KEY POINTS
- This systematic review and random-effects meta-analysis included 62 studies involving 3,278 patients undergoing reirradiation for recurrent or second primary head and neck cancer. The analysis comprised 14 prospective and 48 retrospective studies, identified through a PRISMA-guided search covering publications from 1995 onward, with the search completed in March 2025. The primary endpoint was osteoradionecrosis (ORN) of the mandible and/or maxilla occurring at least 90 days after reirradiation.
- Reirradiation techniques and treatment settings were heterogeneous. External beam radiotherapy was reported in 2,546 patients (77.7%), while 684 (20.9%) received brachytherapy; photon RT predominated, with smaller proton, heavy-ion, and electron cohorts. The pooled median initial RT dose was 65 Gy, the median reirradiation dose was 60 Gy, and the median interval between courses was 29.5 months. Among patients with available treatment-intent information, 1,049 received adjuvant and 2,156 definitive reirradiation.
- The pooled observed incidence of mandibular and/or maxillary ORN was 3.6% (95% CI 2.8–4.6%), with moderate between-study heterogeneity (I²=37.5%). The corresponding 95% prediction interval was 0.9–13.4%, indicating considerable variation in expected event rates between clinical cohorts. When restricted to patients evaluable for late adverse events, the pooled incidence increased to 4.0% (95% CI 3.0–5.3%), with a prediction interval of 1.0–14.5%.
- Anatomical definitions influenced the estimated incidence. When all reported osseous sites were considered, including the skull base, cervical vertebrae, and hyoid bone, pooled ORN incidence increased to 4.7% (95% CI 3.6–6.1%). Restricting analysis to studies explicitly reporting mandibular ORN yielded 3.5% (95% CI 2.7–4.6%). Fourteen of the 62 studies documented ORN outside the mandible, emphasizing that jaw-specific estimates do not capture the entire spectrum of skeletal necrosis after head and neck reirradiation.
- Severe complications were common among cases with detailed toxicity grading. Across the included studies, 89 ORN cases were reported, but individual grades were available for only 42. Among these, 35/42 (83.3%) were grade ≥3, including 20 grade 4 events (47.6%). Another 40 cases were reported only within grouped toxicity categories, and seven were ungraded. The predominance of severe events may reflect underrecognition or underreporting of early-stage ORN rather than the true severity distribution among all affected patients.
- Adjuvant reirradiation was the only significant study-level determinant identified in meta-regression. A higher proportion of patients receiving postoperative rather than definitive reirradiation was associated with greater observed ORN incidence (β=1.091, 95% CI 0.256–1.926; p=0.010). In contrast, no significant associations were identified for reirradiation dose (p=0.355), interval between RT courses (p=0.620), or treatment modality, including protons (p=0.400) and heavy ions (p=0.207). These aggregate associations cannot establish individual risk factors or demonstrate that one modality is safer than another.
- The pooled estimate is not a formal cumulative incidence of ORN. Median follow-up was only 26 months among surviving patients and 20 months across the overall population, while individual time-to-event data and competing-risk information were largely unavailable. ORN was usually a secondary endpoint, diagnostic and grading criteria varied, and funnel-plot asymmetry was substantial (Egger's test p<0.0001), although this may reflect heterogeneity and low event rates rather than publication bias alone. Consequently, late events in long-term survivors may be underestimated.
CLINICAL TAKEAWAY
Mandibular and maxillary osteoradionecrosis was reported in approximately 4% of patients undergoing head and neck reirradiation, providing a useful reference for toxicity counseling and multidisciplinary treatment planning. However, incomplete surveillance, competing mortality, and substantial variation between studies mean that this estimate should not be interpreted as an individual patient's cumulative lifetime risk, and the evidence does not establish reliable modality-specific risk differences.