Integration of ARISCAT Score and Inflammatory Biomarkers for Predictin

Introduction

Esophageal squamous cell carcinoma (ESCC) represents a major public health challenge in East Asia, particularly in China, where it accounts for a significant proportion of cancer-related morbidity and mortality.^1–3 Although the advent of minimally invasive surgical techniques and neoadjuvant multimodal therapies has improved management strategies, the overall prognosis for patients with locally advanced ESCC remains guarded.^4–6 Postoperative complications, especially pulmonary infections such as pneumonia, continue to undermine surgical success and impede long-term survival.^7–9 Postoperative pneumonia is not only common but also clinically consequential, leading to prolonged hospitalization, increased perioperative morbidity, and elevated healthcare costs.^10–12

In recent years, neoadjuvant chemo-immunotherapy has emerged as a promising approach for treating locally advanced ESCC, enhancing pathological complete response rates and increasing the likelihood of curative resection.^13–15 However, this treatment modality may also compromise immune and nutritional reserves, thereby predisposing patients to infectious complications during the perioperative period.^16–18 The McKeown esophagectomy—although effective for complete tumor clearance—adds an additional layer of risk due to its complexity and proximity to vital respiratory structures.^19,20 These factors synergistically heighten the incidence of postoperative pneumonia, necessitating accurate risk stratification tools tailored to this high-risk population.

Currently, postoperative pneumonia is typically diagnosed based on criteria proposed by the Centers for Disease Control and Prevention (CDC) and the American Thoracic Society/Infectious Diseases Society of America (ATS/IDSA), which include new pulmonary infiltrates on imaging, signs of systemic inflammation (such as fever and leukocytosis), and respiratory symptoms (such as purulent sputum or cough with abnormal auscultation findings).^21–24 Although these criteria are widely accepted, they are usually applied retrospectively and offer limited utility in identifying high-risk individuals before surgery. This highlights the need for preoperative tools that can proactively assess pneumonia risk and guide clinical decision-making.

One such tool is the ARISCAT (Assess Respiratory Risk in Surgical Patients in Catalonia) score, which integrates key preoperative variables—including age, hemoglobin concentration, peripheral oxygen saturation, recent lower respiratory infection, and surgical characteristics—into a unified pulmonary risk model.^25–28 Its practical advantages lie in its ease of use, reliance on routinely available clinical parameters, and potential to inform perioperative management. In clinical scenarios, this score may aid in tailoring prehabilitation plans, allocating intensive care resources, and implementing preventive strategies such as pulmonary physiotherapy or prophylactic antibiotics. Despite its validation in general surgical populations, the ARISCAT score has rarely been examined in thoracic oncology, particularly in ESCC patients treated with modern neoadjuvant immunotherapy. Clarifying its predictive utility in this specific setting may help bridge a critical gap in preoperative risk stratification.

To address this, we conducted a retrospective analysis of 312 ESCC patients who received two cycles of neoadjuvant chemo-immunotherapy followed by McKeown esophagectomy. The primary objective was to assess the predictive value of the ARISCAT score for postoperative pneumonia in this specific clinical context. Additionally, we explored the interactions between ARISCAT scores, systemic inflammation indices (NLR, SII), and postoperative recovery metrics such as hospital length of stay. Through this study, we aim to refine risk stratification strategies and inform perioperative decision-making to mitigate pulmonary complications in high-risk ESCC patients.

Methods

Study Design and Population

As illustrated in Figure 1, we conducted a retrospective, single-center cohort study of consecutive patients with pathologically confirmed esophageal squamous cell carcinoma (ESCC) who underwent neoadjuvant chemo-immunotherapy followed by McKeown esophagectomy at the Affiliated Huai’an No.1 People’s Hospital of Nanjing Medical University between January 2018 and December 2023. Inclusion criteria: (1) histologically proven ESCC; (2) completion of two cycles of platinum-based chemotherapy plus a PD-1 inhibitor (Camrelizumab); (3) R0 resection via McKeown approach; and (4) complete baseline, perioperative and outcome data. Exclusion criteria: (1) chronic pre-existing pulmonary disease (eg, COPD), (2) prior thoracic surgery or other active malignancies, or (3) missing key covariates required for modeling. The protocol was approved by the institutional ethics committee (Approval No. KY-2024-374-01); informed consent was waived due to the retrospective design.

Treatment Regimen and Perioperative Management

All patients received a uniform neoadjuvant regimen (two cycles of platinum-based doublet plus Camrelizumab), minimizing therapeutic heterogeneity. Perioperative antibiotic prophylaxis and enhanced recovery protocols followed institutional standards and did not differ meaningfully across patients; detailed antibiotic timing/intensity was not modeled as a covariate and is discussed as a potential source of residual confounding in the limitations.

Outcomes and Definitions

The primary outcome was postoperative pneumonia within 30 days of surgery, diagnosed per CDC/NHSN criteria: new or progressive pulmonary infiltrates on chest radiography/CT plus evidence of systemic inflammatory response (leukocytosis >12×10⁹/L or leukopenia <4×10⁹/L with fever >38 °C and/or altered mental status) and ≥2 clinical features (eg, purulent sputum, change in sputum characteristics, increased secretions requiring suctioning, or dyspnea). All suspected cases were adjudicated by two clinicians.

Data Collection and Timing of Measurements

Demographic characteristics (age, sex, BMI), comorbidities (hypertension, diabetes, coronary heart disease), lifestyle factors (smoking/drinking), cardiopulmonary function (FEV1, FEV1/FVC, LVEF, ECOG), operative variables (operative duration, recurrent laryngeal nerve palsy), and tumor features (site, length, clinical/ypTNM) were extracted from the electronic medical record by trained nurses. NLR was calculated as the ratio of absolute neutrophil count to absolute lymphocyte count. SII was calculated as platelet count × neutrophil count / lymphocyte count.

ARISCAT Calculation and Risk Strata

The ARISCAT score was calculated from seven preoperative variables (age, SpO₂, recent lower respiratory infection, hemoglobin <10 g/dL, incision site/surgical field, operative duration, and urgency) according to published algorithms. For descriptive analyses, ARISCAT was also categorized as low (0–25), intermediate (26–44), and high (≥45). Pneumonia incidence was summarized across strata to demonstrate risk gradient.

Handling of Missing Data and Data Quality

Data entry followed standardized procedures by trained staff. Records with missing key covariates required for multivariable modeling were excluded listwise. For non-critical variables with <5% missingness, we applied multiple imputation by chained equations (MICE) (20 imputations), pooling effect estimates per Rubin’s rules. Variable ranges and distributions were checked to identify implausible outliers.

Statistical Analysis

Continuous variables were assessed for normality using the Shapiro–Wilk test and summarized as mean ± SD or median (IQR), as appropriate. Between-group comparisons were performed with Student’s t-test or the Mann–Whitney U-test. Categorical variables were presented as counts (percentages) and compared using the χ²-test or Fisher’s exact test, as appropriate. Univariable logistic regression was first used to screen candidate predictors of postoperative pneumonia (screening threshold p < 0.10). Variables with clinical relevance and/or meeting the screening criterion were then entered into the multivariable logistic regression model. To avoid reverse causation, length of hospital stay (LOS) was excluded from multivariable modeling and is reported descriptively only. Categorical variables were dummy-coded with clinically sensible reference categories (eg, female sex; upper tumor location; ypTNM). Odds ratios (ORs) with 95% confidence intervals (CIs) and p values were reported. Potential non-linear associations for ARISCAT score, NLR, and SII were examined using restricted cubic splines (RCS). We report the p value for non-linearity and, where applicable, indicate approximate inflection points (eg, NLR ≈ 5.0, SII ≈ 1300). Model discrimination was evaluated using receiver operating characteristic (ROC) curves with area under the curve (AUC) and 95% CIs. Model calibration was assessed using calibration plots (observed vs predicted risk across deciles of predicted probability) and the Brier score as a global measure of prediction error. In addition, single-marker models (NLR-only, SII-only) were fitted to contextualize standalone performance relative to ARISCAT and the combined model. All statistical analyses were performed using R software (version 4.2.2), with two-sided p-values < 0.05 considered statistically significant.

Results

Patient Characteristics

Among the 312 patients with resectable esophageal squamous cell carcinoma who received neoadjuvant chemo-immunotherapy followed by McKeown esophagectomy, 86 (27.6%) developed postoperative pneumonia within 30 days after surgery. Table 1 summarizes the baseline characteristics of the entire cohort, stratified by pneumonia occurrence. There were no statistically significant differences between the pneumonia and non-pneumonia groups in terms of gender (p = 0.587), age (66.0 ± 6.4 vs 66.0 ± 6.7 years, p = 0.937), BMI (22.5 ± 2.6 vs 22.7 ± 2.3 kg/m², p = 0.438), smoking history (p = 0.603), or common comorbidities such as hypertension (p = 0.478), diabetes (p = 0.266), and coronary artery disease (p = 0.331). However, significant differences emerged in several perioperative and inflammatory parameters. The pneumonia group had markedly higher preoperative ARISCAT scores (53.9 ± 10.6 vs 37.4 ± 5.2, p < 0.001), suggesting a higher baseline risk for respiratory complications. Inflammatory markers were also notably elevated: neutrophil-to-lymphocyte ratio (NLR: 8.3 ± 3.1 vs 4.9 ± 2.1, p < 0.001) and systemic immune-inflammation index (SII: 1379.8 ± 547.3 vs 917.2 ± 397.2, p < 0.001). Moreover, patients with pneumonia experienced significantly prolonged hospital stays (17.7 ± 2.8 vs 13.7 ± 2.3 days, p < 0.001).

Table 1 Baseline Clinical Characteristics of ESCC Patients Undergoing Neoadjuvant Chemo-Immunotherapy Followed by McKeown Esophagectomy Stratified by Postoperative Pneumonia Status

Discrimination and Calibration

As shown in Figure 2, the combined model integrating ARISCAT, NLR, and SII achieved the highest discriminatory performance for predicting postoperative pneumonia, with an AUC of 0.962 (95% CI: 0.916–0.992), outperforming ARISCAT alone (AUC = 0.935, 95% CI: 0.884–0.977), NLR alone (AUC = 0.812, 95% CI: 0.732–0.876), and SII alone (AUC = 0.781, 95% CI: 0.659–0.881). Calibration plots (Figure 3) demonstrated acceptable agreement between predicted and observed risks for all models, with the combined model showing the closest alignment to the ideal 45° reference line. Quantitatively, model performance metrics were as follows: ARISCAT—Brier score = 0.168, Hosmer–Lemeshow p = 0.421; NLR—Brier score = 0.174, Hosmer–Lemeshow p = 0.387; SII—Brier score = 0.182, Hosmer–Lemeshow p = 0.359; combined model—Brier score = 0.152, Hosmer–Lemeshow p = 0.628. Collectively, these findings indicate that the combined model not only provided superior discrimination but also achieved the lowest overall prediction error and the most accurate risk estimation among all models assessed.

Figure 2 Receiver operating characteristic (ROC) curves for predicting postoperative pneumonia in ESCC patients after neoadjuvant chemo-immunotherapy and McKeown esophagectomy. (A) ARISCAT score; (B) NLR; (C) SII; (D) Combined model incorporating ARISCAT, NLR, and SII. The area under the curve (AUC) with 95% confidence intervals (CI) is shown for each model, demonstrating that the combined model achieved the highest discriminative ability.

Figure 3 Calibration plots assessing the agreement between predicted and observed risk of postoperative pneumonia. (A) ARISCAT score; (B) NLR; (C) SII; (D) Combined model. The solid line represents model predictions, the dashed diagonal line represents perfect calibration, and shaded areas indicate 95% confidence intervals. All models demonstrated acceptable calibration, with the combined model exhibiting the closest fit to the ideal line. Model performance metrics were as follows: ARISCAT score: Brier score = 0.168, Hosmer–Lemeshow p = 0.421; NLR: Brier score = 0.174, Hosmer–Lemeshow p = 0.387; SII: Brier score = 0.182, Hosmer–Lemeshow p = 0.359; Combined model: Brier score = 0.152, Hosmer–Lemeshow p = 0.628.

Univariable and Multivariable Analyses

Univariable logistic regression identified several variables associated with postoperative pneumonia (p<0.10; Table 2). To avoid reverse causality, length of hospital stay (LOS) was excluded from the multivariable analysis. In the final adjusted model, ARISCAT score (OR = 1.43, 95% CI: 1.26–1.62, p<0.001), NLR (OR = 1.66, 95% CI: 1.31–2.10, p<0.001), and SII (OR = 1.09, 95% CI: 1.02–1.71, p = 0.010) remained independent predictors. NLR demonstrated the most robust effect size, whereas the effect size for SII—although statistically significant—was close to the lower bound of clinical relevance, suggesting its value lies primarily in combination with structured scores like ARISCAT.

Table 2 Univariate and Multivariate Logistic Regression Analyses of Risk Factors for Postoperative Pneumonia in ESCC Patients After Neoadjuvant Chemo-Immunotherapy and McKeown Esophagectomy

Dose–Response Patterns From Restricted Cubic Splines

Restricted cubic spline modeling revealed distinct dose–response patterns for the three predictors (Figures 4–6). For the ARISCAT score, a linear association with postoperative pneumonia risk was observed across the entire range, with no evidence of non-linearity (p for non-linearity = 0.794), indicating a steady risk increase with higher scores. In contrast, both NLR and SII demonstrated threshold-dependent non-linear relationships. Pneumonia risk began to rise sharply when NLR exceeded approximately 5.0 (p for non-linearity < 0.05), while for SII, a similar inflection occurred at around 1300 (p for non-linearity < 0.05). These findings suggest that while ARISCAT operates as a consistent linear predictor, NLR and SII may confer disproportionate increases in risk beyond specific biomarker thresholds.

Figure 4 Restricted cubic spline illustrating the association between ARISCAT score and postoperative pneumonia risk in ESCC patients undergoing McKeown esophagectomy after neoadjuvant chemo-immunotherapy. A linear dose-response trend was observed (p for non-linearity = 0.794), with risk increasing steadily at higher ARISCAT scores. No distinct inflection point was detected across the observed range.

Figure 5 Restricted cubic spline showing the relationship between NLR and postoperative pneumonia risk in ESCC patients treated with neoadjuvant chemo-immunotherapy followed by McKeown esophagectomy. A nonlinear association was observed (p for non-linearity < 0.05), with a marked increase in pneumonia risk when NLR exceeded approximately 5.0, suggesting a potential threshold effect.

Figure 6 Restricted cubic spline depicting the association between SII and postoperative pneumonia risk following McKeown esophagectomy in ESCC patients who received neoadjuvant chemo-immunotherapy. A threshold-dependent nonlinear pattern was noted (p for non-linearity < 0.05), with pneumonia risk rising sharply when SII exceeded approximately 1300.

Discussion

This retrospective cohort study demonstrates that the ARISCAT score is an independent and clinically meaningful predictor of postoperative pneumonia in patients with resectable ESCC undergoing McKeown esophagectomy after neoadjuvant chemo-immunotherapy. In multivariable analyses that excluded length of stay to avoid reverse causality, ARISCAT, the neutrophil-to-lymphocyte ratio (NLR), and the systemic immune-inflammation index (SII) remained significant risk factors. Importantly, a combined model integrating ARISCAT, NLR, and SII achieved superior discrimination (AUC 0.804; 95% CI 0.756–0.852) and the lowest prediction error (Brier 0.152) with acceptable calibration, outperforming each component alone. Restricted cubic spline analyses further clarified risk dynamics: ARISCAT showed a linear dose–response across the observed range (p for non-linearity = 0.794), whereas NLR and SII exhibited threshold-dependent surges in risk at approximately 5.0 and 1300, respectively (both p for non-linearity < 0.05). Together, these findings underscore the multifactorial nature of postoperative pneumonia and support combining a structured clinical score with readily available inflammatory biomarkers to refine preoperative risk stratification in this high-risk population.

To our knowledge, this is the first study to validate the ARISCAT score in a cohort of ESCC patients treated with modern neoadjuvant chemo-immunotherapy, a regimen that is increasingly adopted as standard care. While the ARISCAT model has been widely evaluated in general surgical settings, evidence for its applicability in thoracic oncology—particularly after combined chemo-immunotherapy—has been lacking. Our findings confirm its strong predictive value in this immunologically and surgically complex context, addressing an important gap in the literature and underscoring its clinical relevance in the current treatment era.

Mechanistically, the ARISCAT score captures several biologically relevant domains that synergistically contribute to postoperative pulmonary vulnerability.²⁹ Advanced age is associated with progressive decline in respiratory muscle strength, reduced mucociliary clearance, and immunosenescence—factors that collectively diminish the ability to clear pulmonary pathogens and recover from surgical stress.³⁰ Preoperative anemia, as included in the ARISCAT scoring system, compromises systemic oxygen-carrying capacity, which can exacerbate tissue hypoxia and impair wound healing and immune responses.³¹ Low peripheral SpO₂ may reflect baseline pulmonary dysfunction or subclinical cardiopulmonary disease, further compounding intraoperative and postoperative respiratory compromise.³² Moreover, a recent history of lower respiratory tract infection suggests a primed pro-inflammatory milieu in the lungs, potentially increasing susceptibility to nosocomial pathogens post-surgery.³³ Importantly, the duration and complexity of McKeown esophagectomy—characterized by three-field lymphadenectomy, bilateral thoracic manipulation, and intrathoracic anastomosis—intensify these risks.³⁴ The procedure leads to substantial operative trauma, extensive exposure of mediastinal structures, fluid shifts, and prolonged mechanical ventilation, all of which contribute to increased incidence of atelectasis, pleural effusion, and ventilator-associated infections.

Furthermore, neoadjuvant chemo-immunotherapy introduces immunologic and metabolic perturbations that further complicate the perioperative course.³⁵ Immune checkpoint inhibitors (ICIs), particularly anti–PD-1/PD-L1 agents, while enhancing anti-tumor immunity, may disrupt immune homeostasis by promoting autoimmune-like inflammation or depleting regulatory T-cell populations critical for mucosal tolerance.³⁶ These alterations have been associated with increased risks of immune-related adverse events, including pneumonitis and opportunistic infections in other oncologic settings.³⁷ Concurrent cytotoxic chemotherapy, such as platinum-based doublets, impairs bone marrow function, suppresses mucosal barrier integrity, and depletes neutrophil reserves, thereby reducing microbial defense capacity.³⁸ When these therapeutic effects are superimposed on the tumor-driven immune suppression frequently observed in ESCC—marked by high myeloid-derived suppressor cell (MDSC) activity and T-cell exhaustion—the result is a profoundly immunocompromised state. Additionally, malnutrition and sarcopenia, prevalent in ESCC patients due to dysphagia and catabolic tumor burden, further impair both innate and adaptive immune responses.³⁹ Collectively, this complex interplay of oncologic treatment, tumor biology, and surgical trauma creates a “perfect storm” for the development of postoperative pneumonia.

Markers such as NLR and SII serve as easily obtainable yet biologically meaningful surrogates for systemic inflammatory and immunologic status.⁴⁰ Elevated NLR is indicative of a neutrophil-predominant pro-inflammatory state coupled with lymphopenia-mediated immunosuppression—a profile associated with impaired pathogen clearance and poor surgical outcomes in multiple solid tumor types.⁴¹ Recent meta-analyses have established NLR as a prognostic biomarker not only for postoperative infection but also for survival in ESCC, further highlighting its clinical relevance.⁴² Similarly, the SII, calculated as platelet count × neutrophil count / lymphocyte count, reflects both innate immune activation and thrombopoietic response to systemic inflammation.^43,44 Platelets play a non-trivial role in pulmonary host defense and tissue remodeling but can also promote microvascular thrombosis and immune dysregulation.⁴⁵ Our finding that both NLR and SII were independently associated with pneumonia risk—and displayed non-linear dose-response curves—underscores their value as integrative, dynamic indicators of perioperative immune competence. These markers may prove especially valuable in resource-limited settings where advanced immunophenotyping is not feasible, offering a pragmatic approach to individualized risk stratification and early intervention.⁴⁶ However, it is noteworthy that the adjusted odds ratio for SII in our multivariate model was 1.09, with a 95% confidence interval ranging from 1.02 to 1.71. While statistically significant, this effect size approaches the lower bound of clinical relevance. This suggests that SII alone may have limited predictive power when used in isolation. Accordingly, SII should be interpreted as a complementary marker rather than a standalone predictor, particularly when combined with structured clinical tools such as the ARISCAT score.

This study benefits from a relatively large, homogeneous cohort, contemporary treatment protocol, and rigorous statistical analysis. The integration of traditional logistic regression with RCS modeling allowed for a more granular understanding of risk factor behavior. However, limitations remain. The retrospective design precludes causal inference and may be subject to residual confounding. Additionally, factors such as perioperative antibiotic regimens, chest physiotherapy, and fluid management were not standardized or accounted for. Our single-center experience may limit generalizability, and survival outcomes were not assessed. Despite these limitations, our findings strongly support the clinical utility of the ARISCAT score in thoracic oncology. By combining it with inflammation-based markers and recovery metrics, a more robust and individualized risk model can be established to guide preoperative counseling, enhance perioperative care, and potentially reduce postoperative morbidity.

Conclusion

In this retrospective study of patients with resectable ESCC undergoing McKeown esophagectomy after neoadjuvant chemo-immunotherapy, the preoperative ARISCAT score, neutrophil-to-lymphocyte ratio (NLR), and systemic immune-inflammation index (SII) were identified as independent predictors of postoperative pneumonia. ARISCAT demonstrated a linear association with risk, whereas NLR (>5.0) and SII (>1300) showed threshold-dependent surges. A combined model incorporating these variables achieved superior discrimination and calibration, representing a practical and cost-effective tool for individualized perioperative risk assessment and targeted preventive strategies. Prospective multicenter studies are needed for validation.