Application of a predictive model for postoperative pulmonary complications in patients with non-small cell lung cancer based on cardiopulmonary exercise testing

Abstract

Background

In patients with non-small cell lung cancer (NSCLC), postoperative pulmonary complications (PPCs) significantly increase morbidity and healthcare costs. To improve upon models based solely on static variables, this study aimed to develop a preoperative nomogram integrating cardiopulmonary exercise testing (CPET) parameters for predicting PPCs. The primary objective of this study was to develop a nomogram for predicting PPCs in NSCLC patients using preoperative CPET parameters combined with clinical variables, and to validate its discriminatory power and predictive.

Methods

Data, including clinical and CPET results, were collected from patients who underwent CPET before video-assisted thoracic surgery (VATS) at the Department of Thoracic Surgery, Xuzhou Central Hospital between August 2019 and November 2023. Independent risk factors for PPCs were identified through univariate and multivariate stepwise logistic regressions, and a nomogram prediction model was constructed. The model’s discriminatory power and accuracy were assessed using the concordance index (C-index), calibration curve, receiver operating characteristic (ROC) curve, and area under the curve (AUC) in the validation cohort.

Results

Data from 607 patients were used to construct the nomogram, which included age, intraoperative blood loss, chronic obstructive pulmonary disease (COPD), peak oxygen uptake (VO₂ peak), and the minute ventilation/carbon dioxide production (VE/VCO₂) slope as predictive factors. The model demonstrated good discrimination and accuracy, with a C-index of 0.790 [95% confidence interval (95% CI): 0.743–0.853]. The calibration plot showed strong agreement between predicted and actual PPC probabilities. The ROC curve confirmed the model’s acceptable discrimination ability [area under the curve (AUC): 0.790, 95% CI: 0.605–0.829] in internal validation.

Conclusions

The predictive model for PPCs in patients with NSCLC exhibits strong discrimination and accuracy. It offers valuable support for clinicians in making informed treatment decisions.

Highlight box.

Key findings

• A novel nomogram was developed, incorporating cardiopulmonary exercise testing (CPET) parameters, to accurately predict postoperative pulmonary complications (PPCs) in non-small cell lung cancer (NSCLC) patients, demonstrating strong discrimination (concordance index: 0.790).

What is known and what is new?

• While CPET is recognized for preoperative risk assessment, existing prediction models rely primarily on clinical parameters.

• While the integration of CPET and clinical data has been applied in other medical fields, its use in developing predictive nomograms for PPCs in NSCLC patients remains limited.

What is the implication, and what should change now?

• This tool allows for precise preoperative risk quantification, supporting personalized patient counseling and targeted perioperative management for high-risk individuals.

• Multicenter validation is required, along with the implementation of rehabilitation interventions and enhanced perioperative management to improve postoperative outcomes.

Introduction

Background

According to the Global Cancer Statistics 2023, lung cancer is the leading cause of cancer-related mortality in both men and women over the age of 50 years, surpassing deaths from breast, prostate, and colorectal cancers combined (1). Non-small cell lung cancer (NSCLC) accounts for approximately 85% of all lung cancer cases (2). Treatment options include surgery, chemotherapy, and radiotherapy, with surgery, particularly tumor resection, being the most effective curative approach for early-stage patients (3). Video-assisted thoracic surgery (VATS) has become a common surgical method. While VATS generally results in better patient outcomes, postoperative pulmonary complications (PPCs) remain a significant concern, affecting 15.8–31.7% of patients with NSCLC, nearly one in three surgical candidates (4). PPCs contribute to prolonged hospital stays, increased mortality, and a greater financial burden on patients (5,6). Consequently, preoperative risk assessment is crucial for optimizing perioperative management.

Rationale and knowledge gap

Nomograms, which are visual tools that simplify complex risk predictions, are frequently used by clinicians to aid decision-making (7). Several predictive models and nomograms for PPCs have been proposed in thoracic surgery, but most are based solely on clinical or perioperative variables without incorporating objective cardiopulmonary function (8,9). Previous perioperative assessments have often relied on static pulmonary function tests (PFTs) or American Society of Anesthesiologists (ASA) scores, which require full patient cooperation or are qualitative in nature. Cardiopulmonary exercise testing (CPET) is considered the gold standard for assessing cardiopulmonary function and has gained prominence in recent years. CPET enables objective identification of patients at higher risk for complications and helps determine whether they are fit for major surgeries such as lung resection (10). Its value is widely recognized, with endorsements from organizations such as the American College of Chest Physicians, the European Association for Cardiovascular Prevention & Rehabilitation, and the American Heart Association, all of which recommend CPET for preoperative risk assessment (11,12). Chinese experts have also issued similar recommendations based on consensus (13,14). To address the gap in existing models, our study integrates CPET-derived metrics with clinical risk factors to provide a more individualized risk assessment.

Objective

This study aims to develop a predictive nomogram that combines clinical, pathological, and CPET variables and assess its discriminative ability using the area under the receiver operating characteristic (ROC) curve (AUC) for predicting PPCs in patients with NSCLC. We present this article in accordance with the TRIPOD reporting checklist (available at https://jtd.amegroups.com/article/view/10.21037/jtd-2025-2005/rc).

Methods

Study design and population

This study is a retrospective analysis of prospectively collected clinical and CPET data from patients with NSCLC who underwent surgical resection at our center. It does not involve secondary analysis of a previously published dataset. All consecutive patients who met the inclusion criteria between August 2019 and November 2023 were included in the analysis. A total of 607 patients who underwent VATS at the Department of Thoracic Surgery, Xuzhou Central Hospital, were enrolled.

Inclusion criteria were as follows: (I) postoperative pathological confirmation of NSCLC; (II) age between 18 and 80 years; (III) ability to communicate effectively and complete CPET. Exclusion criteria were as follows: (I) participation in other lung cancer-related research projects within the past 6 months; (II) contraindications to CPET (15).

This study aims to identify factors influencing PPCs and develop a predictive nomogram. PPCs were defined according to previous studies (16), including: respiratory failure requiring intensive care unit admission, intubation, or both; pneumonia (new pulmonary infiltrate with fever treated with intravenous antibiotics); atelectasis requiring bronchoscopy (as determined by the surgical team); pulmonary embolism (diagnosed by computed tomography and treated); and the need for supplemental oxygen at hospital discharge. The study design and reporting follow the TRIPOD guidelines to ensure completeness and transparency in developing and validating the predictive nomogram (17). The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. This study was approved by the Biomedical Research Ethics Review Committee of Xuzhou Central Hospital (approval No. XZXY-LJ-20191114-032). Due to the retrospective nature of the analysis, the requirement for individual patient consent was waived by the ethics committee.

Baseline clinical data

Baseline clinical data, including age, sex, body mass index (BMI), smoking history, preoperative albumin levels, comorbidities, pathological stage, and histology, were extracted from the electronic medical records at the time of the initial visit.

PFT and CPET

The physician calibrated the instrument according to gas circulation and ambient humidity. PFT was conducted to record forced expiratory volume in 1 second (FEV1), forced vital capacity (FVC), and the FEV1/FVC ratio. The subject wore a gas analysis mask, ensuring proper fit, the exercise test was conducted using the HIGHERMed CPET system (Model SMAX58-CE, with static pulmonary function testing capability) manufactured by Nanjing Hanya Co., Ltd. (Nanjing, China), with a cycle ergometer (Model V6001---0004) serving as the exercise equipment. The exercise protocol followed the RAMP method, widely used in clinical practice, which gradually increased the work rate (WR) by 10–30 W/min until exhaustion. Throughout the test, respiratory gas exchange, 12-lead electrocardiogram (ECG), blood pressure, and blood oxygen saturation were continuously monitored. The subject was required to maintain a cadence of 60±5 rpm. After reaching exhaustion, a 3-minute recovery period with no-load exercise was provided before the test concluded without abnormalities. Data collected included peak WR, oxygen uptake at the anaerobic threshold (VO₂AT), peak oxygen uptake (VO₂ peak), peak systolic blood pressure (SBP peak), peak diastolic blood pressure (DBP peak), peak oxygen pulse (O₂pulse), and the minute ventilation/carbon dioxide production (VE/VCO₂) slope.

Statistical analysis

Quantitative data were analyzed using either the independent sample t-test or Mann-Whitney U test, as appropriate, while qualitative data were analyzed using the chi-squared test. Missing values were addressed using median imputation. The selection of variables for the predictive model was performed in two stages. First, univariate logistic regression was used to screen potential variables, with those having a P value <0.2 selected for further analysis. In the second stage, these selected variables were entered into a multivariate logistic regression model to identify independent predictors of PPCs. The final model retained variables with a significance level of p<0.05, and the results are presented as odds ratios (ORs) with 95% confidence intervals (CIs).

A nomogram was developed based on the final multivariate logistic regression model to predict individual PPC risk using the “rms” package in R. The model’s performance was evaluated in terms of discrimination and calibration. Discrimination was assessed using the concordance index (C-index) and the AUC. Calibration, which assesses the alignment between predicted probabilities and observed outcomes, was evaluated using a calibration curve. Both metrics were validated in an independent validation cohort.

All statistical analyses were performed using SPSS software (version 23.0), with nomogram construction and related validation conducted in R. A two-sided P value <0.05 was considered statistically significant.

Sample size considerations adhered to the events-per-variable (EPV) rule. With 5 predictors and an estimated PPC incidence of 20%, this required a minimum of 50 events and a total sample size of at least 250 patients to ensure model stability.

Results

Baseline clinical data

A total of 607 patients were included in the study (patient enrollment process in Figure 1), with 155 patients (25.55%) experiencing PPCs. The demographic and clinical characteristics of the patients are shown in Table 1. Significant differences were found between the PPC and non-PPC groups in terms of age, tumor stage, histology, history of coronary artery disease, chronic obstructive pulmonary disease (COPD), smoking history, and intraoperative bleeding (P<0.05).

Figure 1.

Patient enrollment process. NSCLC, non-small cell lung cancer; SCLC, small cell lung cancer; VATS, video-assisted thoracic surgery.

Table 1. The demographic and clinical features of 607 NSCLC patients.

Variables	All (n=607)	With-PPCs (n=155)	Non-PPCs (n=452)	t/z/χ2	P value
Age (years)	61.82±10.16	65.76±9.34	60.47±10.08	5.747	<0.001
Sex				1.438	0.23
Male	293 (48.3)	81 (52.3)	211 (46.7)
Female	314 (51.7)	74 (47.7)	241 (53.3)
BMI (kg/m2)	24.27±3.06	24.44±2.93	24.22±3.11	0.783	0.43
Smoking				7.421	0.02
No	352 (58.0)	77 (49.7)	275 (60.8)
Yes	194 (32.0)	63 (41.9)	131 (29.0)
Former	61 (10.0)	15 (8.4)	46 (10.2)
Stage				9.364	0.009
I	518 (85.3)	121 (78.0)	397 (87.8)
II	59 (9.8)	24 (15.5)	35 (7.7)
III	30 (4.9)	10 (6.5)	20 (4.5)
Histology				6.871	0.03
AdC	543 (89.5)	130 (83.9)	413 (91.4)
SqCC	52 (8.5)	20 (12.9)	32 (7.1)
Other	12 (2.0)	5 (3.2)	7 (1.5)
Type of surgery				0.165	0.69
Lobectomy	415 (68.4)	108 (69.7)	307 (67.9)
Segmentectomy	192 (31.6)	47 (30.3)	145 (32.1)
CAD				5.997	0.01
Yes	99 (16.8)	35 (22.6)	64 (14.2)
No	508 (83.7)	120 (77.4)	388 (85.8)
COPD				8.229	0.004
Yes	134 (22.1)	47 (30.3)	87 (19.2)
No	473 (77.9)	108 (69.7)	365 (80.8)
DM				3.312	0.07
Yes	73 (12.0)	25 (16.1)	48 (10.6)
No	534 (88.0)	130 (83.9)	404 (89.4)
Preoperative albumin (g/L)	43.48±4.11	42.96±4.65	43.67±3.90	1.708	0.06
Intraoperative blood loss (ml)	100 [50, 100]	100 [50, 150]	100 [50, 100]	6.049	<0.001

Data are presented as mean ± standard deviation, n (%), or median [Q1, Q3]. AdC, adenocarcinoma; BMI, body mass index; CAD, coronary artery disease; COPD, chronic obstructive pulmonary disease; DM, diabetes mellitus; NSCLC, non-small cell lung cancer; PPC, postoperative pulmonary complication; SqCC, squamous cell carcinoma.

The features of CPET and PFT

Table 2 presents the CPET characteristics of the patients. FEV1/FVC, WR peak, VO₂AT, VO₂ peak, and VE/VCO₂ slope were all statistically different between the two groups (P<0.05).

Table 2. Characteristics of patients with NSCLC in CPET.

Variables	With-PPCs (n=155)	Non-PPCs (n=452)	t	P value
FEV1 (L)	2.33±0.61	2.40±0.72	1.005	0.32
FVC (L)	2.77±0.65	2.81±0.77	0.559	0.58
FEV1/FVC	83.92±9.70	85.48±10.02	1.688	0.09
WR peak (W)	97.95±25.26	107.66±31.02	3.882	<0.001
VO2 AT (ml/min/kg)	12.41±2.92	13.59±3.21	4.072	<0.001
VO2 peak (ml/min/kg)	16.68±3.37	18.82±4.00	5.979	<0.001
O2 pulse peak (ml·beat−1)	8.83±2.24	8.98±2.40	0.664	0.51
SBP peak (mmHg)	182.86±24.91	186.81±25.63	1.667	0.10
DBP peak (mmHg)	82.70±14.62	84.60±13.91	1.446	0.15
VE/VCO2 slope	30.96±4.04	27.99±3.43	8.196	<0.001

Data are presented as mean ± standard deviation. CPET, cardiopulmonary exercise testing; DBP, diastolic blood pressure; FEV1, forced expiratory volume in 1 s; FVC, forced vital capacity; NSCLC, non-small cell lung cancer; PPC, postoperative pulmonary complication; SBP, systolic blood pressure; VE/VCO₂ slope, slope of carbon dioxide ventilation equivalent; VO₂ AT, oxygen uptake at anaerobic threshold; VO₂ peak, oxygen uptake at peak; WR, work rate.

Multivariate logistic regression

Logistic regression analysis identified age, intraoperative blood loss, COPD, VO₂ peak, and VE/VCO₂ slope as independent risk factors for predicting PPCs in patients with NSCLC. The ORs for age, intraoperative blood loss, COPD, VO₂ peak, and VE/VCO₂ slope were 1.034, 1.008, 1.776, 0.899, and 1.215, respectively (Table 3).

Table 3. Multivariate logistic regression of NSCLC.

Variables	OR	95% CI	P value
Age (years)	1.034	1.006–1.064	0.02
Intraoperative blood loss (ml)	1.008	1.005–1.012	<0.001
COPD	1.776	1.081–2.918	0.048
VO2 peak (ml/min/kg)	0.899	0.818–0.988	0.03
VE/VCO2 slope	1.215	1.137–1.297	<0.001

CI, confidence interval; COPD, chronic obstructive pulmonary disease; NSCLC, non-small cell lung cancer; OR, odds ratio; VE/VCO₂ slope, slope of carbon dioxide ventilation equivalent; VO₂ peak, oxygen uptake at peak.

Nomogram construction and application

Based on the final regression analysis, a nomogram was constructed that incorporated the five significant risk factors to predict the risk of PPCs (Figure 2). The total score was calculated by summing the points from each factor (age, blood loss, COPD status, VO₂ peak, and VE/VCO₂ slope). The total score could then be projected onto a risk scale to estimate the probability of PPCs. The overall application of this prediction model is presented in Table 4.

Figure 2.

The nomogram of predicting PPCs in patients with NSCLC. COPD, chronic obstructive pulmonary disease; NSCLC, non-small cell lung cancer; PPC, postoperative pulmonary complication; VE/VCO₂ slope, slope of carbon dioxide ventilation equivalent; VO₂ peak, oxygen uptake at peak.

Table 4. The nomogram model scores of predicting PCCs in patients with NSCLC.

Variables	Model score (point)
Age	2.5/10 years (30 years, 0 point)a
Intraoperative blood loss	11.25/100 ml (0 ml, 0 point)b
With-COPD	7.5 pointsc
VO2 peak	2/2 ml/min/kg (32, 0 point)d
VE/VCO2 slope	5/2 (20, 0 point)e

scoring rules for the nomogram: ^a, (age): baseline reference: 30 years old =0 points. The score increases by approximately 2.5 points for every additional 10 years in age. ^b, (intraoperative blood loss): baseline reference: 0 ml =0 points. The score increases by approximately 11.25 points for every additional 100. ^c, (COPD status): with COPD =7.5 points; without COPD =0 points. ^d, (VO₂ peak): baseline reference: 32 ml/min/kg =0 points. The score increases by approximately 2 points for every 2 ml/min/kg decrease below this threshold. ^e, (VE/VCO₂ slope): baseline reference: 20 =0 points. The score increases by approximately 5 points for every 2-unit increase above this threshold. COPD, chronic obstructive pulmonary disease; NSCLC, non-small cell lung cancer; PPC, postoperative pulmonary complication; VE/VCO₂ slope, slope of carbon dioxide ventilation equivalent; VO₂ peak, oxygen uptake at peak.

Calibration and internal validation

The calibration plot for PPCs demonstrated good agreement between predicted and actual probabilities (Figure 3). A C-index of 0.5 indicates no predictive value, while a C-index of 1 indicates perfect agreement between predicted and actual results. The C-index for the model was 0.790 (95% CI: 0.743–0.853), indicating good discrimination and predictive value. The accuracy of individual parameters (age, intraoperative blood loss, COPD, VO₂ peak, and VE/VCO₂ slope) compared to the logistic nomogram was 0.654, 0.654, 0.555, 0.659, 0.717, and 0.790, respectively. The nomogram resulted in fewer missed patients compared to single parameters.

Figure 3.

Calibration plots for predicting PPCs in NSCLC. NSCLC, non-small cell lung cancer; PPC, postoperative pulmonary complication.

As shown in Figure 4, the decision threshold for the optimal Youden index was 0.325, which corresponds to a total score of approximately 65 points. A total score below 65 indicates low risk, while a score above 65 indicates high risk. Additionally, the model demonstrated acceptable discrimination with an AUC of 0.790 (95% CI: 0.605–0.829).

Figure 4.

Receiver operating characteristic curve demonstrated the discriminative ability of nomogram in predicting PPCs. AUC, the area under the curve; PPC, postoperative pulmonary complication.

Discussion

Reducing the incidence of PPCs remains a significant clinical challenge. Current guidelines strongly recommend CPET for assessing surgical risk, particularly for patients undergoing lung resection (11-14). CPET is an effective tool for identifying patients at high risk for PPCs. In addition, nomograms offer a clear, easily interpretable means of conveying outcomes and assist in making more accurate clinical decisions (18).

Based on our current database, this study developed a straightforward and intuitive graphical predictive model that incorporates CPET parameters alongside clinical data to assess the risk of PPCs. This model aims to assist clinicians in making informed management decisions for patients undergoing VATS. In the model, age, intraoperative blood loss, COPD, VO₂ peak, and VE/VCO₂ slope were identified as key predictive factors for postoperative complications. Each of these factors is assigned a score based on its contribution to the outcome. The scores are summed to calculate a total score, which is then transformed into a predicted probability of the outcome event. For example, in our model, a history of COPD contributes approximately 7.5 points, age increases the score by approximately 2.5 points for every 10 years, intraoperative blood loss adds approximately 11.25 points for every 100 ml, VE/VCO₂ slope contributes about 5 points for every 2 units increased, and a decrease in VO₂ peak adds approximately 2 points for every 2 ml/min/kg reduction. As these factors worsen, the total score increases, indicating a higher risk of PPCs (as shown in Table 4). For instance, a 70-year-old patient (approximately 10 points), with a history of COPD (approximately 7.5 points), an intraoperative blood loss of 200 ml (approximately 22.5 points), a VO₂ peak of 16 ml/min/kg (approximately 16 points), and a VE/VCO₂ slope of 32 (approximately 30 points) in CPET, would have a total score of 86 points. This corresponds to a PPC risk of approximately 0.70–0.80, indicating a high likelihood of developing postoperative complications. Our nomogram demonstrated optimal discrimination and excellent calibration in predicting individual PPC risk.

Most existing nomogram models for PPCs primarily use patient history, pathological data, preoperative laboratory tests, PFT, and intraoperative factors as variables. However, to our knowledge, no previous studies have incorporated CPET parameters into a nomogram for predicting PPCs (19,20). Our study is innovative in that it is the first to use CPET data in a nomogram to predict PPCs for patients with NSCLC.

However, several limitations should be noted. First, this is a single-center retrospective study, which may introduce selection bias. Second, external validation through larger, multicenter databases will be necessary to confirm the generalizability of our findings. Future research will conduct multicenter trials for further validation of the model.

The findings align with previous research highlighting the prognostic value of CPET in thoracic surgery. VO₂ peak, representing maximal oxygen uptake, has consistently been shown to predict postoperative complications and long-term survival (21,22). In this study, VO₂ peak emerged as an independent predictor of PPCs through logistic regression (OR: 0.899, 95% CI: 0.818–0.988). A lower VO₂ peak was associated with a higher risk of PPCs (OR: 0.899, 95% CI: 0.818–0.988), supporting its utility in preoperative risk assessment. We hypothesized that lower preoperative VO₂ peak, reflecting reduced exercise tolerance, may increase the risk of PPCs. This is a physiologic inference, supported by the association observed in our cohort.

The VE/VCO₂ slope, a quantitative measure of ventilatory efficiency, was also identified as an independent predictor of PPCs in our cohort. This finding is consistent with prior studies that have linked a higher VE/VCO₂ slope with poorer postoperative outcomes (23,24). Clinically, patients with elevated VE/VCO₂ slope are more likely to experience poor survival and unplanned hospital readmissions in the long term (25,26). Including VE/VCO₂ slope in the nomogram enables more precise preoperative risk stratification, facilitating tailored perioperative management and patient counseling.

Additionally, advanced age and a history of COPD were confirmed as significant predictors of PPCs, consistent with the known pathophysiology where the aging process reduces physiological reserve (27) and pre-existing lung impairment in patients with COPD diminishes their tolerance to surgical stress and anesthesia (28).

Intraoperative blood loss was identified as an independent risk factor for PPCs. This aligns with findings by Li et al. (29), who reported a significantly higher incidence of PPCs, particularly pneumonia and pulmonary atelectasis, in patients with blood loss ≥100 mL compared to those with blood loss <100 mL. The underlying mechanisms suggest that significant blood loss compromises systemic perfusion and tissue oxygenation, contributing to PPCs, as highlighted in major perioperative guidelines (30).

Conclusions

A nomogram was developed to predict PPC risk in patients undergoing VATS. The model demonstrated good calibration and offers clinicians a tool to make individual predictions based on relevant factors, enabling more accurate treatment decisions and perioperative management for patients undergoing VATS.

Supplementary

The article’s supplementary files as

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Biomedical Research Ethics Review Committee of Xuzhou Central Hospital (approval No. XZXY-LJ-20191114-032), and individual consent for this retrospective analysis was waived.

Data Sharing Statement

Available at https://jtd.amegroups.com/article/view/10.21037/jtd-2025-2005/dss

jtd-18-02-77-dss.pdf ^{(100KB, pdf)}

DOI: 10.21037/jtd-2025-2005