A risk score to predict postoperative complications in patients with resectable non-small cell lung cancer

Introduction

Lung cancer is the leading cause of cancer-related death worldwide (1). Lung resection for non-small cell lung cancer (NSCLC) is associated with an increased risk of postoperative complications, which occur in up to one-third of patients (2,3). The mortality rate after lung resection has been reported to be approximately 1% for video-assisted thoracoscopic surgery (VATS) lobectomy and 1.9% for lobectomy by thoracotomy (2,4), with pulmonary and cardiac morbidity being the main complications leading to death (5-7). With the development of minimally invasive surgery, the introduction of neoadjuvant immunotherapy (8) and the perioperative use of immunotherapy combined with chemotherapy (9,10), surgical resection is increasingly being considered for older patients with resectable stage III disease. Studies have shown that lobectomy can be performed for NSCLC in patients aged 75 years or older, with acceptable postoperative morbidity and in-hospital mortality rates when patients are evaluated and selected carefully (11,12).

Several risk scoring systems have been developed to predict the risks of morbidity and mortality after lung resection (13-15). These risk scoring systems use many predictors. Some of them lack specificity for lung resection and are therefore difficult to apply in daily clinical practice (16). Some systems include data from patients who have not been diagnosed with NSCLC nor have undergone anatomical resection. Major complications are often insufficiently classified (14).

In this retrospective analysis, we evaluated the potential risk factors correlated with surgical morbidity and mortality, and on the basis of this assessment, we established a risk scoring system for risk stratification that is useful in clinical practice. The main goal of this project is to offer patients with localized or locally advanced NSCLC a risk-adapted and individualized treatment strategy to reduce the risk of major postoperative complications. We present this article in accordance with the STROBE reporting checklist (available at https://jtd.amegroups.com/article/view/10.21037/jtd-24-1668/rc).

Methods

Study design and patient population

We retrospectively collected and analyzed data from patients with stage 0–IIIA NSCLC and selected patients with stage IIIB NSCLC (T3–4, N2) who underwent tumor resection at the University Hospital Basel between January 2017 and December 2020. The patients underwent either minimally invasive lung resection via VATS or open tumor resection via thoracotomy. Conversion from thoracoscopy to thoracotomy was categorized as open resection. Lung resection involving the chest wall, resection and replacement of the superior vena cava, arterial sleeve resection or patch reconstruction were considered extended tumor resections.

Clinical data were entered into the secuTrial^® database. Postoperative complications were recorded according to the Common Terminology Criteria for Adverse Events (CTCAE) 5.0 and graded according to the Clavien-Dindo classification (17). Grades I and II complications were considered minor complications as they do not require surgical, endoscopic, or radiological intervention. As reported by Seely et al. (3), grades III and IV complications were categorized as major complications. The highest grade of complications was recorded for each patient.

This retrospective study was conducted in accordance with the principles of the Declaration of Helsinki (as revised in 2013). Approval was granted by the institutional review board and the local ethics committee (Ethikkommission Nordwest- und Zentralschweiz) on April 15, 2021 with EKNZ Registration No. 2021-00647. The hospital’s general research consent form was signed by all included patients.

Endpoints

The primary endpoint was any postoperative complication. Postoperative in-hospital mortality was defined as the secondary endpoint.

Statistical analysis

Logistic regression was used for sample size calculation, with an expected event rate of 30% experiencing a major complication. In addition, an odds ratio of 0.4 (or 2.5) was assumed for a given predictor. Approximately 234 patients are needed to achieve a power of 80% and a significance level of 5%. Considering a final model with several predictors, 300 patients were included in the study to meet the overfitting criteria. The descriptive statistics included the frequency (proportion), median with interquartile range (IQR), or mean with standard deviation (SD). Baseline and treatment characteristics were compared using Fisher’s exact test for discrete variables and the Wilcoxon rank sum test or t-test for continuous variables. Various preliminary analyses were performed to select the predictors included in the regression analysis. First, a variable cluster analysis and a redundancy analysis were conducted to identify strongly correlated and redundant cofactors. Clinical and statistical aspects were carefully considered to exclude unimportant and negligible predictors. Furthermore, potentially significant nonlinearities and interactions between variables were analyzed. Sixteen variables were included in a first logistic regression model predicting major complications (grades III–V) vs. minor complications (grades 0–II), despite obvious overfitting, to select the most relevant predictors. The final logistic regression model, adjusted for age and sex, included the diffusing capacity of the lung for carbon monoxide (DLCO), thoracotomy, and extended tumor resection. The back-transformed estimates for the predictors were the odds ratios per given unit or level. In addition, the corresponding 95% confidence intervals (CIs) and P values are indicated. The accuracy of the model was verified via a graphical analysis of the predicted vs. the observed values. The potency of the predictors was displayed by the nomograms. A P value <0.05 was considered significant. All statistical analyses were performed using R version 4.1.3 (18).

Results

After the eligibility of 403 patients who underwent anatomical lung resection for malignancy was assessed, a total of 300 patients with NSCLC were included in our analysis. Patients with malignancies other than NSCLC and patients without signed research consent forms were excluded. The screening criteria for eligible patients and the enrollment, inclusion, and exclusion criteria are presented in Figure 1. Patient demographics, treatment, and disease characteristics are summarized in Table 1. Of 300 patients, 84 (28%) experienced major postoperative complications. Grade III complications occurred in 19.4% of all patients, and 3.6% of patients experienced grade IV complications (Table 2). Detailed information on postoperative complications ≥ grade III can be found in Table 3. The most common minor complications (grade I and II) observed after lung resection included pneumonia, atelectasis, pleural effusion, pneumothorax, air leakage and subcutaneous emphysema (Table 2). The DLCO was significantly lower in the group of patients with major complications than in the group with no or minor complications (63.9% vs. 73.6% of the predicted value, respectively; P<0.001) (Table 1). Extended tumor resection, duration of surgery, and thoracotomy were associated with an increased risk of major postoperative complications (P<0.001). The results suggest that the incidence of major complications was twice as high after neoadjuvant therapy. Among the 33 patients who received neoadjuvant therapy, major complications were observed in 14 patients. Pneumonectomy was performed almost three times as often in patients with major postoperative complications, although the differences did not reach statistical significance (P=0.08 for neoadjuvant therapy, P=0.053 for pneumonectomy; Table 4). There was no significant correlation between patient comorbidities and the incidence of postoperative complications (Table 5).

Figure 1 Study flowchart. NSCLC, non-small cell lung cancer; TNM, tumor-node-metastasis.

Fifteen patients experienced fatal postoperative complications, resulting in a mortality rate of 5%. The mean DLCO was 55.4% (SD 15%) of the predicted value in these patients, and eight patients were known to suffer from chronic obstructive pulmonary disease (COPD). All 15 patients underwent thoracotomy, and extended tumor resection was performed in six patients. The median duration of surgery was 242 minutes (IQR, 185, 289 minutes). Pneumonia and arrhythmia were the most common complications (each in 9 of 15 patients), followed by sepsis and renal failure (each in 5 of 15 patients). Four patients in each group developed pleural empyema and a bronchopleural fistula with an air leak. Respiratory failure was the cause of death in 13 of 15 patients.

The initial multivariate logistic regression analysis with 16 potential predictors revealed that a low DLCO significantly correlated with major postoperative complications (Table 6). Specifically, an increase of 10% in DLCO reduced the probability of experiencing major complications by a factor of 0.80 (95% CI: 0.66, 0.96; P=0.02). Extended tumor resection had the highest odds ratio for major complications (OR: 2.71, 95% CI: 0.79, 9.33; P=0.12). The preliminary analysis revealed a significant correlation between extended tumor resection and major postoperative complications (P<0.001; Table 4). The final logistic regression model adjusted for age and sex included three predictors to avoid overfitting: DLCO, thoracotomy (compared with VATS), and extended tumor resection. Using the estimated coefficients, the probability of occurrence of major postoperative complications can be calculated using the following formula:

1/(1+e−linearpredictor)withlinearpredictor=−1.1457+0.0156∗age+0.3331(iffemale)−0.0244∗DLCO+0.7774(ifopensurgery)+1.5699(ifextendedtumorresection)[1]

On the basis of the formula, but more clearly arranged, individual predictions for the probability of major complications can be represented by a nomogram (Figure 2). For example, the score of a 70-year-old female patient with a DLCO measuring 80% of the predicted value and who underwent planned thoracoscopic lung resection without extended tumor resection was 77 points (20+12+45+0+0), which were calculated from reference points (Figure 3). In the nomogram, a total of 77 points (second line from the bottom; Figure 2) corresponded to a 16% probability of experiencing grades III–V complications (0.16 on the bottom line of the nomogram; Figure 2). The final logistic regression model had an accuracy of 72% (c statistic) in predicting major complications. A visual study abstract is provided in Figure 4.

Figure 2 Nomogram graph demonstrating the probability of experiencing major postoperative complications estimated from the final logistic regression model. The second line from the bottom represents the total points resulting from an addition of the individual points of five predictors. The bottom line presents the estimated probability of experiencing grade III–V complications. DLCO, diffusing capacity of the lung for carbon monoxide; pred., predicted.

Figure 3 Nomogram table with weighting of predictive variables and assignment of reference points to calculate the probability of occurrence of major postoperative complications. DLCO, diffusing capacity of the lung for carbon monoxide; pred., predicted.

Figure 4 Visual abstract. DLCO, diffusing capacity of the lung for carbon monoxide; NSCLC, non-small cell lung cancer; pred., predicted.

Discussion

Despite the small number of patients, the patient cohort represents all anatomical lung resections performed for NSCLC in our center during the study period between January 2017 and December 2020. The database was complete, and no data imputation was required to replace missing variables. During the analysis to develop the risk prediction model, we noticed a lack of correlation between the presumed comorbidities and postoperative outcome, indicating its similarity with an updated Eurolung model (15). The duration of surgery was associated with major postoperative complications; however, this variable is difficult to estimate preoperatively and depends on individual intraoperative findings and surgeons’ experience. For this reason, it was not included in the risk prediction model.

The analysis of mortality revealed that in the 15 patients with fatal outcomes, the DLCO was significantly lower, and more relevant comorbidities, such as COPD and coronary artery disease, were observed. Six patients underwent surgery for stage III disease. Three of these six patients with pathologically confirmed stage III tumors who received neoadjuvant treatment had a major to complete pathological response. Five patients underwent pneumonectomy, which was six times greater than that in the patient group with only minor or no postoperative complications. The number of patients who underwent extended tumor resection was also greater, with 6 out of 15 patients undergoing extended tumor resection. In addition, the duration of surgical procedures was significantly longer in the group of patients who died during their postoperative hospital stay. The high mortality rate may be explained by a combination of the above-mentioned factors.

As the population ages and diagnostics improve, the number of early diagnoses of NSCLC will increase, as will the number of indications for lung resection. If technically feasible, tumor resection should not be withheld because of previous lung surgery, age, or comorbidities. In most patients, pulmonary function can be preserved following minimally invasive sublobar resections. In some cases, a second tumor is detected during the follow-up examination. Compared with open resections, minimally invasive lobar and sublobar resections have been associated with lower complication rates in such cases and are now considered integral parts of lung resection procedures (19).

In light of the above-mentioned facts, risk stratification has gained even greater value in the practice of disciplines involved in the treatment of NSCLC. Several risk prediction models, such as the Thoracoscore, Eurolung and the American College of Surgeons NSQIP Surgical Risk Calculator, have been developed for risk stratification after resection for NSCLC. Studies conducted to validate these models have shown that the predictive accuracy for severe postoperative morbidity and/or mortality is not sufficient (16,20-23).

We aim to prospectively validate our risk prediction model for accuracy and to present the model for risk estimation to the multidisciplinary tumor board. Our model uses only three variables adjusted for age and sex: DLCO, surgical approach (thoracotomy or VATS), and extended tumor resection. This makes it easily applicable in daily practice. The nomogram is simple, and, on the basis of known reference points, the probability of developing major postoperative complications can be quickly assessed.

Limitations

The small number of patients enrolled in this retrospective study and patients whose general research consent was missing could have led to selection bias. Nevertheless, this was a single-center study, and all eligible patients with resectable NSCLC were enrolled consecutively. Furthermore, the surgical procedures were performed during the period before VATS, and parenchymal-sparing resections were widely integrated into clinical practice. However, recently, approximately 70% procedures are minimally invasive. This could explain the relatively high number of major postoperative complications at this tertiary referral hospital, where more complex and larger tumor resections are performed.

Conclusions

We found that patients with high preoperative DLCO and who underwent minimally invasive tumor resection were less likely to experience major postoperative complications. Extended tumor resection in patients with borderline pulmonary function is associated with high risks of postoperative morbidity and mortality. The accuracy in predicting different degrees of complications was 72% and seems appropriate. The prediction of more specific complication grades could improve its predictive ability.

Acknowledgments

We thank the Department of Clinical Research, University Hospital and University of Basel for their consultation and for providing us with the secuTrial^® database. For the data monitoring, we would like to thank Esther Seeberger. We appreciate the patients for allowing us to use their data for scientific purposes. The abstract was presented at the Swiss College of Surgeons Annual Meeting in Basel (7–9^th June 2023) as an oral presentation.

Footnote

Reporting Checklist: The authors have completed the STROBE reporting checklist. Available at https://jtd.amegroups.com/article/view/10.21037/jtd-24-1668/rc

Data Sharing Statement: Available at https://jtd.amegroups.com/article/view/10.21037/jtd-24-1668/dss

Peer Review File: Available at https://jtd.amegroups.com/article/view/10.21037/jtd-24-1668/prf

Funding: This study was funded by University Hospital Basel.

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://jtd.amegroups.com/article/view/10.21037/jtd-24-1668/coif). U.S. is from Statistical Consulting Basel and received consulting fees from Statistical Consulting Basel for the statistical analysis and manuscript writing of this study. The other authors have no conflicts of interest to declare.

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. This retrospective study was conducted in accordance with the principles of the Declaration of Helsinki (as revised in 2013). Approval was granted by the institutional review board and the local ethics committee (Ethikkommission Nordwest- und Zentralschweiz) on April 15, 2021 with EKNZ Registration No. 2021-00647. The hospital’s general research consent form was signed by all included patients.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Sung H, Ferlay J, Siegel RL, et al. Global Cancer Statistics 2020: GLOBOCAN Estimates of Incidence and Mortality Worldwide for 36 Cancers in 185 Countries. CA Cancer J Clin 2021;71:209-49. [Crossref] [PubMed]
2. Falcoz PE, Puyraveau M, Thomas PA, et al. Video-assisted thoracoscopic surgery versus open lobectomy for primary non-small-cell lung cancer: a propensity-matched analysis of outcome from the European Society of Thoracic Surgeon database. Eur J Cardiothorac Surg 2016;49:602-9. [Crossref] [PubMed]
3. Seely AJ, Ivanovic J, Threader J, et al. Systematic classification of morbidity and mortality after thoracic surgery. Ann Thorac Surg 2010;90:936-42; discussion 942. [Crossref] [PubMed]
4. Altorki NK, Wang X, Wigle D, et al. Perioperative mortality and morbidity after sublobar versus lobar resection for early-stage non-small-cell lung cancer: post-hoc analysis of an international, randomised, phase 3 trial (CALGB/Alliance 140503). Lancet Respir Med 2018;6:915-24. [Crossref] [PubMed]
5. Wada H, Nakamura T, Nakamoto K, et al. Thirty-day operative mortality for thoracotomy in lung cancer. J Thorac Cardiovasc Surg 1998;115:70-3. [Crossref] [PubMed]
6. Deslauriers J, Ginsberg RJ, Piantadosi S, et al. Prospective assessment of 30-day operative morbidity for surgical resections in lung cancer. Chest 1994;106:329S-330S. [Crossref] [PubMed]
7. Ginsberg RJ. Lung cancer surgery: acceptable morbidity and mortality, expected results and quality control. Surg Oncol 2002;11:263-6. [Crossref] [PubMed]
8. Forde PM, Spicer J, Lu S, et al. Neoadjuvant Nivolumab plus Chemotherapy in Resectable Lung Cancer. N Engl J Med 2022;386:1973-85. [Crossref] [PubMed]
9. Heymach JV, Harpole D, Mitsudomi T, et al. Perioperative Durvalumab for Resectable Non-Small-Cell Lung Cancer. N Engl J Med 2023;389:1672-84. [Crossref] [PubMed]
10. Wakelee H, Liberman M, Kato T, et al. Perioperative Pembrolizumab for Early-Stage Non-Small-Cell Lung Cancer. N Engl J Med 2023;389:491-503. [Crossref] [PubMed]
11. Kawaguchi Y, Hanaoka J, Ohshio Y, et al. A risk score to predict postoperative complications after lobectomy in elderly lung cancer patients. Gen Thorac Cardiovasc Surg 2018;66:537-42. [Crossref] [PubMed]
12. Dell'Amore A, Monteverde M, Martucci N, et al. Early and long-term results of pulmonary resection for non-small-cell lung cancer in patients over 75 years of age: a multi-institutional study. Interact Cardiovasc Thorac Surg 2013;16:250-6. [Crossref] [PubMed]
13. Brunelli A, Salati M, Rocco G, et al. European risk models for morbidity (EuroLung1) and mortality (EuroLung2) to predict outcome following anatomic lung resections: an analysis from the European Society of Thoracic Surgeons database. Eur J Cardiothorac Surg 2017;51:490-7. [PubMed]
14. Falcoz PE, Conti M, Brouchet L, et al. The Thoracic Surgery Scoring System (Thoracoscore): risk model for in-hospital death in 15,183 patients requiring thoracic surgery. J Thorac Cardiovasc Surg 2007;133:325-32. [Crossref] [PubMed]
15. Brunelli A, Cicconi S, Decaluwe H, et al. Parsimonious Eurolung risk models to predict cardiopulmonary morbidity and mortality following anatomic lung resections: an updated analysis from the European Society of Thoracic Surgeons database. Eur J Cardiothorac Surg 2020;57:455-61. [PubMed]
16. Samson P, Robinson CG, Bradley J, et al. The National Surgical Quality Improvement Program risk calculator does not adequately stratify risk for patients with clinical stage I non-small cell lung cancer. J Thorac Cardiovasc Surg 2016;151:697-705.e1. [Crossref] [PubMed]
17. Dindo D, Demartines N, Clavien PA. Classification of surgical complications: a new proposal with evaluation in a cohort of 6336 patients and results of a survey. Ann Surg 2004;240:205-13. [Crossref] [PubMed]
18. R Core Team. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria. Accessed June 29, 2023. Available online: https://R-project.org/
19. Burt BM, Kosinski AS, Shrager JB, et al. Thoracoscopic lobectomy is associated with acceptable morbidity and mortality in patients with predicted postoperative forced expiratory volume in 1 second or diffusing capacity for carbon monoxide less than 40% of normal. J Thorac Cardiovasc Surg 2014;148:19-28, dicussion 28-29.e1.
20. Qadri SS, Jarvis M, Ariyaratnam P, et al. Could Thoracoscore predict postoperative mortality in patients undergoing pneumonectomy? Eur J Cardiothorac Surg 2014;45:864-9. [Crossref] [PubMed]
21. Nagoya A, Kanzaki R, Kanou T, et al. Validation of Eurolung risk models in a Japanese population: a retrospective single-centre analysis of 612 cases. Interact Cardiovasc Thorac Surg 2019;29:722-8. [Crossref] [PubMed]
22. Gómez de Antonio D, Crowley Carrasco S, Romero Román A, et al. External validation of the European Society of Thoracic Surgeons morbidity and mortality risk models. Eur J Cardiothorac Surg 2022;62:ezac170. [Crossref] [PubMed]
23. Taylor M, Szafron B, Martin GP, et al. External validation of six existing multivariable clinical prediction models for short-term mortality in patients undergoing lung resection. Eur J Cardiothorac Surg 2021;59:1030-6. [Crossref] [PubMed]