But how are radiation oncologists to choose between the alternatives? And having made this choice, how do they decide what dose to prescribe to the tumour? Too high and the patient may suffer serious long-term complications; too low and the tumour may recur, and the patient may die. The answer to both of these questions lies in enlisting the aid of radiobiology - the study of the effects of ionizing radiation on living matter. At times, I feel that we as a profession are too much in love with technology, forever waiting for our next high-tech fix, rather than making the maximum use of the incredible tools we already have.

Ready for use
Radiobiology is a well established scientific discipline, thanks to many years of brilliant and painstaking work by scientists of the calibre of L H Gray, Jack Fowler, Gordon Steel, Don Chapman and others. The linear-quadratic model and its mechanistic interpretation was a major intellectual achievement, as was the teasing out of the radiobiology and clinical impact of hypoxia and dose rate, and the application of all of these to brachytherapy. We now have plausible biomathematical models for the probability of local tumour control and normal-tissue complications, with parameters derived from correlating 3D dose-volume data with clinical outcome.

But isn't it all too difficult to express the aim of radiotherapy in mathematical terms? Not in the slightest. The aim of any radiotherapy treatment administered with curative intent can be stated thus: "Maximize local tumour control for an agreed (acceptable) complication risk." For today's sophisticated, 3D radiotherapy treatment-planning systems this is eminently achievable, provided the functions for tumour control probability (TCP) and normal-tissue complication probability (NTCP) are embedded in the treatment-planning software.

Other recent developments offer the promise of creating images of tumour biology via PET and MRI. This is all very exciting, but how should we make use of these images? Without the guidance of radiobiological modelling, it could end up as inspired guesswork at best. For example, giving a boost of, say, 15 Gy to the hypoxic subvolume of a tumour indicated by PET, and expending considerable effort in achieving this is nonsense. Patients deserve better. Ethics committees should insist on a radiobiological analysis of the potential benefit before approval is given for such a strategy.

Catch-22
Radiation oncologists have for many years hidden behind - or used as an excuse for changing their practice at a snail's pace - the staggeringly cumbersome, unwieldy "holy cow" of the phase III clinical trial, with "moral" support provided (in the UK) by the official NHS mantra of "evidence-based medicine".

One of the arguments used by our clinical colleagues when challenged as to why they don't make active use of predictive models for the probability of complications (NTCP) and local tumour control (TCP) is that they cannot "trust" these models until they have been verified on "our particular patient population".

The logical next step is then to request their follow-up data on patient outcome in order to undertake such a verification. Only for the clinicians to counter with: "But we don't collect such data." And why don't they? "We don't have time for such activities... there's no budget to employ 'data clerks' (the wrong type of professional anyway)... the NHS won't fund such 'research'," and so on. It's a classic Catch-22 situation - and it's simply not good enough.

One oft-met objection is: "But your models don't account for chemo-radiotherapy combined." Well, if our current models don't explicitly take account of combination doses of, say, cisplatinum and radiation, then rather than throwing in the towel, we should be giving a high priority to developing models for this situation, as UK scientists Roger Dale in London and Bleddyn Jones in Birmingham are doing.

Finally, how will we make rational use of the new information long-promised by advances in molecular biology, such as individual radiosensitivity, the genetic predisposition of the patient to higher complication risk, poor vasculature and so on? Only by using a TCP model containing explicit parameters such as radiosensitivity coefficients α and β, along with an NTCP model that includes the tolerance dose for 50% complication rate. One can then choose patient-specific values as and when such information becomes available.

For too long the speciality of radiotherapy has been something of a "black art", nudged along by rules of thumb. With radiobiological guidance, it is now time to move radiotherapy to being a science. The knowledge and the tools are there. We physicists are ready and waiting.