¹⁸F-Fluoro-2-Deoxy-Glucose Uptake Predicts Clinical Outcome in Patients with Gefitinib-Treated Non–Small Cell Lung Cancer

Patients and Methods

Patients. All PET studies using a PET/CT scanner in NSCLC patients who received gefitinib treatment between April 2003 and April 2007 were evaluated retrospectively. Eighty-six patients met the following entry criteria: pathologically proved stage IIIb (with pleural effusion and/or positive supraclavicular lymph nodes) or stage IV NSCLC; PET/CT was applied before any therapy was given; histopathology was reviewed at Korea Cancer Center Hospital in Seoul. Two patients with diabetes mellitus were excluded from the study. Sixteen of the 84 patients were chemotherapy-naive, and 68 patients received up to six cycles of initial cisplatin-based or carboplatin-based chemotherapy combined with one agent of paclitaxel, docetaxel, gemcitabine, or vinorelbine. Patients with progressing disease after initial chemotherapy received subsequent chemotherapy at the discretion of primary physician. As a first to fifth line treatment, gefitinib monotherapy (250 mg/d) was delivered, and in most patients, this was continued until disease progression. Three patients discontinued medication due to development of unacceptable adverse events, such as hepatotoxicity (one patient) and diarrhea (one patient) or refusal (one patient). Chest X-rays were taken every 4 wk. CT scans were done at 4 wk after gefitinib treatment and every 8 wk thereafter. In case of response, confirmation required at least 4 wk of interval. Responses to gefitinib were classified into complete response, partial response, stable disease, and progressive disease using the WHO criteria (15). Performance status was defined according to the Eastern Cooperative Oncology Group classification. The study protocol was reviewed and approved by the institutional review board of the Korea Cancer Center Hospital.

¹⁸F-FDG PET/CT. Pretreatment whole-body ¹⁸F-FDG PET/CT scans were acquired as a part of staging work-up using a Discovery LS PET/CT scanner (GE Medical Systems) with the protocol of PET/CT scanning used at our institution (16). In brief, whole-body CT was done using helical CT, and then an emission scan was done 50 min after injecting 370 MBq of ¹⁸F-FDG i.v. All patients fasted for at least 6 h before PET/CT. The PET images were reconstructed with the OSEM algorithm, and attenuation correction was done using a CT scan. Abnormal FDG uptake was defined as that greater than the background activity in the surrounding tissue, and the intensity of FDG uptake was quantified by calculating the SUV. Region of interest was drawn manually around the primary malignancy with abnormal FDG uptake on transaxial images, which were reconstructed using a Gaussian filter. The SUV was calculated from the amount of FDG injected, total body weight, and soft-tissue uptake in the attenuation-corrected regional images: SUV = (activity / unit volume) / (injected dose / total body weight). The maximum SUV was defined as the peak SUV on one pixel with the highest counts within region of interest. The maximum SUV of the primary site, found on PET image, was selected for further analysis.

Statistical analysis. The categorical variables were analyzed using univariate analysis with a Pearson's χ² test or Fisher's exact test. To obtain the cutoff value of SUVs, which is a continuous variable, receiver operating characteristic curve analysis was done. Using this value, patients were divided into groups of those with low or high SUV. Multivariate logistic regression was used to test the association between significant variables in univariate analysis and gefitinib responsiveness. Nonparametric Kruskal-Wallis test was used when appropriate. The progression-free survival (PFS) and overall survival (OS) were calculated from start of gefitinib administration. Kaplan-Meier estimates of PFS and OS were calculated as described (17). Log-rank tests were done for univariate analysis. In addition to performance status, significant factors in univariate analysis were evaluated using the Cox model (18). Odds ratios, hazard ratios, and their 95% confidence intervals (95% CI) were calculated. Stata version 8.2 was used for statistical analyses. All P values were derived from two-sided tests, and P < 0.05 was considered significant.

Results

Patient characteristics. The baseline characteristics of the patients are shown in Table 1 . Forty-eight (57%) patients were male. The median age was 60 years. Never-smokers comprised 49% of patients. The maximum SUV of 84 patients ranged from 1.6 to 40.1 with median value of 11.3. The response rate (RR) was 25% (21 partial response + no complete response), and the disease control rate was 62% (21 partial response + 31 stable disease). Response to gefitinib was associated with smoking history and pathology (Table 2 ). Never-smokers showed better response than smokers (RR, 41% versus 9%; P = 0.001). Patients with adenocarcinoma showed better response than those with other pathologic types (RR, 35% versus 9%; P = 0.009). RR in female and male patients was 33% and 19%, respectively (P = 0.127). Stages (III versus IV) and performance status (0-1 versus 2-3) were not significantly associated with gefitinib responsiveness.

SUV and other factors. Associations between the SUV and clinical features were evaluated. The SUVs of smokers (median, 14.0; range, 2.7-40.1) were significantly higher than those of nonsmokers (P = 0.005; median, 9.6; range, 1.6-25.2). Patients with adenocarcinoma had low SUVs (median, 8.5; range, 1.6-40.1) compared with patients with other types of histopathology (P < 0.001; median 15.4; range, 4.5-38.7). Female patients showed lower SUVs (median, 9.8; range, 2.4-29.9) than male patients (P = 0.017; median, 13.2; range, 1.6-40.1). Age (<60 years versus ≥60 years), performance status (0-1 versus 2-3), and stage (III versus IV) were not associated with the SUV (data not shown).

SUV and response. We evaluated associations between the SUV and clinical outcomes in gefitinib-treated patients. According to response type (partial response, stable disease, or progressive disease), patients had significantly different SUVs for primary tumor (medians of 6.8, 9.0, and 15.2, respectively; P = 0.001; Fig. 1A ). To classify SUVs into two subgroups showing different outcomes, receiver operating characteristic analysis was done using overall response (Fig. 1B). Receiver operating characteristic analysis suggested that the SUV was a reasonable predictor of tumor response to gefitinib (area under curve ± SD, 0.74 ± 0.06). A lower cutoff would capture more gefitinib-responsive patients. To select gefitinib-responsive patients efficiently, cutoff values required sensitivity of >80% with a loss of specificity (19). Among these values, the SUV of 6.2, which gave maximum sensitivity and specificity (87% and 43%, respectively), was chosen for further analysis. Using this value, 17 patients (20%) had tumors with low SUV (<6.2). These patients showed with higher RR than those with high SUV (53% versus 18%, respectively; P = 0.003). Multivariate analysis, which was corrected for smoking history and pathology, revealed that a high SUV was associated with reduced gefitinib responsiveness (P = 0.018; Table 2).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

A, the maximum SUV of primary lung lesion according to response to gefitinib. B, receiver operating characteristic curve of the SUV predictable for gefitinib-responsiveness. Lower cutoffs indicate a higher probability of response. A cutoff value of 6.2 was selected. C, Kaplan-Meier plots of PFS according to the SUV. D, Kaplan-Meier plots of OS according to the SUV. Horizontal line, median; AUC, area under curve; CR, complete response; PR, partial response; SD, stable disease.

SUV and survival outcome. The median PFS of all patients was 12 weeks (95% CI, 8.1-16.6 weeks). Predictors for PFS were evaluated in univariate and multivariate analysis (Table 3 ). Never-smokers had prolonged PFS compared with ever-smokers (P = 0.002). Adenocarcinoma was a favorable prognostic factor in terms of PFS (P = 0.003). Patients with a low SUV showed longer PFS than those with a high SUV (P = 0.003; Fig. 1C). According to performance status (0-1 versus 2-3), the survival difference did not reach statistical significance (P = 0.175). Age (≤60 years versus >60 years) and sex did not confer statistically different outcomes in terms of PFS (P = 0.987 and 0.427, respectively). When analysis was conducted controlling for smoking history, performance status, and histopathology, patients with a low SUV showed favorable outcomes (P = 0.012).

When OS was analyzed, the median survival time was 30 weeks (95% CI, 21.6-43.6 weeks). Never-smokers had prolonged survival compared with smokers (P = 0.043). Adenocarcinoma and good performance status (≤1) were favorable factors with statistical significance (P = 0.009 and 0.002, respectively). Low SUVs conferred prolonged survival times compared with high SUVs (P = 0.011; Fig. 1D). After adjustment for factors, such as smoking history, performance status, and histopathology, patients with a low SUV had favorable survival (P = 0.043).

Discussion

We retrospectively evaluated PET/CT findings in patients with gefitinib-treated NSCLC. To our knowledge, this is the first study to evaluate SUVs and their association with clinical outcomes. It seems clear that some subgroups might respond to TKI or show a survival benefit in patients with NSCLC. The current study shows that favorable responses or survival occur in patients with no smoking history, adenocarcinoma, and good performance status. Such findings were observed in previous studies for TKI-treated NSCLC patients (20–23). Although epidermal growth factor receptor mutations or gene copy number could predict responsiveness to TKI (7, 8), genetic testing, frequently unavailable in clinical practice, was beyond the scope of this study.

In our study, PET images using a PET/CT scanner were evaluated at the time of diagnosis. Low SUVs were observed in patients with adenocarcinoma, never-smokers, and are female, which were predictable factors in previous studies (20–23). As expected, we observed that SUVs at presentation were different according to response to gefitinib (Fig. 1A). Patients with low and high SUVs were classified using a cutoff value from receiver operating characteristic analysis, and they showed different clinical outcomes. Importantly, the SUVs remained significant in multivariate analysis of tumor responsiveness and survival. These findings suggest that the SUV may be used as an independent predictor of outcomes along with the clinical features reported previously (20–23).

There are some arguments in terms of changing continuous to categorical variables (24). However, when SUVs were analyzed as a continuous variable, SUVs were significantly associated with RR and survival (odds ratio for objective response 1.17, P = 0.005; hazard ratio for PFS 1.04, P < 0.001; hazard ratio for OS 1.04, P = 0.008). It is believed that patient-related factors other than statistical methods might explain the clinical outcomes in this study more accurately.

In our data, primary tumors were larger than 1.7 cm on CT. Considering the resolution of PET/CT at our institution (5.4 mm at the full-width at half-maximum), the results of this study are unlikely to have been altered by partial volume effects (25). In addition to tumor-related factors, SUVs can be affected by various factors (e.g., reconstruction methods, blood sugar level, and the dose of FDG; ref. 26). Further studies for standardized methods to obtain SUV, including optimal cutoff values, should be conducted.

There are rare data to suggest that the SUV at presentation is associated with TKI responsiveness. However, our findings are in line with the result of a previous study that reported a low SUV at the time of diagnosis in a pathologic subtype of bronchioloalveloar carcinoma (27), another predictor of TKI responsiveness (20). Interestingly, we observed that the SUV of never-smokers was lower than that of smokers. Toh et al. reported different clinical features according to smoking history (28). Also, previous studies have suggested that cancer cells in patients with a never-smoking history, compared with patients who are smokers, may undergo different genetic changes, including changes in the TKI-binding region (7, 29). Considering the favorable outcomes in patients with a low SUV (9, 10), we speculate that low SUVs might reflect a less aggressive biology of TKI-responsive tumors. However, to our knowledge, there is no published literature to suggest that genetic changes, such as mutations of the TKI-binding domain might be related to FDG uptake of tumors. Further studies, including laboratory tests, are warranted.

Some authors have reported an association between high SUV and favorable response to cytotoxic agents (14). Similarly, when response to initial chemotherapy was analyzed in 68 patients, objective tumor RRs with initial cytotoxic chemotherapy were higher among patients with a high SUV (56%; 30 of 54 patients) than patients with a low SUV (29%; 4 of 14 patients); however, this difference was not statistically significant (P = 0.072). It seems contradictory that cytotoxic chemoresponsiveness, a favorable prognostic factor (30), may be associated with high SUV, which has been reported as an unfavorable prognostic factor (10) Comprehensive studies need to evaluate that outcomes according to types of chemotherapeutic agents may be affected by SUVs.

In this study, the SUV of primary site was selected for analysis according to the method as previous studies for localized disease was done (10). In fact, we evaluated the SUVs of all cancerous lesions on CT (longest diameter, >1 cm) and in the majority of cases (77 cases), the SUV of primary lesion was the highest one in each patient, as has been observed elsewhere (14). Even when the highest one was used for statistical analysis, clinical significance of the SUV remained (data not shown). However, the SUV of primary site seems to be easily evaluated in clinical practice.

Besides being retrospective with a small sample size, there are some limitations in our study. Timing of gefitinib delivery, another limitation of this work, differed between patients because of various courses of cytotoxic treatment, which has also been reported in previous studies (21, 23). In the current study, groups of chemonaive and chemotherapy-pretreated patients did not show statistically different outcomes. It should be considered that different outcomes according to types of previous treatment were not suggested in previous studies (21, 23).

In conclusion, the present study suggests that low SUVs at presentation can predict favorable response and survival in patients with NSCLC treated with gefitinib. Thus, SUVs might help identify patients who benefit from TKI treatment, but a large prospective study is required to confirm this.