|Home | About | Journals | Submit | Contact Us | Français|
Phase I studies in cancer have changed in recent years. With the advent of new less toxic targeted agents, more patients may now be candidates for new drug studies earlier in the course of their disease. It is to the advantage of the members of the oncology community to know more about the details and requirements for participation in early phase clinical trials so they can advocate for their patients and help them decide when such trials may be an appropriate choice. In order to examine the work intensity of early phase cancer clinical trials, we compared the study requirements of phase I and II protocols.
As a surrogate of study complexity, we examined five parameters—number of physical exams, vital sign determinations, electrocardiograms (ECGs), non-pharmacokinetic laboratory tests, and pharmacokinetic (PK) sampling—in the first four weeks of protocol, in 90 studies (49 phase I and 41 phase II).
From July 2004 through March 2007, there were 49 phase I trials in the Phase I Program, nine phase II studies conducted by physicians appointed in that program, and 32 phase II trials with accessible data in the Department of Thoracic/Head & Neck Medical Oncology. In the phase I versus phase II trials, there were significantly more (p < 0.05) physical exams (mean ± SE = 3.16 ± 0.24 vs. 2.22 ± 0.13), vital signs (5.63 ± 0.61 vs. 2.80 ± 0.26), ECGs (4.36 ± 1.16 vs. 0.80 ± 0.17), non-PK lab tests (18.08 ± 1.31 vs. 10.12 ± 0.65), and PKs (15.14 ± 1.79 vs 1.02 ± 0.53). These values were also significantly different (p<0.005 for each) when comparing medians by non-parametric tests.
While both phase I and phase II trials have substantial study requirements, those for the phase I studies were significantly higher. Successful conduct of early phase clinical trials requires significant research infrastructure.
Phase I cancer clinical trials play a crucial role in drug development. Commonly known as “first-in-human” studies, phase I trials also encompass new combinations and/or new dosing schedules of FDA-approved drugs1. Though several different designs for these trials have been used, the standard design in phase I studies is to enroll cohorts of three to six patients at increasing dose levels until dose-limiting toxicity is reached2-5. Traditionally, the primary endpoints of phase I studies of chemotherapy included determining the pharmacokinetics of the drug6 as well as the maximum tolerated dose7, with subsequent recommendations of a dose and schedule with which to proceed to phase II studies. In contrast to phase I trials, the primary endpoint for phase II studies is efficacy. However, for newer targeted agents, other endpoints for phase I trials, such as optimal biologic dose (e.g., the dose at which the target is best modulated) are becoming increasingly important, as is describing response signals and any early evidence of correlation between these signals and target inhibition and/or patient/tumor characteristics8,9. These endpoints are more complex to obtain because they require measuring target impact in tumor tissue and/or assessing genomic expression patterns in the host or in the tumor10,11. In addition, over the last decade, there has been increasing scrutiny of clinical trials research, with an emphasis on tight adherence to protocol requirements as well as careful monitoring of patient safety12-14. All of these factors have contributed to the impression that phase I trials have become increasingly complex.
Much has been written on the risks and benefits15-17 and the design and importance of phase I trials1-3, as well as the cost of clinical trials versus standard care18-25. There is also literature regarding the work involved in clinical research, but mostly as it regards phase II and III protocols19,26. Until now, little has been done to examine the intensity of the phase I protocol schedule. This information is important because the intensity of phase I trials impacts both the patient and the clinical research staff.
In this study, we analyzed a series of phase I and II trials conducted at M. D. Anderson Cancer Center to further examine and compare the work intensity required for the two different phases of clinical research.
The requirements for all phase I and II protocols managed by the Phase I Program or by physicians with an appointment in that program at M. D. Anderson Cancer Center, as well as the phase II protocols with accessible data managed by the Department of Thoracic/Head & Neck Medical Oncology during fiscal year 2003 through 2007 were reviewed. [For the phase I trials, only studies that were approved by the Internal Review Board and that had accessible data were analyzed. The time frame for these studies was July, 2004 through March, 2007.] After exclusion of 7 trials (5 were phase I, and 2 were phase II), due to either past closure or inaccessible data, 49 phase I trials and 41 phase II trials remained for investigation. Each trial and its requirements were reviewed by tabulating data from the Clinical and Translational Research Center's (CTRC) database, the institution's Protocol Document Online System, the M. D. Anderson mandated protocol abstract, and/or the protocol order sets. All trials had been approved by the institutional review board. In preparing this report, we obtained protocol requirements from sources that contained no patient-specific information.
We analyzed several parameters as representatives of the complexity of requirements. These included: 1) physical examination, 2) vital sign monitoring, 3) electrocardiography monitoring, 4) non-PK laboratory testing, and 5) pharmacokinetic sampling as conducted in the first 28-day cycle of each protocol. To evaluate our findings, we developed a comprehensive database, along with a Microsoft Excel spreadsheet matrix to analyze the number of correlative studies.
The statistical aims included comparison of differences in each of the five variables (physical exams, vital signs, ECGs, non-PK lab draws, and PK timepoints) by: 1) phase of study (i.e., phase I versus phase II), 2) sponsorship within phase (industry, non-sponsored investigator-initiated, NCI/peer-reviewed group), and 3) sponsorship (industry, non-sponsored investigator-initiated, NCI/peer-reviewed group). All statistical analyses were performed by our biostatistician (X.L.).
Summary statistics were used to report the number of observations, mean, standard deviation, standard error, median and range of each of the five variables by phase of study (phase I and phase II), and by sponsor (industry, investigator-initiated, NCI/peer-reviewed group). Non-parametric methods of comparison were implemented. If the number of groups for comparison was two, the Wilcoxon rank-sum test was used. If the number of groups for comparison was more than two, the Kruskal-Wallis test was used. P-values were reported, and differences between protocol phase and sponsor were considered statistically significant if the p-value for a given test was less than 0.05.
All the statistics analysis was performed in SAS® Version 9.1 (SAS Institute, Cary, NC).
We reviewed a total of 90 trials, of which 49 were phase I and 41 were phase II, approved and managed in the same time period. The parameters were those assessed during the first four weeks of the study. For all of the correlative tests we looked at, there were significantly more tests required in phase I trials than in the phase II trials (Table 1). Physical examination timepoints differed between the two groups, with significantly more required in phase I (mean ± SE = 3.16 ± 0.24 vs. 2.22 ± 0.13). The number of vital sign measurements differed significantly between phase I and II, with a greater mean number required in phase I (5.63 ± 0.61 vs. 2.80 ± 0.26). As far as ECGs, the mean number was also significantly greater in the phase I versus phase II trials (4.36 ± 1.16 vs. 0.80 ± 0.18). The mean number of non-PK lab draws differed between phase I and phase II (18.08 ± 1.31 vs. 10.12 ± 0.65). Very notable was the difference in the mean number of PKs between phase I (15.14 ± 1.79) and phase II trials (1.02 ± 0.53). These parameters were also significantly different (p<0.005) for each when comparing medians by non-parametric tests. Therefore, in this study while both phases of clinical trials required significant numbers of correlatives, the requirements for phase I were considerably greater.
Similar to previous studies looking at work requirements26, we also looked at how the sponsorship of different trials affected the number of requirements (Table 2). Of the phase I trials, 38/49 were industry sponsored, 8/49 were non-sponsored, investigator initiated, and 3/49 were National Cancer Institute (NCI) sponsored. As for the phase II trials, 19/41 were industry sponsored, 6/41 were non-sponsored, investigator initiated, and 16/41 were NCI/Department of Defense/or other peer-review group sponsored. (Trials that were investigator initiated, but funded by a sponsor who held the IND, are listed as industry sponsored or NCI sponsored). As noted previously, 78% of the phase I trials were sponsored by private industry with only 6% as NCI or peer reviewed sponsorship – a trend we have witnessed for several years. This limited number of phase I NCI studies precluded analysis of sponsorship within each of the phase I and phase II trials. In that context, for each of the five parameters studied, industry-sponsored trials required an overall greater number of measurements than both the non-sponsored investigator-initiated trials and the NCI/peer-review group trials (Table 3a). However, only the difference in number of electrocardiograms (mean ± SD = 3.93 ± 7.65 vs. 0.86 ± 0.86 vs. 0.58 ± 0.84), and PK timepoints (13.07 ± 12.74 vs. 1.86 ± 5.11 vs. 0.68 ± 2.21) reached statistical significance (Table 3b). Our sense, however, was that NCI trials, when controlled for study phase, were every bit as rigorous as industry-sponsored studies.
New drug development is essential for progress to be made in the treatment of cancer. During the last few years, concern for patient safety in clinical trials has resulted in strict regulatory requirements12-14. In addition, with the advent of targeted agents, correlative studies to understand pharmacodynamic effects and target impact have become crucial to optimal drug development6. These requirements have resulted in the impression of a greater workload assumed by those investigators and institutions that participate in early phase clinical trials.
Because there are few data on this aspect of study conduct for phase I protocols, our analysis focused on them. And, while many activities are required to run a successful phase I clinical trial, including the workload of clinical study coordinators, imaging, etc., we did not specifically address these activities. We instead chose to look at a set of five requirements of current successful, early-phase cancer clinical trials. We chose these variables because of prevalence and their importance in phase I studies. For instance, determining PK is often a primary endpoint6. With more targeted agents being developed, the strategy is evolving to find a less toxic/more effective dose versus the maximal tolerated dose8-11. Frequent ECG monitoring has also become a common requirement to monitor QT interval and rhythms27, and vital signs, physical exams, and non-PK lab draws are for safety assessment12-14, which is another primary endpoint of phase I trials. We found that for all five parameters (physical examination, vital sign monitoring, electrocardiograms, non-PK lab testing, and PKs), there were significantly greater demands in phase I trials versus phase II trials.
These results are supported by earlier findings in the literature which examined requirements of phase II and III studies. For instance, a study by Roche, et al.26 reviewed the time requirements of clinical research associates involved in clinical trial coordination and data collection. They found that the work intensity of phase II and III (and a small number of phase I) studies was greater than had been previously acknowledged. Furthermore, phase and sponsor were found to be significant independent factors predicting an increased workload, after controlling for stage. Similar to our results, they demonstrated that earlier phase studies, and those sponsored by industry were more time-consuming than later phase trials or those that were investigator-initiated.
Issues that have been at the forefront in the past several decades regarding phase I trials are fading with the development of newer, more targeted therapies. Concerns have included the possibility of serious toxicity with minimal benefit from experimental agents. This, as well as concerns about patient safety, have resulted in increasingly stringent regulatory requirements12-14. In the largest analysis of phase I studies, including almost 12,000 patients, Horstmann et al.15, demonstrated that the toxic death rate was less than 0.5%, which is surprisingly low, considering that these patients are, by definition, afflicted by terminal, untreatable cancer. In addition, while prior studies had shown an overall response rate for phase I studies at only 4.4 percent, their data revealed an overall response rate (partial and complete remission) of 10.6%. If less than partial response and stable disease were included, the percent of patients “benefiting” was 44.7%15,16.
Successful clinical trials of today remain the cornerstone of new drug development. Both phase I and phase II clinical trials have recently endured an increase in the number of requirements and workload. The data from our study are consistent with those from previous studies that focused on phase II and III requirements. However, our findings show that phase I studies are even more complex, as shown here by extensive correlative testing and safety monitoring, indicating that these studies require substantial resources and infrastructure.
This publication was made possible in part by Grant Number RR024148 from the National Center for Research Resources (NCRR), a component of the National Institutes of Health (NIH), and NIH Roadmap for Medical Research. Its contents are solely the responsibility of the authors and do not necessarily represent the official view of NCRR or NIH. Information on NCRR is available at http://www.ncrr.nih.gov/. Information on Re-engineering the Clinical Research Enterprise can be obtained from http://nihroadmap.nih.gov/clinicalresearch/overview-translational.asp.