Development of [¹¹C]erlotinib Positron Emission Tomography for In Vivo Evaluation of EGF Receptor Mutational Status

This article is featured in Highlights of This Issue, p. 1

In the Western population, approximately 10% of patients with non–small cell lung carcinoma (NSCLC) harbor an activating mutation in their tumor EGF receptor (EGFR) genes. The most common activating EGFR mutations are exon 19 deletions (particularly delE746–A750) and exon 21 point mutations (particularly L858R; refs. 1, 2). Proper identification of these patients, usually presenting with advanced stage disease, is of clinical importance, as recent trials have shown that treatment with EGFR tyrosine kinase inhibitors (TKI), such as erlotinib or gefitinib, results in a high response rate and significantly longer progression-free survival (PFS) than treatment with classical cytotoxic chemotherapy (3–5).

Identifying patients with activating mutations, however, remains a major clinical challenge. Obtaining representative tumor tissue samples for mutation analysis is often limited by practical issues, such as inability to reach the tumor site, low yields of malignant cells, or tumor heterogeneity. Indeed, results of recent clinical trials have shown that tissue procurement may not be possible in more than one fifth of patients (6). A noninvasive technique may overcome the obstacles associated with tumor tissue sampling. For this, positron emission tomography (PET), which allows for in vivo imaging of physiologic processes, seems to be best suited.

Compared with NSCLC tumors with wild-type EGFR, those with activating EGFR mutations have higher TKI binding affinity (7). Therefore, uptake of radiolabeled TKI may be higher in tumors with activating EGFR mutations. In the present study, [¹¹C]erlotinib, chemically identical to erlotinib itself, was used. Imaging of TKI-sensitive tumor xenografts has already been shown in an in vivo model using this tracer (8).

Although [¹¹C]erlotinib has been used to image NSCLC tumors in a limited number of patients, no quantitative analysis has been described (9, 10). Quantification, however, is essential for an objective comparison of uptake in different tumors. Therefore the aims of the present study were to develop the optimal tracer kinetic model and its corresponding measure for quantification of [¹¹C]erlotinib uptake in NSCLCs, to determine its reproducibility, and to assess whether tumor uptake of [¹¹C]erlotinib correlates with EGFR mutational status.

This study consisted of a number of distinct steps. Patients with and without an activating EGFR mutation underwent 2 dynamic [¹¹C]erlotinib PET scans with continuous arterial blood sampling on the same day. Each [¹¹C]erlotinib scan was preceded by a [¹⁵O]H₂O scan to assess tumor blood perfusion. First, [¹¹C]erlotinib data from all patients, including test–retest information, were used to determine the optimal model for describing erlotinib kinetics. This model was used to assess whether there were differences between tumors with and without activating EGFR mutations. Finally, validity of a noninvasive image derived input function was evaluated in an attempt to simplify the protocol for routine clinical applications.

Ten patients with NSCLCs were included, 5 with and 5 without an activating mutation. Inclusion criteria were histologic diagnosis of NSCLCs, EGFR mutational status as assessed by high resolution melting (HRM), and sequencing of tumor biopsies (11, 12), EGFR expression as determined by immunohistochemistry (IHC), an age of 18 years or older, a life expectancy of at least 12 weeks, presence of a malignant lesion within the chest of at least 1.5 cm diameter as measured by computed tomography (CT), and written informed consent. Exclusion criteria were claustrophobia, pregnancy, lactation, metal implants in the thorax, and use of concurrent or previous treatment with experimental drugs within 30 days before scanning. The study was approved by the Medical Ethics Review Committee of the VU University Medical Center.

[¹¹C]Erlotinib was synthesized under GMP conditions according to the EU directive on radiopharmaceuticals, EudraLex - Volume 4. The precursor was custom synthesized at Syncom. The carbon-11 label was introduced at the 7-methoxyethoxy position of the precursor, 6-O-desmethyl-erlotinib (OSI 420), using [¹¹C]methyl iodide (Supplementary Fig. S1). [¹¹C]methyl iodide was synthesized from [¹¹C]CO₂ according to standard procedures and subsequently distilled in a solution of the precursor in acetonitrile and tetra-n-butylammonium hydroxide. The maximum yield was obtained after 5 minutes at 80°C. After cooling down and dilution with 0.75 mL of water, the mixture was injected onto a Waters SymmetryPrep C18 7 μm 300 × 7.8 mm HPLC column, which was eluted with 70/30 25 mmol/L phosphate buffer (pH 3.5)/acetonitrile mixture. The product, [¹¹C]erlotinib, eluted at 9.5 minutes. This fraction was collected in 20 mL of water for injection. The total solution was passed over a preconditioned (10 mL of sterile ethanol 96% and subsequently 10 mL of water for injection) Waters Sep-Pak tC18 cartridge. This cartridge was washed with 20 mL of water for injection and subsequently the product was eluted from the cartridge with 1.0 mL of sterile ethanol (96%) and 14 mL of a sterile and pyrogen-free solution of 7.1 mmol/L NaH₂PO₄ saline. The final mixture was passed over a Millex GV 0.22 μm filter, yielding a sterile, isotonic and pyrogen-free solution of 2183-3476 MBq of [¹¹C]erlotinib with a (radio)chemical purity more than 98% and a specific activity of 184-587 GB/μmol at end of synthesis (n = 20). The product was analyzed using HPLC with a Phenomenex LUNA C18 5 μm 250 × 4.6 mm column, which was eluted with 55/45 acetonitrile/saline.

All patients underwent 2 scanning sessions on the same day, with an interval of at least 4 hours to allow for radioactive decay of carbon-11. Scans were conducted on a Gemini TF-64 PET/CT scanner (Philips Medical Systems). PET data were normalized and all appropriate corrections were applied for dead time, decay, randoms, scatter, and attenuation. Reconstruction of PET data was conducted using the 3-dimension (3D) RAMLA algorithm with CT-based attenuation correction, resulting in a final voxel size of 4 × 4 × 4 mm³ and a spatial resolution of 5 to 7 mm full width at half maximum. Before scanning, all patients were asked to fast from midnight, but a light breakfast before 8:00 am was allowed. First, a low-dose CT scan (50 mAs, without i.v. or oral contrast) was conducted. To assess tumor perfusion, 370 ± 37 MBq (mean ± SD) [¹⁵O]H₂O was injected intravenously, simultaneously starting a 10-minute emission scan in 3D acquisition mode. Subsequently 256 ± 53 MBq (mean ± SD) [¹¹C]erlotinib (corresponding to a nonpharmacologic dose of 2.2 ± 0.46 μg erlotinib) was injected intravenously, simultaneously starting a 60-minute emission scan in 3D acquisition mode. [¹⁵O]H₂O and [¹¹C]erlotinib emission scans were acquired in list mode and, before reconstruction, sorted into 26 (1 × 10, 8 × 5, 4 × 10, 2 × 15, 3 × 20, 2 × 30, and 6 × 60 seconds) and 36 (1 × 10, 8 × 5, 4 × 10, 2 × 15, 3 × 20, 2 × 30, 6 × 60, 4 × 150, 4 × 300, and 2 × 600 seconds) frames with increasing frame duration, respectively. No corrections for patient motion and/or respiratory motion were applied.

An indwelling cannula was inserted in the radial artery for arterial blood sampling. No blood was withdrawn during [¹⁵O]H₂O PET scans, as it has previously been shown that these scans can be quantified using an image-derived input function (13). During the [¹¹C]erlotinib PET scans, arterial blood was withdrawn continuously at a rate of 5 mL/min for the first 5 minutes followed by 1 mL/min for an additional 35 minutes. This was monitored using an online detection system (Veenstra Instruments), cross-calibrated against the PET scanner (14). In addition, manual arterial samples (7 mL) were taken at 2.5, 5, 10, 15, 20, 30, 40, and 60 minutes after injection of [¹¹C]erlotinib. These discrete samples were used for calibration of the online arterial curve and for measuring plasma-to-whole blood ratios and radiolabeled plasma parent and metabolite fractions as function of time.

Manual blood samples were analyzed for blood and plasma radioactivity concentrations and for radiolabeled fractions (i.e., parent and both polar and nonpolar metabolites) of [¹¹C]erlotinib. Whole blood (0.5 mL) was weighted in duplicate and after centrifuging (5 minutes; 7°C; 4,000 rpm.), plasma was harvested and 0.5 mL plasma was weighted in duplicate. A well-counter, cross-calibrated against the PET scanner was used to determine activity concentrations in whole blood and plasma.

To determine the fractions of [¹¹C]erlotinib and its labeled metabolites in each sample, solid-phase extraction (SPE) and high-performance liquid chromatography (HPLC) were used. About 1 mL of plasma was diluted with 2 mL of water and loaded onto an activated Waters Sep-Pak tC2 SPE column. To separate the parent compound from its metabolites, an HPLC system was used. The eluate was injected onto a Phenomenex Gemini 5 μm 250 × 10 mm with a flow of 3 mL/min. The gradient system was a mixture of acetonitrile (A) and 10 mmol/L ammonium acetate (B) and was programmed as following: t = 0 minute 90% B, t = 15 minutes 30% B, t = 20 minutes 30% B, t = 21 minutes 90% B, t = 25 minutes 90% B. Fractions were collected and measured for radioactivity with a gamma-counter to generate an HPLC profile.

First, images were converted to ECAT7 format. Regions of interest (ROI) were drawn manually within (i.e., smaller than) tumor contours on the CT images, avoiding blood vessels and necrosis as much as possible, using CAPP software (CTI/Siemens), and subsequently projected onto the corresponding PET images. CAPP software was also used to generate tumor time–activity curves (TAC) for all tumor ROI of both [¹⁵O]H₂O and [¹¹C]erlotinib scans.

Pharmacokinetic modeling was performed using in-house software, developed within the Matlab (The MathWorks Inc.) environment. First, using standard nonlinear regression techniques, tumor [¹¹C]erlotinib TAC were fitted to different (i.e., 1-tissue, irreversible 2-tissue, and reversible 2-tissue) compartment models (15) using the measured metabolite corrected plasma time–activity curve as input function. The optimal model was selected on the basis of both Akaike Information and Schwarz Criteria (16, 17). For the 1-tissue compartment model, the outcome measure of tracer uptake is the volume of distribution V_T, and for the irreversible 2-tissue compartment model, it is the influx rate constant K_i. In case of the reversible 2-tissue compartment model, both V_T and the nondisplaceable binding potential (BP_ND) can be used as outcome parameters. The most appropriate parameter was chosen depending on test–retest variability. When fitting to the different compartment models, a blood volume parameter was included to correct for the intravascular contribution to the signal.

A less invasive alternative to using continuous arterial blood sampling in [¹¹C]erlotinib kinetic analyses is the use of an image-derived input function (IDIF), however, results obtained with IDIF need to be validated against full arterial sampling. IDIF reflects the activity concentration in plasma over time and is derived from dynamic PET data of the arterial blood pool. To generate IDIF, ROIs were drawn within the ascending aorta of early [¹¹C]erlotinib frames (typically 30 to 35 seconds after injection) in 10 successive slices, resulting in an approximate volume of 6.3 mm³. By multiplying the arterial whole blood TAC derived from the aorta ROI with a multiexponential function, derived from the best fit to the plasma-to-whole blood radioactivity ratios of the discrete arterial samples, a plasma TAC was obtained. Then, the plasma TAC was corrected for metabolites using a sigmoid function derived from the best fit to the measured parent fractions of the arterial samples. Finally, this metabolite corrected plasma TAC was used as IDIF (18).

Using nonlinear regression, tumor [¹⁵O]H₂O TACs were processed according to the standard single tissue compartment model with IDIF as input function, as described previously (13). This analysis generated blood flow (F) values, reflecting tumor perfusion (mL/cm³/min).

Mutation analysis of EGFR was conducted using HRM and sequencing, as described previously (11, 12).

EGFR protein expression intensity was determined using IHC and scored visually, that is, semiquantitatively, as described previously (19).

Response during erlotinib treatment was measured using Response Evaluation Criteria in Solid Tumors (RECIST; ref. 20).

Statistical analysis was conducted using SPSS software (SPSS for Windows 15.0, SPSS, Inc.). Level of agreement between test and retest scans was determined using the intraclass correlation coefficient (ICC) with a 2-way random-effects model. Values for the ICC range from 0 to 1. Values close to 0 indicate poor agreement between test and retest scans, whereas values close to 1 indicate high agreement. ICC values above 0.70 are considered to have good reproducibility. In addition, the Mann–Whitney test was used to compare V_T and tumor perfusion (F) values between patient groups with and without activating EGFR mutations. Correlations were explored using the Spearman correlation coefficient (r_s). A 2-tailed probability value of P < 0.05 was considered significant. EGFR mutational status was known for all patients before kinetic analysis.

Baseline characteristics of all patients are shown in Table 1. Patients were divided into 2 groups. Group 1 consisted of patients with NSCLCs without activating EGFR mutations. Patient 5 in this group was diagnosed with an adenocarcinoma, harboring only an exon 20 insertion [D770–N771 (ins GG) + N771T]. This is a rare primary oncogenic mutation, insensitive to erlotinib, not considered to be an activating EGFR mutation, and thus classified in group 1 (21). All other patients in this group had a wild-type tumor EGFR. All patients in group 2 had an exon 19 deletion. High EGFR expression in all tumor cells, as assessed by IHC, was reported for 4 out of 5 tumors in group 1 and for all tumors in group 2. Examples are shown in Fig. 1.

Except for one, all patients underwent 2 scan sessions for a total of 19 [¹¹C]erlotinib scans. Patient 9 in group 2 missed the first [¹¹C]erlotinib scan due to technical problems with the arterial cannula.

Before this study, 5 patients had been treated with TKI. In group 1, patients 3 and 4 had received TKI as second-line therapy without experiencing any tumor response. In group 2, patients 7 and 10 initially had achieved a partial response on TKI, but treatment was discontinued because of drug toxicity and disease progression, respectively. Patient 8 achieved an initial partial response on TKI for almost a year. Upon first disease progression, she was treated with cytotoxic chemotherapy showing a partial response. After a second progression, she restarted TKI therapy, this time without tumor response (see Table 1).

After the [¹¹C]erlotinib study, none of the patients in group 1 and 3 patients in group 2 started treatment with TKIs. Patient 6 was treated with TKI for the first time, whereas patients 7 and 10 were treated again. All achieved a partial response. Patients 6 and 10 maintained this partial response for 18 and 4 months, respectively. To date, 20 months after start of treatment, patient 7 still has a partial response (see Table 1).

Tumor sizes ranged from approximately 5 to 100 mm³, with tumor ROI, drawn inside the tumor contours, varying between 2.4 to 91.4 mm³. Analysis of tumor TAC, using both Akaike and Schwarz criteria, showed that the reversible 2-tissue model provided the best fits in 15 of 19 scans. Therefore, further analyses were conducted using this model only. Figure 2 shows fits to a typical tumor TAC using the 3 plasma input models investigated. Fits for all different compartment models were corrected for intravascular activity by including a blood volume parameter. These blood volume fractions were non-zero in most patients (median 0.06; range, 0–0.40) for the defined ROI.

This model provides 2 potential parameters of interest, V_T and BP_ND. V_T was selected as outcome measure because of its superior test–retest variability [ICC for V_T and BP_ND were 0.88 (95% confidence interval, CI, 0.55–0.97) and −0.02 (95% CI, −0.64 to 0.62), respectively]. V_T results are shown in Table 2.

V_T of [¹¹C]erlotinib was higher in tumors with EGFR exon 19 deletions than in tumors without activating EGFR mutations. This result was found both for test and retest scans (P = 0.014 and P = 0.009, respectively; Table 2). Patient 9 in group 2 had a higher fitting error, which was probably due to the presence of lymphoid tissue in the tumor ROI. Representative parametric [¹¹C]erlotinib images, using Logan analysis over the interval 20 to 60 minutes (22), of patients with NSCLCs with wild-type and mutated EGFR are shown in Fig. 3.

V_T values obtained with metabolite-corrected IDIF (Table 2) correlated well with those obtained with full arterial sampling (r_s = 0.96, P < 0.001), considered to be gold standard, as indicated in Fig. 4. Using IDIF data, tumors with EGFR exon 19 deletions also had significantly higher V_T values than those without for both test and retest scans (P = 0.014 and P = 0.009, respectively). The test–retest variability (ICC = 0.85; 95% CI, 0.46–0.96) was similar as obtained with full arterial sampling.

The parent fraction in groups 1 and 2, indicating unchanged [¹¹C]erlotinib in plasma, steadily decreased to 54% ± 2% and 43% ± 7% (mean ± SD) at 60 minutes after injection, whereas polar radiolabeled metabolites steadily increased over 60 minutes to 31% ± 21% and 33% ± 17%, respectively. The fraction of nonpolar metabolites, as assessed by HPLC, was too small to be quantified reliably. Therefore, use of SPE only (i.e., measuring polar metabolites) was sufficient. Results for wild-type and mutated EGFR patients are shown in Supplementary Fig. S2. The fractions do not add up to 100%, because not all blood samples could be measured in all subjects.

Because of technical problems, only one [¹⁵O]H₂O scan was conducted in patients 2 and 9. There was no difference in tumor blood flow between both groups, with P values of 0.111 and 0.413 for test (n = 9) and retest (n = 9) scans, respectively. There was good test–retest consistency (ICC = 0.81, Table 2). Average [¹¹C]erlotinib K₁ values for wild-type and mutated EGFR were not significantly different (P = 0.372) with 0.30 (SD = 0.29) and 0.44 (SD = 0.13), respectively. Correlation between K₁ and flow was good (r_s = 0.70, P = 0.031).

The aim of this study was to develop a method to quantify [¹¹C]erlotinib uptake in NSCLC tumors, which could be used as an in vivo tool for detecting activating EGFR mutations. For quantification of PET tracer studies, the gold standard is kinetic modeling using a measured metabolite-corrected arterial plasma curve as input function. Using this approach, best fits were obtained with the reversible 2-tissue compartment model, providing volume of distribution (V_T) as a measure of [¹¹C]erlotinib uptake. The test–retest design of the present study showed that [¹¹C]erlotinib V_T values were reproducible (Table 2). This model was used in all further analyses.

[¹¹C]erlotinib V_T values obtained using IDIF correlated well with V_T obtained using arterial sampling (Fig. 4). Reproducibility of IDIF based [¹¹C]erlotinib V_T values was also good for all patients (Table 2). These findings validate the use of IDIF in future studies, making this technique better suitable for routine clinical use. Although continuous arterial sampling can now be omitted, image derived whole blood curves still need to be corrected for both plasma-to-whole blood radioactivity ratios and the fraction of radiolabeled metabolites. Even though this needs to be confirmed for [¹¹C]erlotinib, these measurements may be obtained from a number of discrete venous rather than arterial samples, as shown in studies using other tracers (23).

Unintentionally only patients with exon 19 deletions, and no patients with other activating mutations, such as L858R, were included in group 2. V_T of [¹¹C]erlotinib, as determined using both arterial sampling and IDIF, was significantly higher in group 2 than in group 1, for both test and retest scans. Although many factors contribute to tumor [¹¹C]erlotinib accumulation in PET, the differences in V_T may be caused by differences in affinity to erlotinib. Erlotinib competes with ATP to bind at the ATP-binding site of the EGFR kinase domain. Kinase assay studies have found values for the erlotinib dissociation constant (K_i) of 3.3, 6.3, and 17.5 nmol/L, and Michaelis–Menten constant (K_M) values for ATP of 129.0, 10.9, and 5.0 μmol/L for exon 19 deletions, exon 21 point mutations and wild-type EGFR, respectively (7). This means that erlotinib affinity is higher for EGFR with activating mutations than for wild-type EGFR, whereas ATP affinity is lower. Compared with wild-type EGFR, exon 19 deletions and exon 21 mutations exhibit 137-fold and 6-fold higher binding to erlotinib relative to ATP (i.e., K_i/K_M ratios), respectively (7).

Although high affinity of erlotinib for EGFR represents a favorable parameter for [¹¹C]erlotinib as a PET tracer, differences in K_i/K_M ratios between wild-type and mutated EGFR (i.e., >100-fold for exon 19 deletions) are not reflected in the same magnitude in V_T values (i.e., ∼2-fold). Clearly, in vitro studies occur in more simplified circumstances than in vivo studies, lacking complex interactions, such as nonspecific binding and the presence of radiolabeled metabolites that cause underestimation of V_T differences. In addition, in vitro conditions may deviate from normal physiologic conditions. For example, differences in ATP affinity between wild-type EGFR and mutated EGFR are seen at low ATP concentrations, but at higher, near physiologic ATP concentrations, these differences become less significant (7, 24). Other in vivo factors may also contribute to the observed differences in tumor [¹¹C]erlotinib uptake, such as EGFR protein expression and blood flow. However, both tumor EGFR protein expression and tumor blood flow were comparable between tumors with and without an EGFR exon 19 deletion.

In this study, 6 of 10 patients received erlotinib therapy, either before or after scanning. In group 1, only 2 patients were treated with erlotinib. This was before scanning, and none of the 2 low V_T patients achieved tumor response. Three patients in group 2 were treated with erlotinib after scanning. These 3 high V_T patients achieved a partial response. The remaining patient in group 2 (patient 8) developed resistance to erlotinib, which was diagnosed 3 weeks before her [¹¹C]erlotinib PET. Interestingly, her V_T was higher than in the nonmutated group, but the lowest observed in the mutated group, which could be in accordance with her clinically diminished TKI sensitivity. Although the sample size in this study is too small to draw any definite conclusions, these findings support the hypothesis that increased [¹¹C]erlotinib V_T correlates with tumor response to erlotinib. Clearly, these findings need to be confirmed in larger trials.

Recently, results of a qualitative study on 13 patients undergoing [¹¹C]erlotinib scanning were reported (9). Four patients died before response evaluation. Of the 9 patients who were evaluable for tumor response to erlotinib therapy, all 3 patients with high [¹¹C]erlotinib accumulation showed stable disease, whereas 4 of 6 patients with lower accumulation had progressive disease and the remaining 2 had stable disease. Unfortunately, EGFR mutational status was not measured in this study. Nevertheless, these results also indicate that [¹¹C]erlotinib uptake may predict tumor sensitivity to erlotinib therapy.

Clinical response to TKI depends on EGFR affinity to TKI and ATP but also on oncogene addiction of tumor cells. As, in clinical practice, up to 80% of patients harboring activating EGFR mutations respond to TKI, it could be hypothesized that responders have EGFR-driven growth and nonresponders may have other mechanisms that activate proliferation. Whether tumor growth in a patient is fully driven by the EGFR pathway cannot be measured using [¹¹C]erlotinib PET.

This study has a number of limitations. First, being a proof-of-concept study, only a limited number of patients were included. According to the inclusion criteria, all activating EGFR mutations, including exon 19 deletions and exon 21 point mutations, were allowed to enter group 2. Unintentionally, only patients with exon 19 deletions were included. The fact that patients with exon 21 point mutations (L858R) were not included possibly avoided confounding effects caused by differences in TKI affinity between both mutations. Cell-based studies have shown that EGFR with exon 19 deletions are more sensitive to TKI than L858R EGFR (7, 24). On the other hand, clinical trials have shown highly comparable response rates in both mutations. This suggests that TKIs have sufficient affinity to provoke similar initial tumor responses in both types of mutations. [¹¹C]Erlotinib uptake in patients with exon 21 point mutations is unknown and should be investigated in future studies.

Full kinetic modeling and arterial blood sampling were used to derive the volume of distribution of [¹¹C]erlotinib. This approach provides the most accurate assessment of erlotinib uptake, which is important in the present proof-of-concept study. It is, however, less suitable for routine clinical studies. and further studies are needed to assess simplified analytic methods, such as the standardized uptake value (SUV). Preliminary data indicate that an SUV analysis provides a poorer discrimination between both patients groups. On the other hand, tumor-to-blood ratios appear to be more promising. A full comparison is in progress. Simplified methods are also relevant for whole body scans, enabling the study of interlesional heterogeneity.

Differences in V_T between both patient groups could potentially be due to differences in nonspecific binding. The nonspecific component of V_T given by K₁/k₂, however, did not significantly differ between groups. Ideally, this should be confirmed in an independent measurement. One possibility would be to measure V_T after a blocking dose with cold erlotinib. This is, however, not possible in patients and needs to be conducted in future animal studies.

V_T differences between groups appear modest, when compared with the differences in K_i/K_M ratios between mutated and wild-type EGFR. However, V_T values between groups are clearly distinct, with excellent test–retest repeatability, justifying further validation studies with larger numbers of patients, to answer the question whether V_T can be used to effectively distinguish between patients who will and will not benefit from TKI treatment.

In this study, using SPE followed by HPLC, the radioactive parent compound fraction was discriminated from the radioactive metabolite fractions in arterial plasma samples. Clearly, metabolite fractions in tumors were not determined. Previous pharmacology studies have shown that the main metabolite of erlotinib is OSI-420, the precursor to which the C-11 label is introduced in the synthesis of [¹¹C]erlotinib. After losing the C-11 label at the 7-methoxyethoxy position, OSI-420 becomes nonradioactive and does not show up on the PET scan. This metabolism does result in small polar radiolabeled metabolites such as C-11 formaldehyde, C-11 formic acid, and/or C-11 carbon monoxide. These metabolites could cause some bias, that is, an underestimation, of V­_T differences between groups.

Biopsies were not taken at the time of scanning, as indicated in Table 1. No biopsies were taken for the purpose of the present study. This is a relevant limitation, especially for patients with exon 19 deletions who were treated with TKI (i.e., patients 7, 8, and 10) before scanning, as they may have developed secondary mutations, for example, T790M, that could decrease the affinity of EGFR for [¹¹C]erlotinib. In patient 7, no secondary mutations were found in biopsies taken 1 month before scanning. This biopsy was considered representative. In addition, to date (i.e., 22 months), this patient has maintained a partial response to erlotinib. Patient 8 and 10 underwent an additional biopsy, revealing the presence of T790M, 6 and 12 months after scanning, respectively, indicating that disease progression was probably due to this secondary mutation. It is believed that T790M mutations are already present in tumor cells with activating mutations, and that during TKI therapy, a clonal selection takes place. This means that both patients 8 and 10 could have had T790M-bearing tumor cells during scanning, but to which extent and how this changed [¹¹C]erlotinib uptake is unclear.

More studies are needed to validate and substantiate the present findings. In future research, the use of simplified parameters such as SUV should be investigated. In addition, future research needs to focus on heterogeneity in tumor [¹¹C]erlotinib V_T, especially after initiation of TKI therapy and after disease progression occurs. In case of emerging resistance to TKI, [¹¹C]erlotinib PET may reveal residual TKI-sensitive tumor sites, which could justify continuation of TKI within a more complex treatment protocol. As all TKI-treated patients develop resistance, this is an important clinical issue. Furthermore, the use of static whole body [¹¹C]erlotinib scans could provide a means to evaluate distant metastatic lesions and their TKI sensitivity. The present study with [¹¹C]erlotinib can serve as a template for other studies using radiolabeled TKI that target different tyrosine kinases and activate proliferative pathways in a variety of other malignancies.

In conclusion, the present study shows that measuring [¹¹C]erlotinib uptake by V_T using PET is feasible. [¹¹C]Erlotinib V_T is higher in patients with NSCLCs with EGFR exon 19 deletions. Therefore, this noninvasive in vivo method shows promise as a tool for individualizing therapy by identifying those patients who may benefit from TKI therapy.

No potential conflicts of interest were disclosed.

Conception and design: I. Bahce, E.F. Smit, A.A.M. van der Veldt, A.D. Windhorst, P.E. Postmus, A.A. Lammertsma, N.H. Hendrikse

Development of methodology: I. Bahce, E.F. Smit, M. Lubberink, M. Yaqub, A.D. Windhorst, A.A. Lammertsma

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): I. Bahce, E.F. Smit, R.C. Schuit, E. Thunnissen, D.A.H. Heideman, P.E. Postmus

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): I. Bahce, E.F. Smit, M. Lubberink, A.A.M. van der Veldt, M. Yaqub, A.A. Lammertsma

Writing, review, and/or revision of the manuscript: I. Bahce, E.F. Smit, M. Lubberink, A.A.M. van der Veldt, M. Yaqub, A.D. Windhorst, E. Thunnissen, D.A.H. Heideman, P.E. Postmus, A.A. Lammertsma, N.H. Hendrikse

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A.D. Windhorst, R.C. Schuit, E. Thunnissen, D.A.H. Heideman

Study supervision: E.F. Smit, A.A. Lammertsma, N.H. Hendrikse

The authors thank Suzette van Balen and Femke Jongsma for assistance with PET scanning and planning, Ronald Boellaard for advice on tracer kinetic modeling, and the B.V. Cyclotron VU for production and delivery of [¹¹C]CO₂.

This study was conducted within the framework of the Center for Translational Molecular Medicine, Airforce project (grant 03O–103).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.