- Email: [email protected]
Oncology Endpoints
CHAPTER
13
Oncology Endpoints: Time to Progression I. INTRODUCTION The present chapter continues with oncology endpoints, and covers the fourth topic indicated by the check mark. Objective response Overall survival Progression-free survival Time to progression ✓ Disease-free survival Time to distant metastasis. Time to progression (TTP), also known as time to disease progression, is a common endpoint in oncology clinical trials. TTP and progression-free survival (PFS) are defined by FDA’s Guidance for Industry (1) as follows: l
l
l
l
l
l
TTP and PFS have served as primary endpoints for drug approval.TTP is defined as the time from randomizationuntilobjectivetumorprogression;TTPdoesnotincludedeaths.PFSisdefinedas thetimefromrandomizationuntilobjectivetumorprogressionordeath.Theprecisedefinitionof tumor progression is important and should be carefully detailed in the protocol.
But the values for TTP and PFS are often almost identical, as revealed in the following examples from several clinical trials of breast cancer (2). In one of these trials, the values of TTP and PFS, for one particular arm, were both 5.5 months (3). In another trial, the values for TTP and PFS, for one of the arms, were both 6.7 months (4). In yet another trial, the values for TTP and PFS, for one of the arms, were identical,
1
U.S. Dept. of Health and Human Services. Food and Drug Administration. Guidance for Industry. Clinical trial endpoints for the approval of cancer drugs and biologics; April 2005. 2 Burzykowski T, Buyse M, Piccart-Gebhart MJ, et al. Evaluation of tumor response, disease control, progressionfree survival, and time to progression as potential surrogate end points in metastatic breast cancer. J Clin Oncol. 2008;26:1987–1992. 3 Bonneterre J, Dieras V, Tubiana-Hulin M, et al. Phase II multicenter randomized study of deocetaxel plus epirubicin versus fluorouracil plus epirubicin and cyclophosphamide in metastatic breast cancer. Br J Cancer. 2004;91:1466–1471. 4 Bontenbal M, Creemers GJ, Braun HJ, et al. Phase II to III study comparing doxorubicin and docetaxel with fluorouracil, doxorubicin, and cyclophosphamide as first-line chemotherapy in patients with metastatic breast cancer: results of a Dutch community setting trial for the clinical trial group of the Comprehensive Cancer Centre. J Clin Oncol. 2005;23:7081–7088.
Clinical Trials DOI: 10.1016/B978-0-12-391911-3.00013-X
© 2012 Elsevier Inc. All rights reserved.
237
238
Clinical Trials
Table 13.1 The Paccagnella study Arm A (PC arm) (paclitaxel carboplatin)
Objective response TTP Overall survival
20.2% 5.1 months 8.3 months
Arm B (PCG arm) (paclitaxel carboplatin gemcitabine)
Significance of difference between the two study arms
43.6% 7.6 months 10.8 months
P .0001 P .012 P .032
7.4 months (5). Moreover, in another trial the values for PFS and TTP were both 18.4 weeks (6). Please note that TTP does not, but PFS does, use survival time as a trigger for the endpoint. As a matter of first impression, it might be asked, “What is the use of having both types of endpoints?”
II. AGREEMENT OF RESULTS FROM OBJECTIVE RESPONSE, TIME TO PROGRESSION, AND OVERALL SURVIVAL – THE PACCAGNELLA STUDY The following clinical trial provides a straightforward account of various endpoints. In a clinical trial of non-small cell lung cancer (NSCLC), Paccagnella et al. (7) used three endpoints, objective response, TTP, and overall survival. The trial had two arms, as indicated in Table 13.1. Arm A. Paclitaxel and carboplatin (PC) Arm B. Paclitaxel, carboplatin, and gemcitabine (PCG). The results are straightforward (Table 13.1). The triple drug combination was superior, in terms of objective response (WHO criteria), median TTP, and median overall survival. It might be pointed out that the significance (P value) was greater for the endpoint of TTP than for overall survival, indicating an advantage in using TTP as an endpoint. l
l
III. CAN THE VALUE FOR PFS BE LESS THAN THE VALUE FOR TTP? PFS and TTP often have values that are nearly identical. Hence, it might be asked what might cause these values to differ, for example what might cause the value of PFS to 5
Paridaens R, Biganzoli L, Bruning P, et al. Paclitaxel versus doxorubicin as first-line single agent chemotherapy for metastatic breast cancer: a European Organization for Research and Treatment of Cancer randomized study with cross-over. J Clin Oncol. 2000;18:724–733. 6 Brandes AA, Basso U, Reni M, et al. First-line chemotherapy with cisplatin plus fractionated temozolomide in recurrent glioblastoma multiforme: a phase II study of the Gruppo Italiano Cooperativo di Neuro-Oncologia. J Clin Oncol. 2004;22:1598–1604. 7 Paccagnella A, Oniga F, Bearz A, et al. Adding gemcitabine to paclitaxel/carboplatin combination increases survival in advanced non-small-cell lung cancer: results of a phase II–III study. J Clin Oncol. 2006;24:681–687.
Oncology Endpoints: Time to Progression
be less than TTP. The situation where PFS is less than the value for TTP is one where deaths due to drug toxicity (but not from the cancer) occur early on in the trial. This is, “when deaths due to the disease are rare, but those due to the drug are not” (8). Examples of this are likely to be found in clinical trials on the elderly, especially where radiation therapy is followed by chemotherapy (9). It might also be asked when the value for PFS will likely be equal to the value for TTP. PFS will likely be equal to TTP when, at early times in the clinical trial, the tumors in most patients grow to the point where they trigger the RECIST criteria for tumor progression, but where the rate of patient deaths, due to any cause, is low.
IV. TIME TO PROGRESSION MAY BE THE PREFERRED ENDPOINT WHERE, ONCE THE TRIAL IS CONCLUDED, PATIENTS RECEIVE ADDITIONAL CHEMOTHERAPY – THE PARK STUDY In a clinical trial of non-small cell lung cancer (NSCLC), Park et al. (10) used the endpoints of objective response, TTP, overall survival, as well as endpoints set forth in health-related quality of life (HRQoL) questionnaires. Objective response was determined by the WHO criteria. Before randomizing the subjects, all subjects were treated with two cycles of chemotherapy in a run-in period. Only subjects not showing disease progression were randomized to the two arms. Regarding the run-in period, the investigators stated that one reason for including the run-in was to enrich the study with a more homogeneous patient population. This particular rationale is further described in this textbook in the chapter on run-in periods. The Park study had two arms, as follows: Arm A. Six cycles of chemotherapy Arm B. Four cycles of chemotherapy. Patients were monitored for a total of 18 months. Objective response was measured at 3-month intervals by computed tomography. According to TTP, arm A fared better than arm B, as shown in Table 13.2. According to overall survival, there was no significant difference between arms A and B. The investigators were careful to describe reasons for leaving the trial, writing that “[t]reatment was continued until the maximum of four or six cycles was completed, depending on random assignment, unless disease progression or unacceptable toxicity occurred or unless the patient refused further chemotherapy.” Where tumor size or number increased to the point where the endpoint of progression was triggered, treatment with study drugs was discontinued, and the subjects had the option l
l
8
Burzykowski T. E-mail of September 12, 2010. Buyse M. E-mail of September 12, 2010. 10 Park JO, Kim SW, Ahn JS, et al. Phase III trial of two versus four additional cycles in patients who are nonprogressive after two cycles of platinum-based chemotherapy in non small-cell lung cancer. J Clin Oncol. 2007;25:5233–5239. 9
239
240
Clinical Trials
Table 13.2 The Park study
Arm A (6 cycles) Arm B (4 cycles) Significance of difference between arms A and B
TTP (months; median value)
Overall survival (months; median value)
6.2 months 4.6 months P .001
14.9 months 15.9 months P .461
of second-line therapy. As stated, “[s]econd-line chemotherapy was considered at the discretion of the treating oncologist after documentation of progression.” To clarify this point, it is often the case in trial design that where the subject experiences progression (the tumor size or number reaches or passes a limit set for by the RECIST or WHO criteria), treatment with study drug or control is halted, and the subject has the option of receiving second-line treatment. At this point, it is intuitively obvious that this second-line chemotherapy cannot influence the date of progression (for any given patient), but it can influence the survival time. Regarding the finding that the 6-cycle group fared better than the 4-cycle group, in terms of the TTP endpoint, but that there was little difference in the endpoints of overall survival, the authors believed that this discrepancy was likely due to confounding of the survival times by the second-line therapy. In the authors’ words, “[t]he main reason why the TTP benefit did not translate into the survival benefit probably involved the dilution effect of the second-line chemotherapy.” The takehome lesson is that clinical trials should include the endpoint of TTP, where it is expected that many patients will be receiving second-line treatments.
V. THE ENDPOINT OF TTP MAY BE PREFERRED OVER SURVIVAL ENDPOINTS, WHERE DEATHS RESULT FROM CAUSES OTHER THAN CANCER – THE LLOVET STUDY Time to progression may be the preferred endpoint over an endpoint that takes into account survival data, such as the endpoints of PFS, disease-free survival (DFS), or overall survival. This type of preference for TTP is the case where the major cause of death, during the course of the clinical trial, is unrelated to the cancer. The following provides the example of liver cancer, that is, hepatocellular carcinoma (HCC). According to Llovet et al. (11) “although disease- and progression-free survival are appropriate endpoints in other solid tumors, they are particularly unreliable endpoints in HCC research because death resulting from the natural history of cirrhosis might confound detection of potential benefits from effective drugs. That is, a type 11
Llovet JM, Di Bisceglie AM, Bruix J, et al. Design and endpoints of clinical trials in hepatocellular carcinoma. J Natl Cancer Inst. 2008;100:698–711.
Oncology Endpoints: Time to Progression
II error might result from using progression-free survival as an endpoint in a suboptimal population in early phases of drug development.” Consistently, Dancey et al. (12) stated that, “deaths prior to disease progression are not correlated with progression. In situations where many deaths are unrelated to cancer or treatment, TTP can be an acceptable and preferred end-point.” The term Type II error refers to the conclusion that the drug does not work when in fact it actually does work (13). The term Type I error refers to the conclusion that the drug works when in fact it really fails to work. A related problem is that, during the course of a clinical trial, the cause of death is often not reliably collected or documented. Typically, deaths occurring after disease progression are considered malignant, while all deaths occurring prior to disease progression are considered non-malignant (14). The distinction between drug-induced deaths and disease-induced deaths is not often made. In the context of liver cancer, the endpoint of overall survival is expected to be reliable, where examination of study subjects shows that cirrhosis is not a problem. Thus, “[i]n the exceptional circumstances in which these [survival] endpoints are applied, a restrictive selection of patients with well-preserved liver function is recommended to minimize the impact of death unrelated with tumor progression” (15). In a study of hepatocellular carcinoma, Llovet et al. (16) treated subjects with either sorafenib or placebo. The study was designed to capture information on drug efficacy while confounding effects of deaths unrelated to cancer. Since hepatocellular carcinoma occurs mainly in patients with cirrhosis, the inclusion/exclusion criteria required that subjects have well-preserved liver function. However, it was still the case that endpoints tied to survival, such as progression-free survival (PFS), might be suboptimal because of the confusing effect of the cirrhosis. Thus, the Llovet study provides a general lesson regarding the selection of inclusion/exclusion criteria. In the sorafenib group, 7 patients (2%) had a partial response and 211 (71%) had stable disease (according to RECIST), whereas in the placebo group, 2 patients (1%) had a partial response and 204 (67%) had stable disease. There were no complete responses in either group (Fig. 13.1).
12
Dancey JE, Dodd LE, Ford R, et al. Recommendations for the assessment of progression in randomised cancer treatment trials. Eur J Cancer. 2009;45:281–289. 13 Type I error is rejecting the null hypothesis when it is true and Type II error is accepting the null hypothesis when it is false. See, e.g., Biau DJ, Jolles BM, Porcher R. P value and the theory of hypothesis testing: an explanation for new researchers. Clin Orthop Relat Res. 2010;468:885–892. 14 Buyse M. E-mail of September 20, 2010. 15 Llovet JM, Di Bisceglie AM, Bruix J, et al. Design and endpoints of clinical trials in hepatocellular carcinoma. J Natl Cancer Inst. 2008;100:698–711. 16 Llovet JM, Ricci S, Mazzaferro V, et al. Sorafenib in advanced hepatocellular carcinoma. New Engl J Med. 2008;359:378–390.
241
242
Clinical Trials
Figure 13.1 Kaplan-Meier plot of time to progression. Progression was defined by the RECIST criteria. A comparison of arm A (sorafenib) and arm B (placebo) reveals that the study drug was effective against the cancer. Median TTP was 5.5 months (arm A) and 2.8 months (arm B)
Median TTP was 5.5 months in the sorafenib arm and 2.8 months in the placebo arm (HR 0.58; P .001), while median overall survival was 10.7 months in the sorafenib arm and 7.9 months in the placebo arm (HR 0.69; P .001). This demonstrable efficacy of sorafenib was considered to be a therapeutic breakthrough, in view of the inevitably fatal outcome of this type of cancer, and lack of any effective treatment (Fig. 13.2).
VI. THE ENDPOINT OF OVERALL SURVIVAL MAY BE PREFERRED OVER OBJECTIVE RESPONSE OR OVER TTP, WHERE THE DRUG IS CLASSED AS A CYTOSTATIC DRUG – THE LLOVET STUDY Llovet et al. (17) provide a lesson regarding the utility, or actually lack thereof, of objective response, where the drug is cytostatic, rather than cytotoxic. The Llovet study compares the endpoint of objective response with the endpoint of overall survival. The study drug, sorafenib, inhibits angiogenesis and delays the progression of liver tumors in hepatocellular carcinoma. In the Llovet study, response consisted of complete response, partial response, or stable disease, according to the RECIST criteria. Where a drug is classed as a cytostatic drug, endpoints that rely entirely on tumor size or number, such as objective response and TTP, might not do justice to the survival benefit of the drug. With a cytostatic drug, a more reasonable assessment of the 17
Llovet JM, Ricci S, Mazzaferro V, et al. Sorafenib in advanced hepatocellular carcinoma. New Engl J Med. 2008;359:378–390.
Oncology Endpoints: Time to Progression
Figure 13.2 Kaplan-Meier plot of overall survival. A comparison of arm A (sorafenib) and arm B (placebo) reveals that the study drug was effective against the cancer. Median overall survival was 10.7 months (arm A) and 7.9 months (arm B)
drug’s efficacy might be acquired by using endpoints that rely partially or entirely on survival time. In the words of Llovet et al. (18) “current targeted agents may act as cytostatic agents…and possibly improve survival with no measurable change in tumor size.” Thus overall survival might be a better endpoint than endpoints that derived from objective response, such as the endpoint of objective response itself, or the endpoint of TTP, when the drug is classified as one that is cytostatic (not cytotoxic). The Llovet study on sorafenib concluded that the incidence of objective response was relatively low, and the effects on overall survival were high. Consistent with this interpretation, a review article observed that sorafenib does not commonly improve objective response, though it does stabilize cancer and enhances survival (19). Regarding cytostatic drugs, Fleming et al. (20) have reiterated the point that endpoints of objective response and time to progression (TTP) are relatively insensitive to the effects of cytostatic drugs, and that the efficacy of this class of drugs can more effectively be captured by the endpoint of progression-free survival (PFS) or overall survival. Another way around the problem of objective response is to change the technology used for measuring objective response. Instead of just using techniques that measure 18
Llovet JM, Di Bisceglie AM, Bruix J, et al. Design and endpoints of clinical trials in hepatocellular carcinoma. J Natl Cancer Inst. 2008;100:698–711. 19 Almhanna K, Philip PA. Safety and efficacy of sorafenib in the treatment of hepatocellular carcinoma. Onco Targets Ther. 2009;2:261–267. 20 Fleming TR, Rothmann MD, Lu HL. Issues in using progression-free survival when evaluating oncology products. J Clin Oncol. 2009;27:2874–2880.
243
244
Clinical Trials
the dimensions of tumors, techniques of functional imaging such as positron emission tomography (PET) can be used to assess the metabolism of existing tumors (21,22,23).
VII. TIME TO PROGRESSION MAY SHOW EFFICACY, WHERE THE ENDPOINT OF OVERALL SURVIVAL FAILS TO SHOW EFFICACY, WHERE THE NUMBER OF SUBJECTS IS SMALL – THE MCDERMOTT STUDY In a study of melanoma, McDermott et al. (24) used the endpoints of objective response (RECIST criteria), TTP, PFS, and overall survival. There were two study arms. This study used an add-on design. In other words, subjects in one arm received dacarbazine plus placebo, while subjects in the other arm received dacarbazine plus sorafenib. At the time of the clinical trial, dacarbazine was the most commonly used drug for melanoma. In part for ethical reasons, both study arms received dacarbazine. Sorafenib was the add-on drug. Arm A. Dacarbazine plus placebo Arm B. Dacarbazine plus sorafenib. The results are shown in Table 13.3. The study drug worked better than the placebo, as measured by the endpoints of TTP and PFS. However, there was no difference in terms of the endpoints of overall survival. The data provide an instructive account of the use of hazard ratio. The authors stated that, in terms of overall survival, there was “no difference.” The “no difference” result has a hazard ratio of 1.022, which is consistent with the teachings in the statistics chapter in this book. Spruance et al. (25) remind us that the hazard ratio can be greater than 1.0, or less than 1.0, depending on whether it is asked if treatment A works better than treatment B, or if treatment B works better than treatment A. The authors pointed out that the study drug was better than placebo, according to the endpoints of TTP and PFS, but explained the lack of effect in overall survival, by referring to the small number of subjects. In the words of the authors, “[r]andomized l
l
Table 13.3 The McDermott study Arm A. Placebo
Arm B. Sorafenib
Hazard ratio
P value
TTP PFS Overall survival
21.1 weeks 21.1 weeks 51.3
0.619 0.665 1.022
.03 .068 .927
21
11.7 weeks 11.7 weeks 45.6
Avril N, Propper D. Functional PET imaging in cancer drug development. Future Oncol. 2007;3:215–228. Kyle F, Spicer J. Targeted therapies in non-small cell lung cancer. Cancer Imaging. 2008;8:199–205. 23 Weber WA. Positron emission tomography as an imaging biomarker. J Clin Oncol. 2006;24:3282–3292. 24 McDermott DF, Sosman JA, Gonzalez R, et al. Double-blind randomized phase II study of the combination of sorafenib and dacarbazine in patients with advanced melanoma: a report from the 11715 Study Group. J Clin Oncol. 2008;26:2178–2185. 25 Spruance SL, Reid JE, Grace M, Samore M. Hazard ratio in clinical trials. Antimicrob Agents Chemother. 2004;48:2787–2792. 22
Oncology Endpoints: Time to Progression
phase II clinical trials are often limited by their small sample sizes to detect true treatment differences between the study arms. In this study, we observed improvements in PFS and TTP…however, these findings did not translate into an improvement in OS” (26).
VIII. TIME TO PROGRESSION MAY SHOW EFFICACY, WHERE THE ENDPOINT OF OVERALL SURVIVAL FAILED TO SHOW EFFICACY, WHERE THE DURATION OF THE TRIAL WAS TOO SHORT – THE CAPPUZZO STUDY The following study on non-small cell lung cancer (NSCLC) used the endpoints of objective response, TTP, and overall survival. Cappuzzo et al. (27) administered the same drug (gefitinib) to all study subjects. Gefitinib inhibits enzymatic activity of a particular kinase, namely, epidermal growth factor receptor (EGFR) tyrosine kinase. Cytogenetic assays were conducted on all study subjects prior to administering the drug, to determine the number of EGFR genes in tumor cells. Tumor cells were acquired via biopsies, and the number of copies of the EGFR gene in a sampling of tumor cells was determined by way of the fluorescent in situ hybridization (FISH) technique. Twenty-five of the patients were FISH positive for EGFR. Eleven patients were FISH negative for EGFR. “FISH positive” was defined as tumor cells carrying four or more copies of the EGFR gene in 40% of the tumor cells. “FISH negative” was defined as tumor cells carrying four or more copies of the EGFR gene in less than 40% of the tumor cells. Table 13.4 demonstrates the greater efficacy of gefitinib in patients bearing EGFR positive tumor cells (as compared to patients with EGFR negative tumor cells). This Table 13.4 The Cappuzzo study Patients with EGFR positive tumors
Patients with EGFR negative tumors
Significance of difference of endpoint, comparing EGFR and EGFR patients
Objective response TTP Overall survival
9.1% 2.7 months 7.4 months
P .001 P .02 –
26
68.0% 7.6 months Median survival not reached
McDermott DF, Sosman JA, Gonzalez R, et al. Double-blind randomized phase II study of the combination of sorafenib and dacarbazine in patients with advanced melanoma: a report from the 11715 Study Group. J Clin Oncol. 2008;26:2178–2185. 27 Cappuzzo F, Ligorio C, Jänne PA, et al. Prospective study of gefitinib in epidermal growth factor receptor fluorescence in situ hybridization-positive/phospho-Akt-positive or never smoker patients with advanced nonsmall-cell lung cancer: the ONCOBELL trial. J Clin Oncol. 2007;25:2248–2255.
245
246
Clinical Trials
Cumulative percentage of subjects experiencing progression
EGFR positive
EGFR negative Time
Figure 13.3 Schematic representation of Kaplan-Meier plot. This was a 1-arm study. All patients received gefitinib. The two curves show TTP for the subgroup of 25 patients with high EGFR (FISH positive), and TTP for the subgroup of 11 patients with low EGFR (FISH negative)
greater efficacy was demonstrated with the endpoints of objective response, TTP, and overall survival. The Cappuzzo study provides the take-home lesson that, to forestall data acquisition problems due to short follow-up times, it is wise to include endpoints that can be fully captured early in the clinical trial. Suitable endpoints that can be captured early in the trial include objective response (RECIST criteria) and TTP. A schematic representation of the Kaplan-Meier plot demonstrates the striking differences in TTP that were associated with the change in EGFR gene copy number (Fig. 13.3). The authors concluded that EGFR FISH analysis is an accurate predictor for efficacy with gefitinib therapy against lung cancer. To conclude, the Cappuzzo study demonstrates that EGFR is a useful biomarker. Other biomarkers, such as HER2 and KRAS, are described in this book in the chapter on biomarkers and personalized medicine.
IX. METHODOLOGY TIP – ADVANTAGE OF USING AN ENDPOINT THAT INCORPORATES A “MEDIAN” TIME The data of Cappuzzo et al. (28) as disclosed above, were incomplete for the endpoint of overall survival, as indicated in Table 13.4. As explained by the authors, “the median follow-up time was too short…for significant tests of differences in survival outcomes.” But one might ask how a clinical trial can be too short to calculate results of median overall survival, but not be too short to calculate median TTP. The following 28
Cappuzzo F, Ligorio C, Jänne PA, et al. Prospective study of gefitinib in epidermal growth factor receptor fluorescence in situ hybridization-positive/phospho-Akt-positive or never smoker patients with advanced nonsmall-cell lung cancer: the ONCOBELL trial. J Clin Oncol. 2007;25:2248–2255.
Oncology Endpoints: Time to Progression
concerns the median and the mean, which are statistical parameters used for calculating an average. Calculating the mean requires acquisition of all data points. Thus, in the context of a clinical trial, mean survival time calculations could require waiting decades until all subjects die. In contrast, median survival time calculations only require waiting until half of the subjects had reached the endpoint in question. The median is the middle observation, that is, the point at which half the observations are smaller and half are larger (29). In general, median TTP is reached at an earlier time in oncology clinical trials than median survival, because tumor progression is usually a prerequisite for death.
X. SUMMARY To summarize the material presented so far, the Cappuzzo et al. (30) study reveals that TTP may have an advantage over the endpoint of overall survival, where the trial has a relatively short duration. The McDermott et al. (31) study indicates that TTP may have an advantage over overall survival, in the situation where the number of subjects is small. The comments of Llovet et al. (32) reveal that TTP may have an advantage over the endpoint of overall survival, in the situation where deaths result from health factors other than the cancer. The Park et al. (33) study demonstrates the fact pattern where TTP may have an advantage over the endpoint of overall survival, in the situation where subjects receive second-line treatment.
XI. THYMIDINE PHOSPHORYLASE AS A BIOMARKER FOR SURVIVAL – THE MEROPOL STUDY In a study of colorectal cancer, Meropol et al. (34) administered the same chemotherapy to all patients. The treatment was irinotecan plus cepecitabine. Cepecitabine is distinguished in that it is converted in the cell to 5-fluorouracil. This conversion is
29
Dawson B, Trapp RG. Basic and Clinical Biostatistics. 4th ed. New York: Lange Medical Books; 2004:28. Cappuzzo F, Ligorio C, Jänne PA, et al. Prospective study of gefitinib in epidermal growth factor receptor fluorescence in situ hybridization-positive/phospho-Akt-positive or never smoker patients with advanced nonsmall-cell lung cancer: the ONCOBELL trial. J Clin Oncol. 2007;25:2248–2255. 31 McDermott DF, Sosman JA, Gonzalez R, et al. Double-blind randomized phase II study of the combination of sorafenib and dacarbazine in patients with advanced melanoma: a report from the 11715 Study Group. J Clin Oncol. 2008;26:2178–2185. 32 Llovet JM, Di Bisceglie AM, Bruix J, et al. Design and endpoints of clinical trials in hepatocellular carcinoma. J Natl Cancer Inst. 2008;100:698–711. 33 Park JO, Kim SW, Ahn JS, et al. Phase III trial of two versus four additional cycles in patients who are nonprogressive after two cycles of platinum-based chemotherapy in non small-cell lung cancer. J Clin Oncol. 2007;25:5233–5239. 34 Meropol NJ, Gold PJ, Diasio RB, et al. Thymidine phosphorylase expression is associated with response to capecitabine plus irinotecan in patients with metastatic colorectal cancer. J Clin Oncol. 2006;24:4069–4077.
30
247
248
Clinical Trials
Table 13.5 The Meropol study
TTP Overall survival
Primary tumor Metastatic tumor Primary tumor Metastatic tumor
Biopsy positive () for thymidine phosphorylase
Biopsy negative () for thymidine phosphorylase
8.7 months 8.7 months 28.2 months 26.2 months
6.0 months 5.4 months 14.9 months 9.8 months
catalyzed by thymidine phosphorylase. This enzyme is preferentially expressed by colorectal cancer cells, thus lending specificity of cepecitabine’s toxic effects to the cancer cells, rather than to normal tissues. The efficacy of capecitabine is similar to that of 5-fluorouracil. The toxicity of capecitabine is less than the toxicity of bolus 5-fluor ouracil, but similar to that with infusional 5-fluorouracil (35). The goal of the study was to assess the possible correlation of thymidine phosphorylase expression with outcome. Thus, this was a study of a predictive biomarker. The difference between a “predictive biomarker” and a “prognostic biomarker” is detailed in this book in the chapter on biomarkers and personalized medicine. The data demonstrated that increased expression of thymidine phosphorylase is an excellent predictor of increased TTP (Table 13.5). Also, the data demonstrated that increased expression was an excellent predictor of increased overall survival. The data on overall survival were especially dramatic. All patients (52 patients) provided biopsies of primary tumors. But at the time of diagnosis, only 30 of these patients had metastatic tumors. Most of the metastatic colorectal cancer tumors were located in the liver, lung, and lymph nodes. Hence, data from only 30 patients are available for correlating the biomarker expressing on metastatic tumors with the endpoints. Regarding the predictive value of thymidine phosphorylase found in the Meropol study, it has been suggested that the medical community should select capecitabine versus 5-fluorouracil rationally based on analysis of tumor thymidine phosphorylase levels using a simple test for enzyme expression, for example by measuring RNA levels by a PCR-based method, or protein levels by an antibody-based method (36). Meropol’s interesting data on the thymidine phosphorglase biomarker justify the following digressions on this biomarker, on the concept of synergy, and on the issue of mRNA expression versus polypeptide expression.
35
O’Neil BH, McLeod HL. Thymidine phosphorylase and capecitabine: a predictive marker for therapy selection? J Clin Oncol. 2006;24:4051–4053. 36 O’Neil BH, McLeod HL. Thymidine phosphorylase and capecitabine: a predictive marker for therapy selection? J Clin Oncol. 2006;24:4051–4053.
Oncology Endpoints: Time to Progression
XII. DRUG COMBINATIONS THAT INCLUDE CAPECITABINE The Meropol et al. (37) study administered the combination of irinotecan and cepecitabine in a clinical study of colorectal cancer. In a different clinical trial on colorectal cancer using a different drug combination (capecitabine plus oxiplatin), Petrioli et al. (38) also found that higher expression of thymidine phosphorylase is associated with a more favorable clinical response. The common use of drug combinations in oncology, and the frequent decision to change one of the drugs used in a twodrug combination therapy, raises the issue of synergy. In this context, synergy refers to an effect that is more than additive, as it applies to efficacy, and an effect that is more than additive, as it applies to toxicity. Aprile et al. (39) report that taxane drugs stimulate the expression of thymidine phosphorylase, and that this effect accounts for the increased anti-tumor activity of the combination of capecitabine and taxane. Kikuno et al. (40) report the induction of this enzyme, with use of the taxane, paclitaxel.
XIII. METHODOLOGY TIP – DO CHANGES IN mRNA EXPRESSION RESULT IN CORRESPONDING CHANGES IN EXPRESSION OF POLYPEPTIDE? Meropol et al. (41) measured expression of thymidine phosphorylase with immunological assays sensitive to the polypeptide, and with PCR-based assays sensitive to the expressed mRNA. As a general proposition, it is hoped that changes in mRNA expression correlate with changes in the polypeptide. But it must not be assumed that an increase in mRNA results in a corresponding increase in protein. Pennica et al. (42) Haynes et al. (43) Hu et al. (44) Oh et al. (45) Schantz and Pegg (46) and Anderson 37
Meropol NJ, Gold PJ, Diasio RB, et al. Thymidine phosphorylase expression is associated with response to capecitabine plus irinotecan in patients with metastatic colorectal cancer. J Clin Oncol. 2006;24:4069–4077. 38 Petrioli R, Bargagli G, Lazzi S, et al. Thymidine phosphorylase expression in metastatic sites is predictive for response in patients with colorectal cancer treated with continuous oral capecitabine and biweekly oxaliplatin. Anticancer Drugs. 2010;21:313–319. 39 Aprile G, Mazzer M, Moroso S, Puglisi F. Pharmacology and therapeutic efficacy of capecitabine: focus on breast and colorectal cancer. Anticancer Drugs. 2009;20:217–229. 40 Kikuno N, Moriyama-Gonda N,Yoshino T, et al. Blockade of paclitaxel-induced thymidine phosphorylase expression can accelerate apoptosis in human prostate cancer cells. Cancer Res. 2004;64:7526–7532. 41 Meropol NJ, Gold PJ, Diasio RB, et al. Thymidine phosphorylase expression is associated with response to capecitabine plus irinotecan in patients with metastatic colorectal cancer. J Clin Oncol. 2006;24:4069–4077. 42 Pennica D, Swanson TA, Welsh JW, et al. WISP genes are members of the connective tissue growth factor family that are up-regulated in wnt-1-transformed cells and aberrantly expressed in human colon tumors. Proc Natl Acad Sci USA. 1998;95:14717–14722. 43 Haynes PA, Gygi SP, Figeys D, Aebersold R. Proteome analysis: biological assay or data archive? Electrophoresis. 1998;19:1862–1871. 44 Hu Y, Hines LM, Weng H, Zuo D, Rivera M, Richardson A, et al. Analysis of genomic and proteomic data using advanced literature mining. J Proteome Res. 2003;2:405–412. 45 Oh JM, Brichory F, Puravs E, et al. A database of protein expression in lung cancer. Proteomics. 2001;1:1303–1319. 46 Shantz LM, Pegg AE. Translational regulation of ornithine decarboxylase and other enzymes of the polyamine pathway. Int J Biochem Cell Biol. 1999;31:107–122.
249
250
Clinical Trials
and Seilhamer (47) all address the issue of whether or not increased expression of polypeptide correlates with increased mRNA expression. Fu et al. (48) for example, demonstrate the scenario where there is no correlation at all. These publications can be used as a basis for arguing that increased mRNA will likely lead to an increase in protein or, conversely, for arguing that increased mRNA does not necessarily lead to an increase in protein (49).
XIV. CONCLUSIONS The endpoint of TTP does not take into account the survival time. Time to progression is not a measure of survival for any given subject, or for the study population as a whole. As an endpoint, TTP has advantages over endpoints that take survival time into account. An advantage is that data on TTP can be captured at a time earlier than data on overall survival, thus enabling the investigator to arrive at conclusions regarding efficacy before the trial is formally concluded. This advantage is especially important where a significant number of subjects drop out of the trial before they trigger the endpoint of survival (death). A related advantage occurs in the situation where study subjects leaving the trial receive second-line therapy. Similarly, use of TTP as an endpoint, rather than survival, is advantageous where death arises from non-cancer-related causes such as liver cirrhosis. Also note that it is better for a clinical trial to include more endpoints than fewer endpoints, as it cannot be known ahead of time which endpoint will give the most statistically significant results.
47
Anderson L, Seilhamer J. A comparison of selected mRNA and protein abundances in human liver. Electrophoresis. 1997;18:533–537. 48 Fu L, Minden MD, Benchimol S. Translational regulation of human p53 gene expression. EMBO J. 1996;15:4392–4401. 49 See, for example, file histories of U.S. Patent Nos. 7,008,799; 7,144,990; and 7,230,076. These file histories contain the arguments by researchers, and are available from the PUBLIC PAIR device found at www.uspto.gov. The term “file history” refers to the collection of rejections against the claims imposed by the patent examiner and the rebuttals submitted by the inventor or by the inventor’s attorney.
13
Oncology Endpoints: Time to Progression I. INTRODUCTION The present chapter continues with oncology endpoints, and covers the fourth topic indicated by the check mark. Objective response Overall survival Progression-free survival Time to progression ✓ Disease-free survival Time to distant metastasis. Time to progression (TTP), also known as time to disease progression, is a common endpoint in oncology clinical trials. TTP and progression-free survival (PFS) are defined by FDA’s Guidance for Industry (1) as follows: l
l
l
l
l
l
TTP and PFS have served as primary endpoints for drug approval.TTP is defined as the time from randomizationuntilobjectivetumorprogression;TTPdoesnotincludedeaths.PFSisdefinedas thetimefromrandomizationuntilobjectivetumorprogressionordeath.Theprecisedefinitionof tumor progression is important and should be carefully detailed in the protocol.
But the values for TTP and PFS are often almost identical, as revealed in the following examples from several clinical trials of breast cancer (2). In one of these trials, the values of TTP and PFS, for one particular arm, were both 5.5 months (3). In another trial, the values for TTP and PFS, for one of the arms, were both 6.7 months (4). In yet another trial, the values for TTP and PFS, for one of the arms, were identical,
1
U.S. Dept. of Health and Human Services. Food and Drug Administration. Guidance for Industry. Clinical trial endpoints for the approval of cancer drugs and biologics; April 2005. 2 Burzykowski T, Buyse M, Piccart-Gebhart MJ, et al. Evaluation of tumor response, disease control, progressionfree survival, and time to progression as potential surrogate end points in metastatic breast cancer. J Clin Oncol. 2008;26:1987–1992. 3 Bonneterre J, Dieras V, Tubiana-Hulin M, et al. Phase II multicenter randomized study of deocetaxel plus epirubicin versus fluorouracil plus epirubicin and cyclophosphamide in metastatic breast cancer. Br J Cancer. 2004;91:1466–1471. 4 Bontenbal M, Creemers GJ, Braun HJ, et al. Phase II to III study comparing doxorubicin and docetaxel with fluorouracil, doxorubicin, and cyclophosphamide as first-line chemotherapy in patients with metastatic breast cancer: results of a Dutch community setting trial for the clinical trial group of the Comprehensive Cancer Centre. J Clin Oncol. 2005;23:7081–7088.
Clinical Trials DOI: 10.1016/B978-0-12-391911-3.00013-X
© 2012 Elsevier Inc. All rights reserved.
237
238
Clinical Trials
Table 13.1 The Paccagnella study Arm A (PC arm) (paclitaxel carboplatin)
Objective response TTP Overall survival
20.2% 5.1 months 8.3 months
Arm B (PCG arm) (paclitaxel carboplatin gemcitabine)
Significance of difference between the two study arms
43.6% 7.6 months 10.8 months
P .0001 P .012 P .032
7.4 months (5). Moreover, in another trial the values for PFS and TTP were both 18.4 weeks (6). Please note that TTP does not, but PFS does, use survival time as a trigger for the endpoint. As a matter of first impression, it might be asked, “What is the use of having both types of endpoints?”
II. AGREEMENT OF RESULTS FROM OBJECTIVE RESPONSE, TIME TO PROGRESSION, AND OVERALL SURVIVAL – THE PACCAGNELLA STUDY The following clinical trial provides a straightforward account of various endpoints. In a clinical trial of non-small cell lung cancer (NSCLC), Paccagnella et al. (7) used three endpoints, objective response, TTP, and overall survival. The trial had two arms, as indicated in Table 13.1. Arm A. Paclitaxel and carboplatin (PC) Arm B. Paclitaxel, carboplatin, and gemcitabine (PCG). The results are straightforward (Table 13.1). The triple drug combination was superior, in terms of objective response (WHO criteria), median TTP, and median overall survival. It might be pointed out that the significance (P value) was greater for the endpoint of TTP than for overall survival, indicating an advantage in using TTP as an endpoint. l
l
III. CAN THE VALUE FOR PFS BE LESS THAN THE VALUE FOR TTP? PFS and TTP often have values that are nearly identical. Hence, it might be asked what might cause these values to differ, for example what might cause the value of PFS to 5
Paridaens R, Biganzoli L, Bruning P, et al. Paclitaxel versus doxorubicin as first-line single agent chemotherapy for metastatic breast cancer: a European Organization for Research and Treatment of Cancer randomized study with cross-over. J Clin Oncol. 2000;18:724–733. 6 Brandes AA, Basso U, Reni M, et al. First-line chemotherapy with cisplatin plus fractionated temozolomide in recurrent glioblastoma multiforme: a phase II study of the Gruppo Italiano Cooperativo di Neuro-Oncologia. J Clin Oncol. 2004;22:1598–1604. 7 Paccagnella A, Oniga F, Bearz A, et al. Adding gemcitabine to paclitaxel/carboplatin combination increases survival in advanced non-small-cell lung cancer: results of a phase II–III study. J Clin Oncol. 2006;24:681–687.
Oncology Endpoints: Time to Progression
be less than TTP. The situation where PFS is less than the value for TTP is one where deaths due to drug toxicity (but not from the cancer) occur early on in the trial. This is, “when deaths due to the disease are rare, but those due to the drug are not” (8). Examples of this are likely to be found in clinical trials on the elderly, especially where radiation therapy is followed by chemotherapy (9). It might also be asked when the value for PFS will likely be equal to the value for TTP. PFS will likely be equal to TTP when, at early times in the clinical trial, the tumors in most patients grow to the point where they trigger the RECIST criteria for tumor progression, but where the rate of patient deaths, due to any cause, is low.
IV. TIME TO PROGRESSION MAY BE THE PREFERRED ENDPOINT WHERE, ONCE THE TRIAL IS CONCLUDED, PATIENTS RECEIVE ADDITIONAL CHEMOTHERAPY – THE PARK STUDY In a clinical trial of non-small cell lung cancer (NSCLC), Park et al. (10) used the endpoints of objective response, TTP, overall survival, as well as endpoints set forth in health-related quality of life (HRQoL) questionnaires. Objective response was determined by the WHO criteria. Before randomizing the subjects, all subjects were treated with two cycles of chemotherapy in a run-in period. Only subjects not showing disease progression were randomized to the two arms. Regarding the run-in period, the investigators stated that one reason for including the run-in was to enrich the study with a more homogeneous patient population. This particular rationale is further described in this textbook in the chapter on run-in periods. The Park study had two arms, as follows: Arm A. Six cycles of chemotherapy Arm B. Four cycles of chemotherapy. Patients were monitored for a total of 18 months. Objective response was measured at 3-month intervals by computed tomography. According to TTP, arm A fared better than arm B, as shown in Table 13.2. According to overall survival, there was no significant difference between arms A and B. The investigators were careful to describe reasons for leaving the trial, writing that “[t]reatment was continued until the maximum of four or six cycles was completed, depending on random assignment, unless disease progression or unacceptable toxicity occurred or unless the patient refused further chemotherapy.” Where tumor size or number increased to the point where the endpoint of progression was triggered, treatment with study drugs was discontinued, and the subjects had the option l
l
8
Burzykowski T. E-mail of September 12, 2010. Buyse M. E-mail of September 12, 2010. 10 Park JO, Kim SW, Ahn JS, et al. Phase III trial of two versus four additional cycles in patients who are nonprogressive after two cycles of platinum-based chemotherapy in non small-cell lung cancer. J Clin Oncol. 2007;25:5233–5239. 9
239
240
Clinical Trials
Table 13.2 The Park study
Arm A (6 cycles) Arm B (4 cycles) Significance of difference between arms A and B
TTP (months; median value)
Overall survival (months; median value)
6.2 months 4.6 months P .001
14.9 months 15.9 months P .461
of second-line therapy. As stated, “[s]econd-line chemotherapy was considered at the discretion of the treating oncologist after documentation of progression.” To clarify this point, it is often the case in trial design that where the subject experiences progression (the tumor size or number reaches or passes a limit set for by the RECIST or WHO criteria), treatment with study drug or control is halted, and the subject has the option of receiving second-line treatment. At this point, it is intuitively obvious that this second-line chemotherapy cannot influence the date of progression (for any given patient), but it can influence the survival time. Regarding the finding that the 6-cycle group fared better than the 4-cycle group, in terms of the TTP endpoint, but that there was little difference in the endpoints of overall survival, the authors believed that this discrepancy was likely due to confounding of the survival times by the second-line therapy. In the authors’ words, “[t]he main reason why the TTP benefit did not translate into the survival benefit probably involved the dilution effect of the second-line chemotherapy.” The takehome lesson is that clinical trials should include the endpoint of TTP, where it is expected that many patients will be receiving second-line treatments.
V. THE ENDPOINT OF TTP MAY BE PREFERRED OVER SURVIVAL ENDPOINTS, WHERE DEATHS RESULT FROM CAUSES OTHER THAN CANCER – THE LLOVET STUDY Time to progression may be the preferred endpoint over an endpoint that takes into account survival data, such as the endpoints of PFS, disease-free survival (DFS), or overall survival. This type of preference for TTP is the case where the major cause of death, during the course of the clinical trial, is unrelated to the cancer. The following provides the example of liver cancer, that is, hepatocellular carcinoma (HCC). According to Llovet et al. (11) “although disease- and progression-free survival are appropriate endpoints in other solid tumors, they are particularly unreliable endpoints in HCC research because death resulting from the natural history of cirrhosis might confound detection of potential benefits from effective drugs. That is, a type 11
Llovet JM, Di Bisceglie AM, Bruix J, et al. Design and endpoints of clinical trials in hepatocellular carcinoma. J Natl Cancer Inst. 2008;100:698–711.
Oncology Endpoints: Time to Progression
II error might result from using progression-free survival as an endpoint in a suboptimal population in early phases of drug development.” Consistently, Dancey et al. (12) stated that, “deaths prior to disease progression are not correlated with progression. In situations where many deaths are unrelated to cancer or treatment, TTP can be an acceptable and preferred end-point.” The term Type II error refers to the conclusion that the drug does not work when in fact it actually does work (13). The term Type I error refers to the conclusion that the drug works when in fact it really fails to work. A related problem is that, during the course of a clinical trial, the cause of death is often not reliably collected or documented. Typically, deaths occurring after disease progression are considered malignant, while all deaths occurring prior to disease progression are considered non-malignant (14). The distinction between drug-induced deaths and disease-induced deaths is not often made. In the context of liver cancer, the endpoint of overall survival is expected to be reliable, where examination of study subjects shows that cirrhosis is not a problem. Thus, “[i]n the exceptional circumstances in which these [survival] endpoints are applied, a restrictive selection of patients with well-preserved liver function is recommended to minimize the impact of death unrelated with tumor progression” (15). In a study of hepatocellular carcinoma, Llovet et al. (16) treated subjects with either sorafenib or placebo. The study was designed to capture information on drug efficacy while confounding effects of deaths unrelated to cancer. Since hepatocellular carcinoma occurs mainly in patients with cirrhosis, the inclusion/exclusion criteria required that subjects have well-preserved liver function. However, it was still the case that endpoints tied to survival, such as progression-free survival (PFS), might be suboptimal because of the confusing effect of the cirrhosis. Thus, the Llovet study provides a general lesson regarding the selection of inclusion/exclusion criteria. In the sorafenib group, 7 patients (2%) had a partial response and 211 (71%) had stable disease (according to RECIST), whereas in the placebo group, 2 patients (1%) had a partial response and 204 (67%) had stable disease. There were no complete responses in either group (Fig. 13.1).
12
Dancey JE, Dodd LE, Ford R, et al. Recommendations for the assessment of progression in randomised cancer treatment trials. Eur J Cancer. 2009;45:281–289. 13 Type I error is rejecting the null hypothesis when it is true and Type II error is accepting the null hypothesis when it is false. See, e.g., Biau DJ, Jolles BM, Porcher R. P value and the theory of hypothesis testing: an explanation for new researchers. Clin Orthop Relat Res. 2010;468:885–892. 14 Buyse M. E-mail of September 20, 2010. 15 Llovet JM, Di Bisceglie AM, Bruix J, et al. Design and endpoints of clinical trials in hepatocellular carcinoma. J Natl Cancer Inst. 2008;100:698–711. 16 Llovet JM, Ricci S, Mazzaferro V, et al. Sorafenib in advanced hepatocellular carcinoma. New Engl J Med. 2008;359:378–390.
241
242
Clinical Trials
Figure 13.1 Kaplan-Meier plot of time to progression. Progression was defined by the RECIST criteria. A comparison of arm A (sorafenib) and arm B (placebo) reveals that the study drug was effective against the cancer. Median TTP was 5.5 months (arm A) and 2.8 months (arm B)
Median TTP was 5.5 months in the sorafenib arm and 2.8 months in the placebo arm (HR 0.58; P .001), while median overall survival was 10.7 months in the sorafenib arm and 7.9 months in the placebo arm (HR 0.69; P .001). This demonstrable efficacy of sorafenib was considered to be a therapeutic breakthrough, in view of the inevitably fatal outcome of this type of cancer, and lack of any effective treatment (Fig. 13.2).
VI. THE ENDPOINT OF OVERALL SURVIVAL MAY BE PREFERRED OVER OBJECTIVE RESPONSE OR OVER TTP, WHERE THE DRUG IS CLASSED AS A CYTOSTATIC DRUG – THE LLOVET STUDY Llovet et al. (17) provide a lesson regarding the utility, or actually lack thereof, of objective response, where the drug is cytostatic, rather than cytotoxic. The Llovet study compares the endpoint of objective response with the endpoint of overall survival. The study drug, sorafenib, inhibits angiogenesis and delays the progression of liver tumors in hepatocellular carcinoma. In the Llovet study, response consisted of complete response, partial response, or stable disease, according to the RECIST criteria. Where a drug is classed as a cytostatic drug, endpoints that rely entirely on tumor size or number, such as objective response and TTP, might not do justice to the survival benefit of the drug. With a cytostatic drug, a more reasonable assessment of the 17
Llovet JM, Ricci S, Mazzaferro V, et al. Sorafenib in advanced hepatocellular carcinoma. New Engl J Med. 2008;359:378–390.
Oncology Endpoints: Time to Progression
Figure 13.2 Kaplan-Meier plot of overall survival. A comparison of arm A (sorafenib) and arm B (placebo) reveals that the study drug was effective against the cancer. Median overall survival was 10.7 months (arm A) and 7.9 months (arm B)
drug’s efficacy might be acquired by using endpoints that rely partially or entirely on survival time. In the words of Llovet et al. (18) “current targeted agents may act as cytostatic agents…and possibly improve survival with no measurable change in tumor size.” Thus overall survival might be a better endpoint than endpoints that derived from objective response, such as the endpoint of objective response itself, or the endpoint of TTP, when the drug is classified as one that is cytostatic (not cytotoxic). The Llovet study on sorafenib concluded that the incidence of objective response was relatively low, and the effects on overall survival were high. Consistent with this interpretation, a review article observed that sorafenib does not commonly improve objective response, though it does stabilize cancer and enhances survival (19). Regarding cytostatic drugs, Fleming et al. (20) have reiterated the point that endpoints of objective response and time to progression (TTP) are relatively insensitive to the effects of cytostatic drugs, and that the efficacy of this class of drugs can more effectively be captured by the endpoint of progression-free survival (PFS) or overall survival. Another way around the problem of objective response is to change the technology used for measuring objective response. Instead of just using techniques that measure 18
Llovet JM, Di Bisceglie AM, Bruix J, et al. Design and endpoints of clinical trials in hepatocellular carcinoma. J Natl Cancer Inst. 2008;100:698–711. 19 Almhanna K, Philip PA. Safety and efficacy of sorafenib in the treatment of hepatocellular carcinoma. Onco Targets Ther. 2009;2:261–267. 20 Fleming TR, Rothmann MD, Lu HL. Issues in using progression-free survival when evaluating oncology products. J Clin Oncol. 2009;27:2874–2880.
243
244
Clinical Trials
the dimensions of tumors, techniques of functional imaging such as positron emission tomography (PET) can be used to assess the metabolism of existing tumors (21,22,23).
VII. TIME TO PROGRESSION MAY SHOW EFFICACY, WHERE THE ENDPOINT OF OVERALL SURVIVAL FAILS TO SHOW EFFICACY, WHERE THE NUMBER OF SUBJECTS IS SMALL – THE MCDERMOTT STUDY In a study of melanoma, McDermott et al. (24) used the endpoints of objective response (RECIST criteria), TTP, PFS, and overall survival. There were two study arms. This study used an add-on design. In other words, subjects in one arm received dacarbazine plus placebo, while subjects in the other arm received dacarbazine plus sorafenib. At the time of the clinical trial, dacarbazine was the most commonly used drug for melanoma. In part for ethical reasons, both study arms received dacarbazine. Sorafenib was the add-on drug. Arm A. Dacarbazine plus placebo Arm B. Dacarbazine plus sorafenib. The results are shown in Table 13.3. The study drug worked better than the placebo, as measured by the endpoints of TTP and PFS. However, there was no difference in terms of the endpoints of overall survival. The data provide an instructive account of the use of hazard ratio. The authors stated that, in terms of overall survival, there was “no difference.” The “no difference” result has a hazard ratio of 1.022, which is consistent with the teachings in the statistics chapter in this book. Spruance et al. (25) remind us that the hazard ratio can be greater than 1.0, or less than 1.0, depending on whether it is asked if treatment A works better than treatment B, or if treatment B works better than treatment A. The authors pointed out that the study drug was better than placebo, according to the endpoints of TTP and PFS, but explained the lack of effect in overall survival, by referring to the small number of subjects. In the words of the authors, “[r]andomized l
l
Table 13.3 The McDermott study Arm A. Placebo
Arm B. Sorafenib
Hazard ratio
P value
TTP PFS Overall survival
21.1 weeks 21.1 weeks 51.3
0.619 0.665 1.022
.03 .068 .927
21
11.7 weeks 11.7 weeks 45.6
Avril N, Propper D. Functional PET imaging in cancer drug development. Future Oncol. 2007;3:215–228. Kyle F, Spicer J. Targeted therapies in non-small cell lung cancer. Cancer Imaging. 2008;8:199–205. 23 Weber WA. Positron emission tomography as an imaging biomarker. J Clin Oncol. 2006;24:3282–3292. 24 McDermott DF, Sosman JA, Gonzalez R, et al. Double-blind randomized phase II study of the combination of sorafenib and dacarbazine in patients with advanced melanoma: a report from the 11715 Study Group. J Clin Oncol. 2008;26:2178–2185. 25 Spruance SL, Reid JE, Grace M, Samore M. Hazard ratio in clinical trials. Antimicrob Agents Chemother. 2004;48:2787–2792. 22
Oncology Endpoints: Time to Progression
phase II clinical trials are often limited by their small sample sizes to detect true treatment differences between the study arms. In this study, we observed improvements in PFS and TTP…however, these findings did not translate into an improvement in OS” (26).
VIII. TIME TO PROGRESSION MAY SHOW EFFICACY, WHERE THE ENDPOINT OF OVERALL SURVIVAL FAILED TO SHOW EFFICACY, WHERE THE DURATION OF THE TRIAL WAS TOO SHORT – THE CAPPUZZO STUDY The following study on non-small cell lung cancer (NSCLC) used the endpoints of objective response, TTP, and overall survival. Cappuzzo et al. (27) administered the same drug (gefitinib) to all study subjects. Gefitinib inhibits enzymatic activity of a particular kinase, namely, epidermal growth factor receptor (EGFR) tyrosine kinase. Cytogenetic assays were conducted on all study subjects prior to administering the drug, to determine the number of EGFR genes in tumor cells. Tumor cells were acquired via biopsies, and the number of copies of the EGFR gene in a sampling of tumor cells was determined by way of the fluorescent in situ hybridization (FISH) technique. Twenty-five of the patients were FISH positive for EGFR. Eleven patients were FISH negative for EGFR. “FISH positive” was defined as tumor cells carrying four or more copies of the EGFR gene in 40% of the tumor cells. “FISH negative” was defined as tumor cells carrying four or more copies of the EGFR gene in less than 40% of the tumor cells. Table 13.4 demonstrates the greater efficacy of gefitinib in patients bearing EGFR positive tumor cells (as compared to patients with EGFR negative tumor cells). This Table 13.4 The Cappuzzo study Patients with EGFR positive tumors
Patients with EGFR negative tumors
Significance of difference of endpoint, comparing EGFR and EGFR patients
Objective response TTP Overall survival
9.1% 2.7 months 7.4 months
P .001 P .02 –
26
68.0% 7.6 months Median survival not reached
McDermott DF, Sosman JA, Gonzalez R, et al. Double-blind randomized phase II study of the combination of sorafenib and dacarbazine in patients with advanced melanoma: a report from the 11715 Study Group. J Clin Oncol. 2008;26:2178–2185. 27 Cappuzzo F, Ligorio C, Jänne PA, et al. Prospective study of gefitinib in epidermal growth factor receptor fluorescence in situ hybridization-positive/phospho-Akt-positive or never smoker patients with advanced nonsmall-cell lung cancer: the ONCOBELL trial. J Clin Oncol. 2007;25:2248–2255.
245
246
Clinical Trials
Cumulative percentage of subjects experiencing progression
EGFR positive
EGFR negative Time
Figure 13.3 Schematic representation of Kaplan-Meier plot. This was a 1-arm study. All patients received gefitinib. The two curves show TTP for the subgroup of 25 patients with high EGFR (FISH positive), and TTP for the subgroup of 11 patients with low EGFR (FISH negative)
greater efficacy was demonstrated with the endpoints of objective response, TTP, and overall survival. The Cappuzzo study provides the take-home lesson that, to forestall data acquisition problems due to short follow-up times, it is wise to include endpoints that can be fully captured early in the clinical trial. Suitable endpoints that can be captured early in the trial include objective response (RECIST criteria) and TTP. A schematic representation of the Kaplan-Meier plot demonstrates the striking differences in TTP that were associated with the change in EGFR gene copy number (Fig. 13.3). The authors concluded that EGFR FISH analysis is an accurate predictor for efficacy with gefitinib therapy against lung cancer. To conclude, the Cappuzzo study demonstrates that EGFR is a useful biomarker. Other biomarkers, such as HER2 and KRAS, are described in this book in the chapter on biomarkers and personalized medicine.
IX. METHODOLOGY TIP – ADVANTAGE OF USING AN ENDPOINT THAT INCORPORATES A “MEDIAN” TIME The data of Cappuzzo et al. (28) as disclosed above, were incomplete for the endpoint of overall survival, as indicated in Table 13.4. As explained by the authors, “the median follow-up time was too short…for significant tests of differences in survival outcomes.” But one might ask how a clinical trial can be too short to calculate results of median overall survival, but not be too short to calculate median TTP. The following 28
Cappuzzo F, Ligorio C, Jänne PA, et al. Prospective study of gefitinib in epidermal growth factor receptor fluorescence in situ hybridization-positive/phospho-Akt-positive or never smoker patients with advanced nonsmall-cell lung cancer: the ONCOBELL trial. J Clin Oncol. 2007;25:2248–2255.
Oncology Endpoints: Time to Progression
concerns the median and the mean, which are statistical parameters used for calculating an average. Calculating the mean requires acquisition of all data points. Thus, in the context of a clinical trial, mean survival time calculations could require waiting decades until all subjects die. In contrast, median survival time calculations only require waiting until half of the subjects had reached the endpoint in question. The median is the middle observation, that is, the point at which half the observations are smaller and half are larger (29). In general, median TTP is reached at an earlier time in oncology clinical trials than median survival, because tumor progression is usually a prerequisite for death.
X. SUMMARY To summarize the material presented so far, the Cappuzzo et al. (30) study reveals that TTP may have an advantage over the endpoint of overall survival, where the trial has a relatively short duration. The McDermott et al. (31) study indicates that TTP may have an advantage over overall survival, in the situation where the number of subjects is small. The comments of Llovet et al. (32) reveal that TTP may have an advantage over the endpoint of overall survival, in the situation where deaths result from health factors other than the cancer. The Park et al. (33) study demonstrates the fact pattern where TTP may have an advantage over the endpoint of overall survival, in the situation where subjects receive second-line treatment.
XI. THYMIDINE PHOSPHORYLASE AS A BIOMARKER FOR SURVIVAL – THE MEROPOL STUDY In a study of colorectal cancer, Meropol et al. (34) administered the same chemotherapy to all patients. The treatment was irinotecan plus cepecitabine. Cepecitabine is distinguished in that it is converted in the cell to 5-fluorouracil. This conversion is
29
Dawson B, Trapp RG. Basic and Clinical Biostatistics. 4th ed. New York: Lange Medical Books; 2004:28. Cappuzzo F, Ligorio C, Jänne PA, et al. Prospective study of gefitinib in epidermal growth factor receptor fluorescence in situ hybridization-positive/phospho-Akt-positive or never smoker patients with advanced nonsmall-cell lung cancer: the ONCOBELL trial. J Clin Oncol. 2007;25:2248–2255. 31 McDermott DF, Sosman JA, Gonzalez R, et al. Double-blind randomized phase II study of the combination of sorafenib and dacarbazine in patients with advanced melanoma: a report from the 11715 Study Group. J Clin Oncol. 2008;26:2178–2185. 32 Llovet JM, Di Bisceglie AM, Bruix J, et al. Design and endpoints of clinical trials in hepatocellular carcinoma. J Natl Cancer Inst. 2008;100:698–711. 33 Park JO, Kim SW, Ahn JS, et al. Phase III trial of two versus four additional cycles in patients who are nonprogressive after two cycles of platinum-based chemotherapy in non small-cell lung cancer. J Clin Oncol. 2007;25:5233–5239. 34 Meropol NJ, Gold PJ, Diasio RB, et al. Thymidine phosphorylase expression is associated with response to capecitabine plus irinotecan in patients with metastatic colorectal cancer. J Clin Oncol. 2006;24:4069–4077.
30
247
248
Clinical Trials
Table 13.5 The Meropol study
TTP Overall survival
Primary tumor Metastatic tumor Primary tumor Metastatic tumor
Biopsy positive () for thymidine phosphorylase
Biopsy negative () for thymidine phosphorylase
8.7 months 8.7 months 28.2 months 26.2 months
6.0 months 5.4 months 14.9 months 9.8 months
catalyzed by thymidine phosphorylase. This enzyme is preferentially expressed by colorectal cancer cells, thus lending specificity of cepecitabine’s toxic effects to the cancer cells, rather than to normal tissues. The efficacy of capecitabine is similar to that of 5-fluorouracil. The toxicity of capecitabine is less than the toxicity of bolus 5-fluor ouracil, but similar to that with infusional 5-fluorouracil (35). The goal of the study was to assess the possible correlation of thymidine phosphorylase expression with outcome. Thus, this was a study of a predictive biomarker. The difference between a “predictive biomarker” and a “prognostic biomarker” is detailed in this book in the chapter on biomarkers and personalized medicine. The data demonstrated that increased expression of thymidine phosphorylase is an excellent predictor of increased TTP (Table 13.5). Also, the data demonstrated that increased expression was an excellent predictor of increased overall survival. The data on overall survival were especially dramatic. All patients (52 patients) provided biopsies of primary tumors. But at the time of diagnosis, only 30 of these patients had metastatic tumors. Most of the metastatic colorectal cancer tumors were located in the liver, lung, and lymph nodes. Hence, data from only 30 patients are available for correlating the biomarker expressing on metastatic tumors with the endpoints. Regarding the predictive value of thymidine phosphorylase found in the Meropol study, it has been suggested that the medical community should select capecitabine versus 5-fluorouracil rationally based on analysis of tumor thymidine phosphorylase levels using a simple test for enzyme expression, for example by measuring RNA levels by a PCR-based method, or protein levels by an antibody-based method (36). Meropol’s interesting data on the thymidine phosphorglase biomarker justify the following digressions on this biomarker, on the concept of synergy, and on the issue of mRNA expression versus polypeptide expression.
35
O’Neil BH, McLeod HL. Thymidine phosphorylase and capecitabine: a predictive marker for therapy selection? J Clin Oncol. 2006;24:4051–4053. 36 O’Neil BH, McLeod HL. Thymidine phosphorylase and capecitabine: a predictive marker for therapy selection? J Clin Oncol. 2006;24:4051–4053.
Oncology Endpoints: Time to Progression
XII. DRUG COMBINATIONS THAT INCLUDE CAPECITABINE The Meropol et al. (37) study administered the combination of irinotecan and cepecitabine in a clinical study of colorectal cancer. In a different clinical trial on colorectal cancer using a different drug combination (capecitabine plus oxiplatin), Petrioli et al. (38) also found that higher expression of thymidine phosphorylase is associated with a more favorable clinical response. The common use of drug combinations in oncology, and the frequent decision to change one of the drugs used in a twodrug combination therapy, raises the issue of synergy. In this context, synergy refers to an effect that is more than additive, as it applies to efficacy, and an effect that is more than additive, as it applies to toxicity. Aprile et al. (39) report that taxane drugs stimulate the expression of thymidine phosphorylase, and that this effect accounts for the increased anti-tumor activity of the combination of capecitabine and taxane. Kikuno et al. (40) report the induction of this enzyme, with use of the taxane, paclitaxel.
XIII. METHODOLOGY TIP – DO CHANGES IN mRNA EXPRESSION RESULT IN CORRESPONDING CHANGES IN EXPRESSION OF POLYPEPTIDE? Meropol et al. (41) measured expression of thymidine phosphorylase with immunological assays sensitive to the polypeptide, and with PCR-based assays sensitive to the expressed mRNA. As a general proposition, it is hoped that changes in mRNA expression correlate with changes in the polypeptide. But it must not be assumed that an increase in mRNA results in a corresponding increase in protein. Pennica et al. (42) Haynes et al. (43) Hu et al. (44) Oh et al. (45) Schantz and Pegg (46) and Anderson 37
Meropol NJ, Gold PJ, Diasio RB, et al. Thymidine phosphorylase expression is associated with response to capecitabine plus irinotecan in patients with metastatic colorectal cancer. J Clin Oncol. 2006;24:4069–4077. 38 Petrioli R, Bargagli G, Lazzi S, et al. Thymidine phosphorylase expression in metastatic sites is predictive for response in patients with colorectal cancer treated with continuous oral capecitabine and biweekly oxaliplatin. Anticancer Drugs. 2010;21:313–319. 39 Aprile G, Mazzer M, Moroso S, Puglisi F. Pharmacology and therapeutic efficacy of capecitabine: focus on breast and colorectal cancer. Anticancer Drugs. 2009;20:217–229. 40 Kikuno N, Moriyama-Gonda N,Yoshino T, et al. Blockade of paclitaxel-induced thymidine phosphorylase expression can accelerate apoptosis in human prostate cancer cells. Cancer Res. 2004;64:7526–7532. 41 Meropol NJ, Gold PJ, Diasio RB, et al. Thymidine phosphorylase expression is associated with response to capecitabine plus irinotecan in patients with metastatic colorectal cancer. J Clin Oncol. 2006;24:4069–4077. 42 Pennica D, Swanson TA, Welsh JW, et al. WISP genes are members of the connective tissue growth factor family that are up-regulated in wnt-1-transformed cells and aberrantly expressed in human colon tumors. Proc Natl Acad Sci USA. 1998;95:14717–14722. 43 Haynes PA, Gygi SP, Figeys D, Aebersold R. Proteome analysis: biological assay or data archive? Electrophoresis. 1998;19:1862–1871. 44 Hu Y, Hines LM, Weng H, Zuo D, Rivera M, Richardson A, et al. Analysis of genomic and proteomic data using advanced literature mining. J Proteome Res. 2003;2:405–412. 45 Oh JM, Brichory F, Puravs E, et al. A database of protein expression in lung cancer. Proteomics. 2001;1:1303–1319. 46 Shantz LM, Pegg AE. Translational regulation of ornithine decarboxylase and other enzymes of the polyamine pathway. Int J Biochem Cell Biol. 1999;31:107–122.
249
250
Clinical Trials
and Seilhamer (47) all address the issue of whether or not increased expression of polypeptide correlates with increased mRNA expression. Fu et al. (48) for example, demonstrate the scenario where there is no correlation at all. These publications can be used as a basis for arguing that increased mRNA will likely lead to an increase in protein or, conversely, for arguing that increased mRNA does not necessarily lead to an increase in protein (49).
XIV. CONCLUSIONS The endpoint of TTP does not take into account the survival time. Time to progression is not a measure of survival for any given subject, or for the study population as a whole. As an endpoint, TTP has advantages over endpoints that take survival time into account. An advantage is that data on TTP can be captured at a time earlier than data on overall survival, thus enabling the investigator to arrive at conclusions regarding efficacy before the trial is formally concluded. This advantage is especially important where a significant number of subjects drop out of the trial before they trigger the endpoint of survival (death). A related advantage occurs in the situation where study subjects leaving the trial receive second-line therapy. Similarly, use of TTP as an endpoint, rather than survival, is advantageous where death arises from non-cancer-related causes such as liver cirrhosis. Also note that it is better for a clinical trial to include more endpoints than fewer endpoints, as it cannot be known ahead of time which endpoint will give the most statistically significant results.
47
Anderson L, Seilhamer J. A comparison of selected mRNA and protein abundances in human liver. Electrophoresis. 1997;18:533–537. 48 Fu L, Minden MD, Benchimol S. Translational regulation of human p53 gene expression. EMBO J. 1996;15:4392–4401. 49 See, for example, file histories of U.S. Patent Nos. 7,008,799; 7,144,990; and 7,230,076. These file histories contain the arguments by researchers, and are available from the PUBLIC PAIR device found at www.uspto.gov. The term “file history” refers to the collection of rejections against the claims imposed by the patent examiner and the rebuttals submitted by the inventor or by the inventor’s attorney.