3.1 Patient selection

Appropriate patient selection is the first critical step for ensuring good outcomes in re-irradiation cases. Table 1 summarizes postulated and known patient factors which may determine prognosis and outcomes from available published literature [23-31].

Factors found to be significant in influencing treatment outcomes include age at recurrence, with an OS difference found between patients of younger and older than 46 to 50 years old and hazard ratio ranging from 1.02-1.48, depending on the source paper [23-31]. Another important factor is the patient's Karnofsky performance status (KPS), with a hazard ratio of 2.65 for those with KPS ≤70 compared to those with KPS >70 [23-31]. Whether the patient suffered from any grade 3 or higher late toxicities from initial RT was also found to be a significant factor, with a hazard ratio ranging from 1.90 to 2.36 [23-31]. This may be correlated with their performance status, nutritional status, and overall fitness and health status during treatment, which may reflect their ability to tolerate and complete re-irradiation with minimal unscheduled breaks.

Disease characteristics at recurrence are similarly important in influencing prognosis. Patients with early recurrent rT0 to rT2 disease show a 5-year OS rate range of 73.2%-75.8% with re-RT treatment. However, this can decrease to 32.4%-35.1% for rT3 to rT4 disease. Tumor volume of the recurrent disease is another important factor influencing prognosis. Various studies have used different volume cut-offs, ranging from 30 cm³ to 38 cm³, with the hazard ratio for larger volumes of recurrent disease ranging from 1.57 to 1.96. For example, taking the higher volume cut-off of 38 cm³, the 5-year OS rate range for tumor volume ≤38 cm³ was found to be 48.7%-55.9%, whereas for tumors >38 cm³, it can decrease to 15.2%-30.1%. Furthermore, patients with rT3-4 tumors or gross tumor volume (GTV) >30 cm³ may exhibit a significantly higher rate of mucosa necrosis than patients with rT1-2 stage tumors (36.8% vs. 26.1%, P = 0.04) or with GTV >30 cm³ (38.7% vs. 23.0%, P < 0.01) [26]. NPC tumor histology classification also plays a role in prognostication, with WHO type III tumors generally having a better prognosis with retreatment compared to types I and II, due to their more radiosensitive tumor biology.

Other factors that influence outcomes for re-treatment cases include the disease-free interval (DFI) before recurrence, with a hazard ratio of 1.05 for patients whose disease recurred within 25 months. This is likely to be a reflection of the biology of the tumor, suggesting increased radio-resistance. High plasma EBV DNA titer at recurrence is another negative prognostic factor. A recent study found that patients with advanced clinical T and N stage locally recurrent NPC had higher levels of pre-retreatment plasma EBV DNA, and this was further shown to be significant in locoregional recurrence-free survival (LRRFS) (54.2% vs. 75%, P < 0.001), implying such patients were at greater risk of subsequent relapses [31]. Lastly, high previous RT doses delivered to critical OARs could also lead to poorer outcomes after retreatment, likely secondary to limitations on retreatment dose deliverable.

There have been a number of prognostic models published that aimed to correlate the above patient characteristics with survival outcomes and treatment-related morbidity and mortality, in order to more objectively ensure appropriate patient selection [25, 32]. These models used weighted hazard ratios of various parameters to stratify the patients into 2 to 3 risk subgroups, with re-irradiation recommended for the low- to intermediate-risk groups, but more clinical evaluation and caution are needed before offering re-irradiation for the high-risk groups. These prognostic tools can be applied clinically to aid in offering suitable treatment recommendations.

3.2 Re-irradiation doses and fractionation

Currently, there is still no established definitive gold standard dose and fractionation for NPC re-irradiation cases. The available studies are highly variable in terms of timing, technique, and radiation doses; as well as in their use of concurrent chemotherapy and drugs used, and many are retrospective or single-arm, making comparison across different studies difficult. However, available literature largely points to the need for a total EQD2 (Equivalent dose in 2 Gray fractions) dose of at least 60 Gy or more to achieve adequate local control [21, 33-36]. Lee et al. [34] found that the hazard of local failure increased by 1.7% per Gy₁₀ biological effective dose (BED) drop during re-irradiation (assuming an alpha/beta ratio of 10). For rT1 to rT2 recurrences, the 5-year local control rate for patients given >70 Gy₁₀, 60 Gy₁₀-70 Gy₁₀, and <60 Gy₁₀ doses were 40%, 35%, and 14%, respectively. When compared with the group given 60 Gy₁₀-70 Gy₁₀ doses, the local salvage rate in those given <60 Gy was significantly inferior (P = 0.001), but the superiority in local control of those given >70 Gy failed to reach statistical significance (P = 0.229). A similar trend was also observed for patients with rT3 disease [34].

On the other hand, total EQD2 doses greater than 70 Gy (assuming an alpha/beta ratio of 3) could be associated with increased late toxicities, most commonly mucosal necrosis (30.8% to 50%), trismus and dysphagia (∼17.3% to 18.9%), temporal lobe necrosis (17.3% to 22%), and hemorrhage (11.5% to 31%) [34, 36, 37]. Thus, it appears that the optimal EQD2 dose range should lie within a range of 60 Gy to 70 Gy.

The other issue to consider is the optimal fractionation for dose delivery. For most irradiation treatments, including that of initial de-novo NPC treatment, the standard of care is dose delivery at 2 Gy to 2.2 Gy per fraction. In recent years, attention has turned to the consideration of hyperfractionated dose delivery in head and neck re-irradiation cases, including NPC. Hyperfractionated regimens involve delivering a small fraction size of 1.1 Gy to 1.5 Gy, given twice a day with a break of at least 6 hours. This is radiobiologically advantageous for late responding organs such as the spinal cord, brain, mucosa, and other OARs, thus reducing late toxicities of re-irradiation, while still allowing sufficiently high curative doses to be delivered. There have been a number of clinical studies using hyperfractionated radiation doses in head and neck malignancies demonstrating the feasibility of this approach, with no differences in local disease control and acceptable acute and late adverse effects [38-45]. In addition, there are a number of studies looking at locoregional control and late sequelae using hyperfractionation in recurrent head and neck cancers [46-48]. In these studies, a median dose of 60 Gy to 68 Gy was delivered in fractions of 1.2 Gy to 1.5 Gy twice a day, with grade 3 and above toxicities ranging from 14% to 29%.

Lee et al. [49] conducted a non-randomized prospective study comparing hyperfractionation with standard fractionation in locally advanced recurrent NPC. All patients in the study received induction chemotherapy with cisplatin and gemcitabine, as well as weekly cisplatin during the concurrent phase. The hyperfractionated dose delivered was 64.5 Gy at 1.2 Gy per fraction twice a day, while the historical cohort received a standard dose of 60 Gy at 2 Gy per fraction daily. The median local failure-free survival (LFFS) showed a trend in favor of hyperfractionation (28.2 vs. 16.6 months, P = 0.164), though there was no significant difference in overall survival. Grade 3 and above late toxicities were similar, with treatment-related hemorrhage showing a marginally significant reduction in the group receiving hyperfractionated irradiation. Another retrospective review by Karam et al. [50] also showed that re-irradiation for recurrent NPC, delivered mostly with hyperfractionated IMRT, resulted in durable local disease control of 46% with acceptable Grade 3 and above late toxicity of 37%. Even with the use of 3-dimensional conformal RT (3D-CRT), Cho et al. [51] showed in their small case series study that they were able to deliver doses of 59.4 Gy to 69.2 Gy in 1.1 Gy to 1.2 Gy per fraction doses twice daily, achieving good local control with minimal late sequelae.

Another novel approach is the use of stereotactic body radiotherapy (SBRT) to target recurrences with high doses of hypofractionated irradiation. SBRT is usually only suitable for small recurrent tumors, where high doses can be delivered safely without compromising the nearby critical organs. This has the theoretical radiobiological advantage of being more able to overcome clones of radioresistant tumor cells compared to conventional fractionation, which is thought to be one cause of local disease recurrence. Doses used in the literature range from single fractions of 11 Gy to 14 Gy, 18 Gy in 3 fractions, 30 Gy in 5 fractions, to 48 Gy in 6 fractions. In general, most studies had small patient numbers and short follow-up periods of 1 to 3 years. Reported 3-year local control rates ranged from 52% to 89.4% with generally acceptable late toxicity rates [52-56]. However, a number of late toxicities were severe, with some study patients developing fatal hemorrhage and brainstem necrosis.

In conclusion, the optimal EQD2 dose range delivered in NPC re-irradiation should lie within a range of 60 Gy to 70 Gy, preferably with hyperfractionated doses of 1.1 to 1.5 Gy delivered twice a day at least 6 hours apart. SBRT remains a promising option, with more studies needed to determine the ideal re-treatment volumes and OAR doses at retreatment, in order to reduce significant late toxicities.

3.3 Normal tissue dose constraints

The head and neck region contains a large number of radiosensitive organs which can couple with high doses of radiation required to treat de-novo and recurrent NPC and may lead to debilitating normal tissue complications after re-irradiation. Thus, it is particularly important to determine suitable dose constraints to apply to these OARs, focusing especially on vital structures that may result in severe morbidity if damaged. These include the brainstem, cord, optic chiasm, optic nerve, internal carotid arteries, and the temporal lobes.

For late toxicity endpoints, tissues vary considerably in their capacity to recover from radiation damage. Organs such as the skin, mucosa, spinal cord, and nerves have the ability to partially recover from subclinical injury to a magnitude dependent on the organ type, size of the initial dose, as well as the interval between irradiation courses [57]. Table 2 summarizes the recommendations from literature for critical head and neck OARs. The brainstem, spinal cord, and optic chiasm were designated as strict constraints while the optic nerves, internal carotids, and temporal lobes were designated as desirable dose constraints [57-63]. A detailed discussion of dose constraints is beyond the scope of this paper.

Re-irradiation for NPC presents unique challenges pertaining to OAR toxicities. Due to the narrow therapeutic dose range in head and neck retreatment, both acute and late toxicities may be more severe and require careful anticipation and management. Reviewing available literature, we found that re-irradiation was associated with a 7.9% to 16.6% of grade 3 and above acute toxicities, most commonly mucositis, with a smaller proportion of patients experiencing severe xerostomia [23-25, 28]. Also, the most common acute adverse effects were grade 1-2 mucositis and xerostomia, as well as otitis media.

The data concerning late toxicities are more sobering in comparison to the acute toxicity data [4, 21, 23-25, 28, 34, 64]. The common late toxicities were hearing loss (7.5% to 22.2%), significant trismus (8.6% to 22.2%), and xerostomia (9% to 43.2%). Radiation-induced brain damages such as temporal lobe necrosis or other brain changes seen on magnetic resonance imaging (MRI) ranged from 3.1% to 28.5%, and the presentation in such patients varied from asymptomatic to headaches and encephalopathy. Among the more debilitating late toxicities were persistent mucosal necrosis (ranging from 16% to 40.6%), cranial neuropathies (3.4% to 12.6%), and repeated epistaxis from atrophic mucosa. On average, a range of 16% to 40.1% of retreatment patients was found to have developed one or more of these late toxicities. The most common cause of death resulting from retreatment-induced toxicity was massive hemorrhage which ranged from 2% to 23.2% and could be a reflection of case selection and early pre-emptive measures.

3.4 Modalities of RT delivery

Depending on machine availability at each center, RT can be delivered via 3D-CRT or through more advanced photon treatments such as IMRT, volumetric arc therapy (VMAT) or tomotherapy, or via proton therapy. Regardless of the planning method, it is crucial to ensure that an experienced dosimetrist and planning team are on board, as well as stringent quality assurance checks before dose delivery.

For cases of NPC re-irradiation, the principle should be to achieve the most conformal plan, thus allowing desired dose coverage to the target volume while achieving good OAR sparing. In this aspect, the current advanced photon delivery techniques are superior to 3D-CRT techniques and should be the method of choice in such retreatment cases. The trade-off is increased integral dose, especially with VMAT techniques, as compared with 3D-CRT.

Proton therapy and heavy ion therapy, which both exhibit sharp dose fall-off beyond the penetration range, may be used to further reduce OAR dose without compromising target coverage, potentially reducing complications caused by re-irradiation and improving patients’ quality of life [65]. There have been a few retrospective reviews investigating the outcomes and toxicities of these techniques in head and neck re-irradiation over that of IMRT, with encouraging results [66-69]. For example, Hu et al. [70] achieved 1-year OS and PFS rates of 98.1% with carbon ion therapy with aggregate late grade 3 and 4 toxicities of <10%. Longer follow-up periods and larger sample sizes would further establish the role of these irradiation modalities in NPC retreatment.

3.5 The role of chemotherapy in re-irradiation

In cases of NPC re-irradiation, the role of chemotherapy remains to be clearly defined. As with initial radical treatment, possible timings of chemotherapy would be as induction, concurrent with irradiation, or less frequently, as adjuvant treatment.

The main role of concurrent chemotherapy with irradiation is to act as a radiosensitizer. In general, a review of the literature shows that there is uncertainty with regards to its utility in NPC re-irradiation cases. The chemotherapeutic regimen used was largely platinum-based, utilizing either cisplatin or carboplatin alone, or combined with 5-fluorouracil and/or docetaxel. A number of retrospective studies have shown that the use of concurrent chemotherapy together with re-irradiation either showed no benefit or conversely exhibited poorer local control than those on RT alone [71, 72]. Hua et al. [24] showed that chemotherapy, administered either as induction or concurrent, had prognostic significance in univariate analysis, but not in multivariate analysis. However, a caveat in interpreting these studies would be that a majority of these studies are retrospective in nature and exhibit some degree of selection bias. Most of the patients who received chemotherapy had more adverse prognostic factors, more advanced rT stage, or bulky tumors.

Chemotherapy may however have a role as induction treatment before re-irradiation for selected patients. It can be used to downstage tumor recurrences, allowing RT to be delivered more safely, to reduce high dose volume or to decrease the dose delivered to critical OARs. It can also serve as a bridge to RT treatment for patients with tumor recurrences occurring within 9 months to 1 year; where OAR recovery from the initial irradiation may not be sufficient for a second round of RT to be given. A number of studies looking at induction chemotherapy before re-irradiation have generally shown good response rates, especially with the combination of cisplatin and gemcitabine. Chua et al. [73] observed a 75% partial response (PR) rate with 3 cycles of cisplatin and gemcitabine, while Lee et al. [49] observed a PR rate of 70% with cisplatin/gemcitabine and 40% with cisplatin or carboplatin with 5-fluorouracil. The addition of chemotherapy at any timepoint, however, is associated with significant toxicities [28, 35, 74, 75]. Incidence of grade 3 and 4 toxicities such as temporal lobe necrosis, endocrine toxicities, mucosal necrosis, cranial nerve toxicities, and swallowing dysfunction necessitating long-term enteral feeding were all increased compared to re-RT alone, regardless of retreatment with 3D techniques or with IMRT, thus careful patient selection is paramount.

3.6 The role of immunotherapeutic advances in re-irradiation

There has also been great interest recently in the role of immunotherapy in NPC, with promising outcomes [76]. In the recurrent and metastatic settings, the use of immune checkpoint inhibitors alone [77, 78], or combined with chemotherapy [79], have shown good objective response rates of 20%-25% and promising antitumor activity in early phase trials, with tolerable toxicities.

Of relevance to the topic of immunotherapy in locally recurrent cases for retreatment, the National Cancer Institute (NCI) held a clinical trial planning meeting in Arizona, USA in 2018, where a randomized phase II trial of adjuvant immunotherapy following salvage treatment for locally recurrent NPC was proposed [80]. The implementation and outcomes of this proposed trial are eagerly awaited and may represent a shift in how we manage these challenging patients.

ETHICS APPROVAL AND CONSENT TO PARTICIPATE

Not applicable.

CONSENT FOR PUBLICATION

All authors consent to publishing the paper.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflict of interest.