Development and Validation of a Prognostic Nomogram in Lung Cancer With Obstructive Sleep Apnea Syndrome

To analyze the prognostic factors and survival rate of lung cancer patients with obstructive sleep apnea (OSA) by nomogram. The nomogram was established by a development cohort (n = 90), and the validation cohort included 38 patients. Factors in the nomogram were identified by Cox hazard analysis. We tested the accuracy of the nomograms by discrimination and calibration, and plotted decision curves to assess the benefits of nomogram-assisted decisions. There were significant difference in sex, apnea hypopnea index (AHI), Tumor Node Metastasis (TNM), coronary heart disease, lowest arterial oxygen saturation [LSpO2 (%)], oxygen below 90% of the time [T90% (min)], the percentage of the total recorded time spend below 90% oxygen saturation (TS90%) and oxygen desaturation index (ODI4) between lung cancer subgroup and lung cancer with OSA subgroup (P < 0.05). Lung cancer patients with OSA age, AHI, TNM, cancer types, BMI and ODI4 were independent prognostic factor. Based on these six factors, a nomogram model was established. The c-index of internal verification was 0.802 (95% CI 0.767–0.885). The ROC curve analysis for the nomogram show 1-year survival (AUC = 0.827), 3-year survival (AUC = 0.867), 5-year survival (AUC = 0.801) in the development cohort were good accuracy. The calibration curve shows that this prediction model is in good agreement. Decision curve analysis (DCA) suggests that the net benefit of decision-making with this nomogram is higher, especially in the probability interval of <20% threshold. The nomogram can predict the prognosis of patients and guide individualized treatment.

To analyze the prognostic factors and survival rate of lung cancer patients with obstructive sleep apnea (OSA) by nomogram. The nomogram was established by a development cohort (n = 90), and the validation cohort included 38 patients. Factors in the nomogram were identified by Cox hazard analysis. We tested the accuracy of the nomograms by discrimination and calibration, and plotted decision curves to assess the benefits of nomogram-assisted decisions. There were significant difference in sex, apnea hypopnea index (AHI), Tumor Node Metastasis (TNM), coronary heart disease, lowest arterial oxygen saturation [LSpO2 (%)], oxygen below 90% of the time [T90% (min)], the percentage of the total recorded time spend below 90% oxygen saturation (TS90%) and oxygen desaturation index (ODI4) between lung cancer subgroup and lung cancer with OSA subgroup (P < 0.05). Lung cancer patients with OSA age, AHI, TNM, cancer types, BMI and ODI4 were independent prognostic factor. Based on these six factors, a nomogram model was established. The c-index of internal verification was 0.802 (95% CI 0.767-0.885). The ROC curve analysis for the nomogram show 1-year survival (AUC = 0.827), 3-year survival (AUC = 0.867), 5-year survival (AUC = 0.801) in the development cohort were good accuracy. The calibration curve shows that this prediction model is in good agreement. Decision curve analysis (DCA) suggests that the net benefit of decision-making with this nomogram is higher, especially in the probability interval of <20% threshold. The nomogram can predict the prognosis of patients and guide individualized treatment.

INTRODUCTION
Lung cancer is the leading cause of cancer death in the world, at present, the pathogenesis of lung cancer is not clearly. Sleep apnea syndrome (SAS) is a disease with high incidence rate and serious harm to patient's health, obstructive sleep apnea (OSA) accounts for more than 90% of SAS (1). OSA is an independent risk factor of multiple system chronic diseases such as hypertension, coronary heart disease, arrhythmia and stroke (2), which also effects the development of chronic obstructive pulmonary disease (3). OSA also plays an important role in the occurrence and development of tumors (4), lung cancer is the malignant tumor which most closely related with OSA (5).
Through there are comprehensive treatments of lung cancer, the survival rate has been significantly improved, improving the quality of life, especially sleep quality were important for lung cancer patients. It's not only helps to accelerate the physical and mental recovery, but also increases the body's immunity and resistance. Lung cancer is the most common cancer in OSA patients, OSA is usually develop moderate or severe when diagnosis, lung cancer and intermittent hypoxia, apnea, daytime sleepiness forming a vicious cycle, decreased the quality of life and survival time significantly (6).
Currently, nomogram have been developed in the majority of cancer types (7)(8)(9), in this study, we aimed to develop and validate a prognostic nomogram which uses widely available general data and laboratory indicators to improve our ability to predict survival time of lung cancer patients with OSA.

Patient Selection
The retrospectively study included 410 lung cancer patients from 2013 to 2016 in Renmin Hospital of Wuhan University and Tongji Hospital, all subjects received philips YZB/USA 1575-2013 portable sleep recorder to monitor patients' sleep for at least 7 h per night, there are 128 cases diagnosis of OSA, the inclusion criteria were as follows: (1) the pathological diagnosis was confirmed as lung malignant tumor; (2) normal range of blood pressure in patients with hypertension after use of antihypertensive treatment and no hypertension-related complications; after symptomatic treatment, the patients with CHD were stable without complications; the blood glucose level in patients with type 2 diabetes without complications; (3) no drugs possibly influencing the sleep patterns was currently being taken.
Patients suffering from the following diseases or lesions were excluded: respiratory infectious disease, intracranial lesions, pulmonary embolism, rheumatic diseases and other chronic disease diseases that may cause abnormal blood oxygen saturation (10,11). This study was conducted in accordance with the Declaration of Helsinki.

Laboratory Measurements
All subjects received philips YZB/USA 1575-2013 portable sleep recorder to monitor patients' sleep for at least 7 h per night, all data previously listed was send back, analyzed by computer and corrected artificially. The data about apnea hypopnea index (AHI), oxygen desaturation index (ODI4), lowest arterial oxygen saturation [LSpO2 (%)], oxygen below 90% of the time [T90% (min)], and the percentage of the total recorded time spend below 90% oxygen saturation (TS90%) were obtained. All the data was recorded when the first time of hospitalizations, clinical information was extracted from Electronic Medical Record system. All the patients received routine tests at the first time visit in hospital.

Follow-Up
Patients were advised to receive regular follow-ups after completion of the primary therapy according to clinical guidelines. Patients were generally follow-up every 3 months in the first 2 years and annually thereafter for patients without evidence of recurrence in the following 3-5 years. OS was defined as the time from the diagnosis of lung cancer to the time of the last follow-up or death, the follow-up deadline was November 1, 2020.

Statistical Analysis
Statistical analyses were performed by SPSS 25.0 (IBM, Chicago, IL, USA) and R for Windows (version3.4.2, http://www.r-project. org/). The optimal cut-off points in our study were evaluated by minimum P-value from log-rank ×2 statistics using the Xtile program (12) and continuous variables were transformed to categorical variables, regression analysis was used to analyze the risk factors in the development cohort, nomogram was formulated based on the results of univariate and multivariate analysis by the package of rms. Study tested the accuracy of the nomograms by discrimination and calibration both in primary and externa validation cohort. The discrimination of the nomogram was measured by Harrell's C-index (C-index). The calibration curve of the nomogram model for the overall survival were formulated. The total points of each patient were calculated according to the established Cox regression model. Survival curves were depicted by the Kaplan-Meier method. A two-sided P < 0.05 was considered statistically significant.

Clinical Characteristics of All Patients
The clinical characteristics of all patients were evaluated. The characteristics of the 282 lung cancer patients and 128 lung cancer patients with OSA are showed in Table 1. The median age was 59.98 years and only 172 patients (42.0%) were female. Among the 410 patients, there were 73 patients with the small cell carcinoma (17.8%) and 337 patients with the non-small cell carcinoma (82.2%). There was significant difference in sex, AHI, TNM stage, coronary heart disease, LSpO2, T90%, TS90%, and ODI4 between lung cancer subgroup and lung cancer with OSA subgroup (P < 0.05).

Clinical Characteristics of Lung Cancer Patients With OSA
The clinical characteristics of the training and validation sets were evaluated. The characteristics of the 90 patients in the development cohort and 38 patients in the validation cohort are showed in Table 2. The majority of patients are men and the TNM stages were represented, there were no statistically significant difference between development cohort and validation cohort.

Biomarker Selection
All the available information including general data, clinical characteristics and biomarkers were included for univariate and multivariate analysis ( Table 3). In univariate analyses age, AHI,

Development of the Prediction Model
A nomogram is a graphic representation of the solution of an equation that provides a reasonable approximation of the probability of a particular outcome, nomogram was developed to predict for survival using the six independent covariates identified in the multivariate model, the mode explanatory covariables consisted of age, AHI, TNM stage, types, BMI and ODI4. A nomogram was constructed to predict 1-, 3-, and 5-year OS (Figure 1).

Validation of the Predictive Accuracy of Nomograms for OS
After internal verification, the C-index of the nomogram model for OS prediction was 0.802 (95% CI 0.767-0.885).
The calibration curve illustrates how the predictions from the nomogram compare with actual outcomes for the 90 patients. The dashed line represents the performance of an ideal nomogram, in which predicted outcomes mostly match with the actual outcomes. The dots were calculated from sub-cohorts of our dataset and represent the performance of our nomogram based on the six biomarkers of Cox model. The calibration plot for the probability of OS at 1, 3, or 5 years after therapy showed an optimal agreement between the prediction by nomogram and actual observation (Figure 2).

ROC of the Predictive Accuracy of Nomograms for OS
The

Clinical Application of Prognostic Nomogram
Decision curve analysis (DCA) is used to evaluate the predictive effect of nomogram on OS, it is very important for clinical decision-making to accurately judge the impact of clinical features and related indicators on the prognosis of lung cancer patients with OSA (14). Factors not related to prognosis were represented by black line, factors related to prognosis were represented by gray line, the model of nomogram represented by the blue line. Results show net benefit can be obtained when make decisions by nomogram (Figure 4). Using this prediction model to calculate the total score, the median of the total score of all patients was 137, total score ≥137 belong to high-risk group, total score <137 belong to low-risk group. Kaplan-Meier was used to analyze OS survival in low-risk group and high-risk group, results show there was statistically significant difference in high risk group and low risk group (P < 0.05), the model has good accuracy and practicability (Figure 5).

DISCUSSION
Prognostic models can facilitate discussion between physicians and patients, the models help to identify high-risk of OSA patients individualized treatments and clinical trials can be developed and may provide insight into the biology of disease. Nomograms have been developed to predict various clinical end points for patients with all kinds of malignancies. In recent years, multi-national cohort studies found that OSA increases the mortality of cancer, among which lung cancer was the most common malignant tumor (15,16), Dreher et al. proposed that the incidence of OSA in new diagnosed lung cancer was 49%, the incidence of moderate/severe OSA was 17% (17), Perez-Warnisher et al. proposed that the incidence of OSA in lung cancer patients was 77.5%, in which moderate/severe OSA accounted for 41.1% (16), which consistent with the results of this study. Li et al. found that the OS of lung cancer patients with severe OSA was lower than that of patients with mild OSA, suggesting that the occurrence and severity of OSA are risk factors to promote cancer development (18). The univariate and multivariate logistic analysis indicated that age, AHI, TNM stage, cancer types, BMI and ODI4 were risk factors for overall survival. Studies confirmed that age is an important factor in the occurrence of lung cancer and as an independent risk factor for survival and prognosis of lung cancer patients (19). The continuous increase of obstructive sleep apnea with age challenges the current theory that mortality due to obstructive sleep apnea and cardiovascular co-morbidities affect obstructive sleep apnea prevalence at an advanced age (20). Body fat effect obstructive sleep apnea syndrome severity in different age groups, the neck and waist circumferences showed a statistically significant correlation with apnea-hypopnea index in both the full sample and in the ≥40 and <60 years age group, these variables did not show any significant correlation with the other two age groups (<40 and ≥60 years) (21). We found that age is an independent risk factor for lung cancer patients with OSA, and the nomogram score increases fastest in the age range of 50-60 years. The elderly patients with lung cancer complicated with OSA should be closely monitored to prevent the occurrence of disease-related complications.
Apnea hypopnea index (AHI) was the gold standard for the diagnosis of OSA. AHI refers to the average number of apnea and hypoventilation per hour during sleep. AHI also a standard for grading the severity of OSA. The results suggest that the higher AHI, the lower the survival rate of lung cancer patients with OSA and the hypoxia microenvironment promotes the growth of lung tumors. This conclusion has been confirmed by relevant studies. Stimulating the intermittent hypoxia in patients with OSAS induce pulmonary metastasis of melanoma (22), other evidence also suggests that hypoxic microenvironment contributing the development of non-small cell lung cancer (23). In vitro studies further proved that intermittent hypoxia lung cancer cells are more resistant and more prone to metastasis (24), indicating that lung cancer and OSAS promote each other, leading to disease progression and reduced survival.
Li et al. reported that tumor staging is related to the severity of OSA, jointly effect the prognosis of lung cancer patients with OSA (18). This study explored the effect of TNM stage to the prognosis, results show TNM stage was an independent risk factor for the prognosis of lung cancer patients with OSA. In addition, the nomogram score of patients with TNM stage IV increased significantly. TNM stage was not simple linear relationship with OS and prognosis, we should pay attention to the progress of the disease of patients with advanced lung cancer, actively treatment intervention to slow down the progress of the disease and improve the quality of life. According to histological classification, lung cancer can be divided into small cell lung cancer (SCLC) and non-small cell lung cancer (NSCLC), non-small cell lung cancer accounts for 80% of lung cancer. After multi-disciplinary comprehensive treatments in recent years, the 5-year survival rate has been greatly improved. However, only 20-30% of patients were in the early stage when diagnosed, so most of them fail to carry out standardized treatment in time and delayed the best treatment period (25). This study analyzed the prognosis of lung cancer patients with OSA, results show the prognosis of patients with small cell lung cancer was poor and the survival time was significantly reduced, although small cell lung cancer had a good response to treatment, it was often too late to make radical resection.
Obesity is one of the most important risk factors for OSA. The results of this study show BMI is an independent prognostic factor for lung cancer patients with OSA. Obesity not only aggravates the severity of OSA, but also reduces the OS of lung cancer patients with OSA. Obese patients with lung cancer and OSA are at high risk of death. Obesity is related to the increase of throat fat, tongue fat and volume (26,27). Obesity patients always bear severe upper airway stenosis. Abdominal and thoracic fat weaken the longitudinal tracheal traction and pharyngeal wall tension, decrease chest wall compliance, decrease lung capacity, aggravate the severity of OSA. The increased of BMI is accompanied by an increase of the incidence of respiratory events and more severe nocturnal hypoxemia (28), indicating that the higher the degree of obesity, the higher the severity of OSA (29). However, some studies suggest that BMI effects the prognosis of lung cancer by influencing all aspects of physical  fitness of the body (30). Others studies suggest that high BMI is closely related to the better OS of lung cancer patients (31). But smoking is also an important confounding factor of lung cancer, so the influence of BMI on lung cancer needs further explored.
Results of Wisconsin Cohort research show severe sleep disordered breathing increases nearly five times death risk of cancer (32). Lung cancer patients prone to combined with intermittent hypoxia, apnea and daytime sleepiness (6). This study show ODI4 was an independent risk factor for lung cancer patients with OSA. Hypoxia environment plays an important role in the development of lung cancer. On the one hand, adequate oxygenation plays an important role in maintaining the normal function of cells, tissues and organs. Hypoxia is prevalent in tumor tissues, even in the absence of severe respiratory diseases. Hypoxia is the result of high proliferation rate of cancer cells, when the speed of neovascularization is slower than that of tumor growth, the amount of oxygen for metabolism can't be provided. On the other hand, lung cancer patients prone to sleep disorders. In addition to the persistent hypoxia of tumor tissue, vascular compression also promotes intermittent hypoxia and any factors causing intermittent hypoxia and apnea can aggravate OSA.

CONCLUSION
This research has some limitations. The number of samples included is limited, and the follow-up time is long, so incomplete clinical information can't be avoided. It is still necessary to carry out external verification with large sample and multi center.
In summary, age, AHI, TNM stage, cancer types, BMI and ODI4 are clinical factors affecting the prognosis of lung cancer patients with OSA. The nomogram established in this study can be used to predict the prognosis of lung cancer patients with OSA and provide help for patients to formulate individualized treatment strategies.

DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article/supplementary material, further inquiries can be directed to the corresponding author/s.

ETHICS STATEMENT
The studies involving human participants were reviewed and approved by Renmin Hospital of Wuhan University Institutional Review Board. The patients/participants provided their written informed consent to participate in this study.