By: Nick Turner, PhDA largely glossed-over finding from active-controlled trials of checkpoint inhibitors in solid tumors is the delayed crossover of Kaplan-Meier (K-M) survival curves, consistent with increased risk of disease progression and mortality early in checkpoint inhibitor immunotherapy (CPI). This raises the heretical proposition that CPI monotherapy may have a detrimental effect on progression and survival, early in treatment, relative to chemotherapy, in a significant number of patients.

An analysis across multiple solid tumors suggests this is not an uncommon finding. In an examination of data from 29 randomized controlled CPI trials, Winquist et al.1 found that CPI K-M survival curves violated proportional hazard assumptions more frequently than other anti-cancer monoclonal antibodies. This was characterized by early negative crossover of K-M survival curves opposing an overall beneficial survival trend or by delayed separation, an effect that can be large enough to negate any long-term survival advantage of CPI. This is a pattern that is evident across K-M curves from controlled trials of checkpoint inhibitors in several solid tumors, including gastric2, non-small-cell lung cancer (NSCLC)3, renal (favorable risk)4, head and neck5 and bladder6.

This raises important questions regarding the role of CPI monotherapy, particularly in the first-line setting (1L), whether low PD-L1 expression characterizes the at-risk patient, and whether all patients should be treated with CPI-chemotherapy combinations regardless of PD-L1 expression levels, to mitigate this potential early survival liability.

Is PD-1/L1 inhibitor monotherapy detrimental in the low PD-L1 patient?

The divergence of early liability in some patients and sustained benefit in others has particular resonance given the recent warnings from the FDA and EMA against using single-agent checkpoint inhibitors as front-line therapy in urothelial bladder cancer patients who have low PD-L1 expression.

The safety alert was prompted by early data monitoring committee reviews of two ongoing trials (KEYNOTE-361 and IMvigor-130) that identified decreased survival in patients treated with pembrolizumab or atezolizumab monotherapies, compared to patients who received chemotherapy. This raises clear parallels with the negative crossover of K-M survival curves seen in several Phase III trials that point to a potential early detrimental effect of CPI relative to chemotherapy in solid tumors. Taken together, this suggests the early risk identified in PD-L1-low expressers may not be confined to bladder cancer. The mechanism behind this is unclear but could be due to latency of the anti-tumor immune response, particularly in low immunogenic and non-T-cell-inflamed tumors.

The clearest evidence this early survival liability may stretch beyond bladder cancer comes from recently reported studies of CPI monotherapy and combination trials in NSCLC and takes on greater significance as Merck & Co. (MRK) seeks to extend the label for pembrolizumab (Keytruda) in 1L NSCLC to patients with ≥1% PD-L1 expression.

Pembrolizumab was approved as monotherapy in 1L NSCLC based on the results of the KEYNOTE-024 trial7. This reported a 37% reduction in risk of death (HR 0.63, mOS 30.2 months vs. 14.2 months; WCLC 2017) and a 50% reduction in risk of disease progression in patients on pembrolizumab relative to platinum-based chemotherapy. K-M survival curves separated early, implying a rapid onset and sustained benefit of pembrolizumab. All patients in KEYNOTE-024, however, had high levels of PD-L1 expression (≥50%).

In contrast, the negative CheckMate-026 study, which compared nivolumab with platinum-based chemotherapy in 1L NSCLC patients at a lower PD-L1 cut-off (5% or greater), showed risk of disease progression or death was higher in the nivolumab group (HR 1.17)8. There was no separation of curves on overall survival, but the HR 1.07 favored the chemotherapy arm. Similarly, the CheckMate-057 trial (nivolumab vs. docetaxel, 2L NSCLC) also strongly points to an early survival liability of nivolumab compared to docetaxel, with pronounced negative crossover of the K-M survival curves compared to chemotherapy at lower PD-L1 expression levels (<1%, <5% and <10%)9.

This pattern is also evident in MRK’s KEYNOTE-042 study. Although the trial met its primary endpoint, showing an overall survival benefit of single-agent pembrolizumab in 1L NSCLC in subjects with ≥1% PD-L1, sub-group analysis shows, as in KEYNOTE-024, that the survival benefit was carried by high PD-L1 expressers (PD-L1 ≥50%, HR 0.69, mOS 20 months; ≥1% HR 0.81, mOS 16.7 months vs. chemotherapy 12.2 months)3. Moreover, exploratory analysis indicates pembrolizumab has limited clinical benefit in subjects with low-to-moderate PD-L1 expression (1-49%, OS HR 0.92, 95% CI 0.77-1.11) relative to chemotherapy. Key, however, is that across all PD-L1 expression subgroups (but most notably in the TPS 1-49% cohort), patients in the chemotherapy arm fared better over the first 6 to 9 months of the study, K-M curves showing a pronounced negative crossover on both overall and progression-free survival3.

In our opinion, the evidence supports the hypothesis that compared to chemotherapy, there is a negative effect of CPI monotherapy in patients with moderate-to-low PD-L1 expression early in treatment. This is seen in NSCLC, bladder and gastric cancer and potentially also other solid tumors. Moreover, while the clinical data point to a sub-group of patients that derive a clear overall survival benefit from CPI monotherapy, it is also the case that more than half of patients with low PD-L1 expression may have accelerated disease progression on CPI, compared to patients on chemotherapy, and this may translate to earlier mortality in between 20% and 45%.

Whether negative crossover of the K-M curves is identifying a population of patients that receives no benefit from treatment with checkpoint inhibitors or reflects a delayed onset of immune modulation that is detrimental early in therapy is to be resolved and should be given further attention.

Mitigating the risk with immune checkpoint inhibitors

A pointer comes from the 1L NSCLC KEYNOTE-189 (pembrolizumab + chemotherapy) study. The trial showed a remarkable survival advantage for pembrolizumab in combination with pemetrexed and platinum chemotherapy compared to chemotherapy alone, a benefit that was seen across all levels of PD-L1 expression, including PD-L1 1-49% and PD-L1 <1% patient cohorts10. Moreover, the absence of crossover of K-M survival curves in KEYNOTE-189 suggests chemotherapy coupled with 1L CPI may mitigate any potential early detrimental effect of CPI monotherapy in patients with rapidly progressing disease.
We contend that for patients who can tolerate chemotherapy, CPI plus chemotherapy should be standard practice in the 1L setting to mitigate any early survival liability of monotherapy.


  1. Winquist, E., Kuruvilla, S., Nichols, A.C, et al. Clin Oncol, 2018; 36, suppl; abstr 12121.
  2. KEYNOTE-061, Lancet 2018; 392: 123-133; and ASCO18 abstr 4062.
  3. KEYNOTE-042, ASCO18 abstr LBA4 Clin Oncol 36 suppl.
  4. CheckMate 214, ESMO 2017 LBA4.
  5. CheckMate-141, NEJM 2016; 375: 1856-1867.
  6. KEYNOTE-045, NEJM 2017; 376: 1015-1026.
  7. KEYNOTE-024, NEJM 2016; 375: 1823-1833.
  8. CheckMate-026; NEJM 2017; 376: 2415-2426.
  9. CheckMate-057, NEJM 2015; 373: 1627-1639, supplementary appendix.
  10. KEYNOTE-189 NEJM 2018; DOI: 10.1056/NEJMoa1801005 and AACR18.