Effects of Different Liver Diseases on Metabolic Reference in 18F-Fluorodeoxyglucose Positron Emission Tomography/Computed Tomography

Hong Kong J Radiol 2023 Dec;26(4):240-7 | Epub 11 Dec 2023

 

 

Nuclear Medicine Unit, Queen Elizabeth Hospital, Hong Kong SAR, China

 

  Full Article in PDF

Both semiquantitative assessment and qualitative visual interpretation are applied in ¹⁸F-fluorodeoxyglucose positron emission tomography/computed tomography (¹⁸F-FDG PET/CT) for lesion characterisation.[1] In the semiquantitative approach, the maximum standardised uptake value (SUV_max) is calculated, but this depends on multiple factors, including injection time, uptake period, and blood glucose level.[2] Thus, it is difficult to compare the absolute SUV_max between different PET/CT systems.[2] For qualitative visual interpretation, the ¹⁸F-FDG uptake of a lesion is typically graded with respect to mean blood pool and liver uptake of a patient (e.g., score 1: no uptake; 2: less than or equal to blood pool; 3: between blood pool and liver; 4: moderately more than liver; and 5: markedly more than liver[3]). This approach is useful in lesion delineation: a lesion is generally regarded as genuine (i.e., the lesion is true instead of false positive) if its uptake is higher than that of liver and not genuine if its uptake is less than or equal to that of blood pool. This is also useful in treatment response assessment (e.g., a disease is likely deteriorating if the score increases in interval scan). Visual interpretation is the recommended method in different guidelines, including the Deauville criteria for high-grade lymphoma,[3] PERCIST (Positron Emission Tomography Response Criteria in Solid Tumors) 1.0 for solid tumours proposed by the Society of Nuclear Medicine and Molecular Imaging,[4] [5] as well as for vasculitis assessment developed by the European Association of Nuclear Medicine and the Society of Nuclear Medicine and Molecular Imaging.[6] Ideally, blood pool and liver uptake should have minimal variability such that they can be utilised as reliable metabolic references. There are existing procedural protocols standardising patient preparation and acquisition techniques.[7] The aim of this study was to evaluate the liver and blood pool uptake in different liver diseases (hyperbilirubinaemia, cirrhosis, steatosis, and polycystic liver disease) and their potential effect on lesion assessment. Focal liver diseases (e.g., hepatocellular carcinoma and liver metastasis) were not included here as their effects have already been covered in the literature,[8] and we believe that general hepatic metabolism is likely more dependent on systemic liver diseases than focal liver pathologies. Throughout this research, the ratio between the mean SUV of the liver and that of the blood pool (SUV_liver/SUV_{blood pool} ratio) instead of absolute SUV was evaluated because the ratio was more relevant to the grading. Mean SUV (SUV_mean), instead of SUV_max, of liver and blood pool was investigated as an analogue of visual interpretation.

 

Cases of patients who underwent whole-body ¹⁸F-FDG PET/CT in our centre from 1 January 2011 to 31 December 2015 were retrospectively reviewed. The clinical background and investigation results were reviewed, including drinking history, blood test results, radiological images, and endoscopic findings. Continuous data were reported as mean ± standard deviation. Subjects were excluded if: (1) liver malignancy had been diagnosed histologically; or (2) liver malignancy was suspected radiologically (e.g., by ¹⁸F-FDG PET/CT, ultrasound, CT or magnetic resonance imaging) within 12 weeks of ¹⁸F-FDG PET/CT imaging; or (3) no liver function tests were available within 2 weeks of ¹⁸F-FDG PET/CT imaging; or (4) blood glucose level was > 11 mmol/L before ¹⁸F-FDG PET/CT acquisition.

 

A total of 168 adult patients (56.5% male, 43.5% female) with a mean age of 62.3 ± 13.4 years were included. The majority had undergone ¹⁸F-FDG PET/CT for oncological indications: lung (23.2%), lymphoma (14.8%), breast (12.5%), biliary (9.5%), colon (8.3%), renal (6.0%), and other (12.6%) cancers. Some of the subjects (13.1%) showed no evidence of malignancy after thorough workup. Each case was then assigned to one of the following five groups (Table 1):

 

Table 1. Characteristics of the five groups (n = 168)

 

(1) The control group (n = 50): liver function (i.e., serum bilirubin, alanine aminotransferase, aspartate aminotransferase and alkaline phosphatase levels) was normal, and there was no evidence of cirrhosis, steatosis or polycystic liver disease;
(2) the hyperbilirubinaemia group (n = 29): the serum bilirubin level was greater than or equal to the upper limit of the normal level (21 μmol/L) within 2 weeks of ¹⁸F-FDG PET/CT acquisition, and there was no evidence of cirrhosis, steatosis or polycystic liver disease;
(3) the cirrhosis group (n = 27): features of cirrhosis had been documented by means of imaging (e.g., ultrasound, CT or magnetic resonance imaging) or oesophagogastroduodenoscopy. Liver function was normal, and there was no evidence of steatosis or polycystic liver disease;
(4) the steatosis group (n = 52): the mean liver density in Hounsfield units (HU_liver) on CT was lower than that of the spleen (HU_spleen). As SUV measurement is potentially dependent on the distribution of steatosis (e.g., diffuse, focal, multinodular, etc.), this study focused on the patients with diffuse steatosis. Liver function was normal, and there was no evidence of cirrhosis or polycystic liver disease; and
(5) the polycystic liver disease group (n = 10): the liver contained > 20 cysts as defined in the literature.[9] Liver function was normal, and there was no evidence of cirrhosis or steatosis.

 

All ¹⁸F-FDG FDG PET-CT examinations were performed with the same PET/CT scanner (Discovery 710; General Electric, Milwaukee [WI], United States). The mean ¹⁸F-FDG activity administered was 407.0 ± 45.5 MBq. After a mean uptake time of 59.8 ± 6.21 minutes, PET data were acquired from skull vertex to mid thighs in seven to eight bed positions (3 minutes per bed position) with mean axial bed coverage of 15.2 cm per bed and 9-slice bed overlap in twodimensional acquisition mode. Reconstruction using Optimization of Ordered Subset Expectation Maximization was performed with 4.2-mm section thickness in a 128 × 128 matrix and processed through a standard filter. Non-contrast CT data were acquired for anatomical correlation and attenuation correction.

 

The SUV is defined as the activity measured in a volume of interest (VOI) divided by the injected ¹⁸F-FDG dose, based on body weight[10]:

 

 

SUV_liver was measured in a 3-cm–diameter spherical VOI over the right lobe of the liver as recommended in the PERCIST 1.0 criteria.[4] [5] No observable lesion was included in the liver VOI, except for the unavoidable multiple cysts in polycystic liver disease. SUV_{blood pool} was measured in another spherical VOI with diameter > 2 cm in the descending thoracic aorta. Atherosclerotic plaque was avoided in the blood pool VOI as the diseased vessel wall was often ¹⁸F-FDG–avid.[4] [5] The mean HU of the liver and the spleen were recorded in two-dimensional circular regions of interest with diameters > 3 cm. The body weight was routinely recorded on the same day of the ¹⁸F-FDG PET/CT acquisition, with mean weight of 63.4 ± 11.7 kg. Statistical analyses, including two-sided t tests and linear regression, were performed with SPSS (Windows version 20.0; IBM Corp, Armonk [NY], United States). The results were regarded as statistically significant if the corresponding p values were < 0.05.

 

Figure 1a shows a representative maximum intensity projection (MIP) of a control case. The ¹⁸F-FDG uptake in the liver (orange arrow) was homogeneous without discernible hypermetabolic lesions. The degree of uptake was normal and greater than that in the mediastinal blood pool (purple arrow), i.e., SUV_liver/SUV_{blood pool} ratio > 1, and in the spleen. All of the 50 control cases had SUV_liver/SUV_{blood pool} ratios > 1, with a mean SUV ratio of 1.39 (Table 1).

 

Figure 1. Maximum intensity projections of representative subjects of the five groups. (a) A control case showing liver uptake (orange arrow) is normally greater than that in the mediastinal blood pool (purple arrow) [standardised uptake value (SUV) ratio = 1.41]; (b) a hyperbilirubinaemia case with gallbladder tumour causing biliary obstruction (red arrow) [SUV ratio = 1.86]; (c) a cirrhosis case with right lung tumour (blue arrow) [SUV ratio = 1.24]; (d) a steatosis case with prior right nephrectomy for renal cell carcinoma (yellow arrow) [SUV ratio = 1.02]; and (e) a polycystic liver disease case with cold defects in liver and kidneys (green arrows) corresponding to cysts [SUV ratio = 0.79].

 

In the hyperbilirubinaemia group, the serum bilirubin level ranged from 23 to 667 μmol/L (mean = 107). Four cases had elevated aspartate aminotransferase level (> 47 IU/L), four had elevated alkaline phosphatase level (> 140 IU/L), and 18 had both enzymes elevated. Thus, 26 subjects (89.7%) had elevated liver enzyme(s) in addition to the increased serum bilirubin level. The hyperbilirubinaemia cases had a mean SUV ratio of 1.49, which was greater than that of the controls. A two-sided t test showed that the difference in mean SUV ratios achieved statistical significance (p = 0.0053; Table 1). A representative MIP of the hyperbilirubinaemia cases demonstrates the higher degree of contrast between hepatic and blood pool uptake compared with that of the controls (Figure 1b). The cause of hyperbilirubinaemia in Figure 1b was biliary obstruction secondary to a gallbladder tumour (red arrow). To investigate the correlation between bilirubin level and the SUV ratio, linear regression analysis was performed. Figure 2 shows that the SUV ratio was higher with increasing serum bilirubin level (black circles, bottom x-axis), with a corresponding slope of 0.0008 L/μmol and a correlation coefficient of 0.438. The relationship between serum bilirubin level and the SUV ratio was further examined by subdividing the hyperbilirubinaemia cases into two groups: mild (serum bilirubin level: 21-63 μmol/L, i.e., grade 1 to 2 hyperbilirubinaemia as defined by the Common Terminology Criteria for Adverse Events version 5.0[11]) and severe (serum bilirubin level > 63 μmol/L, i.e., grade 3 to 4 hyperbilirubinaemia). Table 2 shows that the mild hyperbilirubinaemia cases had a mean SUV ratio of 1.44, while the severe hyperbilirubinaemia cases had a mean SUV ratio of 1.59 (p = 0.0359).

 

Figure 2. Linear regression plots of liver-to–blood pool standardised uptake value (SUV) ratio against the serum bilirubin level for the hyperbilirubinaemia cases (black circles, bottom x-axis) and the difference between the density of the spleen and the liver in Hounsfield units (HU_spleen–HU_liver) for the steatosis cases (grey squares, top x-axis). The hyperbilirubinaemia cases had a corresponding slope of 0.0008 L/μmoL and a correlation coefficient of 0.438. The steatosis cases had a corresponding slope of -0.0062 and a correlation coefficient of 0.209.

 

Table 2. Subgroup analyses of hyperbilirubinaemia and steatosis groups

 

In the cirrhosis cases, 16 out of the 27 (59.3%) subjects had cirrhotic features documented by more than one modality (e.g., ultrasound, CT and oesophagogastroduodenoscopy). Twenty-three (85.2%) subjects had identifiable causes of cirrhosis (chronic hepatitis B: 51.9%, hepatitis C: 14.8%, chronic alcoholism: 18.5%). The cirrhosis cases had a mean SUV ratio of 1.29, which was less than that of the controls (p = 0.0003; Table 1). A representative MIP of a cirrhosis subject in Figure 1c shows that the visual contrast between liver and blood pool uptake was less than that of the control cases.

 

In the steatosis group, the HU_liver ranged from 5.5 to 55.2 (mean = 37.2) and the difference between HU_spleen and HU_liver (HU_spleen–HU_liver) ranged from 1 to 57 (mean = 16.6). The steatosis cases had a mean SUV ratio of 1.28, which was less than that of the controls (p = 0.0002; Table 1). The mean SUV ratio was still > 1, implying that liver had greater uptake than the blood pool. However, individuals with severe steatosis could have liver uptake as low as that of the blood pool, as illustrated in Figure 1d. To study if the SUV ratio depended on the severity of the steatosis, linear regression analysis was performed. The SUV ratios of the subjects with steatosis are plotted against the HU_spleen–HU_liver in Figure 2. It was observed that the SUV ratios decreased with increasing HU_spleen–HU_liver (grey squares, top x-axis). The corresponding slope in linear regression analysis was -0.0062 and the correlation coefficient was 0.209. The relationship between the steatosis severity and the SUV ratio was further evaluated by subdividing the subjects into mild (HU_spleen–HU_liver ≤ 10) and moderate-to-severe (HU_spleen–HU_liver > 10) steatosis cases in accordance with Jacobs et al’s study.[12] The mild steatosis cases had a higher mean SUV ratio of 1.35, while the moderate-to-severe steatosis cases had a lower mean SUV ratio of 1.25 (p = 0.0201; Table 2).

 

In the polycystic liver disease group, the mean SUV ratio was 0.75. This implies that unlike all the other cases which had mean ratios ≥ 1, the uptake in polycystic liver was generally less than that in the blood pool (Table 1). Figure 1e shows the MIP of a case of polycystic liver disease. The liver uptake was heterogeneous with a cold spot corresponding to a large hepatic cyst (upper green arrow). Table 1 shows that the SUV ratio of the polycystic liver disease cases had the greatest standard deviation (0.44) among all the cases (control group: 0.12, hyperbilirubinaemia group: 0.19, cirrhosis group: 0.10, steatosis group: 0.15) due to the variabilities in size, number and distribution of hepatic cysts with no uptake. Figure 3 shows another subject who had larger and more cysts compared with Figure 1e. The MIP and hybrid images in Figure 3 demonstrate almost no uptake in the right lobe of the liver and the corresponding SUV ratio was 0.12, while the left lobe of the liver (orange arrow in Figure 3b), with fewer and smaller cysts, demonstrated uptake similar to that of the blood pool and the spleen (yellow arrow in Figure 3b). Although the right hepatic lobe is commonly recommended as the standard metabolic reference in many international guidelines,[3] [4] [5] [6] Figure 3 clearly illustrates that the right lobe is less appropriate for reference compared with the left lobe when the right lobe is more diseased.

 

Figure 3. (a) Maximum intensity projection (MIP), (b) ¹⁸F-fluorodeoxyglucose positron emission tomography/computed tomography, and (c) computed tomography images of a patient suffering from polycystic liver disease. The right hepatic region in MIP has no uptake (green arrow in [a]) because of numerous large liver cysts, with a corresponding standardised uptake value ratio of 0.12. The left hepatic lobe (orange arrow in [b]) has uptake similar to that of the spleen (yellow arrow in [b]) because of smaller and fewer cysts compared to the right lobe. A hypermetabolic right lung tumour (purple arrow in [a]) with metastases to lymph nodes (blue arrow in [a]) and the L5 level (red arrow in [a]) are seen.

 

¹⁸F-FDG visual interpretation is advocated for oncological[3] [4] [5] [13] [14] [15] and inflammatory[6] [14] conditions, using liver and blood pool as metabolic references. While the PERCIST 1.0 criteria recommend that diseased liver is generally unsuitable for visual reference, the precise effects of different hepatic diseases on ¹⁸F-FDG uptake have not been entirely elucidated.[4] The current study included a spectrum of liver diseases ranging from biochemical abnormality (hyperbilirubinaemia) to various structural changes (cirrhosis, steatosis, and polycystic liver disease) that can either increase or decrease the SUV ratio. They can be ranked in terms of their mean SUV ratios in descending order of hyperbilirubinaemia, control, cirrhosis and steatosis, and polycystic liver disease.

 

The hyperbilirubinaemia cases showed higher SUV ratios than the controls. This is likely because jaundice implies hepatitis, and inflammation generally leads to increased ¹⁸F-FDG uptake.[15] This hypothesis is supported by our findings that higher serum bilirubin levels were associated with higher SUV ratios (Figure 2). High SUV ratio raises the concern for increasing false negative rate in lesion delineation, particularly if the lesion is only mildly ¹⁸F-FDG–avid. Clinical scenarios of hyperbilirubinaemia, due to biliary obstruction or acute hepatitis, are commonly encountered in oncology practice. In this study, hyperbilirubinaemia was used as an indicator of abnormal liver function and hepatitis. While serum bilirubin can be hepatic or haemolytic in origin, 89.7% of the hyperbilirubinaemia cases in this study exhibited elevated levels of other liver enzyme(s). This finding supports the hypothesis that the observed hyperbilirubinaemia was primarily hepatic in nature.

 

The cirrhosis cases had lower mean SUV ratios compared with the controls, probably as a result of the impaired glucose metabolism in liver fibrosis. Although liver biopsy is the gold standard for the diagnosis of cirrhosis, the procedure is invasive and not commonly employed. The cirrhosis subjects in this study were therefore selected based on radiological and endoscopic findings. For radiological findings, the sensitivity varies from 77% to 82% and the specificity ranges from 68% to 80%.[16] In this study, most cirrhotic cases (85.2%) had identifiable aetiologies of the cirrhosis. The majority (59.3%) also had cirrhotic features documented in more than one investigation.

 

The steatosis cases had lower SUV ratios than the control cases, consistent with previous observations.[17] We demonstrated that a more significant reduction in the SUV ratio can be expected in livers with higher degrees of steatosis. These results can be explained by the lower ¹⁸F-FDG uptake in fat content compared with normal liver parenchyma and the impaired glucose metabolism in steatosis. Abele and Fung’s study[18] showed that the SUV_mean in steatotic patients was lower than that of the controls (2.18 vs. 2.03). While this difference did not achieve statistical significance, the authors suggested that the limited power of their study might not have been sufficient to detect a true difference between the cases (i.e., type II error). The different sample sizes between Abele and Fung’s study (n = 23)[18] and the current report (n = 52) may offer an explanation of this discrepancy. On the other hand, Keramida et al[19] demonstrated no difference in SUV_mean between steatosis cases and the controls; however, their SUV_mean had a complicated adjustment for hepatic fat content and the potential effect of such adjustment on the original SUV magnitude is still unclear. Most clinical scenarios and research studies, including our investigation, had no adjustment. The current study excluded any subject with biochemical or structural abnormality of steatotic livers, while previous studies did not specifically elaborate the biochemical or structural properties.[17] [18] [19] Most importantly, the current study emphasises the SUV_liver/SUV_{blood pool} ratio instead of absolute SUV, because the ratio is more relevant to visual interpretation.

 

A simple cyst is defined as a thin-walled sac containing serous fluid.[20] Therefore, it has lower ¹⁸F-FDG uptake compared with normal hepatic parenchyma. This explains why the mean SUV ratio observed in polycystic liver cases was the lowest among all the cases. While polycystic liver is not considered a reliable metabolic reference, its SUV was still evaluated in a 3-cm–diameter fixed-sized spherical VOI in accordance with the PERCIST 1.0 criteria for equivalent comparison with other cases.[4] [5] The minimum SUV ratio among the polycystic liver cases was 0.12, which is clearly unsuitable as metabolic reference (Figure 3). This report further demonstrates that the hepatic uptake in polycystic liver disease depends on the size, number, and distribution of the cysts. The heterogeneous and variable uptake in the polycystic liver prohibits its application as a reliable metabolic reference.

 

Steatosis, cirrhosis and polycystic liver disease cases showed lower SUV ratios than those in the control cases. Their hepatic uptake could be similar to or even lower than blood pool or splenic uptake. This observation was distinct from the observation in the controls, in which normal liver uptake was always greater than blood pool and splenic uptake. A previous study of high-grade lymphoma indeed has suggested that lymphomatous involvement in spleen should be suspected if the spleen has greater uptake than the liver.[21] Therefore, the deceptively low liver-to–blood pool SUV ratio observed in steatosis, cirrhosis, and polycystic liver disease can potentially lead to higher false positive rates in lesion detection.

 

All disease cases had mean SUV ratios different from the controls and the differences were statistically significant in two-sided t tests. However, such quantitative differences may not always appear conspicuous in the qualitative visual interpretation. For example, the visual contrast between liver/blood pool uptake in the cirrhosis cases was slightly less than that in the control cases. The visual differences between the mild hyperbilirubinaemia/steatosis cases and the controls were also subtle. On the other hand, in severe hyperbilirubinaemia, severe steatosis and polycystic liver disease cases, their visual contrast between liver/blood pool uptake was obviously different from that in the controls. Thus, severe liver diseases can have significant quantitative and qualitative effects on the liver/blood pool references. This can eventually affect the diagnostic accuracy of visual interpretation.

 

The current study had some limitations. First, subtle hepatic tumour or metastasis may be present. To minimise this pitfall, subjects with a histological diagnosis or radiological suspicion of liver malignancy were excluded from this study. Second, the number of cases of polycystic liver disease (10 patients) was lower compared to other groups because of its inherently low prevalence. Third, the liver function tests were obtained within 2 weeks, rather than on the same day, of ¹⁸F-FDG PET/CT acquisition. Fourth, this study demonstrated that all disease cases had SUV ratios different from that of the controls in quantitative aspect. Future study is required to determine if this quantitative difference can be translated to significant changes in qualitative visual interpretation.

 

Liver/blood pool uptake in ¹⁸F-FDG PET/CT can be influenced by various liver conditions, including hyperbilirubinaemia, cirrhosis, steatosis, and polycystic liver disease. As liver diseases progress in severity, their impact on liver/blood pool uptake becomes more prominent. Therefore, radiologists should exercise great caution in the utilisation of liver/blood pool uptake as metabolic references in cases of significant hepatic disease. Clinical history, biochemical function and imaging findings should be thoroughly reviewed before visual interpretation.