Risk estimation of second primary cancers after breast radiotherapy

Abstract

Aims: There is evidence towards the induction of second primary cancers (SPCs) after breast radiotherapy (RT). Organs, such as the lungs and the esophagus, have been identified as common sites for SPC formation. As a result, the current study investigated the risk of secondary carcinogenesis associated with particular RT techniques for breast cancer; including whole breast, segmented breast, partial breast and mammosite brachytherapy.

Methods: In this study, seven breast cancer patients had all major organs contoured on their planning computed tomography (CT) images. Whole breast, segmented breast, accelerated partial breast irradiation (APBI) and mammosite boost treatment plans were generated for each patient using Pinnacle3 treatment planning system. Differential dose-volume histograms were generated for a number of critical structures: bladder, brain and central nervous system (CNS), breast, colon, liver, lung, mouth and pharynx, esophagus, ovary, salivary gland, small intestine, stomach, and uterus. The lifetime attributed risk (LAR) of cancer induction was estimated using the Schneider et al. excess absolute risk models and dose-volume histograms for the above organs.

Results: The sites with the highest LAR estimates were the ipsilateral and contralateral lungs, and contralateral breast for all treatment techniques. For all sites, the LAR estimates for the segmented breast and mammosite treatments were lower than those for the whole breast and APBI treatments. For right-sided target volumes the liver also resulted in high LAR estimates, with all techniques having a LAR greater than 20 per 10 000 person-years (PY), except for mammosite with a mean LAR estimate of 13.2 per 10 000 PY. For left-sided target volumes the stomach also resulted in high LAR estimates, with both whole breast and APBI having a LAR greater than 20 per 10 000 PY, and mammosite the lowest with a LAR of 8.3 per 10 000 PY.

Conclusion: It is concluded that the lungs and contralateral breast showed high LAR estimates.

Breast cancer is the most common malignancy among women worldwide. Surgery, chemotherapy and radiotherapy (RT) are common ways to manage the disease. As the vast majority of breast cancer patients undergo irradiation there is need to evaluate the secondary effects of radiation on these patients. Estimates of these effects are especially important for breast cancer patients of young age [1].

A large body of evidence has accumulated in recent years on the induction of second primary cancers (SPCs) after breast RT [2]. Organs, such as the lungs and the esophagus, have been identified as common sites for SPC formation after breast RT. At present, there is a variety of RT techniques available to patients with breast cancer that may be associated with different morbidity risks [3].

Johansen et al. investigated the risk of contralateral breast SPC induction after RT. In their studies they compared conventional [three-dimensional conformal radiotherapy (3D-CRT)], intensity-modulated RT (IMRT) and intensity-modulated arc therapy (IMAT). It was concluded that in general no increased detriment was found for the IMAT techniques compared with conventional, however, there was a potentially higher risk associated with IMRT treatments [4,5]. Johansen et al. also compared a linear and non-linear risk model, and found that the choice of model significantly affected the interpreted risk, with the non-linear model estimating a lower risk for all three techniques [4,5].

Aziz et al. compared the risk of SPC induction for intraoperative radiotherapy (IORT) to mammosite boost and conventional external beam radiation therapy (EBRT). The risk of SPC induction was estimated using the NCRP report 116 for treatment plans generated on an anthropomorphic phantom. The study reported a reduced risk estimate associated with IORT and mammosite, compared to EBRT [6].

Pignol et al. compared the risk of SPC induction for whole breast, IMRT, APBI, multi-catheter high dose rate (HDR) brachytherapy and permanent breast seed implants with 103Pd seeds. Treatments were modeled for left breast targets using MCNP Monte-Carlo on an anthropomorphic phantom. The NCRP report 116 was used to estimate the risk of breast and lung SPC mortality. It was reported that the HDR brachytherapy technique results in increased risk to the lung compared to the other treatment techniques [7].

Ng et al. studied the risk of lung SPCs associated with whole breast RT, comparing the prone and supine treatment techniques. A spontaneous and radiation-induced carcinogenesis model was developed and used to estimate the SPC induction risk [8–10]. It was reported that the patient setup position significantly affected risk, with the prone breast irradiation resulting in a reduced lung SPC risk [11].

The aforementioned reports commonly focused on a limited number of organs (one or two) and treatment techniques, and generally on treatment plans generated for an anthropomorphic phantom. In the current study the risk of carcinogenesis associated with RT techniques, i.e. whole breast, segmented breast, partial breast and mammosite brachytherapy for breast cancer are estimated for seven patients. The organs investigated include: the contra lateral breast, lung, liver, esophagus, and stomach.

Clinical findings

Patients and setup

In the current study, seven female patients were evaluated. These seven patients were selected for a mammosite trial at the Royal Adelaide Hospital, five patients being treated for their left breast, and two for their right breast. All patients were scanned using a computed tomography (CT) scanner in the prone position, with their arms positioned over their heads. The target volume (PTV) and the organs at risk (OAR) were contoured and checked by a radiation oncologist (RO). For the purpose of this study, the RO contoured various OARs to be evaluated, including bladder, brain and CNS, breast, colon, liver, lung, mouth and pharynx, esophagus, ovary, salivary gland, small intestine, stomach, and uterus.

Treatment planning

For each of the seven patients a treatment plan for each of the techniques was performed. As we were comparing four different techniques, an overall of 28 plans were produced. All treatments were planned on a Pinnacle3 v9.2 (Philips Radiation Oncology Systems, Fitchburg, WI) treatment planning system (TPS). For EBRT, the TPS model for a Varian iX linear accelerator was used. This machine has 6 MV and 18 MV X-ray beams and a 120 multi-leaf collimator (MLC). In the case of the mammosite treatment planning, the brachytherapy module in Pinnacle3 v9.2 was employed to model a 192Ir source. For evaluation the dose grid encompassed the whole CT, with a grid resolution of 0.2 cm × 0.2 cm × 0.2 cm.

Whole breast treatments were defined to be wedged tangential X-ray fields, with energy of either 6 MV, 18 MV or a combination of both. RO contoured baseline structures defined the target area on the patient, with beam isocenters placed approximately in the center of the target breast. MLCs were then utilized to conform to the baselines contoured, and approximately 2 cm beyond the breast. A prescription point was chosen based on the ICRU 50 reference point recommendations [12], with a prescription of 2 Gy per fraction for a total of 25 fractions.

Segmented breasts were defined to be whole breast treatments, where multiple MLC segments could be used, hence a forward planned step and shoot treatment. The use of MLC segments helps to achieve improved dose conformity. No more than three segments were used for all patients.

Accelerated partial breast irradiation (APBI) treatments are a conserving treatment, where the target volume is only a region of the breast, unlike the whole breast irradiation. APBI treatments were planned based on the Trans Tasman Radiation Oncology Group study 06.02 protocol “trefoil” technique, which uses three non-coplanar fields. A 6 MV photon beam was used for all treatment plans with a prescribed dose of 3.85 Gy for 10 fractions.

Mammosite boost treatments consisted of a brachytherapy balloon (∼35 mm diameter) inserted into the tumor bed inside the breast. An Ir-192 source, used for HDR brachytherapy was positioned in the center of the balloon. The prescribed boost dose consisted of 3.5 Gy for 10 fractions, to a depth of 1 cm from the balloon’s surface.

For each plan, the differential dose-volume histograms (DVHs) were computed and exported for analysis. A summary of the treatment techniques is shown in Table 1.

Table 1. Summary of the treatment planning techniques used for generating the breast treatment plans.

Treatment	Irradiation technique	Dose prescription	Mean irradiatedvolume (cm^3)
Whole	2 × 6 MV/18 MV	2 Gy for 25#	1170 ± 205
Segmented	2 × 6 MV/18 MV	2 Gy for 25#	1170 ± 205
APBI	3 × 6 MV	3.85 Gy for 10#	391 ± 130
Mammosite	^192Ir HDR	3.5 Gy for 10#	101 ± 24

Lifetime attributed risk estimate

The lifetime attributed risk (LAR) of cancer induction described by BEIR VII [13] was estimated in this study from the calculated differential DVHs using MATLAB. LAR is defined by Equation (1)(1) : (1)

Equation (1)(1) integrates the age of attainment, agea, from the age of exposure, agex, to the rest of the patient’s life span, accounting for the probability of survival to the age of attainment. While the patient ages ranged from 50 to 70 years, an agex of 30 years was used for all patients. This younger age was used in order to evaluate the higher risk to the younger breast cancer patients (i.e. the worst case scenario). The Australian female survival rates, S, were obtained from the Australian Bureau of Statistics [14]. The Schneider et al. excess absolute risk (EAR) models, were employed in the estimations [15], given in Equation (2)(2) : (2)

Where β_EAR is the initial slope, V(D_i) is the volume of the organ to receive a dose D_i, and V_T is the total volume of the organ. μ(agex, agea) is the modifying function which contains population dependent variables given by Equation (3)(3) . RED represents the risk equivalent dose mechanistic model, which accounts for cell killing and fractionation effects given in Equation (4)(4) : (3) (4)

In Equation (3)(3) , γ_e and γ_a are the age modifying parameters. R is the repopulation/repair parameter and a’ is given by Equation (5)(5) : (5) where D_T and d_T represent the prescribed dose to the target volume and the corresponding dose per fraction, respectively.

The risk of SPC has been investigated in the contra lateral breast, lung, liver, esophagus, and stomach. The model described by Schneider et al. includes a number of organs: female breast, lung, rectum, colon, mouth and pharynx, stomach, small intestine, liver, cervix, bladder, skin, brain and CNS, thyroid and salivary gland [16]. In the case of the breast, the latest reported parameters were applied [17].

For the esophagus, data from [17] was used, which evaluated esophageal cancer risk after breast RT. EAR was 0.58 per 10 000 person-years (PY) per Gy from atomic bomb survivors [18]. The Schneider et al. model parameters were fitted to this data. All of the parameters utilized in this study are shown in Table 2.

Table 2. The age modifying and dose-response model parameters used for the LAR estimate.

Site	γe	γa	βEAR	α	R
Breast	−0.037	1.7	8.2	0.044	0.15
Lung	0.002	4.23	8.0	0.042	0.83
Liver	−0.021	3.6	2.4	0.323	0.29
Stomach	−0.002	1.9	5.2	0.460	0.46
Esophagus	−0.002	1.9	0.58	0.026	0.81

Results

Dose-volume analysis

The organs which received the highest doses were the contralateral breast, both lungs, the liver, the stomach, and the esophagus. Figure 1 shows the volume-normalized cumulative DVHs for the organs which received the highest doses. These show the mean, maximum and minimum DVHs from all the treatment plans generated. Table 3 presents values from the volume-normalized DVHs, where the mean dose (D_50%) and a maximum dose (D_5%) were chosen for comparison. The dose values presented in Table 3 have been converted to equivalent doses in 2 Gy fractions (EQD₂) for comparison [19], where an α/β ratio of 3 was used for the investigated organs. It can be seen that left-sided targets result in significant dose increase to the stomach for all treatment techniques, while right-sided targets result in dose increases to the liver for all treatments techniques. APBI generally resulted in a reduced dose, while whole breast treatments were generally higher.

Figure 1. The volume-normalized cumulative DVHs for the contralateral breast for: (a) right-sided target and (b) left-sided target, the contralateral lung for: (c) right-sided target and (d) left-sided target, the ipsilateral lung for: (e) right-sided target and (f) left-sided target, the liver for: (g) right-sided target and (h) left-sided target, the stomach for: (i) right-sided target and (j) left-sided target, and the esophagus for: (k) right-sided target and (l) left-sided target. The continuous lines depict the mean DVHs, and the broken lines depict the maximum and minimum DVHs.

Table 3. The D_5% and D_50% for the highest OARs for all treatments, separated into right- and left-sided targets. The D_5% and D_50% doses have been converted to EQD₂ doses.

		Right-sided target				Left-sided target
Organ		Whole	Segmented	APBI	Mammosite	Whole	Segmented	APBI	Mammosite
Contralateral breast	D5% (Gy)	0.70	0.33	0.37	0.45	0.74	0.34	1.97	0.37
	D50% (Gy)	0.07	0.07	0.04	0.16	0.10	0.05	0.05	0.15
Contralateral lung	D5% (Gy)	0.45	0.20	0.11	0.50	0.39	0.14	0.10	0.32
	D50% (Gy)	0.08	0.08	0.04	0.24	0.09	0.07	0.04	0.20
Ipsilateral lung	D5% (Gy)	26.98	27.76	7.99	4.04	8.60	1.98	6.42	2.47
	D50% (Gy)	0.54	0.41	0.24	0.94	0.41	0.18	0.21	0.56
Liver	D5% (Gy)	2.24	2.65	1.25	1.85	0.42	0.17	0.10	0.48
	D50% (Gy)	0.34	0.28	0.13	0.39	0.09	0.07	0.04	0.21
Esophagus	D5% (Gy)	0.68	0.22	0.14	0.73	0.36	0.13	0.10	0.42
	D50% (Gy)	0.16	0.14	0.08	0.41	0.17	0.11	0.07	0.34
Stomach	D5% (Gy)	0.24	0.11	0.19	0.31	0.84	0.60	0.19	1.02
	D50% (Gy)	0.08	0.08	0.10	0.15	0.34	0.17	0.10	0.59

For the contralateral breast, little difference is observed between right- and left-sided targets. The whole breast treatments resulted in a higher maximum dose compared to the other treatment techniques. The APBI did result in a significantly higher maximum dose of 2.99 Gy for the left-sided targets. The mammosite boost treatment techniques resulting in a higher mean dose of 0.27 Gy.

In the case of the contralateral lung, both the mammosite boost treatment the whole breast treatment techniques resulted in the highest maximum dose of greater than 0.5 Gy, while the mammosite still having a significantly higher mean dose than the other treatment techniques of greater than 0.3 Gy.

For the ipsilateral lung, the right-sided targets were always observed to result in a higher mean and maximum dose compared to the left-sided targets, especially with the whole breast and segmented breast generally resulting in the highest doses of greater than 30 Gy.

In the case of the liver and the esophagus, the right-sided targets always resulted in a higher dose. For the liver this is expected since this organ is located on the right side of the patient. Generally, the APBI resulted in a lower mean and maximum dose, which is associated with the choice of beam arrangement.

For the stomach the left-sided targets were observed to result in a higher mean and maximum dose, generally over twice the dose of the right-sided targets. The mammosite boost resulted in the highest mean and maximum doses of 0.96 Gy and 1.62 Gy for the left-sided targets.

Lifetime attributed risk

Figure 2 shows the mean, maximum and minimum LAR estimates calculated for each of the treatment techniques. The sites with the highest LAR estimates were the ipsilateral and contralateral lungs, and contralateral breast for all treatment techniques. Table 4 shows the mean LAR values for those sites. Generally, the LAR estimates for the APBI treatments were the lowest and those estimated for the mammosite treatment were the highest among the reported treatment techniques.

Figure 2. The LAR estimated risk where (a) is for left-sided breast targets and (b) is for right-sided breast targets.

Table 4. Summary of the mean LAR estimated risk for all the OARs at highest risk, separated into right- and left-sided targets.

	Mean LAR (per 10 000 PY)
	Right-sided target				Left-sided target
Organ	Whole	Segmented	APBI	Mammosite	Whole	Segmented	APBI	Mammosite
Contralateral breast	13.6	9.6	11.0	16.2	17.3	7.8	17.8	14.7
Contralateral lung	53.7	40.1	20.3	107.1	53.2	29.1	19.2	87.9
Ipsilateral lung	466.0	462.8	292.1	435.9	229.7	183.2	303.0	291.6
Liver	25.6	24.0	13.2	22.7	9.7	6.4	3.4	15.4
Esophagus	1.5	1.0	0.5	2.4	1.3	0.7	0.4	2.1
Stomach	7.3	6.1	2.9	8.5	22.2	14.6	8.3	27.1

For the contralateral breast and lung, there was little difference between the right- and left-sided targets. Generally, the mammosite technique was observed to result in the highest LAR estimate, and the APBI and segmented breast in the lowest LAR estimate. Interestingly, the segmented breast technique resulted in a lower LAR than APBI for the contralateral breast but a higher LAR for the contralateral lung. This is the consequence of the non-coplanar beam arrangement used in the APBI technique.

The LAR estimates for the ipsilateral lung were the highest of all the organs of up to 466 per 10 000 PY, with the contralateral lung having the second highest LAR estimates of up to 107.1 per 10 000 PY. For the ipsilateral lung, interestingly the left-sided targets had a reduced mean LAR estimate compared with the right-sided target. For the right-sided targets, the APBI technique resulted in the lowest LAR of 292.1 per 10 000 PY, while for the left-sided target the APBI technique resulted in the highest LAR of 303 per 10 000 PY.

For the liver the right-sided target led to high LAR estimates, with all techniques having a LAR greater than 20 per 10  000 PY, except for APBI with a mean LAR estimate of 13.2 per 10 000 PY. Among the left-sided targets only mammosite had greater than 10 per 10 000 PY.

For left-sided target volumes the stomach also resulted in high LAR estimates, with both whole breast and mammosite having a LAR greater than 20 per 10 000 PY, and APBI the lowest with a LAR of 8.3 per 10 000 PY. In the case of the right-sided targets, all treatment techniques had a mean LAR of less than 10 per 10 000 PY.

For the esophagus, only whole breast and mammosite techniques resulted in a mean LAR estimate of greater than 1 per 10 000 PY for left-sided target volumes. For right-sided target volumes the esophagus LAR was generally higher, with APBI technique resulting in the lowest LAR.

Discussion

There are several aspects to be taken into account when estimating the risk of cancers from RT. These include the uncertainties associated with the accuracy of the TPS models and that of the risk models themselves. For the EBRT models, the doses to the OARs are generally out of field, for which the TPS models usually do not have a high accuracy as they do not account for head scatter and leakage. Monte-Carlo modeling could be used instead to account for this limitation in the future.

Similarly, for the HDR plan the model is based on TG-43 dose calculation method, which assumes a homogenous water medium. Hence, inhomogeneities are not accounted for. Furthermore, the 192Ir source model has only anisotropic function data for depth below 9 cm, thus TG-43 U1 recommends the same value to be used for depths beyond 9 cm [20,21]. However, this approximation does increase the uncertainties in dose at the aforementioned depths. Monte-Carlo modeling or the TG-186 calculation method [22] can be employed to overcome this limitation.

The use of these SPC risk models should be used with caution as the interpretation of results may be significantly altered by the choice of SPC model used. For example, the use of linear models have been shown to considerably overestimate the risk of SPC induction [23,24].

Real patient treatment plans and anatomical data have been used in this study rather than modeled dose distributions in an anthropomorphic phantom. Data of seven patients was used as only seven patients were recruited into the mammosite clinical trial conducted at the Royal Adelaide Hospital. However, as clinical data has been used, follow-up of these patients can be performed in relation to late effects including the incidence of SPCs and the outcomes can be compared with the predictions of this study.

Conclusion

In summary, a cancer induction model has been simulated in MATLAB to estimate the LAR of SPC formation after exposure to ionizing radiation, and applied to breast cancer RT whole breast, segmented breast, APBI and mammosite boost treatment plans were generated for seven breast cancer patients, and the DVHs used for estimating the risk of SPC induction. The lungs and contralateral breast showed high LAR estimates, results that are in accordance with reported clinical studies.

The liver also showed some large LAR estimates, especially for the right-sided target patients. The stomach showed large LAR estimates for whole breast and mammosite treatments of the left-sided targets. Generally, low LAR estimates, below 1 per 10 000 PY, were observed for the esophagus.

Overall, results show that the APBI technique leads to the lowest risk estimate for SPC formation. Generally, the mammosite and whole breast treatment techniques resulted in a higher LAR estimate.

Disclosure statement

There is no conflict of interest. No funding has been received for this work.