Prognostic and predictive value of neutrophil-to-lymphocyte ratio with adjuvant immunotherapy in stage III non-small-cell lung cancer

Abstract

Background:

Elevated pre-treatment neutrophil-to-lymphocyte ratio (NLR) may reflect immune dysfunction and is negatively prognostic in cancer patients treated with immunotherapy, but it is unclear if NLR is predictive of immunotherapy benefit.

Methods:

We identified stage III non-small-cell lung cancer (NSCLC) patients treated with definitive chemoradiation and adjuvant durvalumab within the national Veterans Affairs system from 2017 to 2021. We compared the prognostic value of NLR measured before durvalumab start to a control group of stage III NSCLC patients treated with definitive chemoradiation alone from 2015–2016 (no-durvalumab group) before the approval of adjuvant durvalumab. We estimated the predictive value of NLR through the statistical interaction of durvalumab group by NLR level. Outcomes included progression-free survival (PFS) and overall survival (OS).

Results:

The primary analysis for NLR included 821 durvalumab patients and 445 no-durvalumab patients. Higher NLR was associated with inferior PFS in both groups (no-durvalumab: adjusted HR [aHR] 1.14 per 7.43 unit increase in NLR, 95% confidence interval [CI] 1.06–1.23; durvalumab: aHR 1.42, 95% CI 1.23–1.64), though this effect was greater in durvalumab patients (p for interaction=0.009). Similar results were found for OS (no-durvalumab: aHR 1.16, 95% CI 1.09–1.24; durvalumab: aHR 1.48, 95% CI 1.25–1.76; p for interaction = 0.010). Absolute lymphocytes, eosinophils, and basophils were not prognostic in either group. Estimates of durvalumab treatment efficacy suggested declining efficacy with higher NLR.

Conclusion:

Pre-treatment NLR is especially prognostic among stage III NSCLC patients treated with adjuvant immunotherapy compared to control patients treated without immunotherapy and may be a predictive biomarker of immunotherapy benefit.

1.1. Introduction

Elevated neutrophil-to-lymphocyte ratio (NLR) in peripheral blood is prognostic in a wide range of malignant and non-malignant conditions.^1–3 Elevated NLR may reflect an imbalance between pro-tumoral neutrophil activation and depletion of anti-tumoral lymphocytes, and consequently, elevated NLR may predict poor responsiveness to immunotherapy.^4,5 Previous studies of NLR’s ability to predict responsiveness to immunotherapy have often lacked appropriate control groups, precluding the estimation of immunotherapy efficacy by NLR level. In this study, we evaluate the prognostic and predictive role of pre-treatment NLR and other peripheral blood markers in a large cohort of stage III non-small-cell lung cancer (NSCLC) patients treated with definitive chemoradiation and adjuvant immunotherapy with the anti-programmed death ligand-1 (PD-L1) therapy durvalumab, compared to a cohort of patients treated with chemoradiation alone before the US Food and Drug Administration approval of adjuvant durvalumab in 2018.

1.2. Methods

1.2.1. Data source.

We identified lung cancer patients using the Department of Veterans Affairs Informatics and Computing Infrastructure (VINCI). VINCI is an informatics platform that facilitates access to patient-level electronic health record information and administrative data housed in the corporate data warehouse (CDW) for all veterans within the VA healthcare system. VINCI also incorporates tumor registry data uploaded from individual VA sites gathered by trained registrars. This study was approved by the local institutional review board.

1.2.2. Patient selection.

We included patients with histologically confirmed stage III NSCLC (AJCC 7^th edition) treated with concurrent chemoradiation, who received at least one dose of adjuvant durvalumab between November 2017 to April 2021. Adjuvant durvalumab infusion dates were identified by intravenous infusion records and manually confirmed by chart review. Staging and treatment data were supplemented with tumor registry data where available. To compare the prognostic value of NLR and other laboratory measurements to patients not treated with durvalumab (“no durvalumab” group), we identified patients with stage III NSCLC (AJCC 7^th edition) treated with concurrent chemoradiation alone between January 2015 and December 2016 before adjuvant durvalumab was approved by the FDA in early 2018. These patients were identified through tumor registry treatment and staging records, and patients with palliative treatment intent or without concurrent chemotherapy treatment were excluded.

1.2.3. Ascertainment of NLR and other laboratory measures.

We used structured laboratory data to ascertain absolute neutrophil, lymphocyte, leukocyte, basophil, and eosinophil counts (K cells/uL). NLR was defined as the ratio of the absolute neutrophil count (ANC) to the absolute lymphocyte count (ALC) in each peripheral blood sample. For analyses evaluating the prognostic and predictive significance of each laboratory measurement (NLR and absolute neutrophils, lymphocytes, leukocytes, basophils, and eosinophils), we selected the closest measurement prior to the date of durvalumab start. Patients without a given laboratory measurement within 30 days before durvalumab start were excluded from that analysis. While this study was designed as a focused exploration of NLR, we included other laboratory measures derived from the routine complete blood count both to explore the contributions of NLR constituents (absolute neutrophils and lymphocytes) and to act as negative internal controls (eosinophils and basophils, for which no association with prognosis or immunotherapy efficacy would be expected).

To evaluate the prognostic value of laboratory measures in patients treated without durvalumab, we applied a nearest-neighbor propensity matching⁶ and imputation procedure to select the target time point of laboratory ascertainment for each patient in the no-durvalumab group. We used multivariable logistic regression (using the baseline covariates listed below) to generate propensity scores for the probability of membership in the no-durvalumab group. Each no-durvalumab patient was then matched (with replacement) to the durvalumab patient with the most similar propensity score. For each matched pair, the lag time from radiation completion to durvalumab start for the durvalumab patient was used to impute the proxy date of durvalumab start for the non-durvalumab patient. This date was then used to search for laboratory measurements within +/− 14 days of the proxy durvalumab start date. This procedure resulted in excellent agreement between groups in the timing of laboratory measurement relative to the radiation completion date (p > 0.14 for all comparisons by one-way ANOVA).

1.2.4. Outcomes and covariates.

The primary outcome measures were progression-free survival (PFS) and overall survival (OS). Date of radiographic progression was determined by manual review of radiological reports by a licensed physician (M.D.G. and K.S.). Date of death was obtained from the VA Vital Status File (drawn from Medicare, Social Security Administration, and the interval VA death registry) and supplemented with the VA Master Patient Index for more recent deaths. For durvalumab patients, survival time was measured from the date of durvalumab start to the date of death from any cause (for OS) or to disease progression or death from any cause (for PFS). For no-durvalumab patients, survival was measured from the proxy date of durvalumab start described above. Patients were censored at the date of last known follow-up, defined as the most recent encounter with a VA provider. Patients with ongoing follow-up past April 15, 2021 were administratively censored at that time.

Demographics including race, sex, and age were obtained through the Master Patient Index. Charlson Comorbidity Index (CCI)^7,8 was calculated from inpatient and outpatient ICD-10 diagnosis codes in the year before the radiation completion date. Smoking status was obtained through Health Factors data.⁹ Concurrent chemotherapy regimen was obtained through intravenous infusion records and supplemented with tumor registry data where available. The reasons for durvalumab discontinuation (classified as completion of planned therapy, progression, immune-related adverse event [irAE], or other) were obtained through manual review of physician notes. Patients were categorized as having durvalumab-related toxicity if the toxicity was possibly, probably, or definitely related to durvalumab in the judgement of the management outpatient oncologist or inpatient physician. Suspected infection was defined as the presence of an order for urinalysis, blood culture, stool culture, urine culture, or sputum culture within +/− 7 days of the laboratory measurement. Steroid therapy was defined as any inpatient or outpatient prescription for oral or intravenous steroid medications within 7 days prior to the laboratory measurement.

1.2.5. Statistical analysis.

Differences in baseline characteristics were assessed with the chi-square test for categorical variables and the t-test for continuous variables. OS and PFS estimates were generated with the Kaplan-Meier method and were compared between groups with the log-rank test in univariable analyses. For presentation purposes NLR was split into 3 groups: 0–3.9, 4.0–7.9, and 8.0+ to incorporate the commonly-used NLR cutoff of 4 and maintain adequate patient numbers in each subgroup.¹⁰ In adjusted analyses, to allow more direct comparison of effects across laboratory measures with different scales and normal ranges, we standardized NLR and other laboratory measures by dividing by the sample standard deviation of each measure. Multivariable Cox regression was performed to estimate the effect of each laboratory measure on survival, adjusting for the laboratory measure (continuous, per 1 standard deviation increase), durvalumab group (binary, durvalumab vs. no durvalumab), the interaction of durvalumab group by laboratory measure, age (continuous, per 10 years), sex (male vs. female), race (African American, Caucasian, or other/unknown), smoking status (current, former, never, or unknown), CCI (0–2, 3–5, 6–8, or 9+), AJCC summary stage (IIIA, IIIB, or III not otherwise specified), suspected infection (binary), steroid treatment (binary), concurrent chemotherapy regimen (carboplatin-paclitaxel vs. other) and histology (adenocarcinoma, squamous cell carcinoma, or other). HRs and 95% confidence intervals [CI] for each laboratory measure in the durvalumab vs. no durvalumab groups were generated from the interaction term by the Wald method. Patients with missing progression data were excluded from the PFS analyses. The logistic regression to generate propensity scores used the same covariate set but without the laboratory measures. Analyses were performed with SAS v9.4 (SAS Institute, Cary, NC) and R v4.0.4 (R Core Team, Vienna, Austria).

1.3. Results

1.3.1. Patient characteristics.

The primary analysis included patients with known NLR within 30 days before durvalumab start. The characteristics of this durvalumab-treated cohort (n=821) and the no-durvalumab comparator group (n=445) are shown in Table 1 and by NLR level in eTable 1. The durvalumab group was more likely to have minimal comorbidity (36% with CCI 0–2 vs 19%, p<0.001) and adenocarcinoma histology (50% vs 34%, p<0.001) compared to the no-durvalumab group but were otherwise similar in baseline characteristics. In both durvalumab and no-durvalumab patients, African American race was associated with lower NLR (eTable 1). There were no other clear associations between NLR level and demographic variables, disease characteristics, or treatment tolerance.

Table 1.

Patient characteristics.

Characteristic	Durvalumab, N = 8211	No durvalumab, N = 4451	p-value2
Age (years)	69 (64, 72)	67 (64, 71)	0.2
Race			0.4
White	610 (74%)	340 (76%)
Black	175 (21%)	82 (18%)
Other/Unknown	36 (4.4%)	23 (5.2%)
Male	784 (95%)	430 (97%)	0.4
CCI			<0.001
0–2	293 (36%)	86 (19%)
3–5	198 (24%)	176 (40%)
6–8	83 (10%)	53 (12%)
9+	247 (30%)	130 (29%)
Smoking			<0.001
Current	284 (35%)	182 (41%)
Former	231 (28%)	174 (39%)
Never	62 (7.6%)	39 (8.8%)
Unknown	244 (30%)	50 (11%)
Summary stage			<0.001
3A	455 (55%)	314 (71%)
3B	344 (42%)	131 (29%)
3 NOS	22 (2.7%)	0 (0%)
Steroid usage	39 (4.8%)	78 (18%)	<0.001
Suspected infection	78 (9.5%)	65 (15%)	0.008
Carboplatin/paclitaxel chemotherapy	587 (71%)	331 (74%)	0.3
Histology			<0.001
Adenocarcinoma	409 (50%)	152 (34%)
Other	27 (3.3%)	49 (11%)
Squamous cell carcinoma	385 (47%)	244 (55%)
NLR (no units)	4.8 (3.2, 7.2)	5.3 (3.3, 8.5)	<0.001
Neutrophils (K cells/uL)	3.70 (2.77, 4.92)	3.80 (2.60, 5.48)	<0.001
Lymphocytes (K cells/uL)	0.78 (0.53, 1.10)	0.70 (0.47, 1.01)	0.062
Leukocytes (K cells/uL)	5.56 (4.36, 6.91)	5.40 (4.00, 7.48)	0.010
Eosinophils (k cells/uL)	0.12 (0.08, 0.23)	0.10 (0.03, 0.20)	0.025
Basophils (K cells/uL)	0.02 (0.00, 0.05)	0.01 (0.00, 0.04)	0.002

¹Median (IQR); n (%)

²One-way ANOVA; Pearson’s Chi-squared test

CCI: Charlson Comorbidity Index; NOS: not otherwise specified; NLR: neutrophil-to-lymphocyte ratio.

In the durvalumab group, NLR was measured on average 2.6 days before durvalumab start (standard deviation [SD] 5.5 days). The timing of NLR measurement relative to the end of radiation therapy was consistent between durvalumab and no-durvalumab groups (durvalumab: mean 48.7 days; no-durvalumab: 46.8 days; p=0.40). NLR at this timepoint was slightly higher in the no-durvalumab group (median 5.3 [interquartile range (IQR) 3.3–8.5] vs. 4.8 [IQR 3.2–7.2], p<0.001).

1.3.2. NLR, ANC, and ALC fluctuate during treatment.

We first assessed the changes in NLR and other laboratory measurements through the course of definitive chemoradiation and durvalumab initiation (Figure 1). Both ANC and ALC declined during concurrent chemoradiation, though the proportional decline from baseline was greater for ALC than ANC; this led to an increase in NLR during chemoradiation from pre-treatment levels in both the durvalumab and no-durvalumab groups. After completion of chemoradiation, both ANC and ALC partially recovered, though ALC remained persistently lower than baseline which resulted in stabilization of NLR at a higher level compared to pre-treatment.

Figure 1. Trajectory of laboratory measurements over time.

This figure shows smoothed estimates of the median value of selected laboratory measurements relative to chemoradiation and durvalumab start. Medians were calculated within +/− 14-day bins and smoothed with generalized additive models. NLR: neutrophil-to-lymphocyte ratio; chemoRT: chemoradiation; K: thousand cells).

1.3.3. Prognostic and predictive value of NLR.

When divided by NLR level, 1-year PFS was lower with higher NLR in the durvalumab group (NLR 0–3.9: 65.2%; NLR 4.0–7.9: 58.7%; NLR 8+: 45.7%; p<0.001 by log-rank; Figure 2A) but not in the no-durvalumab group (NLR 0–3.9: 44.1%; NLR 4.0–7.9: 48.6%; NLR 8+: 39.5%; p=0.14 by log-rank; Figure 2B). Similarly, we observed lower 1-year OS with higher NLR in the durvalumab group (NLR 0–3.9: 82.6%; NLR 4.0–7.9: 75.0%; NLR 8+: 68.1%; p<0.001 by log-rank; Figure 2C), and there was a similar trend in the no-durvalumab group (NLR 0–3.9: 67.5%; NLR 4.0–7.9: 63.1%; NLR 8+: 54.5%, p=0.06; Figure 2D).

Figure 2. Progression-free and overall survival estimates.

This figure shows Kaplan-Meier estimates of progression-free survival for durvalumab (A) and no-durvalumab groups (B), and overall survival for durvalumab (C) and no-durvalumab groups (D). NLR: neutrophil-to-lymphocyte ratio.

In multivariable Cox regression, higher NLR was associated with inferior PFS in both groups (durvalumab: adjusted hazard ratio [aHR] 1.42 per 1 standard deviation [7.55 unit] increase in NLR, 95% CI 1.23–1.64; no-durvalumab: aHR 1.14, 95% CI 1.06–1.23; Table 2), though the risk of progression with higher NLR was relatively greater in the durvalumab group (interaction p-value for NLR by group = 0.009; Figure 3). Similar results were found for OS, with higher NLR associated with inferior survival (durvalumab: aHR 1.48, 95% CI 1.25–1.75; no-durvalumab: aHR 1.16, 95% CI 1.09–1.24; Table 2), though with a relatively greater risk of death with higher NLR in the durvalumab group (interaction p-value for NLR by group = 0.010; Figure 3). In multivariable Cox regression for absolute neutrophils, we found a similar trend of increased prognostic value for PFS in the durvalumab group compared to the no-durvalumab group (Figure 3). Eosinophils, basophils, or lymphocytes were not prognostic in either group. Estimates of durvalumab treatment efficacy (vs. no durvalumab) were generated from the interaction term in each model and suggested decreasing efficacy at higher NLR values (eFigure 1).

Table 2.

Multivariable Cox regression for the effect of NLR on PFS and OS.

	Progression-free Survival			Overall Survival
Characteristic	HR1	95% CI1	p-value	HR1	95% CI1	p-value
NLR (per 7.43 unit increase) by cohort3	—	—		—	—
Durvalumab	1.42	1.23, 1.64	0.0092	1.48	1.25, 1.75	0.0102
No durvalumab	1.14	1.06, 1.23		1.16	1.09–1.24
Age (per 10 years)	1.12	1.00, 1.26	0.043	1.21	1.07, 1.36	0.002
Histology
Adenocarcinoma	—	—		—	—
Other	1.04	0.77, 1.41	0.8	1.12	0.82, 1.55	0.5
Squamous cell carcinoma	0.89	0.76, 1.03	0.12	0.95	0.81, 1.12	0.6
CCI
0–2	—	—		—	—
3–5	1.04	0.84, 1.29	0.7	1.13	0.91, 1.41	0.3
6–8	1.10	0.85, 1.44	0.5	1.02	0.76, 1.37	>0.9
9+	1.04	0.84, 1.29	0.7	1.18	0.93, 1.48	0.2
Summary stage
3A	—	—		—	—
3B	1.35	1.16, 1.57	<0.001	1.30	1.10, 1.54	0.002
3 NOS	1.27	0.68, 2.39	0.5	1.25	0.60, 2.62	0.5
Steroid usage	1.05	0.83, 1.33	0.7	1.07	0.84, 1.37	0.6
Suspected infection	1.28	1.03, 1.60	0.027	1.51	1.21, 1.89	<0.001
Male	1.48	0.92, 2.38	0.11	1.33	0.82, 2.14	0.2
Smoking
Current	—	—		—	—
Former	1.06	0.89, 1.27	0.5	0.96	0.79, 1.17	0.7
Never	1.19	0.91, 1.56	0.2	1.05	0.78, 1.40	0.8
Unknown	1.08	0.86, 1.36	0.5	1.11	0.87, 1.42	0.4
Race
White	—	—		—	—
Black	1.10	0.91, 1.33	0.3	1.07	0.87, 1.32	0.5
Other/Unknown	1.01	0.72, 1.41	>0.9	0.99	0.69, 1.43	>0.9
Carboplatin/paclitaxel chemotherapy	1.04	0.89, 1.23	0.6	1.01	0.84, 1.21	>0.9

¹HR = Hazard Ratio, CI = Confidence Interval.

²p for interaction of group by NLR.

³Hazard ratios for NLR are generated from the interaction term of treatment group and NLR.

Abbreviations: CCI: Charlson Comorbidity Index; NOS: not otherwise specified; NLR: neutrophil-to-lymphocyte ratio.

Figure 3. Forest plot for the relative prognostic value of laboratory measures by group.

This figure shows the adjusted hazard ratios for each laboratory measure expressed in terms of a 1 standard deviation increase in each measure. Hazard ratios for durvalumab (red dots) and no-durvalumab patients (blue dots) were derived from the interaction term of group by laboratory measure in multivariable Cox regressions. Horizontal bars are 95% confidence intervals. NLR: neutrophil-to-lymphocyte ratio; aHR: adjusted hazard ratio; PFS: progression-free survival; OS: overall survival; CI: confidence interval.

1.4. Discussion

In this study of over 1200 stage III NSCLC patients, we show that pre-treatment NLR has greater prognostic value among patients treated with durvalumab compared to NLR measured at a similar time point in patients treated without durvalumab. A similar trend was found for absolute neutrophils but not for other peripheral blood cell types. Estimates of durvalumab treatment effect suggested decreased efficacy with higher NLR. This suggests that proxy immune measures in peripheral blood, particularly measures of neutrophils or the balance of neutrophils to lymphocytes, may predict benefit to immunotherapy.

Many previous studies have demonstrated the prognostic value of NLR among patients with multiple solid tumors,¹ lung cancer in particular,^10–12 and in advanced NSCLC treated with immunotherapy.^13–15 However, NLR is also prognostic in many non-oncologic conditions including pulmonary embolism,² critical trauma,¹⁶ SARS-CoV-2 infection,¹⁷ and septic shock,³ among others. As such it is unclear whether NLR gives unique prognostic information in cancer patients (compared to non-cancer patients) or among patients treated with immunotherapy (compared to other systemic therapies).¹² Previous studies of NLR in oncology have often not included matched control arms to answer these questions directly. Indications for immunotherapy in advanced cancer are rapidly changing, and the proper timing of immunotherapy initiation in the metastatic disease course is also evolving; this complicates the construction of adequate control arms in which patients are matched on baseline characteristics as well as the timing of therapy in their disease course. In this study, since durvalumab was approved for a well-defined stage III NSCLC population and is typically initiated promptly after chemoradiation completion, we were able to evaluate the prognostic value of NLR in patients with similar disease characteristics and at nearly identical timepoints in the disease process. Our finding of greater prognostic value of NLR in patients treated with immunotherapy is therefore suggestive that NLR may be a predictive biomarker of immunotherapy efficacy in this setting.¹⁸

Immunotherapy efficacy is mediated by the innate and adaptive immune system, and tumor development and progression are broadly influenced by systemic inflammation¹⁹; as such it is plausible that peripheral measures of immune system dysfunction may be especially prognostic in patients treated with immunotherapy. Elevated NLR may represent both neutrophil activation, which is linked to cancer progression and inferior immunotherapy efficacy,^20–22 and lymphocyte depletion, which may be linked to poor anti-tumoral immune responses.²³ A recent pan-cancer study linked higher NLR to lower PFS, OS, clinical benefit rates, and response rates among patients treated with immune checkpoint inhibitors, suggesting that NLR may predict immunotherapy efficacy.¹⁸ Other recent studies have suggested that derived NLR (dNLR) and lactate dehydrogenase, combined in the Lung Immune Prognostic Index (LIPI), are prognostic among advanced NSCLC patients treated with immunotherapy but not chemotherapy.^5,24 Our study extends this literature by isolating the unique prognostic role of NLR and neutrophils in non-metastatic NSCLC patients treated with immunotherapy compared to control patients assessed at a similar point in the treatment course, and our data suggest that durvalumab efficacy may decrease at higher NLR values.

Our results are subject to several limitations. First, while NLR and other laboratory measures were assessed at a similar time point in both groups relative to chemoradiation completion, the clinical indication for laboratory testing may differ in each group. For example, peripheral blood markers are routinely checked as a baseline before starting durvalumab, but in patients not treated with durvalumab the peripheral blood may be more likely to be checked in the case of prolonged side effects from chemoradiation, suspected infection, or other adverse clinical scenarios. Though we adjusted for steroid use and suspected infection around the time of laboratory measurement, unmeasured factors may systematically affect the prognostic value of laboratory measures in unpredictable ways. The propensity score matching and imputation of the proxy durvalumab start date for no-durvalumab patients may have also introduced uncertainty into our findings. Radiotherapy dose and fractionation were not available and could represent an unmeasured confounder, given the potential relationship between radiotherapy dose to the immune compartment and chemoradiation outcomes.²⁵ Our results are further limited by their retrospective nature and may be subject to the usual issues of unmeasured confounders and selection biases that may affect the prognostic role of laboratory measures. As such, our results are hypothesis-generating and the predictive role of NLR should be carefully validated prospectively before being used to inform treatment decisions. Despite these limitations, our data suggest that NLR warrants further investigation as a predictive biomarker of immunotherapy benefit.

Highlights.

• Neutrophil-to-lymphocyte ratio (NLR) increases during concurrent chemoradiation.

• Higher NLR predicts poorer prognosis in patients treated with and without adjuvant durvalumab.

• Higher pre-immunotherapy NLR may predict inferior response to immunotherapy.

Availability of data and material:

The data are not available for public use.