Chemotherapy Response Evaluation with ¹⁸F-FDG PET in Patients with Non-Small Cell Lung Cancer

Patient Eligibility Criteria

Between March 2002 and December 2005, all patients in the Radboud University Nijmegen Medical Centre with any stage of NSCLC and measurable tumor lesions according to the Response Evaluation Criteria in Solid Tumors (RECIST)—who were scheduled to undergo induction chemotherapy or chemotherapy in a palliative setting—were offered the opportunity to participate in the study. Initial staging was performed according to the guidelines of the American Society of Clinical Oncology (11). Exclusion criteria were diabetes mellitus and pregnancy. In all patients, treatment decision making was done by a multidisciplinary team, including cardiothoracic surgeons, pulmonologists, medical oncologists, radiation oncologists, pathologists, radiologists, and nuclear medicine physicians. Clinicians had access to the pretreatment staging static whole-body PET scan but were unaware of the results of the dynamic ¹⁸F-FDG PET scans. The study was approved by the Institutional Review Board of the Radboud University Nijmegen Medical Centre, and written informed consent was obtained from each patient.

Quantitative Dynamic ¹⁸F-FDG PET Data Acquisition and Reconstruction

Dynamic ¹⁸F-FDG PET was performed at baseline and after the second or third cycle of chemotherapy, depending on the chemotherapy regimen. The median interval between the clinical-staging ¹⁸F-FDG PET and the baseline dynamic ¹⁸F-FDG PET was 3 d. Patients fasted for at least 6 h before imaging. Intake of noncaloric beverages was permitted. All scans were acquired on an ECAT-EXACT47 PET scanner (Siemens Medical Solutions). The field of view for dynamic acquisition was based on whole-body ¹⁸F-FDG PET and CT scans performed for initial staging, including as many measurable tumor lesions as possible. A 20-min transmission scan was made, using the internal ⁶⁸Ge/⁶⁸Ga sources, to correct for photon attenuation. Approximately 200 MBq ¹⁸F-FDG were injected intravenously over a 1-min period, using a constant-infusion remote-controlled pump (Medrad). The dynamic data acquisition, performed in septa-extended (2-dimensional) mode, was started simultaneously with the injection of ¹⁸F-FDG and consisted of 16 time frames with variable frame length (10 × 30 s, 3 × 300 s, 3 × 600 s) for a total time of 50 min. A more detailed description of the data acquisition, reconstruction, and analysis methods was published previously (12).

Plasma Time–Activity Curves

To measure the blood clearance of ¹⁸F-FDG, plasma time–activity curves were derived from arterial blood sampling when feasible. Arterial blood samples were taken manually at set times to provide an arterial plasma input function as described previously (12). When arterial cannulation was contraindicated or not feasible, an image-based input function was determined by measuring ¹⁸F-FDG counts in a volume of interest (VOI) over the ascending aorta, which accurately determines ¹⁸F-FDG blood levels (12).

Tumor Time–Activity Curves

Tumor time–activity curves were obtained by placing VOIs semiautomatically over the tumor and metastases using a threshold of 50% of the maximum pixel value within the lesion. Necrotic areas were excluded from the VOIs. A volume-weighted mean value of tumor glucose use (metabolic rate of glucose [MR_glu]) of all lesions in each PET scan was derived to provide one value for each study. VOIs drawn on the first ¹⁸F-FDG PET (before initiation of chemotherapy) were copied to the second ¹⁸F-FDG PET (after 2 or 3 cycles of chemotherapy).

Patlak Graphical Analyses

For quantitative measurement of glucose metabolism, Patlak graphical analysis was used to calculate the MR_Glu (expressed in μmol^.mL^−1. min⁻¹) in tumor tissue (13,14). A detailed description of the Patlak graphical analyses and its assumptions have been published elsewhere (12,13). In brief, the Patlak approach takes into account differences in the whole-body distribution of ¹⁸F-FDG at the time of scanning, which may affect the accumulation of ¹⁸F-FDG in the tumor tissue. Furthermore, the MR_Glu in tumor was calculated by multiplication of the slope of the Patlak plot and the basic blood glucose level (expressed in μmol^.mL⁻¹). Blood glucose level was measured (hexokinase method, Aeroset; Abbott Diagnostics) before ¹⁸F-FDG injection. The fractional change in MR_Glu between ¹⁸F-FDG PET at baseline and after 2 or 3 cycles of chemotherapy was calculated.

Standardized Uptake Value (SUV)

¹⁸F-FDG PET scans were also evaluated semiquantitatively by SUV analysis. SUVs normalized to injected activity and patients' body weight were calculated from the mean activity concentration in the tumor VOIs between 40 and 50 min after injection, which is the last time frame of the dynamic scan. A volume-weighted mean value of each PET scan was derived from all lesions to give one average SUV (SUV_avg; hereafter referred to as SUV) for each PET scan. The percentage change in SUV between the ¹⁸F-FDG PET at baseline and after 2 or 3 cycles of chemotherapy was calculated.

Clinical Follow-up

During and after treatment, patients were monitored with clinical examination at regular intervals, chest CT, radiography, and other imaging studies as clinically indicated. Tumor response was evaluated according to RECIST criteria without knowledge of the results of the PET studies. When recurrence was suspected or proven, patients were always restaged. The progression/relapse pattern and cause of death were determined in all cases. The date of local or distant progression was defined as the earliest date at which disease progression was confirmed, either clinically or by imaging or biopsy. Patients who were progression free at the closeout date (November 1, 2006) or who had died from any cause had their time to progression censored at that date. Survival was measured from the date of the baseline PET scan to the date of death. Patients who were still alive at the closeout date had their survival censored to that date.

Statistical Analysis

Survival and progression-free survival served as the standard of reference. Cox proportional hazards regression analysis was used to assess the predictive value of response evaluation with ¹⁸F-FDG PET, as expressed in the percentage change in MR_Glu and SUV between the ¹⁸F-FDG PET at baseline and after 2 or 3 cycles of chemotherapy. The data have been adjusted for tumor stage at the time of the baseline PET scan. Statistical significance was assessed using the Wald χ² test. The overall survival and progression-free survival with respect to MR_Glu and SUV were calculated using Kaplan–Meier estimates. The percentage change in measurements between the first and second scan were stratified by the median value to avoid data-driven significance for the cutoff level. However, stratification at the median value is still a post hoc definition for metabolic response. To avoid this bias, also some predefined, prospective definitions of metabolic response were used. For that purpose the population was categorized according to the definitions for metabolic response of The European Organization for Research and Treatment of Cancer (EORTC), into a group with complete response (CR) or partial response (PR) (Δ < −25%), a group with stable disease (SD) (Δ −25% to +25%), and a group with progressive disease (PD) (Δ > +25%) (15). Furthermore, the population was categorized according to the cutoff values also used for size measurement (RECIST), for which Δ < −30% is CR or PR, Δ −30% to +20% is SD, and Δ > +20% is PD (16). Groups were compared using the Breslow test, which counts losses that occur early in the survival distribution more heavily than losses that occur late, because sample sizes are larger in early months relative to late months. Spearman ρ-correlations were used to determine the comparability between the different methods to measure glucose metabolism. Statistical tests were based on a 2-sided significance level, and the level of significance was set at 0.05.

Patient Characteristics

Sixty consecutive eligible NSCLC patients were included in this prospective study. After the first ¹⁸F-FDG PET, 9 patients were excluded for several reasons: technical issues (n = 1), refusal to undergo a second ¹⁸F-FDG PET scan (n = 2), death before the second ¹⁸F-FDG PET (n = 1), and early discontinuation of chemotherapy due to a significant decline in performance status (n = 5). Therefore, complete ¹⁸F-FDG PET datasets (median interval between baseline and follow-up ¹⁸F-FDG PET, 49 ± 17 d) were available in 51 patients (37 males, 14 females; mean age, 59 y; range, 38–76 y) for analysis of the predictive value. Patient characteristics are summarized in Table 1.

Treatment

Fifteen patients were treated with induction chemotherapy and 36 patients were treated in a palliative setting. Of the latter 36 patients, 32 received first-line chemotherapy and 4 received second-line chemotherapy. The chemotherapy regimens used were carboplatin/gemcitabine (n = 27), gemcitabine/cisplatinum (n = 9), carboplatin/etoposide (n = 9), cisplatinum/vinorelbine (n = 2), cisplatin/etoposide (n = 1), triapine/gemcitabine (n = 1), carboplatin/docetaxel (n = 1), and docetaxel (n = 1). After induction chemotherapy (n = 15), 2 patients underwent surgery, 10 patients were treated with radical radiotherapy, and 3 patients were treated with radiotherapy with palliative intent because of progression during induction chemotherapy based on CT criteria.

Patient Follow-up

No patients were lost to follow-up. Median follow-up time was overall 13 mo (range, 2–48 mo), for surviving patients 20 mo (range, 9–45 mo), and for deceased patients 9 mo (2–48 mo). During this time period, 38 patients had died and tumor progression was diagnosed in 45 of the 51 patients. Median time to progression was 7 mo (range, 0–39 mo). Overall survival after baseline ¹⁸F-FDG PET at 1 y was 57%, at 2 y 36%, and at 3 y 22%. Progression-free survival after baseline ¹⁸F-FDG PET at 1 y was 26%, at 2 y 17%, and at 3 y 8%.

Quantitative Changes in ¹⁸F-FDG Uptake

Mean MR_glu at baseline was 0.146 ± 0.079 μmol^.mL^−1.min⁻¹ (range, 0.028–0.356) and during evaluation was 0.084 ± 0.072 μmol^.mL^−1.min⁻¹ (range, 0.009–0.355). The median fractional change of MR_glu as compared with baseline was −47% ± 55% (range, −93% to +226%). Mean SUV at baseline was 5.27 ± 2.51 (range, 1.37–16.66) and during evaluation was 3.72 ± 2.75 (range, 0.71–16.64). The median fractional change of SUV as compared with baseline was −35% ± 42% (range, −73% to +125%).

Figure 1 shows a typical example of a patient with stage IV NSCLC and tumor lesions who responded to chemotherapy. There is an 82% decrease in MR_glu and a 65% decrease in SUV relative to baseline.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 1. 

Typical example of patient with stage IV NSCLC and tumor lesions who responded to chemotherapy. Relative to baseline (A), there is an 82% decrease in MR_glu and a 65% decrease in SUV on ¹⁸F-FDG PET after 2 cycles of carboplatin/gemcitabine (B).

¹⁸F-FDG uptake measured with simplified methods (SUV) showed a significant correlation with MR_glu as measured with the Patlak analysis (r = 0.84, P < 0.0001) at baseline, after 2 or 3 cycles of chemotherapy (r = 0.93, P < 0.0001), as well as in terms of fractional changes after 2 or 3 cycles of chemotherapy (r = 0.94, P < 0.0001).

Prediction of Survival by ¹⁸F-FDG PET

To generate Kaplan–Meier survival curves, the decrease in tumor glucose metabolism was dichotomized at the median value, which was −47% and −35% for ΔMR_Glu and ΔSUV, respectively. The Kaplan–Meier survival curves for overall survival are shown in Figure 2A (ΔMR_Glu) and Figure 2B (ΔSUV). The difference in survival between both strata was highly significant for both ΔMR_Glu and ΔSUV (P = 0.017 for ΔMR_Glu and P = 0.018 for ΔSUV; Table 2 and Figs. 2A and 2B). For patients with a decrease in MR_Glu lower than the median value, median overall survival was only 8 mo, whereas it was 17 mo for patients with a more pronounced decrease in MR_Glu.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 2. 

Kaplan–Meier estimates for overall survival: Kaplan–Meier analysis of relationship between overall survival and ΔMR_Glu (dichotomized using median value of −47%) (A) and ΔSUV (dichotomized using median value of −35%) (B).

There was also a significant difference in progression-free survival between both groups when dichotomized at the median value of ΔMR_Glu (P = 0.002) and of ΔSUV (P = 0.0009) (Table 2). Patients with a decrease of MR_Glu by >47% were characterized by a significantly longer time to progression (median, 10 vs. 3 mo; P = 0.002). The results obtained for ΔSUV were also evident: median progression-free survival, 11 versus 3 mo. The corresponding Kaplan–Meier curves for progression-free survival are shown in Figure 3A (ΔMR_Glu) and Figure 3B (ΔSUV). There was a highly significant increase in the rates of progression associated with worsening response as assessed by PET on Cox proportional regression analysis adjusting for disease stage at the time of commencement of treatment (P = 0.018 for ΔMR_Glu and P = 0.028 for ΔSUV).

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 3. 

Kaplan–Meier estimates for progression-free survival: Kaplan–Meier analysis of relationship between progression-free survival and ΔMR_Glu (dichotomized using median value of −47%) (A) and ΔSUV (dichotomized using median value of −35%) (B).

Applying predefined, prospective definitions of metabolic response with cutoff levels according to Young et al. (15) (EORTC) and according to the cutoff levels also used for size measurement (RECIST), the complete dataset showed similar highly significant differences in progression-free survival (Table 3). However, for prediction of overall survival, categorization according to these predefined cutoff levels showed only statistically significant differences for RECIST cutoff levels applied to ΔSUV (P = 0.027).