[18F]-3'Deoxy-3'-fluorothymidine positron emission tomography and breast cancer response to docetaxel.

PURPOSE
To establish biomarkers indicating clinical response to taxanes, we determined whether early changes in [(18)F]-3'deoxy-3'-fluorothymidine positron emission tomography (FLT-PET) can predict benefit from docetaxel therapy in breast cancer.


EXPERIMENTAL DESIGN
This was a prospective unblinded study in 20 patients with American Joint Committee on Cancer (AJCC) stage II-IV breast cancer unresponsive to first-line chemotherapy or progressing on previous therapy. Individuals underwent a baseline dynamic FLT-PET scan followed by a scan 2 weeks after initiating the first or second cycle of docetaxel. PET variables were compared with anatomic response midtherapy (after 3 cycles).


RESULTS
Average and maximum tumor standardized uptake values at 60 minutes (SUV(60,av) and SUV(60,max)) normalized to body surface area ranged between 1.7 and 17.0 and 5.6 and 26.9 × 10(-5) m(2)/mL, respectively. Docetaxel treatment resulted in a significant decrease in FLT uptake (P = 0.0003 for SUV(60,av) and P = 0.0002 for SUV(60,max)). Reduction in tumor SUV(60,av) was associated with target lesion size changes midtherapy (Pearson R for SUV(60,av) = 0.64; P = 0.004) and predicted midtherapy target lesion response (0.85 sensitivity and 0.80 specificity). Decreases in SUV(60,av) in responders were due, at least in part, to reduced net intracellular trapping of FLT (rate constant, K(i)). Docetaxel significantly reduced K(i) by 51.1% (±28.4%, P = 0.0009).


CONCLUSION
Changes in tumor proliferation assessed by FLT-PET early after initiating docetaxel chemotherapy can predict lesion response midtherapy with good sensitivity warranting prospective trials to assess the ability to stop therapy in the event of non-FLT-PET response.


Introduction
Taxanes are some of the most commonly used drugs in the treatment of breast cancer today. A molecular imaging biomarker which could provide early accurate prediction of response or resistance in vivo would be of great benefit to oncologists. The cytotoxic activity of taxanes is promoted by stabilizing cellular microtubule assembly (1,2). This leads to inhibition of mitotic cell division (the G 2 -M phase of the cell cycle; ref. 2) and cell death by apoptosis (3). Docetaxel is approved for the treatment of patients with locally advanced or metastatic breast and non-small cell lung cancer (4) who have undergone anthracycline-based chemotherapy and failed because of tumor progression or relapse. The drug has shown effectiveness against tumors resistant to anthracyclines or paclitaxel (5,6) and good response rates (55%-68%) in the neoadjuvant setting (7,8), at the expense of well-known myelosuppressive and other toxicities.
In patients undergoing chemotherapy, systemic toxicity remains the major limitation to adequate dosing. The ability to predict early response after 1 to 2 cycles of docetaxel would avoid unnecessary toxicities in patients unlikely to respond with potential impact on clinical outcomes and quality of life, particularly in the palliative setting. We explored the performance of [ 18 F]-3 0 deoxy-3 0fluorothymidine positron emission tomography (FLT-PET) as an early biomarker of tumor response in this setting. FLT uptake into tumor cells is regulated by equilibrative nucleoside transporter (ENT-1) and thymidine kinase1 (TK1; refs. [9][10][11]; the latter controlled during cell division transcriptionally in the G 1 to S-phase of the cell cycle (12). FLT uptake has been shown to strongly correlate with cell proliferation in untreated patients with breast cancer and further that response by PET imaging is detectable as early as 1 week following fluorouracil-epirubicin-cyclophosphamide (FEC) chemotherapy (13,14). It is, however, unknown whether the antiproliferative activity of taxanes, an effect that is largely G 2 -M in nature, could be detected in patients by FLT-PET. We studied the effects of docetaxel on FLT-PET parameters in this setting.

Patients
This was a prospective, unblinded, single institution study in patients with American Joint Committee on Cancer (AJCC) stage II-IV breast cancer, including primary breast cancer being unresponsive to first-line chemotherapy (via ultrasound scanning done by the same radiologist to minimize operator-dependent errors), or metastatic disease progressing on previous therapy (with either new lesions appearing or >20% increase in maximum diameter of previous target lesions). Patients underwent a baseline dynamic FLT-PET scan for 66.5 minutes followed by a similar posttreatment scan conducted approximately 14 days after the first or second cycle of docetaxel therapy (dose of docetaxel: 100 mg/m 2 ). Primary patients were selected if they were not responding or progressing on chemotherapy. Metastatic patients were those who were documented to have progressive disease on last computed tomographic (CT) scan from prior therapy. The main inclusion criteria were histologically confirmed invasive breast cancer with at least one measurable lesion outside the bone or liver, accurately measured in one dimension (longest diameter recorded) as 20 mm or greater using conventional techniques. This was done to reduce the effect of partial volume effect (15,16). No chemotherapy or radiotherapy was permitted for at least 3 weeks before undergoing first FLT-PET scan. Patients having only bone or liver metastasis were excluded as physiologic uptake of FLT is high in bone and liver and would have obscured any tumor(s) in these organs. The study received ethical approval from the Hammersmith Hospital Research Ethics Committee and a license to inject FLT into subjects from the Administration of Radioactive Substances Advisory Committee. The study was conducted in accordance with the Declaration of Helsinki. All patients gave written informed consent.

Study design and statistical considerations
Previous test-retest reproducibility studies showed that changes in FLT standardized uptake value (SUV) of more than 20% or net irreversible uptake (K i ) of more than 31% could be classified as radiotracer responsive with 95% confidence interval (SD: 10%-15%; ref. 14). Published data indicate a 40% response rate (worse case) to docetaxel therapy in patients with metastatic cancer (6,8). This would dilute the expected "group" effect from 0.31 to 0.124 and inflate the variability for measuring change to 18%. Power calculation (paired statistics) suggested that a number (n ¼ 20) of patients would give to the design a minimum power of 0.80 to detect a group difference in FLT at a 0.05 error rate. Comparisons of tumor results were done on a lesion-by-lesion basis assuming non-Gaussian distribution (Wilcoxon matched pairs test or Mann-Whitney test). The association between FLT retention variables and midtherapy lesion response was explored using Pearson correlation coefficient (R, for continuous data) or c 2 and Bland-Altman plots (for categorical data). Statistical analyses were conducted using GraphPad Prism version 4.00 for Windows (GraphPad Software; www.graphpad.com).

PET imaging protocol and radiology
FLT was manufactured as previously described (13) with 99.6% radiochemical purity and a mean specific radioactivity of 147,584 MBq/mmol. PET scanning was conducted on the GE-Discovery PET-CT scanner (GE-Healthcare) with a 15.7 cm field of view. Patient preparation for PET has been previously described (13). FLT [mean: 210 (AE8) MBq] was injected as a slow bolus in about 5 mL of saline followed by emission scanning for 66.5 minutes. The dose of radioactivity used was in accordance with suggested limits described by Vesselle and colleagues (17). Data were binned into 28 frames and reconstructed. Volumes of interest were drawn on at least 4 slices of tumor (using the combined PET and CT volume) to obtain the activity concentration within the volumes of interest. SUV was calculated by normalizing tumor radiotracer concentration (kBq/mL) to injected radioactivity (in kBq) and body surface area as previously described (13,14). Average SUV (SUV 60,av ) or the maximum pixel value (SUV 60,max ) were determined. Kinetic modeling of PET data was conducted to calculate K i (13,14).
Tumor target lesion size was measured by ultrasound (for primary lesions) or CT (for metastatic disease) prior to treatment and after 3 cycles of docetaxel, and response was assessed according to the Response Evaluation Criteria in Solid Tumors (RECIST) 1.1 (18). Response by FLT-PET was

Translational Relevance
This study investigated the use of [ 18 F]-3 0 deoxy-3 0fluorothymidine (FLT) PET as an imaging biomarker to measure early response to docetaxel chemotherapy in breast cancer. We found that early changes in tumor FLT uptake predicted lesion size reduction midtherapy. This technology will be particularly useful in those patients where early response to therapy may aid definitive treatment. Further verification of this work in a larger population, could permit the ability to stop therapy in the event of non-FLT-PET response sparing patients of cumulative toxicity. defined as more than 20% change for SUV (>31% change for K i ; ref. 14).

Clinical results and visualization of lesions by PET
Twenty patients (11 primary and 9 metastatic; average age, 54 years; average tumor size, 38.4 mm) successfully completed both pre-and posttreatment FLT-PET scans ( Table 1). All lesions were seen on FLT-PET ( Fig. 1). In most patients, posttreatment FLT-PET was conducted approximately 14 days after initiating cycle 1; in 6 patients (patient no. 3, 4, 5, 9, 15, and 17), the posttreatment PET was conducted 2 weeks after initiating cycle 2 of docetaxel either due to patients being extremely unwell from chemotherapy or failure of FLT production. Two individuals (no. 11 and 12) developed grade III toxicity to docetaxel after the first cycle (needing hospitalization for neutropenic sepsis) and treatment was then changed to weekly paclitaxel based on the clinicians decision. Patient no. 11 had 1 cycle of docetaxel and 3 cycles of paclitaxel and patient no. 12 had 1 cycle docetaxel and 2 cycles of paclitaxel prior to RECIST assessment. PET uptake data for these 2 cases have been presented, but these data were not used for further comparison with RECIST response.

PET imaging outcome
The median (AESD) pretreatment SUV 60,av of tumors was 6.47 Â 10 À5 (AE4.1 Â 10 À5 ) m 2 /mL and that for SUV 60,max was 12.6 Â 10 À5 (AE6.11 Â 10 À5 ) m 2 /mL (Supplementary  Table S1). In contrast, the median SUV 60,av in normal breast tissue and normal lung were 1.12 Â 10 À5 (AE0.4 Â 10 À5 ) m 2 /mL and 0.99 Â 10 À5 (AE0.33 Â 10 À5 ) m 2 /mL, respectively. The visual baseline uptake of FLT was low in 3 tumors (patient no. 4, 8, and 9). Of interest to use FLT-PET for assessing early response, single-agent docetaxel chemotherapy resulted in a significant decrease in FLT uptake ( Fig. 1 and Table 2; P ¼ 0.0003 for SUV 60,av and P ¼ 0.0002 for SUV 60,max ; Wilcoxon matched pairs test, for all tumors). The median percentage decrease of SUV 60,av for all tumors was 30.7% (median absolute decrease of 2.32 Â 10 À5 m 2 /mL) and for SUV 60,max , a decrease of 17.5% was observed (median absolute decrease of 3.37 Â 10 À5 m 2 / mL). The median decrease in SUV 60,av in PET responders measured 40.2% versus 10.5% in nonresponders (Mann-Whitney test: P ¼ 0.01). In one metastatic breast tumor of the lung (patient no. 17), the tumor had 30.2% and 41.8% reductions of SUV 60,av and SUV 60,max , respectively, after 2 cycles of docetaxel with complete radiological response on CT after 3 cycles of docetaxel. For comparison, tumor SUV normalized to body weight rather than body surface area are shown in Supplementary Table S2. Comparison of PET outcome with categorical target lesion response We compared PET response with clinical outcome for individual target lesions (Table 2). Overall (excluding patient no. 11 and 12), there was 1 complete responder, 12 partial responders, and 5 patients with stable disease. In 2 patients (no. 6 and 14), new metastatic tumor deposits were detected outside the PET field of view during the subsequent routine clinical cross-sectional imaging. We investigated whether reductions in tumor FLT uptake early after initiating therapy were associated with target lesion size changes after 3 cycles (Fig. 2). There was a significant correlation between FLT-PET early signal change and lesion size midtherapy (Fig. 3A). Pearson correlation coefficient was R ¼ 0.64 (P ¼ 0.004) for SUV 60,av ; this association was less significant for SUV 60,max (R ¼ 0.53; P ¼ 0.02). Categorization of the target lesion size data into two RECIST (19) classes (complete þ partial response vs. stable disease) and of the FLT-PET SUV data into two classes (reduction < 20% vs. reduction > 20%), confirmed a valid association (c 2 test) when the PET measure was SUV 60,av (P ¼ 0.05) but not SUV 60,max (P ¼ 0.16). On the basis of the RECIST stratification (19), 13 of 18 tumors responded to therapy (Table  2), and the median percentage decrease in SUV 60,av and SUV 60,max in these tumors were 40.2% and 34.1%, respec-tively. FLT-PET SUV 60,av correctly detected 11 of the 13 RECIST responders. One tumor (patient no. 9) showed a significant decrease in SUV 60,av of 31.2%, but this was deemed borderline stable disease by tumor size measurement using ultrasonography, with tumor size reduction of 21.8%. This is clearly depicted on Bland-Altman plots devised to further explore the association between categorical target lesion size response and changes in PET semiquantitative parameters (Fig. 3B). The parameter, SUV 60,av , appeared to have better sensitivity to predict categorical lesion size response than SUV 60,max . Using the 20% change in SUV 60,av as (prospective) cutoff point to stratify patients as PET responders (14), the sensitivity and specificity of SUV 60,av to predict categorical tumor size response were 0.85 and 0.80, respectively. SUV 60,max was less robust with sensitivity of 0.62 but equal specificity of 0.80. With respect to predicting no lesion response, both SUV 60,av and SUV 60,max conducted well except in one patient (patient no. 9).

Kinetic analysis of tumor FLT uptake
We assessed the steady-state net irreversible flux constant for transfer of FLT from plasma to tumor (K i ; Table 3) as a means of explaining the SUV data. This was done by a metabolite-corrected modified Patlak graphical approach as previously described (13,20). The median (AESD) K i at baseline was 2.95 Â 10 À4 (AE2.44 Â 10 À4 ) mL plasma/mL tissue/s and with good fits (low SD of fit). Drug treatment significantly reduced K i (Wilcoxon matched pairs test: P ¼ 0.0009) with a median percentage change (AESD) of 51.12% (AE28.39%) mL plasma/mL tissue/s. For patient no. 9, it was not possible to calculate K i accurately due to low uptake that decreased toward the end of the scan (possibly due to NOTE: Negative value implies decrease in PET parameter. Response in parentheses indicates overall response in body after docetaxel chemotherapy.Abbreviations: CR, complete response; PD, progressive disease; PR, partial response; SD, stable disease.  dephosphorylation or contribution of labeled metabolites). The drug did not affect FLT metabolism. The median (AESD) percentage of parent FLT in plasma at 60 minutes was 77.6 (AE0.07)% and 76.5 (AE0.07)%, respectively, before and after docetaxel. Like SUV 60,av , K i also showed a significant association with midtreatment target lesion size change (Pearson R ¼ 0.56; P ¼ 0.02; Table 3). With the exception of patients 6 and 13, K i was found on lesion-by-lesion comparison to be low (>31%) in tumors that had low SUV 60,av (of >20%). This signifies that changes in tumor uptake could be explained largely by net irreversible retention of the radiotracer.

Discussion
The future of taxane therapy includes defining optimal combinations and dosing frequency and strategies to reduce toxicity, as well as trials of novel second-generation taxanes. We showed that early in the course of therapy, an FLT-PET reduction may predict response to docetaxel. FLT-PET has been shown to image proliferation in breast and lung tumors with high reproducibility and accuracy (14,21,22). Furthermore, decreases in FLT uptake depicting reductions in cell proliferation within 1 to 2 weeks have been reported in breast cancer (14) and non-Hodgkin lymphoma (23). Most patients in this present trial showed an early reduction in FLT-PET 2 weeks after commencing the first or second cycle of docetaxel. This FLT-PET response study of taxanes in patients are in keeping with intrinsic reductions in FLT uptake reported by Dittmann and colleagues in breast cancer cells (24) and hormone-refractory prostate cancer xenografts (25). This suggests that the activity of docetaxel is detectable by FLT-PET.
The PET technique estimates phosphorylation of FLT to FLT-phosphate. The wide range of pretreatment FLT uptake values may be explained by the wide variation in proliferation (13,26,27). Other mechanisms including relative use of salvage versus de novo precursors for DNA synthesis could also influence FLT uptake (28,29). We used the SUV at 60 minutes as a semiquantitative measure of this biochemical activity. In most patients, tumor FLT SUV 60,av was higher than the 95% confidence limit of normal breast (1.9 Â 10 À5 m 2 /mL). In patient 8 who received 5 cycles of FEC before entering the study, baseline tumor FLT uptake although detectable above background, was only 1.73 Â 10 À5 m 2 /mL. In general, however, baseline FLT uptake did not appear to be confounded by prior treatment. We used a prospective threshold (14) to define statistically significant FLT response. Overall, the magnitude of tumor FLT reduction was as expected more in clinical responders than nonresponders. Changes in tumor FLT SUV did not differ between patients who had a posttreatment scan after 1 or 2 cycles or between primary and metastatic disease (Mann-Whitney test: P ¼ 0.6).
We found a good association between changes in FLT uptake and changes in lesion size. This was confirmed when the data were categorized (Fig. 3). In this study, 2 tumors showed significant reductions in SUV 60,av but were classified borderline stable disease (Fig. 3). Despite these important differences, changes in SUV 60,av (after 1 or 2 cycles) overall predicted changes in tumor size after 3 cycles with high sensitivity and specificity (0.85 and 0.80, respectively). Analysis of receiver operating characteristics indicated that the optimal sensitivity and specificity (100% and 57.1%, respectively; data not shown) would be achieved when the cutoff value was 11.8%. Notably, using that cutoff point, the high sensitivity was linked to a much lower specificity. Of interest, FLT-PET test-retest reproducibility studies by de Langen and colleagues (30) in non-small cell lung and head and neck cancers suggested that a cutoff value of more than 15% qualified as response. Similarly, Shields and  colleagues reported a cutoff point of 21% as FLT-PET response in lung cancer (31). We did not consider new lesions in the determination of sensitivity and specificity, as only target lesions in the PET field of view were assessed. Future studies should consider whole body PET imaging to permit additional lesions to be detected. Lack of a significant change in FLT-PET will indicate futile treatment and signify a change to alternative treatment. In contrast, an FLT-PET response will indicate reduction in proliferation that probably warrants another assessment (say after 1 additional cycle) to confirm efficacy.
Muzi and colleagues have suggested that, unlike labeled thymidine, FLT phosphorylation may be reversible in somatic tumors making simple variables such as SUV 60,av liable to error (32). Hence, we evaluated the net unidirectional flux (constant, K i ) of FLT from plasma into the intracellular space (via transport and phosphorylation) at steady state. A linear phase was discernible in the graphical plots for most patients indicating net retention. Noisy fits were obtained for tumors (patient no 4, 8, and 9) with low proliferation (SUV 60,av < 3 Â 10 À5 m 2 /mL; K i < 1 mL plasma/mL tissue/s). Patient 9 had an uptake profile characterized by an initial positive linear slope followed by a downward slope, consistent with dephosphorylation of FLT-phosphate. Further compartmental analysis supported this assertion; there was a reasonable association between K i and k 3 (R ¼ 0.42, P ¼ 0.02; data not shown). These data indicate that although simple measures of FLT phosphorylation may have value for widespread clinical use, analysis may be limited in low proliferative tumors, in this case, tumors with SUV 60,av values of 3 Â 10 À5 m 2 /mL (1.2 when normalized to body weight) or less.
There were limitations of this study that are worthy of discussion. Both primary and metastatic patients, who may respond differently to docetaxel (4,5,8), were included in this study potentially impacting on heterogeneity of response. Notably, however, target lesion changes measured after 3 cycles did not appear to differ between the 2 groups (Mann-Whitney test: P > 0.05). While the overall sample size was small (sample size powered to detect a 20%-31% change in PET variables), this represents a single tumor type group of patients and with the same treatment and allowed detailed assessment of both simple measures of uptake, as well as more detailed analysis of radiotracer kinetics. In this study, tumor size reduction after 9 weeks was used for comparison with changes in tumor FLT uptake at 2 weeks after cycle 1 or 2. Given that FLT-PET changes occur rapidly after treatment (14), a potential limitation of the study was the assessment of PET response after 2 cycles in 6 patients (compared with one cycle in the rest). The reduction in tumor SUV was, however, not statistically different between the 2 groups (Supplementary Table S3, and, therefore, reported separately but combined for analysis). For these patients who had FLT studies after 2 cycles, the benefit of using FLT to predict response would be less obvious. Another limitation is the use of RECIST rather than histopathology for comparison. RECIST response was determined at the physician's discretion earlier after cycle 2 in one patient (patient 14) due to suspicion of clinical disease progression. Though this could be seen as a limitation of the study, it did not affect the overall prediction of response with FLT-PET. RECIST after 3 cycles is often used to assess whether a drug has biologic activity, though its correlation with survival is imperfect (18). It will be interesting in future studies to investigate histologic markers of proliferation (primary disease; this will involve repeat biopsies) and/or progression-free and overall survival as endpoints in a larger more homogeneous setting. Finally, on the basis of our prospective cutoff value, two patients would potentially have been denied further docetaxel treatment on the basis of early PET response. This highlights the ethical challenge of defining the ideal cutoff value for response, that is, when sensitivity is not high enough. Decreasing the cutoff value should improve sensitivity but at the cost of specificity.
In aggregate, FLT-PET is a promising imaging biomarker to detect early response to docetaxel in breast tumors and could potentially enable tailoring of therapy to individual patients warranting prospective trials to assess the ability to stop therapy in the event of non-FLT-PET response. As a next step, the promising results should be confirmed in larger trials in a more homogeneous setting and with longer follow-up, with disease-free survival and perhaps overall survival as additional endpoints. SUV 60,av reductions were due, in part, to inhibition of tracer retention and correlated with target lesion size response after 3 cycles of chemotherapy. The outcome of this study may have general application for testing of new taxanes or new taxane regimes.

Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.