Development of New Drugs: Lessons from Clinical Trials |
An ESUN Article
Editor’s Note: Dr. Meyers gave the Nina Axelrad Lecture at the CTOS 14th Annual Meeting in London, United Kingdom in November 2008. The article which follows is a transcription of that talk. Dr. Meyers took an opportunity to modify some of the material after the presentation. We are very grateful to Dr. Meyers for sharing this talk with the ESUN community.
The addition of chemotherapy to the treatment of "pediatric sarcomas" transformed lethal diseases into treatable diseases. We have not seen similar benefits for the addition of chemotherapy to surgery and radiation therapy for the treatment of most soft tissue sarcomas seen in adults. In addition, we have seen a plateau in our ability to improve outcome for the "pediatric" sarcomas, with little or no improvement in outcome in recent years.
All of us are excited by the possibility of biological or targeted therapies. We face an enormous challenge designing appropriate clinical trials to evaluate new agents which have biological targets. When biological agents demonstrate promise in pre-clinical or early phase clinical testing, we will need to consider how best to introduce these agents in the treatment of sarcomas. For diseases in which chemotherapy has an unequivocal benefit we will face the additional challenge of learning how to combine biological agents most effectively and safely with chemotherapy.
I would like to present my thoughts about some of the issues that arise as we consider how to move biological agents forward into our armamentarium of treatment for sarcomas.
Goals of early studies are to obtain pharmacology information about the agent to determine dose ranges appropriate for future studies and to establish toxicity profiles, with starting doses usually based on preclinical toxicology studies.
However, few anti cancer agents are tested in true phase 1 studies in healthy adults. Rather, most anticancer agents are tested in early studies in patients with advanced cancers where, unlike studies in healthy subjects, the intent is to avoid pharmacologically inactive doses. It is worth noting that MTP is so safe that we have been able to administer it to normal volunteers.
Most early studies with biologics as with other anticancer agents have been conducted in patients with late stage cancer.
Do we expect the agent to have an impact on bulky tumors? The example of Imatinib in GIST suggests that some biological agents will be useful in the setting of macroscopic disease. Other agents seem much less likely to work against bulk disease and will have a greater chance to demonstrate efficacy in the adjuvant setting. To some extent we will make this prediction based on the proposed mechanism of action of the biological agent. This information is important in order to set realistic expectations of agents in these early studies.
Especially in the absence of treatment effects, early goals focus on identification of a pharmacologically effective dose or optimal biologic dose rather than maximally tolerated dose (MTD) and biologically active doses may occur well below the MTD. Pharmacologic or biologic activity may be assessed by a variety of means including laboratory and clinical changes that are indicative of in vivo activity.
For MTP evidence of biologic activity in early studies included induction of tumoricidal macrophages, increases in serum cytokines and -the so-called ‘cytokine flu’, including fever, chills and headache. There was anecdotal evidence of antitumor activity in these early studies of late stage patients but no patterns emerged. The MTD was established with the usual criteria of toxicity. We chose a dosing schedule below the MTD because it was associated with evidence of biological activity with a very low rate of grade 3 and 4 toxicities.
There are a number of new monoclonal antibodies against the IGF1 receptor which inhibit the IGF pathway.
In the absence of dose limiting toxicities, potential biomarkers for IGF receptor inhibitors include down regulation of IGF 1R, upregulation of the ligands IGF1 and IGF2, reduction in circulating tumor cells, changes in the phosphorylation status of the receptor, receptor occupancy or other mechanism-related toxicities such as hyperglycemia.
We need to think hard about how we plan to incorporate these agents into clinical trials for sarcomas. We know that we have seen objective responses in a sub set of patients with recurrent sarcomas, including some patients with Ewing sarcoma and osteosarcoma.
Phase 2 studies should be designed to provide the information and confidence that we need to take an agent into large randomized phase 3 studies. Thus from Phase 2 studies we need to obtain sufficient efficacy information to have confidence of the drug’s potential and to estimate the treatment effect we would plan into the phase 3 design. Phase 2 should also further extend the safety profile of the agent to help to determine what might be key safety issues in larger studies.
Phase 2 studies may be single arm or randomized, though randomized phase 3 studies typically provide a better estimate of potential treatment effects.
As we consider our design of phase II trials of biological agents, we must consider the proposed mechanism of action and the likelihood of an objective response. For some biological agents it is more likely that we will expect to see benefit when they are employed against microscopic sub-clinical disease than when they are applied to bulky tumors. For these agents, it makes more sense to achieve a state of minimal disease by surgical resection of bulky disease and then administer a new agent in the adjuvant setting. This is a different paradigm from the conventional phase II design which seeks to demonstrate an objective response in measurable tumor lesions. This requires the use of different end points to assay the effect of anti-tumor agents. This will require the use of progression free survival or overall survival, not objective response to demonstrate the activity of an agent. The expectation, especially for agents that work by activation of the host immune system or host cytokines, is that they will be more likely to demonstrate clinical activity in minimal residual disease, especially since the immune system is often compromised in patients with late stage or bulky disease. This has been demonstrated in many animal models for such agents.
We will also need to consider whether we believe that an agent will exert the greatest effect as a single agent or when given together with other tumor agents, such as chemotherapy. If a novel agent is likely to exhibit synergy with chemotherapy, the best way to detect that benefit is in a phase II trial of the new agent in combination with chemotherapy. We will need to be guided by pre-clinical studies to make these decisions.
Finally, we should consider the natural history of the tumor we hope to treat. For tumors which progress or recur relatively quickly, delaying introduction of a new agent until after completion of conventional therapy will almost certainly be too late to have a major impact on outcome.
Based on the results of MTP in dogs, we thought the best chance to demonstrate anti-tumor activity for this new agent was in osteosarcoma. As we considered the development of MTP, we reasoned that it was unlikely to show marked benefit in the setting of bulky disease. This is a function both of the nature of osteosarcoma, which secretes osteoid and ossifies that matrix, and of data from the preclinical studies of MTP.
The initial phase 2 study of MTP in osteosarcoma performed at the MD Anderson included patients with relapsed pulmonary disease who were rendered disease free by surgery then received single agent therapy. This allowed the agent to be tested in a setting of minimal residual disease but without denying any patient standard of care, since complete surgical excision is the only proven post-relapse treatment with important influence on patient outcome. This study demonstrated benefit of MTP compared to a matched group of historic controls, provided information on the importance of duration of treatment, and evidence of an in vivo effect of MTP in the tumor.
Evidence of the activity of biologic agents that can be obtained from direct observation of tumor tissue is an important piece of information that should not be overlooked while performing early phase studies. In the case of agents like MTP, which depend upon activation of the host immune system or cytokines for anti-tumor activity, evidence of such activation in the tumor can provide strong support for the activity of the agent.
Chemotherapy is an essential part of the treatment strategy for newly diagnosed patients with osteosarcoma. As we considered how to introduce MTP into phase III trials for the treatment of osteosarcoma, we had to decide whether the drug should be given with chemotherapy.
Early preclinical work by Drs Fidler and Kleinerman had suggested that MTP worked through activation of macrophages to become tumoricidal in the autologous setting. If MTP works through effector cells, administration with chemotherapy might be ineffective, since chemotherapy will decrease the numbers of effector cells. While the number of circulating macrophages is certainly decreased by chemotherapy, we had information that tissue fixed macrophages, especially in the lung, survived systemic chemotherapy and even survived cytoreduction and allogeneic BMT. Since the lung is the primary site of metastasis for 85% of patients following treatment for osteosarcoma, we felt that there would be adequate effector cells to respond to MTP. In this case, MTP should be introduced with chemotherapy. More recently Dr Kleinerman has demonstrated that activation of the fas/fas ligand apoptotic pathway may one mechanism by which MTP is effective. If MTP works by activating an apoptotic pathway it might be synergistic with chemotherapy. In this case, too, MTP should be introduced with chemotherapy.
In preparation for a study using the combination, we needed data about the feasibility and safety of combining MTP with the chemotherapy agents commonly used to treat osteosarcoma. We knew from phase I studies that MTP had a very benign toxicity profile. In 1991, we designed a phase II randomized adjuvant therapy trial for osteosarcoma using ifosfamide with and without MTP following resection of recurrent pulmonary nodules. The outcome variable was progression free survival. The sponsor felt that the FDA would not recognized PFS as a valid end-point and declined our study design. Dr Kleinerman and I performed a small trial of ifosfamide and MTP in recurrent osteosarcoma and demonstrated that the safety profile of both agents remained unchanged, and cytokine activation associated with MTP was unchanged when it was administered in combination with chemotherapy. This trial was designed to detect negative impacts of each on the other. One lesson that I take from this experience is that it would have been better to carry out the randomized adjuvant phase II study.
Pre-clinical studies of IGF1R inhibitors have suggested synergy between chemotherapy and IGF pathway inhibition. As we design our phase II studies for these monoclonal antibodies, the same considerations arise. We have observed an extremely favorable toxicity profile for this class of agents. Early data suggest that combination of anti-IGF1R monoclonal antibodies with chemotherapy does not worsen toxicity.
While further single agent studies with anti-IG 1R monoclonal antibodies may not be necessary to gather further safety data, we will need to design phase II trials of these monoclonal antibodies in combination with the agents we commonly use to treat sarcomas. My experience suggests that these would be best designed as randomized studies. End-points for these trials should include objective response, progression free survival and especially in randomized trials, overall survival. We must be vigilant about regulatory considerations as we design these studies. We have been told on both sides of the Atlantic that the regulators will regard PFS analyses as surrogates that will require confirmatory trials. If we are planning phase 3 studies, surrogate endpoints may be acceptable for phase 2 studies, but should the data from randomized Phase 2 be sufficiently convincing, we would certainly want to be able to use it to support approval.
Therefore, we must design every phase II trial to capture overall survival as an end-point in addition to the more conventional endpoints of objective response and progression free survival. This is a major shift in the way we have approached phase II clinical trials, and will require significantly longer follow-up after study enrollment than the much easier endpoints, but this will be essential if phase 2 data will contribute to the regulatory approval that will make these agents available to our patients.
The goal of phase 3 studies is to confirm the efficacy seen in phase 2 studies and draw conclusions about the benefit – risk ratio of a drug for the treatment of a specific disease in a specific population.
As mentioned above, design considerations may be unique for biologics and biological response modifiers (BRMs) in terms of when we give them, what we give them with and how we evaluate them best. The following slides illustrate the factors we considered in designing the phase 3 study of MTP and factors we should consider in designing such studies of IGFR inhibitors.
We saw in our small phase 2 study that MTP could be given with chemotherapy without negative impact on the immune activation by MTP nor on the chemotherapy effects on the tumor.
All preclinical and clinical data had suggested that MTP would work best in the setting of minimal residual disease. For this reason we decided that it would be optimal to delay introduction of MTP until after surgical resection of clinically detectable macroscopic disease.
In osteosarcoma we detect progression or recurrence when macroscopic metastatic nodules, most commonly in the lung, are detected by imaging studies. The appearance of a metastatic nodule in the lung reflects development of a recurrent clone of tumor cells months earlier. The pattern of failure of osteosarcoma suggested that if we waited until completion of all planned chemotherapy to introduce a new agent we would have "lost" one-third to one-half of the patients destined to relapse before we could even introduce the new agent,.
Thus after consideration of the results of early phase clinical trials, consideration of the possible mechanism of action of MTP, and consideration of the pattern of failure of osteosarcoma with chemotherapy alone, we decided to introduce the biological agent immediately after surgery at the same time that chemotherapy was resumed.
The rationale for introducing MTP as early as possible after surgery is illustrated by this Kaplan-Meier plot of the survival and event free survival for patients with localized osteosarcoma treated on INT 0133. Note that the development of macroscopic tumor recurrence, which reflects an event months earlier, occurs relative frequently during the early period of treatment and delaying the introduction of a new agent until nine months after diagnosis would lose a significant fraction of patients before the agent could be introduced.
We attempted to hedge our bets by continuing the administration of MTP after the completion of adjuvant chemotherapy so that there would also be a period of time when MTP was administered without concomitant chemotherapy.
This decision led to a gap shown on this slide in red between the randomization of patients to one of the four treatment regimens and the introduction of the investigational agent approximately 3 months later. This allowed an imbalance in the treatment arms in the response to neoadjuvant chemotherapy, which is usually seen as an important predictor of outcome and which I will address again later. Of note, the current EURAMOS study randomizes patients after surgery and when the response to neoadjuvant chemotherapy is known, thus should avoid such imbalance. Also worth mentioning, however, is that 30% of the patients in the EURAMOS trial are currently not being randomized at the later time point because of patient or investigator unwillingness to accept randomization.
As we carry out phase II trials of inhibition of the IGF pathway, both single agent and in combination with chemotherapy, we should be considering how we will want to design the phase III trials of these agents. We have seen objective responses in sarcomas using anti-IGF pathway monoclonal antibodies, although the response rates are not high. We have pre-clinical evidence for synergy with chemotherapy. This suggests that we should introduce IGF inhibition and chemotherapy concurrently. Concurrent administration would also obviate the problem of delayed randomization. Just as we continue adjuvant chemotherapy after resection or radiation of primary sarcomas to treat microscopic disease, we will want to continue IGF pathway inhibition. Given the extremely favorable toxicity profile of the anti-IGF MoAbs, we may want to continue these agents even beyond the planned termination of chemotherapy, as maintenance therapy for sarcomas in remission.
Of course statistical considerations are critical in the design of randomized trials, especially those intended to support approval for use. Overall survival is the gold standard in oncology trials.
We will absolutely need to design these trials to capture overall survival as well as EFS, and this will require that these studies remain open and active to accrue information about patient vital status much longer than our usual practice.
While there are many phase 3 designs that should be considered when planning a study, I would like to spend a few minutes on a complex approach that is especially appealing in studies of indications with small patient populations – the factorial design – and its benefits and challenges. The use of a factorial design in a clinical trial is a strategy to answer more than one question with a cohort of patients. The intent of a factorial trial is to analyze the contribution of each factor by marginal analysis. This means that we estimate the effect of factor B while ignoring factor A. If there is no interaction between factor A and factor B, then the effect of factor B will be similar within each level (present/absent) of factor A.
The design is based on the assumption that there is no interaction between the two different interventions. It is a widely accepted statistical principle that the test of the hypothesis of no interaction should have a p-value greater than 10% in order to conclude that there is no significant interaction. This rule is more stringent than the 5% threshold we commonly employ for most statistical tests in clinical trials.
To conduct a clinical trial to prove that there is no interaction would be very difficult. Just as a clinical trial to demonstrate equivalence of 2 interventions requires a much larger sample size than a trial to detect a difference between the 2 interventions, a trial to prove no interaction would require a much larger sample size than a trial to detect an interaction.
For designing the phase 3 MTP study in osteosarcoma, we considered that our history had been to conduct one large prospective randomized trial each decade for osteosarcoma. It is worth noting that that history has been validated by subsequent events in the osteosarcoma clinical research community. If we designed a trial to answer a single question, we would not have the opportunity to ask another question for a decade. Thus we designed a factorial trial with two questions. The first question was whether the use of ifosfamide would improve outcome.
At the time we designed the study, there was intense interest in the use of ifosfamide for the treatment of osteosarcoma. Several studies had demonstrated the activity of ifosfamide against OS, both in recurrent disease and in newly diagnosed patients with metastatic disease. There was a widely held presumption that the addition of ifosfamide to the treatment of osteosarcoma would result in improved outcome. In addition, a large study had suggested that multi-drug resistance (mdr) played a role in osteosarcoma resistance to chemotherapy and ifosfamide is not an mdr substrate. By designing the trial as a substitution trial in the induction phase, we hoped to compare ifosfamide to cisplatin using tumor necrosis as an endpoint.
The second question was whether the addition of MTP to maintenance chemotherapy would improve the outcome.
The goal of this design was to answer each question independently, enabling us to assess the efficacy of MTP by comparing the two MTP-containing arms, A+ and B+ to the two No-MTP arms, A- and B-. This was the prospective design of the study. In the course of the study, COG enrolled a total of 662 patients with primary tumors clinically assessed to be resectable and no evidence of clinically detectable metastatic disease
It is worth spending a moment to examine the relationship between sample size and population size and the impact that this relationship has on estimates of error. This is an important issue when we enter into discussion with regulators about the need for a confirmatory trial.
Imagine a large jar which contains 1,000 marbles, 700 red and 300 blue. I chose 70% for this example to simulate the a priori EFS we predicted for patients with localized osteosarcoma in our prospective trial. If we remove a sample of 50 marbles from the jar, we anticipate a mean of 35 red and 15 blue marbles, the same 70% as the population, but the standard error of the mean is large, 6.3%.
If we begin with the same population, but take a much larger sample, 500 marbles, the standard error of the mean is much smaller.
In our phase III prospective randomized trial of the effect of MTP for the treatment of osteosarcoma we enrolled 777 patients in 4 years. We estimate the incidence of pediatric osteosarcoma in North America at 350-400 patients each year. This means that we enrolled roughly half of the pediatric population of patients with osteosarcoma in this clinical trial. Our estimate of effect size are robust, with much smaller errors, than we would see with a smaller sample size. For comparison, a similarly powered trial in breast cancer would have enrolled over 200,000 subjects.
In large scale phase III clinical trials, we perform interim analyses to monitor the safety of subjects in the clinical trial. We want to be sure that we do not subject patient subjects to unexpected or unacceptable toxicity. We want to be sure that we monitor study progression for differences in therapeutic assignment. If one arm of a study is clearly superior, continuation of the study would continue assignment of patient subjects to an inferior treatment. But we must be vigilant to perform analyses at pre-determined points. If we look at the data prior to the final analysis, we need to adjust the statistical test we use to conclude that we have identified differences between study arms.
During the time this trial was accruing patients, NCI implemented requirements that the Cancer Cooperative Groups, including POG and CCG, increase the rigor of the data monitoring committees and implement formal interim analysis plans. Therefore, while the interim analysis plan for this study was implemented following the initiation of the trial, the reason for the implementation was not related to inspection of accruing results.
Furthermore, the specific details of the interim analysis plan were not the result of having inspected the data (Mark Krailo, personal communication). The analyses were performed at predetermined intervals and included assessment of futility or highly significant differences that might suggest ending the study early.
None of the analyses demonstrated results that crossed a monitoring boundary for either the chemotherapy or MTP intervention. In addition, a proportional hazards regression model was used to assess interaction between the chemotherapy and MTP therapy; the interaction was not significant in any of the analyses. At the time we published our first analysis of the outcome of this study, we had not observed enough deaths to activate an analysis of overall survival. There were no interim analyses of overall survival, thus no adjustment for significance is required of our final survival analysis.
The choice of outcome variable or end point is a crucial decision in clinical trial design. Many of our current trials for the treatment of sarcomas use an end point of progression free survival. For newly diagnosed patients we more commonly use event free survival. Many trials in adult cancers employ median survival as a composite measure. For disease where EFS or survival exceeds 50%, median survival is a meaningless term. We have employed 80th percentile survival in some analyses. As I will discuss, overall survival and quality of life are considered the most robust outcome measures and have been preferred endpoints for regulatory approvals.
The NCI has published its position on the relative strength of endpoints from clinical trials. There is a clear hierarchy of study design, with the randomized controlled clinical trial at the top of the list, and double-blind trial design preferred to open-label trials.
In terms of endpoints, the NCI position places overall mortality at the top of the list, and considers all the more commonly used endpoints, including EFS, DFS, PFS, and response, to be inferior measures to indicate the efficacy of treatment.
The FDA convened an expert panel to develop guidelines for the development of new agents in pediatric oncology. The consensus position was that surrogate markers such as PFS and EFS may be used as early means to identify efficacy, but the use of these surrogate markers requires validation and correlation with clinical benefit.
Progression free survival is widely employed in assessing the impact of new agents in recurrent sarcoma. It allows the evaluation of agents in disease states where we cannot achieve a complete remission or minimal residual disease. We believe that prolongation of PFS is a measure of clinical benefit. The information is available quickly. We use it much less commonly in trials of newly diagnosed patients or in trials of the "pediatric" sarcomas. We have also learned that PFS is subject to significant ascertainment bias. We have learned that if we want to use PFS as our primary end point we must specify prospectively the timing of evaluations and we need centralized and blinded review of the imaging studies we employ to assess progression.
To date we have not had good acceptance from regulators that improved PFS, especially from a single trial, is sufficient to justify the registration of a new agent.
Event free survival is widely employed as a surrogate for overall survival. It allows us to incorporate secondary malignancy and toxic deaths into our assessment of outcome. It is less well recognized that EFS, just like PFS, is subject to ascertainment bias. When we design our trials using EFS as an endpoint we need to specify prospectively the evaluation schedule to be sure that evaluations are performed at the same intervals in study arms or sequential trials. We need to design our trials with centralized, blinded review of the imaging studies that we use to assess both continued event free survival or the occurrence of an event.
Challenges to the use of intermediate endpoints, especially in the absence of blinding and central review of images, are illustrated by the analysis of EFS in the phase 3 MTP study.
The clinical endpoints for this study were EFS, as an intermediate endpoint, and OS as the primary study aim. For each analysis, interaction between assigned chemotherapy and MTP was assessed using the proportional hazards regression. A p-value of 0.10 or less was considered evidence of a significant interaction.
We performed our analysis of event free survival, based on the number of events we had observed. At the time we published our first report of study results, we still had not observed half of the anticipated deaths, and therefore the study statistician had not carried out an analysis of survival. The first analysis of survival was performed and published at a later date.
The test for interaction for event free survival does not rise to the prospectively defined level of significance. It is, however, very close. The importance of a negative test is that it allows us to perform the marginal analyses as prospectively defined. We should not lose sight of the fact that the magnitude of difference is greater in regimen B than in regimen A. It is also important to note that the difference between the effect of MTP in regimens A and B is one of magnitude, not one of direction. Another way to describe this is to suggest that there is no qualitative interaction, but there is a suggestion of a quantitative interaction between the MTP intervention and the chemotherapy intervention when we examine event free survival.
Since the test for interaction does not meet the conventional prospectively defined level of significance, we can perform the marginal analyses. If we ignore assignment to receive or not to receive MTP, the addition of ifosfamide to the other three chemotherapy agents had no effect on event free survival as seen in the left panel.
If we ignore the chemotherapy assignment, the addition of MTP to chemotherapy was associated with an improved probability for event free survival, seen in the right panel. The probability that this difference arose by chance is 0.08. This does not meet a conventional level of significance, but does suggest a trend toward improved event free survival.
We did not propose to prove no interaction, and we cannot conclude that there is no interaction. We proposed only to test for the hypothesis of no interaction to determine if it was legitimate to perform the marginal analyses in the prospective trial design. A trial to exclude any possibility of interaction would have required well over 1800 patients.
Despite a negative test for interaction, analysis of EFS by study arm in this trial has created continued questions about the magnitude of difference between the addition of MTP to the two regimens. Since ascertainment bias cannot be ruled out, the importance of this potential interaction (or lack of) cannot be determined.
Survival is the ultimate gold standard end point. There are no issues of ascertainment bias. Survival analysis may require many years of followup. It does not include assessment of quality of life. For some pediatric trials, such as in the area of neuro-oncology, it is clear that some therapies are associated with prolongation of survival, but also with very significant compromise of quality of life. It is very likely that we will need to move to QOL adjusted survival as the ultimate gold standard for oncology studies.
This is an analysis of overall survival for patients with localized osteosarcoma treated on INT0133. This figure displays overall survival by treatment assignment. Overall survival was very similar for three drug chemotherapy and four drug chemotherapy without MTP, regimens A- and B-. The addition of MTP improved survival for both chemotherapy arms.
When we apply the test for interaction to overall survival, there is no hint of interaction. The hazard rations for the addition of MTP to regimen A and regimen B are similar in magnitude, and both show improved overall survival with the addition of MTP to chemotherapy.
When we ignore assignment to receive or not to receive MTP, we see that the addition of ifosfamide to the three other chemotherapy agents had no impact on overall survival.
When we ignore chemotherapy assignment, we see that the addition of MTP to chemotherapy resulted in a significant improvement in overall survival. The probability that this difference arose by chance is 0.03.
In this figure we have depicted the overall survival for the osteosarcoma MTP trial compared to SEER data. Note that survival for patients who were treated with chemotherapy but did not receive MTP is the same as the SEER data. The addition of MTP to chemotherapy resulted in a clinically meaningful and statistically significant improvement in survival with no incremental toxicity. It is important to note that the SEER data provides reassurance that the observed difference is not due to inferior outcome for the patients treated with chemotherapy alone.
Obviously quality of life incorporates survival and adds a dimension related to non-lethal toxicities. This is increasingly considered the optimal end point for studies of cancer therapies. Both survival and QOL require very long followup to make meaningful conclusions about therapeutic interventions. It is critical that we design all of our trials, even phase II trials, to capture survival and QOL outcome data. This will require that we keep studies open much longer than trials which are limited to PFS or EFS endpoints. It will require us to consider issues of patient crossover to novel agents and how such crossover could affect survival after progression of patients randomized to therapy which does not incorporate the novel agents. The need to maintain studies for longer time will place a significant burden on cooperative groups to expend group resources to keep studies open, and institutional investigators who will need to expend time and resources to keep studies open in followup to acquire and submit data. But every regulatory interaction has demanded this type of data to obtain registration of new agents.
A key feature in design of clinical trials is the duration of planned followup. As we have previously noted, we can obtain surrogate endpoints much more quickly in late stage patients than early ones and in all cases, more quickly than survival. However, survival is much more robust and viewed by regulators as the gold standard. We must recognize that we will need these data and anticipate much longer duration of followup than we are accustomed to design into our trials. We must plan to follow every study participant essentially for life.
In our prospective randomized clinical trial to assess MTP in osteosarcoma, the cooperative group made a very unfortunate decision to close the trial to follow-up in March, 2007. This meant that our ability to track patient vital status ceased. The regulators want to hold us to a standard of accountability of 95% completeness for all study participants. We meet that criterion at a median follow up of about 3 years, but not beyond. Institutional investigators have been unwilling to approach patients for additional followup because they interpret study closure to terminate their IRB approval to contact patients. Studies must remain open for followup, probably indefinitely.
This has important implications as we design our phase II and phase III studies of IGF inhibition in sarcomas. We will need to be sure that our phase II trials are designed to capture overall survival, something we have not previously done, and we will need to insure that our phase III trials remain open indefinitely to accrue information about patient vital status which will be essential for regulatory submissions.
Unplanned analyses are frequently performed after the completion of a study to explore new questions or issues that may have been identified during the study or review of the data. Subgroup analyses by demographic characteristics are expected by regulators, but again should be viewed as exploratory. These can provide support for the robustness of the data and new insights into questions that can be asked in future studies.
This Forest Plot illustrates the consistency of results across a broad range of demographic and prognostic factors. Analyses were conducted based on demographic factors such as race, gender, age and prognostic indicators such as LDH, and tumor size. Hazard ratios less than one that favor MTP appear as boxes on the left side of the vertical line and the size of each subset is relative to size of the box.
With one exception, patients 16 years and older, the direction of the treatment effect favors MTP. Further exploratory analyses focused on this age group have produced some interesting results and illustrate the kinds of questions such analyses can generate.
In the ISIS study, the investigators performed a subgroup analysis by the patient’s Zodiac sign. They noted that Libras and Pisces has an inferior outcome. As discussed by the ISIS investigators that include Sir Richard Peto and Sir Richard Doll caution must be used in interpreting subgroup analysis.
In the smaller group of patients older than 16, there was a greater proportion of individuals with unfavorable responses in the MTP arms. Histologic response to induction therapy is one of the best predictors of survival.
The stratification at randomization led to balance between treatment arms in the other important prognostic indicators for survival in osteosarcoma, including LDH level and site of tumor. Histologic response was not determined until after induction therapy and surgery and could not be controlled at randomization.
In the larger group of patients less than 16, a good balance in favorable response was seen.
Shown on the left is an analysis of disease free survival by treatment arm in the phase 3 MTP study excluding patients less than 16 years of age. In this analysis, unlike the ITT analysis of DFS or EFS, any suggestion of interaction has disappeared and these graphs are completely predictive of the overall survival outcome shown in the right panel.
The smaller size of the over 16 group and the excess of poor histologic responders in the MTP arms contribute to the findings seen in the Forest plot and completely account for the suggestion of interaction between MTP and ifosfamide in the EFS analysis.
When we use overall survival to evaluate an intervention in newly diagnosed patients with cancer, there is a concern that variability in post-relapse, or retrieval therapy, might influence subsequent survival. Certainly we have examples of disease where therapy following first relapse can have a marked impact on subsequent survival. In hematologic malignancies, the application of allogeneic bone marrow transplant following relapse after initial therapy can be curative, dramatically altering survival rates.
In the past, we have struggled to demonstrate the benefit of chemotherapy in most sarcomas in front-line treatment. There has been no demonstration of survival benefit for systemic therapy following recurrence.
In osteosarcoma, very large cohorts of patients have been analyzed following recurrence. There is unanimous agreement that there are no survivors if we are unable to perform surgery to remove all sites of clinically detectable recurrent disease. There has been no impact of chemotherapy on post relapse survival, especially in the era during which we conducted INT0133. We analyzed our patients following recurrence after initial therapy on INT0133 and there was no difference in the frequency of patients who underwent surgical resection for metastatic recurrent disease. Survival is the ultimate endpoint, and not subject to ascertainment bias or lack of central review.
When we examine the introduction of new biological agents for the treatment of sarcomas, it will be critical to consider the benefit:risk ratio. We will certainly require a robust indication of benefit. We will want to be sure that we monitor risk in the setting where the agent will be in clinical use, for example, in conjunction with chemotherapy. Agents with favorable toxicity profiles should have a higher priority for introduction into treatment than agents with greater toxicity and risk. Whenever we add to our current therapies, we must consider how we will move beyond this agent to the next improvement. Will the introduction of a new agent compromise our ability to introduce yet additional agents in the future? The better the toxicity profile, the less risk that a new agent will prevent the incorporation of additional treatment strategies in future trials.
Our prospective randomized phase III trial of MTP in osteosarcoma demonstrated a robust 30% reduction in the risk of death. We saw no grade 3 or 4 toxicity attributable to MTP.
This extremely favorable benefit/risk ratio argues strongly in favor for the approval of MTP for the treatment of osteosarcoma.
Our early experience with inhibition of the IGF pathway by monoclonal antibody against IGF1R suggest a very favorable toxicity profile. Phase II studies of these monoclonal antibodies in combination with chemotherapy for carcinomas have shown no increased toxicity from either chemotherapy or the biological. It is probable that we will see a favorable benefit risk ratio in phase III trials in sarcomas.
There are several regulatory issues that have arisen in discussions about the development of biological agents for the treatment of sarcomas.
Do we need to perform our randomized trials in a double blinded manner using placebo controls?
Will we be able to obtain large sample size? Do we need large sample size?
Will we need to perform confirmatory studies after an initial study which detects benefit?
Ethicists who have considered the question have generally concluded that placebos are unacceptable for minors. For example, in some IGF1R monoclonal antibody trials, the agent is administered weekly by intravenous infusion. The administration of weekly infusions of placebo to minors incurs risks of infection generally considered ethically unacceptable.
In our trial of MTP, we required demonstration of fever and chills with the first administration of drug before allowing premedication. What would be an appropriate placebo? Either we would use an agent which did not cause fever and chills, and therefore was obviously a placebo, or use an agent which did cause fever and chills, and by definition put patients at risk.
We have been told by at least one pharmaceutical that the IGF1R formulation has physico-chemical characteristics that make it almost impossible to devise a placebo that would fool the compounding pharmacist.
When we consider sample sizes, we are confronted by the fact that all sarcomas are rare, rare enough to be considered orphan diseases.
Our track record in pediatric oncology is one phase III randomized trial per decade for osteosarcoma and Ewing sarcoma. For both these diagnoses, our most recent clinical trials required development of international consortia to achieve the patient numbers needed to achieve adequate power to address the study questions.
How will we design trials of new agents in sarcoma to have adequate power to address study questions and to have adequate power that will not require confirmatory studies that will delay even further the availability of these new agents to our patients?
Our phase III prospective randomized trial of the addition of MTP to chemotherapy for osteosarcoma is the largest prospective randomized trial in osteosarcoma ever completed.
The sample size was roughly half the population of eligible children.
We demonstrated a robust survival advantage with a favorable benefit:risk ratio.
Mounting a successor study would be very complicated. Many investigators would find it difficult to withhold MTP from young patients with osteosarcoma, so conducting a trial in which patients were assigned not to receive this agent would accrue slowly.
We will certainly be moving forward with trials of IGF1 inhibitors. We will certainly be developing other biological and targeted therapies. We will need to be able to combine these novel therapies with our existing therapies and incorporate them into our standard clinical practice. We have a responsibility to our patients to test every new agent rigorously to insure that we have robust information about benefit and risk. But we also have a responsibility to bring the safe and active new agents to our patients as rapidly as we can. We need to be thoughtful and pro-active as we design our studies to be sure that we acquire all the relevant data that will allow these studies to support an indication for the new agent.
I hope that my description of our experience with MTP has been informative for you and that my thoughts about IGF1R inhibitors have helped clarify for you some of the issues we face as we develop novel therapies for patients with sarcomas.
I wish to thank CTOS and the Axelrads for the honor of allowing me to present to you today, and I welcome any questions.
V5N6 ESUN Copyright © 2008 Liddy Shriver Sarcoma Initiative.