KEY POINTS
- This PRISMA-based systematic review included 51 studies and 10,789 patients treated with curative-intent head and neck radiotherapy between 2013 and 2025. Most studies used modern IMRT-based techniques; 42 of 51 enrolled exclusively IMRT-treated patients.
- Contralateral parotid mean dose was the most frequently and consistently validated dosimetric predictor, evaluated in 16 studies and supported by moderate-certainty evidence.
- Multiple studies reported a marked increase in moderate-to-severe xerostomia when contralateral parotid mean dose exceeded approximately 24–26 Gy. Conversely, keeping mean dose ≤20 Gy was associated with better preservation of salivary function.
- For bilateral parotids, mean dose around 26 Gy remained clinically relevant. One study reported approximately a fivefold higher late xerostomia risk when bilateral parotid mean dose reached or exceeded 26 Gy.
- The review also identified contralateral submandibular gland mean dose as a moderate-certainty predictor. Oral cavity mean dose and bilateral parotid mean dose were supported by lower-certainty evidence, with risk signals generally appearing around the 25–35 Gy range.
- Emerging work suggests that parotid ducts, superficial lobes and stem-cell-rich regions may add predictive information beyond whole-gland mean dose, but definitions and contouring methods remain insufficiently standardized for routine constraints.
- Only 17 of 51 studies were rated high quality, and xerostomia assessment methods and follow-up timing varied substantially. The authors therefore caution that the recurrent numerical thresholds should be treated as planning references rather than universally validated constraints.
CLINICAL TAKEAWAY
Modern evidence still supports aggressive contralateral parotid sparing, with ≤20 Gy a reasonable planning goal and 24–26 Gy representing an important risk range. Submandibular gland and oral cavity dose may refine risk further, but current evidence does not justify replacing established parotid constraints with new universal cutoffs.