Radiobiology of LDR, HDR, PDR and VLDR Brachytherapy - GEC-ESTRO Handbook of Brachytherapy

SECOND EDITION

The GEC ESTRO Handbook of Brachytherapy

PART I: THE BASICS OF BRACHYTHERAPY 5 Radiobiology of LDR, HDR, PDR and VLDR Brachytherapy Erik Van Limbergen, Michael Joiner, Albert Van der Kogel, Wolfgang Dörr

Editors Erik Van Limbergen Richard Pötter

Peter Hoskin Dimos Baltas

Radiobiology of LDR, HDR, PDR and VLDR Brachytherapy

3

THE GEC ESTROHANDBOOKOF BRACHYTHERAPY | Part I: The Basics of Brachytherapy Version 1 - 22/10/2015

5 Radiobiology of LDR, HDR, PDR and VLDR Brachytherapy

Erik Van Limbergen, Michael Joiner, Albert Van der Kogel, Wolfgang Dörr

1. Summary 2. Introduction

3 4 4 5

7. Biological effects of dose inhomogeneity 12 8. Volume, anatomical site, and patient-related effects 12 9. Clinical Results: LDR, MDR, HDR and PDR BT 13 10. Practical Applications 16 11. Key messages 17 12. References 18

3. Radiation Effects and Tissue Responses 4. Dose - Time Patterns in BT 5. The four R’s of Radiobiology and the Dose Rate Effect 6. Mathematical modelling of different dose rates and the EQD2 concept

6

8

1. SUMMARY

Brachytherapy (BT) differs from external beam therapy (EBRT) in two main ways: the distribution of the absorbed dose and the time-dose patterns. Dose distribution: Total dose and dose rate in BT decrease dramatically with the distance from the sources. The dose is pre- scribed to a CTV, i.e. an isodose surface often encompassing a volume of ~20 - 200 cm 3 . The GTV receives about >150% of the prescribed dose. Dose rates: for brachytherapy are divided, somewhat arbitrarily, into three ranges: low-dose rate (LDR) < 1 Gy h -1 , medium dose rate (MDR) between 1 and 12 Gy h -1 and high-dose rate (HDR) >12 Gy h -1 . Changing dose rate in the MDR range causes the most pronounced changes in biological effect. In the present report, terms describing dose-time-distributions specific to cervix brachytherapy, including (mean) dose rate, fractionation, pulse, application and overall time are defined. Moreover, biological mechanisms potentially impacting on effectiveness of the treatment, such as repopulation, re-oxygenation and redistribution, are discussed. The large decrease in dose with distance and the large variation in dose rate and fraction size require a common concept and a joint terminology to facilitate biological comparison between the different brachytherapy schedules feasible. The LQ formalism and EQDX: allows comparison of the predicted effects of a particular BT schedule with other BT and external beam schedules, with regard to both tumour control and normal tissue effects. With a set of assumptions, even the additional ef- fect of chemotherapy may be quantified, but with large uncertainties. This formalism can be safely applied within a range of doses per fraction from 0.5 Gy to 6-10 Gy; it might, however, potentially overestimate the effects at higher doses per fraction For all dose rates the calculations of the effects are strongly dependent on the recovery capacity (related to the α/βvalues) and half- times for recovery T 1/2 that are assumed in the modelling. But uncertainties in the estimates of these values need to be considered. This applies to tumours as well as to normal tissues. For photon irradiations an α/β value of 10 Gy and a recovery half-time of 1.5 h for cervix tumour tissue is generally assumed and of 3 Gy and 1.5 h for late effects in the OAR. For more specific and precise calculations, tissue specific recovery parameters should be used where available. The Equieffective Absorbed Dose Concept, EQDX: has been developed for comparing different irradiation protocols and tech- niques. Equieffective doses delivered in X Gy fractions (EQDX) are defined as total doses that - delivered under different condi- tions, which have to be specified in the context, are assumed to produce the same probability of a specific effect (endpoint), as the resultant total dose given in X Gy fractions. For protocols involving only one type of radiation equi-effective doses can be calcu- lated using the LQ formalism; and assumed values of α/β well as the T 1/2 of recovery if required for the EQDX calculations have to be specified by subscripts: EQDX α/β , T 1/2 . Because recovery is often assumed to be complete between fractions, the reference to T 1/2 may be omitted, depending on the details of the dose delivery. Because of historical precedents and clinical experience EQD2 referring to photon doses of 2Gy /fraction is commonly used. For protocols involving different radiations assumed values of β are required for both radiation types and must be specified. The ICRU/GEC-ESTRO report 88 recommends use of the equi-effective formalism, particularly EQD2, in addition to absorbed doses, to report doses for planning aims, prescriptions and doses delivered.

Radiobiology of LDR, HDR, PDR and VLDR Brachytherapy

4

THE GEC ESTROHANDBOOKOF BRACHYTHERAPY | Part I: The Basics of Brachytherapy Version 1 - 22/10/2015

2. INTRODUCTION

The efficacy of brachytherapy (BT) is attributed to the ability of radioactive sources close to or within the target to deliver higher radiation doses more precisely to the target than external beam radiotherapy (EBRT). As in EBRT, the biological effects depend on total dose deliv- ered, administration parameters such as dose rate, fractionation schedule and overall treatment time, and volume parameters such as total volume treated to certain doses and the dose distri- bution within that treated volume BT. In conventional EBRT, the treated volume is usually large. Varia- tion in dose is kept minimal inside the target volume, aiming at a homogeneous distribution of dose. Dose prescription is usually to a point within the target and deviations within a range of only -5% to + 7% of the prescribed dose are considered to be accept- able (ICRU 50 1993, ICRU 62 1999, ICRU 83, 2010). In BT, the dose is prescribed to an isodose encircling a small tar- get volume (either D100 (100 % of the Minimum Target Dose MTD), D98 % or D90 of the MTD. In contrast with EBRT, the dose distribution is very heterogeneous. It is lowest at the pe- riphery of the target, but much higher doses and dose rates are delivered in the vicinity of the sources. The average dose giv- en to the target volume is therefore always significantly higher than the prescribed dose at the periphery of the target. Hence, doses delivered by BT mean significantly higher integral doses than would be delivered by EBRT for the same nominal doses, fraction sizes and dose rates. Such high integral doses delivered by BT are only tolerated because the volumes treated are usually very small as compared to EBRT. Considering only the dose and dose rate at the reference isodose at the periphery of the implant can therefore be very misleading, and information on dose distribution, such as homogeneity or inhomogeneity indices or full 3-dimensional DVH parameters, such as D90 and D98 for CTV, and 2ccm and 0.1ccm for certain OAR, should be provided (see chapter on Reporting in BT) Time-dose factors may also differ widely between EBRT and BT. In EBRT, the total dose is delivered in small daily exposure times of a few seconds or minutes, allowing for full repair between fractions. The overall treatment time is several weeks. In BT, in contrast, the dose is delivered either continuously (LDR, MDR) or discontinuously (PDR, HDR), and overall treatment times tend to be short (several hours to several days). In this chapter we will describe the radiobiological mechanisms, applicable to clinical BT which may explain the differences in biological effects of dose rates and administration schedules between EBRT and BT. We will provide practical examples and solutions to translate treatment rules between different dose rates. For more detailed radiobiological explanations we refer you to the ESTRO course book of Basic Clinical Radiobiology (Joiner and Van der Kogel 2009).

Fig 5.1: Direct and indirect effects of Irradiation on intracellular targets (DNA)

Fig 5.2: Time course of Irradiation on biological objects

3. RADIATION EFFECTS AND TISSUE RESPONSES

Ionizing radiation (IR) – as the name indicates - induces ion- izations in (biological) matter. These can be either distributed in a loose way, as with sparsely IR (e.g. X-, gamma-rays or elec- trons), or in a concentrated way (densely ionizing radiation, e.g. neutrons, heavy ions), depending on the Linear Energy Transfer (LET), i.e. the energy deposited per unit distance of traversal of the beam. In a cell, these ionizations can hit the critical target, i.e. the DNA, either directly, or they can produce radicals, main- ly from water molecules (“radiation hydrolysis”), which are the major component of a cell (Fig 5.1) In the latter case, subsequent radical reaction chains may reach and damage the DNA indi- rectly. In the case of densely IR, direct effects on the DNA are much more likely than with sparsely IR. This accounts for differ- ences in the Relative Biological Effectiveness (RBE) of different radiation qualities. This chapter will deal with sparsely IR, which is usually administered in BT. In terms of the time course of the effect of IR on biological ob- jects, several phases can be distinguished (Fig 5.2): • A extremely short initial physical phase (about 10 -15 - 10 -6 s), during which the above mentioned ionizations occur, • A chemical phase , again very short (about 1 to 10 -3 s), during which the induced radicals interact with any molecule they meet, resulting in “chain reactions.” In this phase there is a com- petition between natural scavenging reactions, e.g. with sulf- hydryl components (glutathione) or other antioxidants, that

Radiobiology of LDR, HDR, PDR and VLDR Brachytherapy

5

THE GEC ESTROHANDBOOKOF BRACHYTHERAPY | Part I: The Basics of Brachytherapy Version 1 - 22/10/2015

In early responding, turnover tissues (hierarchically structured), such as the bone marrow or the gastrointestinal mucosae, the above mentioned stem cells produce transit cells, which are re- sponsible for overall cell production, which is then counteracted by the physiological cell loss. In late responding tissues, the tissue specific stem cells produce differentiating cells that may also be recruited into proliferation if there is a demand (flexible tissues). In all tissues, there is a combined and complex response of all exposed cell populations at the cellular level. In early respond- ing tissues, the clinical radiation response is largely based on the changes in the proliferating compartment (stem cell loss, defi- cit in transit cells, hypoplasia, ulceration). Therefore, the time to onset of early effects depends mainly on the turnover time of the proliferative compartment, and is largely independent of the radiation dose. For example, oral mucositis occurs at 9 days after the accumulation of a radiation dose of 20 Gy (Van der Schueren et al . 1990). Based on the number of surviving tissue stem cells, early radiation effects usually heal after a (biologically effective) dose-dependent time interval. The more aggressive the treat- ment, the earlier the effect and the longer it takes to heal. In contrast, for late radiation effects, the parenchymal (i.e. stem cell based) response is of varying importance. Here, endotheli- al/vascular changes, such as dose dependent loss of capillaries, vascular occlusions and telangiectasia – all associated with an impairment of perfusion – are one major component. Radia- tion- induced differentiation of fibroblasts, associated with a substantial increase in connective tissue production, is another predominant factor in radiation-induced tissue fibrosis. It must be emphasized that these late tissue reactions are hence based on changes in cellular functions in a complex way, rather than (exclusively) on cell death. With BT, treatment can be delivered in completely different time patterns: either continuously at lower (VLDR/LDR) or medium dose rates (MDR) or fractionated with smaller or larger fraction sizes and intervals (HDR/ PDR) (Fig 5.3) • Low Dose Rate (LDR) BT applies dose rates in the range be- tween 0.4 and 2 Gy/h. However, in routine clinical practice, LDR BT is usually delivered at dose rates between 0.3 and 1 Gy/h. This is compatible with manual or automatic afterload- ing techniques. • Medium Dose Rate (MDR) BT administers doses in the range between 2 Gy/h and 12 Gy/h. MDR can also be delivered by manual or automatic afterloading, • High Dose Rate (HDR) BT delivers the dose at dose rates of 12 Gy/h and more (>20 cGy per minute), and only remote after- loading is feasible because of the high source activity. Two other dose rate schedules are also used in BT at present. • Pulsed Dose Rate (PDR) BT delivers the dose in a large num- ber of small fractions, called pulses, at short intervals, allowing only for incomplete recovery in between. This aims to achieve a radiobiological effect similar to low dose rate over the same treatment time, typically a few days. 4. DOSE - TIME PATTERNS IN BT

inactivate the free radicals, and fixation reactions that eventu- ally lead to stable chemical changes in the DNA. Even at this point, some repair occurs. For example, the restoration of vari- ous DNA lesions, like DNA-strand or DNA-protein crosslinks, base damages, single- (SSB) or double-strand breaks (DSB) can be induced. Multiply damaged sites, i.e. combinations of these damages, appear to be the basic changes leading to critical bio- logical effects, e.g. cell death. • A third, biological phase , much longer (seconds to years and decades), during which the cells react to the inflicted chemi- cal damage and interact with neighbouring cells, but also with the immune system. It begins with the DNA Damage Response (DDR) which is a highly complex and coordinated system of enzymatic reactions that respond to the existing DNA damage. It is estimated that 1 Gy, resulting in around 10 5 ionisations, causes more than 5000 DNA base damages, about 1000 single strand breaks and 20-80 double strand breaks, but will lead only to a 30 % clonogenic cell kill in an average mammalian cell line. This clearly indicates that in order to cope with different types of radiation damage, mammalian cells have a variety of very effec- tive DNA repair systems such as Base Excision Repair (BER) for base damages, Single Strand Break Repair (SSBR) for SSB, and – for DSB - Homologous Recombination (HR), Non Homologous End Joining (NHEJ), and Single Strand Annealing (SSA), which work with varying efficacy and fidelity (Wouters et al. ). These repair processes successfully repair the vast majority of lesions in DNA. Thus only a few lesions are not or inadequately repaired, and may lead to clonogenic cell death. Multiple damaged sites, including DSB, are the main changes contributingto cell death. Clonogenic cell death is defined as loss of the potential of a cell to undergo an unlimited number of divisions. Clonogenic cells may hence be considered as either normal tissue or tumour stem cells. Clonogenic cell death after radiation exposure can occur in various ways and at various times. Three pathways are dominant. Mitotic death may occur at the first or at (3-4) subsequent cell divisions, when the amount of accumulated DNA damage does not allow for completion of the mitosis. Apoptosis is prominent in only a few cell types, such as lympho- cytes and seminal epithelium, as well as in endothelial cells irra- diated at high doses (> 15 Gy). Thirdly, radiation induced differentiation (e.g. of epithelial cells) or senescence (e.g. transformation of fibroblasts into fibrocytes) can occur, which renders the cells non-clonogenic, although still metabolically active, but with different metabolic and also cell-interactive pathways. The “stem cell hypothesis” postulates that the radiosensitivity of a tissue is based on the number and the intrinsic radiosensitivity of the tissue specific stem cells. Whenever these cells divide (in a normal state), they produce a new stem cell (“self-renewal”) and one other cell, which may either have a limited proliferative capacity (“transit cell”) or directly undergo differentiation. With regard to the time course, early (rather than acute) radia- tion effects – by definition occurring during the first 90 days after the onset of radiotherapy – are distinguished from late (chronic) radiation effects that occur after months to many years.

Radiobiology of LDR, HDR, PDR and VLDR Brachytherapy

6

THE GEC ESTROHANDBOOKOF BRACHYTHERAPY | Part I: The Basics of Brachytherapy Version 1 - 22/10/2015

• Recovery processes, which occur – with a certain tissue specif- ic half-time – between radiation fractions, or during exposure with low dose rates. In vivo, these processes are mainly, but definitely not exclusively, based on DNA repair activities. These processes are dominant in late responding tissues. • Repopulation is a process that is occurring in turnover tissues and in some tumour entities, as soon as a certain level of cell depletion has been accumulated (Dörr 2009, Hopewell 2003). The underlying mechanisms are complex (Dörr 1997, 2003). Repopulation is based on an additional production of stem cells, through a switch to symmetrical divisions resulting in two daughter stem cells, an acceleration of these stem cell divi- sions, and a substantial rate of “abortive”, residual divisions of doomed cells. • Redistribution relates to cell cycle effects of IR, with a pre- dominant kill in sensitive cycle phases, synchronisation at check-points, etc. These effects have been extensively studied in in-vitro systems but the relevance of these phenomena for in-vivo tissue effects is highly questionable. • Reoxygenation describes an improvement of the oxygen status of tumour (stem) cells during radiotherapy. In many tumours, hypoxia develops, based mainly on the increasing distance of the tumour cells from their related vessels/capillaries and increasing intratumoral pressure (chronic, diffusion-related hypoxia), and on the inadequate function (e.g. temporary oc- clusion) of the capillaries (acute, perfusion-related hypoxia). During treatment, temporarily occluded vessels re-open, and with tumour shrinkage, the interstitial pressure decreases and the surviving tumour cells move back closer to their vessels. These complex processes promote an increasing radiosensitiv- ity of the hypoxic – and hence cure- and recurrence- relevant – areas with increasing overall treatment time. • Modern EBRT and established techniques such as BT are asso- ciated with an increase in dose inhomogeneity in the OAR; this is frequently depicted as the “volume effect” although this term can be misleading. The response of individual tissues and or- gans, and importantly also their individual response endpoints, to such dose inhomogeneity can be highly varied and complex. The QUANTEC initiative [Marks 2010] made an attempt to summarize current knowledge about volume-related organ tolerance, and a series of studies and analyses has extended this knowledge since then. However, different endpoints of re- sponse in a particular OAR, such as incontinence vs. bleeding after exposure of the rectum, do have different target subvol- umes, radiopathologies, and radiobiological characteristics. These need to be studied experimentally and particularly clini- cally in forthcoming years. There are different time frames for the relevance of these factors. (Fig 5.4) Recovery of sublethal damage (see above) is the fastest process that starts within one hour. Its consequences can be detected af- ter only15-30 minutes and it is completed approximately 6 hours after an exposure, but may take as long as 1 day (e.g. in spinal cord). It is the most significant factor altering radiation effects between 1 Gy/min and 0.3 Gy/h. Half-times of recovery for hu- man tissues and tumours are estimated to be >2-4 h. The possibility of recovery for various tissue responses is de- pendent on the capacity of the tissue (in the linear-quadratic sys- tem described by the α/β ratio) and its time kinetics (halftime, T½). It is effective in the dose ranges between 100 cGy/min (60 Gy/hr) and 0.1 cGy/min (6 cGy/hr). The higher the dose rate,

Fig 5.3: Definitions of time-dose patterns used in brachytherapy for cervix treatment. Overall treatment times (blue) and irradiation times (red) are presented for different types of treatment. (ICRU Report 88, 2015)

• Very Low Dose Rate (VLDR) irradiation by permanent im- plants delivers a high total dose (for example 150 Gy) over several weeks to months. It is evident that the biological effects resulting from the different dose–time patterns will lead to different biological effects. How- ever, they may have the same biological efficacy.

5. THE FOUR R’S OF RADIOBIOLOGY AND THE DOSE RATE EFFECT

Several factors influence the response of normal tissues and tu- mours to therapeutic radiation exposure, which are often sum- marized as the Rs of radiotherapy [Withers 1975, Steel, Dörr and Van der Kogel 2009, Dörr 2015, Dörr and Schmidt 2014). • The intrinsic radiosensitivity of a given organ, which mainly depends on the number and sensitivity of the tissue specific stem cells (the stem cell hypothesis), but also on the sensitivity and interaction of other cell populations (e.g. endothelial cells, fibroblasts) in a complex way.

Radiobiology of LDR, HDR, PDR and VLDR Brachytherapy

7

THE GEC ESTROHANDBOOKOF BRACHYTHERAPY | Part I: The Basics of Brachytherapy Version 1 - 22/10/2015

Fig 5.6 : Increasing biological effectiveness with increasing dose rate. Modified from Hall 1964 HeLa cell survival in function of dose rate (Hall). The arrows indicate the separation between the LDR, MDR and HDR dose rates according to the ICRU 38 definitions (ICRU 1985)

Fig 5.4: The Four R’s of Radiobiology : Repair, Redistribution, Reoxygenation , Repopulation in function of Dose Rate or Treatment Time to deliver 2 Gy (after Steel 1986).According to ICRU LDR ranges from 0.6 to 3.3 cGy / minute, HDR starts from dose rates over 20 cGy / min. (ICRU 1985)

Fig 5.7 : Variation of radiobiological effect in function of changes in dose rate. The highest dose rate effect variation is noticed in the MDR range (Van Limbergen 1987)

Fig 5.5. Cell survival according to dose rate. Increasing the dose rate leads to more cell kill for the same Total Dose. Line A corresponds to a dose rate level where no repair of sublethal dam- age is done. Line B corresponds to a dose rate so low that not only repair but also repopulation is possible. The clinical zones of LDR (40- 200 cGy/h) and HDR (> 12 Gy /h) are indicated

the less time is left for recovery, leading to higher tissue effects at higher dose rates (as demonstrated for an in-vitro situation) up to a dose rate where no recovery is possible at all (line A in Fig 5.5). Further increases in dose rates lead to overkill and no further effects. Lowering the dose rate allows for a progressive increase in recovery, down to a dose rate where maximum recov- ery takes place (line B in Fig 5.5). The differences in biological effectiveness according to dose rate effects are also illustrated in an RBE/dose rate plot (Fig 5.6). This dose rate effect leads to a substantial increase in cell kill by a factor of 2- 4when comparing the effect of LDR to HDR. The variation of biological effect with changing dose rate is smaller (with a slower slope) in the LDR and HDR ranges, and signifi- cantly larger (with a steeper slope) in the MDR range (Fig 5.6). Altering dose rate will hence result in a larger dose rate effect in this MDR region in and around a radioactive source (Van Lim- bergen 1987) (Fig 5.7).

Fig 5.8: The effects of sublethal damage, progression in cell cycle, and repopulation on survival rate, according to dose rate.

Repopulation (see above) is the slowest process.

The relative importance of recovery, repopulation, and reoxy- genation is dependent on the dose rate. Repopulation is possible during low dose rate irradiation or in protracted schedules in combination with external beam irradiation, while recovery has a dominant effect in all the dose rate ranges used in BT. In case of very low dose rate, repopulation of clonogenic tumour cells may play an important role. (Fig 5.8)

Radiobiology of LDR, HDR, PDR and VLDR Brachytherapy

8

THE GEC ESTROHANDBOOKOF BRACHYTHERAPY | Part I: The Basics of Brachytherapy Version 1 - 22/10/2015

6. MATHEMATICALMODELLINGOF DIFFERENT DOSE RATES AND THE EQD 2 CONCEPT

T 1/2 = 30min to 1 h for early-reacting normal tissues and tumours. T 1/2 = 1.5 h for late-reacting normal tissues. Within the time range of conventional low dose rate BT, some- where between 3 and 10 days (0.3 to 1 Gy/hr), recovery kinetics represent an important factor for calculating equi-effective treat- ments (Scalliet 1987). For an α/β ratio of 10 Gy (early effects), the slope of the isoeffect curve (see below) critically depends on the recovery kinetics value. For an α/β ratio of 3 Gy (late effects), recovery kinetics do not play the same central role between 3 and 10 days, and the slope of the isoeffect curve depends much less on T 1/2 value. However, for longer times, as with permanent implants, recovery kinetics become essential for equi-effective dose calculations. Reoxygenation is a relatively slow process, and it could be a dis- advantage in low dose rate irradiation. The total duration of the treatment usually does not exceed a few days, and reoxygena- tion due to the elimination of well oxygenated cells and tumour shrinkage cannot occur by the end of the treatment. However, other and faster mechanisms are implicated (Dörr 2009). One of them is recirculation through initially closed vessels (see above). A temporary increase in blood flow could lead to acute reoxy- genation of hypoxic cells, and the OER associated with low dose rate irradiation has been estimated to be as low as 1.6-1.7 (Bed- ford, Ling 1985). Repopulation is the slowest process and is of significance only for applications lasting more than a few weeks, i.e. with permanent implants. 6.3 Permanent Implants at Very Low Dose Rate Both paladium -103 and iodine -125 encapsulated sources are widely used in permanent implants mainly of prostate cancer. The dose outside the implanted volume falls off very rapidly because both radioactive isotopes emit low energy X-rays in the range 20-30 keV, a major advantage as far as radioprotection is concerned. The relative biological effectiveness (RBE) of radiation varies with radiation quality because of differences in the spatial pat- tern of energy deposition. The range of secondary electrons in water depends upon their initial energy. For example, 20 and 350 keV electrons have a LET of 1.3 keVμm and 0.25 keVμm, corresponding to a range of 9.0 and about 1000 μm, respective- ly. These wide differences account for a measurable variation in biological effectiveness. Compared with cobalt -60 , iodine -125 has a RBE in the range 1.4 – 2.0. Although obtained with different biological systems and endpoints, RBE values of 1.15 - 1.2 are in general observed for high dose and higher dose rate [Scalliet and Wambersie 1988]. On the other hand, values up to 2.0-2.4 (Ling 2000, Scalliet and Wambersie 1988) are observed at low dose or lower dose rate, which is consistent with microdosime- tric data. The lower RBE is relevant to temporary implants with high activity iodine -125 seeds such as eye plaques and the higher RBE to permanent implants with low activity seeds at an initial dose rate of 7 cGy /h. Palladium -103 has a slightly higher LET than iodine -125 . Its initial RBE value at 14 cGy /h is estimated to be1.9 (Ling 2000). Practically, the existence of a RBE larger than 1 implies a differ- ent biological effectiveness per Gy delivered. In this particular

The fractionation/dose rate effect has been studied in many ex- perimental systems and clinical applications of radiotherapy. A shift of the dose effect curves with an increasing number of fractions/decreasing dose per fraction or decreasing dose rate has always been observed. The simplest mathematical model to describe this shift in equi-effective doses is the linear-quadratic (LQ-) model (Bentzen et al. 2012)

d + α/β X + α/β

D =

EQDX

[1]

α/β

6.1 HDR BT The radiobiological processes involved in high dose rate BT are in all respects similar to those involved in fractionated external beam radiation therapy, except for the volume effect and the non-uniform dose distribution, as mentioned earlier. The total effect E can be calculated as follows:

E

= αD + βD 2

[2]

HDR

6.2 LDR BT

The biological effect of IR decreases as the dose rate decreases.

Recovery is a dynamic process, following specific kinetics. For practical purpose, kinetics have been assumed to follow a sim- ple exponential function of time. Kinetics can be described by the half-time of recovery T 1/2 , In conditions of irradiation where recovery can start to take place during exposure, i.e. low dose rate irradiation, the LQ model is modified by incorporation of an incomplete recovery factor g , and equation [3] is modified to

E

= αD + βgD 2

[3]

LDR

g depends in a complex way upon the half-time for recovery T 1/2 and the duration of exposure t according to the relation:

[4]

g = 2 [ t – 1 +exp - t ] / ( t) 2

where μ is a constant, which is dependent on the half time of recovery: μ = Log e 2 / T 1/2 = 0.693/T 1/2 . The value of g is 1 for brief exposures (t tends to 0) and it tends to 0 for very long exposures (it tends to ∞). This modified version of the LQ model is called the “incomplete repair model” (Dale 1985). were estimated experimentally (Ang, Hall, Scalliet 1987, 1988, 1989), but dose rates lower than 1 Gy/h (i.e. continuous irradiation lasting longer than 24 hours) have been rarely used. The few available human data are derived from external irradiation in breast cancer or estimated from BT clinical data (Larra 1977, Leborgne 1996, 1999, Mazeron 1991a, 1991b, Steel 1987, Thames 1990, Turesson 1989). The following approximate values are frequently used although there is no con- clusive evidence from the literature: Recovery T 1/2 for tumours and normal tissues are less well estab- lished than α/β values. Most T 1/2

Radiobiology of LDR, HDR, PDR and VLDR Brachytherapy

9

THE GEC ESTROHANDBOOKOF BRACHYTHERAPY | Part I: The Basics of Brachytherapy Version 1 - 22/10/2015

the decrease in radiation effectiveness due to the reduction in dose rate is partially compensated for by an increase in RBE. Tu- mour shrinkage, when present, also compensates to some extent for radioactive decay by decreasing the distance between adja- cent sources. Complex models are required to describe the inter- play of these various factors (Dale1989, Dale1994). During this long period, repopulation occurs, and the irradia- tion becomes ineffective when the dose rate has decreased to a “critical value”, which is just insufficient to compensate for the effects of repopulation of tumour cells (Dale 1989). This com- pensation dose (M) can be calculated according to the formula: Let us consider, for example, a permanent implant of 125 I sources with an initial dose rate of 0.07 Gy/h. The total dose that will be delivered is 150 Gy. We assume a constant T pot of 6 days. The “critical dose” is 120 Gy, and is reached after 140 days (23.5 times T pot ) when the dose rate has decreased to 0.014 Gy/h. We can the estimate the dose used to compensate for the effects of repopu- lation using formula [6]. It is 47 Gy. The effective dose delivered is then 120 Gy - 47 Gy = 73 Gy (corrections are not made for variations in RBE). In addition, because the dose rate has been continuously very low, the equivalent dose delivered to late re- sponding normal tissues might also be low, but conclusive data are lacking. Experimental data systematically exploring the variation of RBE of palladium -103 and iodine -125 with the dose rate of exposure, relative to iridium -192 and cobalt -60 , are not yet available. It has been suggested from biological modelling that palladium -103 with the higher dose rate could be more effective in high grade pros- tate cancer with a half time of repair shorter than 25 days (King 2000). However this could not be confirmed in a matched pair analysis (Cha 1999) when 125 I and 103 Pd gave equivalent clinical outcomes and survival, regardless of Gleason score and initial PSA. In another phase III Trial (Herstein et.al 2005) more proc- titis was seen with palladium and more urethritis with Iodine but resulting at 12 months in the same IPSS score. 6.4 PDR BT Pulsed Dose BT was developed in the early nineties in order to mimic the biological effect of continuous low dose rate BT, while taking advantage of the stepping source technology developed for high dose rate BT. Source strength was reduced to about 37 GBq (1 Ci) instead of 370 GBq (10 Ci) for an HDR source. The total dose is delivered in the same total time as with continuous low dose rate treatment, but with a large number of small frac- tions (called pulses), typically one per hour, up to one per 4 h, (Fig 5.10). The radiobiological modelling of pulsed dose rate is difficult, due to numerous uncertainties regarding recovery parameters (see above). Theoretical pulsed dose rate protocols, which could sim- ulate a continuous low dose rate treatment, have been worked out (Brenner 91a, 91b, 95 97 Fowler 92, 93, Mason 1994). Their conclusions were quite similar regarding the need to deliver pulses of at least 10 minutes per hour with a source having the lowest possible activity. It must be emphasized once more that these calculations are based on many hypotheses concerning the M= 2Gy t .T pot -1 [6]

Fig 5.9 Survival curve according to the linear-quadratic model.

Fig 5.10: Continuous LDR and several PDR schedules all delivering dose at the average of 50 Gy/h over the whole treatment time. Biological effects however may be very different.

case, iodine or palladium sources are more efficient per Gy for the same dose rate than, for example, external beam irradiation with megavoltage equipment. A second particular feature of permanent implants with iodine and palladium seeds is that the total dose is delivered over an extended period, until the sources are decayed. The actual dose rate ( DR t ) is calculated from the initial dose rate (DR o ) and the half-life constant (λ) as follows: [5] The radioactive half-lives are 60 days for iodine -125 and 17 days for palladium -103 . The initial dose rate at the time of implantation is about 0.08 to 0.1 Gy/h for iodine and 0.18 to 0.2 Gy/h for palla- dium. The corresponding total absorbed doses are 160 Gy (over 1 year) and 115 Gy (over 3 months), respectively. Because of the radioactive decay, the dose rate steadily decreases throughout ir- radiation, with a corresponding increase in RBE. The biological equivalence of the final dose is quite complex to calculate since t = DR o . e -λt

DR

Radiobiology of LDR, HDR, PDR and VLDR Brachytherapy

10

THE GEC ESTROHANDBOOKOF BRACHYTHERAPY | Part I: The Basics of Brachytherapy Version 1 - 22/10/2015

Fig 5.11: Volume Biological Effect of 4 different PDR schedules, compared to CLDR 50 Gy/h. 70 Gy/h in function of pulse size (4 diagrams) and different half lines of repair. When pulse doses exceed 1 Gy isoeffect to CLDR is only present when half line of repair is slower than 30 minutes. After Fowler and Van Limbergen (Fowler 1997)

Fig 5.12 A: Double plane breast implant with 9 source tracks and 7 mean central dose points(a to g). B: Dose rate in central dose point c during a pulse with the stepping source travelling through the nine source channels. Note that the dose in c is delivered by a few dwell positions close to the central dose point c. (with thanks to Barbara Geelen, Delft, The Netherlands)

Radiobiology of LDR, HDR, PDR and VLDR Brachytherapy

11

THE GEC ESTROHANDBOOKOF BRACHYTHERAPY | Part I: The Basics of Brachytherapy Version 1 - 22/10/2015

recovery half times for early and late responding normal tissues as well as tumours (Brenner 95, Fowler92-93). As stated previ- ously, available data on the kinetics of recovery are scanty. An important point is that in the early theoretical calculations (Brenner 91b, 95 97 Fowler 92) as well as in the animal exper- iments (Armour 1997,Brenner 1995,Haustermans 1997,Mason 1994) the ”pulses” were given with external beam equipment, e.g. 50 cGy in 10 minutes (with a dose rate of 5 cGy per min- ute, i.e. 300 cGy/hr or MDR), instead of with a stepping source that “walks through the target” delivering nearly all the dose to the target with a few dwell positions at dose rates higher than 20 cGy/minute or 1200 cGy/hr or definitely HDR.(see Fig 5.12) This has been called the “golf ball” effect by Fowler and Van Lim- bergen (Fowler 1997). Biologically equi-effective doses calculat- ed without accounting for this effect are usually underestimating the tissue toxicity and pulsed dose rate appears “hotter” than expected on the base of the older calculations and experiments. In summary, pulsed dose rate BT behaves biologically like «hy- per fractionated high dose rate with incomplete recovery be- tween the pulses. PDR mimics continuous low dose rate treat- ment only when pulse sizes are small (<0.5 Gy). In those cases, the differential effect to CLDR is less than 10 % (Fig 5.11). With larger pulse sizes and shorter recovery times, the effects in tissues with a lower α/β value are expected to be significantly larger. Mathematical modelling of the total effect E of repeated pulses taking into account incomplete recovery leads to the following equation (Thames 1985): where Hm is the incomplete repair factor which depends upon the number of fractions per day (m), the interval between frac- tions Δ T, the half time of repair T 1/2 and the repair constant μ (for details see Bentzen et al. 2012.) values for human tissues in situ is probably the biggest area of uncertainty of these estimations (Fowler 1993, Fowler 1995). Two regimens have been proposed: • The same total dose as with LDR irradiation, the same total duration, pulses repeated each hour or every two hours with an actual pulse size of not more than 1 Gy/h. This irradiation would reproduce the effects of low dose rate irradiation with a reduction in therapeutic ratio of not more than 10%, whatever the value of T 1/2 for late reacting normal tissues (Fowler 1997) • A fewer number of pulses per day, with intervals between puls- es as long as 3-4 hours (and sometimes a break at night), a sim- ilar total duration and reduced total dose. The equivalent dose was estimated to be acceptable, provided T 1/2 is short for early effects (0.5-1 hour), and long for late effects (3-4 hours) (Bren- ner 1997, Visser 1996). In contrast, a big reduction in therapeu- tic ratio would be observed if an unexpectedly short T 1/2 were observed in late responding normal tissues (compared with tumours) (Fig 5.11). Estimation of equivalent dose is much more complex for PDR treatment than for LDR and HDR treatments, and cannot be rea- sonably done manually. It assumes a constant dose rate during pulses, which may be low or medium according to ICRU defini- tions. It does not take into account the fact that the miniaturised source advances step by step inside the catheters. The dose rate The lack of knowledge for T 1/2 PDR = αD + βd . (1 + Hm). D [7]

in a given point in the target volume therefore varies between low and high values during the pulse, and PDR BT might behave more like HDR than LDR BT with incomplete recovery between the pulses and consequently pulse size rather than dose rate within the pulse will be the dominant factor (Fig 5.11) (Fowler 1997).

6.5 The EQD2 concept In order to compare the biological effects of the different dose rate and fractionation schedules that are used in HDR, PDR and LDR BT , and also to make dose additions to external beam radi- otherapy possible, the GEC-ESTRO promotes use of the equi-ef- fective dose concept of ICRU (Bentzen et al. 2012). Treatment schedules of external beam and BT are recalculated according to the LQ model (with incomplete recovery for PDR and LDR) and are expressed as an equi-effective total dose, as if it were given in 2 Gy fractions (EQD2). This leads to the use of a common language that has been shown to be very practical and useful in the comparisons of HDR, PDR and LDR BT for cervix cancer. The EQD2 formula is a simple formula based on the linear quad- ratic model of radiation effect and on the mono exponential model of recovery kinetics (see above). It includes the tissue- and endpoint-specific recovery parameters α/β and T 1/2 . Equieffectiv- ity can be calculated for each tissue of interest and each endpoint with the relevant parameters, when they are known. However most often average α/β and T 1/2 values are used for early (10 Gy and 1h) and for late reacting tissues (3 Gy and 1.5 h) without any concern for the inherent uncertainties. A fundamental warning is necessary when considering equi-ef- fect calculations based on the LQ model. There are almost no in vivo experimental data exploring the dose rate effect beyond 24 - 30 hours of continuous irradiation. Therefore the mathe- matical models, and more particularly the incomplete recovery model, have not yet been properly validated at dose rates rele- vant to classical LDR. There is also controversy about the validity of the LQ formalism at large doses per fraction (Brenner 2008, Kirkpatrick, et al . 2008), It is believed that the model adequately quantifies the biological effects in most tissues in the dose range of 0.5 Gy to 5 Gy to 6 Gy. At dose per fraction exceeding 6-10 Gy, the LQ formalism may overestimate the biological effect and the LQC model might be more appropriate (Bentzen, et al . 2012, Joiner 2009).These considerations must be taken into account by the clinician when prescribing high fraction sizes as in HDR BT for gynaecological tumours, prostate and breast where a trend to deliver fraction sizes of 6 Gy has become current practice. Moreover, equi-effectivities vary widely with variations of α/β and T 1/2 . Thus a simple “magical” formula equating HDR with LDR is probably a dangerous illusion. The responsible clinician must always make decisions with caution. Of course these parameters can be changed when new data and insights lead to another consensus. It is therefore important that the real physical data on absorbed dose distribution, dose rate, pulse and fraction sizes is available in the treatment report. In practice, starting from the total effect E of a particular frac- tionated external beam or HDR (see formula 1), an equi-effec-

E

Radiobiology of LDR, HDR, PDR and VLDR Brachytherapy

12

THE GEC ESTROHANDBOOKOF BRACHYTHERAPY | Part I: The Basics of Brachytherapy Version 1 - 22/10/2015

tive schedule assuming that irradiation was given at 2 Gy per fraction would become:

EQD2

= D (α/β + d) / (α/β + 2)

[8]

HDR

This is assumed to give a good estimation of the equivalent dose of fractionated external beam and HDR. Starting from formula 3, the EQD2 value for a LDR schedule can then be calculated as:

EQD2

= D (α/β + gDR) / (α/β + 2)

[9]

LDR

For comparing different LDR dose rates we must first translate continuous LDR into an equi-effective fractionated HDR dose using the Liversage formula:

Fig 5.13: Dose and dose gradient distribution within an interstitial implant. The dose is pre- scribed at a peripheral reference isodose (D100 or D90). In the irradiated area higher doses and dose rates lead to a much higher biological effect than expected from physical dose distribution alone.

N = μt/{2 [1-1/μt (1 - e -μt )]}

Where N is number of fractions into which the HDR treatment must be divided in order to be equi-effective to the LDR treat- ment lasting t hours if both total time and total dose remain con- stant. When t exceeds 10 hours, the exponential term becomes negligi- ble and the formula is simplified to:

early reactions, and 186 Gy for late reactions. At the 30 Gy iso- dose, the dose rate is 0.21 Gy/h. The equivalent doses are 28Gy and 24 Gy, respectively. We can then calculate that the biological equivalent doses vary for a physical dose gradient of 4 by a factor 4.7 for early reactions and 6.8 for late reactions. The biological gradient is thus much steeper than the “physical” gradient. Let us now consider a 42 Gy HDR irradiation delivered in 6 frac- tions. At the 84 Gy isodose, the dose per fraction is 14 Gy, and the equivalent doses are 119 Gy and 143 Gy for early and late effects, respectively. At the 21 Gy isodose, the dose per fraction is 3.5 Gy, and the equivalent dose 17 Gy and 14 Gy. We can then calculate that the equivalent doses vary by a factor of 7 for early reactions, and 10 for late reactions. In the concept of the equivalent uniform dose (EUD), the inho- mogeneous dose distribution within one volume is converted to an homogeneous dose which would result in the same survival. Hence the concept is based on a homogeneous distribution of morbidity factors – or tissue tolerance. This might be applicable to early reactions and tumour responses. However for late effects this might be less clear, since these effects are based on a variety of target cells and their interactions.

N =μt/[2 (1-1/μt)]

When t approaches 100 hours, the last term becomes negligible and the formula can be simplified again; it becomes

N = μt/2 and d = 2.9 T 1/2

. DR where DR is the dose rate in Gy/h.

Combining this with formula [10] will simplify the formula to:

EQD2

= D (α/β + 2.9 T

. DR) / (α/β + 2)

[10]

LDR

1/2

Starting from formula [7] EQD2 values for PDR schedules can be calculated as follows:

[11]

EQD2

PDR = D { α/β + (1+Hm )d} / (α/β + 2)

7. BIOLOGICAL EFFECTS OF DOSE INHOMOGENEITY

In a BT implant, the dose gradient distribution is also a dose rate gradient distribution (Fig 5.13). The dose is prescribed at a peripheral reference isodose, the Minimum Target Dose (MTD) = D100, which should encompass the target, or at 90 % of the MTD = D90. Within the irradiated area higher doses delivered at higher dose rates (LDR) fraction sizes HDR) or pulse sizes (PDR) will lead to much greater biological effects than expected from physical dose distribution alone. Let us consider a classical low dose rate continuous irradiation of 60 Gy in 6 days and study the variation in biological effectiveness between 30 Gy and 120 Gy; a dose and dose rate gradient of a factor 4: At the isodose receiving 120 Gy, the dose rate is 0.83 Gy/h. Using formula [9] we can estimate the equivalent dose to be 133 Gy for

8. VOLUME, ANATOMICAL SITE, AND PATIENT-RELATED EFFECTS

It has been shown in animal studies as well as in clinical data that the total dose required to sterilise tumours increases with increasing tumour volume, but that at the same time the sensi- tivity of late responding normal tissues increases with increasing GTV/PTV. The volume of healthy tissues included in the plan- ning target volume is one of the major parameters of treatment morbidity. Mathematical models of the normal tissue compli- cation probability (NTCP), like the Lyman-Kutcher–Burmann model (1989) have been introduced to link the irradiated volume – in addition to the treatment protocol - to the complication rate.

Radiobiology of LDR, HDR, PDR and VLDR Brachytherapy

13

THE GEC ESTROHANDBOOKOF BRACHYTHERAPY | Part I: The Basics of Brachytherapy Version 1 - 22/10/2015

combined with external beam radiation therapy, acting as boost, notably in cancer of the cervix. In other situations, such as for small cancers of the mobile tongue, attempts to replace exclu- sive low dose rate BT by combined external radiotherapy and BT boost led to a decrease in local control, and sometimes to an increase in complications (Benk 1990, Pernot 1994,Pernot 1997,Wendt 1989). One randomised trial has compared a boost with EBRT or BT in breast cancer. Fourquet et al . reported on 255 patients present- ing with large (3 - 7 cm) breast tumours, who were treated with EBRT to the whole breast (58 Gy) with a 20 Gy boost to the tu- mour bed, either with conventional cobalt -60 irradiation or with an interstitial iridium 192 implant (mean dose rate: 0.64 Gy/h) (Fourquet 1995). The 8-year local control rates were 61% and 76%, respectively (p = 0.02). These and other retrospective data confirm the dose effect of the higher EUD resulting from BT as compared to external beam treatments to the same nominal pre- scription dose (see table 5.1) 9.2 Dose rate effects in low dose rate BT It has become clear that dose rate effects are strongly depend- ent on repair capacity and kinetics and that there exist strong differences in dose rate effects between tumour responses and complication rates. Tumour control In squamous cell carcinoma, a tumour assumed to have a high α/β ratio of about 10 Gy, no strong dose rate effects have been reported Pierquin et al . reviewed the local outcome of 263 squamous cell carcinomas of the oral cavity, the lower lip, the skin, and the pe- nis, implanted with iridium 192, to deliver a dose of 70 Gy in 3 to 8 days (Pierquin 1973). They did not find an effect of overall treatment duration on either probability of local control or com- plications within this range. Similar conclusions were drawn by Fu et al. in oral tongue cancers (Fu1976). Larra et al . did not find an effect of overall time between 1 and 10 days on the control of 121 skin carcinomas implanted to a dose of 60 Gy (Larra 1977). The same was concluded in the GEC-ESTRO survey on BT for lip cancer, where no significant dose rate effect was noted in the dose rate range from <40 to >120 cGy/hr. Van Limbergen et al. also observed no strong dose rate effect with differences in local control between 3 and 6 days with tu- mours of the uterine cervix treated with a dose of 60 Gy (Van Limbergen 1985).This observation was confirmed later, by a ran- domised trial in cervix cancer (Haie1994) where no dose rate effect was seen either in tumour cell sterilisation in the postoper- ative specimen or in relapse free survival. Different observations were seen for breast cancer by Mazeron et al . who observed an effect of dose rate on local control in a population of 340 patients with a T1-3 adenocarcinoma of breast treated with a 37 Gy iridium -192 boost. Recurrence rates varied from 31% at dose rates 0.3 – 0.4 Gy/h to 0 % at dose rates 0.8 - 0.9 Gy/h (Mazeron 1991).These findings are compatible with radiobiological modelling from later data of breast cancer hypof- ractionation studies which estimate a lower α/β ratio of about 4.1 Gy for breast cancer. (Yarnold 1994)

For the same CTV, the treated volume in BT is significantly smaller than with EBRT, since in principle no PTV margins have to be taken to cover the target. Assuming a CTV of 4 x 4 x 2 cm, this would need a PTV of 32 cm 3 in case of BT, but would be expanded to 75 cm 3 if a 0.5 cm margin and 144 cm 3 when a 1 cm PTV margin is chosen in EBRT. Thus, treated volumes will increase by a factor 2.3 to 4.3 depending on the chosen PTV margin. This corresponds with clinical data from breast boost implants where the average treat- ed volume is 3 times smaller (50 cm 3 versus 150 cm 3 ) for external beam boosts. However these smaller volumes have not been as- sociated with lower local control rates (Table 5.1). The probability of late effects also depends upon the type of tis- sue involved (see above, e.g. QUANTEC). For example, the same technique of BT applied to oral tongue and floor of mouth carci- nomas leads to a rate of soft tissue necrosis more than two times higher in the floor of mouth than in the oral tongue [Mazeron 1991]. This may be related to the different radiopathologies of the individual tissues/tissue components. Table 5.1 Local recurrence rates of breast cancer treated with RT only or Breast Conserving surgery and RT, either with external beam boost (EBB) with photons or electrons versus interstitial BT boost (BTB). * Randomized phase III trial. Retrospective non randomized data. RT ONLY* EBB BTB Fourquet 1995 5y 30% 16% BCS + RT** Mansfield 1990 10y 18% 8% p= 0.2 Touboul 1995 5y 8.8% 5.5% p= 0.32 Hammer 1995 5y 8.2% 4.3% p= 0.03 Polgar 2001 5y 5.8% 7.7% p= 0.69 Poortmans 2004 5y 4.5% 2.5% p= 0.09 Verhoeven 2015 10y 2.5% 0.7 % p= 0.11

9. CLINICAL RESULTS: LDR, MDR, HDR AND PDR BT

Many clinical data have been accumulated over the years in BT, but there have been very few randomised trials. Nevertheless, these retrospective studies help us understand better the biolog- ical background of BT and devise rules that can be followed in clinical practice.

9.1 BT versus external beam radiation therapy While BT has been a very popular treatment for about a centu- ry, most studies are retrospective. In fact, BT has been used as a standard treatment since the 1920s in many tumours, such as cancers of the cervix, oral cavity, lip, penis, etc. It was originally delivered as the only treatment. Later, it was often successfully