End Points for Adjuvant Therapy Trials: Has the Time Come to Accept Disease-Free Survival as a Surrogate End Point for Overall Survival?
http://www.100md.com
《中华首席医学网》
a British Columbia Cancer Agency, Vancouver, British Columbia, Canada; b Mayo Clinic Cancer Center, Rochester, Minnesota, USA
Key Words. Adjuvant therapy ; Overall survival ; Disease-free survival
ABSTRACT
The intent of adjuvant therapy is to eradicate micro-metastatic residual disease following curative resection with the goal of preventing or delaying recurrence. The time-honored standard for demonstrating efficacy of new adjuvant therapies is an improvement in overall survival (OS). This typically requires phase III trials of large sample size with lengthy follow-up. With the intent of reducing the cost and time of completing such trials, there is considerable interest in developing alternative or surrogate end points. A surrogate end point may be employed as a substitute to directly assess the effects of an intervention on an already accepted clinical end point such as mortality. When used judiciously, surrogate end points can accelerate the evaluation of new therapies, resulting in the more timely dissemination of effective therapies to patients.
The current review provides a perspective on the suitability and validity of disease-free survival (DFS) as an alternative end point for OS. Criteria for establishing surrogacy and the advantages and limitations associated with the use of DFS as a primary end point in adjuvant clinical trials and as the basis for approval of new adjuvant therapies are discussed.
WHAT IS A SURROGATE END POINT?
A surrogate end point has been defined as an alternative end point (such as a biological marker, physical sign, or precursor event) that can be used as a substitute for a clinically meaningful end point that measures directly how a patient feels, functions, or survives [1]. A surrogate end point is selected based on a biologic rationale and may be employed when the primary end point of interest is difficult or expensive to measure and when an alternative, more accessible end point is sufficiently well correlated with the primary to justify its use as a substitute. In cancer trials, potential candidates as surrogate end points for OS include tumor response and progression-free survival (in the presence of measurable disease), or DFS (following curative-intent primary therapy) because these events occur earlier and are typically unaffected by the use of secondary therapies. The potential advantages to the use of surrogate end points are inherent: the effect of an intervention can be evaluated faster, at less cost, and with fewer trial subjects.
Despite the enthusiasm for surrogate end points and the biologic and practical rationale for their use, there are well known examples in clinical medicine of failed surrogate measures, including the use of arrhythmia suppression as a surrogate for cardiovascular mortality [2] and the use of bone mineral density as a surrogate for fracture risk [3]. Fleming and DeMets have described several reasons why a surrogate endpoint may fail [4] : (a) the surrogate is not in the causal pathway of the disease process, (b) the intervention affects only the pathway mediated through the surrogate but does not affect other important causal pathways, (c) the surrogate is insensitive to the treatment effect or is not in the pathway of the treatment’s effect, and/or (d) the treatment has mechanisms of action independent of the disease process. Correlation of a surrogate with the clinical outcome of interest is not solely sufficient to establish the validity of a surrogate end point; proper validation requires an understanding of the causal pathways of the disease process and the mechanism of action of the treatment of interest.
Providing a framework for studying and validating surrogate end points has been the subject of considerable study. Prentice [5] proposed a formal definition of conditions that a surrogate marker be required to satisfy. According to the Prentice criteria, a surrogate marker must be both correlated with the true end point and fully capture the net effect of treatment on the true end point. This approach was advanced by Freedman et al. [6] with the examination of the so-called "proportion-explained;" that is, the proportion of treatment (or exposure) effect explained by the surrogate end point. Buyse and Molenberghs [7] subsequently proposed that a surrogate should be required to fulfill measures of agreement with the true end point at both the individual level and at the population level. Accordingly, a surrogate is said to be "perfect" at the individual level when there is a perfect association between the surrogate and the true end point after adjustment for treatment. At the population level, a surrogate is said to be "perfect" when the effect of treatment on the true end point is equal to the effect of treatment on the surrogate end point. These measures are evaluated on a 0–1 scale, with values approaching 1 indicating effective surrogacy.
As noted by Begg and Leung [8], an important conceptual limitation of these criteria is that a surrogate end point applied in the context of an actual clinical trial cannot be validated unless the true end points are also evaluated; that is, a criterion specific to comparing the treatment effect on both the surrogate and true end point cannot be satisfied unless the true end points are actually observed. This would effectively negate the intent of using a surrogate end point. They, therefore, proposed that the validity of a surrogate end point should be judged by the probability that the trial results based on the surrogate end point alone are concordant with the trial results that would be obtained if the true end point were observed and used for the analysis. Based on these criteria, it is clear that proper validation of a surrogate end point requires data from multiple trials, optimally through a pooled or meta-analysis examining both positive and negative trials to validate both the sensitivity and specificity of the proposed surrogate. The greater the heterogeneity present in the included trials, the stronger the case for the surrogate marker becomes; if an end point is valid in many different heterogeneous settings, it makes it more likely the relationship will be valid in future, yet unspecified trials.
IS DFS A VALID SURROGATE FOR OS?
DFS (or recurrence-free survival) is defined as the time from randomization to the first of either recurrence or relapse, second cancer, or death. Its application as a surrogate end point is most appropriate in settings where it is expected that recurrence of disease presents a major component of mortality in the treated population. This would be the case for most solid tumors, for which secondary therapies at the time of recurrence may prolong survival but ultimately are unlikely to result in a cure. DFS as a surrogate would particularly be advantageous when a sufficiently lengthy interval exists between recurrence and death, thus requiring a longer follow-up for a primary trial end point of OS. The use of DFS as an alternative end point for OS for adjuvant trials has been a topic of much discussion and its application is reviewed below in the context of resected high-risk colon cancer, early breast cancer, and non-small cell lung cancer (NSCLC).
Adjuvant Therapy for Colon Cancer
The benefit of adjuvant chemotherapy for high-risk colon cancer is well established [9, 10]. In a pooled analysis of seven large randomized trials including more than 3,300 patients with stage II and III disease, postoperative 5-fluorouracil (5-FU)-based chemotherapy is associated with a 26% proportional reduction in the risk for death and an overall improvement in the 5-year survival rate from 64% to 71% [10, 11]. In all of these trials, OS was the defined primary end point for efficacy. Also, in these trials, 85% of deaths within 8 years of the colon cancer diagnosis occurred following a recurrence of disease [11]. It is further recognized that the majority of recurrences are documented within 3 years of diagnosis. In order to explore whether the metric of DFS assessed after a 3-year median follow-up is an appropriate alternative end point for OS, Sargent et al. [12] conducted an individual patient data pooled analysis of randomized adjuvant colon cancer trials.
This analysis, known as the ACCENT (Adjuvant Colon Cancer Endpoints) analysis, pooled data from 18 randomized phase III clinical trials, consisting of 43 treatment arms and a total pooled sample size of 20,898 patients. The primary analysis was a comparison of hazard ratios (HRs) for 3-year DFS and 5-year OS within patient, within treatment arm, and within trial. During the 8-year follow-up period, 80% of recurrences were documented within the first 3 years. Comparisons of the 3-year DFS rate and 5-year OS rate by treatment arm, using weighted linear regression, were very highly correlated (R2 = 0.85; Spearman rank correlation coefficient = 0.88). When examined by trial, to determine whether comparison of the experimental and control arms using DFS reaches the same conclusion as using OS, a high level of correlation was again observed (R2 = 0.90; Spearman rank correlation coefficient = 0.94), with a slight attenuation of the treatment effect HR between the two end points. The regression model estimating the relationship between DFS and OS using within-trial hazard ratios was tested using leave-one-out cross-validation; excellent model calibration was demonstrated, with 24 of 25 observed HRs for OS within the 95% prediction interval.
Recognizing that no single method of surrogacy validation is agreed upon, multiple formal analytic approaches were used, including the proportion explained method of Prentice [5] and Freedman et al. [6], trial-level and individual-level measures of agreement [13], and the probability of concordance [8]. The proportion of treatment effect of adjuvant 5-FU-based chemotherapy on OS explained by DFS was >100%. This seemingly counterintuitive finding was the result of a change in the coefficient for treatment in the Cox model from negative to positive when recurrence, which is the dominant predictor of death, was included, and demonstrates a limitation of the proportion explored approach. Trial-level and individual-level correlation were 0.78 (95% confidence interval [CI], 0.60–0.96) and 0.873 (95% CI, 0.869–0.877), respectively. Twenty three of 25 within-trial comparisons (control vs. treatment) obtained the same conclusion, regardless of end point. In summary, strong evidence for surrogacy was demonstrated.
The slight attenuation of treatment effect observed when using DFS versus OS merits additional discussion. While the majority of deaths are preceded by recurrence, and while recurrent disease is typically not salvageable for a curative outcome, deaths from competing causes and the use of palliative yet survival-prolonging therapies may attenuate the treatment effect comparison for OS. Admittedly, the survival impact of palliative interventions in the era of the trials included in the ACCENT analysis (1978–1996) was modest. When extrapolating the DFS–OS treatment effect association to modern-day trials, however, one may expect this attenuation to be magnified in an era of more rigorous surveillance and superior diagnostics leading to earlier detection of recurrent disease, and the availability of more efficacious "palliative" therapies, which not only prolong survival in the setting of recurrent disease but also increase the possibility of salvage for cure [14].
Recently, the efficacy of oxaliplatin in combination with 5-FU/leucovorin for the adjuvant treatment of stage II and III colon cancer was reported using a primary end point of DFS. In the Multicenter International Study of Oxaliplatin/5-FU/Leucovorin in the Adjuvant Treatment of Colon Cancer (MOSAIC) trial, the 3-year DFS rate was significantly superior in the oxaliplatin-containing arm (78.2% vs. 72.9%; p = .002) [15]. A similar magnitude of efficacy was subsequently reported in the National Surgical Adjuvant Breast and Bowel Project (NSABP) C-07 trial (3-year DFS rate, 76.5% vs. 71.6%; p < .004) [16]. Final analysis for the secondary end point of OS is pending for both these trials; however, in an update of the MOSAIC trial results after a median follow-up of 56.2 months, the OS rates at 4 years remain similar (84.9% vs. 82.8%; HR, 0.91; 95% CI, 0.75–1.11; p = .91) [17]. Notable proportions of patients remain alive with recurrence (7.4% and 11.9%) on the oxaliplatin plus 5-FU/leucovorin and 5-FU/leucovorin arms, respectively. While the majority of these recurrences will not be salvageable for cure, this does illustrate the potential for attenuation of the DFS–OS correlation when median survival is prolonged for metastatic recurrence in the current era of superior palliative therapies. Rather than questioning 3-year DFS as an appropriate end point, these considerations may rather indicate that 5 years may be too rapid to truly assess long-term outcome in the modern treatment of colon cancer, further emphasizing the need for a surrogate end point in this setting.
Breast Cancer
While testing for surrogacy has yet to be formally explored in this setting, DFS has evolved as an accepted end point for adjuvant trials in breast cancer. This is perhaps best illustrated in the series of adjuvant studies supporting the efficacy of aromatase inhibitors in postmenopausal women with early breast cancer. In the Arimidex?, Tamoxifen, Alone or in Combination (ATAC) randomized trial of anastrozole alone or in combination with tamoxifen versus tamoxifen alone, anastrozole resulted in a significantly longer DFS than with tamoxifen alone after a median follow-up of 68 months (HR, 0.87; 95% CI, 0.78–0.97). OS (HR, 0.97; 95% CI, 0.85–1.12) and breast-cancer specific survival (HR, 0.88; 95% CI, 0.74–1.05) were not significantly different; however, prolonged follow-up is required to assess the impact on survival given the favorable baseline prognosis of the study population. The authors concluded that "the significant reductions in recurrence associated with anastrozole strongly suggest that a reduction in deaths from breast cancer will eventually be seen." Anastrozole received U.S. Food and Drug Administration (FDA) accelerated approval status based upon the DFS efficacy data, recognizing that the follow-up duration was insufficient for regular approval. Likewise, in the subsequently reported MA-17 randomized trial of letrozole after 5 years of tamoxifen therapy, the primary end point of DFS rate was superior after a median follow-up of 2.4 years (93% vs. 87%; p < .001), but OS rates were similar (p = .25) [18]. In an updated analysis, a survival advantage was reported only in the subset of women with node-positive disease (p = .04) [19]. In interpreting these findings, the American Society of Clinical Oncology Technology Assessment Panel on Adjuvant Use of Aromatase Inhibitors supported the use of aromatase inhibitors either as initial therapy or following a course of tamoxifen, citing that "differences in disease-free survival in the adjuvant setting often lead to differences in OS with prolonged follow-up" [20]. This statement is borne out in the 2005 report of the Early Breast Cancer Trialists’ Collaborative Group overview in which adjuvant chemotherapies and adjuvant tamoxifen associated with recurrence reductions at 5 years were also associated with significant mortality reductions that persisted to a survival benefit at 15 years [21].
More recently, DFS was employed as a primary end point in assessing the efficacy of trastuzumab, a monoclonal antibody that targets human epidermal growth factor receptor 2 (HER-2) in HER-2-positive women with early breast cancer. In a joint analysis of the NSABP B-31 and North Central Cancer Treatment Group (NCCTG)-N9831 trials, trastuzumab in combination with doxorubicin and cyclophosphamide followed by paclitaxel reduced the risk for recurrence by 52% after 2 years of median follow-up (p < .0001). The OS rate at 4 years was also significantly better (91% vs. 87%; p = .015) [22]. Presented at the same 2005 Annual Session of the American Society of Clinical Oncology, the first interim efficacy analysis of the Herceptin? Adjuvant (HERA) trial reported a 46% reduction in recurrence at a 1-year median follow-up (p < .0001) with trastuzumab following adjuvant chemotherapy [23]. The OS rates were similar at the time of this early analysis. The findings from both these trials are significant not only for the magnitude of benefit observed with trastuzumab therapy but also for the proof of concept that DFS may be an acceptable primary end point for adjuvant trials employing targeted therapies. Further follow-up from each of these trials, and likely a formal pooled analysis, will be required to validate DFS as a true surrogate end point in this setting.
NSCLC
Despite earlier conflicting data, recent trials largely support the use of platinum-based adjuvant chemotherapy for early-stage resected NSCLC. The use of DFS as a surrogate end point or as a basis for drug approval in the adjuvant treatment of lung cancer has not yet been examined. As illustrated in Table 1, in the three largest positive randomized trials of adjuvant therapy in NSCLC, the benefit of therapy in terms of DFS appears similar to that observed in terms of OS [24–26]. DFS may be an acceptable end point in this adjuvant setting. However, in the absence of a formal analysis for surrogacy, the reliability of DFS as a predictor of OS in lung cancer remains uncertain at the present time.
DISCUSSION
New therapies need to be effectively and efficiently evaluated in the adjuvant setting, where the treatment goal is increased cure. While OS represents the gold standard metric for establishing efficacy, it requires extended follow-up, which may prevent the timely dissemination of results and consequent implementation of effective curative-intent therapy. For the purposes of regular FDA drug approval, the demonstration of either clinical benefit or an effect on an established surrogate for clinical benefit is required [27]. For an end point that is not sufficiently established to support regular approval, accelerated approval can be granted on the basis that the surrogate is reasonably likely to predict clinical benefit with the requirement that postmarketing studies must verify the clinical benefit. It is acknowledged that DFS may serve as the basis for FDA approval if a high proportion of symptomatic recurrences exist or if a strong correlation with survival exists [28]. Beyond the aforementioned cases for colon, breast, and lung cancer, these conditions may also apply to other settings with expanding adjuvant indications including resected gastric, pancreatic, and bladder cancers.
Can DFS now be routinely applied as an alternative end point to OS for future adjuvant clinical trials? On the one hand, trials would be smaller, faster, and cheaper. By including survival as a component of the composite end point, DFS does account for any long-term effects of a treatment that may ultimately adversely impact survival, such as end-organ toxicities (e.g., cardiotoxicity with trastuzumab) or secondary malignancies. On the other hand, DFS may be compromised in its ability to predict the net treatment benefit on survival as a result of improvements in postrecurrence therapy. In addition, it is unclear whether DFS can adequately reflect the impact of targeted therapies in the adjuvant setting because such therapies are more typically associated with a cytostatic effect; that is, such a therapy may delay but not prevent an eventual recurrence. This may not be a major limitation based upon the adjuvant trastuzumab experience. Ongoing and future adjuvant trials employing targeted therapies will further elucidate this issue. Long-term, simple follow-up for OS will continue to be necessary for adjuvant trials, as OS remains an end point of key clinical and patient interest.
One must also consider the merit of DFS as an end point unto itself. From a physician and patient perspective, it is difficult to question the clinical benefit of preventing or delaying a recurrence independent of its ultimate impact on survival. Alive on therapy is not equivalent to alive free of disease; the impact of symptomatic recurrent disease is significant both in terms of patient morbidity and costs to society. From another perspective, in settings where postrecurrence therapy significantly affects long-term survival, DFS may change as the "preferred" end point because of its unambiguous measure of a new treatment’s impact on the disease process. As an end point less affected by crossover or subsequent therapy, it can be argued that DFS provides a more biologically relevant measure of a new treatment’s impact on the disease process.
In summary, while improving patient OS will always be the goal of adjuvant therapy, in selected settings, DFS has become accepted as a surrogate end point of sufficient merit to satisfy regulatory approval when assessing new treatments in randomized trials. The paths to this outcome were varied: the greatest level of validation is in the setting of high-risk resected colon cancer, where a large international pooled analysis was required, whereas in breast cancer, the extended survival of patients required the use of DFS as a practical necessity with subsequent data supporting surrogacy. In other adjuvant settings, such as early-stage NSCLC, pooled analyses similar to the ACCENT analysis will likely be necessary. The benefit of these efforts can be substantial. Given the high incidence rates of many of these malignancies, the ability to establish and deliver improved therapies by a margin of two or more years earlier, through the use of a surrogate end point, may translate into better outcomes for thousands of patients.
DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
The authors indicate no potential conflicts of interest.
REFERENCES
Temple RJ A Regulatory Authority’s Opinion About Surrogate Endpoints. In: Clinical Measurement in Drug Evaluation. Nimmo WS, Tucker GT, eds. New York: J. Wiley and Sons, 1995.
Echt DS, Liebson PR, Mitchell LB et al. Mortality and morbidity in patients receiving encainide, flecainide, orplacebo. The Cardiac Arrhythmia Suppression Trial. N Engl J Med 1991;324:781–788.[Abstract]
Riggs BL, Hodgson SF, O’Fallon WM et al. Effect of fluoride treatment on the fracture rate in postmenopausal women with osteoporosis. N Engl J Med 1990;322:802–809.[Abstract]
Fleming TR, DeMets DL. Surrogate end points in clinical trials: are we being misled? Ann Intern Med 1996;125:605–613.
Prentice RL. Surrogate endpoints in clinical trials: definition and operational criteria. Stat Med 1989;8:431–440.
Freedman LS, Graubard BI, Schatzkin A. Statistical validation of intermediate endpoints for chronic diseases. Stat Med 1992;11:167–178.
Buyse M, Molenberghs G. Criteria for the validation of surrogate endpoints in randomized experiments. Biometrics 1998;54:1014–1029.
Begg CB, Leung D. On the use of surrogate end points in randomized trials. J R Stat Soc [Ser A] 2000;163:15–28.
NIH consensus conference. Adjuvant therapy for patients with colon and rectal cancer. JAMA 1990;264:1444–1450.
Gill S, Loprinzi CL, Sargent DJ et al. Pooled analysis of fluorouracil-based adjuvant therapy for stage II and III colon cancer: who benefits and by how much? J Clin Oncol 2004;22:1797–1806.
Sargent DJ, Goldberg RM, Jacobson SD et al. A pooled analysis of adjuvant chemotherapy for resected colon cancer in elderly patients. N Engl J Med 2001;345:1091–1097.
Sargent DJ, Wieand HS, Haller DG et al. Disease-free survival versus overall survival as a primary end point for adjuvant colon cancer studies: individual patient data from 20,898 patients on 18 randomized trials. J Clin Oncol 2005;23:8664–8670.
Buyse M, Molenberghs G, Burzykowski T et al. The validation of surrogate endpoints in meta-analyses of randomized experiments. Biostatistics 2000;1:49–67.
Grothey A, Sargent D, Goldberg RM et al. Survival of patients with advanced colorectal cancer improves with the availability of fluorouracil-leucovorin, irinotecan, and oxaliplatin in the course of treatment. J Clin Oncol 2004;22:1209–1214.
Andre T, Boni C, Mounedji-Boudiaf L et al. Oxaliplatin, fluorouracil, and leucovorin as adjuvant treatment for colon cancer. N Engl J Med 2004;350:2343–2351.
Wolmark N, Wieand HS, Kuebler JP et al. A phase III trial comparing FULV to FULV + oxaliplatin in stage II or III carcinoma of the colon: results of NSABP Protocol C-07. Proc Am Soc Clin Oncol 2005;23:3500a.
de Gramont A, Boni C, Navarro M et al. Oxaliplatin-5FU/LV in the adjuvant treatment of stage II and stage III colon cancer: efficacy results with a minimal follow-up of 4 years. Proc Am Soc Clin Oncol 2005;23:3501a.
Goss PE, Ingle JN, Martino S et al. A randomized trial of letrozole in post-menopausal women after five years of tamoxifen therapy for early-stage breast cancer. N Engl J Med 2003;349:1793–1802.
Goss PE, Ingle JN, Martino S et al. Updated analysis of the NCIC.CTG MA.17 randomized placebo controlled trial of letrozole after five years of tamoxifen in postmenopausal women with early stage breast cancer. Proc Am Soc Clin Oncol 2004;23:847a.
Winer EP, Hudis C, Burstein HJ et al. American Society of Clinical Oncology technology assessment on the use of aromatase inhibitors as adjuvant therapy for postmenopausal women with hormone receptor-positive breast cancer: status report 2004. J Clin Oncol 2005;23:619–629.
Early Breast Cancer Trialists’ Collaborative Group (EBCTCG). Effects of chemotherapy and hormonal therapy for early breast cancer on recurrence and 15-year survival: an overview of the randomised trials. Lancet 2005;365:1687–1717.
Romond EH, Perez EA, Bryant J et al. Trastuzumab plus adjuvant chemotherapy for operable HER2-positive breast cancer. N Engl J Med 2005;353:1673–1684.
Piccart-Gebhart MJ, Procter M, Leyland-Jones B et al. Trastuzumab after adjuvant chemotherapy in HER2-positive breast cancer. N Engl J Med 2005;353:1659–1672.
Arriagada R, Bergman B, Dunant A et al. Cisplatin-based adjuvant chemotherapy in patients with completely resected non-small-cell lung cancer. N Engl J Med 2004;350:351–360.
Strauss GM, Herndon J, Maddaus MA et al. Randomized clinical trial of adjuvant chemotherapy with paclitaxel and carboplatin following resection in stage IB non-small cell lung cancer: report of CALGB Protocol 9633. Proc Am Soc Clin Oncol 2004;22:7019a.
Winton TL, Livingston RB, Johnson D et al. A prospective randomised trial of adjuvant vinorelbine and cisplatin in completely resected stage IB and II non-small cell lung cancer Intergroup JBR.10. Proc Am Soc Clin Oncol 2004;22:7018a.
Johnson JR, Temple R. Food and Drug Administration requirements for approval of new anticancer drugs. Cancer Treat Rep 1985;69:1155–1159.
Johnson JR, Williams G, Pazdur R. End points and United States Food and Drug Administration approval of oncology drugs. J Clin Oncol 2003;21:1404–1411.(Sharlene Gilla, Daniel Sa)
Key Words. Adjuvant therapy ; Overall survival ; Disease-free survival
ABSTRACT
The intent of adjuvant therapy is to eradicate micro-metastatic residual disease following curative resection with the goal of preventing or delaying recurrence. The time-honored standard for demonstrating efficacy of new adjuvant therapies is an improvement in overall survival (OS). This typically requires phase III trials of large sample size with lengthy follow-up. With the intent of reducing the cost and time of completing such trials, there is considerable interest in developing alternative or surrogate end points. A surrogate end point may be employed as a substitute to directly assess the effects of an intervention on an already accepted clinical end point such as mortality. When used judiciously, surrogate end points can accelerate the evaluation of new therapies, resulting in the more timely dissemination of effective therapies to patients.
The current review provides a perspective on the suitability and validity of disease-free survival (DFS) as an alternative end point for OS. Criteria for establishing surrogacy and the advantages and limitations associated with the use of DFS as a primary end point in adjuvant clinical trials and as the basis for approval of new adjuvant therapies are discussed.
WHAT IS A SURROGATE END POINT?
A surrogate end point has been defined as an alternative end point (such as a biological marker, physical sign, or precursor event) that can be used as a substitute for a clinically meaningful end point that measures directly how a patient feels, functions, or survives [1]. A surrogate end point is selected based on a biologic rationale and may be employed when the primary end point of interest is difficult or expensive to measure and when an alternative, more accessible end point is sufficiently well correlated with the primary to justify its use as a substitute. In cancer trials, potential candidates as surrogate end points for OS include tumor response and progression-free survival (in the presence of measurable disease), or DFS (following curative-intent primary therapy) because these events occur earlier and are typically unaffected by the use of secondary therapies. The potential advantages to the use of surrogate end points are inherent: the effect of an intervention can be evaluated faster, at less cost, and with fewer trial subjects.
Despite the enthusiasm for surrogate end points and the biologic and practical rationale for their use, there are well known examples in clinical medicine of failed surrogate measures, including the use of arrhythmia suppression as a surrogate for cardiovascular mortality [2] and the use of bone mineral density as a surrogate for fracture risk [3]. Fleming and DeMets have described several reasons why a surrogate endpoint may fail [4] : (a) the surrogate is not in the causal pathway of the disease process, (b) the intervention affects only the pathway mediated through the surrogate but does not affect other important causal pathways, (c) the surrogate is insensitive to the treatment effect or is not in the pathway of the treatment’s effect, and/or (d) the treatment has mechanisms of action independent of the disease process. Correlation of a surrogate with the clinical outcome of interest is not solely sufficient to establish the validity of a surrogate end point; proper validation requires an understanding of the causal pathways of the disease process and the mechanism of action of the treatment of interest.
Providing a framework for studying and validating surrogate end points has been the subject of considerable study. Prentice [5] proposed a formal definition of conditions that a surrogate marker be required to satisfy. According to the Prentice criteria, a surrogate marker must be both correlated with the true end point and fully capture the net effect of treatment on the true end point. This approach was advanced by Freedman et al. [6] with the examination of the so-called "proportion-explained;" that is, the proportion of treatment (or exposure) effect explained by the surrogate end point. Buyse and Molenberghs [7] subsequently proposed that a surrogate should be required to fulfill measures of agreement with the true end point at both the individual level and at the population level. Accordingly, a surrogate is said to be "perfect" at the individual level when there is a perfect association between the surrogate and the true end point after adjustment for treatment. At the population level, a surrogate is said to be "perfect" when the effect of treatment on the true end point is equal to the effect of treatment on the surrogate end point. These measures are evaluated on a 0–1 scale, with values approaching 1 indicating effective surrogacy.
As noted by Begg and Leung [8], an important conceptual limitation of these criteria is that a surrogate end point applied in the context of an actual clinical trial cannot be validated unless the true end points are also evaluated; that is, a criterion specific to comparing the treatment effect on both the surrogate and true end point cannot be satisfied unless the true end points are actually observed. This would effectively negate the intent of using a surrogate end point. They, therefore, proposed that the validity of a surrogate end point should be judged by the probability that the trial results based on the surrogate end point alone are concordant with the trial results that would be obtained if the true end point were observed and used for the analysis. Based on these criteria, it is clear that proper validation of a surrogate end point requires data from multiple trials, optimally through a pooled or meta-analysis examining both positive and negative trials to validate both the sensitivity and specificity of the proposed surrogate. The greater the heterogeneity present in the included trials, the stronger the case for the surrogate marker becomes; if an end point is valid in many different heterogeneous settings, it makes it more likely the relationship will be valid in future, yet unspecified trials.
IS DFS A VALID SURROGATE FOR OS?
DFS (or recurrence-free survival) is defined as the time from randomization to the first of either recurrence or relapse, second cancer, or death. Its application as a surrogate end point is most appropriate in settings where it is expected that recurrence of disease presents a major component of mortality in the treated population. This would be the case for most solid tumors, for which secondary therapies at the time of recurrence may prolong survival but ultimately are unlikely to result in a cure. DFS as a surrogate would particularly be advantageous when a sufficiently lengthy interval exists between recurrence and death, thus requiring a longer follow-up for a primary trial end point of OS. The use of DFS as an alternative end point for OS for adjuvant trials has been a topic of much discussion and its application is reviewed below in the context of resected high-risk colon cancer, early breast cancer, and non-small cell lung cancer (NSCLC).
Adjuvant Therapy for Colon Cancer
The benefit of adjuvant chemotherapy for high-risk colon cancer is well established [9, 10]. In a pooled analysis of seven large randomized trials including more than 3,300 patients with stage II and III disease, postoperative 5-fluorouracil (5-FU)-based chemotherapy is associated with a 26% proportional reduction in the risk for death and an overall improvement in the 5-year survival rate from 64% to 71% [10, 11]. In all of these trials, OS was the defined primary end point for efficacy. Also, in these trials, 85% of deaths within 8 years of the colon cancer diagnosis occurred following a recurrence of disease [11]. It is further recognized that the majority of recurrences are documented within 3 years of diagnosis. In order to explore whether the metric of DFS assessed after a 3-year median follow-up is an appropriate alternative end point for OS, Sargent et al. [12] conducted an individual patient data pooled analysis of randomized adjuvant colon cancer trials.
This analysis, known as the ACCENT (Adjuvant Colon Cancer Endpoints) analysis, pooled data from 18 randomized phase III clinical trials, consisting of 43 treatment arms and a total pooled sample size of 20,898 patients. The primary analysis was a comparison of hazard ratios (HRs) for 3-year DFS and 5-year OS within patient, within treatment arm, and within trial. During the 8-year follow-up period, 80% of recurrences were documented within the first 3 years. Comparisons of the 3-year DFS rate and 5-year OS rate by treatment arm, using weighted linear regression, were very highly correlated (R2 = 0.85; Spearman rank correlation coefficient = 0.88). When examined by trial, to determine whether comparison of the experimental and control arms using DFS reaches the same conclusion as using OS, a high level of correlation was again observed (R2 = 0.90; Spearman rank correlation coefficient = 0.94), with a slight attenuation of the treatment effect HR between the two end points. The regression model estimating the relationship between DFS and OS using within-trial hazard ratios was tested using leave-one-out cross-validation; excellent model calibration was demonstrated, with 24 of 25 observed HRs for OS within the 95% prediction interval.
Recognizing that no single method of surrogacy validation is agreed upon, multiple formal analytic approaches were used, including the proportion explained method of Prentice [5] and Freedman et al. [6], trial-level and individual-level measures of agreement [13], and the probability of concordance [8]. The proportion of treatment effect of adjuvant 5-FU-based chemotherapy on OS explained by DFS was >100%. This seemingly counterintuitive finding was the result of a change in the coefficient for treatment in the Cox model from negative to positive when recurrence, which is the dominant predictor of death, was included, and demonstrates a limitation of the proportion explored approach. Trial-level and individual-level correlation were 0.78 (95% confidence interval [CI], 0.60–0.96) and 0.873 (95% CI, 0.869–0.877), respectively. Twenty three of 25 within-trial comparisons (control vs. treatment) obtained the same conclusion, regardless of end point. In summary, strong evidence for surrogacy was demonstrated.
The slight attenuation of treatment effect observed when using DFS versus OS merits additional discussion. While the majority of deaths are preceded by recurrence, and while recurrent disease is typically not salvageable for a curative outcome, deaths from competing causes and the use of palliative yet survival-prolonging therapies may attenuate the treatment effect comparison for OS. Admittedly, the survival impact of palliative interventions in the era of the trials included in the ACCENT analysis (1978–1996) was modest. When extrapolating the DFS–OS treatment effect association to modern-day trials, however, one may expect this attenuation to be magnified in an era of more rigorous surveillance and superior diagnostics leading to earlier detection of recurrent disease, and the availability of more efficacious "palliative" therapies, which not only prolong survival in the setting of recurrent disease but also increase the possibility of salvage for cure [14].
Recently, the efficacy of oxaliplatin in combination with 5-FU/leucovorin for the adjuvant treatment of stage II and III colon cancer was reported using a primary end point of DFS. In the Multicenter International Study of Oxaliplatin/5-FU/Leucovorin in the Adjuvant Treatment of Colon Cancer (MOSAIC) trial, the 3-year DFS rate was significantly superior in the oxaliplatin-containing arm (78.2% vs. 72.9%; p = .002) [15]. A similar magnitude of efficacy was subsequently reported in the National Surgical Adjuvant Breast and Bowel Project (NSABP) C-07 trial (3-year DFS rate, 76.5% vs. 71.6%; p < .004) [16]. Final analysis for the secondary end point of OS is pending for both these trials; however, in an update of the MOSAIC trial results after a median follow-up of 56.2 months, the OS rates at 4 years remain similar (84.9% vs. 82.8%; HR, 0.91; 95% CI, 0.75–1.11; p = .91) [17]. Notable proportions of patients remain alive with recurrence (7.4% and 11.9%) on the oxaliplatin plus 5-FU/leucovorin and 5-FU/leucovorin arms, respectively. While the majority of these recurrences will not be salvageable for cure, this does illustrate the potential for attenuation of the DFS–OS correlation when median survival is prolonged for metastatic recurrence in the current era of superior palliative therapies. Rather than questioning 3-year DFS as an appropriate end point, these considerations may rather indicate that 5 years may be too rapid to truly assess long-term outcome in the modern treatment of colon cancer, further emphasizing the need for a surrogate end point in this setting.
Breast Cancer
While testing for surrogacy has yet to be formally explored in this setting, DFS has evolved as an accepted end point for adjuvant trials in breast cancer. This is perhaps best illustrated in the series of adjuvant studies supporting the efficacy of aromatase inhibitors in postmenopausal women with early breast cancer. In the Arimidex?, Tamoxifen, Alone or in Combination (ATAC) randomized trial of anastrozole alone or in combination with tamoxifen versus tamoxifen alone, anastrozole resulted in a significantly longer DFS than with tamoxifen alone after a median follow-up of 68 months (HR, 0.87; 95% CI, 0.78–0.97). OS (HR, 0.97; 95% CI, 0.85–1.12) and breast-cancer specific survival (HR, 0.88; 95% CI, 0.74–1.05) were not significantly different; however, prolonged follow-up is required to assess the impact on survival given the favorable baseline prognosis of the study population. The authors concluded that "the significant reductions in recurrence associated with anastrozole strongly suggest that a reduction in deaths from breast cancer will eventually be seen." Anastrozole received U.S. Food and Drug Administration (FDA) accelerated approval status based upon the DFS efficacy data, recognizing that the follow-up duration was insufficient for regular approval. Likewise, in the subsequently reported MA-17 randomized trial of letrozole after 5 years of tamoxifen therapy, the primary end point of DFS rate was superior after a median follow-up of 2.4 years (93% vs. 87%; p < .001), but OS rates were similar (p = .25) [18]. In an updated analysis, a survival advantage was reported only in the subset of women with node-positive disease (p = .04) [19]. In interpreting these findings, the American Society of Clinical Oncology Technology Assessment Panel on Adjuvant Use of Aromatase Inhibitors supported the use of aromatase inhibitors either as initial therapy or following a course of tamoxifen, citing that "differences in disease-free survival in the adjuvant setting often lead to differences in OS with prolonged follow-up" [20]. This statement is borne out in the 2005 report of the Early Breast Cancer Trialists’ Collaborative Group overview in which adjuvant chemotherapies and adjuvant tamoxifen associated with recurrence reductions at 5 years were also associated with significant mortality reductions that persisted to a survival benefit at 15 years [21].
More recently, DFS was employed as a primary end point in assessing the efficacy of trastuzumab, a monoclonal antibody that targets human epidermal growth factor receptor 2 (HER-2) in HER-2-positive women with early breast cancer. In a joint analysis of the NSABP B-31 and North Central Cancer Treatment Group (NCCTG)-N9831 trials, trastuzumab in combination with doxorubicin and cyclophosphamide followed by paclitaxel reduced the risk for recurrence by 52% after 2 years of median follow-up (p < .0001). The OS rate at 4 years was also significantly better (91% vs. 87%; p = .015) [22]. Presented at the same 2005 Annual Session of the American Society of Clinical Oncology, the first interim efficacy analysis of the Herceptin? Adjuvant (HERA) trial reported a 46% reduction in recurrence at a 1-year median follow-up (p < .0001) with trastuzumab following adjuvant chemotherapy [23]. The OS rates were similar at the time of this early analysis. The findings from both these trials are significant not only for the magnitude of benefit observed with trastuzumab therapy but also for the proof of concept that DFS may be an acceptable primary end point for adjuvant trials employing targeted therapies. Further follow-up from each of these trials, and likely a formal pooled analysis, will be required to validate DFS as a true surrogate end point in this setting.
NSCLC
Despite earlier conflicting data, recent trials largely support the use of platinum-based adjuvant chemotherapy for early-stage resected NSCLC. The use of DFS as a surrogate end point or as a basis for drug approval in the adjuvant treatment of lung cancer has not yet been examined. As illustrated in Table 1, in the three largest positive randomized trials of adjuvant therapy in NSCLC, the benefit of therapy in terms of DFS appears similar to that observed in terms of OS [24–26]. DFS may be an acceptable end point in this adjuvant setting. However, in the absence of a formal analysis for surrogacy, the reliability of DFS as a predictor of OS in lung cancer remains uncertain at the present time.
DISCUSSION
New therapies need to be effectively and efficiently evaluated in the adjuvant setting, where the treatment goal is increased cure. While OS represents the gold standard metric for establishing efficacy, it requires extended follow-up, which may prevent the timely dissemination of results and consequent implementation of effective curative-intent therapy. For the purposes of regular FDA drug approval, the demonstration of either clinical benefit or an effect on an established surrogate for clinical benefit is required [27]. For an end point that is not sufficiently established to support regular approval, accelerated approval can be granted on the basis that the surrogate is reasonably likely to predict clinical benefit with the requirement that postmarketing studies must verify the clinical benefit. It is acknowledged that DFS may serve as the basis for FDA approval if a high proportion of symptomatic recurrences exist or if a strong correlation with survival exists [28]. Beyond the aforementioned cases for colon, breast, and lung cancer, these conditions may also apply to other settings with expanding adjuvant indications including resected gastric, pancreatic, and bladder cancers.
Can DFS now be routinely applied as an alternative end point to OS for future adjuvant clinical trials? On the one hand, trials would be smaller, faster, and cheaper. By including survival as a component of the composite end point, DFS does account for any long-term effects of a treatment that may ultimately adversely impact survival, such as end-organ toxicities (e.g., cardiotoxicity with trastuzumab) or secondary malignancies. On the other hand, DFS may be compromised in its ability to predict the net treatment benefit on survival as a result of improvements in postrecurrence therapy. In addition, it is unclear whether DFS can adequately reflect the impact of targeted therapies in the adjuvant setting because such therapies are more typically associated with a cytostatic effect; that is, such a therapy may delay but not prevent an eventual recurrence. This may not be a major limitation based upon the adjuvant trastuzumab experience. Ongoing and future adjuvant trials employing targeted therapies will further elucidate this issue. Long-term, simple follow-up for OS will continue to be necessary for adjuvant trials, as OS remains an end point of key clinical and patient interest.
One must also consider the merit of DFS as an end point unto itself. From a physician and patient perspective, it is difficult to question the clinical benefit of preventing or delaying a recurrence independent of its ultimate impact on survival. Alive on therapy is not equivalent to alive free of disease; the impact of symptomatic recurrent disease is significant both in terms of patient morbidity and costs to society. From another perspective, in settings where postrecurrence therapy significantly affects long-term survival, DFS may change as the "preferred" end point because of its unambiguous measure of a new treatment’s impact on the disease process. As an end point less affected by crossover or subsequent therapy, it can be argued that DFS provides a more biologically relevant measure of a new treatment’s impact on the disease process.
In summary, while improving patient OS will always be the goal of adjuvant therapy, in selected settings, DFS has become accepted as a surrogate end point of sufficient merit to satisfy regulatory approval when assessing new treatments in randomized trials. The paths to this outcome were varied: the greatest level of validation is in the setting of high-risk resected colon cancer, where a large international pooled analysis was required, whereas in breast cancer, the extended survival of patients required the use of DFS as a practical necessity with subsequent data supporting surrogacy. In other adjuvant settings, such as early-stage NSCLC, pooled analyses similar to the ACCENT analysis will likely be necessary. The benefit of these efforts can be substantial. Given the high incidence rates of many of these malignancies, the ability to establish and deliver improved therapies by a margin of two or more years earlier, through the use of a surrogate end point, may translate into better outcomes for thousands of patients.
DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
The authors indicate no potential conflicts of interest.
REFERENCES
Temple RJ A Regulatory Authority’s Opinion About Surrogate Endpoints. In: Clinical Measurement in Drug Evaluation. Nimmo WS, Tucker GT, eds. New York: J. Wiley and Sons, 1995.
Echt DS, Liebson PR, Mitchell LB et al. Mortality and morbidity in patients receiving encainide, flecainide, orplacebo. The Cardiac Arrhythmia Suppression Trial. N Engl J Med 1991;324:781–788.[Abstract]
Riggs BL, Hodgson SF, O’Fallon WM et al. Effect of fluoride treatment on the fracture rate in postmenopausal women with osteoporosis. N Engl J Med 1990;322:802–809.[Abstract]
Fleming TR, DeMets DL. Surrogate end points in clinical trials: are we being misled? Ann Intern Med 1996;125:605–613.
Prentice RL. Surrogate endpoints in clinical trials: definition and operational criteria. Stat Med 1989;8:431–440.
Freedman LS, Graubard BI, Schatzkin A. Statistical validation of intermediate endpoints for chronic diseases. Stat Med 1992;11:167–178.
Buyse M, Molenberghs G. Criteria for the validation of surrogate endpoints in randomized experiments. Biometrics 1998;54:1014–1029.
Begg CB, Leung D. On the use of surrogate end points in randomized trials. J R Stat Soc [Ser A] 2000;163:15–28.
NIH consensus conference. Adjuvant therapy for patients with colon and rectal cancer. JAMA 1990;264:1444–1450.
Gill S, Loprinzi CL, Sargent DJ et al. Pooled analysis of fluorouracil-based adjuvant therapy for stage II and III colon cancer: who benefits and by how much? J Clin Oncol 2004;22:1797–1806.
Sargent DJ, Goldberg RM, Jacobson SD et al. A pooled analysis of adjuvant chemotherapy for resected colon cancer in elderly patients. N Engl J Med 2001;345:1091–1097.
Sargent DJ, Wieand HS, Haller DG et al. Disease-free survival versus overall survival as a primary end point for adjuvant colon cancer studies: individual patient data from 20,898 patients on 18 randomized trials. J Clin Oncol 2005;23:8664–8670.
Buyse M, Molenberghs G, Burzykowski T et al. The validation of surrogate endpoints in meta-analyses of randomized experiments. Biostatistics 2000;1:49–67.
Grothey A, Sargent D, Goldberg RM et al. Survival of patients with advanced colorectal cancer improves with the availability of fluorouracil-leucovorin, irinotecan, and oxaliplatin in the course of treatment. J Clin Oncol 2004;22:1209–1214.
Andre T, Boni C, Mounedji-Boudiaf L et al. Oxaliplatin, fluorouracil, and leucovorin as adjuvant treatment for colon cancer. N Engl J Med 2004;350:2343–2351.
Wolmark N, Wieand HS, Kuebler JP et al. A phase III trial comparing FULV to FULV + oxaliplatin in stage II or III carcinoma of the colon: results of NSABP Protocol C-07. Proc Am Soc Clin Oncol 2005;23:3500a.
de Gramont A, Boni C, Navarro M et al. Oxaliplatin-5FU/LV in the adjuvant treatment of stage II and stage III colon cancer: efficacy results with a minimal follow-up of 4 years. Proc Am Soc Clin Oncol 2005;23:3501a.
Goss PE, Ingle JN, Martino S et al. A randomized trial of letrozole in post-menopausal women after five years of tamoxifen therapy for early-stage breast cancer. N Engl J Med 2003;349:1793–1802.
Goss PE, Ingle JN, Martino S et al. Updated analysis of the NCIC.CTG MA.17 randomized placebo controlled trial of letrozole after five years of tamoxifen in postmenopausal women with early stage breast cancer. Proc Am Soc Clin Oncol 2004;23:847a.
Winer EP, Hudis C, Burstein HJ et al. American Society of Clinical Oncology technology assessment on the use of aromatase inhibitors as adjuvant therapy for postmenopausal women with hormone receptor-positive breast cancer: status report 2004. J Clin Oncol 2005;23:619–629.
Early Breast Cancer Trialists’ Collaborative Group (EBCTCG). Effects of chemotherapy and hormonal therapy for early breast cancer on recurrence and 15-year survival: an overview of the randomised trials. Lancet 2005;365:1687–1717.
Romond EH, Perez EA, Bryant J et al. Trastuzumab plus adjuvant chemotherapy for operable HER2-positive breast cancer. N Engl J Med 2005;353:1673–1684.
Piccart-Gebhart MJ, Procter M, Leyland-Jones B et al. Trastuzumab after adjuvant chemotherapy in HER2-positive breast cancer. N Engl J Med 2005;353:1659–1672.
Arriagada R, Bergman B, Dunant A et al. Cisplatin-based adjuvant chemotherapy in patients with completely resected non-small-cell lung cancer. N Engl J Med 2004;350:351–360.
Strauss GM, Herndon J, Maddaus MA et al. Randomized clinical trial of adjuvant chemotherapy with paclitaxel and carboplatin following resection in stage IB non-small cell lung cancer: report of CALGB Protocol 9633. Proc Am Soc Clin Oncol 2004;22:7019a.
Winton TL, Livingston RB, Johnson D et al. A prospective randomised trial of adjuvant vinorelbine and cisplatin in completely resected stage IB and II non-small cell lung cancer Intergroup JBR.10. Proc Am Soc Clin Oncol 2004;22:7018a.
Johnson JR, Temple R. Food and Drug Administration requirements for approval of new anticancer drugs. Cancer Treat Rep 1985;69:1155–1159.
Johnson JR, Williams G, Pazdur R. End points and United States Food and Drug Administration approval of oncology drugs. J Clin Oncol 2003;21:1404–1411.(Sharlene Gilla, Daniel Sa)