Lung cancer remains leading cancer in terms of both incidence and mortality worldwide [1]. However, the increasing detection of early-stage lung cancer has led to improved survival rates, and surgical treatment has been established as the primary curative approach for these early-stage cases [2]. Patients with lung cancer often have comorbidities, including advanced age, chronic obstructive pulmonary disease (COPD), diffuse interstitial lung disease, or inflammatory lung disease [3]. These comorbidities can result in a high incidence of postoperative complications, which complicates surgical treatment even for early-stage lung cancer. Therefore, it is crucial to assess the potential for postoperative pulmonary complications (PPCs) following lung resection.

COPD is a common pulmonary condition, occurring in approximately 5% of lung cancer patients [4]. The Global Initiative for Chronic Obstructive Lung Disease (GOLD) guidelines recommend diagnosing COPD based on the forced expiratory volume in 1 second (FEV₁) and the forced vital capacity (FVC). According to the most recent GOLD guidelines, COPD is diagnosed in patients with FEV₁/FVC <0.7, and severity is classified based on predicted FEV₁% [5]. Numerous studies have shown that poor FEV₁ is associated with an increased frequency of PPCs. Consequently, risk stratification for PPCs is well-established in patients with moderate or severe COPD [6,7]. However, it remains unclear which mild COPD patients are at elevated risk for PPCs.

Numerous studies have shown the clinical efficacy of diffusing capacity of the lung for carbon monoxide (DL_CO) and the predicted postoperative DL_CO (ppoDL_CO) in predicting postoperative morbidity and mortality in patients with lung cancer [8,9]. There are 3 DL_CO parameters in pulmonary function testing (PFT). First, DL_CO represents the ability of the lung to diffuse carbon monoxide across its membranes. Second, ppoDL_CO represents the predicted DL_CO value after surgery. Third, DL_CO/VA represents the DL_CO divided by the alveolar volume (VA). Therefore, this study aimed to determine which DL_CO value could best predict PPCs after lung resection for lung cancer.

Study population

This study initially evaluated 2,798 patients who underwent segmentectomy or lobectomy at Seoul National University Bundang Hospital (SNUBH) due to non-small cell lung cancer (NSCLC) from January 2011 to December 2019. Institutional Review Board (IRB) of SNUBH approval was received (IRB approval no., B-2402-880-101). The requirement for informed consent was waived because of the retrospective study design. The choice of surgical approach, such as video-assisted thoracic surgery (VATS) or open thoracotomy, and the extent of lung resection were determined by the operating surgeon based on the patient’s clinical stage and condition. Patients who underwent PFT with DL_CO within 1 month of surgery were selected for the study. All PFTs, including DL_CO, were conducted at our institution, and all instruments were calibrated before each test. The COPD diagnosis followed the GOLD guidelines, with FEV₁/FVC less than 0.7 diagnosed as having COPD. Severity was categorized based on FEV₁, with FEV₁ ≥80% classified as mild, 50%≤ FEV₁ <80% as moderate, 30%≤ FEV₁ <50% as severe COPD, and FEV₁ <30% as very severe COPD [5]. For this study, patients with FEV₁/FVC above 0.7 were excluded, resulting in the exclusion of 1,844 patients. Additionally, 246 patients with FEV₁ <80%, including those with moderate and severe COPD, were excluded. Finally, 709 patients were included (Fig. 1).

Figure 1. Flow chart showing the patient selection method. A multivariable logistic regression model was constructed for a total of 709 patients. NSCLC, non-small cell lung cancer; COPD, chronic obstructive pulmonary disease; FEV₁, forced expiratory volume in 1 second; FVC, forced vital capacity; PPC, postoperative pulmonary complication; DL_CO, diffusing capacity of the lungs for carbon monoxide.

Various DL_co measurement definitions

DL_CO is a test that, in simple terms, measures the lung’s ability to transfer air. This value is influenced by factors such as lung cell surface area, blood flow, and others, making it subject to variation based on factors like height and sex. To account for these factors, the concept of VA has been introduced, and DL_CO corrected for VA is measured as DL_CO/VA. Additionally, ppoDL_CO represents the predicted DL_CO value after surgery, and its formula is as follows.

Preoperative DL_CO measurement allows the prediction of ppoDL_CO by excluding the segment removed from the 19 segments and dividing by 19.

The VA is a measure of lung size that is most often determined during the measurement of DL_CO via the single-breath helium dilution technique. The patient breathes normally and then exhales to the residual volume. At the residual volume, a gas mixture (carbon monoxide and helium) is inhaled forcefully to the total lung capacity (TLC), which is held for 10 seconds, after which the patient exhales. The exhaled helium concentration is used to calculate a single-breath estimate of TLC and the initial alveolar concentration of carbon monoxide. The VA is the TLC minus the physiologic dead space. In healthy individuals, the VA equals the TLC. However, in subjects with ventilatory impairment, the VA often is much lower than the TLC [10].

Postoperative pulmonary complications

This study analyzed PPCs occurring within 30 days, including prolonged air leak lasting more than 5 days; pneumonia, which was defined as meeting 3 of 5 characteristics (fever, leukocytosis, new infiltration on chest X-ray, positive sputum culture, or treatment with antibiotics); atelectasis requiring bronchoscopic toileting, bronchopleural fistula, which was defined as a major bronchial air leak confirmed by bronchoscopy; and acute respiratory distress syndrome.

Statistical analysis

Data are presented as means with standard deviations for normally distributed variables, and categorical data are presented as counts and percentages. The independent t-test was used for numerical value comparisons. All p-values were derived from 2-sided tests, and values less than 0.05 were considered statistically significant. Multivariable logistic regression analysis was conducted, including variables such as age, sex, smoking (never versus ever), Eastern Cooperative Oncology Group (ECOG) scale (<2 versus ≥2), neoadjuvant treatment, tumor size on chest computed tomography, clinical N stage (N0 versus N+), approach (VATS versus open), extent of surgery, and 3 DL_CO parameters (DL_CO versus ppoDL_CO versus DL_CO/VA). Stepwise backward elimination was employed for statistical analysis. The Akaike information criterion (AIC) and Bayes information criterion (BIC), which indicate how closely the model’s distribution matches the actual data distribution, were analyzed to assess model fit. Lower values of AIC and BIC are considered more consistent with actual data [11]. All statistical analyses were performed using R software ver. 3.6.1 (R Foundation for Statistical Computing, Vienna, Austria).

Among a total of 709 patients, PPCs developed in 106 patients (15%). Prolonged air leak, which was the most common PPC, occurred in 58 patients (55%), followed by pneumonia in 37 (35%), atelectasis in 14 (13%), empyema in 8 (7%), bronchopleural fistula in 4, and acute respiratory distress syndrome in 2 patients.

The preoperative and intraoperative characteristics of the non-PPC and PPC groups are presented in Table 1. There were no significant differences in age, ECOG performance status, neoadjuvant therapy, tumor size, clinical N stage, or extent of surgery between the 2 groups. However, the PPC group had a significantly higher proportion of men, current smokers, and patients who underwent an open surgical approach. The preoperative DL_CO values were similar between the non-PPC group (103.6±19.3) and the PPC group (100.8±24.2), with no significant difference (p=0.250). However, a significantly greater percentage of patients in the PPC group had DL_CO values below 80% (non-PPC versus PPC, 9.1% versus 18.9%; p=0.005). Although the mean ppoDL_CO values did not differ significantly between the non-PPC group (82.9±16.6) and the PPC group (80.3±19.9, p=0.198), the PPC group had a significantly higher proportion of cases with ppoDL_CO values below 60% (non-PPC versus PPC, 6.5% versus 16.0%; p=0.002). Additionally, the DL_CO/VA values were significantly lower in the PPC group (non-PPC versus PPC, 99.1±18.6 versus 91.5±19.9; p<0.001), and a significantly higher percentage of patients in the PPC group had DL_CO/VA values below 80% (non-PPC versus PPC, 15.1% versus 33.0%; p<0.001).

Table 1 . Preoperative and intraoperative characteristics between non-PPC and PPC groups.

Characteristic	No PPC (N=603)	PPC (N=106)	p-value
Age (yr)	69.1±7.6	70.4±9.2	0.164
Sex (male)	457 (75.8)	95 (89.6)	0.002
Smoking (yes)	448 (74.3)	90 (84.9)	0.032
ECOG scale ≥2	91 (15.1)	15 (14.2)	0.918
Neoadjuvant treatment (yes)	25 (4.1)	9 (8.5)	0.092
DLCO (predicted %)	103.6±19.3	100.8±24.2	0.250
DLCO (predicted %: <80%)	55 (9.1)	20 (18.9)	0.005
ppoDLCO (predicted %)	82.9±16.6	80.3±19.9	0.198
ppoDLCO (predicted %: <60%)	39 (6.5)	17 (16.0)	0.002
DLCO/VA (predicted %)	99.1±18.6	91.5±19.9	<0.001
DLCO/VA (predicted %: <80%)	91 (15.1)	35 (33.0)	<0.001
Tumor size on chest CT (cm)	2.7±1.7	3.0±1.9	0.127
Clinical N positive (yes)	70 (11.6)	12 (11.3)	0.619
Approach (open)	49 (8.0)	20 (18.9)	<0.001
Extent of surgery (lobectomy)	559 (92.7)	97 (91.5)	0.818

Values are presented as mean±standard deviation or number (%)..

PPCs, postoperative pulmonary complications; ECOG scale, Eastern Cooperative Oncology Group scale; DL_CO, diffusing capacity of the lungs for carbon monoxide; ppoDL_CO, predicted postoperative diffusing capacity of the lungs for carbon monoxide; DL_CO/VA, diffusing capacity of the lung for carbon monoxide divided by alveolar volume; CT, computed tomography..

The multivariable analysis incorporated all variables used in the initial univariable analysis, with the inclusion of DL_CO, ppoDL_CO, and DL_CO/VA. The outcomes of the multivariable regression models, which designated a DL_CO below 80% as a predictor, identified significant determinants of PPCs: male sex (p=0.001), open surgical approach (p<0.001), and a DL_CO below 80% (odds ratio [OR], 2.12; 95% confidence interval [CI], 1.16–3.88; p=0.015) (Table 2). Similarly, when ppoDL_CO below 60% was used as a predictor, the multivariable regression model demonstrated significant predictors of PPCs: male sex (p<0.001), open surgical approach (p<0.001), and ppoDL_CO below 60% (OR, 2.83; 95% CI, 1.46–5.48; p=0.002) (Table 3). When DL_CO/VA below 80% was considered as a predictor, the model revealed significant associations with PPCs for the following variables: male sex (p=0.013), open surgical approach (p=0.001), and DL_CO/VA below 80% (OR, 2.14; 95% CI, 1.31–3.48; p=0.002) (Table 4). Finally, when analyzing the 3 DL_CO parameters, the multivariable regression model indicated that ppoDL_CO and DL_CO/VA remained significant factors (Table 5).

Table 2 . Multivariable logistic regression model for postoperative pulmonary complications, adjusting for multiple variables, including DL_CO (model 1).

	OR (95% CI)	p-value
Sex (male)	3.04 (1.56–5.92)	0.001
Neoadjuvant treatment (yes)	2.67 (0.94–7.60)	0.065
DLCO (<80%)	2.12 (1.16–3.88)	0.015
Approach (open)	3.02 (1.61–5.67)	<0.001

DL_CO, diffusing capacity of the lungs for carbon monoxide; OR, odds ratio; CI, confidence interval..

Table 3 . Multivariable logistic regression model for postoperative pulmonary complication, adjusting for multiple variables including ppoDL_CO (model 2).

	OR (95% CI)	p-value
Sex (male)	3.14 (1.61–6.16)	<0.001
Neoadjuvant treatment (yes)	2.58 (0.90–7.40)	0.078
ppoDLCO (<60%)	2.83 (1.46–5.48)	0.002
Approach (open)	3.20 (1.70–6.01)	<0.001

ppoDL_CO, predicted postoperative diffusing capacity of the lungs for carbon monoxide; OR, odds ratio; CI, confidence interval..

Table 4 . Multivariable logistic regression model for postoperative pulmonary complications, adjusting for multiple variables, including DL_CO/VA (model 3).

	OR (95% CI)	p-value
Sex (male)	2.35 (1.20–4.61)	0.013
Neoadjuvant treatment (yes)	2.68 (0.94–7.67)	0.066
DLCO/VA (<80%)	2.14 (1.31–3.48)	0.002
Approach (open)	2.66 (1.46–4.85)	0.001

DL_CO/VA, diffusing capacity of the lung for carbon monoxide divided by alveolar volume; OR, odds ratio; CI, confidence interval..

Table 5 . Multivariable logistic regression model for postoperative pulmonary complication by adjusting multiple variables, including DL_CO, ppoDL_CO, and DL_CO/VA (model 3).

	OR (95% CI)	p-value
Sex (male)	2.63 (1.32–5.24)	0.006
Neoadjuvant treatment (yes)	2.442 (0.84–7.00)	0.103
ppoDLCO (<60%)	2.12 (1.03–4.34)	0.041
DLCO/VA (<80%)	1.76 (1.03–3.00)	0.038
Approach (open)	3.03 (1.60–5.72)	0.001

DL_CO, diffusing capacity of the lungs for carbon monoxide; ppoDL_CO, predicted postoperative diffusing capacity of the lungs for carbon monoxide; DL_CO/VA, diffusing capacity of the lung for carbon monoxide divided by alveolar volume; OR, odds ratio; CI, confidence interval..

To determine which DL_CO value was the best predictor, the AIC and BIC for each model were investigated. As shown in Table 6, DL_CO/VA had the lowest AIC and BIC, indicating that it was the model with the best fit. It was found that 1.8% of patients with DL_CO <80% but DL_CO/VA >80% had PPCs (n=2), whereas PPCs occurred in 16.0% of patients with DL_CO >80% but DL_CO/VA <80% (n=17).

Table 6 . The AIC and BIC for each model.

	AIC	BIC
Model 1: variables with DLCO	576.93	608.88
Model 2: variables with ppoLCO	573.79	605.74
Model 3: variables with DLCO/VA	573.55	600.94

AIC, Akaike information criterion; BIC, Bayesian information criterion; DL_CO, diffusing capacity of the lungs for carbon monoxide; ppoDL_CO, predicted postoperative diffusing capacity of the lungs for carbon monoxide; DL_CO/VA, diffusing capacity of the lung for carbon monoxide divided by alveolar volume..

The reason for conducting this study was that if a patient has a DL_CO of less than 60%, there are concerns about PPCs and dyspnea after surgery; however, in reality, many of these patients did not develop PPCs or dyspnea, making this criterion difficult to apply. In patients without symptoms of dyspnea after surgery, the DL_CO/VA value was found to be greater than the DL_CO value. Conversely, in patients who had severe dyspnea or developed PPCs, the DL_CO/VA value was lower than the DL_CO value. Therefore, we investigated which of the 3 values—DL_CO, ppoDL_CO, and DL_CO/VA—was the best predictor of PPCs in COPD patients with preserved FEV₁. It was found that DL_CO <80%, ppoDL_CO <60%, and DL_CO/VA <80% were significant risk factors for the development of PPCs. Among them, DL_CO/VA was the best predictor of PPCs based on the values of the AIC and BIC as indicators of model fit.

DL_CO measures the lung’s ability to absorb oxygen and eliminate carbon monoxide, and it has shown clinical efficacy in predicting patients’ symptoms or pulmonary complications after surgery [12,13]. The study by Ferguson et al. [14], analyzing 854 lung cancer surgery patients, found that in univariate analysis, hazard ratios (HRs) increased for patients with DL_CO <80% (70%–79%, 1.12; 60%–69%, 1.29; <60%, 1.35). In multivariable analysis, DL_CO emerged as an independent predictor of overall survival for all patients (HR, 1.04; 95% CI, 1.00–1.08; p=0.05). At our institution, the results of PFTs included FVC (measurement, % reference), FEV₁ (measurement, % reference), FEV₁/FVC (%), DL_CO (measurement, % reference), VA, and DL_CO/VA (measurement, % reference). Among them, the DL_CO/VA measures the DL_CO divided by the effective VA, which is the TLC minus the dead space. Because VA represents a more accurate assessment of functioning alveoli, theoretically, DL_CO/VA is the most effective predictor of remaining lung function after surgical resection. Cerfolio et al. [15] conducted a similar study that attempted to elucidate the clinical significance of DL_CO/VA. They found that patients with a normal DL_CO but a low DL_CO/VA had a slightly higher complication rate than patients who had a low DL_CO but a normal DL_CO/VA. In our study, 1.8% of patients with DL_CO <80% but DL_CO/VA >80% had PPCs, whereas PPCs occurred in 16.0% of patients with DL_CO >80% but DL_CO/VA <80%.

This study was designed to explore predictors of PPCs in COPD patients with preserved FEV₁ instead of moderate or severe COPD because patients and caregivers should be adequately informed and medical staff should be prepared for the risk of pulmonary complications in COPD patients with relatively good FEV₁. An exemplary study by Sekine et al. [16] compared the frequency of postoperative respiratory failure between COPD patients and those without COPD. Most pulmonary complications had a higher incidence in COPD patients than in non-COPD patients (p<0.01). In this study, the average FEV₁ for COPD patients was 48.79%±10.9%, and the majority had moderate or severe COPD (p<0.001). Thus, we did not think it was necessary to study the associations with DL_CO in patients with moderate to severe COPD because low FEV₁ alone could adequately predict the risk of PPCs in these patients and their caregivers when deciding to operate.

The similarity between the multivariable logistic regression model and the actual data was assessed using the AIC and the BIC. These are statistical metrics that evaluate the fit of a model to the data and its complexity. Generally, smaller values for both AIC and BIC are indicative of a better-fitting model. It is common practice to consider both metrics together when selecting a model [17]. In this study, AIC and BIC were used to evaluate the model’s fit. Since there are no absolute criteria for these values and they are considered relative measures, the relative values of AIC and BIC were taken into account. DL_CO/VA, which had the smallest AIC and BIC values among the 3 DL_CO-related variables, was determined to be the best variable in predicting PPCs. In summary, the study identified several significant risk factors for PPCs, and the assessment of these factors using a multivariable logistic regression model indicated that DL_CO/VA was the most valuable predictor.

In our institution, we commonly use the 6-minute walk test in conjunction with PFT and DL_CO to predict PPCs. This test is not administered to all patients; rather, it is performed based on an individual’s overall health status and physical fitness. It serves as a tool to evaluate a patient’s capacity to avoid PPCs, such as pneumonia, following lung surgery. The relationship between the 6-minute walk test and PPCs has been the subject of investigation both at our institution and in external studies. For instance, Lee et al. [18] found that patients in the moderate-risk category who covered shorter distances during the test were at a higher risk for postoperative cardiopulmonary complications than those who walked longer distances. In the future, we aim to explore the correlation between the 6-minute walk distance and DL_CO/VA in order to develop more accurate predictors of PPCs.

Limitations

This study has several limitations that need to be taken into consideration. First, DL_CO is a test that is highly influenced by the patient’s condition, and it may have limitations in accurately representing the patient’s true lung function. In other words, variations in the patient’s condition on the day of the test could affect DL_CO values, introducing variability in the prediction of outcomes. Second, the significant difference in the number of patients between the PPC group and non-PPC group and the single-center study design could potentially introduce bias. Third, there was a lack of consideration for operation time, presence of diffuse interstitial lung disease, pack-years of smoking, and other factors that could be expected to affect PPCs.

Conclusions

Different DL_CO parameters can predict PPCs in mild COPD patients after pulmonary resection for NSCLC. Although DL_CO and ppoDL_CO are well-known predictors of PPCs, DL_CO/VA was identified as an even stronger predictor and should be considered for predicting PPCs in mild COPD patients.

Author contributions

Conceptualization: MHK, SC, JL. Data curation: SJ, KK, SC, JHJ, WHJ, BCHS. Formal analysis: MHK, JL, JWS. Methodology: MHK, JL, JHJ. Visualization: MHK, JL, SC. Writing–original draft: MHK, SC. Writing–review & editing: all authors. Final approval of the manuscript: all authors.

Conflict of interest

No potential conflict of interest relevant to this article was reported.

Funding

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.