Prognostic Significance of the Intratumoral Heterogeneity of18F-FDG Uptake in Oral Cavity Cancer

SOO HYUN KWON, MD,1 JOON-KEE YOON, MD, PhD,1 YOUNG-SIL AN, MD, PhD,1

YOO SEOB SHIN, MD, PhD,2 CHUL-HO KIM, MD, PhD,2 DONG HYUN LEE, MD,1

KYUNG SOOK JO, MD,1 AND SU JIN LEE, MD, PhD1*

1Departments of Nuclear Medicine and Molecular Imaging, Ajou University School of Medicine, Suwon, Korea2Otolaryngology, Ajou University School of Medicine, Suwon, Korea

Backgrounds: We evaluated the prognostic value of the intratumoral heterogeneity of 18F‐FDG uptake in oral cavity cancer.Materials and Methods: We enrolled 45 patients who underwent pretreatment 18F‐FDG PET/CT. The intratumoral heterogeneity of 18F‐FDGuptake was represented as the heterogeneity factor (HF), defined as the derivative (dV/dT) of a volume‐threshold function for a primary tumor. Wemeasured the maximum standardized uptake value (SUVmax) and volumetric PET parameters. The relationship between HF and clinical parameters,as well as other PET parameters, was evaluated.Results: The HF was significantly correlated with SUVmax (r¼�0.353, P¼ 0.017), metabolic tumor volume (r¼�0.708, P< 0.0001), and totallesion glycolysis (r¼�0.709, P< 0.0001). A multivariate analysis revealed not only cervical lymph node metastasis (hazard ratio¼ 5.983;P¼ 0.022) but also HF (hazard ratio¼ 2.49� 10�4; P¼ 0.002) to be independent predictors of overall survival. Those patients with HF<�0.13showed a worse prognosis than those with HF ��0.13 (P¼ 0.005).Conclusions: The intratumoral heterogeneity of 18F‐FDG uptake may be a significant prognostic factor for overall survival in addition to cervicallymph node metastasis in oral cavity cancer.J. Surg. Oncol. 2014;110:702–706. � 2014 Wiley Periodicals, Inc.

KEY WORDS: fluorine 18‐fluorodeoxyglucose (18F‐FDG); positron emission tomography/computed tomography (PET/CT);oral cavity cancer; intratumoral heterogeneity; prognosis

INTRODUCTION

Head and neck cancer (HNC) is the sixth leading cause of cancerworldwide, representing 3% of all malignancies, and oral cavitycancer accounts for approximately 30% of HNCs [1]. About 275,000people are newly diagnosed with oral cavity cancer annually, and itsincidence has increased in developing countries [2]. During thelast decade, 18F‐fluorodeoxyglucose positron emission tomography/computed tomography (18F‐FDG PET/CT) has been widely used as auseful imaging tool not only for cancer staging but also for the evaluationof treatment response in HNC [3–5].

Many studies have identified various clinical and radiographicprognostic factors in patients with HNC, although the survival rate hasnot improved over the past three decades [6].Many ongoing studies haveused PET/CT to predict patient prognosis. The most commonly usedPET parameter to show the degree of 18F‐FDGuptake is the standardizeduptake value (SUV), which is an independent prognostic factor inHNC [7,8]. Recently, volumetric PET parameters such as the metabolictumor volume (MTV) and total lesion glycolysis (TLG) have emerged asnovel prognostic factors that may improve risk stratification and theplanning of individualized cancer treatment protocols [9–13].

Malignant tumor cells are composed of heterogeneous components—not only biologic constituents, but also gene expression, metabolic, andbehavioral characteristics. Heterogeneity varies in the same cancer typeand has a wide spectrum even for the same stage because there aredifferences in such properties as the growth rate, vascularity, and necrosiswithin the same tumor cell population [14]. However, studies concerningthe clinical significance of intratumoral heterogeneity are limited, eventhough it is a potential prognostic factor. Intratumoral heterogeneitycould be assessed using objective criteria from 18F‐FDG PET images.Several studies have suggested that the histopathologic characteristics ofmalignancies such as non‐small cell lung cancer, oligodendroglioma, and

squamous cell carcinoma of the head and neck are significantly correlatedwith the intratumoral heterogeneity of 18F‐FDG uptake on PET [15–17].Additionally, the 18F‐FDG heterogeneity of malignant sarcoma andcervical cancer was associatedwith patient prognosis [18,19]. Recently, astudy concerning the intratumoral heterogeneity of 18F‐FDG uptake innasopharyngeal cancer revealed that it was significantly correlated withtumor aggressiveness and was associated with various outcomemeasures [14]. However, to the extent of our knowledge, no study hasevaluated the prognostic value of the intratumoral heterogeneity of PETin oral cavity cancer.

Therefore,we investigated the relationship between tumor heterogeneityand various clinical and PET parameters and then evaluated the prognosticvalue of heterogeneity factors in oral cavity cancer.

MATERIALS AND METHODS

Patients

We reviewed the pretreatment 18F‐FDG PET/CT of 54 consecutivepatients with pathologically proven oral cavity cancer who underwent

*Correspondence to: Corresponding author: Su Jin Lee, M.D., Ph.D.Department of Nuclear Medicine and Molecular Imaging, Ajou UniversitySchool of Medicine. 164, World cup‐ro, Yeongtong‐gu, Suwon 443‐380,Republic of Korea. F: 82‐31‐219‐5950 E‐mail: suesj202@ajou.ac.kr

Received 31 March 2014; Accepted 26 May 2014

DOI 10.1002/jso.23703

Published online 29 June 2014 in Wiley Online Library(wileyonlinelibrary.com).

Journal of Surgical Oncology 2014;110:702–706

� 2014 Wiley Periodicals, Inc.

treatment in our hospital from October 2005 to July 2013. Of these, weexcluded nine patients who had tumors smaller than 1 cm. Ultimately,this study evaluated 45 patients. The ethics committee of our institutionapproved this retrospective study.

All patients underwent contrast‐enhanced neck CT and 18F‐FDGPET/CT scans to evaluate the initial disease stage. Additionally, neckultrasonography and neck magnetic resonance imaging (MRI) wereperformed if clinically indicated. After the staging work‐up, 19 patients(42.2%) underwent surgical treatment only, while 24 patients (53.3%)were treated with surgery followed by adjuvant radiation therapy (RT;n¼ 16) or concurrent chemoradiation therapy (n¼ 8). The remainingtwo patients underwent RT alone. Clinical follow‐up was performedevery 3 to 6 months during the first year. The mean follow‐up durationwas 27.5 months with a range of 0.6–71.2 months.

PET/CT Imaging

All of the subjects fasted for at least 6 h before PET/CT. The bloodglucose level at the time of injection of FDG was<200mg/dl in all of thepatients. PET/CT was performed using two types of dedicated PET/CTscanners: Discovery ST (n¼ 24; GE Healthcare, Milwaukee, WI, USA)andDiscovery STE (n¼ 21; GEHealthcare). Before PET, unenhancedCTwas performed at 60min after a 5MBq/kg FDG injection using 8‐ or 16‐slice helical CT (120keV, 30–100mA in the AutomA mode; sectionwidth¼ 3.75mm), and then an emission scan was acquired from the thighto the head for 2.5min per frame in three‐dimensional mode. Attenuation‐corrected PET images using CT data were reconstructed by an ordered‐subsets expectation maximization algorithm (20 subsets, 2 iterations).

Measurement of Metabolic PET Parameters

All of the PET/CT images were reviewed by two experienced nuclearmedicine physicians using a dedicated workstation (GE AdvantageWorkstation 4.4; GE Healthcare). First, volume‐based metabolicparameters were achieved using volume viewer software, whichrepresents the volume of interest (VOI) automatically using anisocontour threshold method based on the SUV. Various thresholdswere adopted to determine the VOI boundaries: fixed values of SUV at2.5, 3.0, and 3.5 and a relative threshold of 40 to 80% of the SUVmax ofthe tumors. In the present study, MTVs were automatically calculated bysumming the total volumes of the voxels with threshold SUVs of 2.5, 3.0,and 3.5, as well as 40–80% of the SUVmax in the VOI, and wererepresented as MTV (2.5), MTV (3.0), MTV (3.5), and MTV (40–80%),

respectively. TLG was calculated using threshold SUVs of 2.5, 3.0, and3.5, as well as 40–80% of the SUVmax; the results were designated TLG(2.5), TLG (3.0), TLG (3.5), and TLG (40–80%), respectively.Additionally, we evaluated the intratumoral heterogeneity of 18F‐FDGuptake in oral cavity cancer, represented by the heterogeneity factor (HF).HFwas the derivative (dV/dT) of a tumor volume‐threshold function of aseries of SUV thresholds from 40 to 80% for a primary tumor. Figure 1shows representative PET/CT images using various SUV thresholds. TheMTVs at each SUV threshold were recorded, and then the HF wascalculated as the slope from the linear regression on the threshold‐volumecurve (Fig. 2). Thus, HF had a negative value, and a larger negative valuereflected inhomogeneous tissue in the primary tumor.

Statistical Analysis

We expressed statistical data as frequencies and percentages forcategorical variables and as means� standard deviation for continuousvariables. The Kruskal‐Wallis test was used to compare the differencesbetween HF and each tumor stage. Spearman’s rank test was obtained tocorrelate HF and various PET parameters.

Fig. 1. Representative PET/CT images in right tongue cancer using several thresholdmethods. Decreasing tumor volumes are noted by 40% to 80%threshold of SUVmax.

Fig. 2. Threshold‐volume curve by plotting thresholds to volumesautomatically calculated with variable thresholds of 40, 50, 60, 70, and80% of SUVmax. Heterogeneity factor (‐0.18) is calculated as the slopfrom linear regression.

Journal of Surgical Oncology

Intratumoral Heterogeneity of FDG in Oral Cavity Cancer 703

Overall survival was used as a measurement of patient outcome,defined as the time from initial diagnosis to death. Cancer‐ or treatment‐related death was considered as an event. Survival curves were estimatedusing the Kaplan‐Meier method, and differences between low‐ and high‐HF groups were compared using the log‐rank test. Maximally selectedlog‐rank statistics were used to determine the optimal cutoff showing thebest discrimination of survival for HF. The prognostic significance ofPET parameters, including HF and clinical variables, was determinedusing a Cox proportional hazards regression model for univariate andmultivariate analyses of overall survival. All statistical analyses wereperformed using SPSS software (IBM SPSS Statistics 18; IBM Inc.,New York, NY, USA); a P‐value< 0.05 was deemed to indicatestatistical significance.

RESULTS

The patients’ demographic data and clinical characteristics aresummarized in Table I. Among 45 patients, 29 were male (64.4%) and16 were female (35.6%). Their mean age was 57.3� 15.0 years.

According to the seventh edition of the American Joint Committee onCancer (AJCC) staging system, 10 patients had stage I (22.2%), 13 hadstage II (28.9%), 5 had stage III (11.1%), and 17 had stage IV (37.8%)disease. The size of all tumors included exceeded 1 cm; their mean sizewas 2.8 cm (n¼ 43, pathologic tumor size; n¼ 2, tumor size on MRI).

The SUVmax of the primary tumors ranged from 3.4 to 19.6 (mean:10.3� 4.8), and the mean HF of the primary tumors was �0.08� 0.10(range: �0.63 to �0.01). HF was significantly lower (i.e., moreheterogeneous) with advanced tumor (T) stage (P< 0.001) and AJCCstage (P¼ 0.007). A comparison of HF and lymph node (N) stagerevealed no statistical significance (P¼ 0.203).

According to the correlation between HF and metabolic PETparameters, the results were quite similar when the SUV threshold waschanged; therefore, we adopted fixed values of SUV at 3.0 for simplicity.HF was significantly correlated with SUVmax (r¼�0.353; P¼ 0.017),MTV (r¼�0.708; P< 0.0001), and TLG (r¼�0.709; P< 0.0001;Fig. 3).

A total of 33 patients (73.3%) were alive and 12 patients (26.7%) haddied at the last follow‐up. On univariate analysis, the presence of cervicallymph node metastasis (P¼ 0.013), SUVmax (P¼ 0.009), SUVavg

(P¼ 0.008),MTV (P< 0.001), TLG (P< 0.001), andHF (P¼ 0.001) ofthe primary tumor were significant predictors of overall survival(Table II). Multivariate analyses showed that cervical lymph nodemetastasis (hazard ratio: 5.983; 95% confidence interval: 1.288–27.790;P¼ 0.022) and HF (hazard ratio: 2.49� 10�4; 95% confidence interval:1.16� 10�6

–0.053; P¼ 0.002) were independent prognostic factors foroverall survival (Table III).

Those patients with lower HF values (more heterogeneous tumors)had a significantly lower overall survival than those with higher HFvalues (Fig. 4). A cut‐off HF value of �0.13 was determined by amaximally selected log‐rank test.

DISCUSSION

The present studymay be the first to show the prognostic significanceof HF of 18F‐FDG uptake in oral cavity cancer. The current studyindicates that the intratumoral heterogeneity of 18F‐FDG uptake is animportant prognostic factor for overall survival.

In the current study, 18F‐FDG PET/CT revealed that SUVmax,SUVavg, MTV, TLG, and HF of the primary tumor were significantprognostic factors in oral cavity cancer. SUVmax, a semiquantitativemeasure of tumor 18F‐FDG uptake, has been well established to be anindependent prognostic factor in HNC and other malignancies [5,20].However, SUVmax has several limitations because it reflects a singlevoxel value, which may not be representative of tumor character. Toovercome these limitations, volume‐based PET parameters have beeninvestigated. MTV is a volumetric measurement of tumor cells with highglycolytic activity, and TLG is the sum of SUVs within the tumor. Thus,these parameters could indicate the total volume and total activity ofmetabolically active tumor cells. One recent study suggested that TLGwas an independent prognostic factor in tonsil cancer, and another study

TABLE I. Patient Demographics and Clinical Characteristics

Variables Patients, n (%)

Age, mean (range), yr 57.3� 15.0 (27.2–82.1)Gender

Male/Female 29/16Histology

Squamous cell carcinoma 44 (97.8)Adenosquamous cell carcinoma 1 (2.2)

T stageT1 16 (35.6)T2 19 (42.2)T3 3 (6.7)T4 7 (15.5)

N stageN0 25 (55.6)N1 5 (11.1)N2 15 (33.3)N3 0 (0.0)

AJCC stageI 10 (22.2)II 13 (28.9)III 5 (11.1)IVA 17 (37.8)IVB 0 (0)

Treatment modalitySurgery alone 19 (42.2)Surgery þ adjuvant therapy 24 (53.3)RT alone 2 (4.5)

AJCC, American Joint Committee on Cancer; RT, radiotherapy.

Fig. 3. Correlations between heterogeneity factor and (A) SUVmax, (B) metabolic tumor volume, and (C) total lesion glycolysis.

Journal of Surgical Oncology

704 Kwon et al.

revealed MTV to be a significant survival predictor in clinically node‐negative oral cavity cancer patients [9,10]. Several investigations,including ours, have demonstrated that intratumoral heterogeneity wassignificantly associated with metabolic tumor volume [14,19]. Larger‐volume tumors tend to be more heterogeneous or greater heterogeneityleads to larger tumors. However, we found that some large tumors hadlower heterogeneity, whereas some smaller tumors had significantheterogeneity. In addition, heterogeneity varied within the same tumorstage. We think that heterogeneity may provide valuable information ontumor character.

Tumor tissues are heterogeneous at the microscopic level. Variationsin vascularity, hypoxia, proliferation rates, energy metabolites,necrosis, and gene expression might make a tumor moreheterogeneous. Using more refined molecular and genetic probes,tumor cells showed diversity in origin, clonality, and the expression andregulation of the genome and proteomic pathways [18]. Theintratumoral heterogeneity of 18F‐FDG uptake is related to theheterogeneity of the histopathologic characteristics of non‐small celllung cancer, squamous cell carcinoma of the head and neck,oligodendroglioma, and uterine cervical cancer [15–19]. Theunderlying mechanism of intratumoral heterogeneity remainsunclear, but Zhao et al. [21] reported that the expression levels ofglucose transporter (Glut)‐1, Glut‐3, and hexokinase‐II may beincreased in hypoxic tumor tissues and contribute to heterogeneous18F‐FDG uptake. In other research, not only increased glucosetransporter and hexokinase expression but also decreased expressionof glucose 6‐phosphatase by immunohistochemistry was shown to beinvolved in heterogeneous glucose uptake in HNC [22]. Anotherinvestigation proposed that 18F‐FDG uptake is correlated with bloodflow, hypoxia, and cellular proliferation [23]. Additionally, high lactatelevels have been demonstrated to correlate with worse survival inHNC [24].

Several methods exist to assess intratumoral heterogeneity using18F‐FDG PET/CT. Several studies have calculated HF as the derivative(dV/dT) of the volume‐threshold function for each tumor [14,19]. Somestudies have used the coefficient of variation as the ratio between the

standard deviation of the SUVs and SUVmean within the delineatedmetabolically active tumor volume to estimate 18F‐FDG uptakeheterogeneity [25,26]. Because tumor heterogeneity is recorded atvarious intensities on PET imaging, textural analysis could also beapplied. Textural analysis involves a group of computational methodsfor evaluating the position and intensity in digital images and extractinginformation about the relationship between adjacent pixels [27,28].However, no standard method exists for the global measurement ofintratumoral heterogeneity on PET/CT. Thus, we analyzed 18F‐FDGheterogeneity using dV/dT due to its simplicity.

Although pretreatment 18F‐FDG PET/CT is not always performed inall medical centers, the number of cases of HNC has been increasingrapidly. Furthermore, there is increasing evidence supporting theprognostic significance of PET parameters in HNC. Therefore, 18F‐FDGPET/CT can provide important information concerning tumorcharacteristics and it may help clinicians to select risk‐basedtreatment strategies and to stratify high‐risk patients.

The current study has several limitations. First, it was a retrospective,single‐center study. Second, several different treatment modalities wereused in our study. Finally, although we evaluated the primary tumorlarger than 1 cm, 16 of 45 patients (35.6%) had T1 stage disease. PEThas the limitation of spatial resolution, particularly in small‐sizedlesions, possibly leading to the underestimation of volumetric PET

TABLE II. Univariate Analysis for Overall Survival Using cox Proportional‐Hazard Model

Variables Category HR 95% CI P value

Age (1‐yr increase) 1.020 0.979–1.063 0.351Gender Male 1.850 0.500–6.840 0.357Treatment modality Surgery alone vs. surgery þ adjuvant Tx. or nonsurgical Tx 6.943 0.894–53.940 0.064T stage T1 vs. T2‐4 46.880 0.394–5577.134 0.115N stage negative vs. positive 6.809 1.491–31.091 0.013�

AJCC stage I vs. II‐IV 29.497 0.079–11021.887 0.263SUVmax (1‐unit increase) 1.182 1.040–1.340 0.009�

SUVavg (1‐unit increase) 2.138 1.224–3.735 0.008�

MTV (1‐cm3 increase) 1.066 1.029–1.104 <0.001�

TLG (1‐unit increase) 1.010 1.004–1.015 <0.001�

HF (0.01‐unit increase) 1.15� 10�4 5.51� 10�7–0.024 0.001�

AJCC, American Joint Committee on Cancer; MTV, metabolic tumor volume; TLG, total lesion glycolysis; HF, heterogeneity factor.

TABLE III. Multivariate Analysis for Overall Survival Using CoxProportional‐Hazard Model

Variables HR 95% CI P value

N stage (negative vs. positive) 5.983 1.288–27.790 0.022HF (0.01‐unit increase) 2.49� 10�4 1.16� 10�6

–0.053 0.002

HF, heterogeneity factor.

Fig. 4. Kaplan‐Meier curve of overall survival according toheterogeneity factor. Graph shows comparisons of survival betweenpatients with higher (less heterogeneous tumors) and lower (moreheterogeneous tumors) heterogeneity factors (P¼ 0.005).

Journal of Surgical Oncology

Intratumoral Heterogeneity of FDG in Oral Cavity Cancer 705

parameters because of the partial volume effect. Therefore, prospective,multicenter validation studies are mandatory with larger numbers ofpatients with more advanced disease status and a unified treatmentmodality.

CONCLUSIONS

The intratumoral heterogeneity of 18F‐FDG uptake represented byHFmay be a significant predictor of overall survival in oral cavity cancerpatients, in addition to cervical lymph node metastasis. Additionalmulticenter, large‐scale, prospective studies are needed to validatethe prognostic usefulness of this functional biomarker derived from18F‐FDG PET.

REFERENCES

1. Bello IO, Soini Y, Salo T: Prognostic evaluation of oral tonguecancer: means, markers and perspectives (I). Oral oncology2010;46:630–635.

2. Warnakulasuriya S: Global epidemiology of oral and oropharyngealcancer. Oral oncology 2009;45:309–316.

3. Yi JS, Kim JS, Lee JH, et al.: 18F‐FDG PET/CT for detecting distantmetastases in patients with recurrent head and neck squamous cellcarcinoma. J Surg Oncol 2012;106:708–712.

4. Abgral R, Querellou S, Potard G, et al.: Does 18F‐FDG PET/CTimprove the detection of posttreatment recurrence of head and necksquamous cell carcinoma in patients negative for disease on clinicalfollow‐up? J Nucl Med 2009;50:24–29.

5. Ong SC, Schöder H, Lee NY, et al.: Clinical utility of 18F‐FDGPET/CT in assessing the neck after concurrent chemoradiotherapy forlocoregional advanced head and neck cancer. J Nucl Med2008;49:532–540.

6. Murthy V, Agarwal JP, Laskar SG, et al.: Analysis of prognosticfactors in 1180 patients with oral cavity primary cancer treated withdefinitive or adjuvant radiotherapy. J Cancer Res Ther 2010;6:282–289.

7. Kim SY, Roh JL, KimMR, et al.: Use of 18F‐FDG PET for primarytreatment strategy in patients with squamous cell carcinoma of theoropharynx. J Nucl Med 2007;48:752–757.

8. Higgins KA, Hoang JK, Roach MC, et al.: Analysis of pretreatmentFDG‐PET SUV parameters in head‐and‐neck Cancer: TumorSUVmean has superior prognostic value. Int JRadiat Oncol Biol Phys2012;82:548–553.

9. Lee SJ, Choi JY, Lee HJ, et al.: Prognostic value of volume‐based18F‐fluorodeoxyglucose PET/CT parameters in patients withclinically node‐negative oral tongue squamous cell carcinoma.Korean J Radiol 2012;13:752–759.

10. Moon SH, Choi JY, Lee HJ, et al.: Prognostic value of 18F‐FDGPET/CT in patients with squamous cell carcinoma of the tonsil:Comparisons of volume‐based metabolic parameters. Head Neck2013;35:15–22.

11. Choi KH, Yoo IR, Han EJ, et al.: Prognostic value of metabolictumor volume measured by 18F‐FDG PET/CT in locally advancedhead and neck squamous cell carcinomas treated by surgery. NuclMed Mol Imaging 2011;45:43–51.

12. Abd El‐Hafez YG, Moustafa HM, Khalil HF, et al.: Total lesionglycolysis: a possible new prognostic parameter in oral cavitysquamous cell carcinoma. Oral oncol 2012;49:261–268.

13. Ryu IS, Kim JS, Roh JL, et al.: Prognostic significance ofpreoperative metabolic tumour volume and total lesion glycolysismeasured by 18F‐FDG PET/CT in squamous cell carcinoma of theoral cavity. Eur J Nucl Med Mol Imaging 2014;41:452–461.

14. Huang B, Chan T, Kwong DL, et al.: Nasopharyngeal carcinoma:investigation of intratumoral heterogeneity with FDGPET/CT. AJRAm J Roentgenol 2012;199:169–174.

15. Derlon J, Chapon F, Noel M, et al.: Non‐invasive grading ofoligodendrogliomas: correlation between in vivo metabolic patternand histopathology. Eur J Nucl Med 2000;27:778–787.

16. Henriksson E, Kjellen E, Wahlberg P, et al.: 2‐Deoxy‐2‐[18F]fluoro‐D‐glucose uptake and correlation to intratumoral heteroge-neity. Anticancer Res 2007;27:2155–2159.

17. van Baardwijk A, Bosmans G, van Suylen RJ, et al.: Correlation ofintra‐tumour heterogeneity on 18F‐FDG PET with pathologicfeatures in non‐small cell lung cancer: a feasibility study. RadiotherOncol 2008;87:55–58.

18. Eary JF, O’Sullivan F, O’Sullivan J, et al.: Spatial heterogeneity insarcoma 18F‐FDG uptake as a predictor of patient outcome. J NuclMed 2008;49:1973–1979.

19. Kidd EA, Grigsby PW: Intratumoral metabolic heterogeneity ofcervical cancer. Clin Cancer Res 2008;14:5236–5241.

20. Sasaki R, Komaki R, Macapinlac H, et al.: [18F] fluorodeoxyglu-cose uptake by positron emission tomography predicts outcome ofnon–small‐cell lung cancer. J Clin Oncol 2005;23:1136–1143.

21. Zhao S, Kuge Y, Mochizuki T, et al.: Biologic correlates ofintratumoral heterogeneity in 18F‐FDG distribution with regionalexpression of glucose transporters and hexokinase‐II in experimen-tal tumor. J Nucl Med 2005;46:675–682.

22. Zhou S, Wang S, Wu Q, et al.: Expression of glucose transporter‐1and‐3 in the head and neck carcinoma–the correlation of theexpression with the biological behaviors. ORL J OtorhinolaryngolRelat Spec 2008;70:189–194.

23. Pugachev A, Ruan S, Carlin S, et al.: Dependence of FDG uptake ontumor microenvironment. Int J Radiat Oncol Biol Phys 2005;62:545–553.

24. Brizel DM, Schroeder T, Scher RL, et al.: Elevated tumor lactateconcentrations predict for an increased risk of metastases in head‐and‐neck cancer. Int J Radiat Oncol Biol Phys 2001;51:349–353.

25. Hatt M, Cheze‐le Rest C, Van Baardwijk A, et al.: Impact of tumorsize and tracer uptake heterogeneity in 18F‐FDG PET and CT non–small cell lung cancer tumor delineation. J Nucl Med 2011;52:1690–1697.

26. Frings V, de Langen AJ, Smit EF, et al.: Repeatability ofmetabolically active volume measurements with 18F‐FDG and18F‐FLT PET in non–small cell lung cancer. J Nucl Med 2010;51:1870–1877.

27. Rahim MK, Kim SE, So H, et al.: Recent trends in PET imageinterpretations using volumetric and texture‐based quantificationmethods in nuclear oncology. NuclMedMol Imaging 2014;48:1–15.

28. Chicklore S, Goh V, Siddique M, et al.: Quantifying tumourheterogeneity in 18F‐FDG PET/CT imaging by texture analysis. EurJ Nucl Med Mol Imaging 2013;40:133–140.

Journal of Surgical Oncology

706 Kwon et al.