Recurrence at the primary tumor site occurs approximately in 20–50% of patients with HNC and remains the most common cause of treatment failure4,14. Post treatment assessment is crucial because further treatments depend on it15, but an accurate restaging could be a challenging issue due to therapies which can cause relevant tissue changes related to inflammation/fibrosis and consequently distortion of the anatomy.
In this setting, PET/CT is superior to morphological imaging in the assessment of tumor response and detection of recurrence, as demonstrated by two meta-analyses16,17. Moreover, a recent systematic review showed that MTV and TLG could be useful for identifying patients with a higher risk of postsurgical disease progression who could receive early therapeutic intervention to improve their prognosis18.
Many studies showed that failures are often within pre-treatment PET/CT volume. Madani et al. in their series found that most of the local relapses occurred inside the pre-treatment BTV or at its borders (5/9 i.e. 56% of the patients)19. La et al. found 6/7 (86%) local failures within the pre-treatment BTV10 and Soto et al. found a higher rate of local failure within the pre-treatment BTV (8/9 patients, i.e. 89%)9.
Our findings are consistent with these data: failures are completely located in BTVpre in 2/10 patients (20%) and they were mostly inside (61–91%) included in BTVpre (total: 6/10 recurrence within BTVpre) in 4/10 patients (40%). In the latter cases, we could argue that the onset of recurrence was primarily inside the BTVpre and then extended also outside.
Therefore, PET volumes may encompass radio-resistant areas that warrant more aggressive therapy20.
Daisne et al. examined the feasibility of target volume delineation based on the PET/CT uptake and found that, even if PET/CT was the imaging modality that better depicted the real extent of the tumor, it failed to detect a small fraction of the macroscopic tumor extension as CT and MRI6. This finding may explain the occurrence of local recurrence outside the BTV: Soto et al. observed 1/9 (11%) local failure outside the BTV9 and Madani at al. reported 3/9 (33%) local failure outside the BTV19.
In our series, 4 recurrences (40%) were marginal to BTVpre (< 50% of recurrence volume) and three of them were included for less than 15% in BTVpre. In these cases, we could consider that recurrence primarily occurred outside the BTVpre meaning that it could not be fully representative of the real tumor extension.
Madani’s and Soto’s series, and our study as well, are likely to underestimate the true risk of failures outside the BTV because since treatment purpose GTV was defined not only on PET/CT uptake, but also on clinical examination, CT and MRI images9,19.
Thus, PET/CT should not be used as a single imaging modality in target volume definition, but it could be used to identify pre-treatment potentially more aggressive or resistant sub-volumes within the tumor.
Madani et al. reported good results in dose escalating radiation dose on PET avid region within the target volume both in terms of complete response (complete response in 81–86% of patients), 1-year local control (85–87%) and 1-year overall survival (54–82%)19.
PET defined target volume (BTV) is smaller than CT based counterpart (GTV-CT), but the additional information obtained from PET can enlarge the GTV-CT in 25% of cases and this percentage can increase to up to 64% depending on the SUV threshold adopted21. These results confirm the value of using PET/CT as an additional source of information to achieve an adequate target coverage.
Half patients of our series received concurrent cisplatin-based chemotherapy. The two patients with the local recurrence entirely inside the BTVpre did not receive chemotherapy because of comorbidity. In such patients, an intensification of radiotherapy treatment might be a way to improve the therapeutic ratio.
In all the 10 patients, an overlap BTVpre&rec existed between the planning PETpre and the PETrec, which indicates a high probability that recurrence could originate from the BTVpre.
The mean dose received by the 99% of the BTVpre&rec volume was 68.1 Gy, (66.5–69.2 Gy): considering the prescription dose of 70 Gy to the PTV, dose coverage of these volumes was optimal. It is not possible to determine if higher radiation doses to PET/CT based sub-volumes could affect failure rates and outcomes. Our data supports the need to explore this hypothesis because most of failures were mapped inside the BTVpre.
In this regard, Thorwarth et al. published a study on dose painting by numbers (DPBN) in thirteen HNSCC patients exploring the use of PET/CT in identifying radioresistant areas within the tumor 22. For each patient, a conventional IMRT plan, a plan with an additional uniform dose escalation of 10% to the FDG volume (uniDE) and a plan with DPBN were obtained and analyzed according to a map calculated from FMISO-PET. Both dose-escalation approaches were shown to be feasible under the constraint of limiting normal tissue doses to the level of conventional IMRT and for DPBN; the prescription could be fulfilled in larger regions of the target by using DPBN rather than uniDE. Tumor control probability increased from 55.9% with conventional IMRT to 57.7% for the uniDE method. For DPBN a potential increase in tumor control probability from 55.9–70.2% was calculated22.
In our study, recurrences were classified as IF in 6 cases (60%), as EOF in three cases (30%) and as OF in one case (10%). The 6 IF failures were included in high-dose coverage region (D99 between 66.4–68.7 Gy) or inside high gradient of dose region (D99 = 64.8 Gy and D99 = 66.4 Gy); the three EOF failures were only partially included in high dose coverage and high gradient of dose region (D99 between 54.2–61.7 Gy) and the one OF failure was mostly outside high dose coverage region (D99 = 44.7 Gy and D95 = 55.1 Gy).
Our study suggests that patients with IF failure could benefit of a higher dose, and patients with EOF failure could benefit of an improved dose coverage, by an optimization of GTV and CTV delineation. Besides integration of different imaging and clinical data, PET/CT based contouring could also be improved. As shown in Brambilla’s study, radiation oncologists should probably move from visual PET interpretation to an established method for PET volume segmentation and reconstruction algorithms12.
There are some limitations in our study. As reported also in the other literature series, a relatively small number of treatment failures was observed and analysed, limiting the possibility to perform a statistical analysis. Another concern is that failures were heterogeneous with respect to tumor location and stage. Given the number of subjects included, our results should be verified in larger series.
On the other hand, all patients were treated in a very homogeneous way and all PET imaging studies were performed with similar technical characteristics by the same scanner with the patient in treatment position.