Neutrophil-to-lymphocyte ratio in combination with PD-L1 or lactate dehydrogenase as biomarkers for high PD-L1 non-small cell lung cancer treated with first-line pembrolizumab.

The identification of prognostic and predictive biomarkers for high-programmed cell death-ligand 1 (PD-L1) advanced non-small cell lung cancer (aNSCLC) treated with first-line pembrolizumab could support the decision-making about possible combination therapies. To explore the baseline neutrophil-to-lymphocyte ratio (NLR) with the possible addition of PD-L1 tumour proportion score (TPS) level or lactate dehydrogenase (LDH) as possible prognostic biomarkers by a multicenter retrospective exploratory analysis aiming at identifying favourable-risk patients. Baseline NLR was available for all 132 high PD-L1 aNSCLC patients, PD-L1 level and LDH for 81 (61%) and 85 (64%) patients, respectively. NLR, PD-L1 and LDH cut-offs by receiver operating characteristic (ROC) curves were 4.9, 77.5% and 268.5, respectively. Seventy-one patients (54%) had NLR <5; 25 out of 81 NLR <5 (31%) had PD-L1 >80%, 26 out of 85 (31%) NLR <5 and normal LDH (nLDH). Median follow-up was 16.3 months. As compared to NLR >5, significantly better 2-year overall survival (OS) and progression-free survival (PFS) were observed with NLR <5 [62% vs. 41%, P=0.005, hazard ratio (HR) 0.45, and median of 12.0 vs. 5.7 months, P=0.01, HR 0.56, respectively], NLR <5 + PD-L1 >80% (81%, P=0.006, HR 0.20 and median of 14.7, P=0.03, HR 0.44, respectively), and NLR <5 + nLDH (74%, P=0.009, HR 0.25 and median of 14.7, P=0.02, HR 0.40, respectively). NLR <5 and NLR <5 + nLDH significantly associated with PD (P=0.008 and P=0.025, respectively) but not response rate (RR) (P=0.09 and P=0.07, respectively); NLR <5 + PD-L1 >80% both RR (P=0.03) and PD (P=0.02). NLR <5 ± PD-L1 >80% or nLDH could represent easy-to-assess tools to identify high PD-L1 aNSCLC patients with favourable outcome following first-line pembrolizumab monotherapy. 2020 Translational Lung Cancer Research. All rights reserved.

Introduction

The current standard first-line treatment in advanced non-small cell lung cancer (aNSCLC) with programmed cell death-ligand 1 (PD-L1) tumour proportion score (TPS) of at least 50% includes pembrolizumab monotherapy (1), with a 3-year overall survival (OS) rate of 43.7% as compared to 24.9% with chemotherapy, supported by durable disease responses (2). However, pembrolizumab monotherapy is being challenged by the addition of chemotherapy (3) or a different immune-checkpoint inhibitor (ICI) (4). Hence, there is a need for reliable biomarkers to predict patients’ outcome and disease response following pembrolizumab monotherapy (5,6).

The neutrophil-to-lymphocyte ratio (NLR) is a surrogate for tumour-associated inflammation and likely represents the frequency and activity of myeloid-derived suppressor cells (MDSCs), that hinder T-cell proliferation and expansion (7). Lactate dehydrogenase (LDH) has been investigated as a potential inflammatory biomarker in patients with cancer and is associated with poor outcomes in several cancer types (8). In aNSCLC, high NLR and its combination with high LDH, namely the immune prognostic index (LIPI), were correlated with worse outcomes for ICIs, but not for chemotherapy (8,9).

We explored the NLR in combination with PD-L1 or LDH as possible biomarkers for high PD-L1 aNSCLC with first-line pembrolizumab.

We present the following analysis in accordance with the REMARK Guideline (10) (available at http://dx.doi.org/10.21037/tlcr-19-583).

Patients and methods

The study analysis aimed to explore the prognostic value of low NLR, high PD-L1 TPS level and normal LDH, and of the combination of low NLR with high PD-L1 or normal LDH, in high PD-L1 aNSCLC treated with first-line pembrolizumab treated in five European Centers (three in Italy, one in the UK, one in Switzerland), as possible tools to identify patients with favourable outcome.

Patients aged more than 18 years, with histologically confirmed aNSCLC, Eastern Cooperative Oncology Group (ECOG) Performance Status (PS) ≤2, PD-L1 TPS >50%, treated with first-line pembrolizumab were retrospectively assessed. The analysis followed the principles outlined in the Declaration of Helsinki (as revised in 2013) for all human or animal experimental investigations. The analysis involved a real-world series of patients treated according to the clinical practice; ethical approval was waived because all patients in each Center signed a specific consent form for their data collection and sharing with other Institutions.

Three baseline parameters, from fresh routine blood or histopathological samples, were retrospectively analyzed: (I) the NLR, or the ratio between absolute neutrophils and lymphocytes; (II) the PD-L1 TPS level on the histological sample; (III) the blood level of LDH (in units/litre). The blood for NLR and LDH values was yielded within seven days from the ICI treatment start and routinely analyzed by local laboratories. PD-L1 was assessed by the platform and antibody used in each Center (see ).

Table 1

Patients and disease characteristics (No. 132)

Characteristic	No.	% [range]
Age, median	68	[31–85]
Gender
Male/female	87/45	66/34
Tobacco use
Never/current/former	12/42/75	9/32/57
NK	3	2
Histology
Adenocarcinoma	80	61
Squamous	35	27
Sarcomatoid	7	5
Adenosquamous	3	2
Pleomorphic	2	2
Other	5	4
Initial stage
I/II/III	2/3/14	2/2/11
IV	113	86
Previous treatments
Surgery/NACT/RT	7/7/5	5/5/4
Brain metastases	19	14
CK-GK/ WBRT	4/7	21/37
Pembrolizumab	8	42
Steroids (on pembrolizumab)	27	20
ECOG PS
0/1/2	42/68/22	32/52/17
Autoimmune diseasea	7	5%
PD-L1 expression
≥50% (NOS)	51	39
50–70%/70–80%	32/8	24/6
80–90%/90–100%	23/18	17/14
22C3/SP263/E1L3N	101/22/9	77/17/7
EGFR/ALK/ROS1	3/0/0	4b
Radiotherapy (metastases)	41	31
Response to pembrolizumab
CR/PR/SD/PD/NA	1/55/25/36/15	2/42/19/27/11
Discontinuation for toxicity	14	11
Symptomatic PD	40	30
Second-line	37	28
PR/SD/PD	10/7/6	27/19/16

a, psoriasis in 3 (in 2 with associated arthritis), Crohn’s disease in 2, connectivities and myasthenia gravis each in one; b, of 80 patients with adenocarcinoma. CK, cyberknife; ECOG PS, Eastern Cooperative Oncology Group Performance Status; GK, gammaknife; NK, not known; No. Number; NA, not assessable; NACT, neoadjuvant chemotherapy; No., number; NOS, not otherwise specified; PD, progressive disease; PD-L1, programmed cell death-ligand-1; RT, radiotherapy; WBRT, whole brain radiotherapy.

The cut-offs for NLR, PD-L1 and LDH were identified by receiver operating characteristic (ROC) curves.

For the combined biomarkers analysis patients with low NLR were grouped with those with available and high PD-L1 or available and normal LDH, with a possible overlapping between these two cohorts, or high PD-L1 and normal LDH.

The prognostic role in OS and progression-free survival (PFS) was assessed by the two-sided log-rank test; the association with response rate (RR) and progressive disease (PD) by the 2-tailed Fisher exact test.

RR, including complete response (CR) and partial response (PR), and PD, as the best response to the treatment, were assessed by the Response Evaluation Criteria in Solid Tumours (RECIST) criteria version 1.1 (11).

Statistical significance was investigated by Chi-square test and 2-tailed Fisher exact tests with an acceptable significance value of P<0.05. The OS and PFS were calculated from the date of ICI treatment start until death or last date of follow-up, and of progression or death from any cause, respectively, were estimated using the Kaplan-Meier, reported as medians with confidence limits (95% CI) and compared using two-sided log-rank test, with an acceptable significance value of P<0.05. Patients who did not have events at the time of the analysis have been censored. Considering an expected 2-year OS of 52% in this patient population (2) and aimed at finding a difference in the 2-year OS of 25% with the combined biomarker analysis, the required sample size was 113 patients.

Results

Patients’ characteristics are summarized in . One hundred thirty-two patients treated with pembrolizumab from December 2016 to June 2019 were analyzed. Twenty-two patients (17%) discontinued the treatment, 11 due to toxicity; 6 of them because of pneumonia and one due to worsening of pre-existent myasthenia gravis.

NLR was available for all patients, while PD-L1 level and LDH were available only for 81 (61%) and 85 (64%) patients, respectively.

With a median follow-up of 16.3 months (95% CI: 15.0–17.7), the 2-year OS of all patients was 52.0% (95% CI: 49.3–55.0) and the median PFS 9.7 months (95% CI: 6.9–12.5) (see Figure S1A,B). Both OS and PFS were significantly associated with the disease response, CR/PR (2-year OS 89.3%, median PFS not reached) vs. SD (47.5%, 9.7 months) vs. PD (0%, 3.7 months) (P<0.001 for both, respectively) (see and Figure S1C,D).

Table 2

Prognostic value of NLR, PD-L1, LDH and association with disease response

Parameter	No. [%] (95% CI)	No. [%] (95% CI)	Total No. [%] or HR (P value) [95% CI]	Log-rank, P value	χ2, P value	Fishera, P value
NLR	<5	≥5
CR/PR	36 [55]	20 [38]	56 [48]		0.0686	0.0936
PD	13 [20]	23 [44]	36 [31]		0.0048	0.0083
NA	6 [8]	9 [15]	15 [11]
Total	71 [54]	61 [46]	132
2-year OS, %	62.0 (57.7–66.6)	41.2 (38.1–44.6)	0.45 (0.006) [0.26–0.80]	0.005
Median PFS, month	12.0 (9.0–15.0)	5.7 (3.9–7.6)	0.56 (0.01) [0.35–0.89]	0.012
NLR/PD-L1	<5/≥80	≥5/<80
CR/PR	15 [63]	6 [33]			0.0182	0.0308
PD	3 [13]	11 [61]			0.0114	0.0172
NA	1 [4]	2 [10]
Total	25b [31]c	20 [25]c
2-year OS, %	81.0 [72.9–89.5]	42.9 [37.8–48.5]	0.17 (0.007) [0.05–0.62]	0.002
Median PFS, month	14.7 [11.4–18.0]	4.7 [2.2–7.3]	0.32 (0.01) [0.14–0.77]	0.007
NLR/PD-L1	≥5/≥80	<5/<80
CR/PR	10 [67]	5 [26]			0.7919d	1.0000d
PD	3 [20]	9 [47]			0.5277d	0.6580d
NA	1 [6]	1 [5]
Total	16 [20]c	20 [25]c
2-year OS, %	47.7 [39.9–56.1]	39.7 [34.0–46.0]	3.11 (0.11)d [0.78–12.45]	0.09d
Median PFS, month	19.2 [2.3–40.6]	9.5 [5.5–13.5]	1.18 (0.73)d [0.46–3.01]	0.73d
NLR/LDH	<5/<269	≥5/≥269
CR/PR	14 [54]	5 [24]			0.0370	0.0716
PD	4 [15]	10 [48]			0.0163	0.0250
NA	0 [0]	3 [12.5]
Tot.	26e [31]f	24 [28]f
2-year OS, %	74.3 (65.5–83.5)	36.0 (32.3–40.1)	0.20 (0.004) [0.07–0.61]	0.002
Median PFS, month	14.7 [NA]	5.4 (3.0–7.9)	0.35 (0.007) [0.16–0.75]	0.007
NLR/LDH	≥5/<269	<5/≥269
CR/PR	6 [50]	12 [63]			0.8253g	1.0000g
PD	5 [42]	4 [21]			0.0765g	0.1082g
NA	1 [8]	3 [14]
Total	13 [15]f	22 [26]f
2-year OS, %	65.9 (56.8–75.6)	53.0 (47.1–59.4)	0.43 (0.23)g [0.11–1.71]	0.214g
Median PFS, month	6.4 (5.0–7.8)	10.4 (6.0–14.8)	2.06 (0.15)g [0.78–5.45]	0.139g
Response	CR/PR; SD	PD
2-year OS, %	89.3 (84.6–94.2); 47.5 (41.1–54.4)	0.0 [NA]	0.02 (<0.001) [0.008–0.08]; 0.14 (<0.001) [0.06–0.32]	<0.001
Median PFS, month	NR; 9.7 (5.8–13.6)	3.7 (3.1–4.3)	0.08 (<0.001) [0.05–0.16]; 0.18 (<0.001) [0.09–0.34]	<0.001

a, 2-tailed Fisher test; b, 45 (63%) out of 71 patients with NLR <5 had available PD-L1 scores; 25 patients with NLR <5 and PD-L1 >80% included 16 patients with LDH scores whose 14 (87.5%) had normal LDH (nLDH); c, percentages refer to 81 patients with NLR and PDL1 scores; d, comparison between NLR/PD-L1 <5/≥80 and ≥5/≥80 cohorts; e, 48 (68%) out of 71 patients had available PD-L1 and LDH scores; 26 patients with NLR <5 and nLDH included 23 patients with PD-L1 scores whose 14 (61%) had PD-L1 >80%; f, percentages refer to 85 patients with NLR and LDH scores; g, comparison between NLR/PD-L1 <5/<269 and ≥5/<269 cohorts. CI, confidence interval; CR, complete response; HR, hazard ratio; LDH, lactate dehyidrogenase; NLR, neutrophils-to-lymphocytes ratio; No. Number; OS, overall survival; PD, progressive disease; PD-L1, programmed cell death-ligand-1; PFS, progression-free survival; PR, partial response; SD, stable disease.

NLR, PD-L1 and LDH cut-offs by ROC curves were 4.9, 77.5% and 268.5, respectively (see ).

Figure S2

Receiver operating characteristic (ROC) curves of NLR, PD-L1 and LDH on PD. CI, confidence interval; DCR, disease control rate (= complete response, partial response, stable disease); HR, hazard ratio; LDH, lactate dehydrogenase; NA, not assessable; NLR, neutrophils-to-lymphocytes ratio; NR, not reached; OS, overall survival; mo., months; PD, progressive disease; PD-L1, programmed death ligand 1; PFS, progression-free survival.

Seventy-one patients (54%) had NLR <5, 45 (63%) and 48 (68%) of them had available PD-L1 and LDH scores, respectively. Twenty-five patients out of the 81 with NLR and PD-L1 scores (31%) had NLR <5 and PD-L1 >80% and included 16 patients with LDH scores whose 14 (87.5%) had normal LDH (nLDH); 26 patients out of 85 patients with NLR and LDH scores (31%) had NLR <5 and nLDH and included 23 patients with PD-L1 scores whose 14 (61%) had PD-L1 >80%.

NLR <5, PD-L1 ≥80% and LDH <269 significantly associated with both OS and PFS (see ), with 2-year OS of 62.0% versus 41.2% and median PFS of 12.0 versus 5.7 months, for patients with NLR <5 as compared to those with NLR >5 [P=0.005, hazard ratio (HR) 0.45 and P=0.012, HR 0.56, respectively); 65.7% versus 40.6% and 14.7 versus 6.1 months, for PD-L1 ≥80% versus PD-L1 <80% (P=0.007, HR 0.35, and P=0.006, HR 0.44, respectively); and 72.1% versus 44.3% and 12.9 versus 8.0 months, for LDH <269 versus LDH ≥269 (P=0.026, HR 0.41, and P=0.052, HR 0.56, respectively).

Figure 1

Overall and progression-free survival according to NLR <5, PD-L1 ≥80% and LDH <269.

Data on outcomes by NLR, PD-L1 and LDH cut-offs are reported in and . Better 2-year OS and PFS were seen in patients with low NLR <5 and high PD-L1≥80% as compared to those with high NLR and low PD-L1 (81.0% vs. 42.9%, P=0.002, HR 0.17, and median of 14.7 vs. 4.7 months, P=0.007, HR 0.32, respectively); a trend toward better OS but not PFS was observed in patients with low NLR and high PD-L1 as compared to those with high NLR and high PD-L1 (81.0% vs. 47.7%, P=0.09, HR 3.11, and median of 14.7 vs. 19.2 months, P=0.73, HR 1.18, respectively). Better 2-year OS and PFS were seen in patients with low NLR <5 and nLDH <269 as compared to those with high NLR and high LDH (74.3% vs. 36.0%, P=0.002, HR 0.20, and median of 14.7 vs. 5.4 months, P=0.007, HR 0.35, respectively); a trend toward better OS and PFS was observed in patients with low NLR and nLDH as compared to those with high NLR and low PD-L1 (74.3% vs. 65.9%, P=0.214, HR 0.43, and median of 14.7 vs. 6.4 months, P=0.139, HR 2.06, respectively).

Figure 2

Overall and progression-free survival according to NLR <5 ± PD-L1 ≥80% or LDH <269.

As far as the disease response is concerned, low NLR <5 significantly associated with PD (P=0.008) but not RR (P=0.09); low NLR <5 with high PD-L1 >80% both RR (P=0.03) and PD (P=0.02), low NLR <5 with nLDH with PD but not RR (P=0.025 and P=0.07, respectively).

Discussion

We found that low NLR <5, high PD-L1 >80% and LDH <269, especially when low NLR <5 was combined with one of the other two factors, significantly associated with favourable outcomes following first-line pembrolizumab monotherapy in aNSCLC. This underlines the essential role of tumour inflammation and microenvironment (12).

Differently from other reported series, the cut-offs we used for NLR, PD-L1 and LDH were ROC-based and quantitative. This could raise the issue of their external validity, of the need for their validation or just of comparison with those currently reported. As far as the NLR is concerned, our ROC-based cut-off of 5 mirrored what has already been reported in patients with NSCLC (9) and previously validated in those with metastatic melanoma treated with ipilimumab (13). In the LIPI score for NSCLC patients (8), the NLR cut-off used was instead 3, based on a larger and updated series of patients with metastatic melanoma treated with ipilimumab (14). Regarding LDH, there is no current standard cut-off. The upper limit of normal (ULN) value according to the limit of each center was for example adopted for the LIPI score, where the median LDH value of all patients was 248.5 (8). In the largest series of patients with metastatic melanoma treated with pembrolizumab an LDH value of at least 2.5 times ULN resulted as an independent prognostic value (15), while in another series a quantitative cut-off of 480 U/L was reported (16). No validated cut-off of PD-L1 expression levels over the threshold of 50% has been formally explored. In a series of 187 patients with high PD-L1 NSCLCs treated with first-line pembrolizumab, patients’ clinical outcomes were significantly improved in very high PD-L1 ≥90% positive tumours (17). Thus, our ROC analyses suggested that in patients with NSCLC treated with an ICI, 5 is an appropriate NLR cut-off for patients with aNSCLC, a quantitative LDH cut-off of 269 could be considered as an alternative to the ULN and a lower threshold of 80% PD-L1 NSCLC expression could be considered. Furthermore, our combined analyses differed from the LIPI (8), besides the use of ROC-based and quantitative cut-offs, for the inclusion of PD-L1 expression level, which currently represents the only validated predictive biomarker.

Regarding the association with disease response, only PD-L1 ≥80% was associated with both RR and PD, while NLR <5 and nLDH with the only PD. Intriguingly, although high PD-L1 seemed to be associated with RR independently of NLR, a nonsignificant trend toward better OS in patients with low NLR and high PD-L1 as compared to those with high NLR and high PD-L1 was observed. This suggests further investigations in comparative trials to explore the predictive value of PD-L1 >80% against the prognostic value of NLR and LDH.

One limitation of this study is the lack of a control cohort. This does not allow the discrimination between the prognostic and predictive value of the factors we investigated. Ideally, the proper control cohort to explore the predictive value of our factors in this specific setting would be the cohort of patients treated with chemotherapy plus the ICI, which currently is unlikely in the clinical practice for patients with high PD-L1 tumours. We are, therefore, conscious that the output of our analysis was to explore an easy-to-assess lab prognostic tool, which could currently aid the decision-making about the addition of chemotherapy to ICI in patients with high PD-L1 tumours, alongside other clinical factors, while waiting for evidence from randomized clinical trials in this setting and the possible exploration of their predictive value. Identifying patients who are likely to benefit from pembrolizumab monotherapy thanks to low NLR and high PD-L1 or nLDH (approximately one third) would be highly relevant to us. The opposite could be even more interesting. Patients with a high PD-L1 (>80%), but NLR >5 and/or high LDH may warrant a combination strategy to offer them the best outcomes. We advocate the clinical evaluation of this easy-to-assess tool based on the combination of low NLR <5 and high PD-L1 >80% or nLDH <269 for retrospective and prospective analyses in clinical trials, comparing the addition of chemotherapy or a different ICI to single-agent immunotherapy. If the validity of this simple tool were confirmed, it could be of crucial importance in the current therapeutic landscape.

Another possible limitation of our study, other than the lack of a validation set, could be the relatively short follow-up time. However, currently available evidence and approval of ICIs in the first-line treatment of NSCLC are based on a median follow-up time from the 12 available studies of 13.0 months, ranging from 7.7 to 25.2. Particularly, the 4 studies with pembrolizumab in the first-line setting had median follow-up of 11.7 months, ranging from 7.7 to 25.2 (1,3).

As above mentioned, since the present study did not compare immunotherapy to chemotherapy plus ICI, we cannot support the predictive value of PD-L1 >80%, or of low NLR <5, but we can suggest their prognostic value, alongside with that of nLDH.

In conclusion, we suggest that low NLR <5, when combined with high PD-L1 >80% or nLDH <269 represents an easily assessable and affordable tool to explore the outcomes of aNSCLC patients with high PD-L1 likely to benefit from pembrolizumab monotherapy. This tool could have a role in therapeutic decision-making alongside other clinical factors. Its validation in retrospective and prospective randomised trials is warranted.

Overall and progression-free survival of all patients and according to disease response (DCR versus PD).

The article’s supplementary files as