Abstract
Purpose
Previous studies have shown that external beam radiation therapy is associated with an increased risk of second primary cancer (SPC) among prostate cancer (PCa) patients, but the relative risks associated with newer and advanced radiation modalities such as proton beam therapy (PBT) and stereotactic body radiation therapy (SBRT) are unclear. This study aimed to assess the relative probability of SPC among patients treated with these newer modalities compared to intensity-modulated radiation therapy (IMRT).
Patients and Methods
Using the National Cancer Database (NCDB), N0M0 PCa cases diagnosed between 2004 and 2018 were identified. Second primary cancer probabilities were compared among those treated with curative-intent PBT, SBRT, and IMRT. Multivariable logistic regression and inverse probability of treatment weighting were used to generate adjusted odds ratios (aORs) and 95% confidence intervals (CIs).
Results
In total, 133 898 patients were included, with a median age of 69 years and median follow-up of 6.4 years. As their first course of treatment, 3420 (2.6%) received PBT, 121 211 (90.5%) received IMRT, and 9267 (6.9%) received SBRT. Compared with IMRT, PBT and SBRT were associated with lower SPC risk (aORs and 95% CIs, PBT: 0.49 [0.40-0.60], SBRT: 0.57 (0.51-0.63), P < .001). Inverse probability of treatment weighting analyses corroborated these results.
Conclusion
In this large national cohort, PBT and SBRT performed similarly and were associated with reduced SPC risk compared to IMRT when used as the first course of treatment.
Keywords: Localized prostate cancer, Secondary tumor, Radiation therapy, Proton beam therapy, Stereotactic body radiation therapy
Introduction
Radiation therapy (RT) has been associated with an increased risk of developing a second primary cancer (SPC) due to the exposure of normal tissue to radiation.1, 2, 3 The absolute rates of SPC are relatively low in prostate cancer (PCa) survivors,4, 5 but the long survival rates and frequent use of RT make SPC incidence an important topic in PCa research.6, 7
Among PCa survivors, SPCs commonly occur in the immediate surrounding organs (eg, the bladder and rectum) and often correspond with integral radiation exposure.8, 9 To spare normal tissues from exposure, advanced radiation technologies, such as proton beam therapy (PBT) and stereotactic body radiation therapy (SBRT), with high in-field targeting accuracy, steep dose gradients, and reduced out-of-field exposure, are increasingly being utilized.10, 11, 12, 13 The targeted delivery methods of these technologies may improve toxicity and quality of life and are expected to reduce the risk of SPC, but strong evidence is still lacking.11, 12, 13, 14, 15
Here, we perform a retrospective cohort analysis of US men with localized PCa undergoing definitive therapy between 2004 and 2018 to provide a comprehensive comparison of SPC occurrence among patients treated with PBT, SBRT, or intensity-modulated radiation therapy (IMRT) as the initial course of treatment. Previous studies have utilized the National Cancer Database (NCDB) and Surveillance, Epidemiology, and End Results database to assess SPC incidence among PCa survivors treated with radiation,16, 17 but none have performed an extensive comparison of PBT and other radiation modalities with careful handling of the data heterogeneity rooted in the real world setting, as we do here using inverse probability of treatment weighting (IPTW) and subgroup analyses. Accounting for this heterogeneity is important because patients often have limited access to advanced techniques, necessitating longer travel distances for treatment and potentially leading to shorter and inconsistent follow-up times. Therefore, stratification by travel distance and follow-up time subgroups allows for more fair comparisons of SPC incidence due to possible differences in data reporting. We hypothesize that, due to the steep dose gradients and lower overall integral dose achieved with PBT or SBRT, the incidence of SPC would be lower with PBT and SBRT compared with IMRT.
Patients and methods
Data source
A retrospective cohort analysis was performed using the NCDB 2019 Participant User File (PUF) for PCa, which contains data for patients diagnosed with or treated for PCa at Commission on Cancer-accredited facilities between 2004 and 2019.18 The NCDB is a hospital-based registry system that includes more than 70% of all diagnosed cancers in the United States.19
Inclusion and exclusion criteria
Men diagnosed with localized PCa between 2004 and 2018 were identified. Patients with clinical stage N0M0 prostate adenocarcinoma, whose PCa diagnosis was their only primary cancer or the first in the sequence, were included. Patients undergoing PBT, SBRT, or IMRT as their initial course of treatment were eligible. Those treated with more than 1 modality were excluded. Radiation volume was limited to the prostate.
To ensure adequate time for SPC development, patients were restricted to those with at least 2 years of known follow-up time from the start of treatment to death or last contact.7, 16 Based on prior literature,20, 21, 22 analyses were further restricted to patients living within 250 miles of their treatment facility, so that access to treatment within the PBT cohort was similar to access to treatment within other cohorts, thus reducing the potential for bias. Cases with potentially incomplete data (indicated by a reference date flag value of 0 in the NCDB23) and patients who received palliative care were removed. In total, 133 898 patients were included in the study (Table 1).
Table 1.
Selection diagram of the study population.
| Selection and exclusion criteria | Sample size | Excluded |
|---|---|---|
| NCDB prostate PUF cancer cases | 1 868 671 | - |
| Invasive tumor behavior | 1 868 331 | 340 |
| AJCC N0 M0 and exclude any metastasis cases | 1 485 931 | 382 400 |
| The current cancer diagnosis was the only one or the first in the sequence | 1 370 723 | 115 208 |
| Study cohorts as PBT, IMRT, and SBRT | 193 157 | 1 177 566 |
| Radiation volume as prostate (whole or partial) | 191 313 | 1844 |
| Minimum 2 y of follow-up time | 157 817 | 33 496 |
| Travel distance to treatment facility <250 miles | 137 675 | 20 142 |
| Remove cases that can no longer be updated and may not be current | 133 984 | 3691 |
| Remove patients receiving palliative care | 133 898 | 86 |
Abbreviations: NCDB, National Cancer Database; AJCC, American Joint Committee on Cancer; PBT, proton beam therapy; IMRT, intensity-modulated radiation therapy; SBRT, stereotactic body radiation therapy; PUF, Participant User File.
Primary outcome
The primary outcome of interest was an SPC at any primary site, defined using the sequence number variable in the NCDB. The sequence number is updated when the facility learns of an unaccessioned tumor that affects the sequence.23 A value of 00 indicated that the patient had no secondary malignancies, and 01 indicated that the patient had at least 1 secondary malignancy following the current PCa diagnosis. All patients with sequence number values of 02 or greater were excluded because SPC occurrence after PCa diagnosis could not be determined. Detailed information regarding the time to diagnosis and type of the SPC is unavailable in the NCDB, which only records data for the first course of treatment.
Treatment
Treatment modalities were categorized based on the phase I and phase II RT modalities reported in the NCDB. Reported phase I and II beam techniques were used to subcategorize IMRT and SBRT modalities. Only patients receiving RT for curative purposes (total dose ≥60 GyE for PBT and IMRT; 35-50 GyE for SBRT) were included in analyses. The target volume was limited to the prostate only. Patients treated with multiple modalities (eg, combined external beam RT and brachytherapy) were excluded.
Covariates
Facility/patient demographic and clinical factors included age, race/ethnicity, facility type, facility location by region, Charlson-Deyo comorbidity score (categorized as 0, 1, and 2+), T stage, prostate-specific antigen (PSA) value, Gleason score, androgen deprivation therapy (ADT), median income quartiles and educational attainment quartiles (measured as the percent without a high school degree) by zip code for 2008-2012 as surrogates for socioeconomic status, chemotherapy, insurance type, rurality of residence, travel distance to treatment facility (defined as crowfly distance from the treatment facility to the centroid of a patient’s zip code of residence, categorized as short: <12.5 miles, intermediate: 12.5-49.9 miles, and long: 50-250 miles), and follow-up time from the start of treatment to last contact or death (categorized as 2-4, 5-9, and 10+ years).
Statistical analysis
Descriptive statistics and univariate associations (eg, χ2 test, analysis of variance (ANOVA)) were calculated to capture the distribution of covariates. A multivariable logistic regression model was fitted to the data with binarized SPC occurrence as the outcome of interest. All model covariates described previously were included in the model, with a backward selection criterion of 0.05. The fitted model was used to generate adjusted odds ratios (aORs) with 95% confidence intervals (CIs). To avoid multicollinearity in model fitting, the year of diagnosis was not considered in the multivariate analysis due to its high correlation with the length of follow-up.
To further minimize confounding, IPTW, a propensity score (PS)-based approach, was implemented.24, 25 The PS was defined as the conditional probability of being treated given the observed covariates.26, 27 In our case with a 3-level cohort, the PS was estimated using a multinomial logit model containing all previously listed covariates. Average treatment effect matching weights were then calculated using the PS.28 Balance diagnostics were performed using absolute standardized difference, with absolute standardized difference <0.1 indicating sufficient balance. A weighted multivariable logistic regression model for SPC was used to generate PS-weighted aORs and CIs.
A second multivariable logistic regression model with interactions was fitted to explore cohort heterogeneity, in which interaction terms for cohort, travel distance to the treating facility, follow-up time, and Charlson-Deyo comorbidity score were included as additional model effects. This model was used to generate stratified aORs and 95% CIs and to guide a subsequent subgroup analysis among patients with short travel distances and long follow-up times (refer to Table S1 for results). Restricting our analyses to this patient subgroup allowed for more fair comparisons of SPC incidence, as differences in follow-up time and travel distance to the treatment facility can affect data collection/reporting.
An additional subgroup analysis was performed to assess the relative risks of SPC by radiation cohort among patients with no comorbid conditions (refer to Table S1 for results). Patients without comorbidities are less likely to develop new primary malignancies, so SPC occurrence is anticipated to be more closely linked to RT within this subgroup.
All analyses were done in SAS V9.4 (Cary, NC) and SAS macros,29 and the significance level was set at P < .05.
Results
Patient and treatment characteristics
As shown in Table 2, 3420 (2.6%) patients received PBT, 121 211 (90.5%) received IMRT, and 9267 (6.9%) received SBRT as their first course of treatment. The median (interquartile range) age by cohort was 67 (62-72) for PBT, 67 (62-73) for SBRT, 69 (63-74) for IMRT, and 69 (63-74) overall. Median follow-up times by cohort were 5.3, 5.2, and 6.6 years for PBT, SBRT, and IMRT, respectively, and 6.4 years overall. Patients who received PBT tended to travel farther for treatment even after restricting the travel distance to <250 miles and were treated primarily at academic/research facilities located in large metropolitan areas, which differed from the geographic distributions of other cohorts. Patients treated with PBT or SBRT also tended to live in areas with higher median incomes compared to those treated with IMRT. The crude percentages of patients with SPC were lower among the PBT (3.0%) and SBRT (4.3%) cohorts compared to IMRT (8.6%) (Table 2).
Table 2.
Baseline characteristics by treatment cohorts.
| Treatment cohorts |
||||||
|---|---|---|---|---|---|---|
| Covariate | Level | Total N = 133 898 (100%) |
Proton N = 3420 (2.6%) |
IMRT N = 121 211 (90.5%) |
SBRT N = 9267 (6.9%) |
P-valuea |
| Secondary tumor | No | 122 969 (91.8) | 3316 (97) | 110 782 (91.4) | 8871 (95.7) | <.001 |
| Yes | 10 929 (8.2) | 104 (3) | 10 429 (8.6) | 396 (4.3) | ||
| Age at diagnosis | Median (Q1-Q3) | 69 (63-74) | 67 (62-72) | 69 (63-74) | 67 (62-73) | <.001 |
| Race-ethnic groups | Non-Hispanic White | 102 098 (76.3) | 2714 (79.4) | 92 019 (75.9) | 7365 (79.5) | <.001 |
| Non-Hispanic Black | 21 361 (16) | 353 (10.3) | 19 650 (16.2) | 1358 (14.7) | ||
| Asian-Indians-Pac | 2964 (2.2) | 122 (3.6) | 2654 (2.2) | 188 (2) | ||
| Hispanic | 5813 (4.3) | 177 (5.2) | 5437 (4.5) | 199 (2.1) | ||
| Other/Unknown | 1662 (1.2) | 54 (1.6) | 1451 (1.2) | 157 (1.7) | ||
| Facility type | Nonacademic/Research program | 93 520 (69.9) | 401 (11.7) | 88 457 (73) | 4662 (50.3) | <.001 |
| Academic/Research program | 40 363 (30.1) | 3017 (88.3) | 32 741 (27) | 4605 (49.7) | ||
| Facility location | Eastern | 65 513 (48.9) | 846 (24.8) | 58 410 (48.2) | 6257 (67.5) | <.001 |
| Central | 49 332 (36.8) | 536 (15.7) | 46 468 (38.3) | 2328 (25.1) | ||
| Mountain | 4093 (3.1) | 19 (0.6) | 3719 (3.1) | 355 (3.8) | ||
| Pacific | 14 945 (11.2) | 2017 (59) | 12 601 (10.4) | 327 (3.5) | ||
| Charlson-Deyo score | 0 | 113 273 (84.6) | 2960 (86.5) | 102 277 (84.4) | 8036 (86.7) | <.001 |
| 1 | 16 088 (12) | 379 (11.1) | 14 675 (12.1) | 1034 (11.2) | ||
| 2+ | 4537 (3.4) | 81 (2.4) | 4259 (3.5) | 197 (2.1) | ||
| T stage | T2 | 39 009 (29.1) | 1126 (32.9) | 36 139 (29.8) | 1744 (18.8) | <.001 |
| Unknown | 512 (0.4) | 5 (0.1) | 487 (0.4) | 20 (0.2) | ||
| T1 | 87 614 (65.4) | 2181 (63.8) | 77 972 (64.3) | 7461 (80.5) | ||
| T3 | 6306 (4.7) | 106 (3.1) | 6158 (5.1) | 42 (0.5) | ||
| T4 | 457 (0.3) | 2 (0.1) | 455 (0.4) | 0 (0) | ||
| PSA | Unknown | 2510 (1.9) | 28 (0.8) | 2353 (1.9) | 129 (1.4) | <.001 |
| <10 | 89 399 (66.8) | 2657 (77.7) | 79 311 (65.4) | 7431 (80.2) | ||
| 10-20 | 25 586 (19.1) | 537 (15.7) | 23 812 (19.6) | 1237 (13.3) | ||
| >20 | 16 403 (12.3) | 198 (5.8) | 15 735 (13) | 470 (5.1) | ||
| Gleason score | 2-6 | 37 052 (27.7) | 1260 (36.8) | 31 957 (26.4) | 3835 (41.4) | <.001 |
| 7 | 60 655 (45.3) | 1640 (48) | 54 212 (44.7) | 4803 (51.8) | ||
| 8-10 | 34 016 (25.4) | 513 (15) | 32 983 (27.2) | 520 (5.6) | ||
| Unknown | 2175 (1.6) | 7 (0.2) | 2059 (1.7) | 109 (1.2) | ||
| ADT | No | 69 162 (51.7) | 2534 (74.1) | 58 663 (48.4) | 7965 (86) | <.001 |
| Yes | 62 720 (46.8) | 871 (25.5) | 60 682 (50.1) | 1167 (12.6) | ||
| Unknown | 2016 (1.5) | 15 (0.4) | 1866 (1.5) | 135 (1.5) | ||
| Census median income quartiles 2008-2012 | <$38 000 | 22 076 (16.5) | 322 (9.4) | 20 736 (17.1) | 1018 (11) | <.001 |
| $38 000-$47 999 | 29 903 (22.4) | 526 (15.4) | 28 116 (23.2) | 1261 (13.6) | ||
| $48 000-$62 999 | 35 883 (26.8) | 919 (26.9) | 33 004 (27.3) | 1960 (21.2) | ||
| ≥$63 000 | 45 843 (34.3) | 1651 (48.3) | 39 184 (32.4) | 5008 (54.2) | ||
| Percent no high school degree 2008-2012 | ≥21.0% | 21 093 (15.8) | 600 (17.6) | 19 339 (16) | 1154 (12.5) | <.001 |
| 13.0%-20.9% | 33 254 (24.9) | 804 (23.5) | 30 452 (25.1) | 1998 (21.6) | ||
| 7.0%-12.9% | 44 791 (33.5) | 986 (28.8) | 40 899 (33.8) | 2906 (31.4) | ||
| <7.0% | 34 663 (25.9) | 1028 (30.1) | 30 438 (25.1) | 3197 (34.5) | ||
| Chemotherapy | No | 132 083 (98.6) | 3396 (99.3) | 119 498 (98.6) | 9189 (99.2) | <.001 |
| Yes | 493 (0.4) | 18 (0.5) | 473 (0.4) | 2 (0) | ||
| Unknown | 1322 (1) | 6 (0.2) | 1240 (1) | 76 (0.8) | ||
| Primary payor | Other government/Not Insured/Unknown | 12 268 (9.2) | 197 (5.8) | 11 464 (9.5) | 607 (6.6) | <.001 |
| Private | 42 262 (31.6) | 1277 (37.3) | 37 572 (31) | 3413 (36.8) | ||
| Medicare | 79 368 (59.3) | 1946 (56.9) | 72 175 (59.5) | 5247 (56.6) | ||
| Urban/Rural 2013 | Metro | 110 892 (82.8) | 3090 (90.4) | 99 796 (82.3) | 8006 (86.4) | <.001 |
| Urban | 17 877 (13.4) | 120 (3.5) | 16 955 (14) | 802 (8.7) | ||
| Rural | 2307 (1.7) | 13 (0.4) | 2211 (1.8) | 83 (0.9) | ||
| Unknown | 2822 (2.1) | 197 (5.8) | 2249 (1.9) | 376 (4.1) | ||
| Great circle distance (miles) | Short | 84 489 (63.1) | 890 (26) | 78 792 (65) | 4807 (51.9) | <.001 |
| Intermediate | 41 935 (31.3) | 1383 (40.4) | 37 096 (30.6) | 3456 (37.3) | ||
| Long | 7474 (5.6) | 1147 (33.5) | 5323 (4.4) | 1004 (10.8) | ||
| Year of diagnosis | 2004-2008 | 38 195 (28.5) | 681 (19.9) | 36 453 (30.1) | 1061 (11.4) | <.001 |
| 2009-2013 | 50 606 (37.8) | 1318 (38.5) | 45 825 (37.8) | 3463 (37.4) | ||
| 2014-2018 | 45 097 (33.7) | 1421 (41.5) | 38 933 (32.1) | 4743 (51.2) | ||
| Years of follow-up time | 2-4 | 48 291 (36.1) | 1522 (44.5) | 42 389 (35) | 4380 (47.3) | <.001 |
| 5-9 | 57 666 (43.1) | 1514 (44.3) | 52 201 (43.1) | 3951 (42.6) | ||
| 10+ | 27 941 (20.9) | 384 (11.2) | 26 621 (22) | 936 (10.1) | ||
| Median (Q1-Q3) | 6.4 (4-9.4) | 5.3 (3.6-7.9) | 6.6 (4.1-9.6) | 5.2 (3.4-7.8) | ||
| Phase I+phase II dose (Gy) | Mean (Std Dev) | 74 (11.7) | 79.3 (10) | 76.7 (5.6) | 36.4 (1.8) | <.001 |
| Median (Q1-Q3) | 77.4 (74.8-79.2) | 79.2 (78-81) | 77.4 (75.6-79.2) | 36.3 (35-36.3) | ||
| Min-Max | 35-162.6 | 60-162 | 60-162.6 | 35-50 | ||
Abbreviations: IMRT, intensity-modulated radiation therapy; SBRT, stereotactic body radiation therapy; PSA, prostate-specific antigen; ADT, androgen deprivation therapy. Bold values represent statistical significance with p<0.05.
The parametric P-value is calculated by ANOVA for numerical covariates and χ2 test for categorical covariates.
Impact of radiation modality on second primary cancer
As illustrated in Table 3, the overall risk of SPC relative to IMRT, after adjusting for all available covariates, was lower among patients treated with PBT (aOR: 0.49, 95% CI: 0.40-0.60, P < .001) or SBRT (aOR: 0.57, 95% CI: 0.51-0.63, P < .001). The overall risk of SPC was also found to decrease with increasing travel distance (intermediate vs short: aOR: 0.84, 95% CI: 0.80-0.88; long vs short: aOR: 0.58, 95% CI: 0.51-0.65), increase with prolonged follow-up time (5-9 vs 2-4 years: aOR: 1.32, 95% CI: 1.26-1.39; 10+ vs 2-4 years: aOR: 1.38, 95% CI: 1.31-1.46), and increase with Charlson-Deyo comorbidity score (2+ vs 0: aOR: 1.20, 95% CI: 1.08-1.33; 1 vs 0: aOR: 1.15, 95% CI: 1.08-1.22) (all P < .001).
Table 3.
Multivariable logistic regression for secondary tumor.
| Secondary tumor=Yes |
|||||
|---|---|---|---|---|---|
| Covariate** | Level | N* | Odds ratio (95% CI) | OR P-value | Type 3 P-value |
| Treatment cohorts | Proton | 3416 | 0.49 (0.40-0.60) | <.001 | <.001 |
| SBRT | 9247 | 0.57 (0.51-0.63) | <.001 | ||
| IMRT | 121 027 | - | - | ||
| Race-ethnic groups | NH-White | 101 935 | 1.20 (1.13-1.28) | <.001 | <.001 |
| Asian-Indians-Pac | 2962 | 0.90 (0.77-1.06) | .211 | ||
| Hispanic | 5806 | 0.80 (0.70-0.90) | <.001 | ||
| Other/Unknown | 1658 | 0.48 (0.36-0.63) | <.001 | ||
| NH-Black | 21 329 | - | - | ||
| Facility type | Academic/Research program | 40 311 | 0.87 (0.83-0.91) | <.001 | <.001 |
| Nonacademic/Research program | 93 379 | - | - | ||
| Facility location | Pacific | 14 934 | 0.95 (0.89-1.03) | .199 | .005 |
| Mountain | 4072 | 0.87 (0.76-0.98) | .028 | ||
| Central | 49 281 | 1.04 (1.00-1.09) | .062 | ||
| Eastern | 65 403 | - | - | ||
| Charlson-Deyo score | 2+ | 4532 | 1.20 (1.08-1.33) | <.001 | <.001 |
| 1 | 16 071 | 1.15 (1.08-1.22) | <.001 | ||
| 0 | 113 087 | - | - | ||
| T stage | T2 | 38 940 | 1.14 (1.09-1.19) | <.001 | <.001 |
| Unknown | 510 | 1.02 (0.73-1.43) | .893 | ||
| T3 | 6299 | 1.00 (0.90-1.10) | .946 | ||
| T4 | 457 | 1.14 (0.82-1.59) | .433 | ||
| T1 | 87 484 | - | - | ||
| Gleason score | 7 | 60 554 | 0.87 (0.83-0.92) | <.001 | <.001 |
| 8-10 | 33 967 | 0.82 (0.77-0.87) | <.001 | ||
| Unknown | 2175 | 1.19 (1.03-1.36) | .015 | ||
| 2-6 | 36 994 | - | - | ||
| ADT | Unknown | 2015 | 1.05 (0.90-1.23) | .535 | <.001 |
| Yes | 62 618 | 1.11 (1.06-1.17) | <.001 | ||
| No | 69 057 | - | - | ||
| Census median income quartiles 2008-2012 | ≥$63 000 | 45 839 | 0.90 (0.84-0.96) | <.001 | .004 |
| $48 000-$62 999 | 35 880 | 0.95 (0.89-1.02) | .147 | ||
| $38 000-$47 999 | 29 898 | 0.97 (0.91-1.04) | .359 | ||
| <$38 000 | 22 073 | - | - | ||
| Years of follow-up time | 5-9 | 57 562 | 1.32 (1.26-1.39) | <.001 | <.001 |
| 10+ | 27 912 | 1.38 (1.31-1.46) | <.001 | ||
| 2-4 | 48 216 | - | - | ||
| Primary payor | Medicare | 79 264 | 1.16 (1.07-1.26) | <.001 | <.001 |
| Private | 42 183 | 1.03 (0.94-1.11) | .559 | ||
| Other government/Not Insured/Unknown | 12 243 | - | - | ||
| Urban/Rural 2013 | Unknown | 2792 | 1.01 (0.87-1.18) | .854 | <.001 |
| Rural | 2296 | 1.35 (1.16-1.56) | <.001 | ||
| Urban | 17 844 | 1.06 (1.00-1.13) | .070 | ||
| Metro | 110 758 | - | - | ||
| Great circle distance (miles) | Intermediate | 41 844 | 0.84 (0.80-0.88) | <.001 | <.001 |
| Long | 7447 | 0.58 (0.51-0.65) | <.001 | ||
| Short | 84 399 | - | - | ||
| Age at diagnosis | 133 690 | 1.02 (1.02-1.02) | <.001 | <.001 | |
Abbreviations: CI, confidence interval; OR, odds ratio; SBRT, stereotactic body radiation therapy; IMRT, intensity-modulated radiation therapy; NH, Non-Hispanic. Bold values represent statistical significance with p<0.05.
*Number of observations in the original data set = 133 898. Number of observations used = 133 690.
**Backward selection with an alpha level of removal of 0.05 was used. The following variables were removed from the model: percent no high school degree 2008-2012, PSA, and chemotherapy.
Comparison of radiation cohorts after inverse probability of treatment weighting
To further adjust for all observed confounding and achieve covariate balance among treatment cohorts, IPTW was performed with sufficient covariate balance (Figure S1). The unweighted and weighted models yielded similar aOR estimates, but the weighted models produced larger CIs due to the smaller effective sample sizes after PS-matched weighting (Table 3, Table 4). As shown in Table 4, after adjustment, the risk of SPC among patients receiving PBT and SBRT treatment was comparable (aOR: 1.16, 95% CI: 0.80-1.70, P = .436). Proton beam therapy was associated with lower SPC risk compared to IMRT (aOR: 0.59, 95% CI: 0.43-0.82, P = .002). Similarly, SBRT was associated with lower SPC risk compared to IMRT (aOR: 0.51, 95% CI: 0.36-0.72, P < .001).
Table 4.
Adjusted OR and 95% CI for pairwise comparisons of treatment cohorts after inverse probability of treatment weighting.
| Cohort (treatment vs reference group) | Adjusted OR (95% CI) | P-value |
|---|---|---|
| PBT versus SBRT | 1.16 (0.80-1.70) | .436 |
| PBT versus IMRT | 0.59 (0.43-0.82) | .002 |
| SBRT versus IMRT | 0.51 (0.36-0.72) | <.001 |
Abbreviations: OR, odds ratio; CI, confidence interval; PBT, proton beam therapy; SBRT, stereotactic body radiation therapy; IMRT, intensity-modulated radiation therapy.
Travel distance and follow-up time subgroup analysis
In Figure S2, we revealed patterns in the association of SPC risk relative to IMRT for each radiation modality across subgroups defined by travel distance to the treatment facility and length of follow-up time. The assessment of interaction terms in the multivariable logistic regression model allowed us to examine a heterogeneous association between RT cohort and SPC across those subgroups, likely reflecting differences in reporting practices that may bias results. For example, PBT and SBRT had reduced risk overall compared to IMRT, but the magnitude of risk reduction decreased among subpopulations residing closer to the treating facility (Figure S2a, interaction P-value < .001) and did not differ significantly by follow-up time (Figure S2b, interaction P-value = .670).
Based on these results and the need for long follow-up to adequately assess SPC incidence, we performed a subgroup analysis among patients traveling <12.5 miles for treatment who were also followed for at least 10 years (N = 18 449). Among this subgroup, 1807 (9.8%) had SPC. The estimated SPC risk associated with SBRT within this subgroup (aOR: 0.56, 95% CI: 0.38-0.83, P = .004) was similar to that observed in the whole population (Tables 3 and S1). The SPC risk associated with PBT was lower than IMRT among this patient subgroup (aOR: 0.80, 95% CI: 0.32-2.01, P = .634), but this difference was not significant due to the small sample sizes, especially within the proton cohort (N = 78) (Table S1).
No-comorbidity subgroup analysis
Because of the limited SPC data available in the NCDB, we ran an additional subgroup analysis among patients without comorbidities (N = 113 273). Within this subgroup, 9036 (8.0%) had SPC. These patients were more likely to have SPC attributable to SPC because they were at lower risk of developing a new primary malignancy due to their lack of comorbid conditions. Within this subgroup, SBRT and PBT were associated with a similarly reduced risk of SPC compared to IMRT (Table S1). Second primary cancer risk estimates also did not differ significantly from those obtained using the entire population (Tables 3 and S1).
Discussion
Using a national cohort of patients with localized PCa, we show that the relative risk of SPC among patients treated with PBT or SBRT is significantly reduced compared to IMRT. The significance of these findings is that, for men with localized PCa undergoing definitive RT, PBT and SBRT treatment may mitigate the risk of SPC, and subsequent morbidity, compared with other commonly available RT options such as IMRT.
Although the retrospective nature of NCDB-related studies comes with the possibility of residual confounding, we are able to estimate the relative probabilities of SPC among RT cohorts with improved rigor by utilizing advanced statistical techniques that minimize selection bias through covariate balancing and by performing subgroup analyses. These analyses provide additional support for our primary analysis, as they indicate a consistent trend in SPC risk within travel distance, follow-up time, and patient comorbidity subgroups.
The highly targeted nature of PBT and SBRT is likely responsible for the observed reduction in SPC risk, since normal surrounding tissues are exposed to less radiation compared to conventional techniques. Specifically, PBT uses high-energy particles with unique physical properties that allow most of the dose to be deposited at a target depth without a large exit dose12, 13; SBRT utilizes advanced immobilization and imaging techniques with steep dose fall-offs that allow high doses to deliver treatment to well-defined target regions in fewer fractions than conventional photon RT.30, 31 Clinical trials to assess the true nature of these benefits are still ongoing.32
Observational studies that utilize large-scale national cancer registry databases provide alternative approaches to assess these potential benefits, since they include many patients from diverse populations with long follow-up times. Compared to other cancer registries, the NCDB provides more robust reporting of radiation modality technique and dosage information and contains a well-defined sequence number variable that is updated annually to capture the development of new primary tumors among existing cases.16 Together, these features make the NCDB a valuable data source for assessing the relationships between SPC and RT techniques, as is supported by its use in previous studies. For example, a similar study by Dee et al17 found that the SPC probabilities associated with SBRT are comparable to those associated with surgery alone, but PBT was not included in their analysis, and they excluded all cases with <5 years of follow-up. Another study by Xiang et al16 reported much lower relative SPC risks among PBT versus IMRT recipients across 9 solid tumors, including PCa, but their analysis did not control for disease-specific confounding effects. Our previous experience also suggests that selection bias may be substantial for newly developed advanced technologies like PBT and SBRT, due to their limited accessibility, so they must be handled carefully.33 By focusing solely on PCa survivors, expanding our analysis to include patients treated with PBT, and utilizing an analytic strategy that reduces confounding effects, we provide improved rigor in estimating the relative SPC probabilities by treatment modality within this population.
Fundamental differences were observed in the patient populations receiving PBT or SBRT compared to the other treatment cohorts. Proton beam therapy patients overwhelmingly received treatment at academic/research institutions, lived in more affluent areas, traveled longer distances to receive treatment, and had less follow-up time on average than other cohorts (Table 2). Similarly, SBRT patients generally lived in more affluent areas and had shorter follow-up times (Table 2). To adjust for the potential confounding effects of these and other clinical and demographic factors, we used multivariable regression and IPTW. Inverse probability of treatment weighting allowed us to create a pseudo-stratified randomized study population with balanced baseline covariates among the 3 treatment cohorts.
There are many factors unique to PBT, not directly measurable in the NCDB or controlled by multivariable modeling or PS adjustment, that may confound our findings. Specifically, access to PBT is much more limited than access to other treatments due to the small number of proton centers in the United States. There were only 3 operating centers in the country at the start of the study period (2004) and only 31 centers by the last year of treatment included in the study (2018).34 This required patients who received PBT to travel farther for treatment and to bear the associated travel costs. As such, financial factors likely play a role. Proton beam therapy is more expensive on average than other types of treatment,10, 35 which may explain why a high proportion of the PBT cohort lived in regions with median incomes in the upper quartile (Table 2). Insurance coverage of PBT also differs by insurance type/plan; a recent study by McDonald et al36 found that patients identifying as Black, Indigenous, or People of Color are significantly less likely to be on proton-favorable insurance plans, which widens the disparities in access to PBT. Patients treated with PBT may also be less likely to return to the treatment facility for follow-up visits due to the far travel distances, which could lead to an underdiagnosis of SPC within the PBT patient cohort, or may go outside of the NCDB system for follow-up visits and treatments, which could prevent cases from being updated appropriately. Travel distance to the treatment facility was included as a model covariate to try to adjust for this effect.
Proton beam therapy and SBRT are also relatively new technologies compared to IMRT, so they are associated with shorter follow-up times. Both PBT and SBRT require highly trained radiation oncologists, physicists, and therapists for treatment delivery, potentially limiting access to these treatments in the early years of the study. As a result, the referral and request rates of physicians and patients, respectively, for PBT or SBRT treatment were likely low toward the beginning of the study period, resulting in small sample sizes and potentially introducing sources of selection bias that contribute to the short follow-up times observed (median: 6.4 years). This is a major limitation of the NCDB and this analysis, contributing to the need for future studies with longer follow-up times to assess whether these trends hold.
For these reasons, the least-biased comparisons of SPC risk by treatment modality likely occur in the subgroup traveling short distances for treatment with long follow-up times, since the influence of related confounders such as the patient ability to travel, ease of access, and insufficient follow-up time is reduced. Therefore, we performed a subgroup analysis among patients with 10+ years of follow-up and travel distances <12.5 miles. Although this subgroup analysis cannot fully account for the limitations discussed above, it helps to provide a more accurate picture of relative SPC incidence. It also supports the results of the full-population analysis, since SBRT and PBT were associated with a reduced chance of SPC compared to IMRT in this subgroup, and the estimated probabilities were similar in magnitude to those estimated using the whole population (Tables 3 and S1). The power of this analysis was limited by the small number of individuals treated with PBT and traveling short distances with long follow-up times, but the direction of SPC risk relative to IMRT was the same, which validates our primary analysis.
Another major limitation of the NCDB is that comprehensive details related to SPC, such as the date of secondary primary malignancies, type/location of SPC, and attribution to RT, are not recorded in the database. This meant that cases with SPC occurrence at sites distant from RT could not be excluded, reported SPCs could not be verified as true secondary malignancies associated with RT, and time-to-event analyses could not be performed. These drawbacks are highlighted in the editorial by Blanchard et al37 in response to Dee et al,17 and they apply to our analysis as well. However, for our primary analysis, we deemed it reasonable to assume that the incidence of SPC unrelated to RT is the same across cohorts based on previous literature.5, 7, 9, 38
We performed another subgroup analysis among patients without comorbid conditions to try to account for these limitations. Patients without comorbidities are expected to be at lower risk of developing a new primary malignancy and, therefore, are more likely to have SPC attributable to RT treatment. Just as in the previous subgroup analysis, SBRT and PBT were associated with a reduced likelihood of SPC compared to IMRT and the relative risk estimates were similar to those obtained using the entire study population (Tables 3 and S1).
Despite the limitations associated with the retrospective design and NCDB, the large sample size, real-world setting, well-rounded analytic strategy, and consistent findings within our primary and subgroup analyses make this study worthy of further validation using other data sources (eg, Surveillance, Epidemiology, and End Results, Veterans Affairs database, clinical trials, or large provincial registries of single-payer health care systems).
Conclusion
In this large hospital-based data set of US men with PCa who received definitive RT, the probability of SPC occurrence was lower among patients receiving PBT and SBRT compared to IMRT. Further studies, including matured results of ongoing prospective clinical trials, will help to elucidate the true benefits of these advanced technologies and provide more definitive evidence about the relationship between radiation exposure and SPC development across treatment modalities.
Acknowledgments
Author Contributions
Sarah Kulkarni: Conceptualization, Writing- Original draft, Formal analysis, Methodology, Software, Visualization. Sagar Patel: Conceptualization, Methodology, Writing- Original draft, Writing- Review and Editing. Yuxian Sun: Conceptualization, Formal analysis, Software. Ashesh Jani, Theresa Gillespie, Mark McDonald: Writing- Review and Editing. Yuan Liu: Conceptualization, Formal analysis, Funding acquisition, Methodology, Writing- original draft, Writing- Review and Editing, Software.
Declaration of Conflicts of Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Funding and Support
The research reported in this publication was supported in part by the Biostatistics Shared Resource of Winship Cancer Institute of Emory University and NIH/NCI under award number P30CA138292. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. The data used in the study are derived from a deidentified National Cancer Database file. The American College of Surgeons and the Commission on Cancer have not verified and are not responsible for the analytic or statistical methodology employed or the conclusions drawn from these data by the investigator.
Footnotes
Supplementary material associated with this article can be found in the online version at doi:10.1016/j.ijpt.2024.100627.
Supplementary material
Supplementary material
.
Data Availability Statement
National Cancer Database 2019 Participant User File for prostate cancer is available and can be requested through the link here: https://www.facs.org/quality-programs/cancer-programs/national-cancer-database/puf/.
References
- 1.Donin N., Filson C., Drakaki A., et al. Risk of second primary malignancies among cancer survivors in the United States, 1992 through 2008. Cancer. 2016;122(19):3075–3086. doi: 10.1002/cncr.30164. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Moon K., Stukenborg G.J., Keim J., Theodorescu D. Cancer incidence after localized therapy for prostate cancer. Cancer. 2006;107(5):991–998. doi: 10.1002/cncr.22083. [DOI] [PubMed] [Google Scholar]
- 3.Schneider U., Lomax A., Besserer J., Pemler P., Lombriser N., Kaser-Hotz B. The impact of dose escalation on secondary cancer risk after radiotherapy of prostate cancer. Int J Radiat Oncol Biol Phys. 2007;68(3):892–897. doi: 10.1016/j.ijrobp.2007.02.029. [DOI] [PubMed] [Google Scholar]
- 4.Bagshaw H.P., Arnow K.D., Trickey A.W., Leppert J.T., Wren S.M., Morris A.M. Assessment of second primary cancer risk among men receiving primary radiotherapy vs surgery for the treatment of prostate cancer. JAMA Netw Open. 2022;5(7) doi: 10.1001/jamanetworkopen.2022.23025. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Wallis C.J.D., Mahar A.L., Choo R., et al. Second malignancies after radiotherapy for prostate cancer: systematic review and meta-analysis. BMJ. 2016;352:i851. doi: 10.1136/bmj.i851. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Siegel R.L., Miller K.D., Fuchs H.E., Jemal A. Cancer statistics, 2022. CA Cancer J Clin. 2022;72(1):7–33. doi: 10.3322/caac.21708. [DOI] [PubMed] [Google Scholar]
- 7.de Gonzalez A.B., Curtis R.E., Kry S.F., et al. Proportion of second cancers attributable to radiotherapy treatment in adults: a cohort study in the US SEER cancer registries. Lancet Oncol. 2011;12(4):353–360. doi: 10.1016/S1470-2045(11)70061-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Davis E.J., Beebe-Dimmer J.L., Yee C.L., Cooney K.A. Risk of second primary tumors in men diagnosed with prostate cancer: a population-based cohort study. Cancer. 2014;120(17):2735–2741. doi: 10.1002/cncr.28769. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Zhao S., Xie Q., Yang R., et al. High prevalence of secondary bladder cancer in men on radiotherapy for prostate cancer: evidence from a meta-analysis. Cancer Manag Res. 2019;11:587–598. doi: 10.2147/cmar.S185867. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Amini A., Raben D., Crawford E.D., et al. Patient characterization and usage trends of proton beam therapy for localized prostate cancer in the United States: a study of the National Cancer Database. Urol Oncol. 2017;35(6):438–446. doi: 10.1016/j.urolonc.2017.01.013. [DOI] [PubMed] [Google Scholar]
- 11.Henderson D.R., Tree A.C., van As N.J. Stereotactic body radiotherapy for prostate cancer. Clin Oncol (R Coll Radiol) 2015;27(5):270–279. doi: 10.1016/j.clon.2015.01.011. [DOI] [PubMed] [Google Scholar]
- 12.Efstathiou J.A., Gray P.J., Zietman A.L. Proton beam therapy and localised prostate cancer: current status and controversies. Br J Cancer. 2013;108(6):1225–1230. doi: 10.1038/bjc.2013.100. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Bryant C.M., Henderson R.H., Nichols R.C., et al. Consensus statement on proton therapy for prostate cancer. Int J Part Ther. 2021;8(2):1–16. doi: 10.14338/ijpt-20-00031.1. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Fang P., Mick R., Deville C., et al. A case-matched study of toxicity outcomes after proton therapy and intensity-modulated radiation therapy for prostate cancer. Cancer. 2015;121(7):1118–1127. doi: 10.1002/cncr.29148. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Yu J.B., Soulos P.R., Herrin J., et al. Proton versus intensity-modulated radiotherapy for prostate cancer: patterns of care and early toxicity. J Natl Cancer Inst. 2012;105(1):25–32. doi: 10.1093/jnci/djs463. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Xiang M., Chang D.T., Pollom E.L. Second cancer risk after primary cancer treatment with three-dimensional conformal, intensity-modulated, or proton beam radiation therapy. Cancer. 2020;126(15):3560–3568. doi: 10.1002/cncr.32938. [DOI] [PubMed] [Google Scholar]
- 17.Dee E.C., Muralidhar V., King M.T., et al. Second malignancy probabilities in prostate cancer patients treated with SBRT and other contemporary radiation techniques. Radiother Oncol. 2021;161:241–250. doi: 10.1016/j.radonc.2021.06.023. [DOI] [PubMed] [Google Scholar]
- 18.Boffa D.J., Rosen J.E., Mallin K., et al. Using the National Cancer Database for outcomes research: a review. JAMA Oncol. 2017;3(12):1722–1728. doi: 10.1001/jamaoncol.2016.6905. [DOI] [PubMed] [Google Scholar]
- 19.Mallin K., Browner A., Palis B., et al. Incident cases captured in the National Cancer Database compared with those in U.S. population based central cancer registries in 2012-2014. Ann Surg Oncol. 2019;26(6):1604–1612. doi: 10.1245/s10434-019-07213-1. [DOI] [PubMed] [Google Scholar]
- 20.Vetterlein M.W., Löppenberg B., Karabon P., et al. Impact of travel distance to the treatment facility on overall mortality in US patients with prostate cancer. Cancer. 2017;123(17):3241–3252. doi: 10.1002/cncr.30744. [DOI] [PubMed] [Google Scholar]
- 21.Onega T., Duell E.J., Shi X., Wang D., Demidenko E., Goodman D. Geographic access to cancer care in the U.S. Cancer. 2008;112(4):909–918. doi: 10.1002/cncr.23229. [DOI] [PubMed] [Google Scholar]
- 22.Massarweh N.N., Chiang Y.-J., Xing Y., et al. Association between travel distance and metastatic disease at diagnosis among patients with colon cancer. J Clin Oncol. 2014;32(9):942–948. doi: 10.1200/jco.2013.52.3845. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.American College of Surgeons. National Cancer Database Participant User File: 2019 Data Dictionary. Updated November 2021. Accessed May 5, 2023. 〈https://www.facs.org/media/aq3aummh/puf_data_dictionary_2019.pdf〉.
- 24.Chesnaye N.C., Stel V.S., Tripepi G., et al. An introduction to inverse probability of treatment weighting in observational research. Clin Kidney J. 2022;15(1):14–20. doi: 10.1093/ckj/sfab158. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Austin P.C., Stuart E.A. Moving towards best practice when using inverse probability of treatment weighting (IPTW) using the propensity score to estimate causal treatment effects in observational studies. Stat Med. 2015;34(28):3661–3679. doi: 10.1002/sim.6607. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Rosenbaum P.R., Rubin DB The central role of the propensity score in observational studies for causal effects. Biometrika. 1983;70(1):41–55. doi: 10.1093/biomet/70.1.41. [DOI] [Google Scholar]
- 27.Li F., Morgan K.L., Zaslavsky A.M. Balancing covariates via propensity score weighting. J Am Stat Assoc. 2018;113(521):390–400. doi: 10.1080/01621459.2016.1260466. [DOI] [Google Scholar]
- 28.Li L., Greene T. A weighting analogue to pair matching in propensity score analysis. Int J Biostat. 2013;9(2):215–234. doi: 10.1515/ijb-2012-0030. [DOI] [PubMed] [Google Scholar]
- 29.Liu Y., Nickleach D.C., Zhang C., Switchenko J.M., Kowalski J. Carrying out streamlined routine data analyses with reports for observational studies: introduction to a series of generic SAS (®) macros. F1000Res. 2018;7:1955. doi: 10.12688/f1000research.16866.2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Lo S.S., Fakiris A.J., Chang E.L., et al. Stereotactic body radiation therapy: a novel treatment modality. Nat Rev Clin Oncol. 2010;7(1):44–54. doi: 10.1038/nrclinonc.2009.188. [DOI] [PubMed] [Google Scholar]
- 31.Halpern J.A., Sedrakyan A., Hsu W.-C., et al. Use, complications, and costs of stereotactic body radiotherapy for localized prostate cancer. Cancer. 2016;122(16):2496–2504. doi: 10.1002/cncr.30101. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Massachusetts General Hospital. Proton Therapy vs. IMRT for Low or Intermediate Risk Prostate Cancer (PARTIQoL). ClinicalTrials.gov identifier: NCT01617161. Updated October 25, 2023. Accessed November 5, 2023. 〈https://clinicaltrials.gov/study/NCT01617161〉.
- 33.Liu Y., Patel S.A., Jani A.B., et al. Overall survival after treatment of localized prostate cancer with proton beam therapy, external-beam photon therapy, or brachytherapy. Clin Genitourin Cancer. 2021;19(3):255–266.e7. doi: 10.1016/j.clgc.2020.08.009. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.The National Association for Proton Therapy. Proton Therapy Centers in the U.S. Updated 2023. Accessed May 5, 2023. 〈https://www.proton-therapy.org/map/〉.
- 35.Konski A., Speier W., Hanlon A., Beck J.R., Pollack A. Is proton beam therapy cost effective in the treatment of adenocarcinoma of the prostate? J Clin Oncol. 2007;25(24):3603–3608. doi: 10.1200/JCO.2006.09.0811. [DOI] [PubMed] [Google Scholar]
- 36.McDonald M.W., Bates J.E., McCall N.S., et al. Insurance authorization and access to proton therapy for patients with head and neck cancers. Int J Radiat Oncol Biol Phys. 2023;116(2):404–412. doi: 10.1016/j.ijrobp.2023.02.033. [DOI] [PubMed] [Google Scholar]
- 37.Blanchard P., Zelefsky M.J., Bossi A., Chargari C., Cosset J.-M. Second malignancy (SM) in prostate cancer patients treated with SBRT and other contemporary radiation techniques. Radiother Oncol. 2021;164:251–252. doi: 10.1016/j.radonc.2021.10.001. [DOI] [PubMed] [Google Scholar]
- 38.de Gonzalez A.B., Wong J., Kleinerman R., Kim C., Morton L., Bekelman J.E. Risk of second cancers according to radiation therapy technique and modality in prostate cancer survivors. Int J Radiat Oncol Biol Phys. 2015;91(2):295–302. doi: 10.1016/j.ijrobp.2014.10.040. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Supplementary material
Data Availability Statement
National Cancer Database 2019 Participant User File for prostate cancer is available and can be requested through the link here: https://www.facs.org/quality-programs/cancer-programs/national-cancer-database/puf/.