Exploratory analysis of the association of depth of response and survival in patients with metastatic non-small-cell lung cancer treated with a targeted therapy or immunotherapy

Abstract

Background

Response Evaluation Criteria in Solid Tumors (RECIST) permits rapid evaluation of new therapeutic strategies in cancer. However, RECIST does not capture the heterogeneity of response in highly active therapies. Depth of tumor response may provide a more granular view of response. We explored the association between, depth of response (DepOR), with overall survival (OS) and progression-free survival (PFS) for patients with NSCLC being treated with an ALK inhibitor (ALKi) or an anti-PD-1 antibody (Ab).

Methods

Experimental arms from two randomized controlled trials (RCTs) of an ALKi and two RCTs of an anti-PD-1 Ab were separately pooled. Patient responses were grouped into DepOR ‘quartiles’ by percentage of maximal tumor shrinkage (Q1 = 1%–25%, Q2 = 26%–50%, Q3 = 51%–75%, and Q4 = 76%–100%), Q0 had no shrinkage. We carried out a retrospective exploratory responder analysis to evaluate the association between DepOR and OS or PFS using hazard ratios (HR) generated by the Cox proportional hazards model.

Results

In the pooled ALK analysis there were 12, 39, 70, 144, and 40 patients in quartiles 0–4, respectively. The DepOR versus PFS/OS analyses HR were: 0.19/0.94 for Q1 0.11/0.56 for Q2, 0.05/0.28 for Q3, and 0.03/0.05 for Q4. In the PD-1 trials within quartiles 0–4 there were 168, 70, 44, 45, and 28 patients, respectively. The DepOR versus PFS/OS analyses HR were 0.3/0.52 for Q1, 0.22/0.47 for Q2, 0.09/0.07 for Q3, and 0.07/0.14 for Q4.

Conclusions

Our analysis suggests a greater DepOR is associated with longer PFS and OS for patients receiving ALKi or anti-PD1 Ab. Overall, this suggests that DepOR may provide an additional outcome measure for clinical trials, and may allow better comparisons of treatment activity.

Introduction

Oncology therapeutics development relies on assessment of several clinical trial end points. Objective response rate (ORR), an often-used end point, categorizes patients’ responses by Response Evaluation Criteria in Solid Tumors (RECIST) [1]. A limitation of RECIST is that determination of response is a categorical output that defines tumor responses to therapy as progression (≥20% growth), stable disease (19% growth to 29% reduction) versus partial response (≥30%) or complete response (100%) and thus does not capture the full spectrum of benefit that patients derive from therapy. Targeted therapies (TT) directed toward non-small-cell lung cancer (NSCLC) tumors harboring EGFR, ALK, or ROS1 genetic alterations typically demonstrate ORR of 60%–80% or higher in the front-line treatment setting [2–5]. Given the difficulty of improving on these high response rates, the rapid development of TT, and the challenge of monitoring treatment response with immunotherapies (IO), it is important to consider additional outcome measures to characterize the anti-tumor activity of new drugs more thoroughly and efficiently.

Depth of response (DepOR) is an outcome measure that has been used in hematologic malignancies to evaluate changes in M-protein in multiple myeloma, minimal residual disease in leukemia and myeloma, as well as radiographic changes in colon cancer [6–12]. For solid tumors, it is defined as the percent of maximal tumor reduction from baseline of a target lesion. The potential benefit of utilizing DepOR as an outcome measure is that it is a more granular measure of tumor response than ORR and might provide an earlier readout of drug activity than time-to-event end points such as progression-free survival (PFS) or overall survival (OS). In an era of more successful therapeutic strategies, direct measure of tumor response might allow a better determination of log cancer cell killing. It may also provide an earlier assessment of activity for new combination therapeutic approaches to treating cancer, a theme that has been often proposed and studied in preclinical models but has not yet translated into an end point to analyze in lung cancer clinical trials [13].

Recent analyses of NSCLC data from trials of TT have demonstrated an association in clinical trial-level analyses between PFS and ORR [14]. The question remains whether the individual patients who have the largest reduction in tumor burden, DepOR, are the patients deriving benefit with longer OS or PFS. In this study, we sought to determine whether there were patient-level associations between the maximum reduction in tumor target lesions (DepOR) from baseline and PFS or OS in metastatic NSCLC for two different representative drug categories—an anaplastic lymphoma kinase (ALK) inhibitor representing a TT and a programmed cell death-1 (PD-1) inhibitor.

Methods

Trial selection criteria

Data from two randomized trials were pooled for each drug for exploratory analyses (four in total). Within the ALK inhibitor trials 305 of the 345 (88%) patients were included in the analysis and within the PD-1 trials 355 of 427 (83%) were included. Given that across the trials patients were treated with different chemotherapy agents, data only from the experimental arms of the four trials were used. Trials and drugs were chosen based on the availability of randomized data submitted to FDA from March 2013 to March 2016. An ALK inhibitor was chosen as representative of a TT and a PD-1 inhibitor was chosen as representative of an IO.

For the two ALK inhibitor trials, one trial enrolled treatment-naïve patients and the other enrolled patients who had progression after platinum-doublet-based chemotherapy. For both trials, key eligibility criteria included a histologic diagnosis of locally advanced or metastatic NSCLC positive for ALK rearrangements by central testing using a break-apart fluorescence in situ hybridization assay. Patients must have been at least 18 years of age, had measurable disease, and an Eastern Cooperative Oncology Group performance status (ECOG PS) of 0–2. Key eligibility for the two PD-1 inhibitor trials included patients having IIIB or IV NSCLC who had disease recurrence after one prior platinum-containing regimen. Patients were at least 18 years of age, had an ECOG PS of 0–1, and not allowed to have autoimmune disease, symptomatic interstitial lung disease, systemic immunosuppression, or prior immune based therapies.

Outcome measures

Maximum tumor reduction from baseline was compared for PFS and OS.

Statistical analysis

Data from patients with evaluable radiographic scans demonstrating any reduction in tumor size were grouped into four quartiles based on the greatest percent of reduction in tumor target lesions from baseline and compared with data from patients with no tumor reduction. A retrospective exploratory responder analysis was carried out to evaluate the association between the DepOR quartiles and PFS and OS for each drug using hazard ratios (HR). Using both adjusted and unadjusted statistical methods, HRs using the Cox HR model were computed. Kaplan–Meier curves were plotted for PFS and OS by the DepOR quartiles. Only patients who had baseline and at least one post-baseline tumor measurements were included in the analyses.

Results

Patient demographics

In the ALK-positive patient trials, the median age was 51 years, 43% were men, most patients enrolled were white and not located in the USA (Table 1). Most patients had ECOG PS of 0–1 (94%) and were never smokers (63%). Patient demographics and disease characteristics were balanced across the quartiles for most variables with the exception that those patients with 76%–100% tumor reduction were more frequently females and never smokers.

Table 1.

Demographics of the ALK inhibitor cohort

	Overall	No tumor shrinkage	1%–25%	26%–50%	51%–75%	76%–100%
	(n = 305)	(n = 12)	(n = 39)	(n = 70)	(n = 144)	(n = 40)
Age
Median	51	57	53	49	50	50
Range	22, 81	41, 68	25, 73	22, 78	22, 81	29, 73
Gender
Male	131 (43%)	7 (58%)	20 (51%)	32 (46%)	62 (43%)	10 (25%)
Female	174 (57%)	5 (42%)	19 (49%)	38 (54%)	82 (57%)	30 (75%)
Race
White	158 (52%)	2 (17%)	25 (64%)	43 (61%)	70 (49%)	18 (45%)
Black or AA	2 (<1%)	0	1 (3%)	0	1 (<1%)	0
Asian	140 (46%)	10 (83%)	13 (33%)	26 (37%)	69 (48%)	22 (55%)
Other	5 (2%)	0	0	1 (1%)	4 (3%)	0
Region
USA	36 (12%)	0	6 (15%)	10 (14%)	17 (12%)	3 (8%)
Non-USA	269 (88%)	12 (100%)	33 (85%)	60 (86%)	127 (88%)	37 (92%)
ECOG PS
0	119 (39%)	5 (42%)	15 (38%)	30 (43%)	57 (40%)	12 (30%)
1	169 (55%)	6 (50%)	23 (59%)	35 (50%)	78 (54%)	27 (68%)
2	17 (6%)	1 (8%)	1 (3%)	5 (7%)	9 (6%)	1 (3%)
Smoking status
Current/former	114 (37%)	7 (58%)	14 (36%)	26 (37%)	61 (42%)	6 (15%)
Never	191 (63%)	5 (42%)	25 (64%)	44 (63%)	83 (58%)	34 (85%)

Table 2 illustrates the PD-1 inhibitor patient characteristics. The median age was 61 years, 62% were men and 90% of patients were white. The majority (66%) were not from the USA, were identified as current or former smokers (83%), and 66% had non-squamous histology. All patients had an ECOG performance status of 0–1. The demographic distribution across the quartiles was largely stable apart from an increased number of patients in Q4 with non-squamous histology (79%) and more patients in Q4 who were enrolled in the USA (50%).

Table 2.

Demographics of the PD-1 inhibitor cohort

	Overall	No tumor shrinkage	1%–25%	26%–50%	51%–75%	76%–100%
	(n = 355)	(n = 168)	(n = 70)	(n = 44)	(n = 45)	(n = 28)
Age
Median	61	63	61	62	60	64
Range	39, 85	39, 81	40, 85	46, 76	42, 84	48, 81
Gender
Male	221 (62%)	101 (60%)	45 (64%)	28 (64%)	33 (73%)	14 (50%)
Female	134 (38%)	67 (40%)	25 (36%)	16 (36%)	12 (27%)	14 (50%)
Race
White	321 (90%)	156 (93%)	62 (89%)	39 (89%)	39 (87%)	25 (89%)
Black or AA	10 (3%)	4 (2%)	1 (1%)	1 (2%)	4 (9%)	0
Asian	13 (4%)	4 (2%)	3 (4%)	3 (7%)	1 (2%)	2 (7%)
Other	11 (3%)	4 (2%)	4 (6%)	1 (2%)	1 (2%)	1 (4%)
Region
USA	120 (34%)	55 (33%)	20 (29%)	12 (27%)	19 (42%)	14 (50%)
Non-USA	235 (66%)	113 (67%)	50 (71%)	34 (73%)	26 (58%)	14 (50%)
ECOG PS
0	103 (29%)	41 (24%)	21 (30%)	15 (24%)	15 (33%)	11 (39%)
1	252 (71%)	127 (76%)	49 (70%)	29 (66%)	30 (67%)	17 (61%)
Smoking status
Current/former	295 (83%)	132 (79%)	59 (84%)	40 (91%)	39 (87%)	25 (89%)
Never	57 (16%)	34 (20%)	11 (16%)	4 (9%)	5 (11%)	3 (11%)
Missing	3 (<1%)	2 (1%)	0	0	1 (2%)	0
Histology
Sq	120 (34%)	56 (33%)	22 (31%)	18 (41%)	18 (40%)	6 (21%)
NSq	235 (66%)	112 (67%)	48 (69%)	26 (59%)	27 (60%)	22 (79%)

ALK inhibitor pooled analysis

Figure 1A demonstrates the pooled maximal tumor reduction from the two anonymized ALK inhibitor trials. Patients with any amount of tumor size reduction (>0% to 100%) were grouped into DepOR quartiles by the greatest percent decrease from baseline in tumor target lesions. The analysis included 305 patients. Twelve patients (3.9%) had no tumor size reduction; this group serves as the comparator cohort. The distribution of patients within each DepOR quartile was as follows: quartile 1 (1%–25%, n = 39), quartile 2 (26%–50%, n = 70), quartile 3 (51%–75%, n = 144), and quartile 4 (76%–100%, n = 40). Kaplan–Meier curves for each of the DepOR quartiles demonstrate PFS (Figure 1B) and OS (Figure 1C) within each patient group. Table 3 lists the unadjusted and adjusted HR using the Cox proportional hazards model for the responder analysis. Within the ALK inhibitor pooled trial analysis, increasing DepOR demonstrated a stepped improvement in OS; however, only DepOR quartiles 3 and 4 indicated nominally significant results with adjusted HR 0.28 (CI 0.11, 0.73) and HR 0.05 (CI 0.01, 0.28), respectively. The PFS analysis for the ALK inhibitor treated pooled group parallels the OS finding with the exception that (Figure 1C) all DepOR quartiles indicated significant association between DepOR and PFS. Three-dimensional plots further illustrate the relationship between DepOR, patient quartile, and days on therapy (Figure 1D).

Figure 1.

(A) Waterfall plot of the pooled analysis of two ALK inhibitor trials depicting the maximum tumor shrinkage. (B) Kaplan–Meier curve for each quartile base on overall survival probability. (C) Kaplan–Meier curve for each quartile base on progression free survival probability. (D) Three-dimensional plot demonstrating DepOR, patient quartile (left figure), and median days on therapy for each quartile, days on therapy for the pooled ALKi trial patients. On the right, a 3D plot demonstrating DepOR, individual patient, and days on therapy for the pooled ALKi trial patients.

Table 3.

Cox proportional hazard model for depth of response

	Unadjusted OS HR (95% CI)	Adjusted OS HR (95% CI)*	Unadjusted PFS HR (95% CI)	Adjusted PFS HR (95% CI)*
ALK inhibitor DepOR subgroups (n = 305)
No tumor shrinkage (n =12)
1%–25% (n =39)	0.98 (0.36, 2.69)	0.94 (0.34, 2.61)	0.19 (0.09, 0.39)	0.19 (0.09, 0.40)
26%–50% (n =70)	0.66 (0.25, 1.75)	0.56 (0.21, 1.51)	0.12 (0.06, 0.24)	0.11 (0.06, 0.24)
51%–75% (n =144)	0.38 (0.15, 0.96)	0.28 (0.11, 0.73)	0.06 (0.03, 0.12)	0.05 (0.03, 0.11)
76%–100% (n =40)	0.06 (0.01, 0.33)	0.05 (0.01, 0.28)	0.03 (0.02, 0.07)	0.03 (0.02, 0.07)
PD-1 inhibitor DepOR subgroups (n =355)
No tumor shrinkage (n =168)
1%–25% (n =70)	0.54 (0.38, 0.76)	0.52 (0.37, 0.74)	0.30 (0.22, 0.41)	0.30 (0.22, 0.41)
26%–50% (n =44)	0.50 (0.32, 0.78)	0.47 (0.30, 0.74)	0.22 (0.15, 0.32)	0.22 (0.15, 0.32)
51%–75% (n =45)	0.09 (0.04, 0.21)	0.07 (0.03, 0.18)	0.10 (0.06, 0.15)	0.09 (0.06, 0.15)
76%–100% (n =28)	0.14 (0.06, 0.31)	0.14 (0.06, 0.32)	0.06 (0.03, 0.12)	0.07 (0.03, 0.12)

*Adjusted by study, ECOG, and smoking status

PD-1 inhibitor pooled analysis

We also analyzed two pooled, anonymized trials of patients treated with a PD-1 inhibitor. Figure 2A illustrates the results as a waterfall plot. Within DepOR quartiles 1–4 there were 70, 44, 45, and 28 patients, respectively. Of the 355 patients in the pooled analysis, 168 (47%) demonstrated no tumor size reduction. These results do not demonstrate a stepped interval increase in OS or PFS as seen in the ALK trials. Figure 2B and Table 3 illustrate the relationship between OS and DepOR quartile. In the adjusted analysis, all DepOR quartiles had an association between OS and DepOR. Additionally, Q1/Q2 and Q3/Q4 had similar survival curves, indicating that patients with >50% reduction in tumor burden make-up one response group and those patients with ≤50% reduction in tumor burden comprise an alternate response group. When PFS is evaluated in this group, a similar pattern is found (Figure 2C). Three-dimensional plots illustrate the association between DepOR, patient, patient quartile, and days on therapy (Figure 2D).

Figure 2.

(A) Waterfall plot of the pooled analysis of two PD-1 inhibitor trials depicting the maximum tumor shrinkage. (B) Kaplan–Meier curve for each quartile base on overall survival probability. (C) Kaplan–Meier curve for each quartile base on progression free survival probability in the two PD-1 inhibitor trials. (D) Three-dimensional plot demonstrating DepOR, patient quartile (left figure), and median days on therapy for each quartile, days on therapy for the pooled PD-1i trial patients. On the right, a 3D plot demonstrating DepOR, individual patient, and days on therapy for the pooled PD-1i trial patients.

Discussion

In this study, we evaluated the relationship between the DepOR and OS or PFS in metastatic NSCLC patients treated either with a TT or an IO. The results indicate that in metastatic NSCLC, DepOR is associated with OS and PFS with a TT in a molecularly-selected patient population and with a more complex relationship in IO trials.

In the analyses of trials using ALK inhibitors, 96% of patients had some degree of response/reduction in tumor size and there was an incremental increase in survival across quartiles, although in evaluation of OS, only patients in quartiles 3 and 4 had a significantly longer survival time compared with patients with no tumor reduction. All quartiles had a significant association between DepOR and PFS. In the pooled ALK trials, post-study treatments may have impacted overall survival, highlighting one of the limitations of using OS as a clinical end point in cancers for which multiple effective therapies are available. Finally, the reference group is limited by the small number of patients.

In the PD-1 inhibitor pooled analysis, 53% of patients had some degree of tumor size reduction. Patients in all DepOR quartiles had significantly longer PFS or OS compared with patients with no tumor reduction. This finding may be due to the unique mechanism of action of PD-1 inhibitors (indirect tumor cell inhibition) compared with TKIs (direct tumor cell inhibition), and may indicate that DepOR is not as clinically useful for IO therapies. Alternatively, it could reflect that an arbitrary use of DepOR quartiles does not represent the optimal cut-off. For example, Q3 and Q4 did not differ from each other, but did seem to have greater benefit than Q1 and Q2, suggesting that a 50% cut point may be a better threshold for predicting benefit with IO therapies. Finally, a limitation of this analysis was the smaller number of patients in each of the four quartiles compared with the non-responders. For example, only 21% of patients in the IO trials had ≥50% reduction in tumor size compared with 60% of patients in the ALK inhibitor trials.

DepOR has been previously reported as an outcome measure in solid tumors. It was investigated in colon cancer in a post hoc analysis of the CRYSTAL and OPUS trials, which evaluated the efficacy of FOLFIRI+/− cetuximab and FOLFOX4+/− cetuximab in 1535 patients and demonstrated that deeper responses correlated with longer OS and post-progression survival [10]. It was then evaluated prospectively in the phase III TRIBE trial in which first line FOLFOXIRI + bevacizumab and FOLFIRI + bevacizumab were compared in metastatic colorectal cancer. The authors found that DepOR is highly correlated with PFS, OS, and post-progression survival on a patient level [7]. DepOR is also a clinically meaningful outcome in multiple hematologic malignancies and major molecular response evaluation is now a standard outcome measure [9, 11, 12, 15, 16].

A search of Clinicaltrials.gov identified two active, therapeutic trials that are evaluating DepOR as an outcome measure in patients with solid tumors. NCT02296125 is a trial of osimertinib versus gefitinib or erlotinib in patients with locally advanced or metastatic NSCLC and NCT02394834 is a study comparing treatment sensitivity and prognostic factors in FOLFOX6 with either bevacizumab or panitumumab in advanced or recurrent colorectal cancer. These studies may assist in demonstrating the utility of DepOR with additional TT and antibody therapies as well as with additional tumor types.

The association between improved time to event outcomes in patients with larger magnitude of DepOR suggests that for patients to obtain the most benefit from a treatment, maximizing tumor size reduction is critical. Given the high ORR and extended PFS generated by TT such as ALK, ROS1, and EGFR TKIs, using DepOR may provide an earlier read-out of activity for new therapeutics. There are many strategies to maximize initial tumor response, including the use of more potent TT in the first-line setting [17], local therapy such as radiation or surgery to residual disease at maximal response from initial therapy [18], or combination therapy strategies in the first-line setting [19, 20]. Most therapeutic strategies have focused on any response, followed by intervention to address drug resistance after disease progression has occurred. The use of DepOR refocuses attention on the lack of complete response in the vast majority of patients and may allow for initial therapeutic strategies to improve responses and possibly address primary resistance [13].

A limitation of this study was the inclusion of only one type of TT. Thus, it is unclear whether our findings translate to other treatment strategies. However, given its potential utility with chemotherapy in colon cancer, DepOR may have broader applicability. We were unable to evaluate DepOR in the control arms of these trials because patients received different chemotherapy regimens. In this analysis, we only included target lesions as measurements of non-target lesions were not collected. Given this we were unable to make direct survival comparisons to the RECIST response categories. However, we expect that since our analysis is based on RECIST data aggregated in DepOR quartiles the incremental improvement in survival in DepOR quartiles may not be observed in RECIST categories. Additionally, we observed inter-patient variation which may be increased by using only measurable target lesions (Figures 1D and 2D). However, with a large sample size, this variation may average out across a large patient population. Finally, as RECIST is the current gold standard for evaluating response the DepOR analysis, it is subject to the same limitations as RECIST in that it is a marker of tumor burden.

Another limitation of this analysis is that the comparisons in this responder analysis are carried out post-randomization and, therefore, potentially important prognostic values are not balanced between quartiles. Additionally, the patient numbers within each DepOR quartile differ, and in the IO trials, the number of patients with any tumor reduction is relatively low, 187 (53%). Furthermore, the use of DepOR quartiles was arbitrary and may not represent the optimal cut-points for categorization of a continuous variable. Finally, in patients with tumor growth we did not evaluate the survival. Survival in patients without tumor shrinkage likely reflects tumor biology and subsequent therapy instead of treatment efficacy.

Prospective evaluations are needed to understand how to use DepoR as an end point in clinical trials and how it compares with traditional RECIST measures of tumor response. Additional evaluations such as maximal tumor reduction at first imaging, or time to maximal tumor reduction may be complementary. In conclusion, DepOR may be a useful clinical end point to more efficiently evaluate activity of new drugs or assist in the early evaluation of new combination therapies to determine if there is the potential for increased benefit over existing single agent therapies.

Note: This study was previously presented at the ASCO 2016 Annual Meeting.