Purpose:

Tumor growth rate (TGR) represents the percentage change in tumor volume per month (%/m). Previous results from the GREPONET study showed that TGR measured after 3 months (TGR3m) of starting systemic treatment (ST) or watch and wait (WW) was an early biomarker predicting progression-free survival (PFS) in neuroendocrine tumors (NET).

Experimental Design:

Patients from 7 centers with advanced grade (G) 1/2 NETs from the pancreas (P)/small bowel (SB) initiating ST/WW were eligible. Computed tomography (CT)/MRI performed at prebaseline, baseline, and 3(±1) months of study entry were retrospectively reviewed. Aim-1: explore treatment-induced changes in TGR (ΔTGR3m-BL; paired T test), and Aim-2: validate TGR3m (<0.8%/m vs. ≥0.8%/m) as an early biomarker in an independent cohort (Kaplan–Meier/Cox regression).

Results:

Of 785 patients screened, 127 were eligible. Mean (SD) TGR0 and TGR3m were 5.4%/m (14.9) and −1.4%/m (11.8), respectively. Mean (SD) ΔTGR3m-BL paired-difference was −6.8%/m (19.3; P < 0.001). Most marked ΔTGR3m-BL [mean (SD)] were identified with targeted therapies [−11.3%/m (4.7); P = 0.0237] and chemotherapy [−7.9%/m (3.4); P = 0.0261]. Multivariable analysis confirmed the absence of previous treatment (OR = 4.65; 95% CI, 1.31–16.52; P = 0.018) and low TGR3m (continuous variable; OR 1.09; 95% CI, 1.01–1.19; P = 0.042) to be independent predictors of radiologic objective response. When the multivariable survival analysis for PFS (Cox regression) was adjusted to grade (P = 0.004) and stage (P = 0.017), TGR3m ≥ 0.8 (vs. <0.8) maintained its significance as a prognostic factor (P < 0.001), whereas TGR0 and ΔTGR3m-BL did not. TGR3m ≥ 0.8%/m was confirmed as an independent prognostic factor for PFS [external validation; Aim-2; multivariable HR 2.21 (95% CI, 1.21–3.70; P = 0.003)].

Conclusions:

TGR has a role as a biomarker for monitoring response to therapy for early identification of treatment-induced changes and for early prediction of PFS and radiologic objective response.

Translational Relevance

Radiologic assessment of response to therapy remains challenging, especially in specific tumor types, such as neuroendocrine tumors. The GREPONET study shows that tumor growth rate (TGR) is a reliable biomarker able to help with the assessment of response to therapy in neuroendocrine tumors. In addition, the fact that TGR measured at 3 months was validated as a prognostic marker highlights the role of TGR as an early biomarker to predict benefit from treatment and could allow tailoring of patient management and follow-up.

Gastroenteropancreatic neuroendocrine tumors (GEP-NET) are rare and heterogeneous neoplasms (1, 2). Advanced disease is only amenable to palliative treatment and is associated with impaired quality of life (3, 4). Based on available evidence, stage, primary tumor site, and tumor grade are the cornerstone for treatment planning (5, 6). Tumor grade is defined by WHO based on morphology and the percentage of tumor cells with nuclear expression of Ki-67; tumors are classified in grade (G)1 Ki-67 <2%, G2 3% to 20% and G3 >20%. The majority of GEP-NETs demonstrate a slow growth rate and fall into G1 to G2 categories whereas G3 tumors are less frequent. G1 and G2 are characterized by a well-differentiated morphology, whereas G3 are poorly differentiated tumors. Recently, a newly described group of neoplasms, so-called G3NET, has been defined to include patients with well-differentiated tumors with Ki-67 above 20% (7).

Current practice relies on the use of response evaluation criteria in solid tumors [RECIST, version 1.1 (RECIST 1.1); ref. 8] for definition of response/progression to systemic therapies. Unfortunately, the slow tumor growth, and similarly tumor shrinkage in response to therapy, in G1/2 GEP-NETs can confound the applicability of RECIST 1.1 by providing limited information, because monitoring by CT and MRI in most patients usually results in “stable disease.” Although to face this problem, researchers have proposed an alternative cut-off for RECIST 1.1–defined response in NETs (9), identification of slow progression by RECIST v1.1 remains challenging.

Tumor growth rate (TGR) provides quantitative information on the change in tumor volume over time (%/month), based on data from 2 CT/MRI examinations and the time between these investigations (10, 11). Previous studies have explored the role of TGR in cancer and assessment of response to treatment (10, 12, 13), including NETs (11, 14–18). In summary, these studies have suggested that TGR could be of value both as a predictive factor of tumor progression and as a prognostic factor in terms of survival in patients with NET undergoing somatostatin analogue (SSA) treatment.

The GREPONET study aimed to further clarify the prognostic role of TGR in pancreatic and small bowel NETs (11). In this study, TGR was retrospectively assessed for 222 patients diagnosed with G1/2 small bowel and pancreatic NETs across 7 international centers. Both TGR pretreatment (TGR0) and TGR at 3 months (TGR3m) of starting systemic treatment (ST) or watch and wait (WW) were explored and showed that patients with high TGR0 (≥4%/m) and TGR3m (≥0.8%/m) had shorter progression-free survival [PFS; HR = 2.2; 95% confidence interval (95% CI, 1.1–4.3) and HR 3.8; 95% CI, 2.2–6.3, respectively]. In this study, only 30 patients had paired TGR0 and TGR3m data, thus analysis of TGR changes induced by treatment could not be explored in detail due to limited power.

The aim of the GREPONET-2 study was to validate previous findings in an independent cohort, and to further explore TGR for therapy monitoring and follow-up in an expanded cohort of patients in which paired data for TGR0 and TGR3m was available.

Study design

The GREPONET-2 study was designed as a retrospective multicentre study, with ethical approval being obtained (as applicable) in each center. Clinical and radiologic data were collected for consecutive eligible patients.

For each patient, the same imaging technique (CT/MRI) was selected for each time point. All examinations were reviewed locally by one radiologist with an expertise in NETs, blinded for clinical data. An adapted version of RECIST 1.1 so-called RECISTmTGR1.1 (as defined previously; ref. 11) which allowed inclusion of a maximum of 10 target lesions (5/organ) and which used RECIST 1.1 for lesion definition and measurements was utilized.

Calculation of TGR was expressed as the percentage change in tumor volume over one month (%/m). TGR was calculated as previously described (10, 11, 13) using the formula: TGR = 100 × [exp (TG)–1]; TG = [3 × log (D2/D1)]/time (months), where D1 is the tumor size at date 1; D2 is the tumor size at date 2; and time (months) = (date 2–date 1 + 1)/30.44; TG is the tumor growth. Tumor size (D1, D2) was determined using the sum of the longest diameters (SLD) of target lesions only. Two TGRs were calculated for every patient: (i) pretherapeutic TGR (TGR0): calculated comparing baseline and imaging examination performed within 12 months before the baseline examination (pre-baseline); (ii) TGR at 3 (±1) months of starting treatment/follow-up (for patients on “WW”; TGR3m): calculated comparing baseline and examination at 3 months.

Objectives

The GREPONET-2 study addressed 2 main questions:

Aim-1 had as primary objective to evaluate whether starting any ST (including “WW”) produced any absolute intrapatient changes on TGR (ΔTGR3m-BL = TGR3m − TGR0). Additional objectives of Aim-1 were to assess (i) ΔTGR3m-BL by treatment subgroups; (ii) the impact of such change on PFS and overall survival (OS); (iii) the correlation between ΔTGR3m-BL and treatment and other clinical factors (including but not limited to TGR0, primary tumor and grade); and (iv) the identification of an optimum cut-off value (%/month) for ΔTGR3m-BL to predict 12-month PFS.

Aim-2 was to confirm the impact of TGR3m on PFS in an external and independent patient cohort as previously suggested, using the previously defined cut-off of ≥0.8%/m for definition of “high TGR3m” (11).

Study population

Eligible patients were those who met the following inclusion criteria: patients diagnosed with small bowel or pancreatic grade 1/2 NETs (confirmed by histopathology or cytology); disseminated disease (not amenable for curative resection) at study entry; due to start any line of treatment with systemic therapies [including WW and peptide receptor radionuclide therapy (PRRT)]; available CT/MRI pretreatment (maximum of 12 months were allowed; no minimum time was prespecified), baseline and at 3 months (±1 month) of study entry. Only patients with presence of at least one target lesion meeting RECISTmTGR1.1 criteria (as previously described; ref. 11) were eligible. Patients with other concomitant malignancy (advanced stage with disease present at time of study entry), or treated with other therapies before disease progression while followed-up as per the GREPONET-2 study were excluded.

Patients with paired TGR0 and TGR3m data from GREPONET study (11) were eligible for Aim-1. These were, however, excluded for Aim-2 because an independent cohort was required for validation of our previous findings.

Statistical analysis

For TGR0 and TGR3m, previously identified cut-offs [4%/m for TGR0 (14) and 0.8%/m for TGR3m (11)] were used for survival analysis. PFS and OS were defined as the time from starting the new treatment/“watch and wait” period to the time of progression [defined as the time to progression as per radiologic/clinical criteria based in current clinical practice (RECIST v1.1)] and the date of death/last follow-up, respectively.

TGR did not follow a normal distribution, thus nonparametric tests were used for analysis. Statistical Wilcoxon matched-pairs signed-ranks test, Mann–Whitney test, linear regression, and ANOVA were applied as appropriate for comparison of TGR within and between subgroups. Receiver-operating characteristic (ROC) curves were used for identification of the most informative ΔTGR3m-BL cut-off for prediction of 12-month progression rate (for this purpose, patients who were progression-free with a follow-up shorter than 12 months were excluded from the analysis). Kaplan–Meier method and univariate/multivariable Cox regression were used for survival analysis; mutivariable analysis was undertaken including factors identified to be significant in the univariate analysis (P < 0.05). For Aim-2 a sensitivity analysis following the landmark method was performed, by which patients dying or progressing during the first 3 months of follow-up were excluded from the multivariable Cox regression PFS analysis. Chi-square test and univariate/multivariable logistic regression were used to identify factors related with increased chances of patients achieving objective response as best response (defined as complete or partial response by RECIST v1.1) at any point during the study period. Two-sided significance tests with P < 0.05 were considered significant. Stata version 12.0 software was used for the statistical analysis (19).

Patients were identified across 7 international centers. Out of 785 patients screened, 127 were eligible for Aim-1 and 97 for Aim-2 (Fig. 1). Median follow-up was 27.36 months. The majority of patients were ST naïve (72.44%) and started treatment with SSA (37.80%) or chemotherapy (32.28%). CT was more frequently used (76.38%); the majority of target lesions (74.94%) were located in the liver. Median time between pre-baseline and baseline imaging for calculation of TGR0 was 3.38 months (95% CI, 2.92–3.71); for 18 and 9 patients, this time was above 6 and 9 months, respectively.

Figure 1.

Consort diagram. *, Contrast-enhanced CT examinations on a multidetector CT with at least 16 rows and/or contrast-enhanced MRI with nonspecific or liver-specific Gd-based contrast medium. **, Patients included in our GREPONET study were excluded for Aim-2 (validation of TGR as an early biomarker in an external cohort of patients).

Figure 1.

Consort diagram. *, Contrast-enhanced CT examinations on a multidetector CT with at least 16 rows and/or contrast-enhanced MRI with nonspecific or liver-specific Gd-based contrast medium. **, Patients included in our GREPONET study were excluded for Aim-2 (validation of TGR as an early biomarker in an external cohort of patients).

Close modal

Patient characteristics and imaging details are summarized in Tables 1 and 2, respectively. Estimated median OS and PFS were 77.47 months [95% CI, 46.93–not reached (nr)] and 17.31 months (95% CI, 14.25–19.47), respectively; 43.31% of the patients had progressed at 12 months of study entry.

Table 1.

Patient baseline characteristics

Baseline characteristics (Aim-1; n = 127)N%
Age (study entry; years) Median (range) 62.43 (17.82–83.41) 
Gender Male 62 48.82 
 Female 65 51.18 
Familiar NET syndrome Yes 5* 3.94 
Primary tumor Pancreas 75 59.06 
 Small bowel 52 40.94 
Stage Localized/locally advanced 5.51 
 Metastatic 120 94.49 
Grade Grade 1 39 30.71 
 Grade 2 80 62.99 
 Not specified 6.30 
Ki-67 (%) All patients; median (95% CI); mean (SD) 5. 0 (4–8); 6.89 (5.38) 
 Grade 2 patients; median (95% CI); mean (SD) 10 (8–10) 9.32 (4.81) 
Line of treatment 92 72.44 
 18 14.17 
 3/4 17 13.38 
Type of treatment received during study period WW 10 7.87 
 SSA 48 37.80 
 Targeted therapies 23 18.11 
 Chemotherapy 41 32.28 
 PRRT 3.94 
Best response at 3 months (RECIST 1.1) Complete response 0.79 
 Partial response 1.57 
 Stable disease 84 66.14 
 Progressive disease 39 30.71 
 Not specified 0.79 
Best response achieved during the whole study period (RECIST v.1.1) Complete response (CR)/Partial response 18 (includes 1 patient with CR) 14.2% 
Baseline characteristics (Aim-1; n = 127)N%
Age (study entry; years) Median (range) 62.43 (17.82–83.41) 
Gender Male 62 48.82 
 Female 65 51.18 
Familiar NET syndrome Yes 5* 3.94 
Primary tumor Pancreas 75 59.06 
 Small bowel 52 40.94 
Stage Localized/locally advanced 5.51 
 Metastatic 120 94.49 
Grade Grade 1 39 30.71 
 Grade 2 80 62.99 
 Not specified 6.30 
Ki-67 (%) All patients; median (95% CI); mean (SD) 5. 0 (4–8); 6.89 (5.38) 
 Grade 2 patients; median (95% CI); mean (SD) 10 (8–10) 9.32 (4.81) 
Line of treatment 92 72.44 
 18 14.17 
 3/4 17 13.38 
Type of treatment received during study period WW 10 7.87 
 SSA 48 37.80 
 Targeted therapies 23 18.11 
 Chemotherapy 41 32.28 
 PRRT 3.94 
Best response at 3 months (RECIST 1.1) Complete response 0.79 
 Partial response 1.57 
 Stable disease 84 66.14 
 Progressive disease 39 30.71 
 Not specified 0.79 
Best response achieved during the whole study period (RECIST v.1.1) Complete response (CR)/Partial response 18 (includes 1 patient with CR) 14.2% 

aIncludes: 1 Tuberous sclerosis, 2 MEN1, 1 MUTYH FAP, 1 VHL.

Table 2.

Radiological details

n (%)
Type of imaging CT 97 (76.38) 
 MRI 30 (23.62) 
Target lesions Total 387 (100) 
 Liver 290 (74.94) 
 Node metastases 28 (7.23) 
 Lung 3 (0.78) 
 Other 29 (7.49) 
 Not specified 37 (9.56) 
Target lesions (per patient) Mean (SD) 3.07 (1.78) 
Tumor burden per patient (sum of all target lesions) (mm) Mean (SD) 92.86 (74.98) 
n (%)
Type of imaging CT 97 (76.38) 
 MRI 30 (23.62) 
Target lesions Total 387 (100) 
 Liver 290 (74.94) 
 Node metastases 28 (7.23) 
 Lung 3 (0.78) 
 Other 29 (7.49) 
 Not specified 37 (9.56) 
Target lesions (per patient) Mean (SD) 3.07 (1.78) 
Tumor burden per patient (sum of all target lesions) (mm) Mean (SD) 92.86 (74.98) 

Treatment-related changes in TGR: ΔTGR3m-BL

Figure 2 summarizes changes in TGR during the study period. Mean TGR0 and TGR3m for the whole series were 5.39%/m (SD 14.85) and −1.41%/m (SD 11.82), respectively. TGR0 and TGR3m were similar between treatment groups (ANOVA P = 0.6168 for differences in TGR0 between treatment groups; ANOVA P = 0.6315 for differences in TGR3m between treatment groups). Paired intrapatient absolute differences between TGR0 and TGR3m (ΔTGR3m-BL) were statistically significant [mean = −6.81 (SD 19.27); Wilcoxon P < 0.001] for the whole cohort (regardless of treatment received; Fig. 2A). Most patients (90/127; 71.65%) developed some degree of reduction in TGR after 3 months of treatment (Fig. 3); with 5 patients experiencing no change (ΔTGR3m-BL = 0) and 32 patients (25.19%) experiencing an increase in TGR.

Figure 2.

A–E, Changes in tumor growth rate after 3 months of treatment or WW (ΔTGR3m-BL). ANOVA P values are provided for assessment of differences between groups (i.e., differences in ΔTGR3m-BL between treatment groups). Mean change (SD) with Wilcoxon matched-pairs signed-rank test is provided for comparison of TGR0 and TGR3m within categories (i.e., changes between TGR0 and TGR3m for patients treated with SSA).

Figure 2.

A–E, Changes in tumor growth rate after 3 months of treatment or WW (ΔTGR3m-BL). ANOVA P values are provided for assessment of differences between groups (i.e., differences in ΔTGR3m-BL between treatment groups). Mean change (SD) with Wilcoxon matched-pairs signed-rank test is provided for comparison of TGR0 and TGR3m within categories (i.e., changes between TGR0 and TGR3m for patients treated with SSA).

Close modal
Figure 3.

Pairwise comparisons of tumor growth rate at baseline (TGR0) and after 3 months of ST or WW (TGR3m). ΔTGR3m-BL, changes in tumor growth rate after 3 months of treatment or WW; n, number of paired observations; P value refers to Wilcoxon matched-pairs signed-ranks test P value.

Figure 3.

Pairwise comparisons of tumor growth rate at baseline (TGR0) and after 3 months of ST or WW (TGR3m). ΔTGR3m-BL, changes in tumor growth rate after 3 months of treatment or WW; n, number of paired observations; P value refers to Wilcoxon matched-pairs signed-ranks test P value.

Close modal

The most marked reduction in ΔTGR3m-BL were identified in the subgroup of patients treated with SSA, targeted therapies, and chemotherapy (Fig. 2B), whereas no significant changes were identified within the WW subgroup (Wilcoxon P = 0.2411; Fig. 2B). Differences between treatment groups did not reach statistical significance (ANOVA P = 0.6774; Fig. 2B).

ΔTGR3m-BL was similar in grades 1 and 2 patients (mean ΔTGR3m-BL = −4.97 vs. −8.45; ANOVA P = 0.3663; Fig. 2E). In contrast, patients with pancreatic primary (vs. small bowel; mean ΔTGR3m-BL = −9.92 vs. −2.31; ANOVA P = 0.0283; Fig. 2D) and with high TGR0 (≥4%/m; vs. <4%/m; mean ΔTGR3m-BL = −14.16 vs. −0.22; ANOVA P < 0.0001; Fig. 2C) had a more marked reduction of ΔTGR3m-BL. Multivariable linear regression adjusted for grade, primary tumor, TGR0, and treatment (full data not shown) confirmed that high TGR0 [≥4%/m (vs. <4%/m) multivariable linear regression P < 0.001] and primary pancreatic tumor (vs. small bowel; multivariable linear regression P = 0.015) were the only independent factors related to a major reduction in ΔTGR3m-BL regardless of administered treatment (multivariable linear regression P = 0.599) and tumor grade (multivariable linear regression P = 0.202).

Changes in ΔTGR3m-BL and RECIST

The use of TGR in the form of ΔTGR3m-BL identified subtle variations in tumor volume that would otherwise have been missed (i.e., classified as stable disease despite having reduction in tumor size) by only applying RECIST v.1.1 (Fig. 4; gray area). In our series, 56 of the 84 patients (66.67%) who had been classified as having stable disease after 3 months of treatment by RECIST1.1 showed a reduction in ΔTGR3m-BL.

Figure 4.

Variation of tumor growth rate (ΔTGR3m-BL) according to tumor response by RECIST 1.1 after 3 months of treatment or WW. Grey fields highlight patients for whom TGR was able to identify changes indicating response to treatment (reduction of ΔTGR3m-BL) and classified as “stable disease” by RECIST 1.1.

Figure 4.

Variation of tumor growth rate (ΔTGR3m-BL) according to tumor response by RECIST 1.1 after 3 months of treatment or WW. Grey fields highlight patients for whom TGR was able to identify changes indicating response to treatment (reduction of ΔTGR3m-BL) and classified as “stable disease” by RECIST 1.1.

Close modal

Patients who achieved objective response as their best response during the whole study period had lower TGR3m [mean = −10.43 (SD 20.31)] as compared with the other patients [mean 0.01 (SD 9.48)]; Mann–Whitney P-value = 0.004. Such differences were also relevant when the TGR3m cut-off of 0.8%/m was applied: 88.89% of patients who achieved an objective response had TGR3m <0.8%/m (vs. 11.11% with TGR≥0.8%/m, Chi-square P = 0.023). In contrast, no differences in TGR0 (Mann–Whitney P = 0.6957) or ΔTGR3m-BL (Mann–Whitney P = 0.1425) were identified between patient who did versus did not achieve partial response during the whole study period. Notably, only 2 of 18 (11.1%) patients who achieved an objective response as best response during the whole study period, met such criteria at 3 months.

Univariate logistic regression identified following factors to be related to increased objective response rate (full data not shown): pancreatic NET (vs. small bowel NET), treatment-naïve patients (vs. patients with previous treatment), low TGR3m (both as a continuous and binary variable, <0.8%/m vs. ≥0.8%/m) and low ΔTGR3m-BL (understood as marked reduction in TGR after therapy/follow-up). Multivariable analysis confirmed the absence of previous treatment (OR = 4.65; 95% CI, 1.31–16.52; P = 0.018) and low TGR3m (continuous variable; OR = 1.09; 95% CI, 1.01–1.19; P = 0.042) to be related to increased rate of objective response. When included as a categorical variable, low TGR3m (<0.8%/m) showed a trend towards increase objective response rate, however it did not reach statistical significance in the multivariable analysis (OR = 5.20; 95% CI, 0.99–27.12; P = 0.05).

Impact of ΔTGR3m-BL on patient outcome and survival analysis

ROC curve analysis was performed to explore if ΔTGR3m-BL correlated with progression rate at 12 months; AUC was low (AUC = 0.4221; 95% CI, 0.31397–0.53024); thus, no further analysis with this purpose was performed.

Univariate Cox-regression analysis identified that metastatic disease, grade 2 tumors, high Ki-67, high TGR0, high TGR3m and low ΔTGR3m-BL (understood as marked reduction in TGR after therapy/follow-up) correlated with shorter PFS (Table 3; first column). Both high TGR3m and high TGR0 correlated with shorter PFS when these 2 variables were included in the multivariable analysis (Table 3; Model 1). ΔTGR3m-BL did not impact on PFS (Table 3; Model 2/Model 3). Grade 2, metastatic disease and high TGR3m were independent prognostic factors in the multivariable analysis when all variables significant in the univariate analysis were included (Table 3; Model 3). ΔTGR3m-BL did not impact on overall survival (P = 0.511).

Table 3.

Univariate and multivariable Cox regression for PFS (Aim-1)

Univariate analysis (Cox regression)Multivariable analysis (Cox regression)Multivariable analysis (Cox regression)Multivariable analysis (Cox regression)
Model 1Model 2Model 3
HR (95% CI)P valueHR (95% CI)P valueHR (95% CI)P valueHR (95% CI)P value
Age (years) Continuous variable 0.99 (0.98–1.01) 0.286 — — — — — — 
Gender Male (vs. Female) 0.91 (0.61–1.34) 0.624 — — — — — — 
Familiar NET syndrome Yes (vs. No) 1.05 (0.42–2.58) 0.921 — — — — — — 
Primary tumor Pancreas (vs. Small Bowel) 1.29 (0.86–1.95) 0.201 — — — — — — 
Stage Metastatic (vs. loc/loc adv) 4.66 (1.47–14.79) 0.009 4.49 (1.39–14.43) 0.012 4.33 (1.35–13.89) 0.014 4.17 (1.29–13.43) 0.017 
Grade Grade 2 (vs. Grade 1) 1.93 (1.23–3.03) 0.005 2.02 (1.25–3.27) 0.004 1.96 (1.24–3.11) 0.004 2.04 (1.25–3.33) 0.004 
Ki-67 Continuous variable 1.06 (1.02–1.09) 0.003 Included in the form of Grade Included in the form of Grade Included in the form of Grade 
Previous treatment Yes (vs. No) 1.23 (0.83–1.82) 0.295 — — — — — — 
Type of treatment (during study period) Watch and wait 1 (Ref) — — — — — — — 
 SSA 0.68 (0.32–1.44) 0.318 — — — — — — 
 Targeted therapies 0.92 (0.42–2.02) 0.839 — — — — — — 
 Chemotherapy 0.92 (0.44–1.91) 0.824 — — — — — — 
 PRRT 2.46 (0.81–7.52) 0.113 — — — — — — 
Number of target lesions Cont variable 1.09 (0.98–1.23) 0.099 — — — — — — 
Tumor burden (sum all lesions) Cont variable 1.002 (0.99–1.005) 0.196 — — — — — — 
Type of examination MRI (vs CT) 1.14 (0.73–1.79) 0.560 — — — — — — 
TGR0 (%/month) ≥4 (vs <4) 1.96 (1.31–2.92) 0.001 1.68 (1.09–2.59) 0.018 Included in the form of change in ΔTGR3m-BL 1.43 (0.88–2.30) 0.147 
TGR3m (%/month) ≥0.8 (vs <0.8) 2.06 (1.37–3.09) 0.001 2.55 (1.64–3.97) <0.001 Included in the form of change in ΔTGR3m-BL 2.97 (1.84–4.78) <0.001 
ΔTGR3m-BL (x10 %/month) Cont variable 0.89 (0.79–0.99) 0.036 Included in the form of TGR0 and TGR3m 0.96 (0.83–1.04) 0.191 0.89 (0.78–1.01) 0.073 
Univariate analysis (Cox regression)Multivariable analysis (Cox regression)Multivariable analysis (Cox regression)Multivariable analysis (Cox regression)
Model 1Model 2Model 3
HR (95% CI)P valueHR (95% CI)P valueHR (95% CI)P valueHR (95% CI)P value
Age (years) Continuous variable 0.99 (0.98–1.01) 0.286 — — — — — — 
Gender Male (vs. Female) 0.91 (0.61–1.34) 0.624 — — — — — — 
Familiar NET syndrome Yes (vs. No) 1.05 (0.42–2.58) 0.921 — — — — — — 
Primary tumor Pancreas (vs. Small Bowel) 1.29 (0.86–1.95) 0.201 — — — — — — 
Stage Metastatic (vs. loc/loc adv) 4.66 (1.47–14.79) 0.009 4.49 (1.39–14.43) 0.012 4.33 (1.35–13.89) 0.014 4.17 (1.29–13.43) 0.017 
Grade Grade 2 (vs. Grade 1) 1.93 (1.23–3.03) 0.005 2.02 (1.25–3.27) 0.004 1.96 (1.24–3.11) 0.004 2.04 (1.25–3.33) 0.004 
Ki-67 Continuous variable 1.06 (1.02–1.09) 0.003 Included in the form of Grade Included in the form of Grade Included in the form of Grade 
Previous treatment Yes (vs. No) 1.23 (0.83–1.82) 0.295 — — — — — — 
Type of treatment (during study period) Watch and wait 1 (Ref) — — — — — — — 
 SSA 0.68 (0.32–1.44) 0.318 — — — — — — 
 Targeted therapies 0.92 (0.42–2.02) 0.839 — — — — — — 
 Chemotherapy 0.92 (0.44–1.91) 0.824 — — — — — — 
 PRRT 2.46 (0.81–7.52) 0.113 — — — — — — 
Number of target lesions Cont variable 1.09 (0.98–1.23) 0.099 — — — — — — 
Tumor burden (sum all lesions) Cont variable 1.002 (0.99–1.005) 0.196 — — — — — — 
Type of examination MRI (vs CT) 1.14 (0.73–1.79) 0.560 — — — — — — 
TGR0 (%/month) ≥4 (vs <4) 1.96 (1.31–2.92) 0.001 1.68 (1.09–2.59) 0.018 Included in the form of change in ΔTGR3m-BL 1.43 (0.88–2.30) 0.147 
TGR3m (%/month) ≥0.8 (vs <0.8) 2.06 (1.37–3.09) 0.001 2.55 (1.64–3.97) <0.001 Included in the form of change in ΔTGR3m-BL 2.97 (1.84–4.78) <0.001 
ΔTGR3m-BL (x10 %/month) Cont variable 0.89 (0.79–0.99) 0.036 Included in the form of TGR0 and TGR3m 0.96 (0.83–1.04) 0.191 0.89 (0.78–1.01) 0.073 

First column shows data for univariate analysis. Second, third, and fourth columns show results from 3 different multivariable models, each one exploring different TGR combinations (TGR0 and TGR3m in Model 1; ΔTGR3m-BL in Model 2; TGR0, TGR3m and ΔTGR3m-BL in Model 3).

ΔTGR3m-BL, changes in TGR; Loc, localized; loc adv, locally advanced.

Validation of the impact of TGR3m on patient outcome (PFS)

Median PFS for patients eligible for Aim-2 (n = 97) was 16.19 months (95% CI, 11.69–18.59); median follow-up 22.99 months. Baseline characteristics for this patient population is available in Supplementary Table S1. When the previously identified TGR3m cut-off (11) was applied to these cohort, patient population was divided in 2 groups [TGR3m <0.8%/m (68.04%) and TGR3m ≥0.8%/m (31.96%)] with different outcomes. Patients with TGR3m ≥0.8 had shorter median PFS [12.15 months (95% CI, 5.58–16.19)] when compared with those with lower TGR3m [<0.8%/m; median PFS 18.59 months (95% CI, 10.77–20.92)]. Both TGR3m, TGR0 and tumor grade were significant in univariate Cox regression and included in multivariable Cox regression which confirmed high TGR3m [≥0.8%/m (vs. <0.8%/m)] to be an independent factor related with worse PFS (HR = 2.21; 95% CI, 1.32–3.70; P = 0.003; Fig. 5; Supplementary Table S2). Landmark analysis (excluding 6 patients) confirmed these findings [TGR3m (≥0.8%/m [vs. <0.8%/m]) multivariable HR 1.81 (95% CI, 1.04–3.15); P = 0.037].

Figure 5.

Kaplan–Meier curve. Validation of TGR3m as an early biomarker predictor for PFS in an external cohort (Aim-2). Adjusted Cox-regression multivariable analysis HR and P value are provided. Full survival analysis is available in Supplementary Table S2. m, months.

Figure 5.

Kaplan–Meier curve. Validation of TGR3m as an early biomarker predictor for PFS in an external cohort (Aim-2). Adjusted Cox-regression multivariable analysis HR and P value are provided. Full survival analysis is available in Supplementary Table S2. m, months.

Close modal

This study validates the use of TGR3m as an early radiologic biomarker with significant impact on PFS in an external cohort of patients with NET. The impact on PFS identified in this validation cohort was similar to the one previously described by our group (11); thus, validating that high TGR3m relates to shorter PFS in patients with NET in an independent series.

It is worth emphasizing that the median PFS achieved in patients with low TGR3m was shorter in this series compared with those in our previous study. This is likely to reflect that the patient population included in this GREPONET-2 study comprised a population with worse prognosis (shorter PFS and increased number of PFS events). The TGR3m could be helpful for tailoring the follow-up of patients with NET. Based on our findings, those with high TGR3m should undergo more frequent follow-up imaging because of their shorter PFS and increased risk of early progression.

We found that the prognostic role of TGR0 was diminished when the survival analysis was adjusted to TGR3m. TGR0 was found to be a significant factor in the univariate analysis but not in the multivariable analysis, showing that TGR3m reflects not only the tumor biology but also the treatment effect, thus having major clinical significance. Therefore, future studies should focus on TGR3m rather than analyzing TGR0 as an isolated factor.

Our study also explored the impact of treatment on TGR changes (ΔTGR3m-BL) by comparing paired TGR0 and TGR3m data. Changes in ΔTGR3m-BL were identified in patients receiving SSAs, targeted therapies and chemotherapy, as previously shown by other groups (14, 16). As expected, patients managed with WW had similar paired TGR0 and TGR3m. Our study lacked power to identify changes induced by PRRT since only 5 patients were in this subgroup. Interestingly, patients diagnosed with pancreatic NET and those with high TGR0 were more likely to achieve a more marked reduction in ΔTGR3m-BL likely to be reflection of more aggressive disease and therefore an increased antitumor treatment effect.

Based on the survival analysis, the impact of ΔTGR3m-BL on PFS was limited. We were, indeed, unable to identify a ΔTGR3m-BL cut-off with an impact on survival. Our findings showed that a high TGR3m was identified as an independent factor related to shorter PFS, suggesting that TGR3m does have more of a biological relevance than other TGR-derived parameters. We could therefore conclude that achieving a low TGR3m (<0.8%/m) has more relevant impact on survival, regardless of TGR0, ΔTGR3m-BL, or the treatment employed.

A significant proportion of patients would have been classified as stable disease by RECIST v1.1 at 3 months, despite showing a reduction of ΔTGR3m-BL likely to reflect some degree of treatment response (10, 20, 21). This was especially prevalent in the subgroup of patients treated with chemotherapy and targeted therapies. RECIST 1.1 and TGR thus provide different, and probably complementary, information for therapy monitoring (20). TGR is likely to be more informative to detect early subtle response shortly after initiation of treatment (10, 20) in contrast to RECIST 1.1 that at this early time point (3 months of treatment initiation) does not reflect in full the effect of therapy. The fact that TGR3m was significantly lower in patient who later achieved partial response by RECIST 1.1 supports that TGR3m may be used as an early indicator of favorable therapy outcome, not only in terms of PFS but also as an early marker of radiologic objective response (22). Studies in other cancer types also support the role of early changes in TGR as an indicator of patient outcome (13) and also support the role of TGR to identify early progression (23, 24). Further studies are however required to establish how TGR and RECIST 1.1 may be best combined to optimize therapy monitoring. The incorporation of TGR into ongoing clinical trials would provide a robust set of data for such exploratory endpoints (20).

Even though TGR has already shown to be reproducible regardless of type of imaging, reader and site of target lesions used (liver vs. nonliver), some of the disadvantages inherent to TGR apply (25). One of such limitations is the fact that “nontarget” and “new” lesions are not accounted for in the TGR calculation. In addition, other limitations of our study include its retrospective design and the potential existence of a selection bias introduced by the strict inclusion criteria, especially from the imaging requirements point of view. The fact that the study population was not limited to treatment-naïve patients is a limitation worth mentioning. It is likely that this not only incorporated selection bias, but also impacted on the significance of TGR0 for non–treatment-naïve patients. In addition, due to limited sample size, especially for treatment-type-related subgroup analysis, statistical power was limited and could explain, for example, why ΔTGR3m-BL differences between treatment groups did not reach statistical significance. Further prospective studies, together with exploratory analyses of data collected within ongoing clinical NET trials would be ideal for further understanding the role and applicability of TGR. Further research in this topic would also benefit from focusing on a more homogeneous population to the one presented in this study, especially from the time between pre-baseline and baseline imaging (wide time period of 12 months was allowed in this study), type of treatment (multiple type of therapies were included in this work), and line of treatment perspective (future studies should focus on treatment-naïve population).

In summary, our findings confirm the prognostic role of TGR3m as an early radiologic biomarker, which is able to predict not only PFS but also radiologic therapy response. TGR3m is likely to be more informative than RECIST 1.1 for early treatment evaluation at 3 months after treatment initiation, to identify subtle changes in tumor growth and treatment activity. The potential role of TGR3m in the treatment decision needs to be further explored in future prospective studies.

A. Lamarca reports commercial research grants from Ipsen; reports speakers bureau honoraria from Pfizer, Ipsen, and Incyte; is a consultant/advisory board member for Ipsen, Eisai, and Nutricia; and reports other remuneration from Ipsen, Pfizer, Bayer, AAA, SirtEx, Novartis, and Delcath. J. Crona reports receiving other remuneration from Novartis and Ipsen. C. Lopez Lopez reports receiving commercial research grants and speakers bureau honoraria from Ipsen, Novartis, and Pfizer, and is a consultant/advisory board member for Ipsen. L. de Mestier is a consultant/advisory board member for Ipsen, Pfizer, and Novartis. M. Pavel is a consultant/advisory board member for Ipsen, Novartis, Lexicon, and Pfizer. C. Dromain is a consultant/advisory board member for Ipsen. No potential conflicts of interest were disclosed by the other authors.

Conception and design: A. Lamarca, M. Ronot, M. Opalinska, C. Lopez Lopez, H. Vidal Trueba, L. de Mestier, F. Costa, M. Pavel, C. Dromain

Development of methodology: A. Lamarca, M. Ronot, M. Opalinska, C.L. Lopez, L. de Mestier, F. Costa, M. Pavel, C. Dromain

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A. Lamarca, M. Ronot, J. Crona, C. Lopez Lopez, D. Pezzutti, P. Najran, L. Carvhalo, R.O.F. Bezerra, P. Borg, L. de Mestier, N. Scaefer, E. Baudin, A. Sundin, F. Costa, C. Dromain

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A. Lamarca, M. Ronot, S. Moalla, D. Pezzutti, P. Najran, P. Borg, E. Baudin, F. Costa, M. Pavel, C. Dromain

Writing, review, and/or revision of the manuscript: A. Lamarca, M. Ronot, S. Moalla, J. Crona, M. Opalinska, C. Lopez Lopez, P. Najran, R.O.F. Bezerra, N. Vietti Violi, L. de Mestier, N. Scaefer, A. Sundin, M. Pavel, C. Dromain

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A. Lamarca, M. Ronot, M. Opalinska

Study supervision: A. Lamarca, F. Costa, M. Pavel

A. Lamarca received partial funding from ASCO Conquer Cancer Foundation Young Investigator Award and The Christie Charity. The authors thank Ipsen and Solaris for supporting The Knowledge Network initiative that facilitated this international collaboration.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
Dasari
A
,
Shen
C
,
Halperin
D
,
Zhao
B
,
Zhou
S
,
Xu
Y
, et al
Trends in the incidence, prevalence, and survival outcomes in patients with neuroendocrine tumors in the United States
.
JAMA Oncol
2017
;
3
:
1335
42
.
2.
Auernhammer
CJ
,
Spitzweg
C
,
Angele
MK
,
Boeck
S
,
Grossman
A
,
Nölting
S
, et al
Advanced neuroendocrine tumours of the small intestine and pancreas: clinical developments, controversies, and future strategies
.
Lancet Diabetes Endocrinol
2018
;
6
:
404
15
.
3.
Wolin
EM
,
Leyden
J
,
Goldstein
G
,
Kolarova
T
,
Hollander
R
, et al
Patient-reported experience of diagnosis, management, and burden of neuroendocrine tumors: results from a large patient survey in the United States
.
Pancreas
2017
;
46
:
639
47
.
4.
Singh
S
,
Chan
DL
,
Moody
L
,
Liu
N
,
Fischer
HD
,
Austin
PC
, et al
Recurrence in resected gastroenteropancreatic neuroendocrine tumors
.
JAMA Oncol
2018
;
4
:
583
5
.
5.
Lloyd
RV
, et al
WHO classification of tumours: pathology and genetics of tumours of endocrine organs
. Vol. 4th ed. 
2017
:
Lyon
:
IARC
; 
2017
.
6.
Bosman
FT
, F Carneiro, RH Hruban, ND Theise.
WHO classification of tumours of the digestive system
.
World Health Organization
; 2010.
7.
Klöppel
G
,
CA
,
Hruban
RH
, et al
Neoplasms of the neuroendocrine pancreas
.
In: WHO classification of tumours of the endocrine organs
.
Lyon
:
IARC Press
; 
2017
. p.
210
239
.
8.
Eisenhauer
EA
,
Therasse
P
,
Bogaerts
J
,
Schwartz
LH
,
Sargent
D
,
Ford
R
, et al
New response evaluation criteria in solid tumours: revised RECIST guideline (version 1.1)
.
Eur J Cancer
2009
;
45
:
228
47
.
9.
Lamarca
A
,
Barriuso
J
,
Kulke
M
,
Borbath
I
,
Lenz
HJ
,
Raoul
JL
, et al
Determination of an optimal response cut-off able to predict progression-free survival in patients with well-differentiated advanced pancreatic neuroendocrine tumours treated with sunitinib: an alternative to the current RECIST-defined response
.
Br J Cancer
2018
;
118
:
181
8
.
10.
Ferté
C
,
Fernandez
M
,
Hollebecque
A
,
Koscielny
S
,
Levy
A
,
Massard
C
, et al
Tumor growth rate is an early indicator of antitumor drug activity in phase I clinical trials
.
Clin Cancer Res
2014
;
20
:
246
52
.
11.
Lamarca
A
,
Crona
J
,
Ronot
M
,
Opalinska
M
,
Lopez Lopez
C
,
Pezzutti
D
, et al
Value of tumor growth rate (TGR) as an early biomarker predictor of patients' outcome in neuroendocrine tumors (NET)—the GREPONET study
.
Oncologist
2019
;
24
:
1
10
.
12.
Gomez-Roca
C
,
Koscielny
S
,
Ribrag
V
,
Dromain
C
,
Marzouk
I
,
Bidault
F
, et al
Tumour growth rates and RECIST criteria in early drug development
.
Eur J Cancer
2011
;
47
:
2512
6
.
13.
Ferté
C
,
Koscielny
S
,
Albiges
L
,
Rocher
L
,
Soria
JC
,
Iacovelli
R
, et al
Tumor growth rate provides useful information to evaluate sorafenib and everolimus treatment in metastatic renal cell carcinoma patients: an integrated analysis of the TARGET and RECORD phase 3 trial data
.
Eur Urol
2014
;
65
:
713
20
.
14.
Dromain
C
,
Pavel
ME
,
Ruszniewski
P
,
Langley
A
,
Massien
C
,
Baudin
E
, et al
Tumor growth rate as a metric of progression, response, and prognosis in pancreatic and intestinal neuroendocrine tumors
.
BMC Cancer
2019
;
19
:
66
.
15.
Caplin
ME
,
Pavel
M
,
Ćwikła
JB
,
Phan
AT
,
Raderer
M
,
Sedláčková
E
, et al
Lanreotide in metastatic enteropancreatic neuroendocrine tumors
.
N Engl J Med
2014
;
371
:
224
33
.
16.
Ito
T
,
Honma
Y
,
Hijioka
S
,
Kudo
A
,
Fukutomi
A
,
Nozaki
A
, et al
Phase II study of lanreotide autogel in Japanese patients with unresectable or metastatic well-differentiated neuroendocrine tumors
.
Invest New Drugs
2017
;
35
:
499
508
.
17.
Palazzo
M
,
Lombard-Bohas
C
,
Cadiot
G
,
Matysiak-Budnik
T
,
Rebours
V
,
Vullierme
MP
, et al
Ki67 proliferation index, hepatic tumor load, and pretreatment tumor growth predict the antitumoral efficacy of lanreotide in patients with malignant digestive neuroendocrine tumors
.
Eur J Gastroenterol Hepatol
2013
;
25
:
232
8
.
18.
Durante
C
,
Boukheris
H
,
Dromain
C
,
Duvillard
P
,
Leboulleux
S
,
Elias
D
, et al
Prognostic factors influencing survival from metastatic (stage IV) gastroenteropancreatic well-differentiated endocrine carcinoma
.
Endocr Relat Cancer
2009
;
16
:
585
97
.
19.
Stata.v.12., StataCorp
. 
2011
.
Stata Statistical Software: Release 12
.
College Station, TX
:
StataCorp LP
.
20.
Milella
M
. 
Optimizing clinical benefit with targeted treatment in mRCC: "Tumor growth rate" as an alternative clinical endpoint
.
Crit Rev Oncol Hematol
2016
;
102
:
73
81
.
21.
Dienstmann
R
,
Tabernero
J
. 
TGR analysis in phase I clinical trials–letter
.
Clin Cancer Res
2014
;
20
:
2495
6
.
22.
El-Maraghi
RH
,
Eisenhauer
EA
. 
Review of phase II trial designs used in studies of molecular targeted agents: outcomes and predictors of success in phase III
.
J Clin Oncol
2008
;
26
:
1346
54
.
23.
Ferrara
R
,
Mezquita
L
,
Texier
M
,
Lahmar
J
,
Audigier-Valette
C
,
Tessonnier
L
, et al
Hyperprogressive disease in patients with advanced non-small cell lung cancer treated with PD-1/PD-L1 inhibitors or with single-agent chemotherapy
.
JAMA Oncol
2018
;
4
:
1543
52
.
24.
Champiat
S
,
Ferrara
R
,
Massard
C
,
Besse
B
,
Marabelle
A
,
Soria
JC
, et al
Hyperprogressive disease: recognizing a novel pattern to improve patient management
.
Nat Rev Clin Oncol
2018
;
15
:
748
62
.
25.
Dromain
C
, et al
Additional value of tumour growth rate in patients diagnosed with well-differentiated neuroendocrine tumours
.
Abstract 54303, ESGAR
2018
.