[¹⁸F]FDG PET/CT in head and neck squamous cell carcinoma: a head-to-head between visual point-scales and the added value of multi-modality imaging

Head and Neck Squamous Cell Carcinoma (HNSCC) represents the 6th leading cancer worldwide [1]. After the primary treatment, the recurrence rate within 2 years accounts for approximately 60% of patients, with 20–30% experiencing distant metastases, underlining the need for reliable post-treatment evaluation and proper follow-up [2].

In most cases, patients present a locally advanced disease at diagnosis (stage III or IV) with nodal involvement. In this setting, the combined chemo-radiotherapy (CRT) approach or, less frequently, radiotherapy alone (RT) with curative intent is considered the standard of care. After the end of therapy, a clinical assessment is needed to guide the following decision-making [3].

It is well known that treatment-related changes, particularly edema and fibrosis after high radiation dose, could often limit morphological imaging in post-treatment evaluation [4‐6]. Nowadays, fluorine-18 fluorodeoxyglucose positron emission tomography/computed tomography ([¹⁸F]FDG PET/CT) is a validated tool in post-treatment HNSCC evaluation, with a high level of evidence. The most recent guidelines recommend the use of [¹⁸F]FDG PET/CT preferably 12 weeks after the end of treatment, pointing out a PET-guided approach. In case of a negative PET exam, the patient is followed-up with clinical observation; in doubtful cases, the patient can be subjected to clinical observation or can repeat the [¹⁸F]FDG PET/CT in 3–6 months. Finally, a clear PET positivity requests biopsy, primary surgery, or lymphadenectomy after CT or MRI confirmation [3].

In 2016, the PET-NECK, randomized, controlled trial data demonstrated that [¹⁸F]FDG PET/CT is an accurate and cost-effective technique for HNSCC response assessment, sparing unnecessary neck dissection in nearly 80% of patients [7]. Notably, the metabolic assessment showed a high negative predictive value (NPV) assessing its essential role in HNSCC surveillance [8, 9].

Over the last few years, the need to clarify the diagnostic value of [¹⁸F]FDG PET/CT even in doubtful and positive cases has promptly emerged. In order to better distinguish the disease from treatment-related changes, some authors have evaluated the prognostic role of several semiquantitative parameters, such as maximum standardized uptake value (SUV_max), but failed to find out a reliable method [4, 10, 11].

Therefore, there has been growing interest in optimizing the qualitative analysis by proposing different visual scales to standardize the metabolic imaging response. The first scoring system, named the Hopkins Criteria (HC), was introduced by Marcus and colleagues in a retrospective study considering a cohort of 214 patients. The 5-point scale achieved an NPV and overall diagnostic accuracy of 91.1% and 86.9%, respectively. These results were confirmed by the prospective multicenter ECLYPS study, where the use of HC resulted in a NPV of 92.1% [12, 13]. However, considering the studies’ bias as well as the low positive predictive value (PPV), other PET qualitative scores have recently been proposed: the well-known Deauville score (DS) and the novel Cuneo score (CS). In particular, CS, a 6-point visual scale, was introduced to minimize false-negative and false-positive results, resulting in a higher PPV to better discriminate persistent disease from treatment-related changes (e.g. inflammatory reactions after RT) [14]. In addition, each scale takes into account different visual reference regions (e.g. internal jugular vein (IJV); mediastinal blood pool (MBP); liver); only the CS introduces the absence/presence of the local background in PET images evaluation to highlight the possible interference of the post-treatment reaction of local tissue that impacts on visual score interpretation.

The first goal of this study was a head-to-head comparison between the diagnostic performance of the Hopkins criteria, Deauville score, and Cuneo score, together with the assessment of their prognostic role.

Secondly, we extracted SUV_max and ΔSUV_max of the lymph node with the highest uptake and we calculated the product of diameters measured on the co-registered CT images on the restaging PET/CT scan to investigate the possible added value of semiquantitative analysis and morphological data as a useful variable to consider in doubtful cases.

Despite the consolidated role of [¹⁸F]FDG PET/CT in the treatment response evaluation of HNSCC, there is no clear consensus on the optimal interpretation criteria to use in this clinical scenario. Semi-quantitative PET parameters, such as the well-known SUV_max, have not proved to be reliable indicators probably due to the numerous technical and patient variables that could influence this value, especially in the investigated population (e.g. inflammation and post-treatment effects) [4, 15, 16].

Over the last years, considerable efforts have been made to standardize the reporting system in the surveillance of HNSCC and several qualitative assessment methods for predicting regional disease control have been proposed to find out the best [¹⁸F]FDG PET visual scale. However, none of them were approved and routinely used up to date. As known, one of the major limits remains the low sensitivity and PPV.

In our study, the head-to-head comparison between scales showed good sensitivity, specificity, and PPV (87%, 86%, and 76% respectively) and a high NPV (92%) of the first introduced Hopkins scale. Comparing our results with the reference one, published by Marcus and colleagues, we reported almost similar specificity and NPV (92.2% and 91.1%, respectively), but a higher sensitivity and PPV compared to 68.1% and 71.1% in the reference study [12].

Similarly, a 2017 study by Kendi et al. [17] on 69 HNSCC patients showed a sensitivity, specificity, PPV, and NPV of the overall therapy assessment of 66.7%, 87.3%, 33%, 96.5% respectively, underlining the higher sensitivity and PPV reached in our study. The diagnostic accuracy of PET/CT response assessment in our study might be affected by the time interval between treatment and follow-up imaging; the later median time point of imaging post-therapy (20 weeks) compared to other studies may explain slightly higher PPV[18].

More recently, other authors proposed to apply the Deauville score, the most famous 5-point standardized Likert scale used for Hodgkin’s and non-Hodgkin’s lymphoma, to post-treatment HNSCC PET assessment [19]. Benjamin et al. conducted a retrospective analysis on 43 HNSCC patients who underwent organ preservation radiotherapy and applied the DS to [¹⁸F]FDG PET/CT scan. Their study results showed a higher PPV than our (100% vs 70%), but only 4/43 patients experienced disease progression (4 or 5 DS) [4].

Another study by Zhong et al., in a larger cohort of 562 patients, compared the diagnostic performance of the Hopkins and Deauville score for the prediction of locoregional control and PFS. The study confirmed the high NPV of both visual criteria, also achieving a very high PPV compared to our as well as to the existing literature (51–78%). However, the authors attributed this higher value to the later median time point of imaging post-therapy (17 weeks) [20].

To deal with these conflicting results, in 2020 an Italian Multicentric study introduced a novel 6-point visual system called Cuneo score to find out a higher PPV [14]. The reference article compared the three different PET visual scores, reporting the best PPV for Cuneo score (with scores 3 and 4 clustered with 1 and 2, indicative of the absence of disease) for all categories: 42.9% for the primary tumor, 100% for the nodal involvement, and 50% for the cumulative score. In our cohort, primary tumor, lymph node involvement, and the cumulative Cuneo score reached higher PPV values (86%, 83%, and 78% respectively), at the expense of lower values of sensitivity and NPV, compared to Hopkins and Deauville scores. Our results support the hypothesis proposed by Bonomo et al. of considering 6-point visual criteria to minimize false positive PET findings in post-treatment evaluation of HNSCC, resulting in a better PPV. Further prospective studies with a larger sample are warranted to support the Cuneo criteria. However, since none of the mentioned criteria showed to be able to solve equivocal cases (e.g., Fig. 5) a further repeat PET/CT 3–6 months later is still necessary, as suggested by current guidelines [3].

The survival analysis was consistent with the current literature [21]. All the studied scores again showed their high prognostic value in terms of both PFS and OS, underlining the reliability of a prognostic scoring system in HNSCC patients. Two clinical examples extracted from our sample are shown in Figs. 6 and 7, supporting the prognostic role of each visual score.

Regarding the secondary endpoint of the study, both SUV_max extracted from restaging PET and ΔSUV_max were able to discriminate lymph node persistent disease from treatment-related changes, showing a good performance in predicting outcomes. However, their cutoff value found (2.75 for SUV_max, − 6.6 for ΔSUV_max) needs to be considered with caution, confirming the debated role of semiquantitative analysis in literature.

The added value of our study was the evidence that hybrid imaging, combining morphological and metabolic imaging, could improve [¹⁸F]FDG PET/CT interpretation in treatment response evaluation. Notably, the ROC analysis of the combination of the SUV_max and the product of diameters with the highest uptake on the restaging PET scan showed good discriminatory power with an AUC of 0.822. The multivariate analysis also demonstrated that the Cuneo criteria and the product of lymph node diameters could be considered prognostic factors for PFS. To the best of our knowledge, only one other study from 2016 study attempted to consider lymph node diameters (those with short axis) in the restaging PET scan but failed to assess its usefulness [21]. These results could help the visual interpretation of [¹⁸F]FDG PET/CT to clarify indeterminate cases and assign new risk classes. The use of contrast-enhanced CT on PET/CT exams could strengthen our hypothesis to better discriminate lymph node involvement [22].

The unknowledge of HPV-status in more than half of our sample limits our study. To note, HPV status is a well-known prognostic factor in HNSCC. HPV-related HNSCCs, which primarily arise in the oropharynx, have a markedly better prognosis compared to HPV-unrelated [15]. One study suggests that lymph node involvement may take longer to regress in patients with HPV-positive disease [23]. Accordingly, we can speculate that the slightly higher PPV obtained in our study than reported rates, could be related to the unknown HPV status as well as the time interval between the restaging PET/CT and the end of radiotherapy. As reported by some literature data, it is possible that HPV-positive patients with equivocal findings at 3-month [¹⁸F]FDG PET/CT assessment, could avoid a demolitive neck dissection if they undergo a further later PET/CT exam. For positive lymph node disease in HPV-associated patients achieving incomplete response on the 12-week restaging PET/CT, a repeat 16-week PET/CT showed to spare patients from unnecessary surgery [7, 24, 25]. Moreover, the retrospective nature of the study and the relatively small sample size should also be considered.

In conclusion, each PET-based qualitative therapy assessment was statistically related to prognosis, thus demonstrating the reliability of visual point-scale criteria in HNSCC after definitive non-surgical therapy. The new Cuneo score showed the highest specificity and PPV, but the lowest sensitivity and NPV compared to Hopkins and Deauville criteria. Furthermore, the combination of PET data with morphological features could support the evaluation of equivocal cases.