**1. Introduction**

Lung cancer is the most commonly diagnosed malignancy and accounts for approximately 25% of cancer deaths in men and women [1]. The spinal column is the most frequent site for the extrapulmonary metastasis of non-small cell lung cancer (NSCLC), which accounts for 80–85% of lung cancer cases [2]. The lung is also the most common location for primary cancer when a patient presents with spinal metastasis as an initial manifestation of the disease [3]. The incidence of spinal metastasis associated with NSCLC is increasing because of improved survival in these patients based on recent advancements in systemic treatment for NSCLC, such as tyrosine kinase inhibitors (TKIs) for epidermal growth factor receptor (EGFR) mutations [4,5]. Improved survival and increased incidence of spinal metastasis in NSCLC patients render surgical treatment and related decision-making processes for spinal metastasis more important.

Numerous decision-making systems or prognostic models have been introduced to estimate the remaining life expectancies and to sugges<sup>t</sup> appropriate treatment options for patients with spinal metastasis [6]. Authors have used evolving methodologies, such as machine-learning algorithms, to develop a novel prognostic model for spinal metastasis [7]. These models are based on the prognostic factors significantly associated with patient survival in multivariate logistic or proportional hazards regression analyses [8]. Among these factors, the anatomical site for a primary cancer is the most significant prognostic factor,

1

**Citation:** Kim, H.; Chang, S.Y.; Son, J.; Mok, S.; Park, S.C.; Chang, B.-S. The Effect of Adding Biological Factors to the Decision-Making Process for Spinal Metastasis of Non-Small Cell Lung Cancer. *J. Clin. Med.* **2021**, *10*, 1119. https://doi.org/ 10.3390/jcm10051119

Academic Editors: Takashi Hirai, Hiroaki Nakashima, Masayuki Miyagi, Shinji Takahashi and Masashi Uehara

Received: 31 January 2021 Accepted: 4 March 2021 Published: 8 March 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

and is included in all models [9]. However, recent advances in tumor genetics sugges<sup>t</sup> that a simple stratification of primary cancer by the anatomical site is insufficient [10]. Given the extensive evidence in the literature that molecular target therapies significantly improve survival in patients with certain mutations [11], genetic subtype analysis should also be considered when predicting survival in patients with spinal metastasis.

Another biological factor that should be considered in survival prediction is the chronicity of spinal metastasis. Several authors have reported that patients with spinal metastasis at the initial presentation of malignancy (synchronous metastasis) survive longer than those diagnosed with spinal metastasis later during treatment (metachronous metastasis) [3,12]. The development of resistance to previous systemic treatment and the availability of further systemic treatment options have been suggested as potential reasons for the difference in prognosis [13].

The New England Spinal Metastasis Score (NESMS) was recently introduced as a novel prognostic model for patients with spinal metastasis [14]. The NESMS consists of a modified Bauer score component, ambulatory function, and serum albumin (Table 1). The developers of NESMS prospectively validated the system in their following study [15]. However, even the recently developed NESMS system does not consider previously described biological factors when stratifying primary cancer and predicting survival. Therefore, we conducted this study to evaluate the effect of adding biological factors to a validated prognostic model for spinal metastasis—the NESMS. Although multiple prognostic models are available, from conventional scoring systems to novel machine-learning-based models, we chose NESMS because, to the best of our knowledge, it is thus far the only model validated using a well-designed prospective investigation with appropriate power [15].

**Table 1.** The New England Spinal Metastasis Score (NESMS).


#### **2. Materials and Methods**

Consecutive patients who underwent palliative surgical treatment for spinal metastasis of lung adenocarcinoma between March 2012 and October 2018 at the authors' institution were included in the current retrospective study. We included only patients who were biopsy-proven to have adenocarcinoma of the lung and underwent EGFR mutation analysis. Exclusion criteria were as follows: (1) missing data on EGFR mutation analysis results, (2) follow-up period of less than 12 months or unidentified survival period, and (3) patients who died within 2 weeks following surgery due to immediate postoperative complications (Figure 1). The current retrospective study obtained ethical approval and a waiver of informed consent from the institutional review board (IRB No. 2009-060-1155).

**Figure 1.** Flowchart for patient selection. (Abbreviations: NSCLC, non-small cell lung cancer; EGFR, epidermal growth factor receptor).

Surgeries for NSCLC patients with spinal metastasis were performed based on the decisions made during a weekly multidisciplinary tumor board meeting consisting of medical and radiation oncologists, orthopedic and neuro-surgeons, diagnostic radiologists, and pathologists. In general, surgical treatment was considered for patients who were anticipated to have a postoperative survival period longer than 6 months. Surgical indications included (1) metastatic spinal cord compression and (2) spinal instability causing pain that was uncontrolled by medications or radiotherapy. Three different surgeons from the Department of Orthopedic Surgery operated on these patients. We performed all surgeries for palliation.

Patient information was retrieved from electronic medical records and was retrospectively reviewed. Regarding NSCLC and spinal metastasis; we identified the chronicity of spinal metastasis and the positivity of EGFR mutation as primary dependent variables. Spinal metastasis diagnosed at the initial presentation of NSCLC was referred to as synchronous metastasis, and spinal metastasis diagnosed during the course of NSCLC treatment was referred to as metachronous metastasis. Analysis for EGFR mutation was performed using either direct DNA sequencing analysis or peptide nucleic acid (PNA)- mediated real-time polymerase chain reaction (PCR) clamping analysis [16]. Information on pre- and post-operative systemic treatment regimens, including conventional cytotoxic chemotherapy and target therapies, such as TKIs, were also collected. To evaluate the patients' preoperative status, we assessed the preoperative ambulatory status and serum albumin, and applied the NESMS using these variables (Table 1). Preoperative serum albumin within 1 week before surgery and preoperative ambulatory status, which was routinely recorded 1 day before surgery, were selected for the preoperative evaluation. Postoperative survival, defined as the time interval between spinal surgery and either death or the last follow-up, was identified as the primary outcome. Patients' survival beyond 6 months postoperatively was considered the secondary outcome.

Survival probability was estimated using the Kaplan–Meier method (product-limit estimator). The Cox proportional hazard model was applied to develop a prognostic model, and the proportion hazard assumption was checked using log–log plots and the time-bycovariate interaction for each predictor. The Uno's C-index and time-dependent area under the curve (AUC) 6 months postoperatively were utilized to evaluate the discrimination and prediction ability of the NESMS, and the effect of adding two biological factors (chronicity of spinal metastasis and EGFR mutation positivity) into the NESMS. *p*-values were adjusted using the Bonferroni method. All statistical analyses were performed using SAS system for Windows, version 9.4 (SAS Institute, Cary, NC, USA) and R software version 3.6.1 (R Foundation for Statistical Computing, Vienna, Austria). *p*-values less than 0.05 were considered statistically significant.
