Abstract
This study investigated the possibility of early response assessment based on 18F-FDG uptake during radiotherapy with respect to overall survival in patients with non–small cell lung cancer. Methods: 18F-FDG PET/CT was performed before radiotherapy and was repeated in the second week of radiotherapy for 34 consecutive lung cancer patients. The CT volume and standardized uptake value (SUV) parameters of the primary tumor were quantified at both time points. Changes in volume and SUV parameters correlated with 2-y overall survival. Results: The average change in mean SUV in the primary tumor of patients with a 2-y survival was a decrease by 20% ± 21%—significantly different (P < 0.007) from nonsurvivors, who had an increase by 2% ± 22%. A sensitivity and specificity of 63% and 93%, respectively, to separate the 2 groups was reached for a decrease in mean SUV of 15%. Survival curves were significantly different using this cutoff (P = 0.001). The hazard ratio for a 1% decrease in mean SUV was 1.032 (95% confidence interval, 1.010–1.055). Changes in tumor volume defined on CT did not correlate with overall survival. Conclusion: The use of repeated 18F-FDG PET to assess treatment response early during radiotherapy is possible in patients undergoing radiotherapy or sequential or concurrent chemoradiotherapy. A decrease in 18F-FDG uptake by the primary tumor correlates with higher long-term overall survival.
Overall survival rates in lung cancer have improved but nevertheless remain low. State-of-the-art treatment of locally advanced-stage non–small cell lung cancer is chemoradiotherapy. Response assessment in the first weeks of radiotherapy would be useful to tailor the optimal treatment strategy for the individual patient. 18F-FDG PET/CT serves as an imaging technique allowing quantification of tumor response to treatment before morphologic changes become visible on standard CT (1–3).
Several authors have shown the prognostic value of residual 18F-FDG uptake in the primary tumor after treatment—a finding that indicates poor survival (4–7). Early response assessment for chemotherapy has been described by several groups (1,3,8–10), but for radiotherapy only a few studies have been performed. For high-dose stereotactic body radiotherapy, elevated levels after treatment do not always indicate residual tumor (11). In a smaller study, 18F-FDG PET/CT during treatment could already distinguish metabolic responders from nonresponders based on the maximum standardized uptake value (SUV) (12). High 18F-FDG uptake before treatment is known to correlate with poor overall survival (13,14); however, the influence of treatment response during chemoradiotherapy is not well described (4,15–18).
Assessment of response early during treatment is useful in following patients throughout the course of treatment. Chemoradiotherapy is a demanding strategy that has a high burden for the patient. Selection of patients up front who will benefit is difficult, and an early assessment of treatment response could allow adaptation of the treatment to the individual patient.
In the present study, we hypothesized that early changes in 18F-FDG uptake by the primary tumor are a predictive factor for treatment success. We evaluated all non–small cell lung cancer patients in a fixed period who were treated with curative intent using radiotherapy or sequential or concurrent chemoradiotherapy. We investigated the correlation between metabolic response visualized on repeated 18F-FDG PET/CT in the second week of radiotherapy and overall survival.
MATERIALS AND METHODS
Patient Characteristics and Dose Prescription
For all non–small cell lung cancer patients from July 2008 to December 2008 who were scheduled for radical radiotherapy, we performed 18F-FDG PET/CT before therapy and in the second week of radiotherapy. Patients scheduled for stereotactic body radiotherapy were not included in this study. The study was approved by the appropriate Institutional Review Board. Stages II–IV were included in this prospective analysis, and treatment was performed according to the clinical protocol. Treatment consisted of radiotherapy only or sequential or concurrent chemoradiotherapy. The type of chemotherapy for sequential and concurrent chemoradiotherapy was cisplatin–gemcitabine and cisplatin–vinorelbine, respectively. In both cases, a total of 3 cycles of chemotherapy were administered. For concurrent chemoradiotherapy, radiation began at the beginning of the second cycle of chemotherapy. Two cycles of chemotherapy were given during radiotherapy. The treatment consisted of a dose-escalation protocol up to the normal-tissue constraints (19). For concurrent chemoradiotherapy, a 45-Gy dose of radiotherapy was given twice daily with fraction sizes of 1.5 Gy for the first 30 fractions. Afterward, the dose was escalated: daily fractions of 2 Gy were given up to the maximum tolerated normal-tissue toxicity or until the maximum prescribed dose of 69 Gy was reached. For the sequential-chemoradiotherapy patients, 3 cycles of chemotherapy were given and then radiotherapy was started. The radiotherapy consisted of 1.8-Gy fractions given twice daily up to the normal-tissue toxicity or a maximum prescribed dose of 79.2 Gy (19). This radiotherapy scheme was also used for the radiotherapy-only group.
PET/CT
18F-FDG PET/CT (Biograph 40; Siemens Medical Solutions) was performed after 1 cycle (concurrent chemoradiotherapy) or typically 3 cycles (sequential chemoradiotherapy) of chemotherapy but before radiotherapy. Patients were required to fast for at least 6 h before the acquisition. Blood glucose levels were determined for all patients and were lower than 10 mmol/L, and no correction was applied for the blood glucose level. The scan was obtained in radiotherapy position, and the images were used for accurate tumor delineation and treatment planning. Four-dimensional respiration-correlated CT was acquired to visualize possible tumor movement due to respiration, as is necessary for radiotherapy planning purposes. A separate contrast-enhanced CT scan was also obtained during this imaging session. During the second week of radiotherapy, the imaging session was repeated using the same immobilization devices and setup as the pretreatment 18F-FDG PET/CT procedure. For each patient, the injected amount of 18F-FDG (MBq) was equivalent to 4 × body weight (kg) + 20 (MBq), and the patients needed to rest for 60 min before image acquisition could start. The raw PET data were corrected for scatter and decay, were rebinned, and subsequently were reconstructed using ordered-subset expectation maximization in 2 dimensions with 4 iterations and 8 subsets.
Quantitative Analysis
Quantitative analysis of 18F-FDG uptake was performed using SUV. The maximum SUV and mean SUV inside the primary tumor were calculated, and changes in uptake between time points were analyzed. For mean SUV inside the primary tumor, the volume was defined by voxels having uptake greater than 50% of the maximum SUV inside the primary tumor; we called this volume the PET volume. Survival of the patients was grouped according to the response criteria of the European Organization for Research and Treatment of Cancer (EORTC). A 15% or 25% decrease in 18F-FDG uptake, depending on the number of chemotherapy cycles, is associated with partial metabolic response using the EORTC criteria (20), and the PET Response Criteria in Solid Tumors suggest a 30% decrease for response (21). These cutoffs were used in the analysis to separate the groups.
Automatic segmentation methods based on the 18F-FDG PET images were used to define tumor volume (22). Additionally, we used the gross tumor volume of the primary tumor and the possible involved lymph node volume as delineated on the PET/CT scan by the radiation oncologist. These volumes were subsequently used to design the radiotherapy plan. The gross tumor volume of the primary tumor, as well as the total tumor volume including possible involved lymph nodes, was also investigated for correlation with survival.
Endpoint and Statistical Analysis
The endpoint evaluated in this study was the 2-y overall survival rate. All paired analyses were performed using a Wilcoxon signed rank test, with a P value of less than 0.05 indicating statistical significance. Differences between groups were evaluated using a Mann–Whitney U test. Changes in the SUV parameters of the midtreatment scan were calculated relative to the pretreatment scan. For this threshold, a Cox regression and survival analysis was performed. Survival curves were displayed by Kaplan–Meier curves, and survival between groups was compared by the log-rank test.
RESULTS
For 35 non–small cell lung cancer patients, 18F-FDG PET/CT was performed before radiotherapy and in the second week of radiotherapy. Imaging time points and patient characteristics are summarized in Table 1. Excluded from further analysis was a patient who had a complete response after 3 cycles of chemotherapy before radiotherapy and did not show any tumor mass on the repeated scan. Another patient had 2 primary tumors in the lung; for this patient, the lesion with the largest volume was analyzed. Minimum follow-up was 2 y and 2 mo, with an overall 2-y survival rate of 56% for all patients. Figure 1 shows the survival curves for the different therapy regimens in this study.
Patient Characteristics
Survival curves for different patients grouped according to timing of chemotherapy: either no chemotherapy (n = 2), sequential (Seq.) chemoradiotherapy (n = 18), or concurrent (Conc.) chemoradiotherapy (n = 14).
The volume of the primary tumor was on average 61.4 ± 56.7 cm3 and decreased to 56.4 ± 49.2 cm3 in the second week of treatment, a fractional decrease of 5.8% ± 19.3% in volume. The total tumor volume (primary tumor and nodes) for all patients, including those with nodal involvement (n = 24), was 81.9 ± 63.3 cm3 and 77.8 ± 59.3 cm3 for the pretreatment and midtreatment time points, respectively. The average relative decrease in total tumor volume was 4.1% ± 15.3%. The PET volume before treatment was on average 17.5 ± 20.3 cm3; for the midtreatment time point this volume was 16.5 ± 18.0 cm3, a fractional change of 7.5% ± 44.2%.
The maximum SUV and mean SUV inside the PET volume for the entire population was 10.0 ± 4.9 and 6.6 ± 3.2, respectively, for the pretreatment scan, compared with 8.8 ± 4.2 and 5.9 ± 2.9, respectively, for the midtreatment scan. Maximum and mean SUV parameters did not reach significance in a Cox regression analysis. The fractional change between time points was −10.8% ± 22.3% and −10.4% ± 23.6% for maximum and mean SUV, respectively.
The average change in the CT volume of the primary tumor or the total tumor volume including the involved lymph nodes was not significantly different between patients who survived for 2 y and those who did not (P = 0.215 and P = 0.918, respectively) (Table 2). Also, the change in PET volume was not significant (P = 0.215).
Volume and SUV Characteristics for All Patients
The average change in mean SUV inside the PET volume was −20.2% ± 20.5% for patients who survived for 2 y, compared with +2.1% ± 21.9% for patients who did not (P = 0.007). For maximum SUV inside the gross tumor volume, these numbers were −12.9% ± 23.1% vs. −4.2% ± 18.5% (P = 0.015), respectively. A detailed overview of all parameters is given in Table 2. Figure 2 shows an example of 18F-FDG uptake in the primary tumor before and during treatment for both a metabolic responder and a nonresponder.
Patient example of metabolic responder vs. nonresponder for both pretreatment and mid-treatment 18F-FDG PET/CT. SUV window levels are scaled equally per patient.
The EORTC criterion of 15% for partial metabolic response was used to divide the dataset into 2 groups. For patients with a decrease in SUV of more than 15%, median overall survival was not yet reached and 2-y overall survival was 92%, compared with a median overall survival of 19 mo and a 33% 2-y overall survival, respectively, for the patients with a decrease in mean SUV of less than 15%.
Figure 3 shows the Kaplan–Meier curves for the decrease in mean SUV of the primary tumor. Survival between the 2 groups was statistically different, with a hazard ratio of 1.032 (95% confidence interval, 1.010–1.055) per percentage-point decrease in mean SUV. At cutoffs of 15%, 25%, and 30% decrease, a sensitivity of 63%, 47%, and 42%, respectively, and a specificity of 93%, 93%, and 93%, respectively, were reached.
(A) Overall survival plots for metabolic responders (n = 13) and nonresponders (n = 21) defined for decrease of at least 15% in mean SUV of primary tumor (P = 0.001). (B) Survival plots if cutoff of 30% decrease in mean SUV is used: 5 metabolic responders vs. 29 nonresponders (P = 0.026).
Figure 4 shows the survival plots for the change in maximum SUV of the primary tumor. The hazard ratio for percentage decrease in SUVmax was 1.027 (95% confidence interval, 1.005–1.049) per percentage-point change in maximum SUV.
(A) Overall survival plots for metabolic responders (n = 14) and nonresponders (n = 20) defined for decrease of at least 15% in maximum SUV of primary tumor (P = 0.004). (B) Survival plots if cutoff of 30% decrease in maximum SUV is used: 9 metabolic responders vs. 25 nonresponders (P = 0.026).
DISCUSSION
This is one of the first studies showing that repeated 18F-FDG PET early during radiotherapy has added value by being a predictive factor for survival before CT changes become evident (23,24). A decrease in metabolic activity of the primary tumor by as early as the second week of treatment was predictive of survival. The simplicity of calculating the average 18F-FDG uptake inside the primary tumor is one of the factors that could be exploited in clinical practice for individualizing treatment.
The EORTC criteria (20) indicate a partial metabolic response after 1 cycle of chemotherapy if 18F-FDG uptake decreases by more than 15%. Our study confirmed that this percentage correlates with a more long-term endpoint: the 2-y overall survival. The PET Response Criteria in Solid Tumors (21) suggest that a 30% decrease is needed to classify a partial response; however, these criteria are also based on the large variability caused by technical issues, the use of different scanners, and the use of different reconstruction protocols—all decreasing the reproducibility of the repeated imaging. We have minimized these factors by using the same PET/CT scanner, acquisition protocols, and procedure for both imaging time points. Reproducibility (e.g., test–retest) studies, however, have shown a 10%–15% variability using the same equipment in repeated imaging of the same patient on different days (21).
The number of patients (n = 34) in this study was, however, too limited for an in-depth subgroup analysis. A future larger study is necessary that might be able to provide even better prognostic and predictive factors incorporating stage or histologic type.
Van Baardwijk et al. (12) showed that nonresponders had an increase in maximum SUV early in treatment. We did not observe this trend in our population. There could be several reasons for this difference. Our imaging time point was in the second week of treatment, whereas the maximum in their publication was found in the first week. Another likely cause is that in our study almost all patients were treated with sequential or concurrent chemotherapy before or during treatment. Chemotherapy is known to suppress the 18F-FDG uptake signal inside tumors (25,26). Kong et al. quantified changes in 18F-FDG uptake after a 45-Gy dose of radiotherapy (18). They found a correlation between an early metabolic response and a 3- to 4-mo CT-based response, but they did not perform a survival analysis. In addition, Huang et al. recently investigated a group of non–small cell lung cancer patients with repeated imaging after they had received approximately 40 Gy, and short-term outcome was evaluated using the Response Evaluation Criteria in Solid Tumors at 4 wk after the end of treatment (16). Changes in maximum or mean SUV allowed prediction of the treatment outcome. However, these studies had imaging time points in the second half of the radiotherapy course, making treatment adaptation less effective because there is not much time left to adapt or improve the treatment. Our study shows that changes in metabolic activity can already be detected in the second week of treatment, when more than half the treatment time is still available for improvement and individualization.
To fully characterize the changes in SUV uptake, one needs to look at the individual voxels inside the tumor and assess the response per voxel (27). Such an analysis requires a detailed voxel-tracking method, which was outside the scope of this study. Another option for a more comprehensive tumor quantification is the use of multiple tracers, such as 3′-deoxy-3′-18F-fluorothymidine and 18F-fluoromisonidazole in addition to 18F-FDG, as described by Vera et al. (28). Also interesting is an approach described by van Velden et al., who looked at cumulative histograms of SUV distribution and heterogeneity and at the response that might be characterized from these histograms (29). This might be an alternative approach to fully voxel-based response analysis. Also, the definition of PET volume using a cutoff of 50% of maximum SUV is highly sensitive to maximum SUV and might also underestimate the true tumor volume. For the repeated PET scan, therapy-induced reduced activity might again enlarge this volume because of reduced maximum uptake. Newly developed methods, for example, gradient-based, might give a more robust volume definition (30,31).
To use early information about response in clinical practice, one could investigate the added value of an adaptive protocol based on the results of an 18F-FDG PET scan during treatment. For responders, treatment could be continued as planned. In the future, we might envisage designing trials that lower the radiotherapy dose in very favorable responders to decrease toxicity. For nonresponders, treatment could be intensified using a new treatment plan that includes possible volume reductions already observed compared with the planning scan. This change might allow dose escalation of the smaller volume with the same level of normal-lung toxicity. Volume changes in the second week of treatment are, however, small but may trigger additional imaging later during the course of radiotherapy where larger volume changes are reported (18,32–36). However, one has to be careful with this type of dose escalation and shrinking-field approach because microscopic disease might not be treated effectively. Therefore, such an approach should be investigated in a clinical trial. The data presented in this study could serve as an estimation of which cutoff for 18F-FDG decrease should be used. The imaging biomarker (e.g., difference in mean SUV of the primary tumor) might then become a predictive marker that could be used in patient-individualized treatment.
CONCLUSION
Early assessment of treatment response is possible by measuring the decrease in average 18F-FDG uptake in the primary tumor on repeated 18F-FDG PET in the second week of radiotherapy. A large decrease in 18F-FDG uptake early during treatment correlates with improved overall survival.
DISCLOSURE STATEMENT
The costs of publication of this article were defrayed in part by the payment of page charges. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.
Acknowledgments
This study was performed within the framework of CTMM, the Center for Translational Molecular Medicine (www.ctmm.nl), project AIRFORCE number 03O-103. The research was supported by EU FP7 funding (project ARTFORCE) and in part by a KWF fellowship from the Dutch Cancer Society. No other potential conflict of interest relevant to this article was reported.
Footnotes
Published online Aug. 9, 2012.
- © 2012 by the Society of Nuclear Medicine and Molecular Imaging, Inc.
REFERENCES
- Received for publication December 29, 2011.
- Accepted for publication May 18, 2012.