The radiobiological

modelling challenges

of

21st century radiotherapy

Roger Dale Department of Surgery and Cancer

Faculty of Medicine

Imperial College

London, UK

r.dale@imperial.ac.uk

Stockholm. September 2014

This talk will...

• Outline the status of radiotherapy in the 21st century;

• Consider some radiobiological problems introduced by new treatment methods, along with their possible consequences;

• Examine some partial-solutions to the problems posed;

• Emphasise the quantitative importance of the physics base associated with all types of radiotherapy;

• Consider how many Rs are required to describe modern radiotherapy;

• Introduce an elephant!

Lecture outline

An important statistic for the 21st century

“ Radiotherapy remains a mainstay in the treatment of cancer. Comparison of the contribution towards cure by the major cancer treatment modalities shows that of those cured, 49% are cured by surgery, 40% by radiotherapy and 11% by chemotherapy”.

RCR document BFCO(03)3, (2003).

Conclusion: Radiotherapy is a very successful and cost-effective cancer treatment modality, capable of bringing cure and symptom relief to large numbers of patients. Modern developments hold the promise of further improvement in radiotherapy success rates. However, successful application of the emerging techniques will require careful consideration of the associated radiobiology.

Radiotherapy... • Has been used in the treatment of cancer for well over 100 years. As a

consequence, there is now a huge World-wide accumulation of clinical experience in the successful application of radiotherapy;

• In the 21st century, radiotherapy retains a central place in cancer treatment, often in combination with surgery and/or chemotherapy;

• In the past 20 years, technological advances have led to the introduction of many new types of radiotherapy treatment (e.g. IMRT, V-MAT, stereotactic RT, intra-operative RT, high-LET therapy, biologically-targeted RT, permanent seed implantation, etc);

• These are exciting developments and could provide yet more improvement in cancer cure rates;

• However, it is sometimes difficult to use the accumulated clinical experience gained from conventional radiotherapy to determine the optimum use of the new treatment methods;

• This is why an understanding of the radiobiological issues remains fundamentally important.

Standards of dosimetry in radiotherapy are now extremely high. As a result, most treatments can be delivered with an overall dosimetric accuracy which is within +/- 3%. This target is comfortably achievable and is often exceeded.

Confident correlation of clinical outcome with delivered dose is centrally-reliant on this high degree of dosimetric accuracy. This means, for example, that the biological implications of DVHs etc can be explored with a confident acceptance that the “physics” aspect of DVHs is reliable.

In the UK, physical overdoses >10% are legally reportable and can lead to extensive investigation by government officers and enforcement action. This is a strong indication of how good the physics is expected to be.

New treatments (e.g. those involving high-LET therapy, non-homogeneous beams, biologically-targeted radiation, novel fractionation, etc) can introduce many radiobiological uncertainties. Thus the new techniques raise many serious questions regarding our ability to maintain OVERALL (i.e. physical + biological) dosimetric integrity to currently accepted standards.

This is why continued research and investment in radiobiology is essential.

Another reason why radiobiology is important...

The potential for using clinical datasets for the determination of radiobiological knowledge bases for normal tissues has been demonstrated by Rutkowska et al [Phys Med Biol (2010); PhD thesis (2010); Brit J Radiol (2012)]. The information required is: DVHs Reliable input data Binary complication information Several models of normal tissue effect were then used to determine: Critical Functioning Volume (CFV) as a percentage of whole organ volume, and Threshold number of Functional Sub-Units (FSUs) in any one CFV, below which a complication occurs This approach is capable of characterising the volume effect for any selected organ . The derived parameter values were relatively independent of the choice of model, but non-dosimetric confounding factors (e.g. those related to patient lifestyle, co-existing medical conditions, etc) exerted more influence on the results. The model predictions were greatly improved when confounding factors (i.e. those outside the realms of physics dosimetry) were taken into consideration and this should motivate the need to identify, record and quantify such aspects. A similar finding has been found in optimising the application of QUANTEC dose constraints - even better predictions were possible when patient risk factors were additionally taken into account [e.g. Appalt et al, Acta Oncol (2014)].

A recent example...

New external beam techniques...

Type of treatment New problem introduced Consequence

IMRT, VMAT, etc a) There may no longer be a uniform dose across the tumour. b) Although average dose to normal tissues may be reduced, large volumes of normal tissue may receive small doses. c) Individual fraction delivery times may be quite long (i.e. several minutes).

Difficult to relate to the clinical experience from older techniques, in which tumour dose was more uniform. As above. Also, the risk of secondary cancers may be increased. There may be repair during the fraction delivery, meaning that the delivered dose is less effective.

Radiosurgery Dose fractions may be very much larger than conventional fraction sizes

Radiobiological models are less reliable for large fraction sizes. Also repair in long fractions times.

Hadron therapy RBE is not a fixed value – depends on fraction size and type of tissue irradiated ().

Difficult to compute the local RBE effect without voxelelised information about LET distribution.

New brachytherapy techniques... Type of treatment New problem introduced Consequence

Permanent implants a) The radionuclides used may have long half-lives, allowing considerable tumour repopulation to occur during treatment. b) The radionuclides used generally have RBEs which are significantly > 1.0.

For fast-growing, and/or radio-resistive tumours the overall biological effectiveness may be much reduced. The biological effectiveness may be much greater than is predicted from considering only the prescribed dose.

Pulsed brachytherapy The technique involves a large number of closely spaced fractions, each delivered at relatively low dose-rate.

Incomplete-repair between fractions must be allowed for. Additionally, some repair may take place during each fraction.

Biologically-vectored radiotherapy

a) The targeting radionuclide could be a beta- or alpha-emitter possessing an high RBE and with short-range dose delivery characteristics.

b) The critical normal tissue is more likely to be the kidney, rather than a tissue/organ immediately adjacent to the targeted tumour.

Almost impossible to relate to previous clinical experience. Pre-treatment tracer studies required in each patient to estimate the degree of targeting and hence, the activity to administer. As above. Also requires detailed understanding of kidney function and physiology in relation to the range/energy of the specific radionuclide being used.

Intra-operative x-ray radiotherapy (e.g. of the breast or rectum)

Introduces large fraction sizes, very rapid dose fall-off and (because of low x-ray energies) likely RBE > 1).

Modelling problems with large fraction sizes and dose-gradient effects.

Addressing the consequences... Some of the consequences identified in the previous two

slides can be addressed from consideration of the

conventional “5Rs” in conjunction with quantitative

models.

However, some of the other issues require a deeper

understanding of aspects not currently covered by the

5Rs or by the models.

For example...

The 5 (shouldn’t it be 6?) Rs of radiotherapy...

•Repair – intra-cellular processes

•Repopulation – inter-cellular processes

•Re-assortment – cell cycle times/inter-cycle sensitivity variations

•Re-oxygenation – time, dose and structure-dependent

•Radiosensitivity – gross responsiveness

The above phenomena are mostly relevant to tumours, for which TCP is essentially related to the net cell kill. For normal tissues NTCP is not a simple function of net cell kill since it is very dependent on the functional structure of the tissue/organ. Since newer forms of radiotherapy involve non-uniform dose distributions we need to know more about...

•Response characteristics – Late/early, hierarchical structure, FSU distributions, physiology, consequential effects, cancer induction probability, etc.

But even 6Rs would provide only an incomplete picture, because of the elephant...

The elephant in the room...

Many RT schedules involve the addition of chemotherapy. Chemotherapy may be combined with RT in a wide variety of ways. Chemotherapy may therefore be one of the most significant confounding factors of all. Any analysis of radiochemotherapy (RCT) schedules which takes no account of what the chemo is doing risks seriously undermining the quantitative usefulness of any derived radiation parameters.

What is known? (Grégoire and Baumann, 2009)

• For several sites, combined modality treatment provides modest but useful gains (10-20%) in tumour response.

• The improvement comes mostly from increased local control rather than a decrease in distant metastasis.

• Results are generally better for concurrent therapy rather than for sequential therapy.

• However, many of the comparisons cannot be made at equal normal tissue toxicity, or have involved RT schedules which themselves were sub-optimal.

• Many factors introduce bias into inter-comparisons between RT and RT + chemo and the methodology used in most trials limits the assessments.

What modelling has been done?

• There are very many publications dealing with the modelling of chemotherapy.

• Typical approaches: Use complex compartmental (dynamic) models to perform “concentration versus time” analysis of drug dwell in the tumour/tissue. Then find the optimal duration of drug exposure to maximise cell kill.

• Some papers suggest a reasonable correlation between “area under the curve” and tumour response but find little correlation for normal tissue toxicity.

• Hardly any consider mixed modality issues.

A possible starting point...

• In RT the concepts of BED and EQD2 are well-established. Physical doses are converted to additive biological doses (BEDs) which incorporate physical and biological response parameters. This allows quantitative assessment of combined RT treatments, (e.g. XTB + boost; XTB + brachy) and can help in guiding the application of new and novel RT treatments (ion therapy, targeted therapy, etc).

• Clinically prescribed schedule(s) may thus be specified, for each critical end-point, in terms of a “one-number” measure of overall biological effect.

• Chemotherapy prescriptions are typically based on body surface areas, body weight and glomerular filtration rate. Thus, whereas the RT BED values are directly linked to log cell kill, the chemo prescription factors are not.

• In principle this may not be too much of a problem, provided chemo contributions can be classified in terms of end-points already used for RT.

Why express chemo in terms of RT effect?

• Modern RT enjoys high standards of dosimetric precision. This is taken for granted but the sheer importance of this factor is too easily overlooked.

• Without this precision, RT clinical trial results would be much weaker and radiobiological modelling would be of little value.

• Chemo drug delivery to a tumour has nowhere near the precision inherent in RT dose delivery.

• Until better models of chemo quantification are developed it seems reasonable to use existing RT models to quantify the impact of chemo in terms of a radiation- equivalent.

How might chemo interact with RT? • Increase the radiation sensitivity (i.e. increase

and/or .)

• Produce independent cell kill.

• Reduce tumour repopulation (i.e. slow-down or stop the repopulation and/or extend the delay time.)

• Various combinations of the above.

• In principle, all the possibilities may be expressible in terms of an equivalent dose or BED.

Existing quantitative view of chemo...

• It is claimed that chemo alone rarely achieves a SF of better than 10-6 (Grégoire and Baumann, 2009). Given that most tumours require a target cell kill of 10-9 – 10-11 in order to be sterilised, the chemo contribution to RCT appears modest and possibly of small significance.

• In fact this is not true – a cell kill of 10-6 (i.e. one cell in a million remaining) represents a very substantial and clinically detectable contribution.

• However, most current modelling seems to suggest that SF due to chemo alone is >>10-6.

Some typical chemo dose-equivalents already determined...

• For head and neck cancer [Fowler et al (2003, 2008); Pettit, 2013] concurrent chemo adds a BED of ~3 - 9Gy10, i.e. an EQD2 of 3 – 7.5Gy.

• In malignant glioma Jones & Sanghera (2007) estimated that concurrent chemo is equivalent to an EQD2 of 8 – 10Gy.

• These results appear to be consistent with an independent chemo cell kill of ~10-2, rather than 10-6.

A pragmatic radiobiological approach to analysing RCT data...

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

0 20 40 60 80 100 0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

0 20 40 60 80 100 120

Fixed sensitisation Chemo dose-equivalents appear to increase with increasing dose

Fixed independent cell-kill Chemo dose-equivalents appear to decrease with increasing dose

RT-alone response

RCT response

Radiation dose

Clinical response

TCP

Radiation dose

True chemo radiation-equivalent Apparent chemo radiation-equivalent

RT-alone response

CRT response

The methodology in practice...

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

0 20 40 60 80 100

CR

EQD2 (Gy)

Muscle-invasive bladder cancer. Weighted best-fit line for the RT–alone trial data.

Fitted parameters: D50 = 45.92Gy Gamma = 0.532

These were taken to be the definitive parameters describing the MIBC radiation –alone response and were kept fixed in the subsequent RCT analyses.

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

0 20 40 60 80 100 120

Sensitisation

95% CIs

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

0 20 40 60 80 100 120

Com

ple

te r

esponse

(C

R)

rate

Independent cell-kill

EQD2 (Gy)

Best fit

Clinical end-point at which chemo

radiation-equivalent is determined

Apparent radiation equivalent of chemo

Independent cell

kill model

Sensitisation model

Depending on the method of analysis, the derived radiation-equivalent dose may be dependent on the end-point (clinical response) at which it is determined. These “apparent” dose-equivalents may not therefore reflect the “true” values. Projections about how the radiation-equivalent dose varies at different response-rates ARE dependent on the assumed model for chemo action. In principle, standard radiation dose-response equations (e.g. the generalised logistic model) can be modified to include the chemo effect and then fitted to clinical data to help identify (and quantify) the likely modus operandi of the chemo and determine the “true” chemo contribution. However, the noise associated with many clinical datasets demonstrate the huge uncertainties that exist within RCT treatments. In spite of these shortcomings, the derived chemo dose equivalents using this method do appear to be more consistent with a chemo-induced cell kill of ~10-

6, rather than ~10-2. Plataniotis & Dale, IJROBP(2014).

Summing up the problem with RCT...

Derived chemo dose-equivalents would appear to be a pragmatic way of assessing the relative “worth” of the chemo contributions in RCT treatments. If reliable chemo dose-equivalents were obtainable it could help: a) In assessing whether or not a similar clinical effect could be obtained (for example) by RT alone, by RT + boost, by RT + brachy or via the use of high-LET radiations. b) Guide the optimised use of chemotherapy. Unfortunately, derived chemo dose-equivalents may depend on the level of clinical effect at which they were determined and also on the assumed modus-operandi of chemo action. Determination of true (as opposed to apparent) dose-equivalent estimates is therefore very difficult unless the associated datasets are of especially high quality. For this reason the dose-equivalent idea may currently be of limited value. Good quality clinical datasets could help address these issues, but specifically-designed animal and laboratory experimentation would undoubtedly accelerate the process.

Where does radiobiology sit in the 21st Century?

The Radiobiology Melting Pot

Experimental

RadiobiologyPhysics

Clinical

Practice

MELTING POT

Ideas, modelling of

hypotheses, testing,

advancement.

Cancer BiologyImaging

Microdosimetry

Dale, Douglas Lea lecture, (2005)

Radiobiological modelling has proven useful in the analysis of a number of different types of radiotherapy. However, for the types of treatment now being increasingly practiced, a wider pool of knowledge needs to be accessed in order for models to be refined in order to provide the best guidance and predictive value. This may involve the assessment of factors which are not conventionally considered as being part of radiobiological models. Quantitative assessment of RCT should be an integral part of this quest. To go from where we are now to where we want to be in the future requires more support for focussed radiobiology research and, above all, education.

CONCLUSIONS...

Thank You