- Email: [email protected]
β ratio for prostate cancer: What is it, really?
Radiotherapy and Oncology 76 (2005) 1–3 www.thegreenjournal.com
Special commentary
The a/b ratio for prostate cancer: What is it, really? Søren M. Bentzen*, Mark A. Ritter Department of Human Oncology, University of Wisconsin Medical School, WI, USA
Abstract Two recent studies of fractionated external beam radiotherapy in early prostate cancer provide outcome data that allow a statistical estimation of the a/b-ratio of the linear–quadratic model when combined with clinical data on the steepness of the dose–response curve. Results of the large randomized PR5 trial by the Ontario Clinical Oncology Group/National Cancer Institute of Canada yield an estimate of a/b at 1.12 Gy with 95% confidence interval (K3.3, 5.6) Gy. A non-randomized study by Valdagni and colleagues of hyper-fractionation delivered BID versus conventional fractionation yields an a/b-estimate of 8.3 Gy with 95% confidence interval (0.7, 16) Gy. Thus, the confidence interval of this latter study cannot exclude even very low values of a/b. Furthermore, this point estimate may be an over-estimate if incomplete repair plays a role in the BID group of the Italian study. Taken together, the outcomes of these two studies still favor a high fractionation sensitivity of prostate cancer. q 2005 Published by Elsevier Ireland Ltd. Radiotherapy and Oncology 76 (2005) 1–3.
“The researches of many commentators have already thrown much darkness on this subject, and it is probable that, if they continue, we shall soon know nothing at all about it.”—Mark Twain Radiotherapy and Oncology has flown the colors of evidence-based medicine more than most other journals [1,2,11,13,14,16]. Therefore, it is slightly ironic that in the 75th Anniversary Volume of the Journal, a non-randomized, parallel-group study of fractionated radiotherapy in prostate cancer gets selected for prime time. Admittedly, the conclusion of the study by Valdagni and colleagues [17] is a gem: with the current flurry of interest in hypo-fractionated radiotherapy for prostate cancer it is thought-provoking that a quite large, apparently well-conducted albeit nonrandomized, study shows that hyper-fractionated (HFX) radiotherapy—introduced to spare late side-effects—seems to break even with conventional fractionation with respect to PSA-relapse-free rate. A simple calculation shows that the HFX schedule tried should do much worse than the conventional schedule in terms of tumor control if the a/bratio is as low as the 1.5 Gy proposed by several recent analyses [6,8]. Thus, we can understand the Editor of the Journal: this is the kind of ‘man bites dog’ story that would make it to the front page of most journals. But does it really pull the rug under current trials of large dose per fraction in prostate cancer? The Achilles heel of virtually all published estimates of the fractionation sensitivity for prostate cancer, as quantified by the a/b-ratio of the linear–quadratic bioeffect model, is that they are derived from comparisons between the outcomes after brachytherapy and external beam
radiotherapy (EBRT) for prostate cancer. Valdagni’s study is of immediate interest here, because it compares two EBRT fractionation schedules delivered as conformal radiotherapy. Valdagni reports the outcome after a conventional schedule delivering 70–76 Gy (median 74 Gy) with 2 Gy per fraction, one fraction per day in one group of 179 patients and compares this with the outcome after an HFX schedule delivering two daily fractions of 1.2 Gy with a minimum inter-fraction interval of 6 h to a dose of 70–82.8 Gy (median 79.2 Gy) in a concurrent control group of 151 patients. At the time of its activation, this prospective study was in line with the then current radiobiological thinking [9]. Treatment allocation was not randomized but was made according to patient preference. The main tumor endpoint was biochemical relapse-free survival and it was found that there was no statistically significant difference between the two groups with respect to this endpoint. No discussion: this is an interesting observation. Yet, the study by Valdagni et al. has some limitations. The inclusion of low, intermediate and high-risk patients, the use of whole pelvic irradiation in some 40% of cases and the use of a range of radiation doses in both study groups are all factors that tend to increase study heterogeneity and thereby reduce the ability of the study to detect a potential difference between treatment groups. As also pointed out by the authors the median follow-up of 29 months is relatively short, again reducing event rates and thereby lowering statistical power. Finally, it should be noted that adjuvant androgen deprivation was used in nearly half of all cases for a period of 6–24 months after radiotherapy, which again will interfere with the assessment of PSA relapse. All of these circumstances pull in the same direction, making a ‘no significant
0167-8140/$ - see front matter q 2005 Published by Elsevier Ireland Ltd. doi:10.1016/j.radonc.2005.06.009
2
The a/b ratio for prostate cancer
difference’ conclusion more likely. Admittedly, the report shows no major bias in the distribution of baseline characteristics in the two groups and the outcome of the comparison remains intriguing. If we take Valdagni’s study at face value, what would it tell us about the likely value of a/b for prostate cancer? Using a method developed by us and described in detail elsewhere (Bentzen et al., submitted) it is possible to estimate the a/b-ratio and its 95% confidence limits for two non-isoeffective regimens provided that the steepness of the dose-response curve is known. Intuitively, this can be seen as a method where the steepness of the dose–response curve is used together with the observed outcomes to estimate the dose in the conventional group that would have been exactly isoeffective with the dose in the other group and then derive the a/b-ratio based on these two data points. In reality, this can be done using a single closed-form expression. The statistical precision of the study outcome is combined with the statistical precision of the steepness estimate into an overall estimate of the uncertainty of the estimated a/bratio. Probably, the best current estimate of the steepness of the dose–response curve for prostate cancer was recently derived from a study of 235 low-risk and 382 intermediaterisk prostate cancer patients treated between 1987 and 1998 with external beam RT alone at the MD Anderson Cancer Center [7]. In the intermediate risk patients, defining biochemical failure as a PSA rise of R2 ng/mL above the current nadir PSA, g50 was estimated at 1.4 with 95% confidence limits 0.2 and 2.5. In this group of patients, the g50-value using the ASTRO definition of biochemical failure was 2.2, but we assume that the slightly lower steepness is more relevant for the data sets analyzed below as these are more heterogeneous than the MD Anderson data set. A technical remark is that the method uses the local value of g at the relevant level of tumor control, but it can be shown that the steepness of the dose–response curve is completely specified at any response level when g50 is known [5]. Applying the method described above to Valdagni’s data allows a formal estimate of a/b. The difference in event rates between the two groups is quantified by the hazard ratio, HR, for failure after conventional relative to HFX. The estimate is HRZ0.932 with 95% confidence limits 0.49 and 1.77. Note that this point estimate—despite being not statistically significantly different from 1—actually favors the standard schedule, a point worth noting in passing as Valdagni et al. state that their clinical result to date favors the HFX schedule. Thus, the dose in the standard fractionation group that would have produced a level of tumor control equivalent with that of the HFX group is estimated at 73.0 Gy, as compared with the median dose actually delivered of 74 Gy. From this estimate of HR, we obtain a point estimate of a/b from Valdagni’s data at 8.3 Gy. This is indeed on the high side of recent estimates. However, the estimated 95% confidence limits are 0.7 and 16 Gy. In other words, Valdagni’s study does not have sufficient statistical power to rule out even very low a/b-values. In a way, this is all there is to say. There is, however, one further interesting radiobiological speculation. The outcome after a BID schedule in prostate cancer could conceivably be affected by incomplete repair.
Valdagni et al. state that they prescribed a minimum interfraction interval of 6 h, which for logistical reasons probably means that it was close to 6 h in the majority of cases. After the design of Valdagni’s BID schedule, it has become clear that repair half-times for some late normal tissue endpoints are likely to be much longer than originally thought, with estimates ranging between 4 and 5 h [4]. Inspired by the analogy between the radiobiological properties of prostate cancer and late-responding normal tissues, could it be that the repair halftime, T1/2, for prostate cancer is also, say, 4 h? Applying the standard incomplete repair (IR) model with a/bZ1.5 Gy, T1/2Z4 h and an inter-fraction interval of 6 h, the median physical dose to the prostate of 79.2 Gy delivered with 1.2 Gy/F BID would be equivalent to a dose in 2-Gy fractions of EQD2Z70.7 Gy, considerably higher than the 61 Gy estimated under the assumption of complete repair between fractions. Using Guttenberger’s formulation of the IR model [4,10], that corrects for the effect of incomplete repair in the overnight interval between the afternoon fraction of one day and the fraction delivered the following morning (but assuming repair to be complete after a weekend), brings EQD2 of the HFX schedule to 72.4 Gy, very close to the 73.0 Gy estimated to produce the same tumor control in the conventional group as in the HFX group. To summarize, even if we assume complete repair between dose fractions for prostate cancer, Valdagni’s study cannot exclude that a/b is as low as 0.5–1.5 Gy as shown by the lower bound of the 95% confidence interval. Furthermore, if T1/2 for prostate cancer is as long as for some late normaltissue endpoints, Valdagni’s observations are just what you would expect if a/b is in the range of 1–2 Gy. A somewhat overlooked early report on the outcome of a large, and potentially very informative, randomized trial of hypo-fractionated radiotherapy in prostate cancer was presented by Lukka et al. [12] at the 2003 meeting of the Canadian Association for Radiation Oncology (CARO). The Ontario Clinical Oncology Group/National Cancer Institute of Canada (OCOG/NCIC) PR5 trial, compared 66 Gy in 33 F over 6.5 weeks versus 52.5 Gy in 20 F over 4 weeks. A total of 936 patients with T1 or T2 prostate cancer were randomly allocated to the two trial arms with a 1:1 allocation ratio. The primary endpoint was defined as biochemical failure according to ASTRO criteria, clinical local or distant failure, hormonal intervention or prostate cancer death. Median follow-up was 59 months and there had been a total of 460 events at the time of the analysis. The 5-year failure probability was estimated at 48.6% after conventional fractionation and 55.6% after hypo-fractionation. Again, using the data in the abstract for the hazard ratio and its 95% confidence interval, combined with the steepness of the dose–response curve for intermediate-risk prostate cancer from Cheung et al. [7], we can estimate the fractionation sensitivity for prostate cancer at a/bZ1.12 Gy with 95% confidence limits K3.3 and 5.6 Gy. This calculation assumes that there is no significant effect of increasing overall time from 4 to 6.5 weeks on the control of prostate cancer, an assumption that is quite well supported by data from large non-randomized clinical studies [3]. The confidence intervals of the above a/b-ratios illustrate the statistical power required to pin-down this value more precisely. The Valdagni estimate without
S.M. Bentzen, M.A. Ritter / Radiotherapy and Oncology 76 (2005) 1–3
the incomplete repair assumption and the OCOG/NCIC estimates are actually not statistically significantly different (2-tailed PZ0.11). Due to the slightly higher precision of the estimate from the Canadian trial, a best joint estimate would be closer to the value from this trial, namely around 3.0 Gy. However, the substantial biological uncertainties involved in the BID group in Valdagni’s study, discourages a more formal pooled estimate of a/b from the two studies. So, what do we know about the fractionation sensitivity of prostate cancer as estimated from external-beam only studies? At the time of writing the most compelling evidence for a high sensitivity to dose per fraction comes from the large randomized OCOG/NCIC trial. This estimate is a/bZ1.12 Gy with 95% confidence limits K3.3 and 5.6 Gy. This does not conclusively settle the matter though, partly because of the rather wide confidence interval and partly because we need to assume a zero time-factor for prostate cancer to arrive at this estimate. While data in the literature strongly suggest a low time factor for prostate cancer, it cannot statistically be ruled out that it is greater than zero. Thus, we do need more information and this is likely to come from ongoing clinical trials. One of these is the multi-institutional trial by Ritter and colleagues [15] testing dose-per-fraction escalation in a phase I/II trial setting in a planned 300 patients. Due to the careful consideration of early and late toxicity before each dose-per-fraction escalation step, this is a non-randomized study. However, as radiation is delivered with 24 h between fractions and in an overall time that is not directly dependent on dose per fraction some of the biological parameters are better controlled in this case. This trial together with other hypo-fractionation trials in progress all over the World will undoubtedly improve our knowledge of prostate cancer radiobiology. Valdagni and his colleagues are to be congratulated for their timely publication of this interesting clinical study. However, rather than casting doubt on the biological rationale for ongoing hypo-fractionation trials, the Italian study provides a perfect demonstration of why these clinical trials are needed.
*
Corresponding author. Søren M. Bentzen, Department of Human Oncology, University of Wisconsin Medical School, K4/316 Clinical Sciences Center, 600 Highland Avenue, Madison, WI 53792, USA. E-mail address: [email protected] Received 27 May 2005; accepted 2 June 2005; available online 28 June 2005
References [1] Bayley AJ, Catton CN, Haycocks T, et al. A randomized trial of supine vs. prone positioning in patients undergoing escalated dose conformal radiotherapy for prostate cancer. Radiother Oncol 2004;70:37–44.
3
[2] Bentzen SM. Towards evidence based radiation oncology: Improving the design, analysis, and reporting of clinical outcome studies in radiotherapy. Radiother Oncol 1998;46: 5–18. [3] Bentzen SM. Time–dose relationships for human tumors: estimation from non-randomized studiesBeck-Bornholt HP. Current topics in clinical radiobiology of tumours. Medical radiology. Berlin: Springer; 1993. 11–26. [4] Bentzen SM, Saunders MI, Dische S. Repair halftimes estimated from observations of treatment-related morbidity after CHART or conventional radiotherapy in head and neck cancer. Radiother Oncol 1999;53:219–26. [5] Bentzen SM, Tucker SL. Quantifying the position and steepness of radiation dose-response curves. Int J Radiat Biol 1997;71: 531–42 [see comments]. [6] Brenner DJ, Hall EJ. Fractionation and protraction for radiotherapy of prostate carcinoma. Int J Radiat Oncol Biol Phys 1999;43:1095–101. [7] Cheung R, Tucker SL, Lee AK, et al. Dose–response characteristics of low- and intermediate-risk prostate cancer treated with external beam radiotherapy. Int J Radiat Oncol Biol Phys 2005;61:993–1002. [8] Fowler J, Chappell R, Ritter M. Is alpha/beta for prostate tumors really low? Int J Radiat Oncol Biol Phys 2001;50: 1021–31. [9] Fowler JF, Ritter MA. A rationale for fractionation for slowly proliferating tumors such as prostatic adenocarcinoma. Int J Radiat Oncol Biol Phys 1995;32:521–9. [10] Guttenberger R, Thames HD, Ang KK. Is the experience with CHART compatible with experimental data? A new model of repair kinetics and computer simulations Radiother Oncol 1992; 25:280–6. [11] Koper PC, Jansen P, van PW, et al. Gastro-intestinal and genitourinary morbidity after 3D conformal radiotherapy of prostate cancer: observations of a randomized trial. Radiother Oncol 2004;73:1–9. [12] Lukka H, Hayter C, Warde P, et al. A randomized trial comparing two fractionation schedules for patients with localized prostate cancer. Radiother Oncol 2003;69:S7. [13] Macbeth F, Overgaard J. Expert reviews, systematic reviews and meta-analyses. Radiother Oncol 2002;64:233–4. [14] Overgaard J, Bentzen SM. Evidence based radiation oncology. Radiother Oncol 1998;46:1–3 [editorial; comment]. [15] Ritter MA, Chappell R, Fowler JF, Tome WA, Kupelian PA. A multi-institutional phase I/II trial of dose-per-fraction escalation for localized prostate cancer. ASCO Prostate Symposium, Orlando, FL 2005. Abstract 91. [16] Sydes MR, Stephens RJ, Moore AR, et al. Implementing the UK Medical Research Council (MRC) RT01 trial (ISRCTN 47772397): methods and practicalities of a randomised controlled trial of conformal radiotherapy in men with localised prostate cancer. Radiother Oncol 2004;72:199–211. [17] Valdagni R, Italia C, Montanaro P, et al. Is the alpha-beta ratio of prostate cancer really low? A prospective, non-randomized trial comparing standard and hyperfractionated conformal radiation therapy Radiother Oncol 2005;75:74–82.
Special commentary
The a/b ratio for prostate cancer: What is it, really? Søren M. Bentzen*, Mark A. Ritter Department of Human Oncology, University of Wisconsin Medical School, WI, USA
Abstract Two recent studies of fractionated external beam radiotherapy in early prostate cancer provide outcome data that allow a statistical estimation of the a/b-ratio of the linear–quadratic model when combined with clinical data on the steepness of the dose–response curve. Results of the large randomized PR5 trial by the Ontario Clinical Oncology Group/National Cancer Institute of Canada yield an estimate of a/b at 1.12 Gy with 95% confidence interval (K3.3, 5.6) Gy. A non-randomized study by Valdagni and colleagues of hyper-fractionation delivered BID versus conventional fractionation yields an a/b-estimate of 8.3 Gy with 95% confidence interval (0.7, 16) Gy. Thus, the confidence interval of this latter study cannot exclude even very low values of a/b. Furthermore, this point estimate may be an over-estimate if incomplete repair plays a role in the BID group of the Italian study. Taken together, the outcomes of these two studies still favor a high fractionation sensitivity of prostate cancer. q 2005 Published by Elsevier Ireland Ltd. Radiotherapy and Oncology 76 (2005) 1–3.
“The researches of many commentators have already thrown much darkness on this subject, and it is probable that, if they continue, we shall soon know nothing at all about it.”—Mark Twain Radiotherapy and Oncology has flown the colors of evidence-based medicine more than most other journals [1,2,11,13,14,16]. Therefore, it is slightly ironic that in the 75th Anniversary Volume of the Journal, a non-randomized, parallel-group study of fractionated radiotherapy in prostate cancer gets selected for prime time. Admittedly, the conclusion of the study by Valdagni and colleagues [17] is a gem: with the current flurry of interest in hypo-fractionated radiotherapy for prostate cancer it is thought-provoking that a quite large, apparently well-conducted albeit nonrandomized, study shows that hyper-fractionated (HFX) radiotherapy—introduced to spare late side-effects—seems to break even with conventional fractionation with respect to PSA-relapse-free rate. A simple calculation shows that the HFX schedule tried should do much worse than the conventional schedule in terms of tumor control if the a/bratio is as low as the 1.5 Gy proposed by several recent analyses [6,8]. Thus, we can understand the Editor of the Journal: this is the kind of ‘man bites dog’ story that would make it to the front page of most journals. But does it really pull the rug under current trials of large dose per fraction in prostate cancer? The Achilles heel of virtually all published estimates of the fractionation sensitivity for prostate cancer, as quantified by the a/b-ratio of the linear–quadratic bioeffect model, is that they are derived from comparisons between the outcomes after brachytherapy and external beam
radiotherapy (EBRT) for prostate cancer. Valdagni’s study is of immediate interest here, because it compares two EBRT fractionation schedules delivered as conformal radiotherapy. Valdagni reports the outcome after a conventional schedule delivering 70–76 Gy (median 74 Gy) with 2 Gy per fraction, one fraction per day in one group of 179 patients and compares this with the outcome after an HFX schedule delivering two daily fractions of 1.2 Gy with a minimum inter-fraction interval of 6 h to a dose of 70–82.8 Gy (median 79.2 Gy) in a concurrent control group of 151 patients. At the time of its activation, this prospective study was in line with the then current radiobiological thinking [9]. Treatment allocation was not randomized but was made according to patient preference. The main tumor endpoint was biochemical relapse-free survival and it was found that there was no statistically significant difference between the two groups with respect to this endpoint. No discussion: this is an interesting observation. Yet, the study by Valdagni et al. has some limitations. The inclusion of low, intermediate and high-risk patients, the use of whole pelvic irradiation in some 40% of cases and the use of a range of radiation doses in both study groups are all factors that tend to increase study heterogeneity and thereby reduce the ability of the study to detect a potential difference between treatment groups. As also pointed out by the authors the median follow-up of 29 months is relatively short, again reducing event rates and thereby lowering statistical power. Finally, it should be noted that adjuvant androgen deprivation was used in nearly half of all cases for a period of 6–24 months after radiotherapy, which again will interfere with the assessment of PSA relapse. All of these circumstances pull in the same direction, making a ‘no significant
0167-8140/$ - see front matter q 2005 Published by Elsevier Ireland Ltd. doi:10.1016/j.radonc.2005.06.009
2
The a/b ratio for prostate cancer
difference’ conclusion more likely. Admittedly, the report shows no major bias in the distribution of baseline characteristics in the two groups and the outcome of the comparison remains intriguing. If we take Valdagni’s study at face value, what would it tell us about the likely value of a/b for prostate cancer? Using a method developed by us and described in detail elsewhere (Bentzen et al., submitted) it is possible to estimate the a/b-ratio and its 95% confidence limits for two non-isoeffective regimens provided that the steepness of the dose-response curve is known. Intuitively, this can be seen as a method where the steepness of the dose–response curve is used together with the observed outcomes to estimate the dose in the conventional group that would have been exactly isoeffective with the dose in the other group and then derive the a/b-ratio based on these two data points. In reality, this can be done using a single closed-form expression. The statistical precision of the study outcome is combined with the statistical precision of the steepness estimate into an overall estimate of the uncertainty of the estimated a/bratio. Probably, the best current estimate of the steepness of the dose–response curve for prostate cancer was recently derived from a study of 235 low-risk and 382 intermediaterisk prostate cancer patients treated between 1987 and 1998 with external beam RT alone at the MD Anderson Cancer Center [7]. In the intermediate risk patients, defining biochemical failure as a PSA rise of R2 ng/mL above the current nadir PSA, g50 was estimated at 1.4 with 95% confidence limits 0.2 and 2.5. In this group of patients, the g50-value using the ASTRO definition of biochemical failure was 2.2, but we assume that the slightly lower steepness is more relevant for the data sets analyzed below as these are more heterogeneous than the MD Anderson data set. A technical remark is that the method uses the local value of g at the relevant level of tumor control, but it can be shown that the steepness of the dose–response curve is completely specified at any response level when g50 is known [5]. Applying the method described above to Valdagni’s data allows a formal estimate of a/b. The difference in event rates between the two groups is quantified by the hazard ratio, HR, for failure after conventional relative to HFX. The estimate is HRZ0.932 with 95% confidence limits 0.49 and 1.77. Note that this point estimate—despite being not statistically significantly different from 1—actually favors the standard schedule, a point worth noting in passing as Valdagni et al. state that their clinical result to date favors the HFX schedule. Thus, the dose in the standard fractionation group that would have produced a level of tumor control equivalent with that of the HFX group is estimated at 73.0 Gy, as compared with the median dose actually delivered of 74 Gy. From this estimate of HR, we obtain a point estimate of a/b from Valdagni’s data at 8.3 Gy. This is indeed on the high side of recent estimates. However, the estimated 95% confidence limits are 0.7 and 16 Gy. In other words, Valdagni’s study does not have sufficient statistical power to rule out even very low a/b-values. In a way, this is all there is to say. There is, however, one further interesting radiobiological speculation. The outcome after a BID schedule in prostate cancer could conceivably be affected by incomplete repair.
Valdagni et al. state that they prescribed a minimum interfraction interval of 6 h, which for logistical reasons probably means that it was close to 6 h in the majority of cases. After the design of Valdagni’s BID schedule, it has become clear that repair half-times for some late normal tissue endpoints are likely to be much longer than originally thought, with estimates ranging between 4 and 5 h [4]. Inspired by the analogy between the radiobiological properties of prostate cancer and late-responding normal tissues, could it be that the repair halftime, T1/2, for prostate cancer is also, say, 4 h? Applying the standard incomplete repair (IR) model with a/bZ1.5 Gy, T1/2Z4 h and an inter-fraction interval of 6 h, the median physical dose to the prostate of 79.2 Gy delivered with 1.2 Gy/F BID would be equivalent to a dose in 2-Gy fractions of EQD2Z70.7 Gy, considerably higher than the 61 Gy estimated under the assumption of complete repair between fractions. Using Guttenberger’s formulation of the IR model [4,10], that corrects for the effect of incomplete repair in the overnight interval between the afternoon fraction of one day and the fraction delivered the following morning (but assuming repair to be complete after a weekend), brings EQD2 of the HFX schedule to 72.4 Gy, very close to the 73.0 Gy estimated to produce the same tumor control in the conventional group as in the HFX group. To summarize, even if we assume complete repair between dose fractions for prostate cancer, Valdagni’s study cannot exclude that a/b is as low as 0.5–1.5 Gy as shown by the lower bound of the 95% confidence interval. Furthermore, if T1/2 for prostate cancer is as long as for some late normaltissue endpoints, Valdagni’s observations are just what you would expect if a/b is in the range of 1–2 Gy. A somewhat overlooked early report on the outcome of a large, and potentially very informative, randomized trial of hypo-fractionated radiotherapy in prostate cancer was presented by Lukka et al. [12] at the 2003 meeting of the Canadian Association for Radiation Oncology (CARO). The Ontario Clinical Oncology Group/National Cancer Institute of Canada (OCOG/NCIC) PR5 trial, compared 66 Gy in 33 F over 6.5 weeks versus 52.5 Gy in 20 F over 4 weeks. A total of 936 patients with T1 or T2 prostate cancer were randomly allocated to the two trial arms with a 1:1 allocation ratio. The primary endpoint was defined as biochemical failure according to ASTRO criteria, clinical local or distant failure, hormonal intervention or prostate cancer death. Median follow-up was 59 months and there had been a total of 460 events at the time of the analysis. The 5-year failure probability was estimated at 48.6% after conventional fractionation and 55.6% after hypo-fractionation. Again, using the data in the abstract for the hazard ratio and its 95% confidence interval, combined with the steepness of the dose–response curve for intermediate-risk prostate cancer from Cheung et al. [7], we can estimate the fractionation sensitivity for prostate cancer at a/bZ1.12 Gy with 95% confidence limits K3.3 and 5.6 Gy. This calculation assumes that there is no significant effect of increasing overall time from 4 to 6.5 weeks on the control of prostate cancer, an assumption that is quite well supported by data from large non-randomized clinical studies [3]. The confidence intervals of the above a/b-ratios illustrate the statistical power required to pin-down this value more precisely. The Valdagni estimate without
S.M. Bentzen, M.A. Ritter / Radiotherapy and Oncology 76 (2005) 1–3
the incomplete repair assumption and the OCOG/NCIC estimates are actually not statistically significantly different (2-tailed PZ0.11). Due to the slightly higher precision of the estimate from the Canadian trial, a best joint estimate would be closer to the value from this trial, namely around 3.0 Gy. However, the substantial biological uncertainties involved in the BID group in Valdagni’s study, discourages a more formal pooled estimate of a/b from the two studies. So, what do we know about the fractionation sensitivity of prostate cancer as estimated from external-beam only studies? At the time of writing the most compelling evidence for a high sensitivity to dose per fraction comes from the large randomized OCOG/NCIC trial. This estimate is a/bZ1.12 Gy with 95% confidence limits K3.3 and 5.6 Gy. This does not conclusively settle the matter though, partly because of the rather wide confidence interval and partly because we need to assume a zero time-factor for prostate cancer to arrive at this estimate. While data in the literature strongly suggest a low time factor for prostate cancer, it cannot statistically be ruled out that it is greater than zero. Thus, we do need more information and this is likely to come from ongoing clinical trials. One of these is the multi-institutional trial by Ritter and colleagues [15] testing dose-per-fraction escalation in a phase I/II trial setting in a planned 300 patients. Due to the careful consideration of early and late toxicity before each dose-per-fraction escalation step, this is a non-randomized study. However, as radiation is delivered with 24 h between fractions and in an overall time that is not directly dependent on dose per fraction some of the biological parameters are better controlled in this case. This trial together with other hypo-fractionation trials in progress all over the World will undoubtedly improve our knowledge of prostate cancer radiobiology. Valdagni and his colleagues are to be congratulated for their timely publication of this interesting clinical study. However, rather than casting doubt on the biological rationale for ongoing hypo-fractionation trials, the Italian study provides a perfect demonstration of why these clinical trials are needed.
*
Corresponding author. Søren M. Bentzen, Department of Human Oncology, University of Wisconsin Medical School, K4/316 Clinical Sciences Center, 600 Highland Avenue, Madison, WI 53792, USA. E-mail address: [email protected] Received 27 May 2005; accepted 2 June 2005; available online 28 June 2005
References [1] Bayley AJ, Catton CN, Haycocks T, et al. A randomized trial of supine vs. prone positioning in patients undergoing escalated dose conformal radiotherapy for prostate cancer. Radiother Oncol 2004;70:37–44.
3
[2] Bentzen SM. Towards evidence based radiation oncology: Improving the design, analysis, and reporting of clinical outcome studies in radiotherapy. Radiother Oncol 1998;46: 5–18. [3] Bentzen SM. Time–dose relationships for human tumors: estimation from non-randomized studiesBeck-Bornholt HP. Current topics in clinical radiobiology of tumours. Medical radiology. Berlin: Springer; 1993. 11–26. [4] Bentzen SM, Saunders MI, Dische S. Repair halftimes estimated from observations of treatment-related morbidity after CHART or conventional radiotherapy in head and neck cancer. Radiother Oncol 1999;53:219–26. [5] Bentzen SM, Tucker SL. Quantifying the position and steepness of radiation dose-response curves. Int J Radiat Biol 1997;71: 531–42 [see comments]. [6] Brenner DJ, Hall EJ. Fractionation and protraction for radiotherapy of prostate carcinoma. Int J Radiat Oncol Biol Phys 1999;43:1095–101. [7] Cheung R, Tucker SL, Lee AK, et al. Dose–response characteristics of low- and intermediate-risk prostate cancer treated with external beam radiotherapy. Int J Radiat Oncol Biol Phys 2005;61:993–1002. [8] Fowler J, Chappell R, Ritter M. Is alpha/beta for prostate tumors really low? Int J Radiat Oncol Biol Phys 2001;50: 1021–31. [9] Fowler JF, Ritter MA. A rationale for fractionation for slowly proliferating tumors such as prostatic adenocarcinoma. Int J Radiat Oncol Biol Phys 1995;32:521–9. [10] Guttenberger R, Thames HD, Ang KK. Is the experience with CHART compatible with experimental data? A new model of repair kinetics and computer simulations Radiother Oncol 1992; 25:280–6. [11] Koper PC, Jansen P, van PW, et al. Gastro-intestinal and genitourinary morbidity after 3D conformal radiotherapy of prostate cancer: observations of a randomized trial. Radiother Oncol 2004;73:1–9. [12] Lukka H, Hayter C, Warde P, et al. A randomized trial comparing two fractionation schedules for patients with localized prostate cancer. Radiother Oncol 2003;69:S7. [13] Macbeth F, Overgaard J. Expert reviews, systematic reviews and meta-analyses. Radiother Oncol 2002;64:233–4. [14] Overgaard J, Bentzen SM. Evidence based radiation oncology. Radiother Oncol 1998;46:1–3 [editorial; comment]. [15] Ritter MA, Chappell R, Fowler JF, Tome WA, Kupelian PA. A multi-institutional phase I/II trial of dose-per-fraction escalation for localized prostate cancer. ASCO Prostate Symposium, Orlando, FL 2005. Abstract 91. [16] Sydes MR, Stephens RJ, Moore AR, et al. Implementing the UK Medical Research Council (MRC) RT01 trial (ISRCTN 47772397): methods and practicalities of a randomised controlled trial of conformal radiotherapy in men with localised prostate cancer. Radiother Oncol 2004;72:199–211. [17] Valdagni R, Italia C, Montanaro P, et al. Is the alpha-beta ratio of prostate cancer really low? A prospective, non-randomized trial comparing standard and hyperfractionated conformal radiation therapy Radiother Oncol 2005;75:74–82.