Kinetic Analysis of 2-[ 11 C]Thymidine PET Imaging Studies of Malignant Brain Tumors: Preliminary Patient Results

Abstract

2-[¹¹C]Thymidine (TdR), a PET tracer for cellular proliferation, may be advantageous for monitoring brain tumor progression and response to therapy. Kinetic analysis of dynamic TdR images was performed to estimate the rate of thymidine transport (K_1t) and thymidine flux (K_TdR) into brain tumors and normal brain. These estimates were compared to MRI and pathologic results. Methods: Twenty patients underwent sequential [¹¹C]CO₂ (major TdR metabolite) and TdR PET studies with arterial blood sampling and metabolite analysis. The data were fitted using the five-compartment model described in the companion article. Results: Comparison of model estimates with clinical and pathologic data shows that K_1t is higher for MRI contrast enhancing tumors (p > .001), and K_TdR increases with tumor grade (p > .02). On average, TdR retention was lower after treatment in high-grade tumors. The model was able to distinguish between increased thymidine transport due to blood–brain barrier breakdown and increased tracer retention associated with tumor cell proliferation. Conclusion: Initial analysis of model estimates of thymidine retention and transport show good agreement with the clinical and pathological features of a wide range of brain tumors. Ongoing studies will evaluate its role in measuring response to treatment and predicting outcome.

Introduction

MRI and [¹⁸F]FDG PET scans have utility in detecting malignant brain tumors. However, these modalities are not optimal for monitoring tumor progression and response to therapy as MR images can be ambiguous due to blood–brain barrier (BBB) breakdown caused by tumor extensions or as a consequence of therapy [1], and quantitative measures of tumor growth from tumor energy metabolism can be misleading [2,3]. Our group has investigated PET imaging using 2-[¹¹C]thymidine (TdR) as a quantitative, in vivo, method of evaluating tumor cellular proliferation (recently summarized in Mankoff et al. [4] and Krohn et al. [5]. We showed previously that images of [¹¹C]TdR uptake reflect early response to therapy and that patients undergoing chemotherapy showed larger and more consistent decreases in TdR tumor uptake after therapy compared to FDG [6]. Our conclusion is that cellular proliferation imaging may offer significant advantages in monitoring treatment, although kinetic analysis must be performed as static images of thymidine uptake can be misleading due to the high normal background and increased transport across a damaged BBB [7].

We previously described a five-compartment model for 2-[¹¹C]TdR that has been validated for somatic tissues and successfully applied to patient studies [8]. The model accurately estimated the thymidine incorporation rate into DNA in simulations and in an animal model [8,9]. However, since the intact BBB restricts the transport of thymidine [10] and BBB disruptions can be caused by the tumor or as a result of therapy [1], further analysis was necessary to demonstrate that the model could distinguish between increased thymidine transport across a damaged BBB and increased DNA synthesis in proliferating tissue when applied to brain images. In addition, although the BBB restricts thymidine transport, it is much less a barrier to the principal metabolite, [¹¹C]CO₂. Labeled metabolites can therefore result in a high background in thymidine PET brain tumor imaging and the kinetic analysis must be able to separate the contribution of metabolites from thymidine uptake and retention.

Preliminary analysis showed that the model adequately described tracer kinetics in normal brain and brain tumors [7]. In a companion article, we presented studies that tested the detailed behavior of the [¹¹C]TdR model in normal brain and brain tumors [11]. Simulations were performed that validated the ability of the model to estimate thymidine flux constant into tumor DNA and tested whether the model could distinguish between thymidine transport and retention. The model estimated the rate of thymidine incorporation into DNA with a standard error of approximately 10% and the rate of thymidine transport from blood to tissue with a standard error of approximately 15%.

In this article, we report on the application of the model to data obtained from a series of patients with a variety of brain lesions. These studies investigate the performance of the model in a wide range of clinical scenarios: high- and low-grade glial tumors, necrosis, and treated and untreated tumors. We compare the estimates of thymidine transport and retention with clinical and pathological features associated with tumor proliferation and tracer transport (contrast enhancement on MRI, tumor grade, and treatment status), and analyze the relationship between transport and retention in patient data.

Materials and Methods

The Model

The kinetic model for 2-[¹¹C]TdR that was previously validated for somatic tumors is shown in Figure 3 of the companion article [11]. Three compartments describe the behavior of thymidine, CO₂, and non-CO₂ metabolites, driven by three input functions measured by arterial blood sampling and metabolite analysis. The assumptions made to simplify the model are described in a previous article [8].

The thymidine compartment set has a reversible tissue compartment (A) that includes intracellular thymidine and thymidine nucleotides. The fixed thymidine tissue compartment (B) represents thymidine incorporated into DNA and is irreversibly bound. Three rate constants are used to model the kinetics of the thymidine compartment set.

The non-CO₂ metabolite compartment set has a single reversible compartment (C) that includes labeled thymine, dihydrothymine (DHT), and β-ureidoisobutyric acid (BUIB) and can be modeled using two rate constants. The CO₂ metabolite compartment set has a reversible compartment (D) that represents labeled CO₂/HCO₃⁻ that are reversibly transported into tissue, whereas the fixed CO₂ tissue compartment (E) represents labeled CO₂ that has been incorporated into molecules trapped in the tissue. The behavior of CO₂ can be described by four rate constants: K_1c (transport from blood to tissue), k_2c (transport from tissue to blood), k_3c (incorporation into molecules in tissue), and k_4c (transport from molecules into the reversible tissue compartment) [12,13]. As in the studies of Brooks et al. [12], we assume k_3c = k_4c.

The differential equations that characterize the compartmental model are detailed in the companion article. The total image activity (μCi/g) is given by the sum of the activities in each compartment plus the activity in the blood volume contained in the tissue:

Total Tissue Activity = A + B + C + D + E

$\begin{array}{cc}\text{Total Tissue Activity}=\hfill & A+B+C+D+E\hfill \\ & +V_{\mathrm{b}}({[\text{TdR}]}_{\mathrm{b}}+{[\text{Met}]}_{\mathrm{b}}+{[{\text{CO}}_2]}_{\mathrm{b}})\hfill \end{array}$

The thymidine flux constant (retention) is given by the product of the transport rates of thymidine from blood to tissue and from tissue to DNA divided by the sum of the rates of efflux from the reversible tissue compartment, A. Two parameters that describe thymidine incorporation into DNA, K_TdR (the thymidine retention, mL/min/g) and K_1t (the thymidine blood–tissue transport rate, mL/min/g) are important in interpreting [¹¹C]TdR images of brain tumors.

$K_{TdR}=\frac{K_{1t}{k}_{3t}}{k_{2t}+k_{3t}}$

Patient Studies

Twenty patients with either primary or recurrent gliomas were evaluated in this study. Patients who had two tumors and patients who had repeat studies were treated as separate data sets, therefore the kinetic analysis was performed for 26 brain lesions. The pathology diagnoses for the low-grade tumors included eight oligodendrogliomas grade II, six mixed gliomas grade II, one astrocytoma grade II, and two low-grade gliomas not otherwise specified. The high-grade tumors were two anaplastic astrocytomas grade III and one anaplastic mixed glioma grade III. Fourteen of the tumors had been treated and 10 of the lesions were contrast enhancing on MRI. Four patients had more than one lesion. One patient had four scans over the course of tumor progression and treatment, and one patient had a brain lesion that was known to be purely necrotic as a result of radiotherapy of a non-CNS tumor that was adjacent to normal brain, but not in the brain. All patients had MRI scans prior to the PET studies, including gadolinium-enhanced T1-weighted imaging. They provided written informed consent in accordance with the Human Subjects and Radiation Safety Committees of the University of Washington.

The patients underwent sequential [¹¹C]CO₂ and [¹¹C]TdR scans after the acquisition of attenuation images. They were required to fast overnight prior to the studies. Patients also had [¹⁸F]FDG imaging, which was used for region-of-interest (ROI) definition only.

In some cases, this was acquired on a different day and coregistration methods were used to align the TdR and FDG images [14]. A thermoplastic mask and a form-fitting head-holder were used to restrict patient movement during the imaging studies. All patients underwent arterial blood sampling from the radial artery, and the radiopharmaceuticals were administered intravenously in the contralateral antecubital fossa.

2-[¹¹C]TdR was synthesized using the method of Vander Borght with modifications described previously [15,16]. Labeled cyanide was converted to cyanate and was then condensed with ammonia to produce urea, which was azetropically dried. Carrier urea (0.5 μmol) was added at this step [16]. The ring structure was closed by condensing the [¹¹C]urea with 2-methyl-3-methoxy-2-propanoic acid methyl ester in 100 μL oleum, then it was neutralized and converted to thymidine by enzymatic action. The 2-[¹¹C]TdR was purified by high-performance liquid chromatography, producing a yield of 25–40 mCi from a 40-min irradiation. The thymidine used in the patient studies had a radiochemical purity of > 97% and a typical specific activity of > 1 Ci/μmol at injection because of the added urea.

[¹¹C]CO₂ was produced by irradiating nitrogen gas containing 1% O₂ with protons, then using a cold trap at liquid argon temperature to separate the CO₂ from the other species produced, [¹¹C]CO and [¹³N]N₂. The cold trap was flushed with high purity nitrogen gas and then warmed to release CO₂, which was bubbled into 7 mL of PBS USP to produce [¹¹C]CO₂-aqueous. FDG was prepared using the method of Hamacher et al. [17]; it had a radiochemical purity of 99% and a specific activity of nominally 1.5 Ci/μmol.

The dynamic PET images were acquired in 3-D mode on the Advance PET scanner (General Electric Medical Systems, Waukesha, WI) and were reconstructed onto a 128 × 128 × 35 matrix using 4 mm transverse and 8.5 mm axial filters, resulting in a reconstructed image resolution between 5.5 and 6.5 mm in both the axial and transverse directions. The imaging sequence was identical for the CO₂ and TdR studies: 4 × 20 sec, 4 × 40 sec, 4 × 1 min, 4 × 3 min, and 8 × 5 min.

Arterial blood samples were taken at 0.25, 0.50, 0.75, 1, 1.5, 2, 2.5, 3, 4, 5, 6, 7, 10, 13, 15, 20, 25, 40, 50, and 60 min after injection for both CO₂ and thymidine. A 0.2-mL aliquot of each sample was pipetted into a tube containing 0.8 mL of 0.5 N NaOH and the tubes were immediately capped to trap CO₂. A 0.2-mL aliquot of the samples taken at 1, 2, 3, 4, 5, 6, 7, 10, 13, 15, 20, 30, 40, and 60 min was pipetted into tubes containing 0.6 mL of isopropanol and then 0.2 mL of 0.5 N HCl was added; these samples were vortexed and bubbled with argon to remove the labeled CO₂. A 0.4-mL aliquot of the samples taken at 2, 4, and 7 min was used for metabolite analysis using high-performance liquid chromatography to determine the fraction of label present as thymidine versus non-CO₂ metabolites as described previously [18]. All blood samples were counted using a calibrated well scintillation counter (Packard Instrument, Meriden, CT). The time–activity curves (TACs) from the blood sampling were analyzed to produce three blood input functions for the thymidine study (CO₂, TdR, and non-CO₂ metabolites). These manipulations were performed as we have described previously [18], except that the TACs for total blood activity and non-CO₂ blood activity were interpolated instead of fitted, to preserve the shape of the front end of the blood curve. Analysis showed that the difference in the three input functions between the original and new methods was less than 3% of the area under the curve.

The normal brain ROI was generated by drawing around the contralateral normal brain at the level of the centrum semiovale using the image analysis package Alice (HIPG Image Processing Solution, Boulder, CO). A region was drawn around the whole tumor on each slice it was present by reference to the T2 and T1 gadolinium MRI and summed FDG PET images, creating a volume of interest. The ROIs were transferred onto the dynamic CO₂ and TdR image sets and TACs were generated. The model was used to optimize the fit to the blood and tissue TACs and the estimated parameters were recorded for both ROIs for each patient. The model estimates of thymidine transport (K_1t) and retention (K_TdR) were compared with clinical and pathologic features associated with tumor proliferation and tracer transport (tumor grade, prior treatment, and contrast enhancement on MRI) using Wilcoxon rank sum statistics.

Results

Summed CO₂ and thymidine images from a patient with a high-grade glioma and the associated tissue TACs are shown in Figure 1. The initial peak is from the CO₂ injection and the second peak is due to the injection of thymidine.

Figure 1.

Typical patient summed CO₂ and TdR images from a high-grade glioma with corresponding time–activity curves. [¹¹C]CO₂ was injected at t = 0 min and [¹¹C]TdR was injected at t = 85 min.

There was an association between transport (K_1t) and contrast enhancement on MRI (Figure 2A), in the expected direction of increased thymidine transport into the brain with contrast enhancement on Gd-enhanced MRI as an indication of BBB damage (p > .001). There was an overlap between the estimated retention (K_TdR) for contrast-enhancing and non–contrast-enhancing lesions (Figure 2B), suggesting that the flux constant is not simply reflecting transport. The graph of thymidine retention against tumor grade for untreated tumors (Figure 3A) shows that in spite of low population numbers (n = 9 for untreated low grades, n = 3 for untreated high grades), there was a statistically significant trend for increased thymidine retention with increasing tumor grade (p > .02). On average, previously treated patients with high-grade tumors had lower thymidine retention than those who had not yet been treated (Figure 3B). In this group, the difference was not statistically significant (p < .2), possibly due to the small population size (n = 3 for untreated high grades, n = 4 for treated high grades). There was no significant difference between patients with low-grade tumors that had received or had not received treatment (p < .7); however, several of the patients with low-grade tumors and prior treatment had clinical evidence of tumor progression at the time of PET.

Figure 2.

Relationship between contrast enhancement on MRI and thymidine transport (A) and thymidine retention (B).

Figure 3.

Thymidine retention against tumor grade for untreated tumors (A) and tumor grade and treatment status (B).

The graph of transport (K_1t) versus retention (K_TdR) for all lesions studied allows analysis of the relationship between these two important parameters (Figure 4A, shown with a reduced scale in Figure 4B). At low retention and transport rates (K_TdR and K_1t > 0.01 mL/min/g), at the level of normal brain, K_1t and K_TdR closely correlate. Outside this region, retention and transport are not proportional and there are some interesting features. The patient shown with the circle had an extremely aggressive tumor, and the model estimated a high retention and a high transport rate. This patient had a poor outcome with rapid tumor progression. The patient shown with a square had pure necrosis resulting from neutron radiotherapy of an extracranial nasopharyngeal carcinoma. Although the model estimated a nonzero K_TdR, it is in the region of the graph where K_1t is much greater than K_TdR, suggesting a predominant effect of transport rather than retention in DNA. This patient did well clinically, with eventual resolution of MR abnormalities.

Figure 4.

(A) Relationship between K_1t (transport) and K_TdR: (retention) ● aggressive tumor; ▀ pure radionecrosis; ♦ patient scanned over course of tumor progression and treatment—(1) low-grade pre-Rx, (2) transformed pre-Rx, (3) high-grade post-Rx, (4) necrotic. (B) Relationship between K_1t and K_TdR, shown with a reduced scale.

The patient shown with the diamonds was scanned four times over the course of tumor progression and treatment; K_TdR, K_1t, and TdR SUV are shown against scan date in Figure 5. The tumor was classified as a mixed grade II glioma at the time of the first scan, and the model estimated a low level of retention. At the time of the second scan, MRI suggested that the untreated tumor had transformed into a high grade, and the thymidine retention (K_TdR) was also much higher. Partial resection confirmed a mixed grade III glioma. After treatment by surgery, radiotherapy, and chemotherapy (third scan), the model estimated a lower retention level, corresponding with residual contrast enhancement on MRI and a clinical course suggesting treated, but viable, tumor. One year after completing treatment, the patient was clinically well but had increased contrast enhancement on MRI, leading to the clinical question of active tumor versus radionecrosis. The kinetic analysis showed a decreased retention and increased transport rate, suggesting radionecrosis. The patient is doing well a year later without additional treatment or evidence of tumor progression. Figure 5 illustrates the inadequacy of using static thymidine images to monitor tumor progression and response to therapy as the third and fourth scans showed a similar level of uptake (represented by SUV) but very different levels of transport and retention.

Figure 5.

Estimated flux constant, transport rate, and SUV for a single patient scanned four times over course of tumor progression and treatment: (1) low-grade pre-Rx, (2) transformed pre-Rx, (3) high-grade post-Rx, (4) necrotic.

Discussion

The rate of cellular proliferation is an important feature of tumor behavior that is useful in monitoring tumor progression and response to therapy. PET imaging with [¹¹C]TdR provides a noninvasive method of quantifying the rate of cellular proliferation in tumors [9], but static images of uptake in brain tumors can be misleading due to the high normal background and increased transport across a damaged BBB and so kinetic analysis is required [7]. Although thymidine analogues are available that undergo less systemic catabolism and therefore produce images with a lower level of background activity due to labeled metabolites, any kinetic analysis performed for these tracers will still need to be able to separate thymidine in the tumor due to cellular proliferation (increased retention) from that due to a disrupted BBB (increased transport) when applied to brain images.

In a companion article, we presented simulations that showed that our model could accurately estimate the rate of thymidine incorporation into DNA (retention, K_TdR) with a standard error of approximately 10%, and the rate of thymidine transport from blood to tissue (K_1t) with a standard error of approximately 15% [11]. In this report, we present patient data and compare model parameter estimates to clinical and pathological features associated with tumor proliferation and tracer transport.

There was a strong correlation between thymidine transport and contrast enhancement on MRI that was not seen to the same extent with thymidine flux. Higher tumor grades had higher flux levels, and on average, patients who had undergone therapy had lower levels of thymidine flux than those yet to be treated. The model was able to distinguish between necrotic and actively proliferating tissue, and was able to predict tumor progression and response to therapy in selected patients. These results suggest that the model can separate the contributions of transport and retention. However, the relationship between K_TdR and K_1t is a straight line at low rates of transport and retention (K_TdR and K_1t < 0.01 mL/min/g), suggesting that delivery is matched with demand or that the model is transport limited at thymidine flux rates similar to that found in normal brain tissue.

Work currently in progress is comparing thymidine retention (K_TdR) to patient outcome (i.e., survival time) and investigating the extent to which a change in thymidine retention over serial scans is indicative of response to therapy. We anticipate that K_TdR will be a more useful surrogate end-point to tumor therapy than measures derived from FDG PET. Ultimately, parametric imaging using a procedure such as mixture analysis will be performed to produce images of regional flux and transport [19]. Using parametric images will fully depict the quantitative properties of the tumor and remove operator dependency in ROI definition.

Conclusion

Analysis of [¹¹C]TdR PET image data in a wide range of patients with brain lesions demonstrates the robustness of the compartmental model to fit the data and provide estimates of the thymidine retention and transport. Parameter estimates, in general, agree with the clinical and pathological grade of the tumors. Further studies are necessary to demonstrate that thymidine PET images of cellular proliferation measure response to treatment and predict outcome.