EURURO Vol. 71 No. 5

of prostate cancer–specific mortality (PCSM). Since its

introduction, however, the original grading system has

undergone multiple revisions. Gleason patterns 1 and 2 are

rarely assigned, and glands containing cribriform growth

patterns are now classified as Gleason pattern 4. Pierorazio

et al

[2]

proposed a new five-tiered Gleason grade grouping

(GGG) in which Gleason scores 6 comprise GGG1; Gleason

score 3 + 4, GGG2; Gleason score 4 + 3, GGG3; Gleason score

8, GGG4; and Gleason scores 9 and 10, GGG5. Validation

studies have shown that the new system predicts the risk of

biochemical recurrence–free survival and recurrence-free

progression

[2–4]

. To date, however, no studies have

assessed whether the new system predicts PCSM. This

population-based study was undertaken to assess whether

the five-tiered GGG system predicts PCSM using Surveil-

lance, Epidemiology, and End Results (SEER) data.

We identified men diagnosed with prostate cancer

between January 1, 2006 and December 31, 2012

(

= 408 396) from the SEER database, as major Gleason

scoring revisions were adopted at the 2005 International

Society of Urological Pathology Consensus Conference

[5]

. Patients who did not have both primary and secondary

patterns recorded (

= 45 101), patients diagnosed with

prostate cancer at the time of death (

= 31 852) and

patients with missing survival data (

= 55) or of pediatric

age (

= 68) were excluded. The final study cohort totaled

331 320 patients. PCSM data were available up to December

2012.

For patients undergoing radical prostatectomy (RP), the

Gleason pattern was based on a surgical specimen.

Otherwise, the biopsy-based Gleason pattern was used.

We analyzed survival data by cancer treatment received

within 1 yr of cancer diagnosis and by clinical stage (T1,T2,

and T3/4). If a patient receivedmore than one treatment, the

treatment group was assigned using the following hierar-

chy: RP, radiotherapy, other treatments. Patients may have

received subsequent therapy, such as salvage radiation or

prostatectomy, and probably received hormonal therapy

before they died from prostate cancer.

Multivariate competing-risks hazard ratios (HRs) and

confidence intervals (CIs), as well as cumulative incidence

estimates, were computed using the Fine and Gray propor-

tional hazards (PH) model for subdistributions

[6]

. GGG1

(Gleason score 6) was chosen as the reference group. The

multivariate competing-risks PH model accounted for age,

race, marital status, geographic region, diagnosis period

(2006–2009 or 2010–2012), regional lymph node involve-

ment, and metastasis status. Data for prostate specific

antigen (PSA) values at diagnosis were unavailable

[7]

All analyses except for competing risks were performed

using SAS version 9.2 (SAS Institute Inc., Cary, NC, USA). Fine

and Gray competing risks models were fitted using the

stcreg

procedure in STATA version 14 (StataCorp LP, College

Station, TX, USA).

Reclassification of the study cohort according to the

proposed five-tiered GGG system resulted in the following

distribution: 43.3% for GGG1, 29.5% for GGG2, 11.7% for

GGG3, 8.1% for GGG4, and 7.5% for GGG5. Most patients

received either RP (36.4%) or radiation therapy (32.6%) as

primary therapy. During the follow-up period, 9198 of

patients died from prostate cancer and 19 482 died from

other causes. The median follow-up was 38 mo (inter-

quartile range 18–60). The number of patients followed for

at least 5 yr was 40 238 for GGG1, 23 331 for GGG2, 8423 for

GGG3, 5784 for GGG4, and 3803 for GGG5. Characteristics

for the 331 320 men are summarized in Supplementary

Table 1.

Using GGG1 as the reference group, the PCSM hazard

ratio increased with each GGG increment, regardless of the

primary treatment provided. For example, men who

underwent RP the adjusted hazard ratio was 1.13 (95% CI

0.83–1.54) for GGG2, 1.87 (95% CI 1.33–2.65) for GGG3, 5.03

(95% CI 3.59–7.06) for GGG4, and 10.92 (95% CI 8.03–14.84)

for GGG5. The HR approximately doubled for each GGG

increment and differed significantly from each of the others.

Similar patterns were observed for the radiotherapy and

other treatment groups, and for models by clinical stage

( Table 1 )

. As variability in Gleason pattern assessment may

Table 1 – Results of competing-risks regression for prostate cancer–specific survival by Gleason grade grouping (GGG

) d

GGG

By treatmen

t a

By clinical stag

e b

Other treatment

s c

Radical prostatectomy

Radiotherapy

Stage T1

Stage T2

Stage T3/4

HR (95% CI)

GGG1 Reference

Reference

GGG2 2.48 (2.18–2.83

) *

1.13 (0.83–1.54)

1.78 (1.51–2.09

) *

1.92 (1.67–2.21

) *

1.90 (1.64–2.21

) *

1.27 (0.82–1.96)

GGG3 4.74 (4.16–5.40

) *

1.87 (1.33–2.65

) *

2.88 (2.42–3.43

) *

3.46 (2.99–4.02

) *

3.24 (2.78–3.78

) *

2.36 (1.55–3.62

) *

GGG4 7.94 (7.03–8.97

) *

5.03 (3.59–7.06

) *

4.98 (4.24–5.84

) *

5.78 (5.04–6.63

) *

6.19 (5.37–7.13

) *

3.78 (2.49–5.73

) *

GGG5 14.22 (12.61–16.04

) *

10.92 (8.03–14.84

) *

9.88 (8.38–11.63

) *

12.61 (10.99–14.47

) *

10.61 (9.22–12.22

) *

6.96 (4.64–10.46

) *

HR = hazard ratio; CI = confidence interval.

0.05.

Treatment models were adjusted by age, race, marital status, region, clinical stage (T1 vs T2 vs T3/T4), diagnosis period (2006–2009 vs 2010–2012), regional

lymph node metastasis (N0 vs others), and distant metastasis status (M0 vs others).

Clinical stage models were adjusted by age, race, marital status, region, treatment (RP, radiotherapy, or other treatments), diagnosis period (2006–2009 vs

2010–2012), regional lymph node metastasis (N0 vs others), and distant metastasis status (M0 vs others).

Other treatments included watchful waiting, active surveillance, or androgen deprivation therapy.

For patients with RP the Gleason pattern was based on a surgical specimen. Otherwise, biopsy-based Gleason pattern was used.

E U R O P E A N U R O L O G Y 7 1 ( 2 0 1 7 ) 7 6 0 – 7 6 3

761