European Journal of Nuclear Medicine and Molecular Imaging

, Volume 41, Issue 3, pp 452–461

Prognostic significance of preoperative metabolic tumour volume and total lesion glycolysis measured by 18F-FDG PET/CT in squamous cell carcinoma of the oral cavity

Authors

  • In Sun Ryu
    • Department of Otolaryngology, Asan Medical CenterUniversity of Ulsan College of Medicine
  • Jae Seung Kim
    • Department of Nuclear Medicine, Asan Medical CenterUniversity of Ulsan College of Medicine
    • Department of Otolaryngology, Asan Medical CenterUniversity of Ulsan College of Medicine
  • Kyung-Ja Cho
    • Department of Pathology, Asan Medical CenterUniversity of Ulsan College of Medicine
  • Seung-Ho Choi
    • Department of Otolaryngology, Asan Medical CenterUniversity of Ulsan College of Medicine
  • Soon Yuhl Nam
    • Department of Otolaryngology, Asan Medical CenterUniversity of Ulsan College of Medicine
  • Sang Yoon Kim
    • Department of Otolaryngology, Asan Medical CenterUniversity of Ulsan College of Medicine
    • Biomedical Research Institute, Korean Institute of Science and Technology
Original Article

DOI: 10.1007/s00259-013-2571-z

Cite this article as:
Ryu, I.S., Kim, J.S., Roh, J. et al. Eur J Nucl Med Mol Imaging (2014) 41: 452. doi:10.1007/s00259-013-2571-z

Abstract

Purpose

Metabolic tumour volume (MTV) and total lesion glycolysis (TLG) from 18F-FDG PET/CT are emerging prognostic biomarkers in human solid cancers; yet few studies have investigated their clinical and prognostic significance in oral cavity squamous cell carcinoma (OSCC). The present retrospective study evaluated the utility of pretreatment MTV and TLG measured by 18F-FDG PET/CT to predict survival and occult metastasis (OM) in OSCC.

Methods

Of 162 patients with OSCC evaluated preoperatively by 18F-FDG PET/CT, 105 who underwent definitive surgery with or without adjuvant therapy were eligible. Maximum standardized uptake value (SUVmax), MTV and TLG were measured. For calculation of MTV, 3-D regions of interest were drawn and a SUV threshold of 2.5 was used for defining regions. Univariate and multivariate analyses identified clinicopathological and imaging variables associated with OM, disease-free survival (DFS) and overall survival (OS).

Results

The median (range) SUVmax, MTV and TLG were 7.3 (0.7–41.9), 4.5 ml (0.7–115.1 ml) and 18.3 g (2.4–224.1 g), respectively. Of 53 patients with clinically negative lymph nodes, OM was detected in 19 (36 %). By univariate and multivariate analyses, MTV (P = 0.018) and TLG (P = 0.011) were both independent predictive factors for OM, although they were not independent of each other. The 4-year DFS and OS rates were 53.0 % and 62.0 %, respectively. Univariate and multivariate analyses revealed that MTV (P = 0.001) and TLG (P = 0.006), with different cut-off levels, were both independent predictive factors for DFS, although they were not independent of each other, and MTV (P = 0.001), TLG (P = 0.002) and the involved resection margin (P = 0.007) were independent predictive factors for OS.

Conclusion

Pretreatment MTV and TLG may be useful in stratifying the likelihood of survival and predicting OM in OSCC.

Keywords

Oral cavity cancer18F-FDG PET/CTMetabolic tumour volumeTotal lesion glycolysisPrognosis

Introduction

Approximately 30 % of head and neck cancers arise in the oral cavity, with the major pathology being squamous cell carcinoma (SCC). In the US, an estimated 27,450 patients may be newly diagnosed with oral cavity cancer in 2013, and an estimated 5,510 patients die from the disease [1]. Various clinicopathological and biological markers for predicting outcome in oral cavity SCC (OSCC) have been identified [25]. Nonetheless, the survival rate of OSCC has remained substantially unchanged over the past several decades [6]. Pretreatment selection of patients with poor prognosis or who require intensified therapy remains difficult, despite the use of prognostic parameters including clinicopathological factors, anatomical subsite and TNM stage. The identification of novel biomarkers to predict long-term outcome in OSCC is urgent [7].

18F-FDG PET is based on tumour glucose metabolism and serves as a marker of tumour metabolic activity in terms of cell viability and proliferation [8, 9]. Combining 18F-FDG PET with CT provides useful functional and anatomical information, and enables more accurate initial staging, treatment response evaluation and posttreatment surveillance in head and neck cancers [1013]. The maximum standardized uptake value (SUVmax), as an estimate of tumour metabolic activity, is the most commonly used parameter in 18F-FDG PET/CT; however, since it is measured as the maximum pixel count in a region of interest (ROI), SUVmax shows only the highest intensity of 18F-FDG uptake in the tumour and cannot reflect the metabolic activity of the whole tumour [14].

Recently, 18F-FDG metabolic tumour volume (MTV) and total lesion glycolysis (TLG), combining the tumour volume and metabolic activity of the entire tumour, have been introduced as prognostic biomarkers in various solid malignancies [1517]; however, criteria for these volumetric parameters have not yet been established. In particular, the question of which is the better parameter to predict outcome remains unresolved. To our knowledge, few studies have investigated the clinical and prognostic significance of MTV and TLG in OSCC [18, 19]. Therefore, we evaluated the utility of theses metabolic parameters, measured by pretreatment 18F-FDG PET/CT, for predicting survival outcome and occult metastasis (OM) in OSCC.

Materials and methods

Patient population

We reviewed the records of 162 patients with newly diagnosed OSCC who underwent preoperative 18F-FDG PET/CT imaging for initial staging between January 2004 and August 2011. Prior to surgery, all patients were also assessed by physical examination and imaging work-up, including CT and/or MRI of the head and neck. Exclusion criteria were clinical stage I OSCC (44 patients) to minimize the partial volume effect of the primary tumour volume [20], and previous history of head and neck cancer (9 patients). Patients with incomplete surgical operation or with incomplete clinical data (4 patients) were also excluded. All surviving patients were followed for at least 12 months. Thus, 105 patients were eligible for the study.

Data were obtained from medical records, including clinicopathological variables, treatment and follow-up. Follow-up data were available in all patients, with recurrence and survival calculated from the date of initial surgery. Tumours were staged according to the American Joint Committee on Cancer (AJCC) staging system [21]. The study was approved by the Institutional Review Board of Asan Medical Center and the requirement informed consent from each patient was waived.

Treatments

All patients initially underwent wide resection of the primary tumour and neck dissection with curative intent. Standardized modified radical neck dissection involving levels I–V as described by the American Head and Neck Society [22] was performed in 52 patients with clinical nodal metastasis, and bilateral dissection in 22. Bilateral neck dissection was indicated in patients with primary tumours involving the middle line and in those suspected of having metastatic nodes in the contralateral neck. Selective neck dissection involving levels I–III or I–IV was performed in 53 patients without clinical nodal metastasis (cN0). cN0 was defined as no evidence of cervical lymph node metastasis by physical examination, CT and/or MRI, and PET/CT. OM was defined as tumour deposits that were initially undetected by preoperative evaluation and subsequently identified in postoperative pathology. Postoperative chemoradiation therapy or radiotherapy was performed in 70 patients with a median dose of 59.4 Gy (range 40.3–75.4 Gy) in single daily fractions of 1.8–2.0 Gy. Indications for postoperative adjuvant therapy relied on postoperative pathological features such as positive surgical margins, advanced T3/T4 classification, N2 or N3 nodal disease, extracapsular nodal spread, lymphovascular invasion and perineural invasion [2325].

18F-FDG PET/CT imaging

18F-FDG PET/CT scans were obtained with a multislice PET/CT camera system together with a Biograph Sensation 16 (Siemens Medical System, Knoxville, TN), which provided an axial spatial resolution of 6.3 mm in full-width at half-maximum (FWHM), equipped with a 16-slice CT scanner. All PET/CT scans were measured using the same device and using a consistent method, including accumulation time, scanning mode and reconstruction parameters. All patients fasted for at least 6 h before the 18F-FDG PET scan. Their blood glucose concentrations were <150 mg/dl before the scan. Whole-body images were obtained 50–70 min after intravenous injection of 370–688 MBq (10.0–18.6 mCi) of 18F-FDG. CT scanning was performed in spiral mode from the skull base to the proximal thigh at 100 mAs and 120 kV, with a section width of 5 mm and collimation of 0.75 mm. No oral or intravenous contrast medium was used. CT scanning data were obtained for attenuation correction and image fusion, and were followed by a three-dimensional craniocaudal PET emission scan. The emission scans were obtained for the whole-body from head to proximal thigh with the arms up, followed by dedicated images of the head and neck with the arms down. The acquisition times were 2.5 min per bed position using six or seven bed positions for the whole body and 5 min per bed position using two bed positions for the head and neck. The PET data were reconstructed with CT attenuation correction using an attenuation-weighted ordered subsets expectation maximization algorithm (two iterations, 16 subsets) followed by postreconstruction smoothing with a gaussian filter (FWHM, 6 mm), resulting in an image resolution of about 6.3 mm FWHM.

Image interpretation

18F-FDG PET/CT findings were reviewed on the workstation by a board-certified nuclear medicine physician (J.S.K) with more than 15 years of clinical experience in head and neck imaging who identified visible lesions with high tracer uptake and then quantified the 18F-FDG uptake. SUV was used to determine 18F-FDG-PET activity. The SUV was determined using the equation: SUV = A/(ID/BW), where A is the decay-corrected activity in tissue (in millicuries per millilitre), ID is the injected dose of FDG (in millicuries), and BW is the patient body weight (in grams). Spherical or ellipsoidal ROIs were placed over the lesions visible on PET images. The ROIs of lesions that were invisible on PET images were located using the corresponding CT images. The SUVmax was calculated by automatically drawing an ROI over the most intense slice of the primary tumour and the metastatic lymph node visible on PET images.

MTVs were computed from attenuation-corrected PET data using a commercial software package (INFINITT PACS; INFINITT Healthcare Co. Ltd, Seoul, South Korea). 18F-FDG PET data were fed into the workstation in DICOM format and intensity values were automatically converted to SUVs. For MTV calculations, the contouring margins of the tumour were defined using a fixed SUV of 2.5, which was based on the results of previous studies [15, 16, 26]. Using a graphical user interface, ROIs were drawn to enclose three-dimensional coverage of the entire hypermetabolic tumour lesion on all three projections (axial, sagittal and coronal). The tumour volume was then delineated with the SUV 2.5 isocontour. Lymph nodes smaller than 1 cm in diameter were generally not included unless they showed abnormal uptake of 18F-FDG on the corresponding PET images. The software program automatically calculated the volume of the tumour in each ROI. The metabolic volume of the primary tumour and metastatic cervical lymph nodes was then found (Fig. 1), and summed as the total tumour volume. The SUVmean was automatically calculated as the average SUV of the tumour in each ROI. The TLG was defined as MTV × SUVmean [27], and calculated using a summation including both the primary tumour and all metastatic lymph nodes.
https://static-content.springer.com/image/art%3A10.1007%2Fs00259-013-2571-z/MediaObjects/259_2013_2571_Fig1_HTML.gif
Fig. 1

Examples of MTV measurement in two contrasting patients (a and b). The ROIs (circles) were set to include the whole tumour or metastatic disease with metabolic activity, and the software automatically calculated the SUVmax and the MTV. The SUVmax in patients a and b are almost identical (13.93 and 13.80) but their MTVs are very different (11.4 and 6.8 ml). Patient a experienced recurrence at 9 months while at the time of this report patient b had been disease-free for 38 months from treatment

The enhanced CT images and/or MR images were interpreted by a board-certified radiologist with 12 years clinical experience in head and neck radiology. The cervical lymph nodes were considered metastatic if central necrosis, spherical shape, loss of hilar structure or high contrast enhancement was present, if their shortest axial diameter was >1.2 cm in level IB and II or >1 cm in other levels, or if there was a cluster of three or more lymph nodes of borderline size [28, 29].

Statistical analysis

Continuous variables are expressed as medians and ranges, and categorical variables as numbers and percentages. Univariate analysis was used to identify clinical factors and imaging parameters predictive of OM by Fisher’s exact test or the chi-squared test. Multivariate logistic regression analysis was performed to determine statistically significant risk factors for OM. The primary endpoint was to establish whether the exploratory imaging markers, MTV and TLG, provided prognostic information about disease-free survival (DFS) and overall survival (OS). Time periods were calculated from the date of surgery either to the date of a confirmed event (death, any recurrence) or to the last clinical follow-up. Recurrence after achieving locoregional control by curative surgery was defined as the existence of histologically confirmed recurrent tumours at local, regional and/or distant sites. The DFS and OS curves were calculated using the Kaplan-Meier method. The log-rank test was used to compare survival rates according to imaging parameters.

The Cox proportional hazards model was used to evaluate prognostic variables for univariate and multivariate prediction of DFS and OS; tests were based on the likelihood ratio statistic and the estimated hazard ratio (HR) and 95 % confidence intervals (CIs) were calculated. The following variables were considered as potential predictors of DFS and OS on univariate analysis: age, sex, smoking, alcohol drinking, primary tumour site, overall pathological TNM (pTNM) stage, histological grade, resection margin, lymphovascular invasion, perineural invasion, SUVmax for the primary tumour,, MTV and TLG. Variables with P < 0.05 in the univariate analyses were selected for the multivariate analyses. Multicollinearity between MTV and TLG was established using the Pearson correlation coefficient. Receiver operating characteristic (ROC) curve analysis was used to determine areas under the curve (AUC) to estimate the accuracy and predictive value of various imaging biomarkers. Two different cut-off values for MTV and TLG were determined by cross-validated predicted probabilities. The cut-off value for predicting OM was obtained from 53 patients with cN0 and those for predicting DFS and OS were obtained from all 105 patients. The presence of OM and a confirmed event were also used as references, respectively. All tests were two-sided, and P < 0.05 was considered statistically significant. All statistical analyses were performed using IBM SPSS software version 20.0 (IBM, Armonk, New York, NY).

Results

Patient characteristics

The 105 eligible patients consisted of 58 men and 47 women with a median age 57 years (range 22–87 years). Their demographic and clinicopathological characteristics are presented in Table 1. The tongue was the most common site of primary tumour, observed in 78 patients (74 %), and poorly differentiated tumours were observed in 10 patients (9 %). Of 53 patients with cN0, OM was histologically detected in 19 (36 %). Of the 105 patients, 42 (40 %) had advanced T-stage tumours, 65 (62 %) had pathologically positive cervical lymph nodes, and 74 (70 %) were in advanced overall stage III/IV. The median follow-up for surviving patients was 44 months (range 13–104 months). During follow-up after definitive treatment, there were 40 local and/or regional recurrences, 12 distant metastases of which 5 were in patients who also had local and/or regional recurrence, and 35 patients (33 %) died of their disease. Only two patients died from other causes. The Kaplan-Meier estimates of 4-year DFS and OS rates for all patients were 53.0 % and 62.0 %, respectively.
Table 1

Patient characteristics

Variables

All patients (n = 105)

Patients with cN0a (n = 53)

Age (years), median (range)

57 (22–87)

53 (22–87)

Gender, n (%)

 Male

58 (55)

33(62)

 Female

47 (45)

20 (38)

Site of primary tumour, n (%)

 Tongue

78 (74)

42 (79)

 Gingiva

7 (7)

3 (6)

 Floor of mouth

12 (11)

5 (9)

 Retromolar trigone

7 (7)

0 (0)

 Buccal mucosa

1 (1)

3 (6)

Size of primary tumour (cm), median (range)

2.5 (0.5–6.0)

2.5 (0.5–5.0)

Smoking, n (%)

45 (43)

23 (43)

Alcohol drinking, n (%)

54 (51)

30 (57)

Histological differentiation, n

 Well

57

29

 Moderately

38

19

 Poorly

10

5

Pathological TNM stage, na

 T1/T2/T3/T4

23/40/3/39

9/28/1/15

 N0/N1/N2

40/29/36

34/10/9

 Stage I/II/III/IV

10/21/20/54

8/19/7/19

Treatment

 Surgery alone

35 (33)

25 (47)

 Surgery + radiotherapy/chemoradiotherapy

70 (67)

28 (53)

aTumour-node-metastasis stage (6th edn) developed by the American Joint Committee on Cancer

bcN0 was defined as no evidence of cervical lymph node metastasis by physical examination, CT and/or MRI, and PET/CT

Determination of the cut-off values of imaging parameters

The median (range) SUVmax, MTV and TLG were 7.3 (2.56–33.8), 4.5 ml (0.7–115.1 ml) and 18.3 g (2.4–224.1 g). From the ROC curve analysis, the SUVmax, MTV and TLG cut-off values for patients with OM were 8.1, 3.0 ml and 13.9 g, respectively, in cN0 patients (Fig. 2a, b). The SUVmax, MTV and TLG cut-off values for evaluating prognostic values were 11.5, 8.1 ml and 52.0 g, respectively, in all patients (Fig. 2c, d). To validate the selected cut-off point for the ROC curves, we also performed dichotomization analysis of patient survival with various cut-off points including mean and median values. The selected cut-off points for the ROC curves were associated with the lowest P value (P < 0.05). Thus, these were taken as the optimal cut-off points.
https://static-content.springer.com/image/art%3A10.1007%2Fs00259-013-2571-z/MediaObjects/259_2013_2571_Fig2_HTML.gif
Fig. 2

ROC curves for patients with OM (a, b) and for those with a confirmed event (c, d) according to MTV and TLG. a AUC 0.759 (P < 0.001, 95 % CI 0.649–0.869), cut-off value 3.0 for MTV; b AUC 0.747 (P = 0.015, 95 % CI 0.628–0.864), cut-off value 13.9 for TLG; c AUC 0.753 (P < 0.001, 95 % CI 0.654–0.852), cut-off value 8.1 for MTV; d AUC 0.678 (P = 0.004, 95 % CI 0.563–0.792), cut-off value 52.0 for TLG

Prediction of occult metastasis in cN0

Of 53 patients without clinically positive nodes, 19 (36 %) had OM. Clinical factors predictive of OM are presented in Table 2. MTV >3.0 ml and TLG >13.9 g were significantly associated with high OM rates (P < 0.05); however, SUVmax and clinical T classification were not predictive of OM. As expected, there was a significant correlation between MTV and TLG (calculated by multiplying MTV by SUVmean; r = 0.845, P < 0.001). Therefore, two different models including MTV or TLG separately were used for multivariate analysis. In multivariate analysis, MTV (P = 0.018, OR = 4.75, 95 % CI 1.30–17.33) and TLG (P = 0.011, OR = 5.36, 95 % CI 1.46–19.60) were the only independent predictive factors for OM.
Table 2

Univariate and multivariate analyses to identify preoperative variables predictive of OM in clinically lymph node-negative patients with OSCC

Variable

Occult metastasis, n (%)

P valuea

No (n = 34)

Yes (n = 19)

Age at diagnosis (years)

 ≤60

23 (68)

13 (68)

0. 999

 >60

11 (32)

6 (32)

Sex

 Male

20 (59)

13 (68)

0.564

 Female

14 (41)

6 (32)

Smoking

 No

19 (56)

11 (58)

0.887

 Yes

15 (44)

8 (42)

Alcohol drinking

 No

17 (50)

6 (32)

0.253

 Yes

17 (50)

13 (68)

Histological differentiation

 Well/moderately

31 (91)

17 (90)

0.839

 Poorly

3 (9)

2 (10)

cT classification

 cT2/3

27 (79)

11 (58)

0.120

 cT4

7 (21)

8 (42)

SUVmax

 ≤8.1

22 (65)

10 (53)

0.559

 >8.1

12 (35)

9 (48)

MTV (ml)

 ≤3.0

19 (56)

4 (21)

0.021b

 >3.0

15 (44)

15 (79)

TLG (g)

 ≤13.9

20 (59)

4 (21)

0.010b

 >13.9

14 (41)

15 (79)

aAssessed using the two-sided Fisher’s exact test or the χ2 test

bAssessed by a multivariate logistic regression model. Because of multicollinearity (r = 0.845) between MTV and TLG, TLG and MTV were not included in the same multivariate model: MTV > 3.0 ml (P = 0.018, OR = 4.75, 95 % CI 1.30–17.33), TLG >13.9 g (P = 0.011, OR = 5.36, 95 % CI 1.46–19.60). Variables with P < 0.2 in the univariate analyses were selected for the multivariate analysis

Univariate and multivariate analyses

Univariate analysis showed that histological differentiation, MTV and TLG were significantly associated with decreased DFS (P < 0.05) and that resection margin, MTV and TLG were significantly associated with decreased OS (P < 0.05; Table 3). pTNM stage was not significantly associated with decreased DFS or decreased OS, although regression analysis revealed a significant association between pTNM stage and increased MTV (P = 0.005) and TLG (P = 0.002). Patients with MTV >8.1 ml had significantly lower DFS rates (24.0 % vs. 64.7 %, P = 0.001) and OS rates (48.1 % vs. 70.5 %, P = 0.013) than those with MTV ≤8.1 ml (Fig. 3a, c). Patients with TLG >52.0 g also had significantly lower DFS rates (23.1 % vs. 61.1 %, P = 0.004) and OS rates (34.1 % vs. 69.1 %, P = 0.002) than those with TLG ≤52.0 g (Fig. 3b, d); however, SUVmax was not significantly associated with either DFS (P = 0.206) or OS (P = 0.080).
Table 3

Univariate analysis of clinicopathological and imaging variables in relation to DFS and OS

Variable

DFS

OS

HR (95 % CI)

Pa

HR (95 % CI)

Pa

Sex

 Male

1

 

1

 

 Female

1.045 (0.59–1.85)

0.880

1.085 (0.57–2.07)

0.806

Age at diagnosis (years)

 ≤60

1

 

1

 

 >60

0.89 (0.49–1.61)

0.688

1.20 (0.63–2.31)

0.579

Smoking

 Never or stopped

1

 

1

 

 Current user

2.48 (0.23–0.32)

0.442

1.03 (0.54–1.97)

0.923

Alcohol drinking

 Never or stopped

1

 

1

 

 Current user

1.06 (0.52–1.66)

0.936

1.05 (0.55–2.01)

0.873

Site of primary tumourb

 Tongue

1

 

1

 

 Other subsites

1.26 (0.67–2.36)

0.469

1.36 (0.68–2.71)

0.380

Histological differentiation

 Well/moderately

1

 

1

 

 Poorly

2.21 (0.99–4.96)

0.049

1.36 (0.48–3.84)

0.562

Stage

 I

1

 

1

 

 II

1.01 (0.10–1.62)

0.202

1.45 (0.16–12.99)

0.740

 III

1.60 (0.51–5.03)

0.420

3.39 (0.42–27.60)

0.253

 IV

1.66 (0.58–4.74)

0.343

5.21(0.71–38.46)

0.106

Resection margin

 Not involved

1

 

1

 

 Involved

1.54 (0.76–3.10)

0.226

2.51 (1.24–5.10)

0.011

Lymphovascular invasion

 No

1

 

1

 

 Yes

1.31(0.52–3.32)

0.567

1.53(0.54–4.33)

0.424

Perineural invasion

 No

1

 

1

 

 Yes

1.08 (0.41–1.78)

0.688

1.19 (0.56–2.53)

0.644

SUVmax

 ≤11.5

1

 

1

 

 >11.5

1.52 (0.79–2.95)

0.206

1.92 (0. 92–3.96)

0.080

MTV (ml)c

 ≤8.1

1

 

1

 

 >8.1

2.81 (1.53–5.18)

0.001

2.64 (1.35–5.21)

0.005

TLG (g)

 ≤52.0

1

 

1

 

 >52.0

2.97 (1.38–6.41)

0.006

3.30 (1.50–7.24)

0.003

aCox proportional hazards model, P < 0.05

bIncluding gingiva, floor of mouth, retromolar trigone, and buccal mucosa

cThe metabolic volume of primary tumour and metastatic cervical lymph nodes was summed as total MTV

https://static-content.springer.com/image/art%3A10.1007%2Fs00259-013-2571-z/MediaObjects/259_2013_2571_Fig3_HTML.gif
Fig. 3

Kaplan-Meier curves of DFS (a, b) and OS (c, d) according to MTV and TLG in the study population (n = 105). a, cUpper line MTV ≤8.1 ml (n = 81); lower line MTV >8.1 ml (n = 24); P = 0.001 for DFS and P = 0.013 for OS. b, dUpper line TLG ≤52.0 g (n = 94); lower line TLG >52.0 g (n = 11); P = 0.004 for DFS and P = 0.002 for OS

Since a significant correlation was observed between MTV and TLG, two different models including MTV and TLG separately were used for multivariate analyses. MTV (P = 0.001, HR = 2.81, 95 % CI 1.53–5.18) and TLG (P = 0.006, HR = 2.97, 95 % CI 1.37–6.41) were the only independent prognostic factors for DFS, and MTV (P = 0.001, HR = 3.07, 95 % CI 1.54–6.13), TLG (P = 0.002, HR = 3.50, 95 % CI 1.59–7.70) and involved resection margin (P = 0.007, HR = 2.64, 95 % CI 1.29–5.37) were independent prognostic factors for OS (Table 4).
Table 4

Multivariate analyses of clinicopathological and imaging variables in relation to DFS and OS

Variables

DFS

OS

Model Aa

Model Ba

Model Aa

Model Ba

HR (95 % CI)

P valueb

HR (95 % CI)

P valueb

HR (95 % CI)

P valueb

HR (95 % CI)

P valueb

Histological differentiation

 Well/moderately vs. poorly

1.63 (0.70–3.75)

0.252

2.09 (0.93–4.70)

0.074

 

 

Resection margin

 Not involved vs. involved

 

 

2.97 (1.44–6.13)

0.003

2.64 (1.29–5.37)

0.007

MTVb

 ≤8.1 vs. >8.1 ml

2.81 (1.53–5.18)

0.001

 

3.07 (1.54–6.13)

0.001

 

TLG

 ≤52.0 vs. >52.0 g

 

2.97 (1.37–6.41)

0.006

 

3.50 (1.59–7.70)

0.002

aBecause of multicollinearity (r = 0.845) between MTV and TLG, TLG and MTV were not included in models A and B, respectively

bMultivariate analysis using the Cox proportional hazards model. Variables with P < 0.05 in the univariate analyses were selected for the multivariate analysis

Discussion

This study examined the utility of 18F-FDG metabolic parameters for predicting clinical outcomes in OSCC. The most significant result was that MTV and TLG are potential predictors of OM and prognostic factors of survival in OSCC treated with curative surgery with or without adjuvant therapy. We focused on 18F-FDG metabolic parameters that could help identify patients at high risk of OM and recurrence. Current evidence suggests that metabolic tumour burden may be a more important predictor of recurrence than clinicopathological parameters previously known in OSCC.

Various 18F-FDG PET parameters have been investigated in solid tumours [12, 1517]. MTV, which is a measure of the volume of the tumour displaying 18F-FDG uptake and quantifies the overall tumour burden, is theoretically a better predictor of outcome than SUVmax [30]. TLG represents metabolic activity throughout the entire tumour above a minimum threshold designed to exclude background activity [27]. Therefore, volume-based parameters such as MTV and TLG may reflect the metabolic burden of the active tumour more accurately and potentially be better surrogate imaging markers of tumour biology than SUVmax or tumour diameter. In the present study, multivariate analysis showed that MTV and TLG were independent prognostic factors for DFS and OS. The risk of recurrence was more than three times higher in patients with high MTV and/or TLG than in patients with low MTV and/or TLG. High MTV and/or TLG were also significantly associated with reduced OS.

Pretreatment SUVmax has been used to evaluate the likelihood of aggressive disease, metabolic response to therapy, early detection of disease recurrence, and outcome in patients with head and neck cancers [8, 31]. Many studies have shown that elevated SUVmax is associated with a poor clinical course [7, 32, 33]; however, previous work and the present study showed that high SUVmax is not a prognostic factor for recurrence or survival in OSCC [18, 19]. The present study also showed that SUVmax was not significantly correlated with either DFS or OS. Our findings suggest MTV and TLG may be more reliable prognostic predictors than SUVmax in these patients.

About 30 % of patients with cN0 OSCC harbour OM, depending on the size and thickness of the primary tumour and the histological features [34]. Because of frequent metastasis and regional recurrence, elective neck management has been advocated in cN0 patients with OSCC [35, 36]. We found that pretreatment MTV and TLG were the only independent predictors of OM in cN0 patients. The odds ratio of patients with MTV > 3.0 ml and/or TLG >13.9 g was over fivefold higher than that of those with MTV ≤3.0 ml and/or TLG ≤13.9 g. In contrast, SUVmax and clinical T classification were not predictive of OM. Nevertheless, 21 % (4 of 19) of the patients with OM had MTV ≤3.0 ml and/or TLG ≤13.9 g and would not have received proper neck treatment if these parameters had been used alone to determine the treatment plan. Although in the present study MTV ≤3.0 ml and/or TLG ≤13.9 g indicated a lower probability of OM, it is not yet appropriate to make decisions for neck management based on MTV and TLG alone, and further investigation is needed before applying these parameters in clinical practice.

In the present study, two different cut-off values for MTV and TLG were used to predict for DFS, OS and OM. The cut-off values for predicting OM were lower than those for predicting DFS and OS. While MTV and TLG of cN0 patients were the same as in those with primary tumour, MTV and TLG of all patients combined (those with primary tumour and those with metastatic lymph nodes) were included. The differences in cut-off values resulted from the difference in baseline volume between cN0 patients and all patients.

TNM stage and unfavourable histological features are important prognostic factors for clinical outcome; however, except for histological differentiation and involved resection margin, these were not prognostic of recurrence or death in the present study. This result may be explained by the small number of patients and emphasizes the strong predictive value of preoperative MTV and TLG.

Although volume-based parameters have advantages in terms of measuring metabolic tumour burden, the most appropriate segmentation method for estimating MTV and TLG is still debated. In this study, we excluded clinical T1N0 OSCC to minimize the partial volume effect of the primary tumour volume [20]. We used an SUV of 2.5 for tumour contouring based on previous investigations [15, 16, 26]. Although this method may miss some metastatic disease, it can avoid the erroneous inclusion of normal tissue surrounding the tumour. Further, the SUV is affected by many parameters, including ROI definition, image resolution, reconstruction method, parameters for the scanner model, the count noise bias effect, time between tracer injection and imaging, attenuation correction, normalization factors, and plasma glucose level [37]. Therefore, more accurate guidelines for the determination of PET/CT volume-based parameters should be established.

Our study had several limitations, including its retrospective design, the small number of patients and the inherent selection biases. Head and neck SCC are histologically identical but clinically heterogeneous entities that show a disparity in natural course or clinical behaviour depending on primary location. Previous studies of MTV involved the entire head and neck, all stages of cancer, and various treatment modalities, leading to a potential bias. To minimize potential biases, we included OSCC patients who underwent uniform treatment. Our findings are valuable and show the clinical significance of pretreatment MTV and TLG in OSCC.

In conclusion, pretreatment MTV and TLG may be useful in stratifying recurrence risk and survival in OSCC. Patients with high MTV and/or TLG have poor survival. MTV and TLG are the only independent predictors of OM in cN0 patients with OSCC. Pretreatment MTV and TLG could be helpful in selecting appropriate treatment and follow-up strategies. These metabolic parameters could be applied in stratifying patients as low-risk or high-risk and in determining the need for aggressive adjuvant chemoradiation therapy, along with primary resection plus neck dissection, in high-risk patients. Additional large-scale prospective trials are needed to validate the prognostic utility of these functional biomarkers derived from 18F-FDG PET/CT.

Conflicts of interest

None.

Grant support

This study was supported by grants (no. 2013-417) from the Asan Institute for Life Science and Basic Science Research Program through the National Research Foundation (NRF) of Korea funded by the Ministry of Education, Science and Technology (grant no. 2012R1A1A2002039), Seoul, Korea (J.-L. Roh).

Copyright information

© Springer-Verlag Berlin Heidelberg 2013