EURURO Vol. 72 No. 4

short at 2.1 mo and was not PD-L1 dependent. Median

overall survival (OS) was 11.4 mo (95% CI 9.0–not estimable

[NE]) in the IC 2–3+ PD-L1–positive population, 6.7 mo (95%

CI 5.1–8.8) in the IC 1+ PD-L1–positive population, and

6.5 mo (95% CI 4.4–8.3) in patients who did not express

PD-L1 (IC 0). The drug was well tolerated, with 15% of

patients having grade 3 or 4 adverse events

[3]

A recent study with nivolumab in a similar platinum-

refractory population included 265 patients and achieved

Food and Drug Administration approval in February 2017

[11]

. The overall response rate was 19.6% (15.0–24.9%) and

median survival was 8.7 mo (6.1–NE). PD-L1 positivity was

defined as

1% TC expression, and 46% of patients were

positive

[11] .

A numerically higher response rate of 23.8%

(16.5–32.3%) and longer OS of 11.3 mo (8.7–NE) were

apparent in the PD-L1–positive population. One of the

confusing issues with these data is that the biomarker was

not discriminatory in the phase 1 trial

[6]

. These incon-

sistencies impede biomarker development.

Together, both agents look preferable to chemotherapy

in this setting

[3,11]

. Median OS in unselected patients

appears to be modest, underlining the potential importance

of the biomarker

( Table 1

). The atezolizumab data involve

more robust biomarker discrimination and longer follow-

up data on the durability of remission

[3] .

The modest OS

data for the PD-L1–negative population with nivolumab

suggest potential prognostic value

[11]

. Long-term durable

remissions in this population and the shape of the Kaplan

Meier curves suggest that the median values may be

misleading. Hazard ratios (HRs) in randomised trails are

required.

Randomised phase 3 data in platinum-refractory

A recent randomised phase 3 trial showed that pembroli-

zumab significantly outperformed standard chemotherapy

(paclitaxel, docetaxel, or vinflunine) in platinum-refractory

[4] .

The HR for OS (primary endpoint) was 0.73 (95% CI

0.59–0.91). Biomarker positivity (defined as composite TC

and IC expression) was predictive of a response to therapy

(0.57, 95% CI 0.37–88) and prognostic of a poor outcome.

The response rate was also significantly higher in the

pembrolizumab than in the chemotherapy group (21% vs

11%). The adverse event profile favoured pembrolizumab.

Forest plot analysis revealed that all groups favoured

pembrolizumab, except for the never smoker group

[4]

. These are the most robust data to date and compre-

hensively underline the benefits of ICIs in this setting. The

study sets a standard for future randomised trials in this

setting.

Phase 2 data in cisplatin-ineligible,

chemotherapy-naı¨ve metastatic UC

A large proportion of patients with UC are unable to receive

cisplatin-based therapy for various reasons, such as renal

insufficiency, poor performance status, and comorbidities

including neuropathy and hearing loss. These patients

usually receive carboplatin-based treatment instead, with a

median OS close to 10 mo.

A single-arm study investigating atezolizumab in

119 patients showed response rates of 24% (16–32%) with

median OS of 14.8 mo (10.1–NE)

[12]

. Again, median PFS

was short and does not compare favourably with historical

chemotherapy controls, while the median OS is outstand-

ing. Intriguingly, the biomarker did not predict response or

survival, and it is possible that the modest sample size and

differences in treatment patterns and disease character-

istics were responsible for these differences.

A second study with pembrolizumab presented response

data for the same first-line cisplatin ineligible population

(

= 100)

[13]

. The response rate among all comers was 24%

(16–34%). The biomarker was exploratory and measured IC

and TC PD-L1 expression. Different cutoff points were tested

(1% and 10% positivity). Response rates in the biomarker-

positive population, for the same biomarker as in the

positive phase 3 trial, were numerically higher than in the

negative population (37% vs 18%). Results for the first

100 patients compare favourably with the fully enrolled

Table 1 – Immune checkpoint inhibitors in cisplatin-ineligible and cisplatin-refractory metastatic urothelial carcinoma

Cisplatin-ineligible

Cisplatin-refractory

IMvigor210

cohort I

[13]

Keynote-052

[14]

IMvigor210

cohort II

[4]

CheckMate275

[12]

Keynote-045

[5]

Phase

Agent

Atezolizumab

Pembrolizumab

Atezolizumab

Nivolumab

Pembrolizumab

Chemotherapy

Patients (

)

119

100

310

265

270

272

RR, % (95% CI)

All comers

24 (16–32)

24 (16-34)

15 (11–19)

19.6 (15.0–24.9)

21.1

11.4

PD-L1

28 (14–47)

37 (20–56)

26 (18–36)

23.8 (16.5–32.3)

21.6

6.7

MOS, mo (95% CI)

All comers

14.8 (10.1–NE)

–

7.9 (6.6–9.3)

8.7 (6.05–NE)

10.3 (8.0-11.8)

7.4 (6.1–8.3)

PD-L1

12.3 (6.0–NE)

–

11.4 (9.0–NE)

11.30 (8.74–NE)

8.0 (5.0–12.3)

5.2 (4.0–7.4)

TRAEs grade 3/4 (%)

17.8

13.5

47.8

RR = response rate; CI = confidence interval; MOS = median overall survival; TRAEs = treatment-related adverse events; NE = not estimable.

Different methods were used to measure PD-L1 expression. Direct comparisons are not possible.

E U R O P E A N U R O L O G Y 7 2 ( 2 0 1 7 ) 4 7 7 – 4 8 1

479