Background
Regimens containing immune checkpoint inhibitors (ICIs) such as anti-PD1 or anti-PD-L1 antibodies have been the standard-of-care therapy for the treatment of advanced NSCLC (aNSCLC) without identifiable molecular driver mutations. However, even in the first-line setting, less than 50% of the patients respond to this type of therapy [
1‐
3], and the response rates are less than 20% when used as part of 2nd line treatments [
4‐
6]. Subsequently, significant efforts have been made to identify biomarkers that are predictive of treatment response. Although PD-L1 has been widely used as a patient stratification biomarker for making treatment decisions, its predictive power is less than optimal as the response rate in PDL1 positive and negative individuals only modestly differs [
1,
2].
In order to be effective, immunotherapies must encourage a robust innate and/or adaptive immune response towards the patient’s tumor. It is, therefore, reasonable to ask if populations of immune effector cells such as neutrophils and lymphocytes shift over the course of immunotherapy treatment, and if baseline or in-treatment levels of these cells affect ICI response. For example, the neutrophil-to-lymphocyte ratio (NLR) is a recognized prognostic marker. A high NLR at baseline or during treatment correlates with poor prognosis such as shorter overall survival (OS), shorter progression-free survival (PFS), or lack of response to therapy in lung, colorectal, kidney and many other solid cancers [
7]. Recently, blood counts have been heavily studied in relation to melanoma immunotherapy response [
8], and several studies have also explored their utility in the context of NSCLC (Supplementary Table
1) [
9‐
23]. These previous studies have generally focused on pretreatment counts rather than changes over the course of therapy, or have been relatively small when examining post treatment effects with less than 160 individuals being studied. Further, most existing studies have not considered the overall status of the patient, other than the ECOG score or site specific metastases.
In addition to blood counts, many other lab tests assess electrolyte imbalances, kidney function, and liver function. These frequently repeated tests monitor the status of the patient’s health during the course of cancer treatment and progression. Previous studies have shown associations between cancer outcomes and standard lab tests. For example, low baseline serum sodium concentration has been associated with shorter OS in a cohort of 197 NSCLC patients on immunotherapy [
24]. Low pretreatment serum albumin has been associated with shorter PFS and early progression [
25]. Pretreatment lactate dehydrogenase (LDH) levels greater than the upper limit of reference range was associated with shorter OS in a cohort of 161 individuals [
26], and in a meta-analysis (
n = 1136), higher pretreatment LDH levels were correlated with significant shorter PFS and OS for ICI therapy in aNSCLC [
27]. Anemia has been associated with poor survival in cancers in general, and more than 30% of lung cancer patients experience anemia, and its incidence after chemotherapy has been estimated at 80% [
28].
Electronic medical records provide a valuable resource for retrospective analysis of real-world patient data for biomarker discovery and validation. Typically, these real-world patient populations, although extremely large, span many years and, reflecting the evolution of treatment guidelines, are heterogeneous in terms of therapies received and laboratory tests performed. However, in spite of this longitudinal heterogeneity, most cancer patients receive a standard battery of laboratory tests including a comprehensive metabolic panel (CMP) and complete blood counts (CBC) throughout their course of therapy. When analyzed within a large enough cohort, it is possible that these standard tests have additional utility and potential prognostic or predictive power beyond their original intent.
In this study we analyzed the medical records of 11,138 NSCLC patients in the Mount Sinai Health System electronic health record (EHR) database, 249 of whom were treated with the PD-L1/PD-1 ICIs nivolumab, atezolizumab, or pembrolizumab for metastatic disease at any line of therapy. We first tested if we could reproduce previously reported correlations between NLR and clinical outcomes in ICI-treated aNSCLC from our data. We also evaluated additional factors that can influence neutrophil levels such as the timing of chemotherapy over a patient’s cancer journey, infections, administration of white blood cell growth factors (which often occur with chemotherapy treatments), and overall patient health to determine if these effects impact the association of NLR and ICI treatment outcomes. We further analyzed common, readily available lab test results such as CBC and CMP to identify markers that correlate with response and survival independently of NLR. Finally, we developed a novel composite biomarker to better predict ICI treatment response and clinical outcomes in aNSCLC.
Discussion
In this study, we confirmed previously reported association between high NLR and poor clinical outcomes in ICI-treated aNSCLC by elucidating such correlations at baseline as well as during the treatment. To expand on published findings, we further demonstrated the differences of survival in the high vs. low NLR patient populations during treatment is more profound than those at baseline (Fig.
2). Moreover, we show high NLR at baseline and to a lesser extent at 2–8 weeks negates the positive association between PD-L1 positivity and ICI response (Fig.
5a, b). To our knowledge, this has not been previously described in the literature and has clinical implications in that a sustained high level of NLR is particularly detrimental to patient outcomes and it is important to adequately manage patients’ blood counts during the course of ICI treatment.
We showed that results of many routine lab tests also correlate with clinical outcomes in our cohort. Of most interest are HGB, RBC counts and HCT where low levels at both baseline and during the treatment reflecting anemia were associated with shorter OS. The association between anemia and poor clinical outcomes in ICI-treated aNSCLC observed in this study has meaningful clinical implications. In the randomized phase 3 Keynote-189 [
1] and IMpower-130 [
3] studies evaluating pembrolizumab and atezolizumab, respectively, in combination with chemotherapy for the treatment of aNSCLC in the 1st line setting, similar percentage of patients experienced any grade anemia in the combination arm vs. the chemotherapy alone arm. However, in Checkmate-227 [
2] study evaluating combination of two ICIs, nivolumab and ipilimumab for the front line treatment of aNSCLC, only 3.8% of patients in the nivolumab/ipilumumab arm had any grade anemia vs. 33% in the chemotherapy arm [
2]. Therefore, our data suggest for those patients with low baseline HGB, RBC, and/or HCT levels, nivolumab/ipilimumab combination might be a more appropriate regimen than ICI-chemotherapy combinations.
Although we analyzed and reported results from data derived at both baseline and during treatment, the baseline results likely have more impact on clinical decision-making for selecting treatment options. As highlighted above, although baseline NLR does correlate with survival, the statistical significance is minimal. Furthermore, baseline NLR is not associated with response rate (Table
1). Therefore, a primary objective of this study was to develop composite baseline biomarkers by combining NLR and other variables that independently correlate with clinical outcomes. We were able to combine NLR and HGB to further stratify patient populations. We showed that patients with NLR ≥ 5 and mild anemia prior to ICI therapy had the worse survival when compared to other patient groups, and when patients had pretreatment anemia meeting the criteria in NCI’s CTCAE for grade 2 or above, they had poor survival regardless of NLR. Therefore, in the context of clinical practice, managing anemia to elevate patients’ HGB level prior to initiating ICI-based therapy may have clinical benefit and warrants further investigation. Alternatively, applying a non-chemotherapy containing regimen may help to alleviate anemia and improve clinical outcomes. We would like to point out that although the patients in this study received ICI-based therapy in different settings of line of therapy, when we only analyzed those patients treated with ICI as the 1st line therapy, the results were similar to those derived from the entire cohort of 249 patients. Therefore, the results from this retrospective study is clinical meaningful in the current clinical setting where ICI-chemo combinations or nivolumab/ipilimumab combination are the standard-of-care 1st line therapy in advanced NSCLC.
Recently, there are several published studies where composite biomarkers were developed for ICI response [
17,
26,
32‐
34]. While these efforts combined multiple variables into a single numerical score to improve statistical associations with clinical endpoints, their clinical utility is limited largely due to the complexity of the scoring systems. Our composite biomarker of NLR and HGB are based on simple, well established cutoffs, and therefore, can be more easily adopted in clinical practice. Moreover, the composite marker of NLR and HGB is based on CBC tests that are routinely performed in all clinics, and could be particularly useful in certain countries and regions where PD-L1 testing may not be readily available.
We recognize there are significant limitations in this study. First, NLR, anemia, or the composite biomarker of NLR and HGB might be prognostic rather than predictive of ICI response. This type of biomarkers would not help us to understand the underlying mechanisms of innate or acquired resistance to ICI-based therapy to develop novel therapeutics or strategies for combination therapies. Second, the real-world data such as the cohort we analyzed in this study are intrinsically noisy. The medical records span many years with evolving treatment guidelines. The patients are heterogeneous in terms of therapies received and laboratory testing performed. In this 249-patient cohort, some patients received ICI-based therapy as the front line therapy, while others did as the 2nd, 3rd, or even later lines of therapy after disease progression on platinum-containing chemotherapy. Although we applied rigorous statistical methods to harmonize the data and to adjust for variables that may impact the results, the retrospective nature of the work requires replication in other cohorts. Of note, there are 45 patients who received ICI-chemotherapy combination as the 1st line therapy. When we removed these 45 patients from the cohort, correlation between NLR, HGB or the NLR-HGB composite marker with survival still remain (Supplementary Figures
3, and
4). Third, due to data limitations, our determination of HPD was based on TTF < 2 months, which is only an approximation of HPD defined by accelerated tumor growth rate or tumor growth kinetics upon anti-PD1/PD-L1 therapy. Finally, even we began with 11,138 NSCLC patients in MSHS, there were only 249 patients received ICI therapy by the cutoff date. This small sample size further emphasizes the need for additional validation.
Open AccessThis article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit
http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (
http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.