High rates of tumor growth and disease progression detected on serial pretreatment fluorodeoxyglucose-positron emission tomography/computed tomography scans in radical radiotherapy candidates with nonsmall cell lung cancer†
Presented in part at the 13th World Conference on Lung Cancer, San Francisco, California, July 31-August 4, 2009.
The authors studied growth and progression of untreated nonsmall cell lung cancer (NSCLC) by comparing diagnostic and radiotherapy (RT) planning fluorodeoxyglucose (FDG)-positron emission tomography (PET)/computed tomography (CT) scans before proposed radical chemo-RT.
Patients enrolled on a prospective clinical trial were eligible for this analysis if they underwent 2 pretreatment whole body FDG-PET/CT scans, >7 days apart. Scan 1 was performed for diagnosis/disease staging and scan 2 for RT planning. Interscan comparisons included disease stage, metabolic characteristics, tumor doubling times, and change in treatment intent.
Eighty-two patients underwent planning PET/CT scans between October 2004 and February 2007. Of these, 28 patients (61% stage III, 18% stage II) had undergone prior staging PET/CT scans. The median interscan period was 24 days (range, 8-176 days). Interscan disease progression (TNM stage) was detected in 11 (39%) patients. The probability of upstaging within 24 days was calculated to be 32% (95% confidence interval [CI], 18%-49%). Treatment intent changed from curative to palliative in 8 (29%) cases, in 7 because of PET. For 17 patients who underwent serial PET/CT scans under standardized conditions, there was a mean relative interscan increase of 19% in tumor maximum standardized uptake value (SUV) (P = .022), 16% in average SUV (P = .004), and 116% in percentage injected dose (P = .002). Estimated doubling time of FDG avid tumor was 66 days (95% CI, 51-95 days).
Despite improved understanding of the biology of nonsmall cell lung cancer (NSCLC) and the availability of newer targeted therapies, only modest improvements in survival have been achieved. Limited evidence from serial imaging studies suggests that NSCLC may progress rapidly between presentation and initiation of treatment.1-7 However, the accuracy of historical tumor growth data is uncertain because of the shortcomings of earlier imaging methods. Plain radiographs are reliable only for the 2-dimensional measurement of well-delineated primary tumors largely or entirely surrounded by normal lung. The accuracy of serial CT imaging for the assessment of growth rate and extent of NSCLC is limited by its poor sensitivity and specificity with respect to nodal staging, poor visualization of primary tumors in patients with atelectasis, and low sensitivity for the detection of distant metastasis.8, 9
Reliable information on tumor growth rates and the risk of tumor progression over time may be especially important when considering curative treatment for patients with NSCLC, many of whom present with stage III disease and are candidates for radical chemoradiation therapy (chemo-RT). Delays commonly occur during the processes of diagnosis, staging, multidisciplinary consultation, treatment planning, and initiation of therapy. If treatment is delayed the probability of survival may be reduced.4, 7, 10-13 A limited number of serial pretreatment imaging studies have been undertaken in NSCLC patients who are candidates for potentially curative therapy. In a significant study, O'Rourke and Edwards reported that 21% of 29 candidates for radical RT developed incurable disease progression (median 54 days) between diagnostic/staging and RT planning computed tomography (CT) scans.1-7
The most accurate available method for imaging NSCLC is combined positron emission tomography (PET) and CT scanning using a hybrid scanner with 18F-flurorodeoxyglucose (FDG) as the radiopharmaceutical.14, 15 To date, no studies have sequentially used FDG-PET/CT to investigate tumor progression in untreated NSCLC. In this study, each patient had 2 PET/CT scans before proposed radical chemo-RT, separated by a variable interval of time of at least 1 week. This study was part of a larger prospective study in which patients underwent whole body FDG-PET/CT scans in the RT planning position for both target volume determination and staging. Patients with prior staging PET/CT scans that were not acquired in the treatment position or that were acquired too long before treatment commencement were required by protocol to have a further PET/CT scan. This study quantitatively describes tumor growth between scans. It also reports the rate of anatomic disease progression between scans and considers the consequences of using noncontemporaneous imaging studies to select patients for aggressive treatment strategies and target radiation therapy.
MATERIALS AND METHODS
This study is a subanalysis of a prospective clinical trial of PET/CT in RT planning in NSCLC. It had ethics committee approval from the Peter MacCallum Cancer Centre and Monash University. Written informed consent was required. PET/CT scans that were unsuitable for RT planning, either because they were acquired without appropriate positioning and immobilization or because they were >4 weeks old, were required to be repeated. Patients with 2 pretreatment FDG-PET/CT scans were eligible for this substudy. Patients were considered suitable candidates for radical RT after the first PET/CT scan. All RT planning scans were reported by a nuclear medicine physician. Changes in clinical management were made if new clinical information was derived from the second scan.
Patients who underwent 2 PET/CT scans were allocated to the following categories:
Category A: All patients with untreated tumors who underwent 2 PET/CT scans. The first PET/CT scan could be performed at either our center or another center.
Category B: A subgroup of category A comprising only those patients who had both of their PET/CT scans carried out under identical conditions at our center.
For the entire cohort of patients who underwent 2 PET/CT scans before treatment (category A), a separate stage grouping was allocated to each patient after each of the 2 PET/CT scans. A range of endpoints was considered, including changes (if any) in tumor stage (TNM categories and overall stage groupings16-18) and the impact of increasing time intervals between scans on treatment intent and delivery.
Patients in category B were eligible for quantitative analyses of metabolic tumor characteristics and 18F-FDG avid tumor volumes. Tumor metabolism endpoints included changes in maximum and average tumor standardized uptake values (SUVs) over time, changes in the percentage of injected dose, and estimation of doubling time (DTs) according to volumes defined with SUVs individualized for each case and within an isocontour SUV of 3.0.
All initial staging scans were performed on integrated PET/CT scanners, either at 1 of 2 external institutions or at our own center. The external institutions used either a Biograph (Siemens Healthcare, Erlangen, Germany) or a Gemini (Philips Healthcare, Eindhoven, the Netherlands) scanner. Most staging scans and all RT planning scans were carried out at our center using a Discovery LS scanner (GE Medical Systems, Waukesha, Wis). All patients were weighed before scanning. Patients fasted for a minimum of 6 hours and, after an intravenous injection of 296 MBq of 18F-FDG (adjusted for patients' weight), rested before scanning. After injection, we used our validated method of flushing the FDG syringe 3×, which results in <1% of the dose remaining in the syringe. All planning scans utilized a RT compatible rigid palette and laser positioning. The transaxial field of view for CT and PET was 50 cm and extended from lower neck to pelvis.
Reporting and Contouring Procedures
All scans were discussed at a multidisciplinary meeting. Tumor characteristics were determined by a single observer using validated PET analysis software known as Medical Analysis of Region of interest Visualization or MARVn. Two methods were used for contouring tumor volumes; the volume encompassed by a set SUV isocontour of 3.0 (SUV3) and the volume encompassed by an isocontour value individualized for each case (SUVi) by qualitative analysis of an experienced operator of this software. After generation of the initial contour by MARVn, manual editing was performed to ensure that the contoured volumes did not include adjacent FDG avid normal tissues. In this study, the term gross (macroscopic) tumor volume (GTV) referred to the primary tumor alone, and GTV1 was classified as the closest regional lymph node station to the primary tumor, in the path of lymph node drainage.
Disease progression was recorded when a change in disease stage occurred. Disease stage was defined by both TNM characteristics and overall stage grouping. The expected value of the proportion of patients who experienced a change in staging was calculated, and the corresponding 95% confidence intervals (CIs) were calculated using the Clopper-Pearson exact method. The rate of disease progression was assumed to follow an exponential function, so that the probability of progression by time t is given by the expression F(t) = 1 − e−βt. This was equivalent to assuming a constant hazard rate; that is, it was equivalent to assuming that the probability of progression in any time increment, given progression-free survival until the start of that time increment, was constant. The progression rate in the expression above, β, and its 90% and 95% CIs, were then estimated using maximum likelihood estimation.
Tumor volume characteristics were compared using descriptive statistics. For each characteristic, the P value for a significant change between the 2 scans was calculated using the 2-sample matched pairs Student t test. When data did not conform to the normality assumption of the t test, the nonparametric Wilcoxon test was used. Comparisons for which it was found necessary to use the Wilcoxon test are indicated in Table 1. Unifactor linear regression models were developed to calculate tumor volume DTs. The log (base 2) of the ratio between the 2 scan volumes was used as the response variable (using the individualized and standardized SUV settings, in turn), and time between scans (in days) was the explanatory variable. Log volume ratios were used with the assumption made that tumor growth rates are approximately exponential. S-Plus (S-PLUS 2000 Professional Release 2, Mathsoft Inc.) was used for statistical analyses.
Table 1. Characteristics of Primary GTV Uptake on FDG-PET/CT (n=17)
Scan 1, Mean
Scan 1, SE
Scan 2, Mean
Scan 2, SE
Relative Change, % Mean
Relative Change, % SE
GTV indicates gross (macroscopic) tumor volume; FDG, fluorodeoxyglucose; PET/CT, positron emission tomography/computed tomography; SE, standard error; SUVi, individualized standardized uptake value; SUV3, standardized uptake value 3.0; SUVmax, maximum standardized uptake value; SUVavg, average standardized uptake value; %ID, percentage injected dose.
Eighty-two patients entered the FDG-PET/CT planning study between October 2004 and February 2007. Of these, 30 (37%) underwent 2 PET/CT scans before treatment. Two patients were ineligible for assessment of unperturbed tumor growth (1 received chemotherapy, and the other underwent surgery between scans). The remaining 28 (34%) patients were included in category A, including 22 (27%) who underwent both scans at our facility. Seventeen of these 22 (21%) patients were also eligible for category B.
Patient characteristics are presented in Table 2. Of 28 patients in category A, the median age was 70 years (range, 42-87 years). The median interscan period was 24 days (range, 8-176 days). Reasons for obtaining a second PET/CT scan were as follows: 25 (89%) patients were initially surgical candidates and were not positioned specifically for RT during the first scan; the lung cancer was an incidental finding on a scan performed without RT positioning for another purpose in 2 cases; and in 1 case, the patient decided to change management to RT from surgery.
Table 2. Patient Characteristics
No. of Patients Who Underwent 2 PET/CT Scans, n = 28 (%)
No. of Patients Who Underwent 2 PET/CT Scans at PMCC, n = 17 (%)
PET/CT indicates positron emission tomography/computed tomography; PMCC, Peter MacCallum Cancer Center; NSCLC, nonsmall cell lung cancer.
Squamous cell carcinoma
Large cell carcinoma
Overall disease stage
Scanning conditions for the 17 patients in category B who underwent both scans at our center are presented in Table 3. The mean (±SE) difference between rest times was 15 minutes (±5).
Table 3. Scanning Conditions for Patients Who Underwent Scans 1 and 2 at Our Center (n=17)
Scan 1, Mean (±SE)
Scan 2, Mean (±SE)
SE indicates standard error; FDG, fluorodeoxyglucose; MBq, megabecquerel; SUVavg, average standardized uptake value.
Patient weight, kg
FDG dose, MBq
FDG uptake time, min
Disease Progression Between PET/CT Scans
Ten (36%) of the 28 patients underwent a change in overall disease stage and 11 (39%) in TNM stage. The estimated proportion of change in overall stage was 0.36 (95% CI, 0.19-0.56) and in TNM stage was 0.39 (95% CI, 0.22-0.59). The 11 patients who underwent a change in TNM stage included the 10 whose overall stage changed, plus 1 additional patient who was upstaged from T4N2M0 to T4N3M0, therefore remaining stage IIIB. Four (25%) of the 16 patients with lower/mid lobe tumors had disease progression (T, N, or M stage), compared with 7 (58%) of the 12 patients with upper lobe tumors.
The estimated rate of disease progression according to TNM staging is presented in Figure 1. The probability of upstaging (according to TNM criteria) at the median interscan interval of 24 days was 32% (95% CI, 18%-49%). This was consistent for a change in overall disease stage. Increasing estimated rates of disease progression with increasing interscan intervals of 7, 14, 28, and 56 days are also presented in Table 4.
Table 4. Probability of Change in Disease Stage for Various Scan Intervals (n=28)
Scan Interval, d
Progression Rate Estimate
90% CI LL
90% CI UL
95% CI LL
95% CI UL
CI indicates confidence interval; LL, lower level; UL, upper level.
Median interscan interval for this population was 24 days. All progression rates are shown as percentages.
Probability of change in overall stage
Probability of change in TNM stage
After the second PET/CT, treatment intent changed from radical to palliative for 8 (29%) patients. In 5 (18%) cases, this was because of the detection of distant metastasis on the second scan. Serial PET/CT scans are shown in Figure 2 for Case 10, illustrating both local and distant disease progression. Locoregional intrathoracic disease progression was so extensive in 2 (7%) cases that radical RT could not be administered without exceeding dose constraints to normal tissues, mandating a change to palliative RT. In 1 case the disease stage was unchanged (T3N2M0), but a significant decline in performance status mandated a palliative approach. Three (15%) of the 20 patients who remained eligible for radical RT after planning PET/CT required increased target volumes to cover previously unsuspected nodal sites.
Tumor Volumes and Doubling Times
For the 17 patients in category B, the FDG avid tumor volume was contoured on both scans using the SUVi and SUV3 methods. The median SUVi was 3.0 (range, 1.5-4.0), and the same isocontour value was applied for a given patient at both time points. On the basis of the SUVi method, 13 (76%) patients demonstrated an increase in the size of the FDG avid tumor volume, whereas 4 (24%) demonstrated a reduction. Three of these 4 volumes also regressed using the SUV3 method. On the basis of the SUVi method, the median of volumetric progression and regression were 38% (range, 1%-176%) and 15% (range, 5%-31%). As shown in Table 1, the probability of observing a change in volume was statistically significant for both methods. The overall mean relative change in GTV according to the SUVi method was 32% (P = .016) and according to the SUV3 method was 63% (P = .006). Of the 7 (31%) patients with observable tumor in the first lymph node station (GTV1), changes in the volume of these nodes were not significant (P = .986). The expected DT of the FDG avid primary GTVs using the SUVi method was 66 days (95% CI, 51-95 days), as illustrated in Figure 3. Primary GTV growth using the SUV3 method produced an expected tumor volume DT of 62 days (95% CI, 48-88 days).
Metabolic Tumor Characteristics
There were significant increases in all metabolic variables assessed for the primary tumor (GTV) in patients in category B (Table 1). The mean increase in maximum SUV (SUVmax) was 19% (standard error [SE], 7.6%) (P = .022), and average SUV (SUVavg) was 15.6% (SE, 4.7%) (P = .004). The SUVmax was reduced between scan 1 and scan 2 in 4 (24%) of the 17 patients, including 1 of the patients with a small tumor that demonstrated a reduction in SUVi volume between scans. Of the 6 (35%) patients with observable tumor in the first lymph node station (GTV1), the SUVmax, SUVavg, and injected dose increased by 31% (P = .036), 29% (P = .059), and 23% (P = .26) during the interscan period, respectively.
The results of this study are sobering, as they show that approximately a quarter of patients experienced progression to incurable disease between initial staging and treatment planning PET/CT scans, performed at a median of 24 days later. The estimated probability of upstaging by a second PET/CT scan within 24 days was 32%. Without the additional PET/CT scan required by this study, futile radical RT would have been delivered to patients with distant metastasis or extensive nodal involvement. RT fields would have failed to adequately cover tumor in 15% of patients if the initial PET/CT scan was used for target volume definition. Overall, assuming that a geographic miss would render treatment unsuccessful, 37% of patients would have undergone futile radical chemo-RT. No patient had a second PET/CT scan because of suspected disease progression, and there was no apparent bias in patient selection.
These results suggest that disease progression occurs faster in NSCLC than was reported by O'Rourke and Edwards (21% of patients at a median of 54 days4). Because they used noncontrast planning CT scans only, it is reasonable to assume that O'Rourke and Edwards inadvertently underestimated the rate of disease progression in their patients. All patients with progression deemed incurable had squamous cell carcinoma (SCC). This observation was not confirmed in our study, where only 1 of the 8 patients who became a candidate for palliative therapy had SCC. Wang et al showed no effect of histological subtype on delays in starting treatment and overall survival.7
We used an objective semiquantitative SUV-based method to measure tumor metabolism.19 SUV thresholds were individualized for each case (SUVi),20 and a contouring tool (MARVn) was used to minimize bias in tumor volume estimates. Overall, in our cohort, tumor growth was more rapid than in previous studies, with a median tumor volume DT of just 66 days, possibly because the predominantly stage III cancers in our series were more aggressive and grew faster than the lesions described in earlier studies. Previous studies reported DTs of untreated lung cancers of between 90 and 207 days.6, 11, 21 More recently, Eastham and colleagues compared serial FDG-PET scans (not PET/CT) for 11 patients with NSCLC, and reported tumor volume DTs of ≤6.5 weeks (45 days) in 4 patients.2 Experimental solid tumors often show early exponential initial growth with later slowing as they enlarge (Gompertzian growth pattern). Gompertzian growth occurs in breast cancer22 and in some human thoracic tumors, including pulmonary metastases from osteosarcoma.23 In contrast, Blomqvist and colleagues showed only exponential growth in lung metastases from soft tissue sarcomas.24 Because tumor volume determinations were made at only 2 time points in our study, we could not study the shape of the growth curve in NSCLC.
Four tumors showed apparent reductions in metabolically active tumor volumes between scans. Two small lesions (<8 mL) could have been influenced by minor variations in contouring or respiratory excursion and associated misregistration between the CT and PET components. A small variation in contouring of the third case (4.8 mL) caused an apparent volumetric regression of this lesion. The 12% reduction in the FDG avid region observed in the fourth case was attributable to proximity to adjacent consolidated lung tissue and moving diaphragm. Biological factors, such as partial tumor necrosis because of vascular occlusion, could have played a part. Nevertheless, the overall trend was to increasing volume and intensity of FDG avid tumor.
This study is the first to report increasing SUV of primary tumors in patients with NSCLC over time. More accurate measurements of SUV are possible with PET/CT compared with PET imaging alone. Higher SUV levels are associated with more rapid growth, higher rates of recurrence, and a greater risk of distant metastasis. Tumors may also become more hypoxic as they expand.25
Delayed treatment reduces survival in NSCLC,4, 7, 10-12 and tumors are more likely to be controlled by RT when they are <3 cm.26 If our data are confirmed by other groups, they may influence standards of practice for the workup and treatment of lung cancers. In 1993, the Joint Council for Clinical Oncology recommended that potentially curative treatment should ideally start within 2 weeks and no longer than 4 weeks after a decision to treat.27 These strong recommendations may still be too permissive for NSCLC patients who are candidates for potentially curative RT. Patients who progressed with distant metastasis between PET scans would probably not have benefited from earlier treatment. However, it is likely that a significant subset of patients will miss their only chance for cure if treatment for NSCLC is delayed.
There is no lower limit for waiting time below which there is no risk of progression. However, strenuous efforts to promote early diagnosis, workup, and treatment may improve the results of treatment of unresectable NSCLC.28 When radical chemo-RT is planned, up-to-date imaging is essential for accurate tumor targeting. Healthcare providers should consider funding a further FDG-PET scan for RT planning in NSCLC when the initial staging scan can no longer be relied on.