- Email: [email protected]
Temporal Lobe Reactions After Carbon Ion Radiation Therapy: Comparison of Relative Biological Effectiveness–Weighted Tolerance Doses Predicted by Local Effect Models I and IV
International Journal of
Radiation Oncology biology
physics
www.redjournal.org
Physics Contribution
Temporal Lobe Reactions After Carbon Ion Radiation Therapy: Comparison of Relative Biological EffectivenesseWeighted Tolerance Doses Predicted by Local Effect Models I and IV Clarissa Gillmann,* Oliver Ja¨kel, PhD,*,y,z Ingmar Schlampp, MD,* and Christian P. Karger, PhDz *Department of Radiation Oncology and Radiation Therapy, Heidelberg University Hospital, Heidelberg, Germany; y Heidelberg Ion Beam Therapy Center (HIT), Heidelberg, Germany; and zDepartment of Medical Physics in Radiation Oncology, German Cancer Research Center (DKFZ), Heidelberg, Germany Received Sep 24, 2013, and in revised form Dec 2, 2013. Accepted for publication Dec 24, 2013.
Summary This study investigates which version of the local effect model (LEM I vs LEM IV) predicts the relative biological effectiveness (RBE) for temporal lobe reactions more accurately. The derived dose-response curves indicate that the RBE-weighted tolerance doses are 9.5 Gy (RBE) higher for LEM IV than for LEM I. The final decision as to which model better fits clinical experience in photon therapy requires additional analysis of a comparable photon-treated patient collective using the same dosimetric variable.
Purpose: To compare the relative biological effectiveness (RBE)eweighted tolerance doses for temporal lobe reactions after carbon ion radiation therapy using 2 different versions of the local effect model (LEM I vs LEM IV) for the same patient collective under identical conditions. Methods and Materials: In a previous study, 59 patients were investigated, of whom 10 experienced temporal lobe reactions (TLR) after carbon ion radiation therapy for low-grade skullbase chordoma and chondrosarcoma at Helmholtzzentrum fu¨r Schwerionenforschung (GSI) in Darmstadt, Germany in 2002 and 2003. TLR were detected as visible contrast enhancements on T1-weighted MRI images within a median follow-up time of 2.5 years. Although the derived RBE-weighted temporal lobe doses were based on the clinically applied LEM I, we have now recalculated the RBE-weighted dose distributions using LEM IV and derived dose-response curves with Dmax,V-1 cm3 (the RBE-weighted maximum dose in the remaining temporal lobe volume, excluding the volume of 1 cm3 with the highest dose) as an independent dosimetric variable. The resulting RBE-weighted tolerance doses were compared with those of the previous study to assess the clinical impact of LEM IV relative to LEM I. Results: The dose-response curve of LEM IV is shifted toward higher values compared to that of LEM I. The RBE-weighted tolerance dose for a 5% complication probability (TD5) increases from 68.8 3.3 to 78.3 4.3 Gy (RBE) for LEM IV as compared to LEM I. Conclusions: LEM IV predicts a clinically significant increase of the RBE-weighted tolerance doses for the temporal lobe as compared to the currently applied LEM I. The limited available photon data do not allow a final conclusion as to whether RBE predictions of LEM I or LEM IV better fit better clinical experience in photon therapy. The decision about a future clinical application of LEM IV therefore requires additional analysis of temporal lobe reactions in a comparable photontreated collective using the same dosimetric variable as in the present study. Ó 2014 Elsevier Inc.
Reprint requests to: Clarissa Gillmann, Department of Radiation Oncology and Radiation Therapy, Heidelberg University Hospital, Im Neuenheimer Feld 400, 69120 Heidelberg, Germany. Tel: (49) 62215634979; E-mail: [email protected] Conflict of interest: none. Int J Radiation Oncol Biol Phys, Vol. 88, No. 5, pp. 1136e1141, 2014 0360-3016/$ - see front matter Ó 2014 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.ijrobp.2013.12.039
AcknowledgmentsdThe authors thank Michael Scholz, PhD, Thomas Friedrich, PhD, and Rebecca Gru¨n for valuable comments on the manuscript.
Volume 88 Number 5 2014
Introduction As the world’s third facility, the Helmholtzzentrum fu¨r Schwerionenforschung (GSI) in Darmstadt, Germany started treatment of cancer patients with carbon ions in 1997. Within a 10-year pilot project, more than 400 patients underwent irradiation. The encouraging clinical results (1) paved the way for the construction of the Heidelberg Ion-Beam Therapy Center (HIT), a hospitalbased facility that started patient treatment in November 2009. Today, 6 carbon ion facilities are operational worldwide, and 4 more are in the advanced planning or construction phase (2). One of the major advantages of carbon ions relative to photons or protons is the increased relative biological effectiveness (RBE) in the target region relative to the entrance channel. The resulting biological effective dose (also termed RBE-weighted dose) is determined by the product of RBE and absorbed dose. RBE is a complex quantity that depends on physical as well as biological quantities. In treatment planning, it has to be derived either from simple empirical models (3) or from more sophisticated trackstructure models, such as the local effect model (LEM) (4). LEM was developed and clinically applied during the pilot project at GSI. It has subsequently also been incorporated into the treatment planning software at HIT. Thus, until now, the first version of the local effect model, LEM I, has been used to optimize the RBEweighted dose distributions of more than 1500 patients, and clinical results suggest that RBE predictions of LEM I are reasonably accurate (1, 5). On the other hand, animal studies on the tolerance of the rat spinal cord and in vitro studies on cell survival have reported that LEM I underestimates the RBE in the spread-out Bragg peak (SOBP) by 25% and overestimates the RBE for low absorbed doses in the entrance channel by 20% (6, 7). In addition, a comparison of the LEM Iebased RBE-weighted depth-dose profile with that of protons showed that high RBE values may also occur outside the target, implying that LEM I does not predict a biological advantage of carbon ions as compared to protons (8). These findings have raised some doubts about the accuracy of LEM I and motivated further developments (7) that have resulted in a new version of the local effect model, termed LEM IV (9). LEM IV considers the complexity of radiation damage in terms of the microscopic double-strand break distribution on the DNA rather than only the effect of the local dose. The modification results in a more pronounced increase of RBE with linear energy transfer (LET) from the entrance region toward the distal end of the SOBP. Benchmarking tests reported that the RBE-weighted depth-dose profile of LEM IV describes experimental in vivo and in vitro data significantly better than does LEM I (9, 10). To decide which version of the local effect model predicts RBE more reliably, the apparent contradiction between the good clinical results obtained with LEM I on one hand and the better description of experimental in vivo and in vitro data by LEM IV on the other hand has to be further investigated. In this respect, the analysis of clinical data is of major importance. A treatment-planning comparison of LEM I and LEM IV for idealized target geometries (11) showed that the predicted median RBE-weighted doses for a typical target volume are very similar for both models. In contrast to the dose in the target volume, we are now considering the dose in normal tissue. As a first step in this direction, radiation-induced temporal lobe reactions after carbon ion therapy were previously analyzed, and a dose-response curve with the LEM Iebased RBE-weighted dose as an
Clinical characterization of LEM I/IV 1137 independent dosimetric variable was derived (1). The predicted tolerance doses appeared to be comparable to those of photons available at that time, indicating a reasonable accuracy of LEM I. In the present study, we repeat this analysis for the same patient collective and under identical conditions using LEM IV instead of LEM I to calculate the RBE-weighted doses. Tolerance doses predicted by LEM I and LEM IV are reassessed by comparison to recently published clinical data, aiming to answer the question as to whether LEM I or LEM IV gives a more accurate description of the RBE in normal brain tissue.
Methods and Materials The patient collective and endpoint of this study are the same as in the previous analysis (1). Therefore only a brief description is given here.
Patient collective and follow-up This retrospective study includes 59 patients treated with carbon ion radiation therapy for chordoma and low-grade chondrosarcoma of the skull base at GSI in 2002 and 2003. Prescribed doses to the planning target volume (PTV) were 60, 66, and 70 Gy (RBE) for 49, 2, and 8 patients, respectively. Treatment was delivered in 20 fractions of 3.0, 3.3, or 3.5 Gy (RBE). For this study, the resulting total RBE-weighted doses were rescaled to treatments with 2 Gy (RBE) per fraction using the linear-quadratic model (12). To be consistent with the a/b-ratio used by LEM I and LEM IV and with that of our previous study, a/b Z 2 Gy was used (1). After completion of carbon ion treatment, all patients underwent regular clinical follow-up examinations. Median followup time was 2.5 years. Within this period, 10 patients developed temporal lobe reactions (TLR), visible as contrast enhancements (CE) on T1-weighted magnetic resonance images. Five patients developed unilateral and 5 patients developed bilateral TLR. Assuming that each temporal lobe responds independently, 15 of 118 temporal lobes responded with TLR during the follow-up period.
Dosimetric parameters The original treatment plans of all patients were biologically optimized using LEM I with the Treatment Planning for Particles (TRiP98) software (13, 14). Using LEM IV, we performed a recalculation of the biological effective dose based on the LEM Ieoptimized fluence distribution of the original treatment plan for each patient, or, in other words, the individually absorbed dose that was actually applied. The resulting RBE-weighted dose distribution of LEM IV therefore describes the case in which the prediction of LEM I is wrong and that of LEM IV is correct. As LEM IV predicts higher RBEs toward the distal edge of the SOBP, the temporal lobe dose increases. Input parameters of RBE tables were chosen according to the RBE tables AB2 I (LEM I) and AB2 IV (LEM IV) for normal tissue as given in Gru¨n et al (11). Both parameter sets use a/b Z 2 Gy, as this value has been used for all patients treated at GSI and HIT so far. For LEM IV, the set AB2 IV was used because it describes experimental in vivo data better than AB2 I (11).
1138 Gillmann et al.
International Journal of Radiation Oncology Biology Physics From the resulting RBE-weighted dose distributions, the following 16 parameters were determined for each temporal lobe: Dmin, Dmean, Dmedian, Dmax,V-xcm3, and V,D > y Gy (RBE), with x Z 0, 1, 2, 5, 10 and y Z 50, 60, 70, 80, 85, 90, 95. Dmax,V-xcm3 refers to the maximum dose in the remaining temporal lobe volume, excluding the volume of x cm3 with the highest dose. V,D > y refers to the volume of the temporal lobe that receives a minimum dose of y Gy (RBE).
Dose-response analysis In the previous analysis, Dmax,V-1 cm3 and V,D > 85 Gy (RBE) were identified as the most significant DVH variables in predicting TLR. Using the same variables, we recalculated the respective dose-response and volume-response curves for LEM IV. Depending on whether TLR did or did not occur, each individual temporal lobe was classified as responder or as non-responder, respectively, and dose-response and volume-response curves were adjusted. From these curves, tolerance doses (TD) and tolerance volumes (TV) were determined for different effect probabilities. For more details on this analysis, the reader is referred to Schlampp et al (1).
Statistical analysis Differences in dosimetric quantities between LEM I and LEM IV were tested by a paired, 2-sided Wilcoxon test using GNU R software (15). A value of P<.05 was considered statistically significant. Dose-response and volume-response curves were adjusted using the Statistica (16) logistic response model, together with a maximum-likelihood fitting procedure.
Results Dosimetric analysis Figure 1 shows a representative comparison of RBE-weighted dose distributions of both models for the temporal lobe. It can be seen that the dose gradient around the planning target volume (PTV) is stronger for LEM IV than for LEM I. Generally, a large area of the temporal lobe receives less dose in the case of LEM IV. However, small areas located directly adjacent to or overlapping the PTV are subject to higher doses. Figure 2 and Table 1 summarize the dosimetric variables for all patients: The median of Dmax increases by 22% from 74.6 to 95.6 Gy (RBE) for LEM IV as compared to LEM I, and Dmax,V-1 cm3 increases by 8% from 64.8 to 70.2 Gy (RBE). Dmean decreases by 23% from 14.7 to 11.3 Gy (RBE). The volume of the temporal lobe that receives a minimum dose of 50 Gy (RBE) is larger in the case of LEM I. V,D > 60 Gy (RBE) yields similar results for both models. For dose thresholds above 60 Gy (RBE), volume variables are collectively larger in the case of LEM IV.
Fig. 1. Comparison of (a) recalculated (local effect model [LEM IV]ebased) and (b) original (LEM Iebased) relative biological effectiveness (RBE)eweighted dose distributions, together
with a difference map (c). Contoured are the planning target volume (PTV) in red and the right temporal lobe in yellow. (d) Dose-volume histogram for the right temporal lobe. A color version of this figure is available at www.redjournal.org.
Volume 88 Number 5 2014
Clinical characterization of LEM I/IV 1139
Fig. 2. Comparison of dose-volume histogram (DVH) variables of the temporal lobe for local effect models (LEM) IV and I for all temporal lobes. Solid line indicates median; box, first and third quartiles; whiskers, 5% and 95% quantiles; circles, outliers; stars, significant difference between LEM I and LEM IV.
Dose-response analysis Figure 3 shows the resulting dose-response and volume-response curves for radiation-induced TLR based on the variables Dmax,V-1 cm3 and V,D > 85 Gy (RBE). Both curves are shifted toward higher values for LEM IV as compared to LEM I. TD5 increases from 68.8 3.3 to 78.3 4.3 Gy (RBE) from LEM I to LEM IV. Similarly, TD50 increases from 87.3 2.8 to 99.8 3.3 Gy (RBE). Table 2 summarizes the RBE-weighted tolerance doses for different effect levels. Tolerance volumes increase as well for LEM IV as compared to LEM I. The tolerance volume V,D > 85 Gy (RBE) determined at the 50% effect level increases from 2.0 0.4 cm3 to 2.8 0.4 cm3 when going from LEM I to LEM IV.
Discussion For carbon ion radiation therapy, the prediction of the biological effective dose depends on the RBE model involved in
Table 1 Comparison of median values of different dosevolume histogram variables over all temporal lobes calculated by local effect models (LEM) I and IV Median (25%/75% quantiles) Dmax [Gy (RBE)] Dmax,V-1 cm3 [Gy (RBE)] Dmean [Gy (RBE)] V,D > 50 Gy (RBE) [cm3] V,D > 60 Gy (RBE) [cm3] V,D > 70 Gy (RBE) [cm3] V,D > 80 Gy (RBE) [cm3] V,D > 90 Gy (RBE) [cm3]
LEM I
LEM IV
74.6 (71.1/78) 64.8 (59.0/73.4)
95.6 (86.9/103.5) 70.2 (62.2/83.1)
14.3 (11.2/17.2) 4.54 (2.7/8.6)
11.5 (8.3/13.9) 3.47 (2.4/5.7)
1.98 (0.8/1.9)
2.16 (1.3/3.9)
0.22 (0.0/1.9)
1.02 (0.4/2.7)
0 (0.0/0.0)
0.27 (0.0/1.4)
0 (0.0/0.0)
0.02 (0.0/0.3)
Abbreviation: RBE Z relative biological effectiveness.
treatment planning (11). The aim of the present comparative treatment planning study, therefore, was to analyze the clinical impact of a new radiobiological model (LEM IV) by comparing it against a standard model used routinely in clinical practice (LEM I) with respect to predictions of tolerances to radiationinduced TLR.
Systematic differences between LEM I and LEM IV Local maximum doses to the temporal lobe were found to be significantly increased by 20% to 30% for the recalculated LEM IVebased as compared to the original LEM Iebased dose distributions. These differences are a consequence of the stronger increase of RBE with LET predicted by LEM IV. In treatment plans, this is shown by a stronger dose gradient around the target area and a generally observed wider statistical spread of dosimetric variables for LEM IV as compared to LEM I. The reduction of the mean dose to the temporal lobe by 23% implies a correction of the systematic overestimation of RBE in the entrance channel shown by LEM I for experimental in vivo data (6). The underlying reason for the increased temporal lobe doses is related to the fact that LEM IV considers the local density of the microscopic spatial distribution of DNA double-strand breaks in the nucleus as the most relevant measure for the radiation response of cells (9). The local density of double-strand breaks is then related to the local dose. For the definition of an isoeffect required for the calculation of the RBE, it is assumed that similar double-strand break densities lead to similar effects, independent of radiation quality (9). It must be noted that LEM IV does not only include additional model features, but also uses a different set of input parameters (AB2 IV rather than AB2 I). However, as the ratio a/b has the highest impact on RBE and was set to 2 Gy for both, AB2 I (LEM I) and AB2 IV (LEM IV), the higher temporal lobe doses are expected to result from differences in the model versions rather than from differences in the input parameters. A detailed analysis of this issue was beyond the scope of this study. In our analysis, predictions of maximum doses by LEM IV were very sensitive to the exact delineation of the temporal lobe, especially if the temporal lobe overlapped the PTV. On the other hand, if there is a small gap between PTV and temporal lobe, maximum doses to the temporal lobe are lower for LEM IV as compared with LEM I. This systematic uncertainty has to be
International Journal of Radiation Oncology Biology Physics
1140 Gillmann et al.
Fig. 3. Dose-response (left) and volume-response (right) curves for radiation-induced temporal lobe reactions based on the variables Dmax,V-1 cm3 and V,D > 85 Gy (relative biological effectiveness [RBE]). For better visualization, circles indicate empirical incidence rates x/ n (reactions exhibited by x of n temporal lobes) in 10-Gy (RBE) dose-response curve, respectively 0.5-cm3 intervals volume-response curve. considered as a potential bias when comparing the doses of LEM I and LEM IV. As a consequence, differences between LEM I and LEM IV depend strongly on the selected dosimetric variable. Although TD5 and TD50 determined with LEM IV are considerably higher for the variables Dmax, Dmax,V-1 cm3, and Dmax,V-2 cm3 than those determined with LEM I, they are lower if the variables Dmax,V-5 cm3 and Dmax,V-10 cm3 are used for evaluation (Table 3). This raises the question as to which dosimetric variable correlates best with the clinically observed temporal lobe reactions. Regarding the local correlation of hot spots in the dose distribution with the appearance of TLR, one would expect that the maximum dose is suited best, eventually excluding a small volume (eg, 1 cm3) to be independent on very small hot spots in the dose distribution, where the dose calculation algorithm might not be reliable. In carbon ion therapy, the volumes of such hot spots are typically comparable to the voxel size and are thus significantly smaller than 1 cm3. Therefore our analysis focused on this dosimetric variable.
Comparison of LEM-based tolerance doses with literature data In our dose-response analysis, we showed that the biological effective TD5 and TD50 were increased by 9.5 and 12.5 Gy (RBE), respectively, when using LEM IV instead of LEM I. This difference is clinically significant and has to be considered as an uncertainty as long as it is not known which of the 2 LEM versions describes the RBE of normal brain tissue more accurately. To Table 2 Comparison of relative biological effectiveness (RBE)eweighted tolerance doses calculated by local effect models (LEM) I and IV Dmax,V-1 cm3 SE [Gy (RBE)] LEM I TD5 TD10 TD30 TD50 TD80 TD90 TD95
68.8 73.5 82.0 87.3 96.0 101.4 105.8
3.3 2.8 2.4 2.8 4.1 5.0 5.9
LEM IV 78.3 83.8 93.6 99.8 106.0 115.8 121.3
4.3 3.5 2.9 3.3 4.1 5.8 6.8
reduce this uncertainty, the tolerance doses of both models have to be compared with the clinically established values for photon irradiations using the same dosimetric variables in a comparable patient collective. In this respect, the value of TD5 is most important. Unfortunately, only very little information on TD5 is available in the literature. Su et al (17) reported a 5-year incidence of temporal lobe injury to be <5%, if a maximum dose of 64 to 68 Gy was applied to the temporal lobe. However, a large percentage of patients received additional chemotherapy, which can be expected to reduce radiation tolerance (18). Based on a single patient and using a normal tissue complication probability model, Tatsuzaki et al (19) published DVHs for the temporal lobe, from which TD5 values of 70 Gy and 72 Gy (RBE) for photons and protons, respectively, could be estimated (1). A recent study by Pehlivan et al (20) on temporal lobe toxicity after proton irradiation found that the severity of toxicity increased with increasing mean dose to a 2-cm3 volume of the temporal lobe, but no relation to complication probability or tolerance doses was given. Apart from these single studies, attempts have been made to quantify normal brain tolerance doses by retrospectively reviewing clinical data. This first resulted in the well-known “Emamidata” on normal tissue tolerance (21). Within the framework of the Quantitative Analyses of Normal Tissue Effects in the Clinic (QUANTEC) analysis (22), these data were recently refined, and the best estimate of TD5 for brain using 2-Gy fractions was reported to be 72 Gy (60-84 Gy, 95% confidence interval). For the comparison with the QUANTEC data, it has to be considered that the conversion of the total dose into 2-Gy fractions was based on a/b Z 3 Gy, although we used a/b Z 2 Gy to be consistent with the a/b-value in LEM I and LEM IV. If we had converted the fractional dose with a/b Z 3 Gy instead, our total doses would have decreased on average by 4.2% (range, 4.0%5.5% for 3.0, 3.3, and 3.5 Gy [RBE] per fraction), and the TD5values would then have been reduced to 65.9 and 75.0 Gy (RBE) for LEM I and LEM IV, respectively. Although the QUANTEC value is somewhat closer to the LEM IV value, a comparison remains difficult for 3 reasons: The QUANTEC value of 72 Gy is based (with 1 exception) on studies that specify only the prescribed rather than the local brain dose. Although this may be a reasonable approximation for the maximum brain dose in the case of brain tumors, it probably overestimates the temporal lobe dose for nasopharyngeal cancer, which constitutes about two-thirds of the patient cases in the underlying studies. With respect to a comparison to
Volume 88 Number 5 2014
Clinical characterization of LEM I/IV 1141
Table 3 Comparison of TD5 and TD50 calculated by local effect models (LEM) I and IV in dependence on the dosimetric variable TD5 [Gy (RBE)] Dmax Dmax,V-1 cm3 Dmax,V-2 cm3 Dmax,V-5 cm3 Dmax,V-10 cm3
TD50 [Gy (RBE)]
LEM I
LEM IV
LEM I
73.3 68.8 64.4 51.5 34.0
89.6 78.3 69.6 44.3 18.8
99.1 87.3 81.7 69.7 60.8
LEM IV 134.9 99.8 91.1 70.7 51.5
2.
3.
4.
5.
Abbreviation: RBE Z relative biological effectiveness. 6.
our study, the actual brain doses and hence the TD5-value are therefore likely to be overestimated, especially if Dmax,V-1 cm3 instead of the maximum dose is used as dosimetric variable. In our study, large variations in Dmax,V-1 cm3 were seen for the same prescribed dose, so a temporal lobespecific dose variable was selected for our analysis. Although the QUANTEC analysis investigated the risk of late effects per patient, we determined the risk for individual temporal lobes based on the local dose. This, together with other differences in endpoint definition and follow-up time, may introduce bias into the comparison. Finally, the 95% confidence interval of the QUANTEC TD5 is very large, most likely because of uncertainties in the actual brain doses and the pooling of several studies.
7.
8.
9.
10.
In view of these methodological issues, it is currently not possible to decide whether RBE-predictions of LEM I or of LEM IV fit better with clinical experience in photon therapy. To overcome these problems, the TD5 value for photons (or protons, assuming a fixed RBE of 1.1) has to be determined in a comparable patient collective using exactly the same dosimetric variable as in the present study.
11.
Conclusions
14.
Although LEM IV seems to give a better description of experimental in vivo data as compared with LEM I (11), a decision based on clinical data remains more difficult. The derived doseresponse curves indicate a clinically significant increase of the RBE-weighted TD5 by 9.5 Gy (RBE) for LEM IV as compared to the currently clinically applied LEM I. As the comparison of the derived tolerance doses with available photon data is associated with several methodological uncertainties, it is currently not possible to decide definitively whether RBE predictions of LEM I or of LEM IV fit better with clinical experience in photon therapy. The decision as to whether LEM IV may be clinically applied at HIT therefore requires further investigation. Such investigation includes, in particular, the analysis of temporal lobe reactions in a comparable photon-treated patient collective using the same dosimetric variable as in the present study.
15.
References 1. Schlampp I, Karger CP, Ja¨kel O, et al. Temporal lobe reactions after radiotherapy with carbon ions: Incidence and estimation of the relative
12. 13.
16.
17.
18.
19.
20.
21.
22.
biological effectiveness by the local effect model. Int J Radiat Oncol Biol Phys 2011;80:815-823. Particle Therapy Co-Operative Group (PTCOG). Available at: http:// ptcog.web.psi.ch/ptcentres.html and http://ptcog.web.psi.ch/ newptcentres.html. Most recent date of access: 2013/08/08.. Kanai T, Furusawa Y, Fukutsu K, et al. Irradiation of mixed beam and design of spread-out Bragg peak for heavy-ion radiotherapy. Radiat Res 1997;147:78-85. Scholz M, Kellerer AM, Kraft-Weyrather W, et al. Computation of cell survival in heavy ion beams for therapy. Radiat Environ Biophys 1997; 36:59-66. Schulz-Ertner D, Karger CP, Feuerhake A, et al. Effectiveness of carbon ion radiotherapy in the treatment of skull-base chordomas. Int J Radiat Oncol Biol Phys 2007;68:449-457. Karger CP, Peschke P, Sanchez-Brandelik R, et al. Radiation tolerance of the rat spinal cord after 6 and 18 fractions of photons and carbon ions: Experimental results and clinical implications. Int J Radiat Oncol Biol Phys 2006;66:1488-1497. Elsa¨sser T, Kra¨mer M, Scholz M. Accuracy of the local effect model for the prediction of biologic effects of carbon ion beams in vitro and in vivo. Int J Radiat Oncol Biol Phys 2008;71:866-872. Wilkens JJ, Oelfke U. Direct comparison of biologically optimized spread-out Bragg peaks for protons and carbon ions. Int J Radiat Oncol Biol Phys 2008;70:262-266. Elsa¨sser T, Weyrather WK, Friedrich T, et al. Quantification of the relative biological effectiveness for ion beam radiotherapy: Direct experimental comparison of proton and carbon ion beams and a novel approach for treatment planning. Int J Radiat Oncol Biol Phys 2010; 78:1177-1183. Friedrich T, Scholz U, Elsa¨sser T, et al. Systematic analysis of RBE and related quantities using a database of cell survival experiments with ion beam irradiation. J Radiat Res (Tokyo) 2012;54:494-514. Gru¨n R, Friedrich T, Elsa¨sser T, et al. Impact of enhancements in the local effect model (LEM) on the predicted RBE-weighted target dose distribution in carbon ion therapy. Phys Med Biol 2012;57:72617274. Fowler JF. The first James Kirk Memorial Lecture. What next in fractionated radiotherapy? Br J Cancer Suppl 1984;6:285-300. Kra¨mer M, Ja¨kel O, Haberer T, et al. Treatment planning for heavy-ion radiotherapy: Physical beam model and dose optimization. Phys Med Biol 2000;45:3299-3317. Kra¨mer M, Scholz M. Treatment planning for heavy-ion radiotherapy: Calculation and optimization of biologically effective dose. Phys Med Biol 2000;45:3319-3330. R Project for Statistical Computing. Available at: http://www.r-project. org. Version: 2.7. Most recent date of access: 2013/08/08. StatSoft I. STATISTICA fu¨r Windows (Software System fu¨r Datenanalyse) Version 7.1; 2011. Available at: www.statsoft.com. Most recent date of access: 2013/08/08. Su SF, Huang Y, Xiao W, et al. Clinical and dosimetric characteristics of temporal lobe injury following intensity modulated radiotherapy of nasopharyngeal carcinoma. Radiother Oncol 2012;104:312-316. Ruben JD, Dally M, Bailey M, et al. Cerebral radiation necrosis: Incidence, outcomes, and risk factors with emphasis on radiation parameters and chemotherapy. Int J Radiat Oncol Biol Phys 2006;65: 499-508. Tatsuzaki H, Urie MM. Importance of precise positioning for proton beam therapy in the base of skull and cervical spine. Int J Radiat Oncol Biol Phys 1991;21:757-765. Pehlivan B, Ares C, Lomax AJ, et al. Temporal lobe toxicity analysis after proton radiation therapy for skull base tumors. Int J Radiat Oncol Biol Phys 2012;83:1432-1440. Emami B, Lyman J, Brown A, et al. Tolerance of normal tissue to therapeutic irradiation. Int J Radiat Oncol Biol Phys 1991;21: 109-122. Lawrence YR, Li XA, el Naqa I, et al. Radiation dose-volume effects in the brain. Int J Radiat Oncol Biol Phys 2010;76:S20-S27.
Radiation Oncology biology
physics
www.redjournal.org
Physics Contribution
Temporal Lobe Reactions After Carbon Ion Radiation Therapy: Comparison of Relative Biological EffectivenesseWeighted Tolerance Doses Predicted by Local Effect Models I and IV Clarissa Gillmann,* Oliver Ja¨kel, PhD,*,y,z Ingmar Schlampp, MD,* and Christian P. Karger, PhDz *Department of Radiation Oncology and Radiation Therapy, Heidelberg University Hospital, Heidelberg, Germany; y Heidelberg Ion Beam Therapy Center (HIT), Heidelberg, Germany; and zDepartment of Medical Physics in Radiation Oncology, German Cancer Research Center (DKFZ), Heidelberg, Germany Received Sep 24, 2013, and in revised form Dec 2, 2013. Accepted for publication Dec 24, 2013.
Summary This study investigates which version of the local effect model (LEM I vs LEM IV) predicts the relative biological effectiveness (RBE) for temporal lobe reactions more accurately. The derived dose-response curves indicate that the RBE-weighted tolerance doses are 9.5 Gy (RBE) higher for LEM IV than for LEM I. The final decision as to which model better fits clinical experience in photon therapy requires additional analysis of a comparable photon-treated patient collective using the same dosimetric variable.
Purpose: To compare the relative biological effectiveness (RBE)eweighted tolerance doses for temporal lobe reactions after carbon ion radiation therapy using 2 different versions of the local effect model (LEM I vs LEM IV) for the same patient collective under identical conditions. Methods and Materials: In a previous study, 59 patients were investigated, of whom 10 experienced temporal lobe reactions (TLR) after carbon ion radiation therapy for low-grade skullbase chordoma and chondrosarcoma at Helmholtzzentrum fu¨r Schwerionenforschung (GSI) in Darmstadt, Germany in 2002 and 2003. TLR were detected as visible contrast enhancements on T1-weighted MRI images within a median follow-up time of 2.5 years. Although the derived RBE-weighted temporal lobe doses were based on the clinically applied LEM I, we have now recalculated the RBE-weighted dose distributions using LEM IV and derived dose-response curves with Dmax,V-1 cm3 (the RBE-weighted maximum dose in the remaining temporal lobe volume, excluding the volume of 1 cm3 with the highest dose) as an independent dosimetric variable. The resulting RBE-weighted tolerance doses were compared with those of the previous study to assess the clinical impact of LEM IV relative to LEM I. Results: The dose-response curve of LEM IV is shifted toward higher values compared to that of LEM I. The RBE-weighted tolerance dose for a 5% complication probability (TD5) increases from 68.8 3.3 to 78.3 4.3 Gy (RBE) for LEM IV as compared to LEM I. Conclusions: LEM IV predicts a clinically significant increase of the RBE-weighted tolerance doses for the temporal lobe as compared to the currently applied LEM I. The limited available photon data do not allow a final conclusion as to whether RBE predictions of LEM I or LEM IV better fit better clinical experience in photon therapy. The decision about a future clinical application of LEM IV therefore requires additional analysis of temporal lobe reactions in a comparable photontreated collective using the same dosimetric variable as in the present study. Ó 2014 Elsevier Inc.
Reprint requests to: Clarissa Gillmann, Department of Radiation Oncology and Radiation Therapy, Heidelberg University Hospital, Im Neuenheimer Feld 400, 69120 Heidelberg, Germany. Tel: (49) 62215634979; E-mail: [email protected] Conflict of interest: none. Int J Radiation Oncol Biol Phys, Vol. 88, No. 5, pp. 1136e1141, 2014 0360-3016/$ - see front matter Ó 2014 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.ijrobp.2013.12.039
AcknowledgmentsdThe authors thank Michael Scholz, PhD, Thomas Friedrich, PhD, and Rebecca Gru¨n for valuable comments on the manuscript.
Volume 88 Number 5 2014
Introduction As the world’s third facility, the Helmholtzzentrum fu¨r Schwerionenforschung (GSI) in Darmstadt, Germany started treatment of cancer patients with carbon ions in 1997. Within a 10-year pilot project, more than 400 patients underwent irradiation. The encouraging clinical results (1) paved the way for the construction of the Heidelberg Ion-Beam Therapy Center (HIT), a hospitalbased facility that started patient treatment in November 2009. Today, 6 carbon ion facilities are operational worldwide, and 4 more are in the advanced planning or construction phase (2). One of the major advantages of carbon ions relative to photons or protons is the increased relative biological effectiveness (RBE) in the target region relative to the entrance channel. The resulting biological effective dose (also termed RBE-weighted dose) is determined by the product of RBE and absorbed dose. RBE is a complex quantity that depends on physical as well as biological quantities. In treatment planning, it has to be derived either from simple empirical models (3) or from more sophisticated trackstructure models, such as the local effect model (LEM) (4). LEM was developed and clinically applied during the pilot project at GSI. It has subsequently also been incorporated into the treatment planning software at HIT. Thus, until now, the first version of the local effect model, LEM I, has been used to optimize the RBEweighted dose distributions of more than 1500 patients, and clinical results suggest that RBE predictions of LEM I are reasonably accurate (1, 5). On the other hand, animal studies on the tolerance of the rat spinal cord and in vitro studies on cell survival have reported that LEM I underestimates the RBE in the spread-out Bragg peak (SOBP) by 25% and overestimates the RBE for low absorbed doses in the entrance channel by 20% (6, 7). In addition, a comparison of the LEM Iebased RBE-weighted depth-dose profile with that of protons showed that high RBE values may also occur outside the target, implying that LEM I does not predict a biological advantage of carbon ions as compared to protons (8). These findings have raised some doubts about the accuracy of LEM I and motivated further developments (7) that have resulted in a new version of the local effect model, termed LEM IV (9). LEM IV considers the complexity of radiation damage in terms of the microscopic double-strand break distribution on the DNA rather than only the effect of the local dose. The modification results in a more pronounced increase of RBE with linear energy transfer (LET) from the entrance region toward the distal end of the SOBP. Benchmarking tests reported that the RBE-weighted depth-dose profile of LEM IV describes experimental in vivo and in vitro data significantly better than does LEM I (9, 10). To decide which version of the local effect model predicts RBE more reliably, the apparent contradiction between the good clinical results obtained with LEM I on one hand and the better description of experimental in vivo and in vitro data by LEM IV on the other hand has to be further investigated. In this respect, the analysis of clinical data is of major importance. A treatment-planning comparison of LEM I and LEM IV for idealized target geometries (11) showed that the predicted median RBE-weighted doses for a typical target volume are very similar for both models. In contrast to the dose in the target volume, we are now considering the dose in normal tissue. As a first step in this direction, radiation-induced temporal lobe reactions after carbon ion therapy were previously analyzed, and a dose-response curve with the LEM Iebased RBE-weighted dose as an
Clinical characterization of LEM I/IV 1137 independent dosimetric variable was derived (1). The predicted tolerance doses appeared to be comparable to those of photons available at that time, indicating a reasonable accuracy of LEM I. In the present study, we repeat this analysis for the same patient collective and under identical conditions using LEM IV instead of LEM I to calculate the RBE-weighted doses. Tolerance doses predicted by LEM I and LEM IV are reassessed by comparison to recently published clinical data, aiming to answer the question as to whether LEM I or LEM IV gives a more accurate description of the RBE in normal brain tissue.
Methods and Materials The patient collective and endpoint of this study are the same as in the previous analysis (1). Therefore only a brief description is given here.
Patient collective and follow-up This retrospective study includes 59 patients treated with carbon ion radiation therapy for chordoma and low-grade chondrosarcoma of the skull base at GSI in 2002 and 2003. Prescribed doses to the planning target volume (PTV) were 60, 66, and 70 Gy (RBE) for 49, 2, and 8 patients, respectively. Treatment was delivered in 20 fractions of 3.0, 3.3, or 3.5 Gy (RBE). For this study, the resulting total RBE-weighted doses were rescaled to treatments with 2 Gy (RBE) per fraction using the linear-quadratic model (12). To be consistent with the a/b-ratio used by LEM I and LEM IV and with that of our previous study, a/b Z 2 Gy was used (1). After completion of carbon ion treatment, all patients underwent regular clinical follow-up examinations. Median followup time was 2.5 years. Within this period, 10 patients developed temporal lobe reactions (TLR), visible as contrast enhancements (CE) on T1-weighted magnetic resonance images. Five patients developed unilateral and 5 patients developed bilateral TLR. Assuming that each temporal lobe responds independently, 15 of 118 temporal lobes responded with TLR during the follow-up period.
Dosimetric parameters The original treatment plans of all patients were biologically optimized using LEM I with the Treatment Planning for Particles (TRiP98) software (13, 14). Using LEM IV, we performed a recalculation of the biological effective dose based on the LEM Ieoptimized fluence distribution of the original treatment plan for each patient, or, in other words, the individually absorbed dose that was actually applied. The resulting RBE-weighted dose distribution of LEM IV therefore describes the case in which the prediction of LEM I is wrong and that of LEM IV is correct. As LEM IV predicts higher RBEs toward the distal edge of the SOBP, the temporal lobe dose increases. Input parameters of RBE tables were chosen according to the RBE tables AB2 I (LEM I) and AB2 IV (LEM IV) for normal tissue as given in Gru¨n et al (11). Both parameter sets use a/b Z 2 Gy, as this value has been used for all patients treated at GSI and HIT so far. For LEM IV, the set AB2 IV was used because it describes experimental in vivo data better than AB2 I (11).
1138 Gillmann et al.
International Journal of Radiation Oncology Biology Physics From the resulting RBE-weighted dose distributions, the following 16 parameters were determined for each temporal lobe: Dmin, Dmean, Dmedian, Dmax,V-xcm3, and V,D > y Gy (RBE), with x Z 0, 1, 2, 5, 10 and y Z 50, 60, 70, 80, 85, 90, 95. Dmax,V-xcm3 refers to the maximum dose in the remaining temporal lobe volume, excluding the volume of x cm3 with the highest dose. V,D > y refers to the volume of the temporal lobe that receives a minimum dose of y Gy (RBE).
Dose-response analysis In the previous analysis, Dmax,V-1 cm3 and V,D > 85 Gy (RBE) were identified as the most significant DVH variables in predicting TLR. Using the same variables, we recalculated the respective dose-response and volume-response curves for LEM IV. Depending on whether TLR did or did not occur, each individual temporal lobe was classified as responder or as non-responder, respectively, and dose-response and volume-response curves were adjusted. From these curves, tolerance doses (TD) and tolerance volumes (TV) were determined for different effect probabilities. For more details on this analysis, the reader is referred to Schlampp et al (1).
Statistical analysis Differences in dosimetric quantities between LEM I and LEM IV were tested by a paired, 2-sided Wilcoxon test using GNU R software (15). A value of P<.05 was considered statistically significant. Dose-response and volume-response curves were adjusted using the Statistica (16) logistic response model, together with a maximum-likelihood fitting procedure.
Results Dosimetric analysis Figure 1 shows a representative comparison of RBE-weighted dose distributions of both models for the temporal lobe. It can be seen that the dose gradient around the planning target volume (PTV) is stronger for LEM IV than for LEM I. Generally, a large area of the temporal lobe receives less dose in the case of LEM IV. However, small areas located directly adjacent to or overlapping the PTV are subject to higher doses. Figure 2 and Table 1 summarize the dosimetric variables for all patients: The median of Dmax increases by 22% from 74.6 to 95.6 Gy (RBE) for LEM IV as compared to LEM I, and Dmax,V-1 cm3 increases by 8% from 64.8 to 70.2 Gy (RBE). Dmean decreases by 23% from 14.7 to 11.3 Gy (RBE). The volume of the temporal lobe that receives a minimum dose of 50 Gy (RBE) is larger in the case of LEM I. V,D > 60 Gy (RBE) yields similar results for both models. For dose thresholds above 60 Gy (RBE), volume variables are collectively larger in the case of LEM IV.
Fig. 1. Comparison of (a) recalculated (local effect model [LEM IV]ebased) and (b) original (LEM Iebased) relative biological effectiveness (RBE)eweighted dose distributions, together
with a difference map (c). Contoured are the planning target volume (PTV) in red and the right temporal lobe in yellow. (d) Dose-volume histogram for the right temporal lobe. A color version of this figure is available at www.redjournal.org.
Volume 88 Number 5 2014
Clinical characterization of LEM I/IV 1139
Fig. 2. Comparison of dose-volume histogram (DVH) variables of the temporal lobe for local effect models (LEM) IV and I for all temporal lobes. Solid line indicates median; box, first and third quartiles; whiskers, 5% and 95% quantiles; circles, outliers; stars, significant difference between LEM I and LEM IV.
Dose-response analysis Figure 3 shows the resulting dose-response and volume-response curves for radiation-induced TLR based on the variables Dmax,V-1 cm3 and V,D > 85 Gy (RBE). Both curves are shifted toward higher values for LEM IV as compared to LEM I. TD5 increases from 68.8 3.3 to 78.3 4.3 Gy (RBE) from LEM I to LEM IV. Similarly, TD50 increases from 87.3 2.8 to 99.8 3.3 Gy (RBE). Table 2 summarizes the RBE-weighted tolerance doses for different effect levels. Tolerance volumes increase as well for LEM IV as compared to LEM I. The tolerance volume V,D > 85 Gy (RBE) determined at the 50% effect level increases from 2.0 0.4 cm3 to 2.8 0.4 cm3 when going from LEM I to LEM IV.
Discussion For carbon ion radiation therapy, the prediction of the biological effective dose depends on the RBE model involved in
Table 1 Comparison of median values of different dosevolume histogram variables over all temporal lobes calculated by local effect models (LEM) I and IV Median (25%/75% quantiles) Dmax [Gy (RBE)] Dmax,V-1 cm3 [Gy (RBE)] Dmean [Gy (RBE)] V,D > 50 Gy (RBE) [cm3] V,D > 60 Gy (RBE) [cm3] V,D > 70 Gy (RBE) [cm3] V,D > 80 Gy (RBE) [cm3] V,D > 90 Gy (RBE) [cm3]
LEM I
LEM IV
74.6 (71.1/78) 64.8 (59.0/73.4)
95.6 (86.9/103.5) 70.2 (62.2/83.1)
14.3 (11.2/17.2) 4.54 (2.7/8.6)
11.5 (8.3/13.9) 3.47 (2.4/5.7)
1.98 (0.8/1.9)
2.16 (1.3/3.9)
0.22 (0.0/1.9)
1.02 (0.4/2.7)
0 (0.0/0.0)
0.27 (0.0/1.4)
0 (0.0/0.0)
0.02 (0.0/0.3)
Abbreviation: RBE Z relative biological effectiveness.
treatment planning (11). The aim of the present comparative treatment planning study, therefore, was to analyze the clinical impact of a new radiobiological model (LEM IV) by comparing it against a standard model used routinely in clinical practice (LEM I) with respect to predictions of tolerances to radiationinduced TLR.
Systematic differences between LEM I and LEM IV Local maximum doses to the temporal lobe were found to be significantly increased by 20% to 30% for the recalculated LEM IVebased as compared to the original LEM Iebased dose distributions. These differences are a consequence of the stronger increase of RBE with LET predicted by LEM IV. In treatment plans, this is shown by a stronger dose gradient around the target area and a generally observed wider statistical spread of dosimetric variables for LEM IV as compared to LEM I. The reduction of the mean dose to the temporal lobe by 23% implies a correction of the systematic overestimation of RBE in the entrance channel shown by LEM I for experimental in vivo data (6). The underlying reason for the increased temporal lobe doses is related to the fact that LEM IV considers the local density of the microscopic spatial distribution of DNA double-strand breaks in the nucleus as the most relevant measure for the radiation response of cells (9). The local density of double-strand breaks is then related to the local dose. For the definition of an isoeffect required for the calculation of the RBE, it is assumed that similar double-strand break densities lead to similar effects, independent of radiation quality (9). It must be noted that LEM IV does not only include additional model features, but also uses a different set of input parameters (AB2 IV rather than AB2 I). However, as the ratio a/b has the highest impact on RBE and was set to 2 Gy for both, AB2 I (LEM I) and AB2 IV (LEM IV), the higher temporal lobe doses are expected to result from differences in the model versions rather than from differences in the input parameters. A detailed analysis of this issue was beyond the scope of this study. In our analysis, predictions of maximum doses by LEM IV were very sensitive to the exact delineation of the temporal lobe, especially if the temporal lobe overlapped the PTV. On the other hand, if there is a small gap between PTV and temporal lobe, maximum doses to the temporal lobe are lower for LEM IV as compared with LEM I. This systematic uncertainty has to be
International Journal of Radiation Oncology Biology Physics
1140 Gillmann et al.
Fig. 3. Dose-response (left) and volume-response (right) curves for radiation-induced temporal lobe reactions based on the variables Dmax,V-1 cm3 and V,D > 85 Gy (relative biological effectiveness [RBE]). For better visualization, circles indicate empirical incidence rates x/ n (reactions exhibited by x of n temporal lobes) in 10-Gy (RBE) dose-response curve, respectively 0.5-cm3 intervals volume-response curve. considered as a potential bias when comparing the doses of LEM I and LEM IV. As a consequence, differences between LEM I and LEM IV depend strongly on the selected dosimetric variable. Although TD5 and TD50 determined with LEM IV are considerably higher for the variables Dmax, Dmax,V-1 cm3, and Dmax,V-2 cm3 than those determined with LEM I, they are lower if the variables Dmax,V-5 cm3 and Dmax,V-10 cm3 are used for evaluation (Table 3). This raises the question as to which dosimetric variable correlates best with the clinically observed temporal lobe reactions. Regarding the local correlation of hot spots in the dose distribution with the appearance of TLR, one would expect that the maximum dose is suited best, eventually excluding a small volume (eg, 1 cm3) to be independent on very small hot spots in the dose distribution, where the dose calculation algorithm might not be reliable. In carbon ion therapy, the volumes of such hot spots are typically comparable to the voxel size and are thus significantly smaller than 1 cm3. Therefore our analysis focused on this dosimetric variable.
Comparison of LEM-based tolerance doses with literature data In our dose-response analysis, we showed that the biological effective TD5 and TD50 were increased by 9.5 and 12.5 Gy (RBE), respectively, when using LEM IV instead of LEM I. This difference is clinically significant and has to be considered as an uncertainty as long as it is not known which of the 2 LEM versions describes the RBE of normal brain tissue more accurately. To Table 2 Comparison of relative biological effectiveness (RBE)eweighted tolerance doses calculated by local effect models (LEM) I and IV Dmax,V-1 cm3 SE [Gy (RBE)] LEM I TD5 TD10 TD30 TD50 TD80 TD90 TD95
68.8 73.5 82.0 87.3 96.0 101.4 105.8
3.3 2.8 2.4 2.8 4.1 5.0 5.9
LEM IV 78.3 83.8 93.6 99.8 106.0 115.8 121.3
4.3 3.5 2.9 3.3 4.1 5.8 6.8
reduce this uncertainty, the tolerance doses of both models have to be compared with the clinically established values for photon irradiations using the same dosimetric variables in a comparable patient collective. In this respect, the value of TD5 is most important. Unfortunately, only very little information on TD5 is available in the literature. Su et al (17) reported a 5-year incidence of temporal lobe injury to be <5%, if a maximum dose of 64 to 68 Gy was applied to the temporal lobe. However, a large percentage of patients received additional chemotherapy, which can be expected to reduce radiation tolerance (18). Based on a single patient and using a normal tissue complication probability model, Tatsuzaki et al (19) published DVHs for the temporal lobe, from which TD5 values of 70 Gy and 72 Gy (RBE) for photons and protons, respectively, could be estimated (1). A recent study by Pehlivan et al (20) on temporal lobe toxicity after proton irradiation found that the severity of toxicity increased with increasing mean dose to a 2-cm3 volume of the temporal lobe, but no relation to complication probability or tolerance doses was given. Apart from these single studies, attempts have been made to quantify normal brain tolerance doses by retrospectively reviewing clinical data. This first resulted in the well-known “Emamidata” on normal tissue tolerance (21). Within the framework of the Quantitative Analyses of Normal Tissue Effects in the Clinic (QUANTEC) analysis (22), these data were recently refined, and the best estimate of TD5 for brain using 2-Gy fractions was reported to be 72 Gy (60-84 Gy, 95% confidence interval). For the comparison with the QUANTEC data, it has to be considered that the conversion of the total dose into 2-Gy fractions was based on a/b Z 3 Gy, although we used a/b Z 2 Gy to be consistent with the a/b-value in LEM I and LEM IV. If we had converted the fractional dose with a/b Z 3 Gy instead, our total doses would have decreased on average by 4.2% (range, 4.0%5.5% for 3.0, 3.3, and 3.5 Gy [RBE] per fraction), and the TD5values would then have been reduced to 65.9 and 75.0 Gy (RBE) for LEM I and LEM IV, respectively. Although the QUANTEC value is somewhat closer to the LEM IV value, a comparison remains difficult for 3 reasons: The QUANTEC value of 72 Gy is based (with 1 exception) on studies that specify only the prescribed rather than the local brain dose. Although this may be a reasonable approximation for the maximum brain dose in the case of brain tumors, it probably overestimates the temporal lobe dose for nasopharyngeal cancer, which constitutes about two-thirds of the patient cases in the underlying studies. With respect to a comparison to
Volume 88 Number 5 2014
Clinical characterization of LEM I/IV 1141
Table 3 Comparison of TD5 and TD50 calculated by local effect models (LEM) I and IV in dependence on the dosimetric variable TD5 [Gy (RBE)] Dmax Dmax,V-1 cm3 Dmax,V-2 cm3 Dmax,V-5 cm3 Dmax,V-10 cm3
TD50 [Gy (RBE)]
LEM I
LEM IV
LEM I
73.3 68.8 64.4 51.5 34.0
89.6 78.3 69.6 44.3 18.8
99.1 87.3 81.7 69.7 60.8
LEM IV 134.9 99.8 91.1 70.7 51.5
2.
3.
4.
5.
Abbreviation: RBE Z relative biological effectiveness. 6.
our study, the actual brain doses and hence the TD5-value are therefore likely to be overestimated, especially if Dmax,V-1 cm3 instead of the maximum dose is used as dosimetric variable. In our study, large variations in Dmax,V-1 cm3 were seen for the same prescribed dose, so a temporal lobespecific dose variable was selected for our analysis. Although the QUANTEC analysis investigated the risk of late effects per patient, we determined the risk for individual temporal lobes based on the local dose. This, together with other differences in endpoint definition and follow-up time, may introduce bias into the comparison. Finally, the 95% confidence interval of the QUANTEC TD5 is very large, most likely because of uncertainties in the actual brain doses and the pooling of several studies.
7.
8.
9.
10.
In view of these methodological issues, it is currently not possible to decide whether RBE-predictions of LEM I or of LEM IV fit better with clinical experience in photon therapy. To overcome these problems, the TD5 value for photons (or protons, assuming a fixed RBE of 1.1) has to be determined in a comparable patient collective using exactly the same dosimetric variable as in the present study.
11.
Conclusions
14.
Although LEM IV seems to give a better description of experimental in vivo data as compared with LEM I (11), a decision based on clinical data remains more difficult. The derived doseresponse curves indicate a clinically significant increase of the RBE-weighted TD5 by 9.5 Gy (RBE) for LEM IV as compared to the currently clinically applied LEM I. As the comparison of the derived tolerance doses with available photon data is associated with several methodological uncertainties, it is currently not possible to decide definitively whether RBE predictions of LEM I or of LEM IV fit better with clinical experience in photon therapy. The decision as to whether LEM IV may be clinically applied at HIT therefore requires further investigation. Such investigation includes, in particular, the analysis of temporal lobe reactions in a comparable photon-treated patient collective using the same dosimetric variable as in the present study.
15.
References 1. Schlampp I, Karger CP, Ja¨kel O, et al. Temporal lobe reactions after radiotherapy with carbon ions: Incidence and estimation of the relative
12. 13.
16.
17.
18.
19.
20.
21.
22.
biological effectiveness by the local effect model. Int J Radiat Oncol Biol Phys 2011;80:815-823. Particle Therapy Co-Operative Group (PTCOG). Available at: http:// ptcog.web.psi.ch/ptcentres.html and http://ptcog.web.psi.ch/ newptcentres.html. Most recent date of access: 2013/08/08.. Kanai T, Furusawa Y, Fukutsu K, et al. Irradiation of mixed beam and design of spread-out Bragg peak for heavy-ion radiotherapy. Radiat Res 1997;147:78-85. Scholz M, Kellerer AM, Kraft-Weyrather W, et al. Computation of cell survival in heavy ion beams for therapy. Radiat Environ Biophys 1997; 36:59-66. Schulz-Ertner D, Karger CP, Feuerhake A, et al. Effectiveness of carbon ion radiotherapy in the treatment of skull-base chordomas. Int J Radiat Oncol Biol Phys 2007;68:449-457. Karger CP, Peschke P, Sanchez-Brandelik R, et al. Radiation tolerance of the rat spinal cord after 6 and 18 fractions of photons and carbon ions: Experimental results and clinical implications. Int J Radiat Oncol Biol Phys 2006;66:1488-1497. Elsa¨sser T, Kra¨mer M, Scholz M. Accuracy of the local effect model for the prediction of biologic effects of carbon ion beams in vitro and in vivo. Int J Radiat Oncol Biol Phys 2008;71:866-872. Wilkens JJ, Oelfke U. Direct comparison of biologically optimized spread-out Bragg peaks for protons and carbon ions. Int J Radiat Oncol Biol Phys 2008;70:262-266. Elsa¨sser T, Weyrather WK, Friedrich T, et al. Quantification of the relative biological effectiveness for ion beam radiotherapy: Direct experimental comparison of proton and carbon ion beams and a novel approach for treatment planning. Int J Radiat Oncol Biol Phys 2010; 78:1177-1183. Friedrich T, Scholz U, Elsa¨sser T, et al. Systematic analysis of RBE and related quantities using a database of cell survival experiments with ion beam irradiation. J Radiat Res (Tokyo) 2012;54:494-514. Gru¨n R, Friedrich T, Elsa¨sser T, et al. Impact of enhancements in the local effect model (LEM) on the predicted RBE-weighted target dose distribution in carbon ion therapy. Phys Med Biol 2012;57:72617274. Fowler JF. The first James Kirk Memorial Lecture. What next in fractionated radiotherapy? Br J Cancer Suppl 1984;6:285-300. Kra¨mer M, Ja¨kel O, Haberer T, et al. Treatment planning for heavy-ion radiotherapy: Physical beam model and dose optimization. Phys Med Biol 2000;45:3299-3317. Kra¨mer M, Scholz M. Treatment planning for heavy-ion radiotherapy: Calculation and optimization of biologically effective dose. Phys Med Biol 2000;45:3319-3330. R Project for Statistical Computing. Available at: http://www.r-project. org. Version: 2.7. Most recent date of access: 2013/08/08. StatSoft I. STATISTICA fu¨r Windows (Software System fu¨r Datenanalyse) Version 7.1; 2011. Available at: www.statsoft.com. Most recent date of access: 2013/08/08. Su SF, Huang Y, Xiao W, et al. Clinical and dosimetric characteristics of temporal lobe injury following intensity modulated radiotherapy of nasopharyngeal carcinoma. Radiother Oncol 2012;104:312-316. Ruben JD, Dally M, Bailey M, et al. Cerebral radiation necrosis: Incidence, outcomes, and risk factors with emphasis on radiation parameters and chemotherapy. Int J Radiat Oncol Biol Phys 2006;65: 499-508. Tatsuzaki H, Urie MM. Importance of precise positioning for proton beam therapy in the base of skull and cervical spine. Int J Radiat Oncol Biol Phys 1991;21:757-765. Pehlivan B, Ares C, Lomax AJ, et al. Temporal lobe toxicity analysis after proton radiation therapy for skull base tumors. Int J Radiat Oncol Biol Phys 2012;83:1432-1440. Emami B, Lyman J, Brown A, et al. Tolerance of normal tissue to therapeutic irradiation. Int J Radiat Oncol Biol Phys 1991;21: 109-122. Lawrence YR, Li XA, el Naqa I, et al. Radiation dose-volume effects in the brain. Int J Radiat Oncol Biol Phys 2010;76:S20-S27.