38_02

The CTV aimed at 70 Gy (CTV_70) includes the GTV with a 5 to 10 mm margin (if possible) and the whole nasopharynx. The CTV aimed at 60 Gy (CTV_60) covers high-risk local structures (including the parapharyngeal spaces, posterior third of nasal cavities and maxillary sinuses, pterygoid processes, base of skull, lower half of sphenoid sinus, anterior half of the clivus, and petrous tips), and lymphatic regions (including bilateral retropharyngeal nodes, levels II, III and VA). The CTV aimed at 50 Gy (CTV_50) covers the remaining levels IV to VB. The level I nodes can be spared for patients with N0 disease.
The planning target volume (PTV) covers the CTV and the margin needed for systemic and random setup variations. Different centers should gauge the range of variations in their actual practice for determining this margin. With proper immobilization and meticulous care in setup, an expansion margin of 2 mm was used at Pamela Youde Nethersole Eastern Hospital for delineating PTV.
Conventional 2D Treatment Techniques
The classic 2D technique used in Hong Kong is that of Ho (76), which composes two phases. Phase I consists of lateral-opposed facial-cervical fields for the primary tumor and enlarged neck nodes, together with a lower anterior cervical field for the lower cervical lymphatics. Phase II is used after 40 Gy to avoid the spinal cord. This consists of three fields (lateral-opposed plus anterior facial fields) for the nasopharyngeal region and an anterior cervical field for the whole neck. Typical radiologic landmarks and treatment portals are shown in Fig. 38.17. Shrinking field arrangement with cone-down after 50 to 60 Gy should be made whenever possible to maximize protection of critical structures.
P.833



The advantage of the three-field technique in phase II is to minimize the dose to bilateral temporal lobes and temporomandibular joints. However, coverage may be inadequate for tumors with extensive posterolateral extension to the parapharyngeal spaces or caudal extension to oropharynx; supplementary dose via a posterolateral field with avoidance of neurologic structures is given to rectify this deficit (212).
Another 2D technique widely used in other centers is to use lateral-opposed portals throughout. The typical field arrangement used at Mallinckrodt Institute of Radiology (United States) is shown as an example (19) (Fig. 38.18A). Phase I consists of lateral-opposed fields that are angled posteriorly 5 degrees to ensure adequate coverage of the posterior wall of the nasopharynx, while reducing the dose to the contralateral lens and avoiding direct irradiation to the ipsilateral external and middle ear.
The posterior borders of the lateral fields are displaced anteriorly after about 43 Gy to shield the spinal cord. High-energy photons (18 MV) are used in phase II to deliver the last 20 to 25 Gy while diminishing the dose to the mandible and temporomandibular joints. With the shrinking field method, a boost of 5 to 10 Gy is delivered to the nasopharynx through reduced lateral portals for patients with T4 tumors.
The lower neck and supraclavicular fossa are treated with a single anterior field at 2-Gy daily fractions to 50 Gy given dose (Fig. 38.18B). Posterior neck nodes are given a supplementary dose of 5 to 15 Gy with 9-MeV electrons through small lateral fields.
Other options for phase II include arc rotation technique as designed by Wang (219) and the anterior infraorbital oblique fields technique of Fletcher (62).
Three-Dimensional (3D) Conformal Treatment Techniques
Development of computerized 3D treatment plans is an important technical advance for NPC with its typically concave tumor volumes. Several investigators have designed innovative multifield conformal plans; for example, the seven-field technique used at Memorial Sloan-Kettering Cancer Center (United States) (231) and the “Boomerang” technique (Fig. 38.19) used at Peter MacCallum Cancer Institute (Australia) (53). All of the evaluation studies showed better tumor dose coverage while reducing normal tissue dose in comparison with conventional 2D plans (20,95,218).
Leibel et al. (136) from Memorial Sloan-Kettering Cancer Center showed that the target volume underdosed at the 95% isodose level was reduced with 3D plans when compared with 2D plans (7% vs. 22%). With the mean tumor dose increased by an average of 13%, it was estimated that the probability of uncomplicated tumor control would increase by 15%.
However, subsequent analysis of 68 patients for whom this technique was used to deliver a boost of 20 to 26 Gy following phase I conventional 2D treatment for 50 Gy did not show significant improvement; the 5-year L-FFR was 77% and late toxicity grade ≥3 was 25% (231).
More encouraging results were achieved by Jen et al. (85), who retrospectively compared 72 patients treated with 3D conformal technique throughout with 108 patients treated with 2D technique. They reported significant improvement in 3-year L-FFR for T4 (86% vs. 47%), and event-free survival for both stage III (80% vs. 56%) and stage IV (82% vs. 33%). In addition, the Inoltre, il
P.834

incidence of xerostomia at 3 years was significantly reduced, although there were little differences for most other late toxicities.

IMRT Techniques
With the clear advantages of sculpting the high-dose volume with tight dose gradients around the targets, dosimetric studies from different centers all show that IMRT techniques can further improve the conformity of dose distribution for NPC (24,80,89,234). There is little controversy that this technique is advocated for treating NPC if resources permit. With the tight margin now employed, precision in target localization and RT delivery become even more important; all the precautions in treatment preparation and quality control must be strictly followed.
Another potential of interest is the possibility of biologic enhancement by simultaneous modulated accelerated-radiation therapy (SMART) as a new way of delivering accelerated fractionation (AF) schedule, a concept that was first reported by Butler et al. (9) for the treatment of other head and neck cancers with IMRT.
Various methods and dose fractionation schemes for IMRT are being explored by different investigators; Table 38.10 summarizes the key features and the results achieved. Most of the patients treated in these series also received additional
P.835

chemotherapy and/or enhanced RT with boosts or AF. All reported most encouraging early results with local control in excess of 90% at 2 to 4 years.

At the University of California, San Francisco, conventional once-daily fraction was used for all patients. A total dose of 70 Gy at 2.12 Gy/fraction was given to the gross tumor, while the CTV (that included both potential microscopic infiltration and margin for setup error) received 60 Gy at 1.8 Gy/fraction, and the neck with clinically negative findings received 54 Gy at 1.65 Gy/fraction (7,135). Updated results of 118 patients by Bucci et al. (7) confirmed excellent locoregional control of 96%. However, distant failure was still high (28%) despite extensive use of concurrent-adjuvant CRT, the OS was 74% at 4 years. One patient died of torrential epistaxis without tumor recurrence was reported by Lee et al. (135).
At Memorial Sloan-Kettering Cancer Center, treatment was delivered with dynamic multileaf collimation, using seven coplanar 6-MV intensity-modulated fields, positioned every 30 degrees from the posterior and lateral directions. Wolden et al. (230) reported their experience on 74 patients: 59 were treated with AF using the concomitant boost method and 15 by the SMART method. For the latter group, a total dose of 70.2 Gy at 2.34 Gy/fraction was given to the gross tumor, and the “microscopic” PTV received 54 Gy at 1.8 Gy/fraction. The 3-year L-FFR was better than for patients treated by 3D conformal boost (91% vs. 79%), although the difference was not statistically significant.
Further dose escalation with SMART boost in 50 patients with T3 to 4 tumors was reported by Kwong et al. (100) from Queen Mary Hospital (Hong Kong). Their goal was to deliver a total dose of 76 Gy at 2.17 Gy/fraction to the gross tumor. Although the early result for locoregional control was excellent (96% at 2 years), there were serious concerns about late toxicities because 4% of patients had life-threatening bleeding from carotid artery pseudoaneurysm, and another 4% developed temporal lobe necrosis at a median follow-up of 2.1 years.
Kam et al. (90) from Prince of Wales Hospital (Hong Kong) prescribed a total dose of only 66 Gy at 2 Gy/fraction to the gross tumor, but supplemented this with an additional boost to 56% of patients. The 3-year L-FFR for their series of 63 patients was 92%; the incidence of xerostomia (grade 2) was only 23% and hearing loss (grade 3) was 13%. However, other serious late toxicities were substantial; 2% developed osteonecrosis of C1 and C2 vertebrae requiring surgical restoration, 3% developed temporal lobe necrosis, and 23% had endocrine dysfunction.
P.836


Different centers have to work out what is their best affordable technique. Some centers only treat the primary tumor and upper neck with IMRT, while the lower neck is treated with a matching field. However, to avoid dose uncertainty at the match line due to the potential angles of IMRT beams, and to attain better control of dose to all normal tissues at the neck and lung apex, whole-volume IMRT technique is the preferred option.
Figure 38.20 shows the delineation of targets and the technique with nine coplanar beams (6-MV photon) covering the entire region that is currently used at Pamela Youde Nethersole Eastern Hospital. A total dose of 70 Gy at 2 Gy/fraction was given to the PTV for gross tumor, while the PTV for high-risk structures received 61.25 Gy at 1.75 Gy/fraction and the PTV for low-risk structures received 52.5 Gy also at 1.75 Gy/fraction by reducing the field size for the last five fractions. Instead of using SMART, patients with T3-4 tumors are treated with a moderate AF schedule of 2 Gy/fraction, six daily fractions per week (see section Dose Escalation and Altered Fractionation). Figures 38.21 and 38.22 illustrate the tumor targets and dose-distribution plan for patients with early and advanced disease, respectively.
Skillful specification of dose constraints is important for inverse planning. Different dose constraint templates have been designed (80,235). Overstringent control of normal tissue constraints might result in inadequate coverage of tumor targets; optimal balance is critical. An example of dose-constraint guidelines is provided in Table 38.11, which shows the guidelines currently used at Pamela Youde Nethersole Eastern Hospital. Top priority is given to critical neurologic structures, followed by tumor targets, organs with intermediate importance, and finally those with lesser importance. Doses to parotids and cochlea are reduced as much as possible, but without sacrificing coverage of tumor targets. Two sets of acceptance criteria are set, stringent ideals are attempted as far as possible, but safe compromise within tolerance will have to be considered for difficult cases.
Dose Escalation and Altered Fractionation
Excellent local tumor control has also been reported by giving additional boost to patients with early disease treated by conventional 2D technique. The most widely used method is brachytherapy. Different types of applicators have been designed for intracavitary brachytherapy and various isotopes have been used for interstitial treatment. Vikram (216) used permanent implantation with iodine-125. Wang (220) used low dose-rate brachytherapy with cesium sources, while most others used high dose-rate with the advent of after-loading equipment (Fig 38.23).
Table 38.12 summarizes reports on the use of brachytherapy as a boost for dose escalation. Most studies demonstrated that local control up to 90% to 95% could be achieved for T1-2 tumors without excessive late damages. The retrospective comparison by Wang (219) from Massachusetts General Hospital (United States) showed that patients given an
P.837

additional 7 to 10 Gy boost by low dose-rate brachytherapy following 60 to 64 Gy by external beam radiation (EBRT) had significantly higher 5-year L-FFR than those treated by EBRT alone (91% vs. 60%; p <.01). A similar retrospective comparison by Teo et al. (209) showed excellent 5-year L-FFR of 95% in patients given an additional 18 to 24 Gy in three fractions by high dose-rate brachytherapy, but the gain over the group without boost was not statistically significant (95% vs. 90%; p = .17). In addition, the comparison by Ozyar et al. (177) of 106 patients with T1-4 tumors did not show any improvement (86% vs. 94%; p = .23). The exact benefit of dose escalation has yet to be addressed in prospective randomized studies.

One major limitation of brachytherapy is that the dose delivered is adequate only for superficial nonbulky tumors. Furthermore, optimal positioning of the applicators depends both on the individual clinician's skill and the patient's anatomic features. The advent of stereotactic radiosurgery or fractionated radiotherapy, enabling precise delivery of highly conformal RT with rapid dose falloff (Fig. 38.24), provides a valuable alternative for dose escalation. In the update from Stanford University Medical Center by Le et al. (111), excellent 3-year L-FFR of 100% was achieved in 45 patients with T1-4 tumors given a median SRT boost of 12 Gy following conventional RT to 66 Gy. However, despite the addition of CRT in 80% of the patients, the distant failure rate was 31% and OS was 75%. With a median follow-up of 31 months, 7% of patients developed asymptomatic temporal lobe necrosis, 2% developed retinopathy, and 9% developed cranial nerve paresis. The risk was especially high in patients with T4 tumors; late toxicity is a major concern.
Randomized trials on other head and neck cancers have confirmed that altered fractionation is an effective option for improving local control, but data that are specific for NPC are very scant. Separate evaluation is needed because acceleration may not be able to achieve similarly significant benefit for poorly
P.838


P.839


P.840

differentiated carcinoma, and the risk of late toxicity may be more serious because of the proximity of neurologic structures. Extra caution is needed in designing fractionation schedules for NPC.

The use of AF for NPC was first reported by Wang (219). Using 1.6 Gy/fraction twice daily (BID), phase I with lateral-opposed fields was treated to 38.4 Gy, followed by 10 to 14 days rest, and then phase II with arc rotation for another 31.6 Gy with or without a brachytherapy boost of 7 Gy. Retrospective comparison showed that patients with T2-4 tumors treated by AF achieved significantly better 5-year results than those treated by conventional fractionation (CF), both in terms of L-FFR (65% vs. 47%) and disease-specific survival (DSS; 70% vs. 35%), but the benefits for T1 tumors were statistically insignificant. No excessive late damage was incurred.
However, using a similar schedule of 1.6 Gy/fraction BID (with a 4-hour interval) to 67 Gy, Leung et al. (139) observed that the incidence of temporal lobe necrosis was as high as 24% (4/17), compared to none (0/15) in patients treated with CF. In the study by Jen et al. (83) using conventional 2D technique and BID fractions with a 6-hour interval, 76 patients were given 1.2 Gy/fraction to a median dose of 80 Gy, and 12 patients were given AF at 1.6 Gy/fraction to a median dose of 70 Gy. When compared with 134 patients treated with conventional once-daily (QD) fractionation, there was no statistically significant difference in 5-year L-FFR (T1-3 tumors, 93% vs. 86%; T4 tumors, 44% vs. 37%). The results for T4 were especially disappointing. Although the 1.2 Gy/fraction schedule did not incur excessive toxicity, the incidence of symptomatic temporal lobe necrosis was as high as 27% in patients treated with the 1.6 Gy/fraction schedule (82). See section Sequelae of Treatment for further information on the influence of dose fractionation for brain necrosis.
The first randomized trial on AF for NPC by Teo et al. (208) used an uncommon schedule of 2.5 Gy/fraction QD for 8 fractions before randomization to an experimental arm using 1.6 Gy BID for another 32 fractions versus a control arm using 2.5 Gy QD for another 16 fractions. The trial was prematurely terminated because of excessive neurologic toxicities (49% vs. 23%). For this series of 159 patients (62% with T1-2 tumors), the AF arm did not achieve significant improvement in tumor control (5-year L-FFR, 89% vs. 85%).
To minimize the risk of late damage, the more moderate AF schedule of the Danish Head and Neck Cancer Study Group 6–7 Trials using 2 Gy/fraction, 6 fractions per week (175) was tested for NPC by Lee et al. (131). A retrospective comparison of patients irradiated to a total dose of 66 Gy with 2D technique showed that those treated with this AF schedule had significantly higher L-FFR than those treated with conventional five fractions per week. The benefit was significant particularly for T3-4 tumors (87% vs. 62%; p <.01). Furthermore, no significant increase in late toxicity was observed at 3 years (20% vs. 15%).
This schedule was hence used in the subsequent NPC-9902 Trial initiated by the HKNPCSG (132), which aimed to assess the therapeutic benefit by AF and/or concurrent-adjuvant CRT using the Intergroup-0099 regimen for patients with T3-4N0-1M0 nonkeratinizing carcinoma. The trial was stopped early because of slow accrual, but the 189 patients randomized were basically balanced in patient characteristics except for unfavorable gender distribution in the AF-alone arm (90% men). Preliminary results at 3 years showed that AF per se did not show significant improvement in event-free survival (EFS) when compared with CF alone (63% vs. 70%), but AF combined with CRT achieved strongly significant improvement (see section New Attempts with Concurrent Chemoradiotherapy).
Chemotherapy Chemioterapia
Effective systemic therapy is needed for patients with advanced locoregional disease because of the notorious predilection for hematogenous dissemination and the need for further improvement of local control. Although NPC is well known for its
P.841

chemoresponsiveness, review of the clinical trials on the value of chemotherapy shows contradictory results.

Up to the year 2004, there were 11 randomized trials comparing combined treatment versus RT alone published in the English literature. All except the one by Rossi et al. (190) used cisplatin-based regimens; the studied populations included patients with stages II-IVB by the current criteria of the AJCC/UICC Staging System (sixth edition); all patients were treated with conventional RT using 2D technique and CF.
P.842


Of the five trials on induction chemotherapy (14,43,73,81,160), only the trial by the International Nasopharynx Cancer Study Group (81) using cisplatin, epirubicin, and bleomycin achieved significant improvement in EFS (58% vs. 35% at 3 years; p <.01). However, treatment mortality was substantially higher (8% vs. 1%), and no benefit in OS was shown even with longer follow-up (40% vs. 46% at 5 years) (71). The results of adjuvant chemotherapy have been even more disappointing; none of the three trials (32,98,190) achieved significant benefit in any end points.
The first trial that achieved significant survival benefit was the Intergroup-0099 Study (2), using cisplatin (100 mg/m2) on days 1, 22, and 43 in concurrence with RT (70 Gy in 35 fractions) followed by combination of cisplatin (80 mg/m2) and 5-fluorouracil (1 g/m2/day for 96 hours) on days 71, 99, and 127 during the post-RT phase. When compared with RT alone, significant benefit in both 3-year EFS (69% vs. 24%; p <.01) and OS (78% vs. 47%; p = .01) was first reported in 1996, and further confirmed with subsequent update (OS: 67% vs. 37% at 5 years; p <.01) (3). However, controversies remain, particularly regarding the actual magnitude of benefit, because the results of the RT-alone arm were grossly inferior to those achieved by most centers.
The trials on concurrent with or without adjuvant chemotherapy are summarized in Table 38.13. Among the subsequent trials, that by Lin et al. (146) using concurrent cisplatin and 5-fluorouracil, also achieved significant benefit in both EFS and OS, the gain in 5-year survival was 18% (OS: 72% vs. 54%; p = .002). However, in a subsequent reanalysis (147) with retrospective restaging of the accrued patients and more accurate segregation into different risk groups, the benefit was insignificant for the high-risk group.

Two other trials from Hong Kong, that by Chan et al. (11) using concurrent cisplatin and that by Kwong et al. (98) using concurrent uracil and tegafur with or without adjuvant cisplatin-combination, both showed borderline improvement in OS, with survival gain of around 10% (p ≥.06), but no corresponding improvement in EFS (p >.14).
A meta-analysis by Baujat et al. (5), based on updated patient data of 1,753 patients from eight accepted trials (2,11,14,32,43,73,81,98), showed a small but significant benefit by adding chemotherapy: The absolute gain for 5-year EFS was 10% (52% vs. 42%) and for OS it was 6% (62% vs. 56%). The reduction in the pooled HR of death was significant: 0.82, 95% confidence interval (CI): 0.71–0.94; p = .006 (Fig. 38.25). However, the treatment effect was heterogeneous because of significant interaction of timing (p = .03); the survival benefit was essentially confined to the concurrent subset.
More recent reports of confirmatory trials using the Intergroup regimen again showed varying conclusions (Fig. 38.26). The SQNP01 Trial by Wee et al. (222) of patients with stages III-IVB disease supported that the Intergroup regimen could achieve significant improvement in both EFS and OS for Asian patients; the 3-year survival gain was 15% (OS: 80% vs. 65%; p = .006). The preliminary results of the NPC-9901 Trial of patients with N2-3 diseases by the HKNPCSG (124) also showed
P.843

significant improvement in 3-year EFS (72% vs. 62%; p = .027). However, with the favorable results by RT alone and aggressive salvage of treatment failures, no corresponding survival benefit was yet observed (78% vs. 78%), although the improvement in tumor control might translate into significant survival benefit with longer follow-up.

One serious concern about the Intergroup regimen is its efficacy for distant control. Preliminary results of the NPC-9901 Trial (124) showed that the improvement in D-FFR for patients with N2-3 disease was minimal (76% vs. 73%; p = .47). Furthermore, both the series treated by IMRT from University of California, San Francisco (7,135) and that by stereotactic boost from Stanford University Medical Center (111) showed disappointingly high incidence of distant failure (≥25%), despite achievement of excellent locoregional control (≥93%) and extensive use of the Intergroup regimen (≥75% of patients).

Another concern about CRT is tolerance and the risk of toxicities. No treatment-related mortality was observed in the Intergroup-0099 Trial, but both Asian confirmatory trials (124,222) reported 1% mortality by the same regimen. The L'
P.844

NPC-9901 Trial (124) is the first trial that assesses the impact on late toxicity. Besides the expected increase in acute toxicities (84% vs. 53%; p <.01), the CRT arm also had significantly higher incidence of major late toxicities (28% vs. 13% at 3 years; p = .02) (Table 38.14). This was mostly due to increased otologic toxicities (14% vs. 8%), peripheral neuropathy (2% vs. 0%), and endocrine dysfunction (4% vs. 1%). The majority of toxicities were grade 3 in severity. Damages to neurologic structures were rare, but one patient (0.6%) in the CRT arm developed palsy of the last four cranial nerves and died of aspiration pneumonia. On the whole, the Intergroup regimen could be accepted as tolerable, but longer follow-up is needed for full assessment.

Even among American patients, the proportions who could complete the scheduled concurrent and adjuvant chemotherapy of the Intergroup regimen were only 63% and 55%, respectively (2). In addition to concern about the poor compliance and tolerance during the post-RT period, the contribution of the adjuvant component is also questionable, as none of the individual trials (32,98,190) or meta-analyses (5) showed any significant benefits.
However, since the currently available positive data are largely based on the Intergroup regimen, this remains the recommendation most commonly used; but patients should be duly informed that with improving RT technologies, the differential gain in survival might be smaller than initially thought, and there is increased risk of toxicities (both acute and late).
New Attempts with Concurrent Chemoradiotherapy
Building on the current achievement by concurrent CRT, different approaches for further improvement of treatment results have been explored (Table 38.15).
One strategy is to enhance the effectiveness of RT by changing the fractionation from CF to AF. Retrospective studies combining AF with concurrent-adjuvant CRT had shown reasonable tolerability and encouraging preliminary results (86,145,232). In a series of 50 patients (44% stage IVA-B) treated at Memorial Sloan-Kettering Cancer Center using the concomitant boost schedule to deliver 70 Gy in 6 weeks combined with the Intergroup regimen, Wolden et al. (232) showed that 3-year L-FFR of 89% and OS of 84% could be achieved. The study by Jian et al. (86) using a hyperfractionation schedule of 1.2 Gy BID to 74.4 Gy combined with CRT, achieved 3-year OS of 100% for T3 and 63% for T4.
The only randomized trial that attempted to study this combined strategy was the NPC-9902 Trial (132) of patients with T3-4N0-1M0 diseases. The fractionation schedule was 2 Gy per fraction, five fractions per week in CF arms and six fractions per week in AF arms. The preliminary results showed that CRT using the Intergroup regimen combined with AF achieved significantly better EFS than CF alone (94% vs. 70% at 3 years; p = .008) (Fig. 38.26). However, the sample size was small and follow-up was relatively short; the possibility of occult bias and chance effect could not be totally excluded. Hence, the findings could be taken only as hypothesis-generating, and further confirmation is needed.
Another strategy is to enhance the effectiveness of chemo-therapy by changing the timing from concurrent-adjuvant to an induction-concurrent sequence. Although induction chemotherapy per se did not achieve significant survival benefit, data from the meta-analyses by Baujat et al. (5) showed that this could significantly reduce the risk of locoregional failures by 24% and distant failures by 35% (Fig. 38.25). Another advantage is that patients' compliance and tolerance are substantially better during the induction phase than the adjuvant phase (133). This early use of potent combinations of cytotoxic drugs at full dose might be particularly advantageous for NPC with extensive locoregional infiltration, as this could shrink the primary tumor to give wider margin for irradiation (Fig. 38.27).
All five phase II studies using induction-concurrent CRT with CF reported encouraging preliminary results (1,12,87,174,189). Using a combination of cisplatin, 5-fluoruracil, and
P.845

epirubicin as induction chemotherapy and cisplatin in concurrence with RT to 60 Gy, Rischin et al. (189) achieved excellent 4-year results in 35 patients (40% stage IV) with OS of 90%, distant control of 94%, and locoregional control of 97%.

For the most difficult stage IV patients with locoregional disease infiltrating or abutting neurologic structures, a more aggressive approach combining induction-concurrent CRT with AF has been explored at Pamela Youde Nethersole Eastern Hospital. Two different induction regimen have been tested:
P.846

Cisplatin and 5-fluoruracil was used in the first pilot study by Lee et al. (133), and a newer combination of cisplatin and gemcitabine was used by Yau et al. (238). These were then followed by cisplatin in concurrence with RT to 70 Gy using the six fractions per week AF schedule. Given the grave prognosis of this notorious group in the past, 3-year OS of 71% (133) and 76% (238), respectively achieved in the two studies, were very encouraging. Further confirmation of efficacy is warranted.
Persistent/Recurrent NPC
Early Detection and Diagnosis
Despite improving control rate with primary treatment for NPC, local failure remains a problem for patients with advanced T-category. Distinction should be made between persistent disease (tumors that do not completely regress following primary treatment) and recurrent disease (tumors that re-emerge after initial complete regression) because the therapeutic considerations and prognosis are different.
As it takes time for tumors to regress following RT, one difficult decision is when to consider residual tumors as genuine persistence and proceed with salvage treatment. Kwong et al. (96) performed serial biopsies from the nasopharynx in 617 patients and showed that the percentage of positive biopsies dropped spontaneously from 29% in the first week after completion of RT to 12% by the ninth week and then rose again. The 5-year L-FFR was 82% for patients who achieved early histologic remission (<5 weeks), 77% for those with delayed remission (5 to <12 weeks), but only 54% for those with persistent tumors at 12 weeks, despite subsequent salvage treatment. The optimal time for intervention remains uncertain; avoidance of unnecessary overtreatment and excessive delay in treatment are both important, and an observation period of 8 weeks is a reasonable balance commonly used.
Early detection of locoregional failure is crucial for a better chance of salvage. However, both CT and MRI have relatively low sensitivity and specificity in detection of persistent/recurrent disease. Generally, MRI is superior to CT. Gong et al. (68) showed that recurrent tumor exhibited higher signal intensity on T1-weighted spin-echo images, whereas radiation fibrosis showed low or medium intensity in T1 and T2 images. The high-intensity feature, however, is not specific for tumor, and may be seen with radiation edema or infection. Technetium-99 m MIBI SPECT may be a useful tool. Kostakoglu et al. (94) showed that this was superior to MRI performed at 3 to 6 months post-RT in diagnosing complete response. The advent of FDG-PET is a valuable development. Yen et al. (241) compared FDG-PET and MRI in 67 NPC patients 4 to 70 months after completion of RT and showed that FDG-PET was superior to MRI in all aspects, including sensitivity (100% vs. 62%) and specificity (93% vs. 44%).

Preliminary data suggest that circulating cell-free DNA of EBV may be another useful tool for early detection of treatment failure. A longitudinal study by Lo et al. (151) showed that elevation of EBV-DNA titer was noted in patients with relapse up to 6 months before detectable clinical disease. However, this measurement is more sensitive for distant metastases than locoregional recurrence; up to one third of patients with locoregional recurrence did not show elevated EBV-DNA copies (229).
Additional Radiation for Persistent Disease
Brachytherapy has been widely used for locally persistent disease after a full course of EBRT (Table 38.16, part A). Excellent results with 5-year L-FFR in the range of 87% to 95% for patients with initial T1-2 a tumors have been reported (101,109,140,209,246). There is preliminary evidence suggesting that patients with initial T2b tumors could also be effectively treated by brachytherapy (143).
Stereotactic RT is a valuable advance for delivering additional EBRT. Yau et al. (239) studied 755 patients with T1-4 tumors and showed that 7% had positive biopsies 8 weeks after completion of primary RT. The 21 patients treated with fractionated stereotactic RT to a median dose of 15 Gy achieved a 3-year L-FFR of 82%, a result that was very close to corresponding L-FFR of 86% in the contemporary cohort with complete remission, and was substantially better than corresponding L-FFR of 71% in 24 patients treated with high dose-rate brachytherapy to a median dose of 20 Gy.
Reirradiation for Recurrent Disease
Aggressive salvage treatment should be attempted because long-term survival can be achieved for a substantial proportion of patients with early locoregional recurrence and useful palliation for those with extensive disease. However, there is a high risk of normal tissue damage. It is crucial to restrict the irradiation of normal tissue to a minimum.
The most important prognostic factor is the TNM stage of the tumor at the time of recurrence. Thorough restaging,
P.847

including metastatic work-up, is needed. A study of 891 patients with local recurrence by Lee et al. (125) showed that 54% of patients also developed regional and/or distant failures. Another significant factor is reirradiation dose: Most series using EBRT with conventional 2D technique showed that doses ≥60 Gy were associated with better outcome (119,186,221).
A retrospective comparison by Lee et al. (116) of the symptomatic late toxicity rate in 487 patients with two courses of EBRT versus 3,635 patients with one course showed that the summated total biologic dose tolerated (BED-Σ) was higher than that expected with a single-course treatment (BED-1), suggesting partial recovery of normal tissues (particularly for patients with reirradiation after an interval >2 years). Assuming an α/β ratio of 3 Gy, the BED-Σ that incurred 20% toxicity at 5 years was 129% that of BED-1.
Brachytherapy has been widely used for treatment of recurrent NPC (Table 38.16, part B). Early-stage recurrent NPC could be effectively salvaged by brachytherapy alone (101,109). Kwong et al. (101), using interstitial implants with radioactive gold grains, reported a 5-year L-FFR of 63%; complications included headache (28%), palatal fistula (19%), and mucosal necrosis (16%). Law et al. (109), using iridium mold, achieved excellent local salvage up to 89%, but the complication rate was 53%.
The combination of brachytherapy and EBRT is useful, particularly when conventional 2D technique is used. Lee et al. Lee et al. (119) showed that patients reirradiated by combined modes had higher salvage rate than those by EBRT or brachytherapy alone: The 5-year L-FFR was 45%, 32%, and 29%, respectively. Similar pattern of superiority by combined method was reported by other investigators (64,141,186,219,221).

Stereotactic radiosurgery or fractionated stereotactic radiotherapy is another useful tool for retreatment of local recurrence. Control rates ranging from 53% to 86% have been reported (22,44,50,178). For advanced recurrence with extension beyond the nasopharynx, this method will give better dose coverage than brachytherapy. A higher salvage rate by adding stereotactic radiation (17,44,236) as a boost after EBRT has been reported. Although most series reported a low risk of complications, massive hemorrhage with potential fatal outcome has been described (44). To minimize this risk, radiosurgery should be avoided when there is direct tumor encasement of the carotid artery or when a high cumulative dose has already been delivered.
Table 38.17 summarizes the treatment outcome and severe late complications by external reirradiation. Past series using 2D technique achieved 5-year survival rates in the range of 21% to 41%, and the incidence of temporal lobe necrosis ranged from 2% to 27%. The use of 3D conformal radiotherapy showed improving results. Chang et al. Chang et al. (17) showed that none of the patients reirradiated by 3D technique developed temporal lobe necrosis compared with 14% in those reirradiated by 2D technique. Zheng et al. (247), reported a very encouraging 5-year local salvage rate of 71%, but the actuarial rate of late toxicities (grade 4) was still as high as 49%.
Preliminary reports using IMRT for reirradiation show encouraging short-term results. Using IMRT to deliver 68–70 Gy, Lu et al. (155) reported 100% salvage rate without any
P.848

severe late complications in a series of 49 patients with a median follow-up of 9 months. Using IMRT to a median dose of 54 Gy in 31 patients (with or without induction chemotherapy and stereotactic boost), Chua et al. (48) reported a locoregional salvage rate of 56% and late complications (grade 3) of 25% at 1 year. Longer follow-up is clearly needed.
Chemoradiotherapy may also improve treatment outcome for recurrent NPC. Using gemcitabine and cisplatin as induction chemotherapy followed by reirradiation with IMRT in 20 patients (95% rT3-4), Chua et al. (42) reported a 1-year local salvage rate of 75%. In a study of 35 patients (66% rT3-4), Poon et al. (185) reported a 1-year EFS of 42% by concurrent cisplatin followed by adjuvant chemotherapy with cisplatin and 5-fluorouracil.
Surgical Treatment Trattamento chirurgico
For patients who develop local or nodal persistent/recurrent disease without distant metastasis, surgical salvage is an important option to consider. For patients with nodal failure following RT, lymph node involvement is often extensive, Radical neck dissection is the recommended salvage procedure (224). A study by Wei et al. (226) showed that salvage with radical neck dissection could achieve a 5-year nodal control rate of 66% and a disease-free survival of 37%. For those with tumor extending beyond the lymph node confines and involving nearby structures, addition of after-loading brachytherapy to the tumor bed following radical neck dissection might be useful (225).
For selected patients with persistent/recurrent disease localized in the nasopharynx, surgical salvage by nasopharyngectomy is an option. As the nasopharynx is located in the center of the head, adequate exposure for oncologic extirpation of the tumor is a great challenge. A number of approaches have been employed. These include an infratemporal approach from the lateral aspect (60), transpalatal, transmaxillary and transcervical approaches from the inferior aspect (58,166), and an anterolateral approach (227). As all the patients concerned have undergone prior radical RT, the associated morbidities of trismus and palatal fistula are common, but the mortalities associated with these surgical procedures have been low.
Recurrent NPC is frequently located in the pharyngeal recess on the lateral wall. Direct access to this region is essential for complete tumor extirpation. Wei and Sham (228) advocated the anterolateral approach or the maxillary swing approach for surgical salvage of localized failure in the nasopharynx. Following facial incisions and the appropriate osteotomies, the maxilla bone is swung laterally while attached to the anterior cheek flap as one osteocutaneous unit (Figs. 38.28 and 38.29). The nasopharynx with the persistent/recurrent tumor and its vicinity including the paranasopharyngeal region are then widely exposed for oncologic resection. At completion of nasopharyngectomy, the maxilla is returned and fixed to the rest of the facial skeleton with miniplates.

Wei et al. (228) studied 161 patients with salvage nasopharyngectomy using this approach performed at Queen Mary Hospital (Hong Kong) for recurrent NPC following primary treatment by radical RT. Twelve patients also had prior brachytherapy as a salvage procedure. Preoperative assessment showed the tumors of all patients were recurrent stage T1. Negative tumor resection margins, confirmed by frozen section, were achieved in 78% of patients, and the remaining patients had microscopic tumor at the internal carotid artery or the skull base detected during surgery, making further resection impossible. All patients recovered from this anterolateral approach nasopharyngectomy and were discharged. Associated morbidities included trismus of different degrees in 60% and palatal fistula in 25% of patients. Recent modification of the palatal incision has eliminated the problem of palatal fistula (168). In concurrence with other reports, satisfactory long-term results could be achieved for persistent/recurrent tumor that can be completely removed surgically. The 5-year local salvage rate was around 65% and disease-free survival was 54% (57,223).
Results of Treatment Risultati del trattamento
The specific results of various new treatments have been summarized in the preceding respective sections. This section focuses on the overall results in major series of patients treated in the past two decades (Table 38.18).
Two of the representative series from the past, 5,037 patients treated at Queen Elizabeth Hospital (Hong Kong) during 1976–1985 (129) and 378 patients treated at MD Anderson Cancer Center (United States) during 1954–1992 (67,192), both reported very similar results with DSS of around 50% at 5 years
P.849