Impact of radiation therapy dose escalation on prostate cancer outcomes and toxicities

Abstract

OBJECTIVES:

Freedom from biochemical failure (FFBF) is a common primary outcome of randomized controlled trials (RCTs) of prostate cancer (PCa). We aimed to determine how increasing the PCa biologically equivalent dose (BED) of external radiation therapy (RT) is correlated with FFBF and overall patient outcomes: overall survival (OS), distant metastasis (DM), and cancer specific mortality (CSM); as well as genitourinary (GU) and gastrointestinal (GI) toxicities.

METHODS:

We performed a meta-analysis of 6,884 PCa patients from 12 RCTs of external beam RT. Mixed effects regression models were used to estimate weighted linear relationships between BED and observed percentages of 5- and 10-year outcomes. For toxicities, a subset analysis of using 3D conformal vs. intensity modulated RT was performed.

RESULTS:

Increasing BED correlated with improved FFBF: 10-year absolute improvement of 9.6% and 7.2% for low- and intermediate -risk patients, respectively (p < 0.05); but not with improvement of OS, DM, or CSM at either time point. BED escalation was not correlated with increased acute toxicities; it was correlated with increased late GI toxicities in patients treated with 3D-CRT (1.5% increase over BED range, p < 0.01). IMRT patients had significantly fewer late toxicities, despite being treated at higher BED.

CONCLUSIONS:

RT BED escalation has resulted in significantly improved PCa FFBF at up to 10 years; but not with improvement in OS, DM, or CSM. Thus, FFBF is a poor surrogate of overall patient outcomes for trials of RT. Late toxicities were less frequent with IMRT than with 3D-CRT, even at higher BED.

INTRODUCTION

Prostate cancer (PCa) is the second most prevalent solid tumor diagnosed in men of the United States and Western Europe.¹ External beam radiation therapy (EBRT) is a cornerstone of curative management of PCa. Biochemical failure (BF), defined in the contemporary era as a prostate specific antigen (PSA) that rises by 2 ng/mL after the nadir after EBRT, is one way to measure EBRT outcomes.

BF is a marker for disease recurrence: local, regional, or distant metastasis (DM); and subsequent prostate cancer specific mortality (PCSM). For example, the Memorial Sloan Kettering group estimated that the median time from BF to DM is 5.4 years, and from BF to PCSM is 10.5 years.² Thus, among almost all randomized controlled trials (RCTs) of EBRT dose escalation, the primary outcome measure has been freedom from BF (FFBF), a harbinger for these more serious patient outcomes.

Nonetheless, the relationship between FFBF and other patient outcomes (DM, PCSM, or overall survival [OS]) is not fully understood. An analysis of the National Cancer Database (NCDB) revealed that treatment to a dose ≥ 75.6 Gy is associated in improved OS;³ on the other hand, individual RCTs of EBRT dose escalation have not revealed a benefit in OS. We aimed to determine how increasing the PCa biologically equivalent dose (BED) of EBRT is correlated with FFBF and overall patient outcomes OS, DM, and CSM; as well as toxicities.

METHODS AND MATERIALS

Definitions and evidence acquisition

We defined EBRT as (1) conventionally fractionated EBRT (CFRT): 76 to 80 Gy in 1.8 - 2 Gy fractions; or (2) hypofractionated EBRT (HFRT, about 50 to 66 Gy in 2.1 to 3.5 Gy fractions). We defined inclusion criteria for the literature search using the Population, Intervention, Control, Outcome, Study Design (PICOS; Table 1) approach. We conducted a systematic search using the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA; Figure 1) literature selection process.⁴ We use the term “BF” because this is used by the Radiation Therapy Oncology Group (RTOG) and American Society for Radiation Oncology (ASTRO).⁵

Table 1.

Population, Intervention, Control, Outcome, Study Design (PICOS) inclusion criteria.

Population	Men with localized (T1-T2, N0-Nx, M0) and locally advanced (T3-T4, N0-Nx, M0) prostate cancer
Intervention	CFRT and HFRT, either with 3D-CRT or IMRT
Control	A control group of CFRT, either as a lower total dose (when compared to dose escalated CFRT), or when compared to HFRT
Outcomes
Efficacy	FFBF at 5-year and at 10-year actuarial FU, stratified by risk groups Phoenix definition preferred ASTRO definition may only be used if there is ≥5 year median FU May also have other patient outcomes: OS, DM, CSM @ 5-year and @ 10-year actuarial FU
Safety	Acute RTOG (or similar scale) toxicities, GI and GU Late RTOG (or similar scale) toxicities, GI and GU
Study design
	Large (n > 150), prospective, start of enrollment year ≥ 1990

Abbreviations: 3D-CRT: 3 dimensional conformal radiation therapy; ASTRO: American Society for Radiation Oncology (3 consecutive rises); BED: biologically equivalent dose; CFRT: conventionally fractionated radiation therapy; CSM: cancer specific mortality; FFBF: freedom from biochemical failure; DM: distant metastasis; FU: follow-up; GI: gastrointestinal; GU: genitourinary; HFRT: hypofractionated radiation therapy; IMRT: intensity modulated radiation therapy; OS: overall survival; RTOG: Radiation Therapy Oncology Group

Figure 1.

Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) flow diagram literature selection process.

The meta-analysis included 6,884 PCa patients (n) from 12 randomized controlled trials (N) with enrollment started after 1990, evaluating CFRT escalation^6-13 or HFRT.^14-18 The trials had ≥ 5-year median and actuarial follow-up. Patients were treated with three dimensional conformal RT (3D-CRT) or intensity modulated RT (IMRT). Retrospective studies were excluded. The number of manuscript publications associated with each study varied (some focusing on outcomes, others on toxicities). For reference, RCTs are listed in Supplementary Table 1.

The EBRT studies included reached a biologically equivalent dose (BED_1.5) of 180 - 200 Gy, assuming an α/β of 1.5. Although SBRT may achieve BEDs_1.5 > 200 Gy, studies using SBRT were not included because: (1) we wanted to focus on CFRT and HFRT (which have similar mechanisms of cell death,^19-21 vs. SBRT²²); and (2) most studies of SBRT do not have reported FFBF, DM, CSM, and toxicities at 10 years with > 5 year median follow-up time.²³ Studies using brachytherapy were excluded because: (1) analyzing only EBRT provides a more homogenous patient population; (2) there is heterogeneity associated with different number of fractions and dose per fraction of high dose rate brachytherapy;²⁴ and (3) studies of low dose rate brachytherapy could not be a accurately captured by the BED calculation model used in fractionated approaches (described below).

Androgen deprivation therapy (ADT) was prescribed to patients in certain studies (detailed in Supplementary Table 1). Unfortunately, we cannot discern which patients received ADT, or if these patients were “favorable intermediate” vs. “unfavorable intermediate” risk.²⁵ Nonetheless, dose escalation studies of EBRT have demonstrated benefit of dose escalation up to ~180 - 200 Gy,²⁶ among all risk types, with or without ADT.^6-13

Statistical analysis

BEDs were calculated for patients of various risk groups, at various α/β ratios (at 1.5, 3.0, and 10), based on the following formula:

$$\text{BED}=(nd[1+d∕(\alpha ∕\beta )])$$

For reference, sample curves of BED vs. α/β ratios of some of the fractionation regiments used in this analysis are compared in Supplementary Figure 1. To report outcomes, we calculated the BED_1.5 based on the RT fractionation regimen from each study (Supplementary Table 1). We reported FFBF rates among low-, intermediate-, and high-risk group patients; and the OS, DM, and CSM, as reported within each study. We chose to use 5-year and 10-year actuarial time points because we anticipated them to show a change in FFBF, and perhaps see a subsequent change in DM and PCSM

We related the BED to FFBF, OS, DM, and CSM. The Phoenix (i.e. nadir + 2 ng/mL) definition was used for FFBF. If this was unavailable, but the study had at least 5 years of FU, then the ASTRO (i.e. 3 consecutive PSA rises) definition could be used – only one study fell in this category.¹⁴

To report severe (i.e. Grade 3-4) early genitourinary (GU) and gastrointestinal (GI) toxicity, we calculated the BED₁₀ based on the RT fractionation regimen from each study. For late toxicity, we calculated the BED_3.0. We analyzed early and late toxicity rates with three dimensional conformal RT and intensity modulated RT (i.e. 3D-CRT and IMRT), in both the acute and late settings, as a function of the BEDs.

This analysis is predicated on a few assumptions. First, we assume that modeling of radiation effect is correct for both CFRT and HFRT. This is a reasonable assumption, since most radiobiological models of PCa using doses used in this analysis have reported similar PCa cell death profiles and consistent α/β ratios.^19-21 Cancer cells treated with high doses per fraction (e.g. > 10 Gy, as with studies of SBRT) are hypothesized to die by means (e.g. lipid membrane phosphorylation) that are not explained by typical radiobiological models;²² thus, those studies were excluded from this analysis. Second, we assume that the delivered dose was delivered as planned and marginal misses that would have effectively decreased the effectiveness of the prescribed doses are negligible. Unfortunately, implanted dosimeters were not mandatory on the trials to assess for these misses; presumptively, if marginal misses had been present, all patients would have been affected. Third, we assume that ADT effects are independent of radiation dose effects, and that PSA rise is never due to testosterone recovery. This assumption would primarily apply to 40% of the high risk patients and 7% of intermediate risk patients in this meta-analysis.^{7, 9, 16-18}

Mixed effects meta-regression models were used to estimate weighted linear relationships between BED and percentages of FFBF, OS, DM, and CSM; and the observed percentages of patients experiencing toxicities. The weight applied to a given study’s published effect estimate was the ratio of the number of patients analyzed in that study divided by the total number of patients over all studies used for the meta-estimate of that effect. This approach is preferable to one with fixed effects in the context of patient care decision making.²⁷ We saw clear evidence of heterogeneity for some outcomes and less so for others. Having meta-analyzed a considerable number of outcome variables here, we chose to assume heterogeneity and apply one consistent robust approach which may have been somewhat conservative for some outcomes where the data were highly consistent.

Results were summarized by slopes representing expected changes in FFBF, OS. DM, CSM, or toxicity percentages per 10-unit change in BED. Separate meta-regression models were fitted for 3D and IMRT for GU and GI toxicities. The models were used to estimate the expected outcome and toxicity percentages at these selected BEDs along with 95% confidence intervals (CIs). Statistical analyses were conducted with SAS version 9.4 (SAS Institute, Cary, NC, USA).

RESULTS

Study characteristics

There were 6,884 patients treated with EBRT from 12 studies (Supplementary Table 1); of these men, 5,056 were treated on trials evaluating CFRT escalation, and 1,828 were treated on trials evaluating CFRT vs. HFRT. The median patient age among the studies was 69, and the range of the median ages was 67 to 75. RCT arms were well-balanced with respect to patient age, and age was used in the RCT multivariate analyses.

Studies using CFRT^6-13 had some of the earliest enrollment times, from 1993 to 1998;⁶ on the other hand, those using or HFRT^14-18 had some of the most recent enrollment times, up to 2010.¹⁸ The total number of fractions ranged from 19 to 44; the dose per fraction ranged from 1.8 to 3.4 Gy. The BEDs_1.5 ranged from 144 to 197 Gy (median: 173 Gy). There was overlap in the values of BED_1.5 attained by studies using CFRT or HFRT (Supplementary Figure 1). ADT was used in 40% of the high risk patients and 7% of intermediate risk patients, these patients were primarily treated on select trials that did not prohibit ADT use.^{7, 9, 16-18}

The primary outcome measure among all studies was BF; every study but one¹⁴ used the Phoenix definition. All studies reported toxicity rates. There were three studies that used IMRT: one of these studies treated patients with both 3D-CRT and IMRT,¹³ while the other three studies used IMRT exclusively.^{17, 18} Studies using IMRT exclusively had accrual start dates after the year 2000.

Outcomes

Weighted regression slope coefficients for 5- and 10-year outcomes are shown in Table 2. For FFBF, 10 Gy increments in BED_1.5 from 140 to 200 Gy was associated with statistically significant 5.3% and 7.5% absolute improvements in percent FFBF at 5-years for intermediate- and high-risk patients, respectively (p < 0.01). Positive associations were similar at 10 years (Figure 2, lower left), though the p-value among the high-risk was not significant (p = 0.10). BED escalation was not associated with improvement of OS, DM, or CSM at the 5-year time point (p-values all >0.05). Similarly, BED escalation was not associated with improvement of OS, DM, CSM at the 10-year time point (p-values all >0.05). The lack of an association between OS, CSM, and DM, and BED_1.5 can be visualized in Figure 2 (right panel).

Table 2.

Outcomes and toxicities of increasing BED on outcomes for prostate cancer radiotherapy among EBRT trials.

			5-year values			10-year values
Outcomes			% range	% value if slope is near 0	Slope (p-value)	% range	% value if slope is near 0	Slope (p-value)
Biochemical	FFBF low-risk		70-100		5.5 (0.04)	45-90		9.6 (<0.01)
	FFBF intermediate-risk		70-95		5.3 (<0.01)	40-60		7.2 (0.01)
	FFBF high-risk		40-80		7.5 (<0.01)	20-50		5.8 (0.10)
Patient overall	OS		82-97	87	−0.2 (0.60)	66-83	71	0.7 (0.34)
	DM		1-14	4	0.2 (0.66)	†	†	†
	CSM		0-8	2	−0.1 (0.68)	1-14	5	0.1 (0.87)
Toxicities (RTOG)			Range		Slope (p-value)
Acute (BED10)	GU	3D-CRT	0-13	5	−0.3 (0.76)
		IMRT	0-8	4	5.4 (0.47)
	GI	3D-CRT	0-6	2	−1.0 (0.19)
		IMRT	0-3	2	1.0 (0.73)
Late (BED3)	GU	3D-CRT	0-13	5	0.4 (0.31)
		IMRT	0-4	3	0.4 (0.71)
	GI	3D-CRT	0-10		1.5 (<0.01)
		IMRT	1-3	2	−0.2 (0.60)

Abbreviations: 3D-CRT: 3 dimensional conformal radiation therapy; BED: biologically equivalent dose; CSM: cancer specific mortality; DM: distant metastasis; FFBF: freedom from biochemical failure; GI: gastrointestinal; GU: genitourinary; IMRT: intensity modulated radiation therapy; OS: overall survival; RTOG: Radiation Therapy Oncology Group

Note:

^†denotes insufficient data reported

p-values < 0.05 are boldfaced.

Figure 2. Plots of outcomes vs. BED_1.5 of trials included in this analysis.

Radiotherapy BED escalation has resulted in improved FFBF of low-, intermediate-, and high-risk patients at 5 and 10 years (top left, lower left). However, BED escalation has correlated with improved patient outcomes of OS, DM, or CSM at 5 and 10 years (top right, lower right).

Toxicity

Toxicity analysis is shown in Figure 3. For acute toxicity, 10 Gy increments of BED₁₀ from 66 to 96 (among 3D-CRT patients) were associated with risk differences of −0.3% (p = 0.76) and −1.0% (p = 0.70) in genitourinary and gastrointestinal toxicities (range, 0% - 13%), respectively. For IMRT patients, BED₁₀ was between 81 to 96 Gy and 10 Gy increments of BED₁₀ were associated with risk differences of 5.4% (p = 0.47) and 1.0% (p = 0.73) for acute GU and GI toxicities, respectively.

Figure 3. Plots of toxicities vs. BED₁₀ and BED_3.0 of trials included in this analysis.

The advent of IMRT in the early 2000s coincided with radiotherapy BED escalation using CFRT and HFRT; thus trials using 3D-CRT and IMRT were analyzed separately. Radiotherapy BED escalation was not correlated with increased acute GU or GI toxicities in patients treated with 3D-CRT or IMRT. Radiotherapy BED escalation was correlated with increased late GU and GI toxicities in patients treated with 3D-CRT. All patients treated with IMRT had significantly less late GU and GI toxicity, despite being treated at higher BEDs. Trend lines are plotted, and the corresponding slopes are listed in Table 2.

For late toxicity, 10 Gy increments of BED₃ from 98 to 133 (among 3D-CRT patients) were associated with risk differences of 0.4% (p = 0.31) and 1.5% (p < 0.01) in genitourinary and gastrointestinal toxicities (range, 0% - 13%), respectively. For IMRT patients, BED₃ was between 121 to 134 Gy; 10 Gy increments of BED₁₀ were associated with risk differences of 0.4% (p = 0.71) and −0.2% (p = 0.60) for late GU and GI toxicities, respectively.

DISCUSSION

Since the 1990s, randomized controlled trials of PCa EBRT have used FFBF as the primary outcome measure. BF has been shown to lead to DM and PCSM over subsequent years in other analyses; thus, it was postulated that if BF were improved in a PCa RCT, overall patient outcomes would also improve. We explored this hypothesis; our results demonstrate two important findings. First, we report the imperfections in using PSA as a surrogate for patient outcomes: although FFBF rates have improved with increasing BED, overall patient outcomes (i.e. DM, CSM, OS) did not improve. Second, we demonstrated that in the era of BED escalation, technological advances of RT delivery (namely with IMRT) have kept acute toxicities low, and they have lowered the late toxicities when compared to the previous era.

PSA is the most studied marker of PCa. PSA is used for screening, risk stratification, as an indicator of treatment success for individual cancer patients, and as a surrogate for patient outcomes in clinical research. The natural history of PCa is relatively long. Thus, waiting for DM or PCSM would make trials lengthy; the methods of trials may be outdated by the time the trials are published.²⁸ Additionally, DM or PCSM can be influenced significantly by secondary interventions and thus may not reflect impact of primary therapy. Finally, the age of diagnosis and competing comorbidities may impact the development of DM and PCSM, given the long time it takes for these endpoints to be reached.

Our analysis included patients enrolled on RCTs, with an endpoint of BF (Table 1, Figure 1). The patients were treated during an era of BED escalation. BED was escalated by both increasing the number of fractions of CFRT (which was possible with technologies, namely with IMRT²⁹) and hypofractionation (Figure 2). Select intermediate- and high-risk patients (8% and 41%, respectively) were also treated with ADT.

With respect to patient outcomes, we showed that increasing the BED was associated with improved FFBF of low- and intermediate-risk patients at 10-year endpoints (Figure 2, left panel; Table 2). However, BED escalation did not affect improved overall patient outcomes: DM, CSM, or OS (Figure 2, right panel; Table 2). The lack of an association between BED and other outcomes may be secondary to several causes. First, patients who had BF had a true cancer recurrence, can receive systemic therapy, thus prolonging their time until documented DMs or death. Second, the patients with BF may have had PSA production from normal prostatic tissue (e.g. from benign prostatic hypertrophy [BPH]). Third, PSA testing frequency may alter the ability of the source RCTs to detect a biochemical event, which may be a reason behind the negative findings in FFBF.

Kalbasi et al.³ noted that dose-escalated EBRT (≥ 75.6 Gy) was associated with improved OS at 7 years in the NCDB, which was not seen in this analysis. This discrepancy may be due to a number of reasons. First, the authors may have included patients with very advanced local disease. Second, they cannot explain why their patients received a particular dose. For example, patients selected to receive a lower dose may have had comorbidities (making them ineligible for the RCTs in our meta-analysis). Additionally, patients in the < 75.6 Gy treatment group may have initially been prescribed ≥ 75.6 Gy, but had acute toxicities, secondary to either advanced local disease or comorbidities. Moreover, the NCDB analysis included only patients in the US; it did not include HFRT regimens; and it did not provide data on DM, CSM, or toxicities, which are available in our analysis. Thus, while we value the analysis of Kalbasi and colleagues, we caution clinicians in using further RT dose escalation on a supposition of improved OS.

Our findings have clinical implications. First, future studies of EBRT should not be powered to solely detect a difference in FFBF. It is unclear what endpoint should be chosen, and the Intermediate Clinical Endpoints in Prostate Cancer (ICECaP) endeavor has been assembled to explore endpoints that would be recognized by regulatory agencies in early trials.³⁰ Future studies should assess other biomarkers; or, novel imaging techniques (e.g. multiparametric magnetic resonance imaging, single-photon emission tomography, small-molecule positron emission tomography), which may help to differentiate benign tissue from recurrence and locate the source of recurrence (i.e. local vs. distant).³¹

The second important finding of this analysis is that technology (i.e. IMRT) has significantly improved severe (i.e. grade 3-4) RTOG late toxicities, despite treatments at higher BEDs in the previous era of 3D-CRT. For patients, this means less frequency, urgency, nocturia, dysuria, hematuria, ulceration, and stricture (which would typically managed with surgery). The post-prostatectomy Surveillance, Epidemiology, and End Results–Medicare analysis noted no difference between 3D-CRT and IMRT in the post-operative setting;³² thus, the technologies may impact toxicity differently in the primary vs. post-operative setting.

This study has limitations. First, we unfortunately do not have individual patient data and could not perform a pooled analysis. Some studies (e.g. the Dutch CKVO-96-10⁷) do not report 5-year FFBF for the high-risk subset, and these outcomes could not be incorporated into our model. It would be useful to compare the effect of biochemical progression in the hypofractionated studies as an independent group as well as the conventionally fractionated studies as a group. However, of the 12 studies that are included in our meta-analysis, only 5 had a hypofractionated arm;^14-18 if we compared CFRT vs. HFRT, we would have limited data points from the HFRT studies, precluding meaningful analysis and conclusions. We cannot comment on which patients benefit from ADT, we cannot analyze patient comorbidities, and we cannot analyze individual patient cancer characteristics (e.g. Gleason score, PSA, T-stage, % positive biopsies); instead, we analyze the risk groups used among the trials.

Certain outcomes (e.g. DM, CSM, OS) are typically reported for all patients in a RCT, not among the individual risk groups; thus, we could not perform subset analyses. Next, we have only evaluated trials using EBRT. We excluded trials of primary brachytherapy, surgery, or post-operative RT. It is unclear if our results would hold true in those settings.

We do not know what types of salvage systemic therapies these patients received. Newer studies may have better outcomes because novel systemic therapies are available to patients with BF. Although we assessed follow-up at 10 years, longer data (e.g. at 15 or 20 years) is needed because in certain cancers (e.g. breast), rates of local control at 5 years have been shown to affect outcomes at 15 years. Similarly, compared to meta-analyses of breast literature that contained 10,000 patients, or the NCDB analysis of 300,000 patients,³ our study may be underpowered.

There are other more important predictive factors for patient outcomes and toxicities besides BED alone, including patient age, ADT use, and initial GU symptoms. Additionally, the racial breakdown of patients is unknown; race affects outcomes³³ and toxicities.³⁴ The RTOG toxicity scores in this analysis are reported by physicians and not patients. RTOG scores do not include the evaluation of anorectal symptoms such as urgency of defecation and fecal incontinence, and they have few discrete values (i.e. Grades 0 - 5). The studies in this analysis generally did not use patient-reported quality of life measures.

CONCLUSION

BED escalation with either CFRT or HFRT has resulted in significantly improved FFBF at up to 10 years; but not with improvement in OS, DM, or CSM. PSA may be a poor surrogate for patient outcome in the contemporary era. Late toxicities were less frequent with IMRT than with 3D-CRT, even at higher BED.

Supplementary Material

Supplementary Figure 1

Supplementary Figure 1. A plot of BED curves for α/β ratios of 1.5 - 10 for several radiotherapy schedules.

Radiotherapy dose escalation and alternate fractionation techniques (i.e. with hypofractionation HFRT]) have enabled delivery of a high BED to the prostate while minimizing the dose to the normal tissues, thereby increasing the therapeutic ratio. This plot compares BED curves for α/β ratios of 1.5 to 10 Gy for some of the CFRT and HFRT regimens included in this meta-analysis (listed in Supplementary Table 1).