EURURO Vol. 71 No. 4

513 692

sensitivity analysis including men regardless of estimated
life expectancy but satisfying all other inclusion criteria of
PIVOT. A detailed description of cohort selection, life
expectancy calculation, and sensitivity analysis is provided
in Supplement 1.
Following exclusions, 294 109 men (83%) underwent RP,
whereas 61 257 (17%) underwent observation within the
NCDB between 2004 and 2012 ). The mean age of the
NCDB cohort was 60.3 yr (vs 67.0 yr in PIVOT), and 93% of
NCDB men had a CCI of 0 (vs 56% in PIVOT; both p < 0.001)
. Overall, 42% of men allocated to the RP arm
harbored D’Amico low-risk PCa in PIVOT; nationally, 32% of
patients treated with RP within the NCDB had low-risk
disease (data not shown). Similar results were noted in the
sensitivity analysis (Supplement 1, Supplementary ).
The NCDB contains approximately 29 million records
from >1500 facilities accredited by the American College of
Surgeons Commission on Cancer (CoC), representing nearly
70% of all incident cancer diagnoses in the United States[3_TD$DIFF].
[4_TD$DIFF]Therefore, in all likelihood, it would provide a highly
accurate depiction of cancer care in the United States. Our
analysis found that men enrolled in PIVOT were significantly
older and had more comorbidities than their NCDB counter-
parts. Notably, level 1 evidence also suggests that older men
with limited life expectancy are less likely to benefit from RP
; the PIVOT study may have been underpowered to detect
survival differences in younger and healthier men. Further-
more, nearly 42% of men allocated to the RP arm of PIVOT
had low-risk PCa compared with 32% nationally within the
NCDB. Although it is plausible to expect a difference in
treatment selection when comparing a randomized trial
with registry data, it nonetheless highlights the differences
in treatment selection for contemporary US men diagnosed
with PCa. Finally, PIVOT was conducted predominantly
across multiple US Department of Veterans Affairs (VA)
hospitals. Although the PIVOT investigators noted that men
enrolled in the trial had baseline characteristics similar to
those who declined participation over the study period
(‘‘early’’ PSA testing era from 1994 to 2002) a recent
study based on an analytic cohort of 35 954 VA men
diagnosed with PCa during the same period as PIVOT found
significantly better overall survival for men in their cohort
than those in PIVOT , raising questions about the external
validity of PIVOT for other surgical VA cohorts. However, this
study did not capture contemporary treatment selection,
and its conclusions may be limited to only the VA
population. Notably, VA patients are generally more likely
to have multiple health issues compared with the general US
population, predisposing them to greater overall mortality
. Although an inclusion criterion for PIVOT was
physician-estimated life expectancy >10 yr, the 10-yr all-
cause mortality rate was nearly 35–40% for men in PIVOT
In contrast, as our results demonstrate, contemporary
American men diagnosed with PCa were healthier, and
Huland and Graefen alluded to similar findings in a
contemporary cohort of European men undergoing RP
(whose 15-yr other-cause mortality was approximately
15%) Notably, although the NCDB records data from
only CoC-accredited facilities within the United States,
patients within the NCDB have been shown to be compara-
ble to those in Surveillance, Epidemiology and End Results
(SEER; a population-based tumor registry) and the general
US population in terms of age, stage of disease at diagnosis,
and socioeconomic status .
Important limitations of our analyses include the
inability (1) to differentiate watchful waiting from active
surveillance within the observation cohort (the variable for
active surveillance was added in NCDB for patients
diagnosed from 2010 onward) and (2) to identify men
Table 1 – Comparison of baseline tumor-specific characteristics of
355 366 men with clinically localized cT1–2NxM0 prostate cancer
treated with radical prostatectomy or observation within the
National Cancer Database (2004–2012) with those in the Prostate
Cancer Intervention Versus Observation Trial
Characteristic NCDB
(n = 355 366)
PIVOT
(n = 731)
p value
Age, yr, n (%)
40–49 22 374 (6.3) 4 (0.6) <0.001
50–59 138 419 (39) 71 (9.7)
60–69 161 234 (45) 414 (57)
70–75 33 339 (9.4) 242 (33)
Race, n (%)
Non-Hispanic black 42 594 (12) 232 (32) <0.001
Non-Hispanic white 283 057 (80) 452 (62)
Other 29 715 (8.4) 47 (6.4)
CCI, n (%)
0 330 226 (93) 410 (56) <0.001
1 22 584 (6.4) 211 (29)
>1 2556 (0.7) 107 (15)
PSA, ng/ml, n (%)
<4.0 63 473 (18) 82 (11) <0.001
4.0–10.0 213 158 (60) 397 (54)
10.1–19.9 29 449 (8.3) 176 (24)
20.0–49.9 11 080 (3.1) 75 (10)
Unknown 38 206 (11) 1 (0.1)
Clinical T stage, n (%)
cT1NOS 1539 (0.4) 0 (0.0) <0.001
cT1a 1391 (0.4) 15 (2.1)
cT1b 962 (0.3) 14 (1.9)
cT1c 256 323 (72) 368 (50)
cT2a 32 340 (9.1) 181 (25)
cT2b 9964 (2.8) 91 (12)
cT2c 52 847 (15) 57 (7.8)
Unknown/other NA 5 (0.7)
Biopsy Gleason score, n (%)
<4 1450 (0.4) 157 (22) <0.001
5–6 167 416 (47) 358 (49)
7 148 572 (42) 133 (18)
8–10 20 961 (7.6) 51 (7.0)
Unknown 10 967 (3.1) 32 (4.4)
Histologic grade, n (%)
Well differentiated 1450 (0.4) 157 (22) <0.001
Moderately well differentiated 315 988 (89) 508 (70)
Poorly differentiated 26 961 (7.6) 51 (7.0)
Unknown 10 967 (3.1) 15 (2.0)
D’Amico risk category, n (%)
Low risk 118 355 (33) 296 (40) 0.001
Intermediate risk 132 609 (37) 249 (34)
High risk 82 643 (23) 157 (22)
Unknown 21 759 (6.1) 29 (3.9)
CCI = Charlson-Deyo comorbidity index; NA = not assessed; NCDB = National
Cancer Database; NOS = not otherwise specified; PIVOT = Prostate Cancer
Intervention Versus Observation Trial; PSA = prostate-specific antigen.
Men within the NCDB were selected to approximate the inclusion criteria of
PIVOT. Percentages may not add up to 100% because of rounding of decimal
places.
E U R O P E A N U R O L O G Y 7 1 ( 2 0 1 7 ) 5 1 1 – 5 1 4 513