Response heterogeneity as a new biomarker of treatment response in patients with neuroendocrine tumors

Figures & data

Figure 1. Overview of the assessment of PFS prediction using the response heterogeneity measurements in patients with NETs enrolled in CLARINET with multiple lesions (n = 148).

A Cox regression analysis assessed the association of response heterogeneity with NET progression at each patient visit at weeks 12, 24, 36, 48 and 72. Cox models were also trained on patient samples of available data at patient visits (week 12: n = 148; week 24: n = 131; week 36: n = 108; week 48: n = 90; week 72: n = 70). A three-variable model (combining the response heterogeneity output with the SLDr and change in logCgA between each patient visit and baseline), was compared with a two-variable model (based on SLDr and change in logCgA between each patient visit and baseline only), and a model based on SLDr alone for the prediction of PFS at patient visits at weeks 12, 24, 36, 48 and 72. The surrogacy of the models, as measured by the proportion explained metric, was also compared at patient visits at weeks 12, 24, 36, 48 and 72.

CgA: Chromogranin A; NET: Neuroendocrine tumor; PFS: Progression-free survival; SD: Standard deviation; SLDr: Sum of longest diameter ratio.

Figure 2. CT images of patients with lesions demonstrating high heterogeneity and low heterogeneity at baseline and week 12.

(A) CT scans of lesions demonstrating low response heterogeneity at baseline and week 12. The response heterogeneity value, calculated as the SD of the LDr of the target lesions, was 0.01 for this patient. (B) CT scans of lesions demonstrating high response heterogeneity at baseline and week 12. The response heterogeneity value was 0.11 for this patient.

CT: Computerized tomography; LDr: Longest diameter ratio; SD: Standard deviation.

Table 1. Baseline demographics and patient disease characteristics for the patients in the RAISE response heterogeneity analysis.

Variable	Lanreotide (n = 69)	Placebo (n = 79)
Male, n (%)	31 (44.9)	40 (50.6)
Prior treatment for NET, n (%)	9 (13.0)	6 (7.6)
Origin of NET, n (%)
Pancreas	26 (37.7)	35 (44.3)
Midgut	25 (36.2)	31 (39.2)
Hindgut	7 (10.1)	3 (3.8)
Unknown or other	11 (15.9)	10 (12.7)
Progressed at baseline, n (%)	0	0
Tumor grade, n (%)
Grade 1: Ki-67 0–2%	45 (65.2)	55 (69.6)
Grade 2: Ki-67 3–10%	24 (34.8)	22 (27.8)
Missing data, n (%)	0	2 (2.5)
Number of lesions at week 12, mean (IQR), range	4.6 (3.0–5.0), 2–10	4.8 (3.0–6.0), 2–10

Baseline demographics and patient disease characteristics from the subset of patients enrolled in CLARINET with multiple lesions who were therefore eligible for the analysis of response heterogeneity in RAISE. Patients generally had stable disease; no patients in either treatment arm had disease which had progressed at baseline between 3 to 6 months before randomization. Ki-67 index data were missing for two patients in the placebo group.

IQR: Interquartile range; NET: Neuroendocrine tumor.

Table 2. Response heterogeneity and NET progression at patient visits over 72 weeks in the subset of patients with multiple lesions enrolled in CLARINET.

Visit	Week 12	Week 24	Week 36	Week 48	Week 72
Number of patients	148	131	108	90	70
Mean value for response heterogeneity	0.05	0.08	0.09	0.09	0.09
SD	0.06	0.07	0.08	0.07	0.07
HR for response heterogeneity	1.48	1.39	1.72	1.38	1.71
95% CI	1.20–1.82	1.14–1.69	1.32–2.24	0.98–1.94	0.87–3.36
p-value	<0.001	0.001	<0.001	0.063	0.120
HR for response heterogeneity with control for SLDr	1.28	1.45	1.81	0.92	1.30
95% CI	1.04–1.59	1.07–1.95	1.20–2.72	0.62–1.36	0.63–2.68
p-value	0.020	0.016	0.005	0.683	0.477

A Cox regression analysis stratified by the number of lesions evaluated the association between response heterogeneity at weeks 12, 24, 36, 48 and 72 and progression after that timepoint. p-values for response heterogeneity have been reported alone and with SLDr as a control covariate.

CI: Confidence interval; HR: Hazard ratio; NET: Neuroendocrine tumor; SD: Standard deviation; SLDr: Sum of longest diameter ratio.

Figure 3. Reported c-index values for the prediction of PFS for the three-variable model, the two-variable model and SLDr alone at patient visits through 72 weeks.

C-index values are reported for the prediction of PFS in the subset of patients with multiple lesions enrolled in CLARINET at patient visits at weeks 12, 24, 36, 48 and 72 for the entire subset. The figure shows the comparison between the three-variable model (incorporating response heterogeneity with SLDr and change from baseline in logCgA), the two-variable model (combining SLDr and change from baseline in logCgA) and SLDr alone.

CgA: Chromogranin A; PFS: Progression-free survival; SLDr: Sum of longest diameter ratio.

Figure 4. Surrogacy of the three-variable and two-variable models, and SLDr and PFS alone, at patient visits through 72 weeks.

Surrogacy of the models was evaluated by the proportion explained metric. This metric quantifies the proportion of treatment effect on PFS that could be accounted for by the variables included in the models (SLDr, change from baseline in logCgA, response heterogeneity) and SLDr and PFS alone at earlier timepoints. The figure shows surrogacy of the three-variable model (incorporating response heterogeneity with SLDr and change from baseline in logCgA), the two-variable model (combining SLDr and change from baseline in logCgA), and SLDr and PFS alone at patient visits at weeks 12, 24, 36, 48 and 72.

CgA: Chromogranin A; PFS: Progression-free survival; SLDr: Sum of longest diameter ratio.