Correlations between DW‐MRI and 18F‐FDG PET/CT parameters in head and neck squamous cell carcinoma following definitive chemo‐radiotherapy

Abstract Background Posttreatment diffusion–weighted magnetic resonance imaging (DW‐MRI) and 18F‐fluorodeoxygluocose (18F‐FDG) positron emission tomography (PET) with computed tomography (PET/CT) have potential prognostic value following chemo‐radiotherapy (CRT) for head and neck squamous cell carcinoma (HNSCC). Correlations between these PET/CT (standardized uptake value or SUV) and DW‐MRI (apparent diffusion coefficient or ADC) parameters have only been previously explored in the pretreatment setting. Aim To evaluate stage III and IV HNSCC at 12‐weeks post‐CRT for the correlation between SUVmax and ADC values and their interval changes from pretreatment imaging. Methods Fifty‐six patients (45 male, 11 female, mean age 59.9 + − 7.38) with stage 3 and 4 HNSCC patients underwent 12‐week posttreatment DW‐MRI and 18F‐FDG PET/CT studies in this prospective study. There were 41/56 patients in the cohort with human papilloma virus‐related oropharyngeal cancer (HPV OPC). DW‐MRI (ADCmax and ADCmin) and 18F‐FDG PET/CT (SUVmax and SUVmax ratio to liver) parameters were measured at the site of primary tumors (n = 48) and the largest lymph nodes (n = 52). Kendall's tau evaluated the correlation between DW‐MRI and 18F‐FDG PET/CT parameters. Mann‐Whitney test compared the post‐CRT PET/CT and DW‐MRI parameters between those participants with and without 2‐year disease‐free survival (DFS). Results There was no correlation between DW‐MRI and 18F‐FDG PET/CT parameters on 12‐week posttreatment imaging (P = .455‐.794; tau = −0.075‐0.25) or their interval changes from pretreatment to 12‐week posttreatment imaging (P = .1‐.946; tau = −0.194‐0.044). The primary tumor ADCmean (P = .03) and the interval change in nodal ADCmin (P = .05) predicted 2‐year DFS but none of the 18F‐FDG PET/CT parameters were associated with 2‐year DFS. Conclusions There is no correlation between the quantitative DWI‐MRI and 18F‐FDG PET/CT parameters derived from 12‐week post‐CRT studies. These parameters may be independent biomarkers however in this HPV OPC dominant cohort, only selected ADC parameters demonstrated prognostic significance. Study was prospectively registered at http://www.controlled-trials.com/ISRCTN58327080


| BACKGROUND
Head and neck squamous cell carcinoma (HNSCC) is the seventh most common cancer worldwide. 1 Patients with advanced loco-regional disease can be treated with radiotherapy, or combined chemoradiotherapy (CRT), but 25% to 50% will have residual disease that requires further intervention. 2,3 Early posttreatment detection of residual tumor is required in order to optimize the outcomes of salvage surgery. 4,5 Unfortunately, clinical assessment and structural imaging are limited in their ability to detect loco-regional residual or recurrent disease due to treatment-related soft tissue distortion. [6][7][8] In order to overcome the shortcomings of structural imaging in this clinical scenario, metabolic imaging with 18F-fluorodeoxygluocose ( 18 F-FDG) positron emission tomography (PET) in combination with computed tomography (PET/CT) has become widely utilized. 2,9-12 18 F-FDG PET/CT is reported to be a highly sensitive technique for detection of HNSCC in the post-CRT setting. Semiquantitative analysis of maximum standardized uptake value (SUV max ) with 18 F-FDG PET-CT has been shown to have prognostic significance, with increased SUV max being associated with treatment failure. [13][14][15][16][17][18] Quantitative diffusion-weighted magnetic resonance imaging (DW-MRI) with measurement of the apparent diffusion coefficient (ADC) is another functional imaging technique, which may be used to help distinguish tumor from posttreatment changes and has been applied to the early posttreatment assessment of HNSCC. The majority of studies have found that an increased posttreatment ADC or greater rise in ADC from the pretreatment to the intratreatment or posttreatment studies is a predictor of treatment success. [19][20][21][22][23][24] It is still unclear whether 18 F-FDG and ADC values provide similar information with regard to viable tumor cells, or whether the two are entirely unrelated biomarkers. Both ADC values and SUV max have shown significant correlation with different histopathological parameters although this may depend on tumor grade. [25][26][27][28] There is currently no comparative data on the ability of post-CRT quantitative 18 F-FDG PET/CT and DW-MRI to predict post-CRT residual disease, with current evaluations having been limited to comparing these parameters in the pretreatment and intratreatment settings. [29][30][31][32][33][34][35][36][37] On the one hand, since a post-CRT reduction in SUV max and interval increase in ADC are both markers of treatment success, it would seem logical to expect them to negatively correlate. On the other hand, they reflect different biological processes, so the possibility of independent biomarkers, which are complementary in stratifying the probability of residual disease, should also be explored.
Our hypothesis was that there would be a significant negative linear relationship between ADC and SUV values on 12-week post-CRT studies in patients with stage III and IV HNSCC. Thus, our primary objective was to determine any correlation between ADC and SUV max values on 12-week post-CRT studies, and between their interval changes from pretreatment to 12-week post-CRT studies. Our secondary objective was to evaluate these posttreatment ADC and SUV max values and their interval changes for their ability to predict

| MRI processing and analysis
Processing and analysis of diffusion imaging was performed offline. cases. Free hand ROIs were placed on the assessable primary tumor and/or largest lymph node using the OsiriX Draw tool with the images magnified to a standard 300%. They were defined on the diffusionweighted imaging (DWI) b = 800 map as a focus of increased signal relative to background, but with access to the other MRI sequences.
Areas of necrosis (non-enhancement and high B0 map signal) and peri-tumoral oedema (avid enhancement and high B0 map signal) were avoided.
ROIs were placed on the baseline and 12-week posttreatment MRI studies in sequence. If there was no longer a 6 mm short axis focus of DWI signal on the posttreatment studies at the location of the initial lesion, a standardized 6 mm diameter circular ROI was placed at its original location by reference to the other sequences and these ROIs were termed "nonmeasurable." All ROIs were then translated directly to a calculated ADC map generated from the b100 and b800 images using the OsiriX ADCmap v1.9 plugin (https://web. stanford.edu/bah/software/ADCmap/). ADC mean and SD were recorded with ADC min calculated as ADC mean − one SD (rather than absolute ADC min ).  Post-filter: Gaussian 5.0 mm FWHM.

| 18 F-FDG PET/CT data processing and analysis
On pretreatment 18  The same method for VOI measurement was applied on 12-week posttreatment 18 F-FDG PET-CT imaging and SUV max was recorded.
The VOIs were chosen with MRI guidance, and always correlated with areas of increased tracer uptake on the pretreatment images. If there was reduced or no uptake on the posttreatment images relative to background, the posttreatment MRI images were referenced and a 6 mm VOI was placed at the same site as the MRI ROI. If necrosis was identified within a lesion, the area of necrosis was excluded as much as possible, and VOI was placed in the area of most intense tracer uptake. Areas of normal physiological uptake were avoided.
A freehand region of interest (ROI) was placed over the right lobe of the liver, at approximately its largest diameter, to record background liver SUV max. 31 This was performed in order to calculate the SUV max ratio to liver parameter.

| Statistical analysis
The Shapiro-Wilks normality test determined a proportion of the ADC mean and liver SUV max data to show a significant deviation from normal (with a multiple comparison corrected threshold). Therefore, the primary descriptive statistics focused on the nonparametric median ± interquartile range) and the primary correlation was performed with the nonparametric Kendall's Tau method.
The correlation between ADC mean and SUV max at the tumor and nodal sites on the 12-week post-CRT studies, and the interval change from pretreatment imaging to 12-week posttreatment studies was analyzed. The threshold for statistical significance was set at P < .05.
This study provided 95% power to detect a true "moderate" correlation of tau = 0.34 at this threshold.
The same statistical approach was used to extend the comparison to alternative parameters (ADC min , SUV max ratio to liver) on the Scatter plots of 18 F-FDG PET/CT and MRI measures were produced to demonstrate any correlation with 95% confidence intervals.
The Mann-Whitney test was used for a univariate analysis comparing tumor and nodal ADC and SUV max parameters with the dichot-

| Participants
The participant flowchart is summarized in Figure 1.
There were 70 participants initially enrolled in the study. However, five patients were subsequently withdrawn and a further nine participants did not attend for one of the posttreatment studies  Table 2. There was a majority of human papilloma virus-associated oropharyngeal cancer (HPV OPC) participants in this prospectively recruited cohort.
Since pretreatment 18 Table 3.  3 and 4). Table 4 demonstrates the correlation of MRI parameters and PET/CT parameters, with Table 5
It would therefore be useful to establish whether the post- F-FDG PET/CT has been questioned. 10,52 As an alternative, 18 F-FDG uptake may be measured relative to normal tissue/background tissue, and quantitative interpretative criteria such as the Porceddu, Hopkins, and Deauville scoring systems are based on this approach. [53][54][55] Zhong et al evaluated these scoring systems and found that they demonstrated high specificity, PPV, and NPV. 56 Tumor uptake exceeding liver tracer uptake is indicative of disease in all these criteria, and hence we decided to include tumor SUV to liver ratio as another 18 F-FDG PET/CT parameter. Some previous studies have demonstrated total lesional glycolysis (TLG) to be a superior predictor of HNSCC treatment outcomes 57 ; however, this requires an assessment of metabolic tumor volume, which was not possible in many of the posttreatment cases, where there was no definable FDG uptake and a standardized small VOI was placed. Second, with respect to the DW-MRI parameters, both ADC mean and ADC min were evaluated in this study since each has previously been applied to previous comparisons of quantitative DW-MRI with 18 F-FDG PET/CT on pretreatment imaging. 35,36 It is of note that while selected ADC parameters were able to predict treatment outcomes in this study, none of the SUV max parameters proved prognostically useful in this study. Some previous studies have shown that posttreatment SUV max is a less accurate predictor of outcome in HPV OPC cohorts. 58,59 In addition, the unexpectedly high rate of HPV-OPC participant recruitment also resulted in low rate of treatment failure such that it was suboptimally powered for the com- There are potential shortcomings with the study methodology.
First, there are greater challenges with the accurate placement and measurement of ADC and SUV in the posttreatment setting. For T A B L E 4 Correlation of MRI-DWI parameters (ADC min and ADC mean ) against PET parameters (SUV max , SUV max : liver-to-target ratio) at 12 weeks, and absolute interval change between pretreatment and 12 weeks posttreatment values

CONFLICT OF INTEREST
The authors have stated explicitly that there are no conflicts of interest in connection with this article.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from the corresponding author upon reasonable request.
T A B L E 6 Two-year disease-free survival and comparison of ADC and SUV max parameters in participants with and without 2-year diseasefree survival