Surrogate endpoints in second-line treatment for mCRC: A systematic literature-based analysis from 23 randomised trials

Abstract

Purpose. To evaluate progression-free survival (PFS), overall response rate (ORR) and disease control rate (DCR) as potential surrogate endpoints (SEP) for overall survival (OS) in second-line treatment for metastatic colorectal cancer (mCRC).

Methods. A systematic literature search of randomised trials of second-line chemotherapy for mCRC reported from January 2000 to July 2013 was performed. Correlation coefficients weighted by number of patients in the treatment arms between median PFS, ORR and DCR with median OS were estimated.

Results. Twenty-three trials reflecting 10 800 patients met the inclusion criteria. Median PFS and OS across all trials were 4.5 months and 11.5 months and median ORR and DCR were 11.4% and 65%, respectively. PFS showed moderate correlation with OS [R_PFS = 0.73; 95% confidence interval (CI) 0.61–0.82]. In contrast, ORR only weakly correlated with OS (R_ORR = 0.58; 95% CI 0.38–0.72, n = 22). Despite a small number of studies (n = 10) reporting on DCR, moderate correlation with OS was observed (R_DCR = 0.74; 95% CI 0.56–0.86).

Conclusion. Based on the available trial-level data, PFS may serve as an appropriate SEP in second-line chemotherapy for mCRC. A small number of studies revealed moderate correlation of DCR with OS that justifies further investigation.

With the introduction of innovative targeted treatments and combination chemotherapy regimens in the treatment of metastatic colorectal cancer (mCRC) surrogate endpoints (SEP) for treatment activity are intensively discussed among investigators, clinicians and drug approval authorities [1–4]. According to Piedbois and Croswell, an appropriate SEP “is an endpoint that can be used as a substitute for the true endpoint” and is ideally reached in shorter duration [2]. However, while the validation of SEPs in mCRC is of emerging and urgent clinical interest, generally accepted validation criteria are currently lacking [5,6].

Several literature-based approaches have been made to evaluate SEPs in first-line treatment, but to the best of our knowledge, have not been estimated in second-line treatment where innovative drugs are usually applied [7–10]. In the setting of previously treated patients with advanced disease, progression-free survival (PFS) and overall response rate (ORR) are widely used as the primary endpoint, with unpredictable surrogacy for the ‘true endpoint’ of overall survival (OS). For the patient, chemotherapy is generally undertaken to provide improved quantity and/or quality of survival, where OS remains the gold standard for assessment of efficacy. However, OS is similarly regarded to be influenced by innovative and effective targeted drugs that are introduced into mCRC in the second- and further-line setting [11].

The purpose of this study was to perform a comprehensive literature-based quantitative analysis of second-line treatment for mCRC, to determine whether PFS, ORR and DCR are correlated with OS, and to evaluate their applicability as SEPs.

Material and methods

Literature search and data extraction

Data from randomised controlled trials including previously treated patients who underwent second-line chemotherapy for mCRC published between January 2000 and July 2013 were identified through a systematic search in the National Library of Medicine medical literature database via PubMed gateway and the Cochrane Library using the keywords ‘metastatic colorectal cancer’ and ‘second-line randomised trials’. Furthermore, congress abstracts presented at the annual meetings of the American Society of Clinical Oncology (ASCO) between January 2000 and July 2013 and bibliographies of published overviews were examined [12–14]. A sample size of at least 100 patients per trial and reports of either PFS or time to progression (TTP) along with OS were required. ORR and DCR were also included, when reported in the trial. Only reports in English language were considered.

Trials investigating surgical resection of resectable/unresectable metastatic disease, trials limited to elderly/younger patients, pooled data reports or non-interventional trials (registry data), and trials using additional hepatic chemotherapy infusion were excluded. In contrast to our previous analysis in first-line treatment, trials involving non-approved drugs for mCRC by the European Medical Association (EMA) were included to reflect the condition of innovative drugs being introduced. Patients suffering from gastrointestinal cancers other than colorectal cancer were not considered.

Data for inclusion and exclusion were reviewed by four investigators (C.G., C.S., D.P.M. and V.H.), following the systematic literature search process in Figure 1 according to the PRISMA statement [15].

Figure 1. PRISMA diagram/literature search.

The reported data on sample size, chemotherapy treatment regimen, median PFS or TTP (months), ORR (%), DCR (%) and median OS (months) and corresponding hazard ratios (HR) were extracted for each trial arm. A total of six trials used TTP as the primary endpoint. Unlike PFS, TTP is defined to use disease progression as the only event of interest neglecting death from any other cause. Following previous analyses of colorectal and metastatic breast cancer, TTP and PFS were used reciprocally and were referred to as PFS [6,9,16]. Response evaluation was performed according to RECIST 1.0 criteria in 16 trials, RECIST 1.1 in one trial and according to WHO criteria in five trials. In one trial information on response evaluation was missing, because response was not an endpoint [30]. ORR was defined as complete response (CR) and partial response (PR). DCR was defined as CR, PR and stable disease (SD).

Statistical methods

Based on summary statistics provided in the publications, Pearson's correlation coefficient between median PFS (logarithmic scale) and median OS (logarithmic scale) was calculated and weighted by number of patients in the treatment arms. For analysis of ORR, logit scale was correlated to median OS in a logarithmic scale. By analogy, DCR (logit scale) and median OS (logarithmic scale) were also calculated. The evaluated endpoints PFS, ORR and DCR would be claimed an acceptable surrogate end point for OS if R was close to 1, indicating a strong correlation between the endpoints and OS adapting the line of reasoning outlined in individual patient analyses [17].

For all weighted correlation coefficients 95% confidence intervals (CI) were obtained using the percentile bootstrap with 10 000 replications. A p-value of p < 0.05 was considered statistically significant. All analyses were performed using the R version 2.13.2 statistical package (R Development Core Team Vienna, Austria, 2012. Available from: http://www.r-project.org).

Results

Within the search period from January 2000 until July 2013, a total of 23 trials including 10 800 patients were available and met the inclusion criteria (Table I).

Table I. Trials included in the analysis.

Trial	Treatment	Number of patients	Overall response rate (%)	Disease control rate (%)	Progression free survival (months)	Overall survival (months)
Bendell et al. 2011 [27]	FOLFIRI + Axitinib	49	25.5		5.7	12.9
	FOLFIRI + Bevacizumab	51	23.5		6.9	15.7
	FOLFOX + Axitinib	36	22.2		7.5	17.1
	FOLFOX + Bevacizumab	35	20.0		6.4	14.1
Bennouna et al. 2013 [19]	Chemotherapy + Bevacizumab	409	5.0		5.7	11.2
	Chemotherapy	411	4.0		4.1	9.8
Cohn et al. 2013b [28]	FOLFIRI + Conatumumab	51	14.0	69.0	6.5	12.3
	FOLFIRI + Ganitumab	52	8.0	67.0	4.5	12.4
	FOLFIRI + Placebo	52	2.0	65.0	4.6	12.0
Cohn et al. 2013a [20]	FOLFIRI + Panitumumab	91	28.0		7.7	18.0
	FOLFIRI + Bevacizumab	91	16.0		9.2	21.4
Cunningham et al. 2004 [29]	Iriotecan + Cetuximab	218	22.9		4.1	8.6
	Cetuximab	111	10.8		1.5	6.9
Eng et al. 2013 [21]	Irinotecan + Cetuximab + Tivantinib	62	45.0	81.7	8.3	19.8
	Irinotecan + Cetuximab + Placebo	60	33.0	71.9	7.3	16.9
Fuchs et al. 2003 [30]	Irinotecan weekly	95			4.0	9.9
	Irinotecan q21d	196			3.0	9.9
Giantonio et al. 2007 [31]	FOLFOX4 + Bevacizumab	286	22.7		7.3	12.9
	FOLFOX4	291	8.6		4.7	10.8
	Bevacizumab	243	3.3		2.7	10.2
Haller et al. 2008 [32]	IROX	317	22.0	78.0	5.3	13.4
	Irinotecan	310	7.0	55.0	2.8	11.1
Hoehler et al. 2013 [33]	FOLFOX/FOLFIRI + Sorafenib	50	25.6		5.2	9.6
	FOLFOX/FOLFIRI	51	12.2		5.6	12.7
Kim et al. 2009 [34]	Irinotecan	245	15.5		4.4	14.3
	FOLFOX4	246	28.0		6.2	13.8
Masi et al. 2013 [35]	FOLFOX/FOLFIRI + Bevacizumab	92	21.0		6.7	16.5
	FOLFOX/FOLFIRI	92	18.0		5.2	16.0
Muro et al. 2010 [36]	FOLFIRI	213	16.7		5.1	18.2
	IRIS	213	18.8		5.8	19.5
Peeters et al. 2010 [22]	FOLFIRI + Panitumumab WT	303	35.0	74.0	5.9	14.5
	FOLFIRI WT	294	10.0	65.0	3.9	12.5
	FOLFIRI + Panitumumab MT	238	13.0	68.0	5.0	11.8
	FOLFIRI MT	248	14.0	63.0	4.9	11.1
Peeters et al. 2013 [37]	FOLFIRI + Trebananib	95	14.0	59.0	3.5	11.9
	FOLFIRI + Placebo	49	0.0	69.0	5.2	8.8
Rothenberg et al. 2003 [38]	LV5U2	151	2.6	42.3	2.7	8.7
	Oxaliplatin	156	3.2	39.7	1.6	8.1
	FOLFOX4	152	13.8	69.1	4.6	9.8
Rothenberg et al. 2008 [39]	FOLFOX4	314	18.0		4.8	12.6
	XELOX	313	20.0		4.7	11.9
Rougier et al. 2002 [40]	Irinotecan/LV5FU2	35	11.4	60.0	3.2	12.2
	Oxaliplatin/LV5FU2	33	21.2	79.7	4.7	11.5
	Irnotecan/Oxaliplatin	33	15.2	60.6	4.2	11.0
Schoemaker et al. 2004 [41]	Irinotecan q3w	41	7.0	51.0	2.7	9.4
	Irinotecan weekly	37	5.0	67.0	3.5	7.1
	Irinotecan q2w	46	11.0	65.0	3.8	8.6
	Irinotecan cont. q3w	44	2.0	45.0	1.7	6.4
Sobrero et al. 2008 [23]	Irinotecan + Cetuximab	648	16.4	61.4	4.0	10.7
	Irinotecan	650	4.2	45.8	2.6	10.0
Tsavaris et al. 2003 [42]	Irinotecan q3w	60	22.0		4.5	7.0
	Irinotecan weekly	60	25.0		6.0	9.0
Van Cutsem et al. 2011 [43]	FOLFOX4 + Vatalanib	426	18.5		5.6	13.1
	FOLFOX4 + Placebo	429	17.5		4.2	11.9
Van Cutsem et al. 2012 [44]	FOLFIRI + Aflibercept	612	19.8	85.7	6.9	13.5
	FOLFIRI + Placebo	614	11.1	76.0	4.6	12.1

FOLFIRI, irinotecan + infusional FU/LV, FOLFOX, oxaliplatin + infusional FU/LV; FU, fluorouracil; IRIS, irinotecan + S1; IROX, irinotecan + oxaliplatin; LV, leucovorin; Lv5FU2, leucovorin 200 mg/m², 2-h infusion, 5-fluorouracil (5-FU) 400 mg/m² bolus, 600 mg/m² 22-h continuous infusion, days 1 and 2; Mayo, 5-FU 425 mg/m² + LV 20 mg/m² daily for 5 days every 4–5 weeks; MT, mutant, OS, overall survival; PFS, progression-free survival; XELOX, capecitabine + oxaliplatin; WT, wild-type.

Trial results according to the year of publication showed a moderate increase in OS and ORR, but stable PFS over time in both, experimental and standard treatment regimens. Across the 23 trials, median PFS was 4.5 (1.6–8.3) and median OS 11.5 (range 6.4–19.8) months. Median ORR was 11.4 % (0.0–45.0%) and median DCR 65.0% (39.7–85.7%) (Table II).

Table II. Descriptive median response rates (ORR, DCR) and median survival rates (PFS, OS).

	No. trials	No. patients	Median	Range
ORR	22	10 509	11.4%	0.0–45.0%
DCR	10	5383	65.0%	39.7–85.7%
PFS	23	10 800	4.5 months	1.6–8.3 months
OS	23	10 800	11.5 months	6.4–19.8 months

DCR, disease control rate; ORR, overall response rate; OS, overall survival; PFS, progression-free survival.

Correlation between PFS and OS

For the analysis of PFS, data from 23 trials representing 10 800 patients were available. PFS showed a moderate correlation with OS (R = 0.73, 95% CI 0.61–0.82; p < 0.001) (Table III, Figure 2).

Figure 2. Correlation between PFS and OS. The solid line represents the change in PFS according to a change in OS; dotted lines indicate pointwise 95% confidence intervals. Symbol size is proportional to number of patients.

Table III. Correlation between progression-free survival (PFS), overall response rate (ORR), disease control rate (DCR) with overall survival (OS).

	No. trials	No. patients	R	95% CI	p-Value
PFS	23	10 800	0.73	0.61–0.82	< 0.001
ORR	22	10 509	0.58	0.38–0.72	0.003
DCR	10	5383	0.74	0.56–0.86	0.007

95% CI, 95% confidence interval; DCR, disease control rate; ORR, overall response rate; PFS, progression-free survival.

Correlation between ORR and OS

For evaluation of ORR, a total of 10 509 patients in 22 trials were included in the analysis. ORR showed only a weak correlation with OS (R = 0.58, 95% CI 0.38–0.72; p = 0.003) (Table III, Figure 3).

Figure 3. Correlation between ORR and OS. The solid line represents the change in ORR according to a change in OS; dotted lines indicate pointwise 95% confidence intervals. Symbol size is proportional to number of patients.

Correlation between DCR and OS

Data from 5383 patients from 10 trials were available for analysis of DCR. A moderate correlation between DCR and OS was found, with the limitation of a small sample size (R = 0.74; 95% CI 0.56–0.86; p = 0.007) (Table III, Figure 4).

Figure 4. Correlation between DCR and OS. The solid line represents the change in DCR according to a change in OS; dotted lines indicate pointwise 95% confidence intervals. Symbol size is proportional to number of patients.

Discussion

The present analysis explored currently applied clinical endpoints during the 2000–2013 period, finding stagnation in PFS compared to only modest increases in OS and ORR in the setting of second-line chemotherapy. This coherency was present in both the standard and experimental trial arms. However, it is important to note that several recent trials incorporating innovative targeting drugs have led to encouraging improvements.

The use of PFS instead of OS as a primary endpoint in clinical trials for first- and second-line chemotherapy for mCRC is widely accepted. In accordance with data regarding first-line treatment, the present study shows that PFS can be considered as a reliable surrogate for OS, thus accelerating the reporting of trial results and facilitating the more rapid introduction of new drugs into clinical practice. In contrast, evaluation of the main response-driven endpoint ORR did not indicate ORR as an appropriate surrogate endpoint. However, analysis of DCR revealed a correlation in the same range of PFS. Limited to small sample size of 10 available trials, DCR may be considered for further investigation as a potential SEP in second-line chemotherapy.

There are several limitations to the present analysis. First, our study was based on aggregated data derived from a literature-based review, while the gold standard for investigation of SEPs requires individual patient data. Accordingly, our meta-analytic approach can only be regarded as a first step towards this issue and has to be confirmed at the individual patient level [18]. Secondly, due to the small number of published trials in the second-line scenario, we were not able to differentiate according to the different mechanisms of drug activity. In this regard, the inhibitions of VEGF versus EGFR-related signal transduction may have induced different effects on OS as well as on its potential surrogate parameters. As an example, the addition or continuation of bevacizumab in the second-line treatment did not provide a clinically measurable increase of response parameters, but resulted in a significant prolongation of PFS and OS [19,20]. In contrast, the application of EGFR-targeting agents demonstrated both, a statistically significant increment of response and a prolongation of survival parameters [20–23]. Recently published data have raised the discussion that in the era of biologicals PFS may not be the optimum surrogate endpoint in first-line treatment of mCRC [4,24,25]. Therefore, re-evaluation according to the different mechanisms of drug activity is also necessary in the second-line setting as soon as a larger set of trials is available.

The present analysis of 23 second-line trials indicated PFS as an applicable SEP for OS in the second-line treatment for mCRC. In addition, DCR rather than ORR may be further investigated as an additional SEP. Innovative endpoints in mCRC including tumour- or patient-related biomarkers, as well as endpoints measured early on during treatment, and investigation of these potential SEPs are of interest with regard to the promising agents emerging [26].

Acknowledgements

No direct or indirect funding was received for this study. No funding bodies had any role in the study design, data collection, analysis, decision to publish or preparation of the manuscript. The authors were personally salaried by their institutions during the period of writing (though no specific salary was set aside or given for the writing of this paper).

Declaration of interest: Clemens Giessen has received travel support form Roche. Ruediger Paul Laubender has received honoraria, research and travel support from Merck. Sebastian Stintzing has received honoraria and travel support from Merck and Roche. Dominik Paul Modest has received travel support from Amgen and Roche and received honoraria from Amgen, Merck and Roche. Volker Heinemann has received honoraria, is member at advisory boards and received financial support for research from Amgen, Merck, Sanofi-Aventis and Roche. All remaining authors have declared no conflicts of interest.