Prospective Study of FLT PET for Early Interim Response Assessment in Advanced Stage B-cell Lymphoma

Abstract

Purpose

Current clinical and imaging tools remain suboptimal for early assessment of prognosis and treatment response in aggressive lymphomas. Positron emission tomography (PET) with ¹⁸F-fluorothymidine (FLT) can be used to measure tumor cell proliferation and treatment response. In a prospective study in patients with advanced stage B-cell lymphoma we investigated the prognostic and predictive value of FLT PET in comparison to standard imaging with FDG PET and clinical outcome.

Patients and Methods

65 patients were treated with an induction/consolidation regimen consisting of four cycles of R-CHOP-14 followed by 3 cycles of ICE. FLT PET was performed at baseline and at interim (iPET) after 1–2 cycles of therapy. FDG PET was performed at baseline, after cycle 4, and at the end of therapy. The relationship between PET findings, progression free survival (PFS) and overall survival (OS) was investigated.

Results

With a median follow-up of 51 months, PFS and OS were 71% and 86% respectively. FLT iPET, analyzed visually, using a 5-point score, or semi-quantitatively, using SUV and ΔSUV, predicted both PFS and OS (p<0.01 for all parameters). Residual FLT SUVmax on iPET was associated with an inferior PFS (HR: 1.26, p=0.001) and OS (HR: 1.27, p=0.002). Using FDG PET, findings in the end of treatment scan were better predictors of PFS and OS than findings on interim scan. Baseline PET imaging parameters, including SUV, proliferative or metabolic tumor volume, did not correlate with outcome.

Conclusion

FLT PET after 1–2 cycles of chemotherapy predicts PFS and OS, and a negative FLT iPET may potentially help design risk-adapted therapies in patients with aggressive lymphomas. In contrast, the positive predictive value of FLT iPET remains too low to justify changes in patient management.

Over the past decade, several changes in the management of diffuse large B-cell lymphoma (DLBCL) have occurred. For instance, the addition of the chimeric monoclonal antibody rituximab (R) to the standard CHOP (cyclophosphamide, doxorubicin, vincristine, prednisone) chemotherapy backbone has improved 5-year overall survival (OS) by approximately 15% (1,2). However, DLBCL shows substantial heterogeneity in its clinical behavior, and new management strategies, including the early identification of poor responders, are needed to improve patient outcome.

The role of interim PET (iPET) with FDG after a few cycles of chemotherapy in identifying patients with poor outcome has been investigated (3–9). Our group studied the utility of FDG iPET as part of a chemotherapy program consisting of induction with R-CHOP-14 followed by consolidation with ICE (ifosfamide, carboplatin, etoposide) (4). In that study, we biopsied sites with residual FDG uptake on iPET after 4 cycles of chemotherapy. All patients with a negative biopsy (or a negative iPET) received consolidation with 3 cycles of ICE, whereas patients with positive biopsy received RICE for 3 cycles followed by high dose therapy and autologous stem cell rescue. Notably this study demonstrated a high rate of false positive FDG iPET, which was also described by other groups (10). In the current study we, therefore, investigated if PET imaging with the proliferation marker ¹⁸F-fluorothymidine (FLT) (11) after 1 or 2 cycles of therapy could provide better prognostic and predictive information than FDG PET does in patients with advanced stage large cell lymphoma.

MATERIALS AND METHODS

Study Design And Patient Population

After providing written informed consent, 65 patients with advanced stage, CD20-positive DLBCL, primary mediastinal large B-cell lymphoma, or follicular lymphoma grade 3B were enrolled in a prospective study (http://clinicaltrials.gov/show/NCT00712582). The Institutional Review Board approved the study. All patients were eligible for autologous transplantation and met inclusion criteria as reported previously (4). Patients underwent staging with contrast enhanced CT (covering chest, abdomen, and pelvis) and FDG-PET/CT (444 ± 44 MBq). Additional FDG scans were performed after 4 cycles and 6 weeks after completion of chemotherapy. Plasma glucose prior to FDG injections was 92±15 mg/dl. FLT-PET/CT (296 ± 30 MBq) was performed at baseline and at interim (cohort 1 post cycle 1, planned n=30; cohort 2 post cycle 2, planned n=30). Initial therapy for all patients consisted of dose-dense R-CHOP-14 for 3 cycles (rituximab 375 mg/m² day 1 and 3, cyclophosphamide at 1000 mg/m², doxorubicin 50 mg/m², vincristine 1.4mg/m² [uncapped] and prednisone 100 mg oral daily days 2–6). Cycle 4 consisted of CHOP alone (same doses) and was followed by FDG iPET. Consolidation consisted of ICE chemotherapy (3) for patients with an initial proliferation index <80%, whereas patients with proliferation index ≥ 80% received augmented RICE (rituximab 375 mg/m2 day1, etoposide 200 mg/m2 every 12 hours × 3 doses, ifosfamide 5 g/m2/day by continuous infusion × 2, carboplatin AUC 5 day 3 [maximum 800 mg]) for 3 cycles.

PET/CT Imaging And Analysis

Patients fasted for 6 hours prior to radiotracer injection. PET/CTs from mid skull to upper thighs were obtained 60 min after injection on GE-Discovery scanners and were analyzed using the PET VCAR program (GE Healthcare). Volumetric regions of interest were placed in reference regions (mediastinal blood pool and normal liver) and over all sites of abnormal uptake in lymph nodes, soft tissue organs, or bones. All scans were read by two nuclear medicine physicians (HS, SG). For visual analysis of FDG iPET we employed a 5 point scale (12); grades 1–3 were considered normal. Grades 4 and 5 were considered abnormal, defining grade 4 as uptake considerably (approximately 20%) higher than liver and grade 5 as increase in number or intensity of FDG-avid lesions. Abnormal FLT uptake was defined as intensity higher than that in surrounding local background. FLT response was classified as: grade 1, no residual uptake; grade 2, uptake ≤ local background; grade 3, uptake slightly > local background; grade 4, improvement from baseline with at least one lesion showing uptake clearly higher than local background; grade 5, increase in lesion size, number or intensity. Grades 1–3 were considered complete response, grade 4 partial response, and grade 5 progression. FLT analysis was restricted to lymph nodes and soft tissue organs other than liver (high physiologic uptake in liver and bone marrow precludes reliable assessment of these organs). The metabolic tumor volume (volume of FDG-avid lymphoma, MTV) and total proliferative volume (volume of FLT-avid lymphoma, TPV) were defined using PET VCAR and applying a 42% threshold (13). ΔSUVmax ([SUV_baseline – SUV_interim]/SUV_baseline) was calculated by (a) comparing a single disease site with the highest uptake in each scan (9) and (b) comparing the summed SUV of the 5 most avid disease sites in each scan. Bone lesions were excluded from SUV-based analysis to enable head to head comparison of FDG and FLT findings.

Statistical Analysis

Overall survival (OS) time was calculated from the start of treatment until the date of death or last follow-up. Progression-free survival (PFS) was calculated from the start of treatment until the date of disease progression, death, or last follow-up. Survival analysis was performed using the Kaplan–Meier method for dichotomous variables and compared by using the log-rank test. Cox proportional hazard regression was used to assess continuous variables. For all iPET scans, patients were dichotomized by residual uptake: grade 1–3 versus grade 4 and 5. A sensitivity analysis was performed to identify the thresholds for change in SUV and/or maximum SUV. This was performed by using maximal χ² test for progression-free survival and/or overall survival. We used Kaplan-Meier analysis with the log rank test to examine the relationship between PFS, OS, and the PET 5-point scale response. Due to limited power, multivariate survival analyses could not be performed. P-values < 0.05 were considered significant. Analyses were performed using SPSS Statistical Software (version 22.0, 2013, IBM Corp.) or R software (version 3.0.1, R-Foundation for Statistical Computing, Vienna, Austria).

RESULTS

Patient Characteristics

All 65 patients enrolled to the protocol were analyzed for PFS and OS (Table 1). Two patients progressed prior to FDG iPET, one patient progressed between iPET and end of planned therapy, and one patient developed drug related toxicity. The remaining 61 patients completed treatment according to protocol. Accordingly, 63 patients were analyzable for FDG iPET, and 61 patients for FDG end of treatment PET. For logistic reasons (unavailability of radiotracer, scheduling difficulty in patients with rapidly progressive disease, or withdrawn consent) only 52 patients underwent FLT scanning at baseline, and 55 underwent interim FLT scanning.

Table 1.

Clinical and Demographic Characteristics

Variable	Value
Gender
Female	36 (55%)
Male	29 (45%)
Median Age (years)	54.7 (range: 20.8–71.7)
Stage
IIX	9 (14%)
III–IV	56 (86%)
Ki-67
Median	70%
Range	20–90%
< 80%	43 (68%)
≥80%	22 (32%)
IPI Score
Low	8 (12.3%)
Low-intermediate	16 (24.6%)
Intermediate-high	21 (32.3%)
High	20 (30.8%)
Extranodal sites
> 1	39 (60%)
Bone marrow biopsy positive	16 (25%)
Histology
DLBCL	51 (78%)
PMBCL	13 (20%)
Follicular 3B	1 (2%)

Values outside parentheses are medians for continuous variables and counts for categorical variables. PMBL – primary mediastinal B-cell lymphoma. X denotes diameter > 10cm.

Baseline Imaging Characteristics

All patients had FDG-avid disease (Table 2). FLT uptake was always lower than FDG uptake, regardless of disease site. Discrepant findings (FDG+/FLT-) were observed for 90 of 2,860 regions (3%), mainly because disease in bones (69 regions) or liver (6 regions) was not detected on FLT scan. Other regions of FDG+/FLT− disease were noted in lymph nodes (n=11), lungs (n=3), and spleen (n=1). However, all of these patients presented with multiple other disease sites that were clearly FLT-positive. There were no patients with disease only recognized on FLT PET/CT.

Table 2.

PET imaging findings

Parameter	Time Point	N	Median	Range
FLT single SUVmax [g/ml]	Baseline	52	13.0	2.7 – 41.4
	Interim	55	2.5	1. 2 – 17.2
FLT sum of hottest 5 SUVmax [g/ml]	Baseline	52	32.3	2.7 – 127.2
	Interim	55	7.6	1.2 – 68.0
FLT ΔSUVmax [%]	Baseline to Interim	50	↓77.1%	↑0.6% – ↓92.7%
FLT TPV* [cm3]	Baseline	52	139.4	1.3 – 2,080.9
FDG single SUVmax [g/ml]	Baseline	65	23.4	6.1 – 48.0
	Interim	63	2.3	1.4 – 29.6
	Final	61	2.0	1.2 – 25.7
FDG sum of hottest 5 SUVmax [g/ml]	Baseline	65	59.3	8.5 – 192.4
	Interim	63	7.8	1.4 – 72.9
	Final	61	6.9	1.8 – 98.4
FDG ΔSUVmax [%]	Baseline to Interim	63	↓ 88.9%	↑0.3% – ↓96.4%
	Baseline to Final	61	↓89.8%	↓74.4 – ↓96.1%
FDG MTV [cm3]	Baseline	65	225.6	8.4 – 3,452.6

^*excluding bone and liver disease, for which volumes cannot be measured accurately due to high physiologic background activity

Interim PET/CT During Chemotherapy

Fifty-five patients underwent FLT iPET after one cycle (n=27) or after 2 cycles (n=28) of R-CHOP. Time between end of cycle and FLT iPET was 9.8 ± 1.9 days for cohort 1 and 9.4 ±1.2 days for cohort 2. Complete proliferative response (CPR) was observed in 29 patients (grade 1, n= 21; grade 2, n= 7; grade 3, n=1), including three patients without baseline FLT PET: since aggressive lymphoma is FL-avid at baseline (14–16), lack of uptake in residual nodes on iPET was considered CPR. The other 26 patients showed partial response (grade 4, n= 22) or progression (grade 5, n= 4). CPR tended to be more common after cycle 2 (78% versus 44% after cycle 1, Suppl. Table 1). However, differences were not significant and therefore data were analyzed jointly. Eighteen patients showed diffusely increased splenic uptake attributable to recent GCSF administration. The median FLT ΔSUV was 77%.

On FDG iPET, performed 18±1 days after cycle 4, complete metabolic response (CMR) was noted in 38 patients (grade 1, n=24, grade 2, n=10, and grade 3, n=4). Grade 4 (partial response) was noted in 24 patients, and grade 5 (progression) in one patient. The median FDG ΔSUV was 89%. Among patients with FDG uptake grade 1–3, the median ΔSUV was 89% (range: −74% to −96%) as compared 80% (range +7.3% to −95%) in patients with grades 4 or 5. Similar numbers were derived when analyzing the 5 lesions with the most intense uptake at baseline and at interim in each patient (median ΔSUV of 76% and 88% for FLT and FDG respectively).

Biopsy Findings And Patient Management After Interim FDG Scan

Twenty-one patients with persistent FDG uptake on iPET (SUV range 2.1–27.4) underwent biopsy to verify imaging findings. Eleven of these sites were residual mediastinal or mesenteric masses. Residual disease was only proven in two patients (SUV 5.3 and 27.4); the other 19 biopsies (SUV 2.1 – 6.6) showed inflammation only. Only the 2 patients with a positive biopsy underwent high dose therapy and ASCT; all remaining patients received ICE-based consolidation only. Only six of the 21 patients had a positive FLT iPET (one biopsy-positive, five biopsy-negative).

Survival Analysis And Assessment Of Prognostic Factors

At a median follow-up for surviving patients of 51 months (range: 35–71 months), the Kaplan-Meier estimates for the proportion of patients alive and progression-free were 87% and 71% respectively (Fig.1). None of the clinical parameters were associated with outcome including: age, gender, B-symptoms, elevated LDH, poor KPS, stage, extranodal site involvement (> 1 site), Ki-67 or IPI score.

Figure 1.

Progression-free and overall survival for the entire patient population

Findings on baseline FLT or FDG PET (in particular SUVmax, FDG-MTV and FLT-TPV) were not associated with patient outcome. Further analysis was therefore focused on PET response parameters. Tables 3 and 4 show the relationships between response on FLT iPET and patient outcome. FLT uptake by visual analysis (grade 1–3 versus grade 4–5) was predictive of PFS and OS (Fig. 2). A χ² analysis (Table 4) revealed optimal cut-points for residual SUV (4.6) and ΔSUV(36%). Analysis of PFS according to these cut-points was highly predictive (Fig. 3). Nevertheless, in view of the small sample size these cut-points should be interpreted with caution and require further validation.

Table 3.

Univariate analysis for progression-free survival

Parameter	HR	P-value
FLT single SUVmax at baseline	0.99	0.91
at interim	1.26	0.001
FLT sum of hottest 5 SUVmax, at baseline	1.003	0.633
at interim	1.063	0.001
FLT ΔSUVmax baseline-interim	0.966	0.001
FLT PTV baseline	1.0	0.84
FDG single SUVmax, at baseline	0.98	0.53
at interim	1.14	0.021
in final scan	1.18	0.023
FDG sum of hottest 5 SUV max, at baseline	1.00	0.930
at interim	1.057	0.014
in final scan	1.048	0.010
FDG ΔSUVmax baseline-interim	0.97	0.021
FDG ΔSUVmax baseline-final	0.96	0.015
FDG MTV baseline	1.00	0.91

Table 4.

Maximal chi-square statistics for progression free survival

Variable	Estimated Cut-Point	M	P-value
FLT single SUVmax, at baseline	15.5	1.07	0.936
at interim	4.6	3.58	0.003
FLT sum of hottest 5 SUVmax at baseline	24.3	2.52	0.119
at interim	6.05	3.65	0.004
FLT ΔSUVmax	35.6	3.57	0.004
FDG single SUVmax, at baseline	37.9	1.64	0.625
at interim	3.8	1.86	0.327
FDG sum of hottest 5 SUVmax, at baseline	34.9	1.87	0.44
at interim	7.8	2.52	0.108
FDG ΔSUVmax	83.1	1.99	0.344

Figure 2.

PFS (A) and OS (B) as stratified by 5-point visual score on FLT iPET

Figure 3.

PFS as stratified by SUVmax on FLT iPET (A) and by FLT ΔSUV (B)

Regarding FDG iPET, residual uptake (visual grade 1–3 versus grade 4–5) predicted PFS (Fig. 4) but not OS (not shown). To evaluate the impact of ΔSUV we analyzed both PFS and OS at the median. There was a significant improvement in PFS (Table 3) and a trend to improved OS for patients with ΔSUV greater than the median. Similar results were obtained when analyzed by residual SUVmax on iPET: patents with SUVmax below median experienced better PFS and a trend to better OS. We then investigated if a cutoff could be determined that optimized the prognostic significance of ΔSUV delta and SUVmax (Table 4). Of note, analyzing a total of 5 lesions in each pair of scans was no more informative or predictive than analysis confined to the single hottest lesion per scan. We also evaluated the prognostic value of the previously proposed post cycle 4 FDG ΔSUV 77% (9). In our data set, only 4 patients showed a ΔSUV < 77%. Therefore, neither this cut-off nor other proposed cut-offs for FDG iPET provided meaningful separation of prognostic groups. We then applied the estimated cut-point for FDG ΔSUV (83%) from our maximal X²analysis (M: 1.99, p=0.34). Thus, while FDG ΔSUV predicted PFS in the univariate analysis, we could not identify a clear cut-off to separate prognostic groups because there was considerable overlap: ΔSUV was 89% in patients without progression (range: −74% to −97%) and 76% in patients who progressed (range: +7% to −94%).

Figure 4.

PFS as stratified but 5-point score on FDG iPET

Among parameters analyzed for the end of treatment FDG scan, residual SUV and ΔSUV from baseline to final scan were both associated with both PFS (HR: 1.18 and 0.96 respectively, each p<0.05) and OS (HR: 1.20 and 0.96 respectively, each p<0.05). The final visual score was predictive of PFS (p=0.03) but not OS.

DISCUSSION

In the current study, early FLT iPET had a high negative predictive value (NPV), with a negative scan clearly identifying patients with good prognosis. This information might help optimizing risk-adapted therapy for patients with advanced stage aggressive lymphoma. In contrast, the positive predictive value (PPV) of FLT iPET, although somewhat better than the PPV for FDG iPET, remains too low to justify changes in patient management. Contrary to expectation and suggestions in the literature, volumetric parameters (FLT-TPV, FDG-MTV) were not associated with patient outcome when using our induction/consolidation treatment regimen.

FLT is a proliferation marker for PET imaging (11), with high correlation between Ki-67 and FLT-SUV reported in lymphoma (17). One might therefore expect high baseline FLT-SUV or high FLT-TPV to be markers of poor prognosis, but our findings do not support this hypothesis. However, visual inspection of residual FLT uptake on iPET predicted both PFS and OS. Quantitative parameters, such as residual FLT SUV and ΔSUV on iPET, also predicted patient outcome, although cut-points identified in χ² analysis require independent validation in a larger dataset. The optimal timing of FLT iPET remains to be determined. Since no prior prospective study had identified an optimal time point for iPET with FLT, we investigated two early time points and studied two cohorts. CPR tended to be more common after cycle 2; thus, while complete resolution of uptake can be observed earlier, scanning after cycle 2 may be more meaningful for risk stratification. Regardless, early response assessment with FLT iPET may potentially inform patient management. For example, patients with negative FLT iPET can very likely be cured with standard chemotherapy. In fact, if the high NPV of FLT iPET is confirmed in future studies, some patients with high ΔSUV might potentially be cured with an abbreviated chemotherapy regimen, maintaining excellent outcome but reducing treatment related toxicity. Thus, FLT PET may be useful to guide a de-escalation of therapy strategy. An early negative FLT iPET may also be reassuring and help eliminate biopsies of sites with residual FDG uptake on iPET as this uptake is very likely false positive. In contrast, the PPV of FLT does not appear sufficiently high to justify any escalation of therapy strategy (including therapy with novel agents) without biopsy confirmation. Regardless, FLT iPET appears more accurate (in particular post cycle 2) in predicting patient outcome than FDG PET is. Of note, a recent multicenter study showed that early iPET with FDG after one cycle of therapy cannot safely discriminate prognostic groups in large cell lymphoma (18).

Experience with FLT PET in lymphoma is limited. In some studies (14,15) FLT ΔSUV on iPET after one cycle emerged as a predictor of survival, but baseline FLT-SUV did not. An optimal cut-point was not reported. In another study including 50 patients with DLBCL (16), residual SUV > 1.9 and ΔSUV< 65% after one cycle of R-CHOP identified patients with worse PFS and OS in ROC analysis. We suggest caution when interpreting these suggested cut-points. Larger studies and standardization of time points and imaging techniques are required before any particular SUV number of ΔSUV can be applied for clinical decision making. It is difficult to compare our study to these prior investigations because of differences in patient population, lack of established criteria for interpretation of FLT iPET, and differences in statistical approaches to data interpretation. Nevertheless, the overall (albeit limited) evidence suggests FLT iPET as a promising marker for early response assessment in aggressive lymphomas.

FDG iPET is used routinely to assess treatment response in DLBCL (9,19). In some studies, iPET provided meaningful prognostic information (3,5,8,9), but in other studies, the end of treatment scan proved more informative (6,7,20). We previously (4) reported a low PPV of FDG iPET, at least in part resulting from rituximab-induced inflammation in the R-CHOP regimen (10) and from the use of iPET interpretation criteria that are now obsolete (12,21). In the current study, FDG iPET again showed limited PPV, despite altering the drug regimen and timing of iPET vis-à-vis preceding cycle, and despite using updated iPET interpretation criteria. Higher PPV were reported with standard regimens that are probably less effective (9), but FDG iPET remains a suboptimal test in the setting of an induction/consolidation regimen. In contrast, FDG PET performed after completion of chemotherapy provided better prognostic information.

Measuring ΔSUV may improve the predictive value and interobserver agreement in reading FDG iPET (5,8,9,22). Cut-offs of 66% for iPET after 2 cycles (5), and 70%–92% after 4 cycles (5,9,23,24) have been proposed. We could not confirm any of these cut-points in the setting of our induction/consolidation regimen. While ΔSUV appears highly reproducible, visual assessment remains the first essential step in scan interpretation; ΔSUV should not be used in isolation.

There is growing interest in studying volumes of PET-avid disease (25–28). Direct comparison of published data is hampered by considerable heterogeneity in methodologies and patient populations. For instance, Sasanelli et al. (26) segmented FDG-MTV in 114 patients with large cell lymphoma using a 41% threshold. Their raw data (median, 315ml; range, 4 – 2,650ml) are similar to our findings. However, whereas an MTV of 550 ml, derived from ROC analysis, proved significant in their analysis, we did not find any association between median MTV and patient outcome. Similarly, FLT-TPV did not prove prognostic in our analysis. It is conceivable that any prognostic value TPV might convey in other settings was overcome by the effectiveness of our induction/consolidation therapy regimen.

Our prospective study has some limitations. For logistic reasons, not all patients underwent FLT PET as planned, and FLT and FDG iPET were performed at different time points. Performing additional FDG iPET after one or two cycles was not feasible for logistic and financial reasons and because of dosimetry concerns. The overall sample size was also limited by current cost for FLT and funding. FLT was of limited use for disease in bone and liver (because of high physiologic uptake); nevertheless, FLT predicted for early failure in a setting where FDG did not (4). Although FLT is widely available from commercial vendors, it remains an investigational agent.

CONCLUSION

FLT iPET was a clinically meaningful predictor of treatment response, which may potentially help design risk-adapted therapies in patients with aggressive lymphomas.