Author + information
- Received September 26, 2011
- Revision received November 26, 2011
- Accepted November 29, 2011
- Published online June 5, 2012.
- Jan Bucerius, MD⁎,†,‡,§,
- Venkatesh Mani, PhD⁎,†∥,
- Colin Moncrieff⁎,†,
- James H.F. Rudd, MD, PhD¶,
- Josef Machac, MD#,
- Valentin Fuster, MD, PhD∥,⁎⁎,
- Michael E. Farkouh, MD, MSc∥,†† and
- Zahi A. Fayad, PhD⁎,†∥,⁎ ()
- ↵⁎Reprint requests and correspondence:
Prof. Zahi A. Fayad, Translational and Molecular Imaging Institute, Mount Sinai School of Medicine, One Gustave L. Levy Place, P.O. Box 1234, New York, New York 10029
Objectives In this study, the impact of noninsulin-dependent type 2 diabetes mellitus on carotid wall 18F-fluorodeoxyglucose (FDG) uptake in patients with documented or suspected cardiovascular disease was evaluated.
Background Inflammation is a pivotal process in the progression of atherosclerosis, which can be noninvasively imaged by FDG positron emission tomography (FDG-PET).
Methods Carotid artery wall FDG uptake was quantified in 134 patients (age 60.2 ± 9.7 years; diabetic subjects, n = 43). The pre-scan glucose (gluc) level corrected mean of the maximum standardized uptake value (SUV) values (meanSUVgluc), mean of the maximum target-to-background ratio (meanTBRgluc), and single hottest segment (SHSgluc) of FDG uptake in the artery wall were calculated. Associations between FDG uptake, the presence of risk factors for atherosclerosis, and diabetes were then assessed by multiple regression analysis with backward elimination.
Results The study demonstrated a significant association between diabetes and FDG uptake in the arterial wall (diabetes meanSUVgluc β = 0.324, meanTBRgluc β = 0.317, and SHSgluc β = 0.298; for all, p < 0.0001). In addition, in diabetic patients, both body mass index ≥30 kg/m2 (meanSUVgluc β = 0.4, meanTBRgluc β = 0.357, and SHSgluc β = 0.388; for all, p < 0.015) and smoking (meanTBRgluc, β = 0.312; SHSgluc, β = 0.324; for all, p < 0.04) were significantly associated with FDG uptake.
Conclusions Type 2 diabetes was significantly associated with carotid wall FDG uptake in patients with known or suspected cardiovascular disease. In diabetic patients, obesity and smoking add to the risk of increased FDG uptake values.
Cardiovascular disease (CVD) remains one of the leading causes of death in the United States, accounting for 1 in every 2.8 deaths (1). It is also known that diabetes mellitus is associated with a high risk of CVD. Atherosclerotic plaque inflammation plays a central role in atherosclerotic plaque progression, vulnerability, and thrombogenicity. The exact mechanisms underlying the association between diabetes and atherosclerotic disease are not well known (2). Epidemiological studies suggest that type 2 diabetes mellitus is not merely an independent risk factor for atherosclerotic CVD but is also associated with increased levels of inflammatory biomarkers (3). In line with these findings, efforts have been made to use noninvasive imaging to quantify vessel wall inflammation and to provide further evidence of the impact of clinical risk factors such as diabetes on atherosclerosis.
Carotid 18F-fluordeoxyglucose (FDG) positron emission tomography (PET) has been shown to reflect the metabolic rate of glucose, a process known to be enhanced in inflamed tissue. Uptake of FDG has been shown to be significantly associated with both the degree of macrophage infiltration and the levels of inflammatory gene expression in plaques (4–6). Several clinical CVD risk factors have also been shown to exhibit a significant association with carotid wall inflammation (7–9). However, data regarding the impact of diabetic disease on vessel wall inflammation remain inconclusive. Some trials failed to show any association between diabetes and atherosclerosis as assessed by FDG-PET (8,9), whereas others have observed a significant association between diabetic disease and vascular inflammation (7,10). Tumor PET studies have also shown that FDG accumulation is diminished during hyperglycemia (11–13). This effect however, has not been evaluated in vascular PET imaging.
The aim of the present study was to assess the impact of non–insulin-dependent type 2 diabetes on carotid wall FDG uptake while seeking to avoid some of the limitations of previous studies. Therefore, we designed a cross-sectional study in a larger sample population and performed FDG-PET imaging with protocols optimized for vessel wall visualization/analysis (7,9,14).
This was a cross-sectional study, investigating the impact of non–insulin-dependent type 2 diabetes on the prevalence of carotid wall inflammation assessed by FDG-PET. This study, approved by the institutional review board, was conducted from February 2006 until October 2010 at the Mount Sinai School of Medicine, New York. All subjects gave written informed consent.
Inclusion criteria were as follows: male and female subjects with a diagnosis of CVD or subjects with multiple CVD risk factors were recruited. Cardiovascular disease was defined as a previous myocardial infarction, stroke, transient ischemic attack (TIA), history of peripheral artery disease, or a history of a coronary revascularization procedure. Patients with insulin-dependent type 2 diabetes and patients with type 1 diabetic disease were excluded from the study, as were subjects with fasting glucose levels ≥11.1 mmol/l or previous carotid surgery.
Questionnaire and biometric and biochemical measurements
Presence of cardiovascular risk factors, use of medication, and family history of CVD were assessed by a questionnaire. Presence of hypertension was defined as a history of systolic blood pressure >140 mm Hg, or a diastolic blood pressure >90 mm Hg. Diabetes was defined as documented diagnosis of type 2 diabetic disease and the use of antidiabetic treatment (diet or oral, no insulin treatment). Weight and height were measured to calculate body mass index (BMI). Smoking was defined as smoking at least 1 cigarette on a daily basis. Fasting glucose levels were obtained by finger stick blood glucose measurements (Accu-Chek Advantage, Roche Diagnostics, Indianapolis, Indiana) before FDG administration.
FDG-PET/computed tomography imaging
The FDG-PET/computed tomography (CT) imaging was performed after an overnight fast using a Lightspeed discovery ST 16-slice PET/CT scanner (General Electric Healthcare, Milwaukee, Wisconsin). The FDG was administered intravenously (557.6 ± 84.0 MBq), and patients rested comfortably for 97 to 193 min (136.7 ± 21.2 min) before the scan of the neck was started. Subjects were placed into a head holder for imaging of the carotids. A low-dose CT scan (140 kV, 80 mA, and 4.25-mm slice thickness) was performed for attenuation correction and coregistration. Images from 1 bed position (15.5 cm) with coverage extending inferior to the internal auditory meatus were acquired in 3-dimensional mode using a 128 × 128 pixel matrix for 15 min. >No CT contrast agent was administered. The total radiation dose from participating into this study was approximately 12 mSv.
Image analysis was performed on a dedicated commercially available workstation (Extended Brilliance Workspace V220.127.116.1106, Philips Medical Systems, Cleveland, Ohio). An experienced reader (J.B.) analyzed all scans. Methodology for analysis and reproducibility of the measurements have been previously reported (15).
Briefly, arterial FDG uptake was quantified by manually drawing a region of interest around each artery (common carotid arteries) on every slice of the coregistered transaxial PET/CT images. Next, the maximum arterial standardized uptake value (SUV [highest pixel activity within the region of interest]) was determined. The SUV is the decay-corrected tissue concentration of FDG in kBq/ml, adjusted for the injected FDG dose and the body weight of the patient. By averaging the maximum SUV values of all arterial slices of the left and right carotid artery, a meanSUV value was derived for the carotid arteries.
The meanSUV values were also corrected for patient's fasting pre-scan glucose levels to account for a competitive impact of glucose (gluc) and FDG using an established formula (16). The measured glucose content was normalized for an overall population average of 5.0 mmol/l (16) as follows: meanSUVgluc = meanSUV × patient's blood glucose (mmol/l)/5.0 mmol/l.
The arterial target-to-background ratio (TBRgluc) was calculated by normalizing the SUVgluc for blood pool activity by dividing the SUVgluc value in the artery by the average blood mean SUV estimated from both jugular veins. The TBRgluc is a blood-normalized arterial SUV, considered to be a reflection of arterial FDG uptake and reflective of underlying macrophage activity (14). For evaluation of the FDG blood pool activity, at least 6 3- to 4-mm regions of interest were placed in consecutive slices of both jugular veins and averaged.
The arterial TBR values obtained were then averaged to derive a meanTBRgluc for both carotid arteries. Additionally, we identified the glucose-corrected single hottest slice (SHSgluc), defined as the highest TBRgluc value of the carotid arteries.
All continuous variables are expressed as mean ± SD, and categorical data as absolute numbers and percentages throughout this paper. To also assess a potential relation between continuous BMI values of the patients and the different FDG uptake parameters instead of assessing the impact of increased BMI values ≥30 kg/m2, Pearson's or Spearman's rho correlation coefficients (r) were calculated, depending on normal distribution, between the BMI values between the uncorrected- and glucose-corrected FDG uptake parameters in the entire study population as well as in the 2 subgroups of patients with and without type 2 diabetes. In general, normal distribution of data was tested for all of the different statistical calculations using the Kolmogorov-Smirnov test.
The study population was divided into 2 subgroups of patients, namely, patients with and patients without diabetes. Furthermore, to test for the effect of antidiabetic medication on the degree of carotid wall inflammation, we compared the different FDG uptake parameters between patient receiving oral antidiabetic drugs and patients receiving dietary treatment only. For all comparisons of subgroups of patients, the Student t test or the Mann–Whitney U test was performed to compare continuous variables depending on normal distribution. Categorical variables were compared between diabetic and nondiabetic patients by using Fisher's exact test.
Multiple Regression With Backward Elimination and Linear Regression With the Enter Method
[Significant variables of the backward elimination entered a consecutive linear regression model in a block in a single step. To look at the relationship between each of >1 independent variable on 1 dependent variable, multiple regression analyses were used in the present study.]
A multiple linear regression analysis with backward elimination was used to assess the association between the cardiovascular risk factors and glucose-corrected FDG uptake parameters (meanSUVgluc, meanTBRgluc, and SHSgluc) in the entire study population as well as in both subgroups of patients with and without diabetes (17,18). The FDG uptake parameters were treated as the response variables (dependent) and cardiovascular risk factors as the explanatory (independent) variables for the regression analysis. The explanatory variables included were as follows: age >65 years, male, BMI ≥30 kg/m2, statin use, type 2 diabetes, history of CVD, smoking, alcohol use, hypertension, and family history of CVD. To evaluate a potential beneficial effect of exercise on carotid wall inflammation, exercise was 1 of the explanatory variables in the regression analyses. After this, the ENTER regression method was used to determine independent predictors of the response variables. For this method, all of the explanatory variables of the backward elimination model that showed a significant association with the FDG uptake value were retained and entered into the regression model in a block in a single step. This entry method was preferred over the forward selection of variables, because after excluding all of the explanatory variables without a significant association with the different carotid wall FDG uptake values, only a few significant variables were left for a relatively low number of cases. A similar analysis approach was also followed for the FDG uptake values obtained without the applied glucose correction (meanSUV, meanTBR, SHS) (16). Throughout this paper, all results of the multiregression models were given with the standardized regression coefficient (β), the 95% confidence interval, and the p value for the estimate of the statistical significance.
The patients were grouped into tertiles based on their FDG-PET uptake parameters. Pearson chi-square tests were performed to compare the prevalence of clinical variables across the groups of patients classified by tertiles of the different FDG uptake parameters (meanSUVgluc, meanTBRgluc, and SHSgluc).
Analysis of variance with the appropriate adjustment for multiple comparisons was performed to compare FDG uptake values between different levels of fasting glucose in nondiabetic patients and patients with diabetes (according to the recommendations by the International Diabetes Federation IGF/IGT consensus statement ). Post-hoc analyses were performed using the Tukey test. All statistical analyses were performed using the SPSS statistical package, version 16.0 (SPSS, Chicago, Illinois).
In all, 134 patients were included in the study. In 3 of the patients, FDG-PET analysis could not be performed because of high FDG uptake in the thyroid, affecting the visualization and analysis of FDG uptake within the carotid arteries, leaving 131 eligible patients for image analysis. On average, 7.56 ± 2.49 slices for the left common carotid artery and 7.36 ± 2.5 slices for the right common carotid artery were analyzed to derive the FDG uptake parameters. Table 1 shows the characteristics of the whole population, both diabetic and nondiabetic subjects. There were some differences in demographics between the diabetic group and nondiabetic group (history of percutaneous coronary intervention and use of beta-blockers being higher among diabetic patients; exercise, family history of CVD, and cigarettes per day in current smokers all being higher among nondiabetic patients); otherwise, the 2 groups were similar. As expected, fasting glucose was higher in the diabetic population compared with that of nondiabetic subjects.
FDG-PET imaging results
The imaging analyses of the groups (Table 2), however, showed that glucose-corrected FDG-PET parameters (meanSUVgluc, meanTBRgluc, SHSgluc) were significantly higher in the diabetic group compared with the nondiabetic group. This difference was not observed in FDG-PET parameters that were not corrected for pre-scan glucose. Dividing patients into 2 groups according to their pre-scan glucose values by using the median of these values, we also found significantly higher values for all glucose-corrected FDG uptake parameters in diabetic patients in the group of higher (5.6 to 10.6 mmol/l) pre-scan glucose levels. In the group of patients with lower pre-scan glucose levels (2.8 to 5.5 mmol/l) all glucose-corrected FDG uptake parameters were also found to be higher in diabetic patients; however, these differences failed to be statistically significant (Table 2). In diabetic patients, no significant differences were observed for all glucose-corrected FDG parameters between patients treated with oral antidiabetic drugs and patients on dietary treatment only (meanSUVgluc p = 0.313, meanTBRgluc p = 0.314, SHSgluc p = 0.122).
FDG-PET imaging and CVD risk factors
Table 3 shows the results of the multiple linear regression analysis with backward elimination to identify clinical risk factors associated with glucose-corrected FDG-PET uptake parameters (i.e., measures of plaque inflammation). Diabetes showed the strongest relationship with all FDG-PET uptake parameters (meanSUVgluc, meanTBRgluc, SHSgluc) followed by BMI >30 kg/m2. Only risk factors that had a p value <0.10 were retained in the model for the ENTER regression (all significant and therefore retained variables entered this model in a block in a single step) and are shown in Table 3. After the ENTER regression analysis, it was found that diabetes, BMI >30 kg/m2, and alcohol use (except of meanSUVgluc values) were independent predictors of plaque inflammation as measured by FDG-PET imaging (Figs. 1 to 3)⇓⇓⇓. The ENTER regression also showed that meanTBRgluc and SHSgluc were inversely associated with a family history of CVD (Figs. 2 and 3).
The multiple regression analysis for non–glucose-corrected FDG-PET data is presented in Online Table 1.
Multiple regressions in subgroups
Multiple linear regression analyses in the subgroup of patients with and patients without diabetic disease is presented in Table 4 after a similar procedure as described previously. In diabetic subjects, BMI ≥30 kg/m2 continued to be significantly associated with all of the 3 glucose-corrected FDG-PET uptake parameters. In nondiabetic subjects, hypertension showed the strongest association with the FDG-PET uptake parameters. These and other associations are shown in Table 4.
Again, results for the regression analysis for non–glucose-corrected FDG-PET data are presented in Online Table 2.
Online Table 3 shows clinical characteristics stratified by tertiles of meanSUVgluc, meanTBRgluc, and SHSgluc. The prevalence of type 2 diabetes and BMI ≥30 kg/m2 were both significantly higher at higher tertiles of the 3 glucose-corrected FDG-PET uptake parameters.
Correlation between continuous BMI values and FDG uptake parameters in the entire study population and the subgroups
Positive significant correlations were found between the continuous BMI values and all of the uncorrected and glucose-corrected FDG uptake parameters in the entire study population and the 2 subgroups (r > 0.25, p < 0.02 for all), except TBRmax values in the total study population and in nondiabetic patients, SUVmax values in diabetic patients, as well as TBRmaxgluc and SHSgluc values in nondiabetic patients.
Distribution of FDG-PET uptake parameters according to pre-scan glucose levels
Online Figure 1 depicts significantly increasing FDG uptake parameters (meanSUVgluc, meanTBRgluc, and SHSgluc) by increments of fasting glucose levels in patients with type 2 diabetes. Remarkably, FDG uptake values in nondiabetic subjects were similar (SHSgluc) or even slightly, but not significantly, higher (meanSUVgluc, meanTBRgluc) compared with those of diabetic patients with fasting glucose levels within the normal range (<6.1 mmol/l) (Online Fig. 1).
The aim of our study was to determine if the presence of type 2 diabetes was related to carotid wall FDG uptake. This relation might therefore represent a link between diabetic disease and carotid wall inflammation in a population of patients with known CVD or multiple risk factors for it. We used a cross sectional study design in a larger sample population than previous studies (7–10) and performed FDG-PET imaging with protocols optimized for vessel wall FDG uptake (9,14). Our results demonstrate that diabetes was significantly associated with the FDG uptake in the carotid wall. Additionally, we showed that obesity was also related to carotid wall inflammation as depicted by FDG-PET. In the nondiabetic group, hypertension was the leading variable associated with inflammation measured by FDG-PET uptake. We also identified increasing fasting glucose levels in diabetic patients to be significantly associated with increments of the FDG uptake, which might be indicative of a higher propensity for carotid wall inflammation with increasing degrees of hyperglycemia.
The rationale for choosing to perform glucose correction of the FDG uptake is based on several oncology studies, which suggest that elevated pre-scan glucose levels can influence significantly the tumor's uptake of FDG during PET imaging (12,20,21). One potential pitfall to using a glucose correction, however, is a resultant increase in the variability of the SUV measurements (22). The role of glucose correction of FDG uptake in noncancer lesions is not well understood. However, we believe, because the mechanism of uptake of FDG into inflammatory cells is the same as for tumor cells, that the same correction should be applied. In accordance with the European Association of Nuclear Medicine procedure guidelines for tumor PET imaging, we have also presented the results without glucose correction (Online Appendix) (16).
Multivariate regression analyses revealed an unexpected negative association between diabetes and the uncorrected meanSUV values of the carotids in our study. This finding is in contrast to the well-known clinical impact of diabetes on CVD. However, future studies still need to be performed to investigate whether the corrected or uncorrected FDG uptake values are more sensitive surrogate markers for carotid wall inflammation by correlating both FDG uptake parameters with the histological assessment of vascular inflammation.
Impact of circulation time on FDG uptake
The optimal circulation time before imaging plaque inflammation has still not been definitively established. Typically, a circulation time between 1 h and 3 h is used by most groups (4,5,23,24). To exclude an impact of the FDG circulation time on the FDG uptake in patients with and patients without diabetes, we compared the FDG circulation time between both groups of patients. As we did not find a statistically significant difference, the differences of the FDG uptake between diabetic patients and nondiabetic patients cannot be explained by different FDG circulation times in both groups.
Type 2 diabetes and carotid wall inflammation
We observed that type 2 diabetes and obesity (BMI) were independently associated with increased FDG-PET uptake values. In patients with type 2 diabetes, obesity and smoking added additional risk for increased FDG uptake in the carotid wall. In nondiabetic patients, however, only hypertension was found to be significantly associated with carotid wall inflammation as depicted by all FDG uptake parameters.
Our results are in agreement with some of the previously published studies evaluating the association between cardiovascular risk factors and vessel wall inflammation. In a case-control study of patients with type 2 diabetes, impaired glucose tolerance, and controls, Kim et al. (10) reported higher TBRmax values in both study groups compared with the control group. As in the present study, they also observed increasing prevalence of diabetes with increments of maximum TBR values as depicted by tertiles. However, glucose correction was not used in their study, and that may have resulted in an underestimation of FDG-PET uptake values.
Previous studies have demonstrated the link between circulating insulin levels and its effect on the over-expression of glucose transporter protein types (10,11). Tahara et al. (7) have also shown that carotid inflammation was associated with several cardiovascular risk factors, including a homeostasis model assessment of insulin resistance. They found obesity, as assessed by waist circumference and use of hypertensive medication, to be significantly associated with carotid wall inflammation. However, fewer subjects in their population had CVD, and the cardiovascular risk profile for their population was much lower primarily because their retrospective analysis was done among cancer patients. In our study, we found a relationship between hypertension and carotid wall inflammation measured by FDG-PET in the nondiabetic group and a relationship between BMI and carotid wall FDG uptake for both diabetics and nondiabetic patients.
Relationship between fasting glucose levels and carotid wall FDG uptake
Studies have shown that hyperglycemia leads to increased oxidative stress producing endothelial dysfunction (25). Several observational studies have shown an association between levels of glycemia and macrovascular events in patients with diabetes (26,27). Early data from the UK Prospective Diabetes Study suggested a protective effect of improved glucose control on CVD incidence and mortality (26,29). Results from our study seem to support these findings, as we found similar meanTBRgluc values in patients with diabetes and fasting glucose levels <6.1 mmol/l compared with nondiabetic patients (p = 0.985). We also observed higher FDG uptake parameters with increasingly poorer glycemic control.
Relationship between hypertension and carotid wall FDG uptake
Two recently published trials prospectively investigated the impact of hypertension on cardiovascular risk in patients with and patients without diabetic disease (28,29). Both trials found hypertension to be associated with a higher risk of CVD among diabetic patients but failed to show a significant interaction between diabetes and increased blood pressure. We found that hypertension showed a significant association with carotid wall FDG uptake only in the nondiabetic subgroup.
Firstly, we did not obtain serum lipid levels, markers of glucose metabolism, or serum inflammatory markers. Secondly, this is a cross-sectional study. Therefore, we could not address whether there was a causal relationship between the presence of diabetic disease and vessel wall inflammation. Thirdly, it is unknown whether vessel wall FDG uptake is predictive of progression of disease or future cardiovascular events in diabetic patients. Longitudinal studies are currently under way to establish such a relation. Finally, in the present study, image analyses were performed by only 1 reader, which might reduce statistical noise related to interobserver variation but might raise concerns regarding intraobserver bias; however, previous reports demonstrated that this method has good interobserver and intraobserver reproducibility (15).
In the present study, we show that type 2 diabetes has a significant impact on the FDG uptake in the wall of the carotid arteries. Obesity (BMI ≥30 kg/m2) and smoking are also significantly associated with FDG-PET uptake parameters in diabetic patients. For nondiabetic patients, hypertension was significantly associated with carotid wall inflammation. Furthermore, the degree of the carotid wall FDG uptake increases with increments of fasting glucose levels in diabetic patients. Whether the glucose-corrected FDG uptake parameters are indicative of vessel wall inflammation has to be determined by future studies.
The authors wish to thank Ash Rafique, RT, BS, CNMT, for his assistance with the image acquisition.
For supplemental tables and figures, please see the online version of this article.
Impact of Non-Insulin-Dependent Type 2 Diabetes on Carotid Wall 18F-Fluorodeoxyglucose Positron Emission Tomography Uptake
This work was partly supported by the NIHR Cambridge Biomedical Research Centre (to Dr. Rudd). Partial support was also provided by National Institutes of Health, National Heart, Lung, and Blood InstituteR01 HL071021 (to Dr. Fayad) and R01 HL078667 (to Drs. Farkouh and Fayad). All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- body mass index
- computed tomography
- cardiovascular disease
- positron emission tomography
- single hottest segment
- standardized uptake value
- target-to-background ratio
- transient ischemic attack
- Received September 26, 2011.
- Revision received November 26, 2011.
- Accepted November 29, 2011.
- American College of Cardiology Foundation
- Roger V.L.,
- Go A.S.,
- Lloyd-Jones D.M.,
- et al.,
- American Heart Association Statistics Committee and Stroke Statistics Subcommittee
- Libby P.,
- Ridker P.M.,
- Maseri A.
- Tawakol A.,
- Migrino R.Q.,
- Bashian G.G.,
- et al.
- Rudd J.H.,
- Warburton E.A.,
- Fryer T.D.,
- et al.
- Tahara N.,
- Kai H.,
- Yamagishi S.,
- et al.
- Rominger A.,
- Saam T.,
- Wolpers S.,
- et al.
- Rudd J.H.,
- Myers K.S.,
- Bansilal S.,
- et al.
- Kim T.N.,
- Kim S.,
- Yang S.J.,
- et al.
- Rudd J.H.F.,
- Myers K.S.,
- Bansilal S.,
- et al.
- Draper N.,
- Smith H.
- Langen K.J.,
- Braun U.,
- Rota Kops E.,
- et al.
- Lindholm P.,
- Minn H.,
- Leskinen-Kallio S.,
- Bergman J.,
- Ruotsalainen U.,
- Joensuu H.
- Menezes L.J.,
- Kotze C.W.,
- Hutton B.F.,
- et al.
- Rudd J.H.,
- Elkhawad M.,
- Fayad Z.A.
- Ceriello A.
- Eguchi K.,
- Pickering T.G.,
- Hoshide S.,
- et al.