Abstract
In this study, we compared the accuracy of the rate of myocardial glucose use (rMGU) measured using PET and 1-11C-glucose with the rate measured using PET and the more conventional tracer 18F-FDG. Methods: PET measurements of myocardial tracer uptake (K, in mL/g/min) and rMGU (in nmol/g/min) were obtained with 1-11C-glucose and 18F-FDG in 21 dogs using kinetic modeling and the Patlak graphical method, respectively. Eighteen dogs were studied during hyperinsulinemic-euglycemic clamp performed either at rest or combined with phenylephrine, dobutamine, intralipid infusion, or intralipid infusion and dobutamine. Three dogs were studied during intralipid infusion alone under resting conditions. Arterial-coronary sinus sampling was performed to measure the K of both tracers (n = 14) and rMGU by the Fick method (n = 21). Results: PET-derived values for K from either 1-11C-glucose or 18F-FDG correlated closely with directly measured tracer K values (glucose: y = 0.98x + 0.01, r = 0.79, P < 0.001; 18F-FDG: y = 0.74x + 0.03, r = 0.77, P < 0.001). In contrast, correlation with K values of unlabeled glucose measured directly was better for 1-11C-glucose (y = 0.92x + 0.02, r = 0.96, P < 0.0001) than for 18F-FDG (y = 0.66x + 0.05, r = 0.72, P < 0.01) (P < 0.001 for comparison of correlation coefficients). As a consequence, PET-derived values for rMGU correlated more closely with Fick-derived measurements of unlabeled glucose using 1-11C-glucose (y = 0.82x + 168, r = 0.97, P < 0.0001) than with 18F-FDG (y = 0.81x + 278, r = 0.79, P < 0.001) (P < 0.001 for comparison of correlation coefficients). Conclusion: Over a wide range of conditions, PET-derived measurements of rMGU are obtained more accurately with 1-11C-glucose than with 18F-FDG.
Quantification of the rate of myocardial glucose use (rMGU) plays a key role in furthering our understanding of normal cardiac development and physiology and of a variety of cardiac disease processes such as myocardial ischemia, left ventricular hypertrophy, and cardiomyopathy (1–4). To date, PET using 18F-FDG has been the most common method to noninvasively quantify rMGU. Recently, our group reported that rMGU could be quantified by PET using glucose radiolabeled in the 1-carbon position with 11C (1-11C-glucose) (5).
Both tracers offer significant advantages and disadvantages for quantifying rMGU with PET. For example, the myocardial kinetics of 18F-FDG have been well characterized, and this tracer has been used to quantify rMGU with PET for nearly 20 y (6–10). The acquisition scheme is relatively straightforward. The synthesis procedures for producing 18F-FDG have been automated and produce large amounts of radiopharmaceutical. When coupled with the relatively long physical half-life of 18F (approximately 110 min), this advantage permits offsite production of 18F-FDG with delivery to multiple imaging centers. The demonstrated efficacy of 18F-FDG uptake to differentiate malignant (high uptake) from benign (low uptake) tumors has led to the rapid growth of clinical PET. That being said, quantifying rMGU with PET and 18F-FDG suffers from several disadvantages. The most prominent is that because 18F-FDG is a glucose analog, its initial uptake and phosphorylation differ from those of glucose. Consequently, a correction factor that accounts for these differences, called the lumped constant (LC), must be known to accurately quantify rMGU with 18F-FDG. However, it is well established that the LC can vary with the availability of insulin and fatty acids in plasma and with the level of blood flow (11–15). Although an approach has been devised to correct for the variability in the level of plasma fatty acids and insulin, the accuracy of the method has been validated only in isolated perfused hearts, not in vivo (16). Other disadvantages of 18F-FDG include the limited metabolic fate of 18F-FDG in tissue, precluding determination of the metabolic fate (i.e., glycogen formation vs. glycolysis) of the extracted tracer and glucose, and limitations on the performance of serial measurements of rMGU because of the relatively long physical half-life of 18F (approximately 110 min).
PET with 1-11C-glucose also offers several advantages over PET with 18F-FDG for quantifying rMGU. Because glucose has been labeled with 11C, its metabolic fate in myocardium is identical to that of glucose. Consequently, no corrections are necessary to account for differences in the myocardial kinetics of 1-11C-glucose relative to unlabeled glucose. An appropriate compartmental model has been developed to characterize the kinetics of this tracer and permits accurate measurements of rMGU (5). Another advantage of 1-11C-glucose is the short physical half-life of 11C (approximately 20 min), making serial measurements of rMGU possible. However, there are several disadvantages to using this tracer. Most notable is the limited experience in measuring rMGU with 1-11C-glucose, with only 1 study reported in the literature (5). The acquisition scheme to quantify rMGU using 1-11C-glucose is more complex than that using 18F-FDG because of the need to correct the arterial input function for the production of 11CO2 and 11C-lactate. Unlike the procedure for 18F-FDG, the current synthesis procedure for 1-11C-glucose is not as routine and produces much less radiopharmaceutical. The short physical half-life of 11C, coupled with the low production yield, requires an on-site cyclotron for PET studies using this tracer. Finally, except for measurements of the rate of cerebral glucose use (17,18), 1-11C-glucose has not yet been shown to be useful for other noncardiac applications, such as oncologic applications.
Because of the advantages and disadvantages of the 2 methods to measure rMGU, understanding the relative accuracy of PET using either 18F-FDG or 1-11C-glucose to quantify rMGU is important. Accordingly, the goal of this study was to determine the accuracy of measurements of rMGU using these 2 methods in a well-controlled canine model over a wide range of rMGU and plasma insulin and fatty acid availability.
MATERIALS AND METHODS
Animal Preparation
All animal experiments were conducted in compliance with the Guidelines for the Care and Use of Research Animals established by the Animal Studies Committee at Washington University. The animals fasted for 12–18 h before they received subcutaneous premedication with 1–2 mg/kg of morphine sulfate. Thirty to 60 min after the premedication, the animals were anesthetized with sodium thiopental (12.5 mg/kg) and were then given intravenous α-chloralose (72 mg/kg). The animals were intubated and ventilated with room air and supplemental oxygen to maintain arterial blood gases within the physiologic range. After the induction of anesthesia and intubation, catheters were placed in the thoracic aorta through the femoral arteries for monitoring arterial blood pressure and for arterial sampling. One femoral vein was cannulated to obtain venous samples and to administer drugs. The right external jugular vein was exposed for placement of a coronary sinus catheter. The coronary sinus catheter was placed under fluoroscopic guidance, and its location was verified by contrast injection and measurement of the oxygen saturation of coronary sinus blood.
To define the myocardial kinetics of 1-11C-glucose and 18F-FDG over a wide range of substrate, hormonal, and cardiac work conditions, we performed the following interventions on 21 dogs. Three dogs were studied at rest during the intravenous infusion of 20% Intralipid (Pharmacia Inc., Clayton, NC) at a rate of 1 mL/min. This intervention was designed to significantly raise plasma fatty acid levels and, thus, lower rMGU. The results from these 3 dog studies were reported previously (5). The other 18 dogs were studied during a hyperinsulinemic-euglycemic clamp (HIC) as described previously (5). Nine dogs were studied during the clamp at rest, a condition that would be expected to increase rMGU. To further increase rMGU, 3 dogs were studied during the coadministration of phenylephrine (0.84–1.6 μg/kg/min), a predominantly α-adrenergic agonist. These dogs were supplemented with intermittent boluses of intravenous atropine (0.5 mg) to maintain the heart rate at preinfusion rates. Two dogs were studied during coadministration of the clamp and dobutamine (10 μg/kg/min), a predominantly β-adrenergic agonist. It was anticipated that the peripheral lipolytic effects of β-adrenergic stimulation would markedly increase plasma fatty acids levels and decrease rMGU. To determine further how well each tracer performed when both plasma insulin and fatty acids were increased, 2 dogs were studied during coadministration of the clamp and lipid infusion. To assess the additive effects of increased cardiac work, 2 dogs were studied during the clamp, lipid infusion, and dobutamine. After completion of the PET study, the animals were euthanized with an overdose of anesthesia and KCl.
Imaging Protocol
1-11C-glucose was synthesized according to a published procedure (19,20). All PET studies were performed on an ECAT-EXACT 962 HR+ scanner (Siemens, Knoxville, TN). The electrocardiogram, hemodynamics, and blood gas values were monitored throughout the study. A transmission scan was obtained to correct for photon attenuation. To measure myocardial perfusion, 14.8 MBq/kg of 15O-water were administered as an intravenous bolus, and dynamic PET data were acquired for 5 min. After allowing for decay of 15O radioactivity, 555–740 MBq of 1-11C-glucose were administered as a bolus, followed by 60 min of dynamic data collection. After allowing for decay of the 11C-activity, the 15O-water scan was repeated. Then, 296–370 MBq of 18F-FDG were administered as a bolus, followed by 60 min of dynamic data collection. Plasma 11C activity and 18F-FDG activity were measured in paired arterial and coronary sinus blood samples that were obtained during the imaging sessions for 14 dogs. This procedure permitted direct measurements of myocardial extraction of both 1-11C-glucose and 18F-FDG. To measure rMGU and myocardial oxygen consumption (MVO2) on the basis of the myocardial extraction of glucose and oxygen, respectively, arterial and coronary sinus blood samples for glucose and oxygen were obtained at the beginning, middle, and end of the scanning intervals for all 21 dogs. To verify the stability of the substrate environment, fatty acids, lactate, and insulin were obtained at the same intervals. In addition, arterial and coronary sinus plasma samples were obtained during 1-11C-glucose imaging for all 21 dogs to measure the production of 11CO2 and 11C-lactate. Arterial levels of these labeled glucose metabolites were used to obtain a correct arterial input function for modeling of the myocardial kinetics of 1-11C-glucose.
Analysis of PET Data
Myocardial 15O-water, 1-11C-glucose, and 18F-FDG images were reoriented to generate standard short-axis and long-axis views. Composite PET images of myocardial 11C or 18F activity representing 30–60 min after injection were generated for placement of regions of interest (5). To generate myocardial time-activity curves, regions of interest encompassing the anterolateral-anteroinferior wall (5–7 cm3) were placed on 3–5 midventricular short-axis slices of the composite image. To obtain the input function, a small region of interest (1 cm3) was placed within the left atrial cavity on a midventricular slice in the vertical long-axis orientation. Within these regions of interest, myocardial and blood time-activity curves were generated for 15O-water, 1-11C-glucose, and 18F-FDG. To avoid contamination from right ventricular blood radioactivity, septal regions were omitted. Subsequently, values for myocardial blood flow (MBF, mL/g/min), myocardial glucose uptake for 1-11C-glucose and 18F-FDG (K, mL/g/min), and glucose use from either 1-11C-glucose or 18F-FDG K values (rMGU, nmol/g/min) were calculated for each region of interest and averaged to obtain 1 mean value for MBF (from 15O-water) and K and rMGU (for 1-11C-glucose and 18F-FDG) per dog. These values were then compared with global measurements of left ventricular uptake of tracer or unlabeled glucose performed using the Fick method. To compare the regional variability of PET-derived 1-11C-glucose and 18F-FDG K values, the regional coefficient of variation (COV%) for K was calculated as the ratio of the regional SD to the mean, multiplied by 100.
Measurement of MBF.
Applying the image-analysis routine to the time-segmented data allowed generation of blood and myocardial time-activity curves for each segment. From these data, MBF was quantified using a compartmental modeling approach developed and validated previously at this institution (21–23).
Measurement of rMGU Using 1-11C-Glucose.
PET blood (corrected for 11CO2 and 11C-lactate) and myocardial time-activity curve data were analyzed using a 4-compartment model developed and validated previously in our laboratory (5). In brief, vascular 1-11C-glucose enters the interstitial and cytosolic component of the myocardium (compartment 1) at a rate of K1 (mL/g/min). Once the tracer is in compartment 1, either it diffuses back into the vasculature at a rate of k2 (min−1) or it is phosphorylated and metabolized to pyruvate (compartment 2) at a rate of k3 (min−1). Label can either form glycogen (compartment 3) at a rate of k4 (min−1) or be metabolized through anaerobic or oxidative pathways (compartment 4) at rate of k5 (min−1). Tracer entering the metabolic compartment is assumed to be unidirectional, and washout of 11CO2 and 11C-lactate to the vasculature is assumed to be proportional to flow. Myocardial time-activity curves were corrected for partial-volume effects a priori using the fractional tissue recovery coefficient obtained from the 15O-water model. The model transfer rate-constants (K1, k2–k5) and the spillover fraction (22) were estimated using well-established numeric methods (24,25). Steady-state conditions were assumed, and K and rMGU were then calculated from these rates as follows:
Eq. 1
Eq. 2 where Glb is plasma glucose level, in nmol/mL.
Measurement of rMGU Using 18F-FDG.
Measurements of rMGU using 18F-FDG kinetic data were done by 2 different methods, the Patlak method and compartmental modeling. Because it is the most commonly used method to quantify rMGU by PET and 18F-FDG, the Patlak method was chosen to compare rMGU measurements obtained from 1-11C-glucose with those derived from 18F-FDG. However, because 18F-FDG kinetics are described more accurately by compartmental modeling than by the Patlak method when 18F-FDG egress or dephosphorylation is present, compartmental modeling was also implemented to compare rMGU measurements obtained from the Patlak method with those obtained from 18F-FDG kinetic modeling.
Myocardial glucose uptake was estimated from PET-derived blood and myocardial 18F-FDG time-activity curves using Patlak graphic analysis (8). This method, which can be applied to kinetic models in which the tracer is assumed to be irreversibly trapped, can be used to calculate net tracer uptake, with K as the slope of the linear equation:
Eq. 3 where Cm(t) (cpm/g/mL) is myocardial activity at time t, Cb(t) (cpm/g/mL) is blood activity at time t, K is net tracer uptake (mL/g/min), ∫0tCb(t) ds (cpm × min) is the integral of plasma activity from time zero to time t, and W is a parameter depending on the steady-state volume of the reversible compartments and the effective plasma volume.
Myocardial time-activity curves were corrected for partial-volume effects a priori using the fractional tissue recovery coefficient obtained from the 15O-water model. To estimate K from Equation 3, PET blood and myocardial 18F-FDG time-activity curves were used to generate 6 sets of x–y data points representing data at 2, 5, 9, 18, 36, and 56 min after injection of tracer. To show the linearity of the dataset, data points were plotted on an x–y plot and points that visually diverged from a straight line were excluded from the analysis. On average, the last 2 sets of data points, representing data collected at 36 and 56 min after 18F-FDG injection, were excluded from the analysis. No early points were excluded. To account for the differences in uptake and phosphorylation of 18F-FDG and glucose, K values obtained from Equation 2 were corrected by an LC value of 0.67 (6). Furthermore, because more recent studies have shown that the value of the LC varies with varying substrate, hormonal, and work environment (11–15), K values were also corrected by a variable LC, as defined by Wiggers et al. (26), as follows:
Eq. 4 where Rp is a known ratio between the rates of phosphorylation of 18F-FDG and glucose (0.43), Rt is a known ratio between the rates of membrane transport of 18F-FDG and glucose (2.26), K (mL/g/min) is net 18F-FDG uptake (estimated from Patlak analysis), and K1 is unidirectional 18F-FDG transport (mL/g/min) (estimated from kinetic modeling of 18F-FDG kinetics). Using Equation 2, rMGU was calculated from uncorrected and corrected K values.
A previously validated 3-compartment kinetic model, originally developed for calculation of the rate of cerebral glucose use and subsequently modified for the heart, was used (7). The individual rate constants describing the forward (K1, mL/g/min) and reverse (k2, min−1) transport of 18F-FDG across the sarcolemma, the hexokinase-mediated phosphorylation of 18F-FDG (k3, min−1) and dephosphorylation of 18F-FDG (k4, min−1), and the spillover fraction (as a fifth parameter) were estimated by fitting the model to the myocardial tissue time-activity curves using a nonlinear least-squares algorithm. All of the myocardial time-activity curves were precorrected for the partial-volume effect using correction factors calculated for the measurement of MBF with 15O-water. Then, for each region of interest, 18F-FDG tracer uptake (K, mL/g/min) was calculated from the estimated rate constants twice, assuming that 18F-FDG egress or dephosphorylation was negligible (i.e., k4 = 0) and that 18F-FDG egress or dephosphorylation was not negligible (i.e., k4 ≠ 0).
Measurement of Radioactive Plasma Metabolites and Myocardial Tracer Uptake
Total 11C-activity and 18F-FDG activity in arterial and coronary sinus plasma were measured in a well counter (model 1000; Canberra, Meridien, CT). Total 1-11C-glucose activity was calculated from 11C-activity as follows: Total 11C-acidic metabolites (11C-lactate and 11CO2) were first separated from neutral sugars by trapping them in an ion exchange column (AG1-X4 resin, 100–200 mesh, formate form). The plasma sample was then eluted through this column with 5 mL of water. Both resin and eluate were counted for radioactivity to obtain the total acidic metabolites and nonmetabolized glucose present (1-11C-glucose). To measure 11C-lactate and 11CO2 from total acidic metabolites, a similar serum sample was also deposited in a test tube, acidified with 6N HCl, sparred with N2 for 10 min to eliminate the 11CO2, and counted for radioactivity against a sample kept under basic conditions. The count difference between the test tubes was used to calculate the percentage of 11CO2 in the plasma sample. To obtain the percentage of 11C-lactate, the percentage of 11CO2 was subtracted from the total percentage of acidic metabolites in the resin. The myocardial extraction fractions for 1-11C-glucose and 18F-FDG were calculated as the difference between the arterial and coronary sinus plasma values divided by the arterial plasma values for 1-11C-glucose and 18F-FDG activity, respectively. Myocardial uptake of the tracers (K, mL/g/min) was then calculated as the product of the extraction fraction (unitless) and MBF (mL/g/min).
Measurement of Myocardial Uptake and Use of Unlabeled Glucose
The extraction fraction for unlabeled glucose was calculated from each of the 3 sets of blood samples taken during the glucose study as the ratio of the difference between plasma levels of arterial and coronary sinus glucose divided by the arterial glucose level. Glucose uptake (K, mL/g/min) was then calculated as the product of glucose extraction fraction (unitless) and MBF (mL/g/min), and glucose use (nmol/g/min) was calculated as the product of glucose uptake and the arterial plasma glucose level (nmol/mL). A mean value per study was obtained by averaging values obtained from the 3 sets of blood samples.
Measurement of Plasma Insulin and Unlabeled Substrates
Plasma insulin levels were measured by radioimmunoassay (27). Glucose and lactate were assayed enzymatically using a commercially available glucose-lactate analyzer (YSI, Yellow Springs, OH). Fatty acids in plasma were determined by capillary gas chromatography and high-performance liquid chromatography (5).
Statistical Analysis
All data are presented as the mean ± SD. Correlations were calculated by linear regression. Differences in slopes were compared using a 2-tailed t test. Correlation coefficients were compared using a Fisher z transformation and a 2-tailed Z test. Because of the small sample size of most groups studied, no statistical analyses were performed on the group means presented in Tables 1–4.
Hemodynamics, MBF, and MVO2 During 1-11C-Glucose Study
RESULTS
Hemodynamics, Myocardial Perfusion, and MVO2
Measurements of hemodynamics, MBF, and MVO2 for the 6 groups of dogs during both the 1-11C-glucose and the 18F-FDG scans are shown in Tables 1 and 2. As expected, an increase in the rate-pressure product (RPP) was observed in the dogs receiving phenylephrine with atropine supplementation, reflecting elevations in both systolic blood pressure and heart rate. Primarily because of an increase in heart rate, the RPP was increased in dogs undergoing both the HIC and dobutamine administration. The heart rate and systolic blood pressure were increased in the groups receiving Intralipid either alone at rest or during the coadministration of dobutamine, consistent with the increase in α1-receptor sensitivity associated with infusion of lipid emulsions (28). This resulted in an increased RPP. In dogs receiving intralipid and the insulin clamp, the RPP did not increase, primarily because the heart rate did not increase. The etiology of the lack of increase in heart rate is unclear. In general, the increased levels in RPP were paralleled by increases in both MBF and MVO2. Values for MBF and MVO2 tended to be higher in the dogs receiving dobutamine than in the dogs receiving phenylephrine, a finding that may well reflect the combined effects of vasodilation due to β2-adrenergic stimulation and the increased oxygen demands of fatty acid metabolism. Also of note, there was a trend toward a lower level of MBF despite an increase in the RPP in the dogs receiving intralipids under resting conditions. This blunting of MBF may reflect the inhibitory effects of high plasma fatty acids on endothelial function (29). Finally, it should be noted that similar hemodynamic, MBF, and MVO2 responses to each intervention were observed during the 1-11C-glucose and 18F-FDG imaging studies.
Hemodynamics, MBF, and MVO2 During 18F-FDG Study
Plasma Substrate Levels and Myocardial Substrate Use
Shown in Tables 3 and 4 are the arterial plasma levels for glucose, free fatty acids, lactate, and insulin levels, as well as rMGU values measured during the 1-11C-glucose and 18F-FDG scans for all interventions. Plasma glucose and lactate levels were similar between the groups. Plasma fatty acid levels were higher in all groups receiving a lipid infusion. The fatty acid levels were the lowest in the dogs undergoing the insulin clamp at rest, consistent with the lipid-storing effects of this hormone. As designed, plasma insulin levels were increased in all groups undergoing the insulin clamp. The insulin levels were lowest in dogs studied at rest during the infusion of lipids. These changes in substrate and insulin plasma levels, as well as the changes in myocardial work induced by the different physiologic interventions performed, resulted in a wide range of rMGU levels. Dogs studied during intralipid infusion at rest had the lowest rMGU levels. Increases in rMGU were observed during HIC and were offset by the concomitant infusion of intralipid. As expected, the highest levels of rMGU were observed in dogs undergoing the insulin clamp and receiving phenylephrine. Levels of plasma substrates, insulin, and rMGU during the different interventions were similar during the 1-11C-glucose and 18F-FDG imaging studies.
Plasma Substrate Levels, Insulin Levels, and Myocardial Glucose Use During 1-11C-Glucose PET Study
Plasma Substrate Levels, Insulin Levels, and Myocardial Glucose Use During 18F-FDG PET Study
PET Data Analysis
Representative myocardial images of 1-11C-glucose and 18F-FDG are shown in Figure 1. As expected, differential myocardial uptake of both tracers increased with increasing levels of rMGU (mean rMGU values are shown in Tables 3 and 4). Because of the low myocardial uptake of tracer that occurred with intralipid infusion, myocardium could not be differentiated from blood at rest with intralipids (IL/R). Because myocardial tracer uptake increases with HIC, myocardium could readily be differentiated from blood during the clamp at rest (HIC/R) and even more so during phenylephrine (HIC/PHEN).
Myocardial 18F-FDG and 1-11C-glucose short-axis images obtained 30–60 min after injection of tracer in dogs studied during either IL/R, HIC/R, or HIC/PHEN. Black is lowest activity and white is highest activity. A = anterior; I = inferior; L = lateral; LV = left ventricle; RV = right ventricle; S = septum.
The ability of PET to measure myocardial uptake of the 2 tracers is shown in Figure 2. PET-derived measurements of myocardial K for either 1-11C-glucose (Fig. 2A, y-axis) or 18F-FDG (Fig. 2B, y-axis) correlated closely with directly measured values of tracer uptake (Fig. 2A, x-axis: 1-11C-glucose K; Fig. 2B, x-axis: 18F-FDG K). There were no significant differences between the regression coefficients of 1-11C-glucose and 18F-FDG (0.79 vs. 0.77, P = not statistically significant [NS]), and neither slope, 0.98 or 0.74, was significantly different from unity.
(A) Correlation between PET-derived measurements of myocardial uptake of 1-11C-glucose and direct measurements of myocardial uptake of 1-11C-glucose. (B) Correlation between PET-derived measurements of myocardial uptake of 18F-FDG and direct measurements of myocardial uptake of 18F-FDG. D = dobutamine.
Shown in Figure 3 are the correlations of K values for 1-11C-glucose and 18F-FDG with unlabeled glucose, all measured by arterial and coronary sinus sampling. The K values for 1-11C-glucose agreed closely with the K values for unlabeled glucose (Fig. 3A). The K values for 18F-FDG, either uncorrected (Fig. 3B) or corrected for differences in transport and phosphorylation between 18F-FDG and glucose using an LC value of 0.67 (Fig. 3C), also correlated with the K values for unlabeled glucose; however, the correlation was significantly less than that observed with 1-11C-glucose (P < 0.001 for the comparison of correlation coefficients of 0.96 and 0.72). Moreover, K values for 18F-FDG corrected using a variable LC calculated from Equation 3 (LCv) (Fig. 3D) did not significantly improve the correlation with K values for unlabeled glucose (r = 0.61, P < 0.05). Of note, neither 1-11C-glucose nor 18F-FDG (corrected by LC) underestimated the myocardial uptake of glucose (P = NS for slope values for both tracers, compared with unity). However, using the LCv underestimated the myocardial uptake of unlabeled glucose (P < 0.001 for a slope of 0.24, compared with unity [Fig. 4D]).
(A) Correlation between direct measurements of myocardial uptake of 1-11C-glucose and unlabeled glucose. (B–D) Correlation between direct measurements of myocardial uptake of 18F-FDG and unlabeled glucose for uncorrected 18F-FDG K values (B), values corrected by LC (C), and values corrected by LCv (D). D = dobutamine.
Correlation between PET-derived (y-axis) and Fick-derived (x-axis) measurements of rMGU from 1-11C-glucose (A), from 18F-FDG before correcting PET values (B), from 18F-FDG after correcting PET values by LC (C), and from 18F-FDG after correcting PET values by LCv (D). D = dobutamine.
Figure 4 shows the correlation between PET-derived values for rMGU using 1-11C-glucose (Fig. 4A) or 18F-FDG (Figs. 4B–4D) with values for rMGU measured directly from plasma glucose levels using the Fick method over a wide range of substrate and insulin availability and levels of cardiac work. PET-derived values for rMGU obtained with 1-11C-glucose correlated closely with directly measured values of rMGU (Fig. 4A). PET-derived values for rMGU obtained with 18F-FDG before and after correction of data by LC values also correlated with direct measurements of rMGU (Figs. 4B–4D). Of note, using the LCv correction did not improve the correlation (Fig. 4D). Moreover, the correlation was significantly closer when 1-11C-glucose, as opposed to 18F-FDG, was used, regardless of whether an LC was used or the type of LC used (Figs. 4B–4D) (P < 0.001 for comparison of correlation coefficients). Measurements of rMGU with 1-11C-glucose were underestimated by ∼18% (Fig. 4A, P < 0.001 compared with unity). Similarly, PET with 18F-FDG also underestimated rMGU by ∼19%–75% depending on the whether an LC correction was used and the type of LC used (Figs. 4B–4D, P < 0.001 compared with unity).
Table 5 summarizes the correlation between PET-derived uptake values for 1-11C-glucose and 18F-FDG (dependent variables) and directly measured glucose uptake values (independent variable). Correlation coefficients for 18F-FDG uptake values derived from either the Patlak method or kinetic modeling assuming no 18F-FDG egress or assuming 18F-FDG egress were all significantly worse than those obtained for 1-11C-glucose K values (r = 0.78, 0.82, and 0.81, respectively, for 18F-FDG PET vs. r = 0.97 for glucose PET, P < 0.001). However, there were no significant differences among the correlation coefficients for the different 18F-FDG uptake values (r = 0.78, 0.82, and 0.81, respectively; P = NS).
Linear Regression Analysis of PET-Derived K Values and Directly Measured Glucose K Values
To assess the regional variability of the PET measurements using either tracer, mean K and rMGU values were obtained from 3–5 regions of interest per study. Regional variation of PET-derived myocardial K values for 1-11C-glucose and 18F-FDG were similar, with COV% values of 8.0% ± 5.7% and 9.4% ± 8.11%, respectively (P = NS).
DISCUSSION
This study provided new findings on the quantification of rMGU by PET. Over a wide range of plasma substrate and insulin availability and of levels of cardiac work, myocardial uptake of 1-11C-glucose more closely paralleled uptake of unlabeled glucose than did uptake of 18F-FDG. As a consequence, PET measurements of rMGU were more accurate when the tracer was 1-11C-glucose, as opposed to 18F-FDG.
As expected, noninvasive PET measurements of myocardial uptake for both 1-11C-glucose and 18F-FDG correlated well with direct measurements of tracer uptake (Figs. 2A and 2B). Moreover, the observation that the myocardial uptake kinetics of 1-11C-gluose more closely paralleled glucose uptake when compared with the myocardial kinetics of 18F-FDG was not surprising. d-glucose labeled in the 1-carbon position with 11C should have the same transport and biochemical characteristics as unlabeled d-glucose. Because uptake of d-glucose far surpasses uptake of l-glucose in muscle, uptake of 1-11C-glucose should be nearly identical to that of unlabeled glucose (30). Indeed, the close and direct correlation between direct measurements of 1-11C-glucose and unlabeled glucose uptake supports this assumption (Fig. 3A). In contrast, 18F-FDG is an analog of glucose whose kinetics parallel only the transmembranous transport and hexokinase-mediated phosphorylation of glucose. To account for the differences between 18F-FDG and glucose, the LC was devised so that PET measurements of rMGU could be performed (6–10). However, values for the LC have been shown to depend markedly on the availability of various plasma substrates such as lactate and fatty acids, the level of insulin, and the level of blood flow (11–15). Variability in the LC likely contributed to the poorer correlation of direct measurements of myocardial 18F-FDG uptake with those of unlabeled glucose.
A solution has been proposed to calculate the LC on the basis of the application of assumptions underlying Michaelis-Menten kinetics to the conventional 3-compartment model describing myocardial 18F-FDG kinetics (16,26). Although initial results obtained in isolated perfused hearts are promising, the approach assumes that the ratios of 18F-FDG to glucose for transport and initial phosphorylation are constant—an assumption that is unproven. Moreover, the approach does not account for LC variability due to changes in MBF (15). In the present study, variable LC values calculated from Equation 3 ranged from 0.49 ± 0.04 in the intralipid intervention at rest, which had the highest levels of plasma fatty acids, to 1.75 ± 0.36 in the clamp/phenylephrine group. Furthermore, results from multiple linear regression analysis showed a significant correlation between LCv values with MBF and plasma glucose, free fatty acid, lactate, and insulin levels (r = 0.84, P < 0.001), with free fatty acids and insulin being significant negative predictors of LCv values (P < 0.0001 and P < 0.05, respectively). Thus, in the present study, the pattern of LCv changes with changing plasma levels of insulin and fatty acids is consistent with what is known about the impact of these variables on the relationship between 18F-FDG and glucose transport and phosphorylation (11–14). The poor correlation between myocardial uptake of 18F-FDG and unlabeled glucose when LCv was used (Fig. 3D) suggests that this approach may be of limited value in vivo, a setting in which the integrative effects of plasma insulin and fatty acid levels and MBF are operational.
Although the myocardial kinetics of 1-11C-glucose more faithfully trace those of glucose than do the myocardial kinetics of 18F-FDG, this advantage would be minimized if the myocardial kinetics of 1-11C-glucose could not be quantified accurately with PET. Clearly, compartmental modeling is more demanding with 1-11C-glucose than it is with 18F-FDG. The arterial input function must be corrected for the production of 11CO2 as well as 11C-lactate. Moreover, the complexity is also exacerbated by the more diverse metabolic fates of 1-11C-glucose, compared with the metabolic fates of 18F-FDG. Finally, because of significant egress of tracer from the myocardium, more simplified approaches to quantify tracer uptake, such as Patlak graphical analysis, as is done with 18F-FDG, cannot be used with 1-11C-glucose. Despite the more complex kinetics of 1-11C-glucose compared with those of 18F-FDG, PET quantification of myocardial glucose uptake was more accurate with 1-11C-glucose than with 18F-FDG (Fig. 3). As a consequence, PET measurements of rMGU were more accurate when 1-11C-glucose, as opposed to 18F-FDG, was used as the tracer, regardless of whether an LC correction was implemented or the type of correction used (Fig. 4).
There was slight, ∼18%, underestimation of rMGU by PET with 1-11C-glucose (Fig. 4A). Given that direct measurements of 1-11C-glucose uptake did not underestimate unlabeled glucose uptake (Fig. 3A) and that PET did not underestimate 1-11C-glucose uptake (Fig. 2A), it is unlikely that the underestimation of rMGU either was due to tracer effects or was the result of compartmental modeling. Indeed, the underestimation of rMGU appears to be mainly due to underestimation of the highest glucose use value, 3,000 nmol/g/min. Removal of this value resulted in a similar regression value of 0.95 (vs. 0.97), with a slope of 0.92 (P = NS from a slope value of 1). Values of rMGU obtained from uncorrected Patlak K values were significantly underestimated (Fig. 4B). Correction of the K values by a constant LC value of 0.67 resulted in a regression slope not different from 1 (Fig. 4C). In contrast, correction of K values by LCv worsened the underestimation of rMGU (Fig. 4D). As already mentioned, this result was most probably due to the limited value of the LCv approach in a complex environment in which substrate and insulin plasma levels, as well as myocardial perfusion, play key roles in the competitive myocardial kinetics of 18F-FDG and glucose. In fact, further investigation of the LCv values obtained in our study showed that those obtained from the interventions that resulted in high glucose uptake values (greater than 0.15 mL/g/min) were the key contributors to the underestimation of rMGU. Exclusion of these studies from the regression analysis (all 3 clamp/phenylephrine studies and 3 clamp/rest studies) resulted in improved correlation and slope between rMGU measured with 18F-FDG and direct measurements of rMGU (y = 0.94x + 71, r = 0.82, P < 0.0001, n = 15). Thus, correction of 18F-FDG kinetics using tracer kinetic information as described by Botker et al. (16) appears to break down for high values of glucose uptake.
The results of previous studies on animals and humans have shown a closer correlation between 18F-FDG and glucose, regardless of whether myocardial uptake or rMGU was measured, than we observed in the current study (Figs. 3 and 4). This worse correlation in our study was not due to the method of analysis chosen. The correlation between myocardial 18F-FDG uptake and unlabeled glucose uptake was similar regardless of whether compartmental modeling or the Patlak method was used to analyze myocardial 18F-FDG kinetics (Table 5). However, in the earlier studies, the level of MBF was not varied and fatty acid levels were either raised (by fasting) or reduced (by adding insulin). In contrast, in the current study, we measured myocardial 18F-FDG kinetics while varying all 3 of these parameters simultaneously.
We studied the accuracy of rMGU using these 2 tracers over a wide range of substrate availability, insulin levels, and cardiac work. As mentioned previously, these interventions were chosen because studies have shown that these factors may alter the LC significantly (11–15). Furthermore, these interventions have physiologic and clinical relevance. Marked elevations in levels of plasma fatty acids either alone or together with hyperinsulinemia are observed during intense exercise or in non-insulin-dependent diabetes mellitus, respectively (31,32). Moreover, under conditions of stress, one could also see an increase in MBF, further increases in plasma catecholamines, and then increases in plasma fatty acids due to the lipolytic effects of these hormones (33). Although the superiority of 1-11C-glucose over 18F-FDG for quantifying rMGU for PET has been shown under these conditions, it is unknown whether similar advantages will be observed under pathologic conditions such as left ventricular hypertrophy, diabetic heart disease, and myocardial ischemia and reperfusion.
CONCLUSION
This study has demonstrated that PET measurements of rMGU are more accurate when 1-11C-glucose is used instead of 18F-FDG. The improved accuracy most likely reflects the more faithful tracing of myocardial glucose uptake with 1-11C-glucose than with 18F-FDG. However, the results of the current study do not negate the wealth of information about the usefulness of PET and 18F-FDG to measure either relative or absolute rMGU in heart disease. For example, the key role of relative estimates of rMGU measured by PET and 18F-FDG in differentiating viable from nonviable myocardium are well established. Moreover, quantitative measurements of rMGU by PET and 18F-FDG have furthered our understanding of the role of alterations in myocardial metabolism in a variety of disease processes, such as myocardial ischemia and reperfusion, dilated cardiomyopathy, and diabetic heart disease. However, the results of the current study do suggest that in situations in which absolute rates of rMGU are required, PET imaging with 1-11C-glucose may provide an advantage over PET imaging with 18F-FDG.
Acknowledgments
We thank Mary Russo for her administrative assistance. This study was supported by grant HL13581 from the National Institutes of Health.
Footnotes
Received Oct. 22, 2001; revision accepted Mar. 25, 2002.
For correspondence or reprints contact: Pilar Herrero, MS, Cardiovascular Imaging Laboratory, Mallinckrodt Institute of Radiology, 510 S. Kingshighway Blvd., St. Louis, MO 63110.
E-mail: HerreroP{at}mir.wustl.edu