The idea is to transition from inferring treatment plan quality from dose-based parameters that are assumed to correlate with biological response, to optimizing radiotherapy plans based on estimates of the biological outcome itself.

"These models have been in the literature for a long time now, but they are not being actively used to change what we do," Nahum told delegates at the recent Clatterbridge course on Radiobiology and Radiobiological Modelling in Radiotherapy. "We need to move beyond just looking at dose distributions and see what data we can extract that will be meaningful for treatment outcome."

Conformal radiobiology

Nahum explained how the increased conformality conferred by modern radiotherapy techniques reduces the volume of high dose delivered to critical organs-at-risk. As a result, there's now a need to quantify this reduction in risk, using a radiobiological model for normal tissues that quantifies what happens to the tolerance of an organ when the fraction of its volume that is irradiated is changed.

Such a model, which determines NTCP as a function of both the dose to normal tissue and the irradiated volume, can then be added to the tumour control curve. This provides a tool for quantifying the difference in risk between two treatment plans (which may even correspond to different modalities such as photons and protons) and may reveal, for example, that different tumour doses can result in the same predicted level of complications (known as isotoxic doses). It's then possible to calculate the difference in tumour control that these doses would provide. "We see an improvement in control when we go from one technology to another, with the same complications; this is what we want these models to do," Nahum explained.

So how can this idea of "conformal radiobiology" be applied to optimize radiotherapy? According to Nahum, treatments should be optimized using the predictions of reliable NTCP models, starting with a treatment plan derived from dose-based criteria and then adjusting it until NTCP reaches the required level.

He presented an example in which a constant prescription of 55 Gy in 20 fractions was delivered to lung tumours in 25 patients, resulting in a wide spread in estimated NTCP (pneumonitis) between patients. Adjusting the dose to give all patients the same level of complication risk (10%, for example) resulted in a spread in prescription dose from 35 to 85 Gy.

This individualization of dose based on normal-tissue constraints translated into a broader spread of TCP, including many higher probabilities, and the average TCP increased from 44.2 to 59.5% with no increase in average NTCP between the original fixed tumour doses and the new variable ones. "All you have done is customize the dose to the tumour volume individually using an NTCP model," said Nahum. "No new technology is needed, just simple software that treatment planning manufacturers should add into their systems."

Nahum noted that this idea – defined as "level one" biological optimization – is now being put into practice in the UK in two current clinical trials: I-START and IDEAL-CRT.

Higher-level optimization

Increasing the complexity one step further, level two optimization involves individualization of both the dose and the number of fractions on an isotoxic basis. This can be achieved using the BioSuite software developed at Clatterbridge. Nahum presented an example in which BioSuite was used to demonstrate the advantages, for a particular patient, of moving to two fractions per day. "This radiobiological approach can help you pick out patients, for example, for stereoablative radiotherapy," he added.

Nahum then described level three optimization, in which expressions for TCP and NTCP are used in the objective function of the inverse planning process, thereby also modifying the dose distributions. For instance, the computer could be instructed to maximize TCP for a fixed NTCP of 4%, or to minimize NTCP for a fixed TCP of 95%. He showed an example of a biologically optimized lung tumour plan (created using the Pinnacle Research Interface) in which the TCP was increased from 40 to 60% while the NTCP remained constant.

Nahum concluded his presentation by mentioning the even "higher" levels that may be achievable: level four biological optimization, in which patient-specific functional imaging data are added to radiobiological inverse planning (dose-painting); and level five optimization, where individual patient biological data are also incorporated.

While all of these approaches are ultimately dependent upon the reliability and accuracy of the NTCP and TCP models used, replacing current treatment protocols with individualized prescriptions should maximize the clinical benefit of radiotherapy. Big improvements can be achieved "just by using a bit of maths and radiobiological intelligence," said Nahum.

•  Next year's course on Radiobiology and Radiobiological Modelling in Radiotherapy is scheduled for 23–27 February 2014.