Potential Effect of Robust and Simple IMRT Approach for Left-Sided Breast Cancer on Cardiac Mortality

Introduction 

Breast-conserving therapy (BCT) is the standard therapy for early-stage breast cancer. Postoperative radiotherapy (RT) after lumpectomy lowers the rate of local relapse. Although no clear survival benefit could be demonstrated in early series, more recent analyses, among them the most recent Early Breast Cancer Trialists' Collaborative Group overview, have suggested such a benefit 1, 2.

Any survival advantage as a consequence of improved local control and prevention of secondary metastases was probably offset by the increased cardiac mortality secondary to radiation-induced heart disease in early series because of technically inferior treatment techniques (large heart volume irradiated in patients with left-sided cancer owing to a lack of three-dimensional [3D] treatment planning) 3, 4, 5, 6, 7, 8, 9 and generally larger target volumes that more frequently included the parasternal lymph nodes.

Three-dimensional planning has reduced the cardiac dose and thus cardiac mortality 7, 8, 10, 11, 12, but the potential for cardiac toxicity has recently increased again owing to the more widespread use of anthracyclines/taxanes/trastuzumab in the adjuvant treatment of early-stage breast cancer (13).

Toxicity modeling using data from patient series treated for breast cancer and for Hodgkin's disease has suggested that a reduction of the maximal dose/greater doses to the left heart might be particularly beneficial, even at the expense of an increase in the heart volume treated to lower doses 14, 15. Patients with a heart anatomically close to the thoracic wall, which, in the pre-computed tomography era was quantified using the maximal heart distance (16), in particular, might benefit from such a strategy. Although partial breast RT offers a possibility to reduce cardiac exposure, it is experimental. Also, because a reduction of the treated volume by heart blocks might result in an increased relapse risk (17), partial breast RT might not be suitable for all patients.

Intensity-modulated RT (IMRT) offers the possibility to spare the left heart, especially when the internal mammary chain is to be treated (18), but also in the seemingly “simple” situation of breast-only therapy 18, 19, 20, 21, 22.

Although several problems encountered with IMRT are secondary to the highly conformal dose distribution (e.g., robustness to interfraction movement), others are specific to breast treatment (e.g., coverage of the “flash region” in the conventional tangent setup, high sensitivity to intrafraction movements). Various strategies have been devised to overcome these problems, such as the virtual bolus concept (19) and the manual extension of the fluence matrix into air (23).

An aperture-based approach with manual or semiautomatic segment creation and monitor unit (MU) optimization (instead of full inverse treatment planning) might help to address these problems favorably. It has been successfully implemented clinically at our department for selected high-risk patients with left-sided breast cancer.

We compared this approach to three-dimensionally planned tangent RT in patients with left-sided breast cancer with extremely unfavorable chest–heart geometry. The physical characteristics of the dose distribution, the technique's robustness to positioning errors, and the requirements for treatment time are reported. To assess the possible effect of IMRT on survival in this high-risk population, normal tissue complication probability (NTCP) calculations were performed using the relative seriality model of Kaellman et al. (25) (parametrized by Gagliardi et al. [14]), estimating the normal tissue response to nonuniform dose distributions and modeling the organ as a combination of serial and parallel subunits.

Results 

The dose distributions in the traverse, coronal, and sagittal plane for the IMRT plan and conformal 3D plan are shown in Fig. 1. The planes shown were not at the isocenter but most clearly showed the difference between the conventional and IMRT plans.

View Large Image Download to PowerPoint

Fig. 1. Dose distributions in transversal, coronal, and sagittal plane for (a) intensity-modulated radiotherapy and (b) conformal three-dimensional plans.

Figure 2 shows the heart DVHs of all 14 patient plans for IMRT and 3D conformal RT, and Table 3 provides an overview of the relevant plan parameters (i.e., the mean, median, and partial-volume doses to the heart, left ventricle, lung, contralateral breast, and spinal cord), reported as the mean ± standard deviation of the individual values for each patient.

View Large Image Download to PowerPoint

Fig. 2. Heart dose–volume histograms of all 14 patient plans for (a) intensity-modulated radiotherapy and (b) three-dimensional conformal radiotherapy. Although heart volume treated to doses >30 Gy was minimal for intensity-modulated radiotherapy plans, mean and median heart dose were increased compared with three-dimensional conformal radiotherapy plans.

Table 3. Clinically relevant plan parameters

	Risk structure partial dose volume	Mean (Gy)	Median (Gy)	Maximum (Gy)	Dose to 60% of volume (Gy)	Dose to 30% of volume (Gy)
	Heart
	IMRT	8.52 ± 1.08	6.44 ± 1.02	35.5 ± 4.33	5.91 ± 1.68	9.22 ± 2.21
	3D-CRT	6.85 ± 2.51	2.77 ± 0.64	49.56 ± 1.43	2.8 ± 0.59	3.63 ± 0.97
	Left ventricle
	IMRT	9.7 ± 2	7.86 ± 2.27	33.97 ± 3.9	6.77 ± 2.04	11.08 ± 3.29
	3D-CRT	10.86 ± 4.12	3.75 ± 1.66	49.14 ± 1.24	3.17 ± 0.97	6.22 ± 3.48
	Right breast
	IMRT	5.4 ± 1.18	4.4 ± 0.91	23.39 ± 4.95	4 ± 0.99	5.75 ± 1.53
	3D-CRT	1.15 ± 0.76	0.61 ± 0.7	18.82 ± 17.52	0.37 ± 0.6	1.26 ± 1.08
	Right lung
	IMRT	4.24 ± 1.26	3.22 ± 1.23	19.89 ± 5.92	2.70 ± 1.27	5.52 ± 1.64
	3D-CRT	0.45 ± 0.15	0.08 ± 0.02	3.99 ± 1.23	0.06 ± 0.05	0.32 ± 0.42
	Left lung
	IMRT	9.84 ± 1.35	6 ± 1.68	42.37 ± 2.4	4.58 ± 1.23	10.1 ± 2.97
	3D-CRT	8.36 ± 3.13	2.5 ± 0.57	51.8 ± 1.68	2.15 ± 0.5	4.15 ± 1.48
	Spinal cord
	IMRT	1.26 ± 0.86	8.1 ± 25.5	4.22 ± 1.9	0.75 ± 0.64	1.59 ± 0.95
	3D-CRT	0.25 ± 0.43	0.16 ± 0.24	1.4 ± 1.67	0.1 ± 0.15	0.3 ± 0.75

Abbreviations: IMRT = intensity-modulated radiotherapy; 3D-CRT = three-dimensional conformal radiotherapy. Data presented as mean, median, and partial volume dose ± standard deviation to target and organs at risk for each plan delivered to total dose of 50 Gy as mean dose to planning target volume.

For each individual patient, IMRT consistently reduced the maximal dose to the heart, as well as the fraction of the heart volume that received >30 or >40 Gy. The maximal dose to the left ventricle was reduced by a mean of 30.9% (49.14 ± 1.24 vs. 33.97 ± 3.9 Gy). The mean dose to the left ventricle was reduced by an average of 10.7% (10.86 ± 4.12 vs. 9.7 ± 2 Gy) because decreasing the maximal dose reduced the mean dose in the small volume of the left ventricle effectively. The average volume of the heart exposed to >30 Gy was reduced from of 45 ± 16 cm3 to 5.84 ± 5.98 cm3. The average volume exposed to >40 Gy was reduced from 35 ± 13.8 cm3 to 0.17 ± 0.46 cm3. The coverage of the PTV was only slightly inferior with IMRT (95% of PTV covered with 45 Gy) than with tangent treatments (95% of PTV covered with 47 Gy). The homogeneity of the IMRT plans was acceptable, with low fraction volumes of the treated breast that received >110% of the prescribed dose (3.8% ± 1.87%).

The IMRT plans increased the mean heart dose to 8.52 ± 1.08 Gy compared with 6.85 ± 2.51 Gy with 3D-conformal RT. The same was true for the median heart dose (6.44 ± 1.02 Gy with IMRT vs. 2.77 ± 0.64 Gy with 3D-conformal RT). The mean risk of excess cardiac mortality was dramatically reduced from 6.03% (conventional) to 0.25% (IMRT) according to the relative seriality model.

The mean total body (integral) dose increased (7.11 ± 1.76 Gy vs. 3.56 ± 1.61 Gy for the partial body volume scanned with computed tomography for treatment planning), and the same was observed for the contralateral breast (from 1.15 ± 0.76 Gy to 5.4 ± 1.18 Gy). Also, a clear increase occurred in the mean dose to the left lung, which, given the overall low numbers, was probably clinically irrelevant.

Figure 3 shows the DVHs for the original plan with the DVHs for the PTV superimposed after a 1-cm movement of the isocenter anteriorly, posteriorly, and caudally, without major effects on the dose distribution (dose to 95% of PTV >43.5 Gy for posterior isocenter displacement and >45.6 Gy for caudal displacement). The small reduction in the PTV dose for the anterior isocenter displacement (dose to 95% of the PTV still 42.5%) was secondary to the steep dorsal dose gradient toward the heart.

View Large Image Download to PowerPoint

Fig. 3. Dose–volume histograms for original plan with dose–volume histograms after 1-cm movement of isocenter anteriorly, posteriorly, and caudally superimposed, without major effects on dose distribution.

The IMRT time was similar for all patients, with an average number of segments of 59 (range, 40–78) and an average number of MUs of 584 (range, 443–683), resulting in average treatment times of <15 min (dose rate, 600 MU/min). The conventional plans had an average number of MUs of 248 (range, 224–344).

Discussion 

As reviewed by several groups 2, 7, 8, 15, 27, 28, 29, 30, 31, the analyses of patients treated with RT for early-stage breast cancer, postoperatively after resection of lung cancer or for Hodgkin's disease suggested that the benefit of RT with regard to survival was counteracted by what seemed to be radiation-induced cardiac mortality. The cardiac dose of breast RT has decreased over the years owing to improvements in technique and the paradigm shift toward breast-conserving therapy (12). Although, initially, cardiac toxicity seemed to even outweigh the benefits of adjuvant RT (32), the most recent Oxford overview finally indicated an RT-associated survival advantage (2). Even more recently, however, the more widespread use of anthracyclines, taxanes, and trastuzumab in the adjuvant setting (13) has again increased the potential for cardiac toxicity.

Even with tangent treatments after breast-conserving surgery, RT to the left breast itself could still be problematic in a subset of patients 16, 33 in whom the heart is situated anteriorly in the concavity of the target volume owing to the size and shape of the thoracic cage. Therefore, either a significant amount of the left heart will be irradiated with the prescription dose or underdosage of some of the PTV will have to be accepted. Although partial breast RT might be an option for these patients, as long as the precise indications for partial breast RT have not been identified, it seems prudent not to compromise on whole breast coverage, if possible (17).

Tangent fields with an optimized angle and optimized fluency 16, 34, 35 can only optimize the tradeoff between PTV coverage and the dose to the heart, because it results in extended blocking of the heart.

Full IMRT has its special strength when geometrically complex treatment situations such as bilateral cancer, pectus excavatum (19), or the indication to cover the parasternal and supraclavicular/axillary lymph nodes arise 18, 21, 36. It yields concave dose distributions, as shown by groups from Sweden, Germany, and Canada 18, 19, 37 with special characteristics of the dose distribution compared with tangents, including a reduction of the volume treated to high doses and increasing the volume treated to lower doses (i.e., an increase in the mean and median heart dose).

Although a significant reduction in the maximal dose to the heart could be shown for patients in whom the internal mammary chain was to be treated (18), we also demonstrated this reduction in patients with “unfavorable” geometry in whom the internal mammary chain was not included. The aperture-based treatment planning approach that we used proved to be fast, very MU efficient, and robust with regard to smaller positioning errors. Although we did only assess PTV coverage as a function of patient shift, Gagliardi et al. (37) also studied the influence of positioning errors on NTCP and found only minimal differences for patient displacements of 3 and 5 mm. For our approach, therefore, as with any other IMRT technique creating steep dose gradients, the thoracic wall where the gradient toward the heart is created must be positioned correctly, preferably by image guidance. Small interfraction shifts of the mobile breast will be compensated for by our geometrically robust technique. The short treatment time will render additional immobilization to prevent intrafraction movement unnecessary.

Modeling cardiac toxicity is complicated for several reasons, as discussed in the review by Gagliardi et al. (15), including organ at risk definition (heart or left ventricle or myocardium), scarce dosimetric data for historical techniques, and a latency of symptoms of >10 years). Although more reliable predictors of endpoints such as pericarditis or myocardial perfusion exist, modeling cardiovascular disease, arguably a very relevant endpoint, is particularly difficult using currently available data. An assumption that has been made for modeling is the hypothesis of a homogeneous radiation sensitivity of the heart or the myocardium volume. According to this assumption, Gagliardi et al. (15) were able to plot dose–response curves for different partial volumes of the heart. As stated by Gagliardi et al., this assumption underlying their model is not necessarily true, as indicated by the different model parameter values obtained for patients with breast cancer vs. patients with Hodgkin's disease. An analysis of the data from patients treated for Hodgkin's disease with RT only using the same method of analysis used for breast cancer patients yielded a dose–response curve for cardiac mortality that seemed to be shallower than that for breast RT. Remembering that the dose distribution for Hodgkin's disease is almost complementary to the one for breast RT (emphasizing the dose to the right heart and sparing the outermost left part), this also points to the direction of an increased radiation sensitivity of the left anterior cardiac wall. It is, therefore, an appealing hypothesis to link the dose to the left anterior descending coronary artery to the incidence of lethal cardiac toxicity 38, 39.

Because considerable controversy exists about the crucial question regarding to what extent lower and, especially, intermediate doses to the heart are toxic 40, 41, 42, 43, 44 and what parameter is the most important for cardiac toxicity (e.g., the dose to the left heart or the mean heart dose), the determination of a threshold of volume treated to prescription doses, beyond which the dose characteristic created by IMRT might be beneficial, will be instrumental for the optimal use of multifield IMRT. The relative seriality model could be used to determine such a threshold and therefore form the basis of a hypothesis that then could be tested clinically.

We applied this model to our patient cohort that would have the greatest risk with tangent RT. Our findings indicated a dramatic reduction in the risk of excess cardiac mortality, along the lines of the sample calculations of Gagliardi et al. 14, 37. Although the mean excess NTCP (cardiac mortality) for tangent RT in their series of 100 consecutive patients was ∼2%, they reported a subset of patients with an excess NTCP of 4–5%. Given that we worked with a cohort of patients selected because of their unfavorable anatomy, our series is probably comparable to their group with the greatest NTCP. Because the patient for whom the reduction in NTCP using IMRT was modeled was drawn from this high-risk cohort, the similar order of magnitude of the NTCP reduction (3–4% to 0.05% for their sample patient and 6.03% to 0.25% [mean values] for our patients) makes sense (37). On the basis of these data, we now treat such patients with multifield IMRT clinically.

Because of the long latency period of clinical symptoms, the introduction of this treatment paradigm should be accompanied by studies with surrogate endpoints for cardiac toxicity. Although, to date, no correlation between such surrogate studies and clinical toxicity has been found (simply because these studies were initiated <10 years ago), perfusion defects in heart muscle irradiated to high doses can be observed 6–24 months after RT and are stable for >5 years 45, 46, 47. Such perfusion measurements, using single photon emission tomography or dynamic magnetic resonance imaging and functional studies with ultrasonography or magnetic resonance imaging could prove to be good predictors of clinical cardiac toxicity and might indicate early whether a new treatment is really beneficial with regard to cardiac toxicity. Current efforts with regard to a more detailed analysis of historic patient cohorts will provide valuable information regarding the dose–volume parameters most important for the development of cardiac toxicity (12).

The increase in the mean dose to the contralateral breast must be included in the decision. Although the effect of the contralateral breast dose in breast RT is not completely clear 2, 48, 49, evidence from the atomic bomb survivor series has shown that doses in the range of 1–4 Gy increase this risk (50), although the absolute excess risk will probably be small compared with the base risk of 15% after >15 years.

Conclusion 

The results of our study have shown that for left-sided breast cancer, multifield IMRT significantly reduces the maximal dose to the left ventricle. This could significantly reduce the cardiac mortality for a high-risk population, at the expense of a larger heart volume treated to intermediate doses, although the clinical significance of this is still unclear. The choice of therapy modality might be determined from predictive models such as the relative seriality model that has been fitted to cardiac mortality data. The outcome must be closely monitored, with functional studies providing early assessment of cardiac perfusion/function. The presented IMRT technique is robust toward the influence of motion on coverage of the anterior breast volume.