Why a Gray of protons isn't a Gray of photons — biological effectiveness along the Bragg curve
A Gray (Gy) of absorbed physical dose is a purely physical quantity — energy deposited per unit mass. But cells do not respond to energy deposition uniformly; they respond to the pattern in which that energy is deposited along individual particle tracks. This is why "1 Gy of protons" and "1 Gy of photons" do not produce the same amount of cell killing.
RBE is defined as the ratio of a dose of reference radiation (conventionally 250 kVp X-rays or ⁶⁰Co gamma rays) to the dose of a test radiation (here, protons) required to produce the same biological endpoint — typically a defined level of clonogenic cell survival in vitro:
RBE = D(reference) / D(test radiation), for equal biological effect
Because the endpoint, cell line, dose level, and radiation quality all influence the ratio, RBE is not a single fixed physical constant — it is an experimentally measured, context-dependent number. A proton beam depositing exactly the same physical dose as a photon beam will, in general, produce a larger biological effect, because RBE for protons is greater than 1.
RBE is a ratio of doses at equal biological effect, not a ratio of effects at equal dose — a subtlety that matters when translating laboratory survival curves into clinical prescriptions.
Despite decades of radiobiology showing that proton RBE varies with LET, dose per fraction, tissue type, and position along the beam track, essentially all proton therapy centers worldwide prescribe using a single generic RBE of 1.1. This value was adopted in the early 1990s as a pragmatic, conservative average derived from a meta-analysis of in vitro and in vivo experiments spanning the plateau region of clinical proton beams, where variation is modest.
The rationale is practical: a single multiplicative factor is simple to implement in treatment planning systems, reproducible across institutions, and adequate for the plateau region, which comprises the vast majority of the irradiated volume in a typical spread-out Bragg peak (SOBP). The generic RBE model treats every voxel of the dose distribution identically — physical dose × 1.1 = RBE-weighted (biologically effective) dose — regardless of local LET.
Across the plateau region, where entering protons are still fast and lose energy relatively slowly, LET stays low and the generic RBE = 1.1 assumption is a reasonable approximation, well supported by data. The problem is the last few millimeters of proton range — the distal Bragg peak and the distal fall-off immediately beyond it — where protons decelerate sharply and LET rises by an order of magnitude or more within a very short distance.
In this narrow zone, decades of cell-survival experiments show RBE values well above 1.1, commonly cited in the range of 1.3 to over 1.7, and higher still in some low-α/β tissue models. Because this zone is spatially compressed at the end of every proton beam's range, it is easy for a treatment plan to place it directly inside or adjacent to a critical structure without the planning system's dose display showing any warning.
Linear Energy Transfer (LET) quantifies how densely a charged particle deposits energy per unit path length (keV/µm) as it traverses tissue. Unlike photons, whose secondary electrons deposit energy sparsely and fairly uniformly, protons deposit energy according to the Bethe-Bloch relationship — slower protons transfer energy much more densely than fast ones, which is precisely what happens as they approach the end of their range.
The Bethe-Bloch formula describes the stopping power (energy loss per unit path length, -dE/dx) of a charged particle in matter, and to a good approximation it scales inversely with the square of particle velocity. A proton entering tissue at high energy travels fast and loses energy gradually — hence the long, relatively flat plateau in the depth-dose curve.
As the proton decelerates, its stopping power increases nonlinearly, culminating in the sharp, narrow Bragg peak where nearly all of the remaining kinetic energy is deposited within a few millimeters. LET follows this same trajectory: low and fairly uniform through the plateau, then rising steeply through the last several millimeters before the particle stops entirely. This is a direct physical consequence of decelerating charged-particle transport, not a biological phenomenon — biology only enters once we ask what that dense ionization pattern does to DNA.
LET can rise from roughly 1–2 keV/µm in the entrance plateau to more than 20–30 keV/µm in the final millimeter of proton range — a track-structure change with no equivalent in conventional photon beams.
At low LET, ionization events along a particle track are spaced widely apart relative to the diameter of a DNA double helix (~2 nm) or a nucleosome (~10 nm) — most energy-deposition "hits" occur in isolation, striking different, unrelated locations within the nucleus. At high LET, ionizations bunch closely together along a much shorter stretch of track, so a single passing particle can deposit multiple ionization events within a span of just a few nanometers — directly inside or immediately adjacent to a single DNA helix turn.
This difference in spatial "hit density" — not merely the total number of ionizations — is the central microdosimetric explanation for why high-LET radiation is more damaging per unit of physical dose than low-LET radiation, and it is why LET, rather than dose alone, is the physical quantity most directly correlated with RBE.
Because a proton beam is a mixture of primary protons at many different residual energies and a spectrum of secondary particles (recoil protons, secondary electrons), any voxel in a patient contains a distribution of LET values, not a single number. Treatment planning research therefore typically computes the dose-averaged LET (LETd) — the mean LET across all track segments crossing a voxel, weighted by the dose each segment contributes.
LETd maps computed alongside physical dose maps are the input used by variable-RBE models (see Stage 4) to estimate spatially resolved biological effectiveness, and are an active area of proton treatment-planning research aimed at flagging regions — like distal edges near organs at risk — where the generic RBE = 1.1 assumption is least reliable.
The biological consequence of dense ionization is not simply "more damage" — it is qualitatively different damage. High-LET tracks produce clustered DNA lesions: multiple double-strand breaks (DSBs), single-strand breaks, and base damage within one or two helical turns of DNA. These "complex" or "clustered" DSBs are substantially harder for the cell's repair enzymes to resolve correctly than the simple, isolated breaks typical of low-LET irradiation.
In the low-LET plateau region, an ionizing track typically produces a single, isolated double-strand break, spatially separated from other lesions by tens to hundreds of nanometers. The cell's two principal DSB repair pathways — non-homologous end joining (NHEJ) and homologous recombination (HR) — evolved to handle exactly this kind of damage efficiently. Repair proteins (Ku70/80, DNA-PKcs, MRN complex, RAD51 and others) can locate, bind, and rejoin a clean break with high fidelity within minutes to hours, and the vast majority of such breaks are repaired correctly, with no lasting genomic consequence.
Near the distal Bragg peak, a single dense proton track can produce two or more DSBs within one or two DNA helical turns, along with adjacent single-strand breaks and oxidized/abasic base lesions — a "locally multiply damaged site." Repair machinery faces several compounding problems here: multiple competing break ends in close proximity increase the chance of end mis-joining (producing deletions, translocations, or other chromosomal aberrations); the local chromatin is destabilized by clustered base damage that must be independently resolved by base-excision repair enzymes; and repair complexes can physically collide or stall when trying to act on immediately adjacent breaks. The net effect is slower, less accurate, and more frequently unsuccessful repair — mechanistically explaining why the same physical dose produces more cell killing when LET is high.
It is the spatial clustering of lesions — not the total energy deposited — that overwhelms DNA repair fidelity. This is the accepted mechanistic basis for why RBE increases with LET.
The link between clustered damage and clinical RBE is not merely descriptive — it is the physical basis for every predictive RBE model in use today (see Stage 4). Cell lines and tissues with a low intrinsic α/β ratio (slower-repairing, more "shoulder" in their survival curve — typically late-responding tissues such as central nervous system, spinal cord, and optic pathway) show proportionally larger RBE increases at a given LET than high-α/β, fast-repairing tissues such as most tumors and early-responding normal tissues. This is because low-α/β tissue is already more reliant on high-fidelity repair to survive irradiation, so a shift toward less-repairable clustered damage disproportionately increases cell killing.
Treatment planning systems convert physical dose into "RBE-weighted dose" (units: Gy(RBE), formerly "cobalt Gray equivalent" or CGE) so that clinical prescriptions remain comparable to photon experience. The generic model multiplies every voxel by 1.1. Variable-RBE models instead compute a spatially resolved RBE from local dose-averaged LET, producing a weighted-dose curve that tracks the generic model through the plateau but diverges sharply at the distal edge.
The generic model is deliberately simple: RBE-weighted dose at every point equals physical dose multiplied by a fixed 1.1, independent of LET, tissue, dose per fraction, or depth. Plotted alongside the physical depth-dose curve, this produces a weighted-dose curve of identical shape, uniformly scaled up by 10% everywhere — including the plateau, the Bragg peak, and the distal fall-off alike.
This approach has the virtue of consistency and decades of clinical outcome data behind it, but it implicitly assumes the distal few millimeters behave no differently than the entrance plateau — an assumption radiobiology data do not support.
Several phenomenological models estimate a spatially variable RBE from local dose-averaged LET and tissue radiosensitivity (α/β from the linear-quadratic survival model), fitted to compiled in vitro survival data:
• Wedenberg et al. (2013): RBE increases linearly with LETd and, critically, this increase is inversely related to the tissue α/β ratio — low-α/β (CNS-like) tissue shows a much larger RBE rise than high-α/β tissue at the same LET.
• McNamara et al. (2015): a refined linear-quadratic formulation fitted to an expanded experimental dataset, now among the most widely used variable-RBE models in proton research and increasingly in commercial planning-system research modules.
Both models agree qualitatively: RBE stays close to 1.1 through most of the plateau, then rises — sometimes substantially — through the last few millimeters before the distal edge, exactly where LETd increases most sharply.
Because RBE rise is inversely coupled to α/β, the same physical dose distribution can imply a meaningfully larger biological overshoot in low-α/β tissue (e.g. brainstem, spinal cord) than in a typical tumor.
When the generic-weighted and variable-weighted dose curves are plotted together, they are nearly indistinguishable through the plateau — confirming the generic model is a reasonable approximation where most of the target volume lies. The divergence appears only in the final few millimeters of range, where the variable-RBE curve climbs above the flat generic curve, sometimes by 20–50% or more in the most sensitive tissue models, before both curves fall to zero at the true physical end of range.
The clinical stakes of variable RBE are concentrated in one specific, avoidable scenario: a treatment plan that intentionally or unavoidably stops the proton beam's distal edge inside, or immediately adjacent to, a critical organ at risk (OAR). In that narrow zone, the true biological dose may exceed the planned physical-dose-times-1.1 estimate — a mismatch that has been implicated in reported cases of unexpected toxicity.
Modern intensity-modulated proton therapy (IMPT) plans can, in principle, place a beam's distal fall-off edge extremely close to a serial, dose-limiting organ — precisely because the sharp Bragg peak lets planners "paint" dose right up to a boundary. That geometric precision is proton therapy's core advantage. But it is also where the generic RBE = 1.1 assumption is weakest: the distal few millimeters are exactly where LET, and therefore true RBE, rises most steeply.
Several published case series and modeling studies have raised concern that unexpected brainstem or optic-pathway toxicity in pediatric and adult proton patients may be partly attributable to distal-edge RBE effects not captured by the generic model, particularly when the distal edge was placed directly within or at the boundary of these structures.
The concern is not that proton therapy is unsafe — its physical dose conformality is a major advantage — but that the distal few millimeters deserve explicit biological scrutiny during planning, not just physical-dose evaluation.
The most robust, immediately actionable mitigation does not require any new RBE model at all: avoid planning beam arrangements whose distal fall-off terminates inside or at the surface of a critical, low-α/β structure. Where feasible, planners can select beam angles so that the higher-LET distal edge is arranged to fall in less critical tissue, or so that the OAR is only ever traversed by the lower-LET plateau region of one or more beams — accepting a small increase in entrance/plateau dose (already well-modeled at RBE ≈ 1.1) in exchange for avoiding the poorly-characterized distal zone in the sensitive structure altogether.
A more advanced, actively researched approach directly optimizes the treatment plan against LETd or a variable-RBE-weighted dose objective, not physical dose alone — sometimes called "LET-painting" or biologically robust optimization. In this approach, the optimizer is penalized for placing high-LETd, high-variable-RBE regions inside OAR contours, effectively steering the distal edges of individual beams away from sensitive structures during the optimization itself, rather than checking for the problem only after a physical-dose-only plan is finished.
This remains a research and early-clinical-adoption area rather than a fully standardized part of routine planning, but it represents the logical next step beyond a single fixed RBE = 1.1: treating biological effectiveness as a spatially resolved planning objective, on par with physical dose conformality.