Assessment of tumour hypoxia, proliferation and glucose metabolism in head and neck cancer before and during treatment

Abstract

Objective:

The aim of the study was to assess the feasibility of multitracer positron emission tomography (PET) imaging before and during chemoradiation and to evaluate the predictive value of image-based factors for outcome in locally advanced head and neck cancers treated with chemoradiation.

Methods:

In the week prior to the treatment [¹⁸F]−2-flu-2-deoxy-D-glucose (FDG), [¹⁸F]−3'-flu-3'deoxythymidine (FLT) and [¹⁸F]-flumisonidazole (FMISO) imaging was performed. FLT scans were repeated at 14 and 28 Gy and FMISO at 36 Gy. Overall survival, disease-free survival and local control were correlated with subvolume parameters, and with tumour-to-muscle ratio for FMISO. For every tracer, total metabolic tumour volume was calculated.

Results:

33 patients were included. No correlation was found between pre-treatment maximum standardised uptake value for FDG, FLT, FMISO and outcomes. Tumour volume measured on initial CT scans and initial FLT volume correlated with disease-free survivall (p = 0.007 and 0.04 respectively). FDG and FLT metabolic tumour volumes correlated significantly with local control (p = 0.005 and 0.02 respectively). In multivariate Cox analysis only individual initial TMRmax correlated with overall survival.

Conclusion:

PET/CT imaging is a promising tool. However, various aspects of image analysis need further clinical validation in larger multicentre study employing uniform imaging protocol and standardisation, especially for hypoxia tracer.

Advances in knowledge:

Monitoring of biological features of the tumour using multitracer PET modality seems to be a feasible option in daily clinical practice.

Evaluation of hypoxic subvolumes is more patient dependent; thus, exploration of individual parameters of hypoxia is needed. tumour-to-muscle ratio seems to be the most promising so far.

Introduction

Despite the rapid progress in radiotherapy, treatment outcomes of locally advanced head and neck cancers (LAHNC) remain unsatisfactory. Morphological assessment of the tumour proved insufficient to determine the treatment outcome and existing research results are conflicting.^1–3

Glucose metabolism, hypoxia^4,5 and proliferation^6,7 remain the most extensively investigated molecular features of tumours visualised by positron emission tomography (PET). [¹⁸F]−2-flu-2-deoxy-D-glucose (FDG) is a marker reflecting glucose metabolism. It was also investigated whether FDG uptake intensity may be an indirect indicator of hypoxia,⁸ however, existing research is not conclusive.⁹ The drawback of FDG is its high affinity to inflammatory regions, which can influence the assessment of response during radiotherapy.

[¹⁸F]−3'-Flu-3' deoxythymidine (FLT) is an indirect proliferation tracer—the highest uptake of FLT is detected in highly proliferative subvolumes of the tumour. Lymph node assessment using FLT in LAHNC is still controversial due to the high false-positive rate.⁷ Although highly proliferative regions seem to be the most sensitive to radiation, it is not clear what dose of radiation is required to sterilise the tumour of proliferating cells.

The role of hypoxia as a main cause of treatment failure in LAHNC has been already shown in multiple studies..^4,5,10,11 [¹⁸F]-flumisonidazole (FMISO) is the most clinically validated hypoxia tracer so far, in spite of its drawbacks—especially the low contrast between hypoxic volumes and the background.

The first attempt to asses multitracer response assessment in head and neck tumours was made by Thorwarth et al who analysed pre-treatment FDG and FMISO images.⁹ The predictive value of FDG and FLT parameters in nasopharyngeal cancer was tested by Qi with promising results¹²

In search for predictive factors of the outcome a number of different parameters were studied so far, including maximum standardised uptake value (SUVmax), metabolic volumes and sub volumes of the tumour and lymph nodes assessed with various tracers.

The aim of this study was to assess the feasibility of the multi tracer PET imaging before and during standard treatment of LAHNC and to correlate potential predictive factors with treatment outcome.

Patients and methods

Patients

39 patients with biopsy-proven squamous cell carcinoma of oral cavity, oropharynx, larynx or hypopharynx and eligible for radical chemoradiation were included in this prospective study. The study was supported by KBN N N402 352138 grant from State Committee for Scientific Research and accepted by the institutional Ethics Committee. All patients provided written informed consent.

All patients underwent diagnostic imaging including contrast-enhanced CT as well as clinical examination. In all cases, p16 status was assessed.

Treatment

All but eight patients were treated with chemoradiation, consisting of planned three courses of cisplatin 100 mg/m² on days 1, 22 and 43 concomitantly with intensity modulated radiotherapy (IMRT, Tomotherapy, Accuray Inc., Sunnyvale, CA) for the tumours with 1 cm margin to a total dose of 70 Gy/35 fractions, 60 Gy for volumes of high risk of microscopic spread of tumour, and 50 Gy for elective volumes. Gross tumour volumes (GTVt) were delineated on contrast-enhanced CT images. Separate nodal GTVn have been created for analysis of the involved lymph nodes. Clinical target volumes (CTV) were adapted individually. Planning target volumes (PTV) were added to the CTVs as three-dimensional uniform margins of 3 mm according to the set-up and internal motion errors for head and neck treatment, calculated for the institution.¹³

Response of the tumour was evaluated 3 months after treatment by clinical examination and contrast-enhanced CT. Complete response (CR) was defined as complete dissapearance of the tumour and partial response (PR) as 50% or greater decreasing of volume of the primary. Stabilsation (ST) was defined as less than 50% decreasing tumour volme and progression (PRO) as increasing volume of primary.

OS was defined as time from diagnosis to cancer-related death, DFS was defined as time to loco-regional or distant disease progression or death whichever comes first after complete (CR) or partial remission (PR) or stabilization. Local control (LC) was defined as time from completion of treatment resulted in CR, PR or stabilization to relapse or progression of tumour at primary site or regional lymph nodes.

Imaging

In the week preceding the treatment, FDG, FLT, FMISO PET/CT (Gemini TF TOF 16, Philips Healthcare Inc., Andover, MA) imaging was performed every 3 days, starting with FDG, followed by FLT and FMISO, to overcome potential interference with tracers. All images were obtained in the treatment position. PET calibration and QA procedures were provided by the manufacturer, according to the European Association of Nuclear Medicine (EANM) guidelines.¹⁴

FDG scans were performed 60 min post injection, (p.i.) following standard protocol.

For FMISO imaging, patients were scanned 120 min p.i., FMISO scans were repeated once during treatment, at 36 Gy.

FLT scans were performed 30 min p.i. Scans were repeated twice during treatment: at 14 and 28 Gy.

For each tracer, the average injected activity was reduced to 200 MBq in order to protect patients’ bone marrow in view of the planned multiple PET scan procedures. To avoid insufficient image statistics the scanning time was increased up to 160% of the standard scanning time—on average 3 min per bed position. This procedure was first tested first with a dedicated phantom to determine the scanning time required to avoid compromising on the number of coincidence events taken for the reconstruction.

PET images were reconstructed with ordered-subset expectation-maximisation (OSEM) algorithm, with correction of scatter and attenuation. Standardised uptake value (SUV) was calculated.

For image analysis rigid co-registration of CT and PET images was performed.

Image analysis and definition of hypoxic subvolumes

For images analysis, MIM Maestro software (MIM Software Inc., Cleveland, USA) was used. Tumour volumes were first delineated on the contrast-enhanced CT and matched with FDG PET/CT images. Subsequently metabolic tumour volume (MTV) was semi-automatically delineated on FDG scans using 50% of the maximum signal intensity in the tumour. Images with FDG contour were co-registered with FLT and FMISO scans before and during the treatment.

FLT sub volumes were automatically segmented inside the basic FDG volume using 50% isocontour of the maximum signal intensity.

FMISO uptake and hypoxic subvolumes automatic segmentation was performed using fixed threshold levels of ≥1.5 and 1.6 for SUVmax and individually calculated tumour-to-muscle ratio (TMR).

To determine individual FMISO uptake in muscles, a region of interest in the contralateral semispinalis neck muscle was delineated for each patient. The impact of an individual TMRmax on outcome was also evaluated for each patient.

All the analyses presented above were performed separately for the tumour and the lymph nodes involved. Total metabolic tumour and lymph nodes volume (MTV) for every tracer was also calculated and correlated with outcome.

Statistics

All statistical calculations were performed using STATISTICA (StatSoft, Inc. Statistica v. 12, Tulsa, OK, 2014). Normality of distribution was determined by the Shapiro–Wilk test. Spearman correlation was used for non-normally distributed data. Wilcoxon test was employed to assess the statistical significance of identified differences. Univariate and multivariate Cox proportional hazards models were used to evaluate predictors of outcome.

In multivariate analysis each of the variable significant in univariate analysis, i.e N, CT tumour volume (VCT), VCT 28 GY, VFLT28 Gy, TMRmax, FLT MTV was adjusted for known clinical prognostic factors like tumour stage (T), nodal status (N) and concomitant use of cisplatin in treatment.

Each statistical test in this study was evaluated at the significance level α = 0.05.

To evaluate impact of FDG, FLT and FMISO pre-treatment MTV on OS and LC, receiver operating characteristics (ROC) analysis for every parameter was performed. Cut-off values were obtained by application of Youden index and used for stratification of patients in Kaplan–Meier plots. For pretreatment VCT median value was used.

The multiple testing correction was done using Benjamini–Hochberg procedure at false discovery rate (FDR) 0.05. Results of this testing were added in the Supplementary Tables 1 and 2

Results

39 consecutive patients were prospectively included in the study in years 2010–2015. Two patients were lost for follow-up. In four patients, it was not possible to perform all planned imaging before treatment. Due to technical reasons or patients’ consent withdrawal all 3 PET/CT imaging procedures during radiotherapy were performed in 20 patients.

33 patients were eligible for analysis of pre-treatment images, and 20 for the analysis of all planned images. Patient characteristics are presented in Table 1.

Table 1.

Patients characteristics

Characteristic N = 33	Value
Age (years)
Range	50–75
Median	59
Gender
Male	24
Female	9
Primary Site
Oropharynx	7
Hypopharynx	10
Oral cavity	9
Larynx	7
Tumour classification
T2	7
T3	4
T4	22
N0	6
N1	5
N2	21
N3	1
Follow-up
Median FU	19,5 mts
Range	1–79 mts
Human papillomavirus status:
Positive	4
Negative	26
Unknown	3
Treatment
RT	8
RTCT	25

RT, radiotherapy;RTCT, chemoradiation.

Median follow-up was 19.5 months (range 1–79 months).

Five patients completed treatment without complete remission: four patients with partial remission (PR) and one patient with stable disease (SD). No progression of tumour was detected at first control after treatment.

At the time of anlaysis, events as follows were identified: 13 patients died from cancer, 8 local treatment failures and 1 locoregional failure. Three patients experienced lung metastases. There was no non-cancer-related death at time of analysis.

The 2-year OS for whole group was 65%.

Treatment and imaging compliance

All patients treated with chemoradiation received two courses of cisplatin, at day 1 and day 22 of treatment. None of them received planned three courses of cisplatin due to side-effects.

Eight patients were treated with radiotherapy only because of symptoms of renal failure and/or comorbidities before radiotherapy.

Imaging before treatment

Fixed threshold method for FMISO images before treatment allowed to identify hypoxic tumours in 33% (11/33) of patients for both 1.6 and 1.5 threshold levels. Median tumour volumes were 1.2 cc (0.1–18.5cc) and 0.85 (0.1–14.4) for 1.5 and 1.6 respectively.

Median individual TMRmax before treatment was 1.36 (0.9–2.9) for the whole group.

Imaging during treatment

Median CT tumour volume measured at consecutive CT scans decreased during treatment in all patients, from baseline 26.5 cc (0.2–174.2 cc) to 7.4 cc(0–55.7cc) after 36 Gy.

Significant decrease in FLT SUVmax (p < 0.001) was found during treatment for both time points. FLT uptake was still detectable in tumour at 28 Gy in 11 patients. In two of them, FLT SUVmax decreased in the first measurement, however an increase of this parameter was noted at 28 Gy. In one of these patients treated with radiotherapy only, rapid recurrence was observed 6 months after treatment completion.

A decrease in FMISO uptake was observed during treatment as a reduction of hypoxic volumes (p < 0.001). In 4 out of 20 patients (20%) residual FMISO volume was detected after 36 Gy. Minimal tumour response after treatment or early local failure was found in three patients. Due to the small number of events, evaluation of the impact of this observation on treatment outcome was not possible.

Median individual TMRmax during treatment was 1.22 (0.9–2.7). In six patients, an increase in the TMRmax was observed, two completed the treatment without complete remission of the tumour, and further two of them had residual hypoxic volumes after 36 Gy.

Impact on outcome

Tumour sub volumes and SUVmax measured before and during treatment as well as correlation with the treatment outcomes are presented in Table 2.

Table 2.

Variables measured before and during treatment and impact on OS, DFS and LC, (p < 0.05^*); Spearman Rank Correlation with time to event or censoring

Variable	No. of patients	Median/range	pOS	R Spearman	OS Above/below median (mts)	pLC	R Spearman	LC Above/below median (mts)	pDFS	R Spearman	DFS Above/below median (mts)
VCT	33	26.5 (0.2–174.3)	0.05	−0.316	15.5/39	0.01a	−0.407	10/35	0.007a	−0.461	9.5/25
VCT1 14 Gy	26	18.8 (0–74.8)	0.13	−0.301	14/32	0.04a	−0.413	8/24	0.13	−0.323	8/25
VCT2 28 Gy	24	12.6 (0–68.9)	0.03a	−0.44	13/36	0.03a	−0.463	6.5/31	0.007a	−0.611	6.5/31.5
VCT3 36 Gy	20	7.4 (0–55.7)	0.03a	−0.504	13/21	0.04a	−0.464	8.5/18	0.02a	−0.609	8.5/18
V[18F]LT	33	16.8 (0.9–66.6)	0.17	−0.225	15/35	0.03a	−0.277	8/24	0.04a	−0.374	8/24
V[18F]LT 14 Gy	24	6.4 (0–47.9)	0.59	−0.113	16/43	0.32	−0.222	10/34	0.2	−0.290	10.5/34
V[18F]LT 28 Gy	22	0.3 (0–19.6)	0.4	−0.184	14.5/17	0.16	−0.320	10/13	0.22	−0.303	9.5/14
V[18F]DG	33	18.7 (0.5–108)	0.24	−0.192	17/36	0.04a	−0.336	13/24	0.05	−0.340	11/13
V[18F]MISO 1.5	33	1.2 (0.1–18.5)	0.83	−0.037	19/NA	0.38	−0.160	15/NA	0.33	−0.191	14/NA
V[18F]MISO 1.5 36 Gy	20	0 (0–3.4)	0.08	−0.404	17/NA	0.07	−0.416	13/NA	0.16	−0.364	14/NA
V[18F]MISO 1.6	33	0.85 (0.1–14.4)	0.76	−0.053	17/NA	0.23	−0.214	13/NA	0.14	−0.278	14/NA
V [18F]MISO 1.6 36 Gy	20	0 (0–1.6)	0.28	−0.258	17/NA	0.37	−0.215	13/NA	0.16	−0.364	14/NA
[18F]MISO SUVmax	33	1.7 (1.5–2.5)	0.66	−0.074	17/27	0.24	−0.207	13/24	0.06	−0.347	13/24
[18F]MISO SUVmax 36 Gy	4/20a	1.5 (0.7–1.8)	0.69	−0.097	19/14	0.51	−0.158	15.5/12	0.64	−0.125	14/13
[18F]LT SUVmax	33	6.5 (2.6–16)	0.2	−0.210	14.5/37.5	0.1	−0.277	10–29.5	0.2	−0.231	7.5/2.5
[18F]LT SUVmax 14 Gy	21	3 (0.5–6.4)	0.25	−0.257		0.24	−0.282	9/31	0.09	−0.405	9/32
[18F]LT SUV max 28 Gy	11/20a	3.2 (1.9–4.5)	0.85	−0.059	17/34 14/15	0.9	−0.038	2/12	0.51	0.250	3/11
[18F]DG SUVmax	33	11.3 (2.9–18.3)	0.85	−0.033	17/21	0.45	−0.136	13/21	0.23	−0.229	8.5/13
TMRmax	33	1.4 (0.9–2.9)	0.065	−0.315	16.5/27	0.03a	−0.364	13/20	0.04a	−0.375	13.5/21
TMRmax 36 Gy	20	1.2 (0.9–2.7)	0.38	−0.214	19/16	0.38	−0.214	16.6/13	0.58	−0.147	12.5/14
[18F]DG MTV	33	23.2 (2.4–106.7)	0.11	−0.263	15.5/36	0.005a	−0.446	6.5/24	0.04a	−0.355	6.5/14
[18F]LT MTV	33	20.1 (0.9–135.5)	0.15	−0.237	16/36	0.02a	−0.392	12/24	0.05a	−0360	9/24
[18F]MISO MTV	33	0 (0–14.9)	0.81	−0.042	17/NA	0.32	−0.181	13/NA	0.25	−0.224	14/NA

[¹⁸F]-FDG, [¹⁸F]-2-flu-2-deoxy-D-glucose; [¹⁸F]-FLT, [¹⁸F]-flu-3'deoxythymidine; [¹⁸F]-FMISO, [¹⁸F]-flumisonidazole; LC, local control; OS, overall survival; SUVmax, maximum standardised uptake value; VCT, CT tumour volume.

V – baseline tumour volume in CT, FDG, FLT, FMISO.

V14Gy, 28 Gy, 36 Gy – tumour volume measured at indicated time point during treatment.

VFMISO 1.5, 1.6 – hypoxic tumour volume based on indicated threshold levels.

MTV FDG, FLT, FMISO – metabolic tumour volume.

SUVmax– baseline maximum standardised uptake value.

TMRmax – baseline Tumour-Muscle-Ratio based on SUVmax.

TMRmax 36 Gy – Tumour-Muscle-Ratio based on SUVmax at 36 Gy.

OS, LC, DSF above/below median – time of survival or control for values of variables above or below median.

^anumber of patients with residual uptake of tracer at selected time point/all patients analysed in selected time point

No statistically significant correlation was found between CT, FGD, FLT and FMISO tumour volumes determined before treatment and the OS. No difference in any outcome was found between patients below and above the median parameter values, except for CT tumour volume, where median (16.8 cc) significantly stratified the patients by DFS but not OS (Figure 1a,b).

Figure 1.

(a) Kaplan–Meier curves for OS stratified according to median baseline VCT. (b) Kaplan–Meier curves for DFS stratified according to median baseline VCT. DFS,disease-free survival; OS, overall survival; VCT, CT tumour volume

There was no correlation between pre-treatment SUVmax for all tracers and the outcome of the treatment.

Pre-treatment VCT and total FDG MTV and FLT MTV correlated with LC and DFS, but not with OS. In an attempt to determine tumour volumes that allowed to stratify patients according to prognosis, ROC curves were generated for FDG and FLT MTVs. Generated ROCs achieved sensitivity 0.8 and specificity 0.6 for LC and 0.8 and 0.5 respectively for DFS. The results for FLT MTV had low statistical power making it impossible to draw definite conclusions. Kaplan–Meier curves for FDG MTV are presented on Figure 2a and b.

Figure 2.

(a) Kaplan–Meier curves for LC stratified according to cut-off value 24.45 for baseline FDG MTV. (b) Kaplan–Meier curves for DFS stratified according to cut-off value 12.5 for baseline FDG MTV. [¹⁸F]-FLT, [¹⁸F]-flu-3'deoxythymidine;DFS, disease-free survival; LC, local control; MTV, metabolic tumour volume

VCT in the second and third week of treatment were significant determinants of the outcome for LC (p = 0.03 and p = 0.04 respectively) and DFS (p = 0.007 and p = 0.02 respectively) as well as for OS (p = 0.03 for both weeks).

The only hypoxia parameter significant for all measured outcomes was baseline TMRmax before treatment. Baseline TMRmax correlated with both LC and DFS (p = 0.03 and 0.04 respectively) and was the only important parameter for OS, in univariate and multivariate Cox regression analysis in combination with known clinical prognostic factors—T and N stage and concomitant cisplatin (Table 3). The other models did not reach statistical significance (Supplementary Table 2)

Table 3.

Cox regression analysis for overall survival

Variable	Univariate model		Multivariate modela
	Hazard ratio (95% CI)	p	Hazard ratio (95% CI)	p
T	1.34 (0.65–2.86)	0.39	1.05 (0.44–2.51)	0.92
N	2.62 (1–6.87)	0.04a	2.59 (0.95–7.02)	0.06
Cisplatin	0.55 (0.18–1.69)	0.3	0.48 (0.13–1.78)	0.27
VCT	1.01 (1–1.02)	0.01a
VCT1 14 Gy	1.02 (0.99–1.06)	0.11
VCT2 28 Gy	1.04 (1–1.08)	0.01a
VCT3 36 Gy	1.03 (0.99–1.08)	0.05
V[18F]LT	1.03 (0.99–1.07)	0.06
V[18F]LT1 14 Gy	1.02 (0.98–1.07)	0.26
V[18F]LT2 28 Gy	1.14 (1.02–1.28)	0.02a
V[18F]DG	1.01 (0.99–1.04)	0.07
[18F]DG SUVmax	1.08 (0.94–1.24)	0.26
TMR max	6.36 (1.67–24.17)	0.006a	4.95 (1.09–22.51)	0.04a
TMR max 36 Gy	2.87 (0.58–14.14)	0.19
[18F]DG MTV	1.01 (0.99–1.03)	0.06
[18F]LT MTV	1.02 (1–1.04)	0.02a
[18F]MISO MTV	1.06 (0.92–1.22)	0.38

CI, confidence interval; [¹⁸F]-FDG, [¹⁸F]-2-flu-2-deoxy-D-glucose; [¹⁸F]-FLT, [¹⁸F]-flu-3'deoxythymidine; [¹⁸F]-FMISO, [¹⁸F]-flumisonidazole; MTV, metabolic tumour volume; SUVmax, maximum standardised uptake value; TMR, tumour-to-muscle ratio;VCT, CT tumour volume.

Only TMR max remained significant when adjusted for known clinical factors (Tumour and nodal status and concomitant using cisplatin) in a multivariate model.

V—baseline tumour volume in CT, FDG, FLT, FMISO.

V14Gy, 28 Gy, 36 Gy—tumour volume measured at indicated time point during treatment.

VFMISO 1.5, 1.6—hypoxic tumour volume based on indicated threshold levels.

MTV FDG, FLT, FMISO—MTV.

SUVmax—baseline maximum standardised uptake value.

TMRmax—baseline TMR based on SUVmax.

TMRmax 36 Gy—TMR based on SUVmax at 36 Gy.

^aMultivariate analysis results for the only model in which baseline TMRmax was significant.

The pre-treatment median TMRmax did not stratify patients with regard to OS, ROC analyses were performed. Generated TMRmax cut-off point of 1.23 allowed to classify patients according to the probability of survival with statistical significance (p = 0.01). However, this result should be treated with caution due to the small group of patients and suboptimal ROC parameters, i.e. sensitivity 0.9 and specificity 0.4. Nevertheless, 8 out of 9 patients with TMRmax before treatment lower or equal 1.23 were still alive and disease free at the end of follow-up period, and one of them survived at least 2 years without treatment failure.

Four patients were human papillomavirus positive, two of them diagnosed with oropharyngeal cancer however the group was too small to calculate correlations.

Discussion

The aim of this study was to assess the feasibility of multitracer imaging before and during treatment as well as to investigate whether biological changes of the tumour determine treatment outcome in LAHNC patients. To the authors’ knowledge, this is the largest cohort of patients with advanced LAHNC with triple tracer PET evaluation before and during treatment studied so far. In line with study by Thorwarth et al, this study did not find a significant association between baseline tumour hypoxia parameters such as volume and SUVmax measured by FMISO PET/CT and treatment outcome.⁹ Our results support findings of Zips et al that the characteristic of tumour during treatment, especially with regard to hypoxia, is a stronger prognostic factor than at baseline.⁵ The number of events in presented study was, however too low to obtain conclusive results. Reports concerning the significance of the parameters obtained from PET imaging are conflicting. Higgins et al emphasised the importance of pre-treatment FDG SUVmean (but not SUVmax) as a prognostic factor of DFS in LAHNC patients treated with chemoradiation and targeted therapy.¹⁵ Schwartz reached different conclusion, stating that primary tumour FDG SUVmax is a promising prognostic factor for HNSCC.¹⁶ An overview of the studies analysing FDG SUVmax was presented by Schinagl who suggested that these conflicting findings might be a result of heterogeneity in terms of both treatment and delineation of target volumes—with or without lymph nodes—as well as of different segmentation methods.² Currently, there is no clear-cut answer as to whether pre-treatment FDG SUV max is a prognostic factor of OS.

Results regarding pre-treatment tumour volumes are equally disparate.^17–19 In the presented research none of FDG, FLT or CT tumour volumes at baseline were predictive for OS, but significant correlation was found with LC and DSF. Moreover, presented study revealed that the total FDG MTV and FLT MTV before treatment correlated inversely with LC and DFS, but contrary to findings by Lim and Trang, not with OS.^20,21 Prognostic impact of FDG PET in head and neck tumours was evaluated in meta-analysis of 25 studies by Bonomo et al, confirming the significance of pre-treatment MTV volume for LC and OS.²²

There are relatively few studies performed with the use of FLT tracer. Hoeben demonstrated that in case of accelerated chemoradiation treatment and weekly low dose of cisplatin, high FLT SUVmax correlates with better outcome.⁶ In presented study such effect was not observed, perhaps due to fact that accelerated radiotherapy helped to overcome the effect of repopulation in the tumour. 2 Gy daily fraction increased the likelihood of negative impact of tumour repopulation in the tumours with higher initial FLT uptake on the treatment outcome.

An interesting finding is the observed increase in the FLT uptake in the tumour at the second time point of 28 Gy. For the majority of patients, FLT uptake was undetectable or residual at this time point except for two patients where an increase in the FLT uptake at the second time point was identified. Both patients with increased FLT uptake after 28 Gy failed to cure.

A similar effect was reported by Yue in the study of radiotherapy of oesophageal cancer after a break in the treatment and it most likely reflects tumour repopulation.²³

Tumour hypoxia was investigated as a predictive factor in number of studies, most of which confirmed the impact of hypoxia and tumour reoxygenation on the outcome.

The main challenge in analysing the impact of hypoxia on treatment outcome is the definition of hypoxic volumes. For FMISO, there is no consensus regarding fixed thresholds of FMISO SUVmax and TMRmax which vary between 1.25 and 1.9.^24–27 Threshold levels for this study were selected based on levels most often reported in the available literature. Applying different segmentation methods resulted in fewer hypoxic volumes observed in comparison to other authors.²⁵ As the volume determined as hypoxic may vary between patients, the more individualised parameter—TMRmax—was applied in this study. Several authors reported TMRmax or tumour-to-background ratio (TBRmax) as a potential predictive factor of treatment outcome, but the results are conflicting.²⁸

In presented study there was a significant correlation between TMRmax, LC and DFS. In multivariate Cox analysis, only TMRmax reached statistical significance for OS, in combination with known clinical prognostic factors—T and N stage and concomitant cisplatin in one model.

Finding the level of TMRmax that would allow robust stratification is difficult and depends on multiple factors.^29,30 Presented study did not succeed in identifying such a level, although with ROC analysis an attempt was made to discriminate patients based on 1.23 TMRmax value. The result is in line with published literature,²⁹ however unsatisfactory ROC parameters must be considered.

The increase of hypoxic volume during treatment was observed in one patient only which stands in contrast to reports by other authors.⁵ A possible explanation is the chosen timepoint of hypoxia measurement (at 36 Gy only), which falls later than the previously reported increase in FMISO uptake. Residual FMISO uptake was noted in four patients and three of them failed treatment. In six patients TMRmax increased during treatment, but the increase did not correlate with treatment failure.

For the presented study, a commercially available automatic segmentation system was used. It should be noted that in the available research various segmentation algorithms (both in-house and commercially available) were employed. Additionally, imaging procedures vary between centres, especially for tracers other than FDG. There are still no accepted standards for automatic segmentation for various tracers—especially in the assessment of subvolumes during treatment, where decrease in FMISO or FLT uptake is expected.

This study is limited in the smaller number of patients than assumed in the study protocol, which resulted in low statistical power. Due to the small sample size and since multiple correlations were evaluated, some of the associations are inevitably expected to be false positives (see Supplementary Table 1). Unadjusted p-values were reported, as the main aim of this exploratory analysis was to highlight markers plausibly associated with the outcomes of interest and identify candidates for more rigorous, confirmatory study.

Conclusions

Evaluation and monitoring of biological features of the tumour using multitracer PET modality seems to be a feasible option in routine clinical practice.

PET/CT imaging is a promising tool; however, many aspects of image analysis need clinical validation in a larger study with uniform protocol of imaging and standardisation—especially for hypoxia tracer. Evaluation of hypoxic subvolumes is even more difficult due to more individual character of hypoxia, which calls for further exploration of more individual parameters such as TMR.