Linear-quadratic radiobiology: comparing fractionation schemes on a common dose scale
Radiotherapy has delivered curative treatment in small daily doses for nearly a century, long before the biological reason was fully understood. Splitting a large tumoricidal dose into many small fractions exploits a simple asymmetry: normal tissue repairs sub-lethal DNA damage between fractions more efficiently than tumor cells do, widening the gap between tumor control and normal-tissue injury — the therapeutic ratio.
Conventional fractionation rests on four radiobiological principles, classically summarized as the "4 Rs": Repair of sub-lethal damage between fractions (mostly benefiting normal tissue, which typically repairs faster and more completely than tumor cells), Repopulation of surviving clonogens during the course (a liability for slow-growing but rapidly-accelerating tumors if treatment is prolonged), Redistribution of cells into radiosensitive phases of the cell cycle between fractions, and Reoxygenation of previously hypoxic — and thus radioresistant — tumor regions as oxygenated cells nearer the surface are killed off.
A fifth "R," intrinsic Radiosensitivity, is often added to describe the baseline genetic differences in how cell lines respond to a given dose. Together these mechanisms explain why delivering 70 Gy in one exposure would be catastrophic for normal tissue, while the same total dose split into 35 daily 2 Gy fractions is tolerable and often curative.
Standard fractionation (1.8–2.0 Gy/day, 5 days/week) became the historical default largely because it empirically balanced tumor control probability against late normal-tissue toxicity — decades before the linear-quadratic model gave it a quantitative footing.
Tissues are broadly classified by how quickly radiation damage manifests. Early-responding tissues — most tumors, skin, and mucosa — are populated by rapidly dividing cells; damage appears within days to weeks as cells fail to divide. Late-responding tissues — spinal cord, lung, kidney, and other slowly-proliferating organs — show damage only months to years after treatment, once damaged stem/stromal cells that were not actively dividing eventually attempt division or the tissue microvasculature deteriorates.
This distinction matters clinically because late effects (fibrosis, necrosis, myelopathy) are typically irreversible, while early effects (mucositis, dermatitis) usually resolve with time and supportive care. Fractionation schedules are therefore designed with late-responding organs-at-risk as the binding constraint on total dose and fraction size, since exceeding their tolerance produces damage that cannot be undone.
Empirical fractionation worked clinically for decades, but clinicians needed a way to compare non-standard schedules — different total doses, different fraction sizes, different numbers of fractions — on equal footing. Simple total physical dose is a poor comparator because 60 Gy in 30 fractions and 60 Gy in 20 fractions are biologically very different treatments despite an identical number of Gray delivered.
The linear-quadratic (LQ) model, developed through the 1970s–80s from cell-survival curve data, provided exactly this quantitative bridge. It captures cell kill as a function of both dose per fraction and total fractions, and forms the basis for the Biologically Effective Dose (BED) calculations explored in the following stages — letting radiation oncologists convert any fractionation scheme onto one common scale.
The linear-quadratic (LQ) model describes the fraction of cells surviving a radiation dose as SF = exp(−αD − βD²). Its two terms represent two physically distinct kinds of lethal damage, and the ratio between their coefficients — α/β — is the single number that determines how sensitive a tissue is to changes in fraction size.
The α coefficient describes lethal damage produced by a single radiation track directly causing an unrepairable double-strand DNA break — its contribution to cell kill scales linearly with dose (−αD). The β coefficient describes lethal damage that results from two separate, individually sub-lethal lesions (each potentially repairable on its own) that combine into a lethal lesion if they occur close together in space and time — its contribution scales with the square of dose (−βD²), because the probability of two independent hits interacting rises quadratically.
At low doses per fraction, the linear α term dominates and cell kill is roughly proportional to dose. As dose per fraction rises, the quadratic β term grows much faster, so survival curves bend downward — the "shoulder-then-bend" shape seen on a semi-log plot of surviving fraction against dose.
The α/β ratio is the dose (in Gy) at which the linear and quadratic components of cell kill contribute equally (αD = βD² when D = α/β). Below this dose, killing is dominated by single-hit (α) damage; above it, multi-hit (β) damage dominates.
Most tumors and early-responding (acutely reacting) normal tissues have high α/β ratios, typically around 10 Gy — their survival curves are relatively straight/shallow on a log plot, meaning they are not very sensitive to changes in fraction size; total dose is the main driver of effect. Late-responding normal tissues (spinal cord, lung parenchyma, kidney, subcutaneous tissue) generally have low α/β ratios, around 2–5 Gy — their curves bend sharply, meaning large fraction sizes disproportionately increase biological damage relative to what the same total physical dose would produce in small fractions.
This differential sensitivity is precisely why conventional small-fraction radiotherapy protects late-responding tissue: at 2 Gy/fraction, the quadratic penalty for low-α/β tissue is modest. Push fraction size to 10+ Gy, and that same low-α/β tissue suffers dramatically more damage per Gray than a high-α/β tumor does — unless normal tissue can be geometrically spared, which is exactly the engineering problem that stereotactic techniques solve (Stage 5).
Most tumors behave like early-responding tissue with α/β ≈ 10 Gy, but prostate adenocarcinoma is an important, extensively validated exception: its α/β ratio is unusually low, estimated around 1.5–3 Gy — closer to that of late-responding normal tissue than to typical tumors. This means the prostate tumor itself becomes progressively more sensitive, relative to surrounding normal tissue, as fraction size increases.
This single biological fact is the rationale behind moderate and ultra-hypofractionated prostate radiotherapy regimens: because tumor and dose-limiting normal tissue (rectum, bladder) have similar low α/β ratios, hypofractionation does not erode — and may even improve — the therapeutic ratio, unlike in most other tumor sites. This insight, confirmed by large randomized trials, fundamentally changed prostate radiotherapy practice over the past two decades.
Hypofractionation delivers a course of radiotherapy in fewer fractions of larger individual size than conventional treatment — for example 5 Gy × 5 fractions instead of 2 Gy × 35. It shortens treatment, reduces patient burden and cost, but the same linear-quadratic mathematics that make it efficient also mean the therapeutic ratio must be checked carefully against the α/β of every tissue involved.
Fewer visits mean less time off work, less transportation burden, lower per-course healthcare cost, and reduced machine time — a meaningful gain in radiotherapy departments serving large populations with limited linear accelerator capacity. Biologically, a shorter overall treatment time can also reduce the opportunity for accelerated repopulation of surviving tumor clonogens partway through a long course, which is particularly relevant for rapidly proliferating tumors.
However, none of these practical advantages matter if the regimen is not radiobiologically equivalent — or better — for tumor control while remaining within normal-tissue tolerance. This is precisely the comparison the BED framework in Stage 4 is built to make explicit and quantitative rather than left to intuition.
Because the β (quadratic) term grows with the square of dose per fraction, increasing fraction size increases biological effect faster in low-α/β tissue than in high-α/β tissue. A hypofractionated scheme that appears to deliver a modest physical dose can therefore still impose a disproportionately large biological burden on late-responding normal tissue compared to an equivalent tumor-focused benefit — narrowing, rather than widening, the gap between cure and complication.
This is why blind hypofractionation is not automatically safe: it must be paired either with a favorable tumor α/β (as in prostate cancer, Stage 2) or with techniques that geometrically reduce the normal-tissue dose independent of radiobiology — precise beam shaping, image guidance, and steep dose gradients, all of which become essential at the SBRT extreme explored in Stage 5.
Randomized evidence (not just theory) now supports moderate hypofractionation: the UK START trials established 40–41.6 Gy in 15–13 fractions as non-inferior to 50 Gy in 25 fractions for breast cancer, and the CHHiP trial validated 60 Gy in 20 fractions against 74 Gy in 37 fractions for prostate cancer, both now standard of care.
Clinicians select fraction size and number by working backward from tolerance doses of the most dose-limiting nearby organ-at-risk, using its α/β to compute the maximum tolerable BED, then checking whether the resulting scheme still delivers a tumoricidal BED to the target. Modern treatment planning systems perform this BED/EQD2 arithmetic automatically for every relevant structure, but the underlying logic is exactly the linear-quadratic calculation demonstrated interactively here.
Moderate hypofractionation (roughly 2.5–4 Gy/fraction) is now routine for breast and prostate cancer. More extreme hypofractionation requires additional technological safeguards — image-guided setup, motion management, and sharp dose falloff — which is the domain of stereotactic body radiotherapy covered in the final stage.
Biologically Effective Dose (BED) rescales any fractionation scheme onto a tissue-specific biological scale, derived directly from the linear-quadratic survival equation. EQD2 takes BED a step further, re-expressing it as the total dose that would produce the same effect if delivered in the familiar unit of 2 Gy fractions — the everyday language radiation oncologists use to compare, combine, and re-plan treatments.
Starting from SF = exp(−αD − βD²) for a total dose D delivered in n equal fractions of size d (D = nd), the natural logarithm of survival is −ln(SF) = αD + βDd = αnd + βnd². Dividing through by α gives −ln(SF)/α = nd[1 + d/(α/β)] — a quantity with units of dose that summarizes total biological effect independent of the specific α value, depending only on total physical dose (nd), fraction size (d), and the tissue's α/β ratio. This quantity is defined as the Biologically Effective Dose:
BED = nd × [1 + d/(α/β)]
Because BED is derived directly from cell survival, two regimens with equal BED (for a given α/β) are predicted to produce equal biological effect on that tissue — even if their total physical dose, fraction size, and number of fractions all differ.
BED must always be quoted with the tissue α/β used to calculate it — the same physical regimen produces a different BED for tumor (α/β≈10) than for late-reacting normal tissue (α/β≈3), which is exactly why comparing schemes requires computing BED twice, once per tissue.
BED values are algebraically convenient but not intuitive as a stand-alone number — a raw BED of "84 Gy" means little without context. EQD2 solves this by re-expressing BED as the equivalent total dose delivered in 2 Gy fractions, the fraction size clinicians think in by default:
EQD2 = BED / [1 + 2/(α/β)] = nd × [1 + d/(α/β)] / [1 + 2/(α/β)]
EQD2 lets a radiation oncologist directly compare a modern hypofractionated regimen to the historical 2 Gy/fraction literature that most dose-tolerance data was originally built on, and — critically — lets them sum doses from two separate courses (e.g. an initial treatment plus a later re-irradiation) on the same additive scale, something raw physical Gray cannot safely do across different fraction sizes.
Using the calculator above: conventional 70 Gy in 35 × 2 Gy gives a tumor BED of 84 Gy (α/β=10) and a normal-tissue BED of 116.7 Gy (α/β=3). A hypofractionated 25 Gy in 5 × 5 Gy gives a tumor BED of only 50 Gy but a normal-tissue BED of 66.7 Gy — both schemes' normal-tissue BED-to-tumor-BED ratio can be compared directly to judge which better preserves the therapeutic margin for a given clinical scenario.
EQD2/BED accounting is essential in re-irradiation: when a patient who previously received radiotherapy needs treatment again to an overlapping or adjacent volume, the cumulative biological dose to at-risk normal tissue (particularly spinal cord) must be estimated in EQD2 terms, factoring in a degree of tissue recovery over time, to keep the total within a tissue's lifetime tolerance and avoid late complications like myelopathy.
Stereotactic body radiotherapy (SBRT) pushes hypofractionation to its extreme: one to five fractions of 15–20+ Gy each, delivered with sub-millimeter targeting accuracy and steep dose gradients that spare adjacent normal tissue geometrically rather than radiobiologically. At these enormous per-fraction doses, the linear-quadratic model itself starts to lose accuracy — a limitation clinicians must understand, not just a mathematical curiosity.
SBRT regimens deliver ablative biological doses to small, well-defined targets — typically early-stage lung tumors, oligometastases, spine lesions, and increasingly prostate cancer — using highly conformal beam arrangements, rigid immobilization, image guidance immediately before each fraction, and motion management for organs that move with breathing. The result is a dose distribution that falls off extremely steeply outside the target volume, so that normal tissue just a few millimeters away receives only a small fraction of the prescription dose.
Because the geometric sparing does most of the protective work, SBRT can safely deliver a per-fraction dose that would be radiobiologically catastrophic to normal tissue if that tissue actually received the full prescription — the LQ-predicted BED to normal tissue is only relevant to the small volume genuinely inside the high-dose region.
The LQ model was fit to cell-survival data collected mostly in the 1–6 Gy per fraction range, and its two-parameter exponential mathematics has no natural upper dose limit — extrapolated to 18–34 Gy in a single fraction, it predicts survival fractions many orders of magnitude below what is biologically plausible or clinically observed. Above roughly 8–10 Gy per fraction, several additional mechanisms not captured by simple clonogenic-cell LQ kinetics become clinically significant: endothelial cell apoptosis causing microvascular collapse and secondary tumor cell death from ischemia, direct stromal and connective-tissue damage, and stimulation of anti-tumor immune responses (increased tumor antigen release and immune cell infiltration) following very high single doses.
Several modified models (universal survival curve, generalized LQ with saturation) have been proposed to better fit high-dose-per-fraction data, but none has yet fully replaced the LQ/BED framework in routine clinical planning — it remains the practical standard, used with the explicit caveat that its numeric BED predictions above ~10 Gy/fraction are directionally useful but quantitatively less reliable.
The LQ model's "always-bending-downward" survival curve has no plateau, but real high-dose cell-survival data increasingly straightens out at very high doses — meaning the model likely overestimates direct clonogenic kill at SBRT-level fraction sizes, even though SBRT is nonetheless highly clinically effective, in part via the additional indirect mechanisms described here.
Lined up side by side: conventional fractionation (2 Gy × 35) achieves a high cumulative tumor BED through many small, normal-tissue-sparing fractions over seven weeks. Moderate hypofractionation (5 Gy × 5) achieves a comparable or lower tumor BED in a fraction of the time, at some cost to the normal-tissue margin unless the tumor's own α/β is favorably low. SBRT (1–5 fractions of 15–20+ Gy) achieves an enormous nominal BED to the target — often several hundred Gy equivalent — compressed into days, made tolerable only because image-guided conformal delivery keeps that dose almost entirely confined to the tumor volume itself.
Common validated SBRT regimens include 3 × 18 Gy or a single 34 Gy fraction for early-stage inoperable lung cancer, and 5 × 7–8 Gy for spine metastases — each chosen from prospective trial data balancing ablative tumor BED against strict normal-tissue dose-volume constraints, rather than derived from LQ extrapolation alone.