|
Advertisement |
|
|
||
 |
|
|||||
|
|||||
|
 |
||||||
|
||||||
|
||||||
|
© 2003 American Society for Clinical Oncology Developmental Chemotherapy and Management of Recurrent Ovarian CancerFrom the Fox Chase Cancer Center, Philadelphia, PA. Address reprint requests to Michael A. Bookman, MD, Fox Chase Cancer Center, 7701 Burholme Ave, Philadelphia, PA 19111; email: ma_bookman{at}fccc.edu.
Despite improvements in median and overall survival using a combination of platinum and paclitaxel, long-term survival rates for patients with advanced epithelial ovarian carcinoma (EOC) remain disappointing, and the development of more effective primary therapies remains a priority. In particular, several nonplatinum agents have demonstrated activity in phase II trials among patients with recurrent platinum-resistant EOC. These agents include gemcitabine, topotecan, liposomal doxorubicin, and prolonged oral etoposide. Preclinical models have indicated a biologic basis for combinations of these agents with platinum, which has been attributed to inhibition of pathways involved in DNA repair. However, efforts to develop multidrug combinations with platinum and paclitaxel have encountered substantial bone marrow toxic effects, necessitating significant dose reductions and prompting exploration of alternative schedules and sequences of drug administration. In this regard, the Gynecologic Oncology Group (GOG) and other organizations have conducted pilot studies in previously untreated patients to define combinations that are suitable for group-wide phase III trials. With international collaboration, GOG has launched a five-arm trial that will compare four new combinations against carboplatin and paclitaxel. The selection of candidate regimens for this trial illustrates the challenges of drug development in EOC, emphasizing the role of phase II trials in patients with recurrent disease and indicating potential strategies to help evaluate newer biologic and molecular-targeted reagents.
THE PRIMARY treatment of epithelial ovarian cancer (EOC) has evolved from single alkylating agents (eg, melphalan) to cisplatin and cisplatin-based combinations, followed by incorporation of paclitaxel, but not without a degree of controversy. EOC is considered to be a chemoresponsive neoplasm, with initial response rates to systemic therapy exceeding 80% when integrated with primary cytoreductive surgery. However, among women with advanced-stage disease at diagnosis, long-term survival remains poor because of eventual tumor recurrence and emergence of drug-resistant disease. These observations have stimulated a growing international program for development, evaluation, and registration of new cytotoxic and biologic agents, including organoplatinum derivatives, taxanes, nontaxane antimicrotubular agents, nucleoside analogues, antifolates, and inhibitors of topoisomerase and other enzymes involved in DNA synthesis and repair, many of which are derived from natural sources. In view of the central role of platinum compounds, there has been particular interest in agents that may accentuate the platinum response through inhibition of DNA repair (Table 1  ). Recent clinical trials have focused on combinations with topotecan, gemcitabine, liposomal doxorubicin, and oral etoposide as potential candidates for incorporation in front-line phase III trials. However, efforts to develop tolerable multidrug combinations with platinum and paclitaxel have encountered substantial hematologic toxic effects, prompting exploration of alternative schedules and sequences of drug administration. Recent studies have also explored molecular-targeted reagents directed against growth factors, cellular receptors, angiogenesis, signal transduction pathways, apoptosis, and cell-cycle regulation. Finally, it is widely anticipated that molecular profiling of tumor genotype, DNA methylation status, mRNA expression, patterns of protein synthesis, and pharmacogenomic pathways of drug metabolism will better predict individual tumor response and host toxicity. These advances in drug development and tumor biology will overwhelm our traditional approach to clinical trials, which has been guided by dose-limiting toxicity rather than biologic targeting or optimization of tumor response.
When evaluating clinical response data, it is important to remember that maximal cytoreductive surgery is a standard component of initial disease management and, when successful, is associated with improved long-term clinical outcomes. Among women with advanced EOC, approximately 60% will undergo optimal cytoreduction after initial surgery and 40% will have macroscopic residual disease that exceeds 2 cm, but only half of those women (20% overall) will have disease that is measurable with radiographic imaging. In addition, although CA-125 levels are frequently elevated at diagnosis, a rapid decline during the first few weeks of treatment reflects the net effect of cytoreductive surgery and initiation of chemotherapy. As such, interpretation of clinical response rates in the context of small phase II trials of platinum-based chemotherapy will have limited predictive value. This ability to achieve optimal cytoreduction reflects the tendency of ovarian cancer to remain clinically confined to the peritoneal cavity, characterized by superficial implants without distant metastatic or parenchymal disease. Thus, patients with tumors amenable to cytoreduction will generally have well-differentiated lesions without deep-tissue invasion or dense adhesions, and this contributes to an improvement in long-term outcomes. This unusual biology has also prompted evaluation of various intraperitoneal treatment modalities, including cytotoxic chemotherapy, radioisotopes, biologics, and monoclonal antibodies, as well as the use of whole-abdomen external-beam radiation. Characterization of the local peritoneal milieu from the perspective of growth factors, angiogenesis, and the host immune response has identified some unique features that may have an effect on the pathogenesis of EOC and guide future treatment strategies,1 but these features are largely beyond the scope of this article.
There are a number of scientific, statistical, clinical, and practical criteria that contribute to the effect of phase III trials, including toxicity, complexity, quality of life, costs, and timing. These points are important to consider in the design of future clinical trials to most efficiently use our limited clinical resources and obtain maximal long-term benefit for our patients. No phase III trial is perfectly symmetric or free of confounding variables. In addition, each trial needs to be considered within the historical context in which it was developed. Although tempting, historical comparisons between trials (over time) have the potential for misinterpretation because of evolving practices, with an effect on case selection, clinical management, and treatment outcomes. These factors include changes in supportive care, such as the widespread availability of multiple hematopoietic growth factors; changes in the timing, sensitivity, and variety of diagnostic studies to confirm disease recurrence; liberalized patient selection criteria for primary and secondary cytoreductive surgery together with a reduction in second-look laparotomy; expanded availability of investigational and noninvestigational treatment options for management of recurrent disease; and a reduction in the extent of symptomatic disease at recurrence. The experience with several phase III trials serve to illustrate these points, including the addition of anthracyclines, the role of intraperitoneal chemotherapy, early adoption of paclitaxel in combination with carboplatin, and combined agents versus sequential single agents. The impact of doxorubicin was evaluated in combination with cisplatin and cyclophosphamide in EOC, and the combination achieved a small but significant effect on long-term survival evident in a meta-analysis of four randomized trials2 without a significant clinical benefit that was detectable within each individual trial. Although the magnitude of the benefit was small (change in median survival of approximately 3 months), a significant P value of .02 was achieved because of the large number of assessable participants (N = 1,194). These data were published in 1991 and did not substantially alter clinical practice in the United States because of concerns about added toxicity and interest in emerging data with paclitaxel. However, the topic was revisited in the postpaclitaxel era through two phase III trials comparing a three-drug combination with epirubicin against carboplatin and paclitaxel.3,4 Preliminary results from these studies indicate that epirubicin adds hematologic and mucosal toxic effects with a small improvement in tumor response rate but without an advantage in either progression-free or overall survival. The use of intraperitoneal chemotherapy is a second illustrative case. Two trials that incorporated intraperitoneal cisplatin documented a modest but statistically significant improvement in survival compared with intravenous cisplatin in patients who underwent optimal initial cytoreductive surgery.5,6 However, once again, these trials have not substantially altered clinical practice. The role of intraperitoneal therapy is subject to considerable scientific and clinical bias, and the use of a peritoneal catheter adds complexity and potential toxicity, even though it can reduce the nonhematologic toxic effects associated with intravenous cisplatin. In addition, with the widespread adoption of intravenous carboplatin, the nonhematologic toxic effects associated with cisplatin are avoided, which negates some of the potential advantages afforded by intraperitoneal cisplatin. The first trial5 accrued more slowly than expected and was expanded to include more patients with minimal residual disease. This prolonged the overall duration of the study and time required for availability of mature results, which were ultimately published after promising data with other new agents, such as paclitaxel, were already being disseminated. Of interest, the subpopulation with small-volume disease that was postulated to have the greatest potential for benefit from intraperitoneal therapy did not show any benefit, raising questions regarding limited intraperitoneal exposure and the importance of noncytotoxic effects on the peritoneal cavity, such as inflammation and fibrosis, that might have an effect on larger-volume disease. The second trial6 included two additional cycles of intravenous dose-intense carboplatin on the experimental arm, which was a novel idea but that complicated any analysis of potential benefit. A third randomized trial7 has confirmed an advantage in disease-free survival but, again, at the expense of increased hematologic and nonhematologic toxic effects, failing to resolve the role of intraperitoneal therapy and promoting discussion of conducting a fourth trial. To date, all the phase III studies have been limited to the Gynecologic Oncology Group (GOG) because there has not been any international enthusiasm for evaluating this particular question. Bias can also occur in the opposite direction, promoting early adoption of new regimens, such as carboplatin and paclitaxel, before mature results from phase III trials are achieved. Early experience with single-agent paclitaxel in recurrent disease generated considerable enthusiasm, and data were eagerly awaited from front-line trials. Empiric combinations with cisplatin were rapidly developed and evaluated in GOG-111, in which a total of 410 previously untreated women with suboptimal stage III and IV EOC were randomly assigned to receive either cisplatin plus cyclophosphamide or cisplatin plus paclitaxel (at 135 mg/m2 over 24 hours). The cisplatin plus paclitaxel regimen was judged to be superior to cisplatin plus cyclophosphamide, with an increased median progression-free interval (18 v 13 months, respectively) and overall survival (38 v 24 months, respectively).8 Favorable data were confirmed by a European-Canadian intergroup trial (OV-10) with similar arms, although paclitaxel was administered at 175 mg/m2 over 3 hours.9 In both trials, the use of paclitaxel, which was an investigational agent at the time, was largely restricted to initial protocol-mandated therapy, and only a small number of patients in the control arm ever received paclitaxel for management of recurrent or progressive disease. This prospective control over the choice of subsequent chemotherapy would tend to maximize any differences in clinical outcomes attributed to the use of paclitaxel in primary therapy and undoubtedly contributed to the positive survival data reported from these important randomized trials. On the basis of the favorable results and manageable toxic effects, the combination of paclitaxel with cisplatin rapidly emerged as the standard of care for advanced-stage EOC. However, in view of the nonhematologic toxic effects and required hydration associated with cisplatin, as well as the inconvenient 24-hour infusion schedule for paclitaxel, there was considerable interest in switching to an outpatient regimen with carboplatin and 3-hour paclitaxel. Despite concern about bone marrow toxicity, combinations of paclitaxel with carboplatin proved relatively easy to develop. Phase I studies in lung cancer and EOC documented that both agents could be safely administered in combination at nearly full single-agent doses without a requirement for hematopoietic growth factors. The relative convenience of this outpatient regimen facilitated rapid adoption by the oncology community, even before the completion of phase III randomized trials. Ultimately, GOG-158 documented that carboplatin and paclitaxel (3-hour administration) was not inferior when compared with cisplatin and paclitaxel (24-hour administration),10 validating a regimen that was already widely adopted within the community. However, the noninferiority statistical design required more than 800 assessable patients. Meanwhile, other studies have indicated that single-agent or sequential therapy may have long-term outcomes that are equivalent to front-line combinations. For example, in GOG-132, patients with suboptimal (stage III to IV) disease were randomly assigned to initial therapy with cisplatin (100 mg/m2), paclitaxel (200 mg/m2 throughout 24 hours), or a combination of cisplatin (75 mg/m2) and paclitaxel (135 mg/m2 throughout 24 hours).11 Although not incorporated in the actual trial design, many patients undergoing single-agent therapy crossed over to receive the other agent at the time of disease progression, including approximately 24% of patients that unexpectedly crossed over before reaching a defined progression end point. This reflected widespread use of CA-125, which was not accepted as evidence of progressive or recurrent disease on the trial. Initial response rates to single-agent paclitaxel (43%) were inferior to each of the arms with cisplatin (67%). In addition, patients who received initial paclitaxel and whose disease progressed were still able to achieve a 60% response rate when treated with second-line cisplatin. Not surprisingly, progression-free survival favored the two arms with cisplatin, but overall survival for all three arms was equivalent. This indicates that sequential therapy with two active single-agents in women with advanced disease is acceptable and perhaps as equally effective as using a front-line combination. Further evidence along these lines is provided by the International Collaborative Ovarian Neoplasm (ICON) 3 trial, which randomly assigned more than 2,000 patients between a reference arm (either single-agent carboplatin or a combination of carboplatin, doxorubicin, and cyclophosphamide) and an experimental arm consisting of carboplatin and paclitaxel.12 No significant improvement in progression-free interval or overall survival was apparent with the combination, reinforcing the option of sequential single agents for women with advanced disease. The apparent discordant results between GOG-111 and OV-10 compared with GOG-132 and ICON3 have prompted a lively editorial debate, raising important issues about trial design, primary end points, treatment crossover, and availability of second-line therapies. In addition, because only a small percentage (< 5%) of EOC patients enter phase III trials in the United States each year, there are ample opportunities for selection bias that could influence study outcomes and interpretation. For example, GOG-111 and GOG-132 were both written to enroll identical patients with suboptimal advanced-stage disease (≥ 1 cm of residual disease). However, investigator enthusiasm for paclitaxel was high, and a favorable group of suboptimal patients was rapidly enrolled onto GOG-111. Early experience with the combination of paclitaxel and cisplatin reinforced that enthusiasm, leading some investigators to conclude that single-agent therapy, as studied in GOG-132, was potentially undesirable. This slowed accrual for GOG-132 skewed the population toward patients with more advanced suboptimal disease and encouraged early cross-over before documented disease progression, limiting the power of GOG-132 to discern any differences in treatment outcome among the three arms. Thus, although phase III trials represent our best vehicle for objective evaluation of new treatment strategies, the results of phase III trials will not always alter clinical practice because of feasibility, toxicity, complexity, timing, and bias. In addition, by their design, large phase III trials often lag behind clinical advances and can sometimes be pre-empted by early adoption of new interventions based on extrapolation from smaller studies.
Developmental chemotherapy in the setting of ovarian cancer includes organoplatinum derivatives, taxanes, nontaxane antimicrotubular agents, nucleoside analogues, antifolates, and inhibitors of topoisomerase and other enzymes involved in DNA synthesis and repair, many of which are derived from natural sources. In view of the central role of platinum compounds, there has also been particular interest in nonplatinum agents that may accentuate the platinum response through inhibition of DNA repair and interference with removal of platinum-DNA adducts (Table 1  ). In addition, recent studies have included molecular-targeted reagents directed against growth factors, cellular receptors, angiogenesis, signal transduction pathways, apoptosis, and cell-cycle progression. Although chemotherapy is the focus of this article, selected patients with recurrent disease may benefit from secondary cytoreductive surgery or palliative radiation. Because of the propensity for diffuse peritoneal spread, the extent of recurrent disease is often underestimated by conventional imaging modalities. This has led many oncologists to reserve surgery for symptomatic patients with localized bowel obstruction or other disease-related symptoms. However, some patients may develop a localized pelvic recurrence amenable to cytoreduction, which can be followed by chemotherapy or observation. In addition, it should be noted that radiation therapy can be useful for palliative management of pain or bleeding associated with bulky pelvic disease. In any discussion of treatment for recurrent ovarian cancer, clinicians, patients, and family members can benefit from a shared understanding of basic goals in the setting of a chronic relapsing illness. Although clinically meaningful objective and subjective responses can occur in patients with resistant disease, and prolonged progression-free and overall survival may be observed in patients with platinum-sensitive disease, available data do not support the conclusion that therapy has a realistic potential to be curative in these settings. Thus, evaluation of new treatment interventions in the context of clinical trials is appropriate and of high priority.
Although a major focus of GOG has been the design and execution of definitive phase III trials to evaluate various aspects of front-line therapy, GOG also maintains an active developmental chemotherapy program in collaboration with the National Cancer Institute (NCI) and the pharmaceutical industry. This effort is guided by scientific prioritization through committees to define new regimens for evaluation in future phase III trials. Although there are nearly 25,000 new cases of ovarian cancer diagnosed in the United States each year, few individual centers manage enough patients to conduct their own phase II trials within a reasonable time frame. Regulatory, scientific, and administrative details require thoughtful prioritization of new concepts in the cooperative group setting, and GOG provides an effective multicenter infrastructure to rapidly complete studies once they are activated. In general, new investigational agents will undergo initial phase I evaluation outside GOG but will then receive formal phase II evaluation in patients with recurrent disease. Agents with significant antitumor efficacy will then be selected for phase I studies in previously untreated patients to develop safe and tolerable combinations with platinum or taxane derivatives (Fig 1
After successful phase I development, a regimen can then move directly to a definitive phase III trial. In general, formal phase II evaluation of each new platinum-based combination is not thought to be warranted within GOG because the expected response rate from platinum and cytoreductive surgery exceeds 75% in this population. Although small phase II studies in patients who have newly diagnosed disease may provide valuable information on feasibility and toxicity, the significance of an isolated response rate or progression-free interval is difficult to assess because of the effect of patient selection factors and other sources of potential bias. In the setting of recurrent disease, patients have been divided into prognostic categories based on the platinum-free interval, which predicts for resistance to platinum and other agents. With the incorporation of nonplatinum agents in front-line therapy, eligibility for phase II trials has been modified to reflect the treatment-free interval, defined as the time from completion of initial chemotherapy (including nonplatinum agents) until there is clinical evidence of recurrent disease. Although arbitrary and imperfect, this simple categorization facilitates the design and interpretation of phase II trials with reference to the existing database of prior studies. Ideally, new agents are selected for evaluation in the specific patient population that will yield the most important clinical data based on potential risks and benefits. For example, a new platinum compound that has been engineered to overcome drug resistance would have highest priority for evaluation in a chemotherapy-resistant population, provided the risk of cumulative nonhematologic toxicity was acceptable. Conversely, a novel biologic agent that inhibits signal transduction is probably better evaluated for efficacy and molecular targeting in patients with longer treatment-free intervals and potentially chemoresponsive disease. Several chemotherapeutic agents exhibit schedule-dependent host toxicities, whereas the effect of schedule on tumor response is less clear. Sometimes these relationships are defined through preclinical models or phase I studies, but in most cases, information on optimal dose and schedule is not available when planning phase II and III clinical trials. In addition, traditional phase I paradigms place undue emphasis on defining dose-limiting toxicity, which usually consists of neutropenia, rather than optimization of tumor biologic targeting or efficacy. As one might expect, the importance of scheduling dependencies or other treatment variables is accentuated during the development of combination regimens, particularly if there is any biologic interaction among the individual components. For example, combinations of platinum with either topotecan or gemcitabine have each demonstrated sequence-dependent effects on hematologic toxicity, requiring a substantial reduction in initial drug dose levels and raising questions about adequate tumor exposure. Conversely, paclitaxel can be combined with carboplatin at full doses with an apparent reduction in hematologic toxicity, prompting some investigators to describe a platelet-sparing effect13,14 and raising questions about the potential for negative tumor interactions.15 In view of the toxicity encountered with some drug combinations, investigators have evaluated alternative strategies for drug delivery, including weekly treatment with lower individual doses, abbreviated treatment duration, cycle-by-cycle alternations, and use of sequential doublets composed of different agents rather than a conventional fixed triplet regimen. The optimum approach has not been defined, and each strategy has potential limitations. Better preclinical models that are predictive of human drug interactions would be helpful to minimize host toxicity while maximizing antitumor effects. Finally, our clinical trial paradigms are being challenged by a fast-growing panel of biologic and molecular-targeted reagents. In addition, it is widely anticipated that molecular profiling of tumor genotype, DNA methylation status, mRNA expression, patterns of protein synthesis, and pharmacogenomic pathways of drug metabolism will better predict which tumors might respond to which reagents in which patients. These advances in drug development and tumor biology have the potential to overwhelm our traditional approach to clinical trials and will require increased attention as we move forward.
Compared with the large number of single-arm phase II trials, as well as front-line randomized phase III trials, there are only a small number of randomized phase III trials that have been completed in the setting of recurrent disease. In part, this reflects the need to efficiently screen individual agents for activity, followed by efforts to quickly integrate active agents with front-line therapy. Although there are certainly valid scientific and clinical questions that would merit the expense and complexity of a large phase III trial, it has been difficult to support these efforts in the absence of pharmaceutical sponsorship, which has usually been focused on more restrained designs geared toward specific Food and Drug Administration (FDA) drug registrations. Similarly, only a small number of trials have addressed important issues, such as consolidation and maintenance after first remission or the management of small-volume residual disease after primary therapy, which remain the subject of considerable bias in the absence of definitive data. From a logistics perspective, several of these ambitious studies have been difficult to complete because of evolving standards of care, limited insurance reimbursement, and stringent criteria for interim analysis. For example, based on promising phase II data with intraperitoneal interferon alfa in patients with small-volume residual disease, the Southwest Oncology Group (SWOG) initiated a phase III trial of intraperitoneal interferon alfa versus observation in patients with a laparotomy-documented complete remission after initial therapy. Because of declining rates of second-look surgery, accrual was slower than expected, even when joined by GOG, and the trial was eventually closed as nonfeasible. In another example, an attempt was made to evaluate the role of high-dose consolidation with hematopoietic progenitor-cell support in a prospective phase III randomized trial organized by GOG and the NCI. However, after 2 years, the study was terminated because of poor accrual, reflecting physician bias and limitations of insurance reimbursement for high-dose therapy in patients with solid tumors. In a third example, SWOG and GOG jointly evaluated the role of maintenance therapy with paclitaxel for 3 months versus 12 months for patients in clinical complete remission after initial therapy. This trial accrued well, but it was terminated with 50% accrual at the time of a planned interim analysis based on a significant improvement in median progression-free survival for the patients who received 12 months of paclitaxel but without any demonstrated benefit in overall survival or quality of life.16 One could argue that a short-term benefit in progression-free survival was not unexpected, given the inequality in treatment duration between the two arms. It is unfortunate that the trial was closed before full accrual, which limits the opportunity to evaluate overall survival because of study termination and treatment crossover. Among the more innovative ongoing phase III trials is an attempt to address the timing of second-line therapy based on clinical evidence of disease progression versus early intervention on the basis of an elevated CA-125 level, sponsored by the Medical Research Council (MRC) in the United Kingdom (OV-05). The MRC has also completed a trial of single-agent carboplatin versus a combination of carboplatin and paclitaxel for patients with platinum-sensitive recurrent disease (ICON4), with preliminary results expected in 2003. Results are also expected from a similar trial sponsored by the Arbeitsgemeinschaft Gynakologische Onkologie (AGO) in Germany, which compares single-agent carboplatin with a combination of carboplatin and gemcitabine. The AGO has also completed accrual for a study of consolidation with single-agent topotecan compared with observation without further therapy after completion of primary chemotherapy with platinum and paclitaxel. Within GOG, a decision was made to initiate an ambitious phase III trial with dual randomization to evaluate secondary cytoreductive surgery versus no further surgery and carboplatin versus topotecan with crossover at time of disease progression (GOG-202). Each of these trials addresses important clinical questions, and it is hoped that they will be successful in reaching their accrual goals, which will expand the opportunity for other phase III initiatives after primary surgery and chemotherapy.
As already discussed, we currently have a diverse array of cytotoxic agents directed against a variety of cellular targets with unique mechanisms of action, nonoverlapping pathways of resistance, and distinct toxicity profiles. In addition, we are beginning to see clinical activity emerge from a fast-growing panel of biologic and molecular-targeted reagents. Finally, it is widely anticipated that molecular profiling of tumor genotype, DNA methylation status, mRNA expression, patterns of protein synthesis, and pharmacogenetic pathways of drug metabolism will better predict which tumors might respond to which reagents in which patients. These advances in drug development and tumor biology have overwhelmed our traditional approach to clinical trials. The potential permutations and combinations that merit evaluation are daunting when viewed within the traditional phase I/II framework, but offer exciting opportunities for innovative design of trials that incorporate biologic and clinical end points. Ideally, phase I trials should determine optimal drug dosing in relationship to tumor targeting and host metabolism rather than single-cycle dose-limiting toxicity. Randomized phase II trials could then further define the safe and effective dose and answer questions related to schedule, sequence, biologic targeting, and new combinations with a biologic basis for synergy. Traditional (nonrandomized) phase II trials are designed and powered based on a surrogate end point that is thought to reflect clinical benefit, such as response rate, disease stabilization rate, biologic (tumor marker) response rate, or progression-free interval. Although each of these end points may reflect antitumor activity, they are viewed as surrogate end points in comparison with overall survival, which provides a true measure of clinical benefit. In general, each phase II study tests a single hypothesis, such as response rate, to determine whether the new agent or regimen crosses a threshold based on historical data in the same patient population with an appropriate degree of power (type 2 error) and precision (type 1 error). Through this process, new agents are selected for further development based on a comparison against historical controls. Randomized phase II trials allocate patients among two or more treatment arms to minimize potential differences in prognostic factors or other variables. Each arm is then independently tested against the same historical threshold value. Using this approach, one or both arms can be selected for further clinical development without a direct comparison between the treatment arms. This strategy has been successfully applied by the National Cancer Institute of Canada (NCIC) to evaluate different doses and schedules of topotecan17 and liposomal lurtotecan.18 However, there is some concern that properly designed noncomparative phase II trials may be inappropriately interpreted as comparative studies after presentation. Obviously, a clinical trial’s organization cannot mandate how trials will be interpreted in the public domain. However, each group can articulate a clear and consistent design within the protocol document and associated publications and can educate the oncology community regarding the intended role of these trials. A second area of controversy concerns comparative randomized phase II trials. Although underpowered to compare overall survival, a phase II trial could be adequately powered to compare surrogate end points, such as response rate, progression-free interval, toxicity, or biologic targeting. Obviously, because of the smaller number of patients, these comparisons would be subject to a wider margin of error and reduced precision, but they could still provide a valid comparison that would guide choice of regimens for fully powered phase III trials. This approach works best with two experimental arms, which reduces the tendency to draw inferences between standard and experimental regimens. A special case occurs when a comparative randomized phase II trial includes a control arm, such as cisplatin, versus cisplatin combined with a targeted biologic agent, such as an inhibitor of the epidermal growth factor receptor. Whether the addition of a growth factor receptor inhibitor on the experimental arm can improve clinical outcomes compared with chemotherapy alone is an important contemporary question. In addition, it would be difficult to find adequate clinical and financial resources to test all known inhibitors with all combinations of chemotherapy in a set of fully powered multiarmed phase III trials. Thus, there is a need for efficient screening of new combinations to select which combinations show promise and should undergo further testing. In this context, the use of an internal control arm seems justified because of changes in treatment practice and supportive care over time, which may influence the expected response rate, toxicity profile, or progression-free interval.
The GOG has used the phase II trial as a mechanism to screen new single agents or specific combinations for activity in women with recurrent disease. In general, these trials have used a two-stage accrual design that affords early stopping in the event of inactivity. The primary study hypothesis is usually based on a response rate of interest to demonstrate activity and a response rate of no further interest consistent with inactivity based on historical thresholds from prior studies. The sample size for each stage is determined by the desired precision (type I error, alpha) and power (type II error, beta) of the analysis. For example, in patients with platinum-resistant disease, the response rate of interest for further development would typically be 25%, with a response rate of less than 10% being of no further interest. When combined with a desired precision of 0.1 and a power of 0.9, this yields a sample size of approximately 25 patients in the first stage and 20 to 25 additional patients in the second stage, which would be activated only if more than two responses are observed during the first stage. With response rate (complete and partial) as the primary end point, patients are required to have measurable disease according to Response Evaluation Criteria in Solid Tumors, as adopted by NCI and GOG. The occurrence of disease stabilization is reported as a favorable clinical outcome but is of uncertain significance in a small phase II study with variability in patient selection and natural history, which may include indolent tumors. It has been argued that tumor response rate may not be the best primary end point for evaluation of newer targeted biologic or antiangiogenic compounds because these compounds may exhibit cytostatic growth-arresting behavior rather than direct tumor cytotoxicity. The importance of this assertion is unclear. However, it has resulted in alternative phase II designs that emphasize the progression-free interval and disease stabilization rate rather than response rate. Historical baselines for these end points are less robust and were largely obtained in patients with bidimensional measurable disease, which likely has a different response rate and time to progression compared with small-volume disease not measurable by radiographic criteria or minimal disease that is only detectable on the basis of an elevated CA-125 level or other serum tumor marker. Nonetheless, as more of these studies are conducted, it will be possible to refine our expectations and apply these end points in the target patient population.
As already mentioned, one of the difficulties with interpreting the reported activity of individual agents after small phase II studies has been variability in the treated population. It is now well recognized that there is considerable heterogeneity among patients who receive second-line treatment, and this can lead to wide variations in the anticipated objective response rates to second-line platinum-based therapy, ranging from less than 10% to greater than 40%.19 Characterization of the treated population makes it easier for results to be appropriately interpreted with reference to historical expectations. Platinum-sensitive ovarian cancer is defined as the recurrence of active disease in a patient who has achieved a documented response to initial platinum-based treatment and has been off therapy for an extended period. For the purposes of study design and interpretation, the GOG has arbitrarily divided sensitive disease from resistant disease at 6 months.20 Platinum-resistant ovarian cancer is defined as disease that has responded to initial chemotherapy but demonstrates recurrence within a relatively short period after the completion of treatment. Within the GOG, patients with documented recurrence within 6 months of completing initial therapy should be considered platinum resistant. Persistent ovarian cancer is the finding of residual disease in a patient who has completed, and perhaps partially responded to, initial chemotherapy. For example, this would include a microscopic-positive second-look laparotomy in a patient who began chemotherapy with suboptimal or optimal residual disease after initial cytoreductive surgery. However, more than half of patients with advanced ovarian cancer begin chemotherapy in clinical complete remission after initial cytoreductive surgery. In this setting, persistence of small-volume disease may actually represent resistant, rather than responsive, tumor deposits, which further contributes to heterogeneity in this population. Refractory ovarian cancer occurs in a patient in whom therapy has failed. This includes patients with either stable disease or actual disease progression during primary therapy, which occurs in approximately 20% of patients. As might be expected, this group has the lowest response rate to platinum and other second-line interventions. In many studies of second-line treatment, patients with resistant, persistent, and refractory disease are considered one group, whereas individuals with platinum-sensitive recurrent disease are considered separately. Even within these more resistant populations, there is still considerable heterogeneity with regard to overall drug sensitivity. Although the difference between drug-sensitive and drug-resistant disease was initially described in relationship to platinum-based therapy, these observations can be applied in a general sense to other chemotherapy regimens. However, the patterns of resistance and mechanisms of action for many drugs, such as paclitaxel, are distinct from platinum, and resistance to a combined regimen does not always imply resistance to individual components. In addition, these definitions are largely based on the response to front-line therapy. With the diverse agents that are currently available to manage recurrent disease, the relevance of response to initial platinum-based therapy has become less clear. Platinum, however, remains the most important drug used in the therapy of advanced ovarian cancer, and selected patients with platinum-sensitive disease may benefit from re-treatment with either cisplatin or carboplatin. Eventually, most patients with recurrent ovarian cancer manifest platinum-resistant disease, in which case the potential risks of cumulative toxic effects outweigh any potential benefits. In some cases, prolongation of the platinum-free interval can improve the likelihood of response to re-treatment with platinum because of partial reversal of acquired drug resistance. This phenomenon has been documented in a small number of patients with cisplatin-refractory disease who have responded to re-treatment with carboplatin after extension of the platinum-free interval using non–platinum-based chemotherapy.21 Fortunately, a number of agents have recently been described with moderate activity against platinum-resistant disease, and these agents may contribute to the extension of the platinum-free interval, opening an opportunity for eventual rechallenge with platinum.
Platinum-Sensitive Disease There is no absolute treatment-free interval that completely separates individuals with regard to potential sensitivity to platinum or paclitaxel after response to initial therapy.22,23 In general, the longer the platinum-free interval, the greater the expectation of durable response to re-treatment. Patients who remain disease free for more than 2 years after primary therapy have the greatest likelihood of benefit and are usually re-treated with a combination of platinum and paclitaxel,24,25 as in de novo disease. Although patients with a treatment-free interval of less than 12 months are frequently considered candidates for re-treatment with platinum and/or paclitaxel, there is currently no evidence from prospective randomized trials that combination chemotherapy achieves superior outcomes with regard to survival or quality of life compared with the use of sequential single agents in this population. In addition, early re-treatment with platinum places the patient at risk for cumulative hematologic (carboplatin) and nonhematologic (cisplatin) toxic effects that can limit further therapy and diminish the overall quality of life. Ideally, such patients should consider participation in phase II clinical trials to evaluate new treatment strategies because they are better able to tolerate multiple cycles of therapy and more likely to respond to a given treatment regimen compared with patients with more extensive prior therapy.
Persistent Disease Several treatment approaches can be considered for the chemoresponsive population. First, patients may continue to receive the drugs to which their disease has responded, with or without interval cytoreductive surgery. The optimal number of chemotherapy cycles to be administered in this setting is unknown, and it remains to be established whether continued treatment in a responding patient will favorably affect long-term clinical outcomes. Alternatively, patients may be treated with one of a number of agents demonstrated to have activity in the second-line setting. For patients who have achieved a major response to initial treatment but who have persistence of small-volume residual disease, it is not unreasonable to consider a dose-intensive approach to maximize response and extend the progression-free interval. Strategies that have been used in this setting include intraperitoneal therapy26 and high-dose intravenous chemotherapy with hematopoietic progenitor cell support.27–29 Again, it must be emphasized that data do not currently exist from randomized controlled trials to demonstrate that any therapeutic approach, including a variety of dose-intensity strategies, has curative potential in the second-line treatment of advanced ovarian cancer. As a result, the potential significant toxic effects of these regimens must be kept in mind when considering possible strategies in this clinical setting.
Platinum-Resistant and Refractory Disease Within this resistant population with measurable disease, objective response rates (complete and partial) have ranged from 5% to 30%, with few agents exceeding 15%. Although the duration of these responses has generally been short (< 8 months), occasional patients may continue to respond or maintain stable disease for a more extended interval. In this palliative setting, the development of stable disease is also considered to be a positive but limited outcome, which is achieved in an additional 25% to 40% of patients. In the absence of dose-limiting toxic effects or clinical evidence of progressive disease, it is reasonable to continue therapy in such patients, depending on their overall quality of life and performance status. In view of the number and diversity of active agents currently available for second-line treatment of ovarian cancer, clinicians and patients often consider treatment beyond the second-line setting, which requires adequate vital organ function, hematologic reserve, and performance status. Unfortunately, there are limited data to predict activity of individual agents against tumors that are refractory to multiple chemotherapy regimens. In these circumstances, it is reasonable to choose an agent with a different mechanism of action and limited potential for serious toxic effects. Response should be evaluated after two or three cycles. Once again, the attainment of stable disease with acceptable levels of toxic effects is a valid clinical end point and would justify continuation of therapy. Ultimately, it is anticipated that tumor molecular profiling, metabolic imaging, or enhanced drug sensitivity testing may help match patients with particular treatment regimens not only in the front-line setting but also during the management of recurrent disease.
Overview of the GOG Development Within the GOG, new agents and combinations are evaluated in the phase I and phase II settings. Generally, phase I studies are limited to the evaluation of new platinum-based combinations in patients with previously untreated disease. These studies use standard dose-escalating models and fixed-dose expanded cohorts to verify feasibility and small pilot studies to examine issues such as high-dose therapy with stem-cell support. Promising combinations that seem to be feasible can advance directly from phase I to phase III evaluation without a formal phase II study because of the known activity of platinum-based therapy in this setting.
In contrast, phase II trials have generally focused on the evaluation of new investigational agents in patients with recurrent disease under sponsorship from the Cancer Therapy Evaluation Program of the NCI or the pharmaceutical industry. These studies are sequentially activated in separate queues for patients with platinum-sensitive (treatment-free interval of greater than 6 months) and platinum-resistant (treatment-free interval of < 6 months) disease. There is also a concurrent queue to evaluate biologic agents, including monoclonal antibodies, antiangiogenesis agents, and other novel compounds (Table 2
Paclitaxel Paclitaxel has been established as an important initial component of ovarian cancer chemotherapy and should also be considered in the management of patients with recurrence. Of interest, the mechanisms of acquired drug resistance are different between paclitaxel and platinum, and not all patients with platinum-resistant disease are resistant to paclitaxel, even if paclitaxel was included in their front-line treatment program.
As a component of initial platinum-based chemotherapy, paclitaxel is most commonly administered as a 3-hour intravenous infusion at 175 mg/m2. In patients with recurrent disease, phase III randomized trials have evaluated dose-intensity (135 v 175 mg/m2 and 175 v 250 mg/m2) and infusion duration (3 v 24 hours) without a clear advantage to either higher doses or prolonged infusion.30,31 In nonrandomized studies, a prolonged 96-hour infusion at a total dose of 120 to 160 mg/m2 was reported to have good activity in breast cancer but was evaluated in recurrent ovarian cancer after multiple prior therapies with unimpressive results.32 Overall, it seems that variations in paclitaxel dose and schedule have little bearing on tumor response but do have a clear effect on the spectrum and severity of hematologic, mucosal, and neurologic toxic effects (Table 3
However, the demonstration of a further reduction in toxic effects and apparent maintenance of efficacy among women with paclitaxel-resistant disease using a 1-hour weekly infusion at 60 to 90 mg/m2 has raised considerable interest, even in the absence of randomized data.33 This particular schedule is associated with minimal noncumulative hematologic and nonhematologic toxic effects, including a reduction in alopecia. Peripheral neurotoxicity is not uncommon, particularly with higher cumulative doses in patients with pre-existing neuropathy, but this can usually be managed with treatment breaks or dose reduction. A number of studies are also being initiated with alternative formulations of paclitaxel as liposomes, polymer-based microspheres, polyglutamated conjugates, and albumin conjugates. In general, these formulations reduce or eliminate the need for Cremephor EL (a solubilizing agent containing ethyl alcohol and polyoxethylated castor oil), which is responsible for most hypersensitivity reactions. In addition, alternative formulations may achieve more prolonged serum or peritoneal concentrations after intermittent dosing. Thus far, although some of these agents may improve the toxicity profile associated with paclitaxel, none are thought to have clinical activity that is superior to native paclitaxel. In addition, attention has focused on entirely different compounds, such as the epothilones, which have the same molecular target but a different chemical source and structure.
Docetaxel
In front-line combination therapy, data have been reported on clinical response rate, progression-free survival, and toxicity from a Scottish Randomized Ovarian Cancer of the Scottish Gynaecological Cancer Trials Group phase III trial, which compared carboplatin with docetaxel at 75 mg/m2 (1-hour infusion) versus carboplatin and paclitaxel at 175 mg/m2 (3-hour infusion).39 There was a clear difference in the pattern of toxic effects, with a reduction in neuropathy and an increase in myelosuppression among women who received docetaxel. However, there has been no apparent difference in clinical efficacy on the basis of response rate, CA-125 reduction, or progression-free survival. Thus, the potential preclinical advantages of docetaxel have not yet translated into improved clinical efficacy.
Topoisomerase-I Inhibitors On the basis of interpatient variability in drug clearance, frequency of dose-limiting hematologic toxic effects, and the lack of evidence for a tight dose-response relationship, it has become common practice to initiate therapy at 1.25 mg/m2 daily for 5 consecutive days. Choice of the starting dose is guided by toxic effects, extent of prior therapy, and estimated renal clearance. Fortunately, because hematologic toxicity is noncumulative and generally emerges during the first cycle, it can be readily managed with dose reduction or introduction of hematopoietic growth factors, which then safely permits continuation of therapy over multiple cycles.
In an attempt to improve the tolerability of single-agent topotecan and to facilitate the development of combination regimens, the GOG has conducted phase II studies of single-agent topotecan with alternative schedules of drug administration (Table 5
Preclinical studies with established cell lines have evaluated the optimal schedule and sequence of drug exposure using various in vitro assays. In almost all cases, the combination of topotecan and cisplatin has exhibited synergistic toxicity. However, only a few cell lines have shown schedule dependence, generally favoring the administration of cisplatin before topotecan.50 Other systems have favored synchronous drug administration51 or shown no difference. Phase I trials have examined drug sequencing of doublets including paclitaxel-topotecan52 and topotecan-cisplatin, confirming that neutropenia and thrombocytopenia are the major dose-limiting toxic effects. When the sequence of topotecan and cisplatin was alternately evaluated in the same patient, it was clear that administration of cisplatin on day 1 was more toxic than administration on day 5.53 The mechanism for this effect was not entirely explained, but it was attributed to changes in topotecan clearance as a consequence of subclinical renal dysfunction after cisplatin administration, a finding that has not been confirmed. In a phase I trial using oral topotecan (5 days) in combination with cisplatin (75 mg/m2 on day 1 or 5), hematologic toxic effects increased when cisplatin was administered on day 1, resulting in a recommended topotecan dose of 1.25 mg/m2 daily compared with 2 mg/m2 daily by the reverse sequence.54 Using intravenous topotecan in patients with untreated lung cancer, it was recommended to use a reverse sequence with topotecan 1.50 mg/m2 daily for 5 days followed by cisplatin 50 mg/m2 on day 5.55 As one might expect, attempts to develop a tolerable triplet combination have been more problematic. An ongoing phase I study (GOG-9602) is evaluating the combination of cisplatin, paclitaxel, and topotecan in previously untreated patients.56 Starting doses were 50 mg/m2 of cisplatin and 175 mg/m2 of paclitaxel using a 3-hour infusion. The dose of topotecan could not be escalated above 0.50 mg/m2 daily, which is one third of the usual single-agent dose, and the study was revised to evaluate a shortened 3-day regimen of topotecan. Other studies have confirmed the increased hematologic toxicity of this triplet regimen. For example, Herben et al57 recommended a phase II regimen with topotecan at 0.30 mg/m2 daily (days 2 to 6), paclitaxel at 110 mg/m2 (day 1), and cisplatin at 75 mg/m2 (day 2) followed by mandatory granulocyte colony-stimulating factor. Similar experiences have been encountered with carboplatin-based triplets, prompting some investigators to suggest that such combinations are impractical without hematopoietic stem-cell support.58 However, despite decreased dose-intensity, some carboplatin-based triplets have shown promising activity in advanced ovarian cancer, as illustrated by a phase II trial of topotecan, 1 mg/m2 daily (days 1 to 3), followed by carboplatin, area under the curve (AUC) 5, and paclitaxel, 175 mg/m2 on day 3.59 As an alternative, a weekly combination of paclitaxel (85 mg/m2), cisplatin (40 mg/m2), and topotecan (0.75 to 2.50 mg/m2) was tested in patients with ovarian and lung cancer, achieving a favorable overall response rate with acceptable toxic effects at a recommended phase II topotecan dose of 2.25 mg/m2 weekly.60 Although phase I trials of weekly single-agent topotecan were inactive, a dose-escalating, single-agent, phase II study in patients with recurrent ovarian cancer documented limited activity,61 which would support evaluation of a well-tolerated weekly combination regimen that might maximize synergistic interactions. As already indicated, full (or increased) doses of this triplet combination would require hematopoietic support. One promising high-dose combination can be safely administered over multiple cycles after a single harvest of peripheral-blood progenitor cells62 and is being formally evaluated in a pilot study with combined end points of feasibility and efficacy (GOG-9903). In view of the difficulty of developing a well-tolerated platinum-based triplet combination, efforts have focused on the use of doublets, individually or in sequence, to maximize effective drug exposure. For example, a phase II trial of the NCIC found the combination of cisplatin (50 mg/m2, day 1) and topotecan (0.75 mg/m2 daily, days 1 to 5) to be feasible when administered for four cycles followed by four cycles of paclitaxel-cisplatin.63 This is the basis of an ongoing NCIC and European Organization for Research and Treatment of Cancer phase III randomized trial (OV-16) comparing cisplatin and topotecan for four cycles followed by four cycles of carboplatin and paclitaxel versus eight cycles of carboplatin and paclitaxel.
In a phase I trial reported by Gordon et al,64 a doublet of topotecan and carboplatin was alternated cycle by cycle with paclitaxel and carboplatin in patients with newly diagnosed EOC. Carboplatin (AUC 4 or 5) was given before a 3-day regimen of topotecan in view of the reported synergy of this sequence. At a topotecan dose level of 1 mg/m2 daily, two of six patients exhibited dose-limiting toxic effects. Similar to the experience with cisplatin, postponing the administration of carboplatin was associated with more effective drug delivery at higher doses with decreased bone marrow toxic effects,65 but the preferred schedule for clinical development has not been established. Although it is clear that sequence has a substantial effect on hematologic toxic effects, the potential effect of sequence on tumor response is unknown. The GOG evaluated both sequences of topotecan in combination with carboplatin to select the optimal doublet regimen for phase III studies (Fig 2 |
TOP
ABSTRACT
INTRODUCTION