RADIATION THERAPY FOR LOCALIZED OR LOCALLY ADVANCED PROSTATE ADENOCARCINOMA AND MYELOMA INCIDENCE IN A POPULATION-BASED COHORT

Authors


Rohit P. Ojha, 44 Binney Street, SM 271 Boston, MA 02115, USA.e-mail: rohit_ojha@dfci.harvard.edu

Abbreviations
EBRT

external beam radiation therapy

SEER

Surveillance, Epidemiology and End Results

ICD-O-3

International Classification of Diseases for Oncology 3rd Edition

DAG

directed acyclic graph

HR

hazard ratio

IQR

interquartile range

SIR

standardized incidence ratio

MGUS

monoclonal gammopathy of undetermined significance

INTRODUCTION

Myeloma is the second most common haematological malignancy in the USA [1]. This plasma cell malignancy is characterized by extensive genetic and chromosomal abnormalities [1]. Myeloma is difficult to treat and confers a poor prognosis; the median survival is 3–7 years, depending on therapeutic course [1]. Ionizing radiation, such as that used in radiation therapy, may have a role in myelomagenesis through radiation-induced genomic and chromosomal instability [2,3]. Radiation therapy (external beam radiation therapy [EBRT] or brachytherapy) is one of the main therapeutic options for localized or locally advanced prostate cancer [4,5]. Prostate irradiation may result in unintentional radiation exposure to the pelvis, which contains a high concentration of active bone marrow in adults [6], an exposure that might be relevant to myeloma incidence and a potential threat to patients with prostate cancer who are treated with radiation therapy. We, therefore, investigated the effect of radiation therapy (EBRT alone or brachytherapy alone) on myeloma incidence in a population-based cohort of 168 612 patients with localized or locally advanced prostate adenocarcinoma.

METHODS

The data used for this analysis have been previously described [7]. Briefly, we used data from nine Surveillance, Epidemiology and End Results (SEER) registries [8] to identify eligible patients. Men with newly diagnosed localized or locally advanced prostate adenocarcinoma between January 1988 and December 2003, with extended follow-up until December 2004, who were treated with EBRT, brachytherapy, surgery, or no definitive therapy and survived >1 year after prostate cancer diagnosis were eligible for our analyses. Patients with localized or locally advanced prostate adenocarcinoma were identified according to the International Classification of Diseases for Oncology 3rd Edition (ICD-O-3) histology code for adenocarcinoma (8140) [9] and the SEER historic stage designation of ‘local/regional’, which allows consistent definitions of stage over time [10].

We used the ICD-O-3 definition of myeloma for our outcome, which includes solitary plasmacytomas and multiple myeloma [9]. The SEER database contained information regarding initial therapy for each patient. These data were used to create a categorical variable for initial therapy that consisted of mutually exclusive categories for EBRT alone, brachytherapy alone, surgery alone and no definitive therapy (i.e. no radiation therapy or surgery [reference category]).

A minimal sufficient set of covariates for which to adjust in the analyses were identified a priori using the back-door criterion in a directed acyclic graph (DAG) [11–13], which encoded risk factors for myeloma incidence and clinical characteristics that guide treatment decisions for localized or locally advanced prostate adenocarcinoma [4,5,14–21]. Our DAG (Fig. 1) indicated that adjustment for age at prostate cancer diagnosis, race, prostate cancer grade and comorbidity could reduce confounding bias when estimating the effect of radiation therapy on myeloma incidence. Consequently, age at diagnosis was included as a continuous variable in our analyses. The patient’s race was categorized as White (reference category), Black or Other. Prostate cancer grade was categorized according to the American Joint Classification on Cancer guidelines for grading tumours (Grade I: well-differentiated [reference category]; Grade II: moderately differentiated, Grade III: poorly differentiated; Grade IV: undifferentiated) [22]. Comorbidity was defined as physician-determined presence of comorbidity at the time of diagnosis that precluded surgery as a therapeutic option.

Figure 1.

Proposed influence structure for the relationship between EBRT and myeloma incidence among patients with localized or locally advanced prostate adenocarcinoma. LHCs, lymphohaematopoietic cancers

Cox proportional hazards regression, with censored observations, was used to estimate the hazard ratios (HRs) and corresponding 95% CIs of myeloma incidence after EBRT alone, brachytherapy alone and surgery alone compared with no definitive therapy after adjusting for age at diagnosis, race, grade and comorbidity. Furthermore, we estimated the effect of EBRT on myeloma incidence by duration of follow-up (≤10 years/>10 years) based on evidence of increased radiation-induced myelomagenesis after 10 years [3]. Insufficient data were available to estimate the long-term (>10 years) effect of brachytherapy on myeloma incidence. Person-time was measured in years from the date of prostate cancer diagnosis. Patients who did not develop myeloma were censored at the time of last follow-up, incident malignancy other than myeloma, or death. The proportionality assumption was evaluated by graphing and examining interaction terms in the model; no violations were detected.

RESULTS

Our study population consisted of 168 612 men with localized or locally advanced prostate adenocarcinoma. Patients who were not treated with definitive therapy were older (mean age = 73.4, SD = 9.0) than patients treated with EBRT (mean age = 70.6, SD = 7.0), brachytherapy (mean age = 66.7, SD = 7.8), or surgery (mean age = 65.7, SD = 9.1). The brachytherapy group had the lowest proportion of Black men (brachytherapy = 8.7%, surgery = 9.6%, EBRT = 12.1%, no definitive therapy = 12.3%) and the lowest proportion of high grade tumours (poorly differentiated or undifferentiated) at diagnosis compared with the other treatment groups (brachytherapy = 7.1%, surgery = 18.4%, EBRT = 21.4%, no definitive therapy = 23.5%).

This cohort yielded 344 incident myeloma cases during 1 064 820 person-years of follow-up after prostate adenocarcinoma diagnosis. The brachytherapy group accrued the shortest duration of follow-up (median = 3.8 years, inter-quartile range [IQR]= 2.4, 5.7) compared with other treatment groups, whereas the surgery group accrued the longest duration of follow-up (median = 6.6 years, IQR = 3.6, 10.2). The highest proportionate mortality during follow-up was observed in the no definitive therapy group compared with the other treatment groups (no definitive therapy = 33.6%, EBRT = 26.0%, surgery = 21.5%, brachytherapy = 5.8%). Table 1 provides detailed baseline and follow-up characteristics for this cohort by treatment group.

Table 1.  Characteristics of patients diagnosed with localized or locally advanced prostate adenocarcinoma in the SEER database, 1988–2003
VariableEBRT (n = 41 986)Brachytherapy (n = 10 259)Surgery (n = 84 031)No definitive therapy (n = 32 336)
Baseline    
 Age, mean (SD)   70.6 (7.0) 66.7 (7.8)   65.7 (9.1)   73.4 (9.0)
 Race, n (%)    
  White33 089 (80.5)8 960 (87.3)72 629 (86.4)26 289 (81.3)
  Black 5 060 (12.1)  888 (8.7) 8 043 (9.6) 3 992 (12.3)
  Other  3 117 (7.4)   411 (4.0) 3 359 (4.0) 2 055 (6.4)
Grade, n (%)    
 Well differentiated 4 296 (10.2)  659 (6.4)12 733 (15.2) 3 922 (12.1)
 Moderately differentiated28 689 (68.3) 8 871 (86.5)55 899 (66.5)20 829 (64.4)
 Poorly differentiated 8 858 (21.1)  721 (7.0)15 098 (18.0) 7 464 (23.1)
 Undifferentiated, anaplastic   143 (0.3)    8 (0.1)   301 (0.4)    121 (0.4)
Comorbidity, n (%) 1 248 (3.0)   87 (0.8)     0 (0.0)   647 (2.0)
Myeloma cases, n (%)    90 (0.2)    7 (0.1)   184 (0.2)    63 (0.2)
Total person-years contributed to cohort258 71744 126596 642165 335
Myeloma incidence density, cases/person-years35/100 00016/100 00031/100 00038/100 000
Median (IQR) duration of follow-up, years    5.6 (3.2, 8.7)3.8 (2.4, 5.7)    6.6 (3.6, 10.2)    4.4 (2.6, 7.1)
Lost to follow-up, n (%)   367 (0.9)101 (1.0)     852 (1.0)   539 (1.7)
Deceased during follow-up, n (%)10 922 (26.0)600 (5.8)18 102 (21.5)10 855 (33.6)

The adjusted relative hazards of myeloma incidence following EBRT and surgery were similar to the no definitive therapy group (EBRT: HR = 0.97, 95% CI 0.70, 1.35; surgery: HR = 1.02, 95% CI: 0.75, 1.39), whereas the relative hazard of myeloma incidence following brachytherapy was lower than the no definitive therapy group (brachytherapy: HR = 0.60, 95% CI: 0.28, 1.33). Furthermore, the effect of EBRT on myeloma incidence appeared to decrease with prolonged follow-up (≤10 years: HR = 1.15, 95% CI: 0.83, 1.61; >10 years: HR = 0.52, 95% CI: 0.13, 2.19). Table 2 details the relative hazards and corresponding 95% CIs by treatment type.

Table 2.  Relative hazards of myeloma incidence by therapeutic approach among patients with localized or locally advanced prostate adenocarcinoma
TreatmentUnadjusted HR (95% CI)Adjusted HR* (95% CI)
  • *

    Adjusted for age at diagnosis, race, grade and comorbidity.

EBRT0.86 (0.62, 1.19)0.97 (0.70, 1.35)
 ≤10 years follow-up1.01 (0.73, 1.41)1.15 (0.83, 1.61)
 >10 years follow-up0.52 (0.13, 2.19)0.52 (0.13, 2.19)
Brachytherapy0.45 (0.20, 0.98)0.60 (0.28, 1.33)
Surgery0.73 (0.54, 0.97)1.02 (0.75, 1.39)
No definitive therapy1.00 (Reference)1.00 (Reference)

DISCUSSION

This report extends our efforts to evaluate biologically plausible haematological consequences of radiation therapy among patients with prostate cancer, while considering the potential for differential radiosensitivity of cells derived from haematopoietic progenitors [23] and expands on our previous findings [7] of increased relative and absolute risks of acute myeloid leukemia following EBRT among patients with localized or locally advanced prostate adenocarcinoma. Our results indicate that neither EBRT alone nor brachytherapy alone increases the relative hazard of myeloma incidence among patients with localized or locally advanced prostate adenocarcinoma. The point estimate for brachytherapy actually suggests an inverse relation to myeloma incidence. However, this point estimate may be misleading because of sparse-data bias (known to induce a bias away from the null[24]) attributable to few incident myeloma cases in this group. Our results also indicate that the effect of EBRT on myeloma incidence is not increased among patients followed >10 years after prostate cancer diagnosis.

Our overall results are consistent with the few previous analyses[25–27] that evaluated myeloma incidence following radiation therapy among patients with prostate cancer. However, the distinction between comparison groups in previous analyses and our analysis should be emphasized within the context of the ultimate question being posed by this investigation: Would patients with localized or locally advanced prostate adenocarcinoma, who were treated with radiation therapy, have assumed the risk (of myeloma incidence) of patients with localized or locally advanced prostate adenocarcinoma who were untreated, if the former group had not been treated with radiation therapy? This question represents a counterfactual contrast, which provides a framework for drawing meaningful inferences from data [24,28–30]. Clearly, this ideal comparison is unachievable because the same person cannot be simultaneously exposed and unexposed to radiation therapy. Comparison groups are therefore substituted to represent this ideal as closely as possible. Previous analyses [25–27] estimated standardized incidence ratios (SIRs) by comparing myeloma incidence between patients with prostate cancer treated with radiation therapy with the general population, whereas our analysis estimated HRs of myeloma incidence by comparing patients treated with radiation therapy with patients who were not treated with definitive therapy. Comparisons with the general population are poor representations of the counterfactual contrast in this scenario; a lack of exposure to radiation therapy does not necessarily qualify the general population as a valid comparison group when estimating treatment effects. Individuals in the general population who do not have a primary diagnosis of localized or locally advanced prostate adenocarcinoma are ineligible for prostate irradiation, therefore, the general population is a non-comparable entity. Estimates derived from comparisons with the general population offer limited insight regarding the effect of radiation therapy on myeloma incidence among patients with prostate cancer. Furthermore, factors other than those used to standardize the comparison groups contribute to disparate baseline risk of myeloma between patients with prostate cancer and the general population. Consequently, residual confounding would threaten the validity of previous estimates even if the general population were an appropriate comparison group [24].

Cuzick et al. [3] suggested that an increased relative risk of myeloma after radiation therapy for solid tumours may not be evident until 10–30 years after radiation, but this long empirical induction period poses a challenge when evaluating patients with prostate cancer because of older age at diagnosis and the resulting reduced sample sizes with prolonged duration of follow-up. For example, the results by McMaster et al. [25] indicated a marked increase in the relative risk of myeloma following EBRT for patients who survived ≥20 years (SIR = 4.11). However, this SIR estimate was based on four myeloma cases and was probably unreliable because of sparse data. Our stratified analysis indicated a decreased relative hazard of myeloma incidence following EBRT for patients with localized or locally advanced prostate adenocarcinoma followed >10 years, but this estimate also lacks durability, evident by the large confidence limit ratio [31]. Ultimately, the combination of a long empirical induction period and older age at diagnosis may further diminish the concern regarding myeloma incidence after radiation therapy for most patients with localized or locally advanced prostate adenocarcinoma, but additional data are needed for patients with long life expectancy after prostate adenocarcinoma diagnosis.

Although our results did not indicate an increased relative hazard of myeloma incidence after radiation therapy among patients with localized or locally advanced prostate adenocarcinoma, recent evidence may provide insight into our findings. Iwanaga et al. [32] reported a positive association between ionizing radiation and monoclonal gammopathy of undetermined significance (MGUS), a premalignant plasma cell disorder, and a longitudinal study by Landgren et al.[33] indicated that MGUS consistently precedes myeloma incidence. These combined findings suggest that a positive relation between EBRT and myeloma incidence is plausible. However, MGUS is related to several malignant and non-malignant causes of death other than myeloma [34]. Consequently, our estimates for the effect of radiation therapy on myeloma incidence could have incurred a bias toward the null (i.e. no apparent effect of radiation therapy) if myeloma incidence were associated with loss to follow-up from MGUS-related competing risks and this loss occurred more frequently among patients treated with radiation therapy than patients who were not treated with definitive therapy (i.e. selective loss to follow-up). Unfortunately, we were unable to evaluate quantitatively the potential impact of MGUS-related competing risks because data regarding MGUS are unavailable in the SEER database. Adequately powered longitudinal studies which determine MGUS and myeloma status might be able to provide further insight into this phenomenon.

In summary, our results indicate that neither EBRT nor brachytherapy increases the relative hazard of myeloma incidence among patients with localized or locally advanced prostate adenocarcinoma. Despite a previous suggestion that myeloma incidence may be a concern >10 years after radiation therapy [3], our results indicate that the effect of EBRT on myeloma incidence is not increased among patients followed >10 years after prostate cancer diagnosis. These results are particularly important for raising awareness that our previous findings [7] regarding increased relative and absolute risks of acute myeloid leukemia after EBRT among patients with localized or locally advanced prostate adenocarcinoma should not be generalized to other haematological outcomes. Independent evaluations of potential haematological outcomes that consider the underlying etiological pathways and this unique population at risk may reveal considerable variation in relative risks and clinical relevance.

ACKNOWLEDGMENTS

We thank Lori Fischbach for comments on an earlier version of this manuscript. The authors declare no financial or non-financial competing interests. This research was granted exempt status by the University of North Texas Health Science Center Institutional Review Board.

CONFLICT OF INTEREST

None declared.

Ancillary