the combination was associated with a 1.7-month improvement in PFS (HR = 0.73; 95% CI, 0.56 to 0.95; P = .017), again with no improvement in OS or QOL. The overall response rate in the entire cohort was higher for the combination compared with single-agent PLD (27.6% and 18.8%, respectively; P = .008). No improvement in PFS, OS, or QOL was observed for platinum-resistant patients.Importantly, some toxicities were worse with the PLD and trabectedin combination compared with PLD alone, including neutropenia requiring growth factor support (42% v 17%, respectively), hyperbilirubinemia (15% v 5%, respectively), and nonfatal congestive heart failure-related diagnoses (n = 6 v n = 1 patients, respectively). In contrast, the incidence of hand-foot syndrome (HFS) was higher in the single-agent PLD arm compared with the combination (18.5% v 3.9%), undoubtedly related to the use of 50 mg/m² of PLD in the control arm as opposed to the reduced dose of 30 mg/m² of PLD in the trabectedin-containing arm. In this regard, many of us do not use single-agent PLD at a starting dose of 50 mg/m² because it is associated with a higher risk of HFS compared with the more commonly used PLD dose of 40 mg/m² (which seems to be equally effective, albeit on the basis of cross-trialcomparisons).^14–16 One wonders whether the QOL for patients treated with single-agent PLD would have actually been superior to that of the PLD and trabectedin combination had the better-tolerated PLD dose of 40 mg/m² been chosen for comparison.

The study by Monk et al¹¹ is interesting because it is the largest randomized trial to address the potential value of a non–platinum-containingcombination in patients with recurrent ovarian cancer. It is also important because it forces us to critically consider what level of evidence is required to approve a new treatment option in this disease setting. It is reasonable to judge the value of a new regimen on whether it improves OS, QOL, or both.¹⁷If improvement in PFS does not translate into an OS or QOL advantage, the rationale for using PFS as the sole metric for defining the value of a new treatment option becomes difficult to justify (unless the new regimen is shown to be equally effective but better tolerated compared with the control). If we do not yet have mature OS data, can we be confident that a statistically significant increase of 6 weeks in median PFS in this study can serve as a surrogate for improved OS in recurrent ovarian cancer? At the moment, we cannot. In contrast to recurrent disease, data involving 5,826 patients enrolled on 14 randomized trials in the first-line setting strongly suggest that PFS might serve as a reliable surrogate for OS.^18,19 This conclusion is supported by the tight correlation observed between the HRs for PFS and OS when such studies are collectively evaluated.^20,21 A similar analysis has been conducted in advanced colon cancer, with identical conclusions.²² This has led the FDA/American Society of Clinical Oncology/American Association for Cancer Research public workshop on clinical trial end points in ovarian cancer to conclude that “the strength of PFS as a surrogate for OS is especially clear in first-line therapy. PFS seems to correlate with OS, especially when a large effect on PFS is seen. A small increase in PFS may not correlate with OS, however. …”²⁰ In contrast, an equivalently robust analysis has not yet been performed to convincingly demonstrate the value of PFS as a surrogate for OS in recurrent disease. This is due to the relative lack of well-powered, mature randomized trials in recurrent disease that would allow a convincing correlation to be made between the HRs for PFS and OS. Indeed, the current lack of evidence to support the use of PFS as a reliable surrogate for OS in recurrent disease is acknowledged by the FDA/American Society of Clinical Oncology/American Association for Cancer Research workshop as follows: “Data supporting the validity of PFS as a surrogate for OS in second- and third-line therapy are less clear than is the case for first-line therapy. Additionalstudy of this issue would be worthwhile.”²⁰ Given the uncertainty regarding whether PFS can serve as a reliable surrogate for OS in recurrent ovarian cancer, it would seem prudent to await mature data from the study of Monk et al¹¹ before assuming that a 6-week improvement in PFS will have a clinically meaningful impact on OS. This is especially important in view of the increased toxicity and lack of improvement in QOL observed with the PLD and trabectedin combination. Furthermore, it is likely that a minimum amount of improvement in PFS will need to be observed in such studies before it can be assumed to translate into a clinically meaningful OS benefit. In this regard, it is interesting to note that the median PFS improvement in ICON4 was 3.0 months, which is double the amount of PFS improvement seen in the current trial.

There is another aspect to assessing PFS in the setting of a randomized phase III trial that should be considered, quite apart from whether it is a good surrogate for OS in recurrent disease. It involves whether bias exists in either the timing or frequency of evaluations as a function of whether the patientwas enrolled on the control or experimental arm. The trial of Monk et al¹¹ was not placebo-controlled, meaning that both the patient who received PLD alone and her physician were aware that she was randomly assigned to the control arm. From the patient’s perspective, there might be concern that she is receiving inferior therapy, making it more likely for her to report symptoms to her physician. Her physician may likewise share the same concern, which would lower the threshold for obtaining follow-up studies to evaluate these symptoms. When this situation exists in a randomized trial, there is a risk of evaluation bias, whereby PFS is artifactually determined to be shorter in the control arm because the patient and her physician may be tempted toperform assessment earlier and more frequently than prescribed by the study.²³ A placebo-controlled trial might minimize the likelihood of this bias, but it does not eliminate it completely, because it is often easy to know whether the patient is receiving experimental therapy (as would have been the case in this trial, on the basis of the increased hematologic toxicity of the combination, as well as the use of different doses and schedules of PLD in the two treatment groups). The possibility of evaluation bias is especially important to consider when the magnitude of PFS improvement is roughly in the same range as the interval between protocol-specified assessments, as was the case in this trial. I cannot say whether evaluation bias exists in this study, butmoving forward it would be useful for all phase III trials to report such data for consideration. Unless the possibility of evaluation bias can be excluded, small PFS differences in a randomized trial must be viewed cautiously.

Finally, it is sometimes argued that PFS is a useful metric because, in contrast to OS, it is not influenced by the use of subsequent therapies that may confound interpretation of OS. In the setting of recurrent ovarian cancer, I do not find this reasoning to be convincing. In the Monk et al¹¹ trial, patients randomly assigned to the control arm (single-agent PLD) were not crossed over to trabectedin at the time of disease progression and therefore could not have derived benefit from this agent. Such patients receive a variety of standard agents on disease progression after they leave the controlled environment of the clinical trial. Let us assume that the survival of these control patients can be “rescued” through the use of standard drugs afterwards, such that an OS benefit is not demonstrated in this trial with long-term follow-up. In my view, such an observation serves to place the alleged value of the combination into a real-world perspective that has practical relevance for my patients. Why should we be interested in the combination if an alleged OS advantage is so tenuous as to be completely negated by whatever else is done once the patient leaves the study using standard drugs readily available to practicing oncologists? I propose that if the combination regimen cannot stand on its own and confer an OS benefit in the context of other subsequent treatment options, and if the combination is associated with increased toxicity and does not confer a QOL benefit, then perhaps we should not be interested in using it. Again, I am referring to a study design in which cross-over did not occur, as was the case for the trial under discussion. Under those circumstances, if an OS advantage in favor of the combination cannot be demonstrated with long-term follow-up, it does not matter to me whether the trabectedin combination failed or whether other salvage therapies succeeded. What matters is that my patient unfortunately did not live longer, and did not live better, despite receiving the more aggressive and complicated regimen.

These considerations suggest that relying exclusively on PFS in the recurrent disease setting may be misleading, especially for relatively small improvements in PFS using a regimen that has increased toxicity. In July 2009, the Oncology Drugs Advisory Committee did not recommend approval of the PLD and trabectedin combination studied by Monk et al¹¹ on the basis of several considerations, including lack of an OS advantage and increasedtoxicity of the regimen compared with PLD alone. The FDA is appropriately waiting for mature OS data from this trial before reevaluating this regimen. Two months later, on the basis of the same data, this combination was approved for use in platinum-sensitive relapse in Europe by the European Medicines Agency. The fact that two regulatory agencies can come to different conclusions about the same study shows how confusing it can be to assess a new regimen in the recurrent disease setting. If an OS advantage in favor of the combination eventually emerges from this trial, the trabectedin and PLD regimen may well represent an important option for patients with platinum-sensitive relapse who cannot tolerate platinum as a result of neuropathy or hypersensitivity. Even so, in patients with platinum-sensitive relapse who have no contraindications, platinum-based therapy should still be considered the treatment of choice because we do not yet know how PLD and trabectedin might compare with a regimen such as paclitaxel and carboplatin in that setting. Regardless of whether the trabectedin and PLD regimen is ultimately FDA-approved, it goes without saying that Monk et al¹¹ should be congratulated for conducting an important study of an interesting drug. Such trials are difficult to perform, require an extensive amount of time and resources, and also require the participation of hundreds of altruistic patients who are our partners in moving the field forward. However, as difficult as these studies are to conduct, we must always remain objective about their results and how they affect the lives of our patients.

source:JCO