Implications of Ambient Glucose Variation on the Target-to-Background Ratio of Hepatic Tumors By ¹⁸FDG-PET Imaging

INTRODUCTION

2-deoxy-2-(¹⁸F)-fluoro-D-glucose (¹⁸FDG) positron emission tomography (PET) is now a well-established imaging modality, with greater emphasis being placed on reporting the standardized uptake value (SUV), as this quantitation is easily understood by referring physicians as opposed to the variable esoteric descriptions of visual interpretation. SUVs can also provide a simple and reproducible method for evaluating disease progression and treatment response.[1] As a result, there have been numerous investigational studies on semi-quantitative SUV analysis, particularly in the area of stratifying benign from malignant activity. These studies have typically relied upon establishing wide divergent separation of the mean SUVmax between two population groups (e.g. disease and disease-free).

We have previously described one such method using the tumor-to-liver background ratio that dramatically improved the accuracy of ¹⁸FDG PET/CT in characterizing hepatic lesions over standard visual interpretation or correlative imaging alone.[2] However, one of the unanswered questions remaining is how variations in serum glucose (Glc) at the time of radiotracer injection would impact the accuracy of this method, because it was well established right from the advent of ¹⁸FDG PET that ambient glucose levels can adversely affect tumor uptake and decrease the sensitivity.[3456] Given the already normally high background ¹⁸FDG liver activity, this potential pitfall is even more apparent when evaluating hepatic tumors.

While elevated Glc at the time of image acquisition is a known problem, there have been only a limited number of studies quantifying its effects, particularly when evaluating hepatic lesions.[7] As more studies are performed investigating semi-quantitative analyses of ¹⁸FDG PET to evaluate unknown lesions and track treatment response, it will become more vital to understand the impact of Glc on SUV measurements.[89] The purpose of this study was to investigate how variations in ambient glucose levels would affect the divergent SUV between malignant hepatic tumors and normal liver tissue, and to derive a method for SUVmax correction in the setting of elevated glucose levels.

PATIENTS AND METHODS

Patients

After receiving permission from our Human Investigations Committee, 86 patients with biopsy-proven malignant hepatic lesions identified on ¹⁸FDG PET were retrospectively analyzed. There were 38 male patients and 48 female patients. The age range was 42–95 years. Patients with many different types of primary malignancies were included, with the breakdown summarized in Figure 1. There was no limitation placed on the size of the lesion as long as it was described on an imaging study. If the lesion was identified on an imaging study other than PET, then these patients were only included if the PET scan was performed within 2 weeks to avoid any significant interval change. Patients were excluded if they were undergoing chemotherapy, had any attempted surgical resection of the lesion in question or any directed therapy such as trans-arterial chemo-/radio-embolization or radiofrequency ablation. Patients with a body mass index (BMI) of over 35 kg/m² were excluded to preserve consistent image quality.

Close

Figure 1

Chart showing the total numbers of lesions by tumor types that were analyzed in the study.

Export to PPT

RESULTS

The mean administered dose of ¹⁸FDG was 16.37 ± 1.81 mCi. The mean uptake time between injection and imaging was 91.5 ± 13.8 min. The mean patient BMI was 25.40 ± 3.56 kg/m². The mean Glc for all the patients at the time of imaging was 106 ± 26 mg/dl (range 70-212 mg/dl). The mean SUV for malignant lesions was 8.45 ± 6.55. The mean SUV for background liver was 1.99 ± 0.58.

Using the linear regression of (ln SUV)/(ln Glc), the C value was 2.21 and 0.68 for hepatic tumors and background liver, respectively. The mean glucose sensitivity (g) for all metastatic lesions was −0.636 ± 0.144 compared with −0.0536 ± 0.0583 for background liver (P = 0.00092). The closer g is to −1, the more sensitive it is to ambient Glc.[11] The rate of change of SUV over Glc (Equation 1) was about two orders higher for metastatic lesions at −0.0561 SUV/(mg/dl) of glucose than for background liver at −0.0003 SUV/(mg/dl) (P < 0.00001).

The differences in the mean SUV of malignant lesions and background liver across various simulated Glc are plotted in Figure 2. The simulation studies using Equation (5) showed that the SUV for metastatic lesions was more dramatically affected than for background liver by Glc levels beyond the normal range, showing a sharp decline as Glc reached 120 mg/dl. SUVs measured in patients with elevated Glc would have been significantly higher if they had been within the target normal Glc range at the time of injection. The linear regression analysis after logarithmic transformation of Equation (1) is graphically represented in Figure 3, with the slopes of the regression lines equaling the rate of SUV change over Glc level derived from Equation (1) for each tissue type.

Close

Figure 2

Linear plot of the mean SUV of each group against the various simulated glucose levels. The malignant lesions show a sharp acceleration in the rate of decline of SUV as Glc crosses 120 mg/dl. Background liver reveals very little change as Glc increases.

Export to PPT

Close

Figure 3

Linear regressions following logarithmic transformation of the simulated data. The slopes of graphs match the mathematically derived rates of change from Equation (1).

Export to PPT

DISCUSSION

The quality of ¹⁸FDG PET imaging can be affected by a number of patient variables at the time of acquisition. One of the most important factors is the Glc since the radiotracer is a glucose analog.[13] In fact, the Society of Nuclear Medicine and Molecular Imaging strongly recommends reporting the Glc in the technical details of interpretations as a best practice measure.[14] It has been proven that the rate of ¹⁸FDG uptake will vary depending on tissue type.[15] Additionally, there have been prior studies demonstrating ambient Glc variation does affect uptake differently in several primary tumor types, including head and neck, lung, lymphoma, and pancreatic cancers.[10161718] However, the effects were not well quantified in such a way that could be easily incorporated into clinical practice.

In practice, glucose sensitivity is one of several variable factors that influence SUV, perhaps the most controllable factor aside from uptake time. Yet, the serum glucose level (Glc) has been somewhat neglected, apart from the tentative effort taken to control it within a physiologic range deemed “acceptable”. This is despite its potential for correction using Equation (4) to work backward and solve for the SUV under ideal Glc levels. This study quantified the impact ambient Glc had on SUV measurements, and demonstrated that the effective variability (g) was far more significant (by a factor of 100) in malignant hepatic lesions compared to normal liver, despite the liver being primarily involved in glucose metabolism. This difference in g is perhaps the best way to underscore the impact of Glc on the sensitivity of SUV. The data presented would appear favorable for universal applicability due to the wide variety of tumor types studied and the demonstrated volatility of liver metastases to variable ambient Glc. It also offers a simple method of correction for SUV at different Glc values, using the dimensionless exponent g.

Previously proposed semi-quantitative analyses of suspicious hepatic lesions, that relied on comparing tumor uptake to background tissue, would be adversely affected by elevated Glc, as there would be decreased separation between tumor SUV and normal liver SUV.[1920] Conversely, the study did prove that background liver uptake is an effective “control tissue” to perform such ratio (tumor to background) -driven semi-quantitative analyses due to its relative stability across a wide range of Glc. The rates of change derived from Equation (1) indicate that background liver is quite insensitive to variation in ambient Glc, while malignant lesions will be subject to far more volatility. Figure 2 shows that the SUV of a malignant hepatic lesion will significantly decrease, with an acceleration occurring at abnormal Glc >120 mg/dl, whereas background liver activity remains virtually constant.

The possibility of glucose-corrected SUV measurements could have very important clinical applications, especially in patients undergoing follow-up FDG PET scans during and after treatment. During treatment, patients are given various medications that may impact ambient glucose levels, such as steroids, to minimize treatment side effects. Patients can also have diseases, such as diabetes or occult glucose intolerance, which become much more difficult to control during the treatment process. Additionally, endogenous insulin sensitivity can be altered by body weight fluctuations from cancer and its therapies during and after the treatment course. These complicating factors can make glucose regulation very difficult. Yet, it may be still necessary to somehow compare the patient's clinical situation at a time of high glucose to prior scans when glucose control had been more feasible.

In order for valid temporal comparison of ¹⁸FDG PET scans, one could reason that the SUV should be obtained under similar conditions.[21] Using the quantitative corrections for tumors, it may be possible to know whether a decrease in SUV of a patient presently being treated is due to treatment response or just due to a temporal differences in Glc. Additionally, differences in the normal liver uptake between scans on the same patient could alert the interpreter that changes in SUV could be reflective of biologic rather than pathologic changes. Application of this type of glucose correction could save scans repeated for this technical factor and improve accuracy when interpreting serial PET scans to monitor tumor progression or treatment response.

The glucose sensitivity of different phenotypes may be an interesting area for future investigation. Unfortunately, we did not have enough number of cases of each tumor type to fully evaluate this factor. It will also be important to determine how patient weight, radiotracer dose variation, and uptake time between injection and image acquisition would affect SUVmax of hepatic lesions. Differences in diabetic patients would seem to be a reasonable consideration in this project; however, our institution gives detailed and individualized instructions for diabetic patients to ensure they are not in a high insulin state at the time of injection. Patients with non-fasting or hyperinsulin status at the time of injection were easily identified, and excluded on the basis of not having reasonable imaging quality. Therefore, there was no reason to stratify diabetic patients as a separate sub-group.

CONCLUSION

Malignant liver lesions have significantly greater sensitivity than background liver tissue to variations in serum glucose level at the time of radiotracer injection. As the glucose level rises, the measured SUV of the malignant lesions may decrease rapidly, while the background liver tissue SUV remains relatively constant. Consideration of glucose sensitivity not only provides the framework for correcting SUV from potential ambient glucose variation to provide more accurate temporal comparison, but also justifies the use of the simple tumor to normal liver ratio for evaluating suspicious hepatic lesions by ¹⁸FDG PET imaging.