Construction and Validation of a Risk Prediction Model for Postoperative Nausea and Vomiting in Patients with Liver Cancer

Abstract

Background

Postoperative nausea and vomiting (PONV) is a common and distressing complication following liver cancer surgery. This study aimed to develop and validate a risk prediction model for PONV in patients undergoing hepatectomy for hepatocellular carcinoma.

Methods

This prospective study enrolled patients who underwent liver resection between February 2024 and December 2024. Based on risk factors identified through univariate and binary logistic regression analyses, a nomogram prediction model was constructed. The model’s discrimination was evaluated using the area under the receiver operating characteristic curve (AUC-ROC) and the consistency index (C-index). Calibration was assessed with calibration curves, and internal validation was performed via the bootstrap method.

Results

A total of 512 patients were included in the modeling cohort. The incidence of PONV was 47.5%. Significant predictors incorporated into the nomogram included age, gender, duration of surgery (min), time of hepatic portal vein occlusion during operation (min), and history of PONV. The model demonstrated an AUC of 0.717 (95% CI: 0.673–0.761), with a sensitivity of 69.9% and a specificity of 62.0% at the optimal cut–off value of 0.413. Bootstrap internal validation yielded a C-index of 0.717, and the calibration curve indicated good agreement between predicted and observed outcomes.

Conclusion

The developed risk warning model shows an acceptable predictive effect in identifying the risk of PONV in patients with liver cancer. It can assist clinical medical staff in early risk assessment and individualized intervention, and has certain predictive value.

Introduction

Postoperative nausea and vomiting (PONV) represents a frequent and distressing complication following surgeries conducted under general anesthesia, with reported incidences ranging from 20–40% in the general surgical population.^1,2 Despite significant advancements in anesthetic techniques and the availability of effective antiemetics, PONV continues to adversely affect patient recovery. It can lead to considerable discomfort, delay the initiation of oral intake, cause electrolyte imbalances, increase the risk of wound-related complications, and pose a threat of pulmonary aspiration.³ These adverse sequelae often result in prolonged hospitalization and elevated healthcare expenditures.

Patients undergoing liver cancer resection are particularly susceptible to PONV. The perioperative profile of these patients encompasses both well-established general risk factors⁴ (eg, history of motion sickness or prior nausea and vomiting) and surgery-specific determinants. The intrinsic characteristics of major hepatectomy–such as prolonged operative duration, extensive manipulation of internal organs, and substantial fluid shifts–independently elevate the risk of PONV. Furthermore, the pathophysiological context of liver cancer often involves varying degrees of pre-existing hepatic dysfunction,⁵ which may impair the metabolism of anesthetic and antiemetic agents, potentially compromising their clinical efficacy.⁶ Notably, time of hepatic portal vein occlusion during operation has been identified as one of the most potent triggers of PONV in this surgical setting.⁷

While several validated risk prediction models are widely utilized in general surgical settings^8–10—such as the Apfel (area under the curve [AUC]: 0.70; 95% confidence interval [CI]: 0.67–0.72),¹¹ Koivuranta (AUC: 0.71; 95% CI: 0.69–0.73), and Palazzo (AUC: 0.68; 95% CI: 0.65–0.70) scores —their applicability and predictive accuracy in specialized populations, such as patients undergoing liver cancer resection, remain uncertain. Our previous single-center prospective study¹² specifically conducted external validation of the Apfel and Koivuranta risk scores in patients with liver cancer undergoing hepatectomy, and the results confirmed that these two classic general surgical PONV risk scores exhibited poor discriminative ability (AUC=0.612 and 0.628, respectively) and unsatisfactory calibration in this specific population, failing to meet the clinical needs of PONV risk assessment for liver cancer patients. These established tools may not adequately incorporate procedure-specific variables relevant to hepatic resection, including parameters of liver function, extent of hepatectomy, or anesthetic considerations related to hepatic physiology. Therefore, it is imperative to develop a specialized multidimensional risk prediction model integrating both general and hepatocellular carcinoma-specific risk factors for PONV. This predictive nomogram can be readily embedded into standardized perioperative protocols for PONV prophylaxis. Specifically, patients stratified into the high-risk group via this model can receive intensified multi-agent antiemetic prophylaxis comprising agents with distinct pharmacological mechanisms, such as 5-HT₃ receptor antagonists, neurokinin-1 receptor antagonists, and dexamethasone. Furthermore, anesthesiologists may adopt total intravenous anesthesia or opioid-sparing analgesic regimens for high-risk patients to mitigate the emetogenic effects of volatile anesthetics and opioids. Postoperatively, early identification of high-risk patients enables nurses to strengthen symptom surveillance and promptly administer rescue antiemetics, which collectively reduces the overall incidence of PONV.

The main objective of this study is to construct and validate a risk prediction model for PONV in patients undergoing liver cancer surgery. By systematically identifying and incorporating key predictors, we aimed to provide clinical healthcare providers with a reliable, patient-specific risk prediction tool that can guide individualized preventive strategies throughout the perioperative period. This nomogram is designed to help clinical medical staff identify individuals at high risk of PONV, formulate targeted preventive measures, optimize resource allocation, and ultimately improve postoperative outcomes and patient satisfaction.

Materials and Methods

Study Subjects and Design

In this prospective study, we consecutively enrolled patients who underwent liver cancer surgery at two tertiary hospitals in Shanghai between February and December 2024. The inclusion criteria were: (1) diagnosis of hepatocellular carcinoma and scheduled for elective hepatectomy under general anesthesia; (2) age ≥18 years; (3) conscious with normal communication ability; (4) provision of written informed consent by the patient or their legal guardian. Exclusion criteria included: (1) receipt of perioperative chemotherapy or immunotherapy; (2) occurrence of major perioperative complications; (3) postoperative admission to the intensive care unit (ICU); (4) pre‑existing cognitive or communication impairments; (5) death due to severe complications during the study period; and (6) incomplete or invalid clinical data. The study protocol was reviewed and approved by the Institutional Review Board of Zhongshan Hospital, Fudan University (approval number: B2020-074).

Research Variables and Definitions

PONV

PONV are common complications following surgery. Based on clinical presentation, PONV was categorized into the following grade¹³: Grade I, no nausea or vomiting; Grade II, mild nausea without vomiting; Grade III, vomiting without expulsion of gastric contents; and Grade IV, intolerable nausea and vomiting, often accompanied by expulsion of gastric contents. In this study, the absence of PONV was defined as Grade I, while the occurrence of PONV was defined as Grade II or higher. PONV occurrence and severity were assessed by trained investigators at 2 hours, 6 hours, and 24 hours after surgery. Nausea was evaluated through direct questioning of the patients, and vomiting was assessed by the investigators’ observation of the characteristics of vomitus to determine the grading. Patients were considered to have experienced PONV if they presented with Grade II or higher symptoms at any of these assessment time points.

Independent Variables

To identify potential factors influencing PONV, our research team conducted a comprehensive review of the relevant literature. Based on this review, a preliminary set of influencing factors was compiled, including demographic characteristics (age, gender, smoking history, body mass index (BMI), history of hypertension, and history of diabetes), patient-specific factors (history of motion sickness and history of migraine), surgical variables (surgical method, duration of surgery, and intraoperative fluid infusion volume), and anesthesia-related factors (anesthesia method). In addition, through consultation with hepatobiliary surgery specialists, two surgery-specific variables–intraoperative occlusion of the hepatic portal vein and time of hepatic portal vein occlusion during operation–were identified as potentially relevant to liver resection and were subsequently incorporated into the analytical framework.

Collection of Data and Quality Control

A standardized data collection protocol was implemented. Given the inherently subjective nature of nausea and vomiting, all researchers underwent standardized training to ensure consistency and minimize inter-rater variability. At each of the three postoperative assessment time points (2 hours, 6 hours, and 24 hours after surgery), trained investigators assessed the presence and severity of PONV. Nausea was evaluated through direct questioning of the patients, and vomiting was assessed by observing the characteristics of vomitus to determine the grading according to the predefined criteria. Data were collected using structured questionnaires, face-to-face interviews, and reviews of medical records. Prior to data collection, all participants were fully informed about the study objectives and procedures. Only clarifying questions regarding uncertainties were addressed, and no leading questions were posed. Simplified language was used when necessary to assist participants with comprehension or recall difficulties in accurately completing the reports. All data were independently entered by two researchers and subsequently cross-checked and logically verified. Any discrepancies were resolved by consulting the original records, thereby ensuring data accuracy and integrity.

Statistical Methods

SPSS 29.0 software was used for statistical analyses. Continuous variables with a non-normal distribution are presented as medians with interquartile ranges and were compared using non-parametric tests. Categorical data are expressed as numbers with percentages and were compared using the chi-square test. Binary logistic regression analysis was performed to identify independent risk factors and to construct a predictive model. Based on these factors, a nomogram was developed using R version 4.5.1. The discrimination of the model was evaluated by plotting the receiver operating characteristic (ROC) curve with the “pROC” and “ggplot2” packages, and the area under the curve (AUC) was calculated. The optimal cutoff value, along with the corresponding sensitivity and specificity, was determined. Calibration was assessed visually using a calibration curve and statistically with the Hosmer-Lemeshow goodness-of-fit test, both implemented with the “rms” package. A two-sided P value < 0.05 was considered statistically significant.

Results

Incidence of PONV in Patients with Liver Cancer

A total of 512 patients with liver cancer after surgery were included in this study, of which 243 patients had nausea and vomiting, and the incidence of nausea and vomiting was 47.5%.

Participant Characteristics

Chi-square test and non-parametric test showed that 12 variables such as age, gender, smoking, motion sickness, diabetes, and duration of surgery(min) were statistically significant (P<0.05), as detailed in Table 1.

Table 1.

Prevalence and Related Variables of Postoperative Nausea and Vomiting in Patients with Liver Cancer [n=512]

Variable		No Nausea and Vomiting n=269	Nausea and Vomiting n=243	χ2	P-value
Gender	Male	234(87)	151(62.1)	42.265a	<0.001
	Female	35(13)	92(37.9)
Age(year)	<60	109(40.5)	129(53.1)	8.104a	0.003
	≥60	160(59.5)	114(46.9)
Smoking	No	146(54.3)	163(67.1)	8.746a	0.002
	Yes	123(45.7)	80(32.9)
Motion sickness	No	246(91.4)	197(81.1)	11.797a	<0.001
	Yes	23(8.6)	46(18.9)
Hypertension	No	185(68.8)	175(72)	0.643a	0.240
	Yes	84(31.2)	68(28)
Diabetes	No	244(90.7)	208(85.6)	3.222a	0.049
	Yes	25(9.3)	35(14.4)
Migraine	No	254(94.4)	212(87.2)	8.051a	0.004
	Yes	15(5.6)	31(12.8)
BMI (kg/m2)	<24	127(47.2)	130(53.5)	2.018a	0.091
	≥24	142(52.8)	113(46.5)
Duration of surgery(min)	<120	132(49.1)	97(40.1)	4.161a	0.025
	≥120	137(50.9)	145(59.9)
Intraoperative fluid infusion volume(mL)	≤1000	99(36.8)	74(30.5)	2.301a	0.077
	>1000	170(63.2)	169(69.5)
Intraoperative occlusion of the hepatic portal vein	No	121(45)	72(29.6)	12.811a	<0.001
	Yes	148(55)	171(70.4)
Time of hepatic portal vein occlusion during operation(min)	<15	148(55)	92(37.9)	15.094a	<0.001
	≥15	121(45)	151(62.1)
History of PONV	No	265(98.5)	209(86)	29.055a	<0.001
	Yes	4(1.5)	34(14)
Anesthesia method	Endotracheal tube + epidural anesthesia	159(59.1)	152(62.6)	1.200a	0.549
	Endotracheal tube + nerve block	96(35.7)	76(31.3)
	Tracheal tube	14(5.2)	15(6.2)
Surgical method	Laparoscopic surgery	111(41.3)	96(39.5)	0.164a	0.377
	Open surgery	158(58.7)	147(60.5)
Liver cirrhosis	No	110(40.9)	129(53.1)	7.627a	0.004
	Yes	159(59.1)	114(46.9)

Note: ^arepresents the chi-square test.

Binary Logistic Regression Analysis of Factors Influencing the Occurrence of PONV in Patients with Liver Cancer

In this study, the presence or absence of nausea and vomiting was used as the dependent variable (0= No nausea and vomiting, 1= Nausea and vomiting), and the statistically significant variables in the univariate analysis were used as independent variables for logistic regression analysis (see Table 2 for details of variable allocation). Eleven variables were significantly associated with the risk of developing PONV at the significance level of P<0.05 and were subsequently included in the logistic regression model, as shown in Table 3. The model was further optimized by stepwise regression (excluding variables with P≥0.05).

Table 2.

Logistic Regression Analysis Factor Assignment

Variable	Assignment
Dependent variable (Y)	0 = No nausea and vomiting; 1 = Nausea and vomiting
Independent variable (X)
X1: Gender	0 = Male; 1 = Female
X2: Age(year)	0 = ≥60; 1 = <60
X3: Smoking	0 = No; 1 = Yes
X4: Motion sickness	0 = No; 1 = Yes
X5: Hypertension	0 = No; 1 = Yes
X6: Diabetes	0 = No; 1 = Yes
X7: Migraine	0 = No; 1 = Yes
X8: BMI	0 = <24;1 = ≥24
X9: Duration of surgery(min)	0 = <120; 1 = ≥120
X10: Intraoperative fluid infusion volume(mL)	0 = ≤1000; 1 = >1000
X11: Intraoperative occlusion of the hepatic portal vein	0 = No; 1 = Yes
X12: Time of hepatic portal vein occlusion during operation(min)	0 = <15; 1 = ≥15
X13: History of postoperative nausea and vomiting	0 = No; 1 = Yes
X14: Anesthesia method	0 = Endotracheal tube + epidural anesthesia; 1 = Endotracheal tube + nerve block; 2 = Tracheal tube
X15: Surgical method	0 = Laparoscopic surgery;1 = Open surgery
X16: Liver cirrhosis	0 = No; 1 = Yes

Table 3.

Logistic Regression Analysis of Influencing Factors of PONV in Patients with Liver Cancer

Variable	Regression Coefficient (β)	Standard Error (SE)	Wald χ2	P-value	OR	95% CI
Constant	−1.278	0.209	37.531	<0.001	0.279
Gender	1.288	0.237	29.569	<0.001	3.625	2.279–5.766
Age(year)	0.537	0.194	7.629	0.006	1.710	1.169–2.502
Duration of surgery(min)	0.398	0.196	4.116	0.042	1.490	1.014–2.189
Time of hepatic portal vein occlusion during operation (min)	0.532	0.195	7.470	0.006	1.703	1.162–2.494
History of postoperative nausea and vomiting	1.986	0.550	13.058	<0.001	7.288	2.482–21.405

Predictive Model Development

Based on the results of the binary logistic regression analysis, a risk prediction model for nausea and vomiting was constructed in R Studio (version 4.5.1) with 5 factors, namely age, gender, time of hepatic portal vein occlusion during operation (min), duration of surgery (min), and history of PONV, as shown in Figure 1.

Figure 1.

Nomogram of the predictive model.

Evaluation and Validation of the Model

Discriminability Analysis

The results showed that the C-index (AUC value) of the model was 0.717 (95% CI: 0.673–0.761) (P<0.001). At the optimal cut-off value of 0.413, the sensitivity and specificity of the model were 69.9% and 62.0%, respectively, as shown in Figure 2 and Table 4. Given that the model was developed based on the training set, internal validation was performed using the Bootstrap resampling method with 1000 repetitions. After internal validation, the C-index (AUC value) of the model was 0.717, confirming that the model still possessed a certain degree of discriminative ability in the Bootstrap samples.

Figure 2.

ROC Curve.

Table 4.

Analysis of the Diagnostic Value of the Model

AUC	95% CI	P-value	Cut-off	Sensitivity	Specificity	PPV	NPV
0.717	0.673–0.761	P<0.001	0.413	0.699	0.620	0.671	0.649

Calibration Analysis

The calibration curve was adopted to evaluate the calibration ability of the model (as shown in Figure 3). The results of the Hosmer-Lemeshow test showed that χ²=7.032, df=7, P=0.426, P>0.05, indicating that there was no significant difference between the predicted probability and the actual probability of the endpoint event. After internal validation using the Bootstrap resampling method with 1000 repetitions, the absolute error was calculated as 0.017. As illustrated in Figure 3, the calibrated curve was closely aligned with both the ideal curve and the original curve, with its slope approximating 1. Therefore, it was concluded that there was a good consistency between the actual and predicted values of the endpoint event occurrence.

Figure 3.

Calibration Curve.

Evaluation of Clinical Utility

The decision curve analysis (DCA) was used to assess the clinical utility of the model (as shown in Figure 4). The results demonstrated that the model had a wide threshold probability interval. When the threshold probability ranged from 23% to 99%, the application of this model yielded a relatively high clinical net benefit, which was significantly higher than that of the None line and the All line, indicating that the model had good clinical application value. Additionally, as presented in Figure 4, the cut-off value (0.413) derived from the aforementioned ROC analysis fell within the threshold probability range of the decision curve of this predictive model, thus confirming that the predictive model possessed clinical validity.

Figure 4.

Clinical Decision Curve.

Discussion

The Incidence of PONV in Patients with Liver Cancer Is Relatively High

Surgical intervention represents the primary treatment modality for liver cancer. PONV are common adverse events following hepatic surgery, exerting substantial negative impacts on both physical and psychological recovery and imposing a considerable burden on clinical and public health systems.¹⁴ Despite their significant prognostic implications, PONV remains under-assessed in routine clinical practice. Consequently, enhancing awareness of PONV and strengthening postoperative symptom management in liver cancer patients is imperative. Early identification of high-risk individuals is essential, and timely interventions should be implemented to mitigate PONV occurrence, thereby improving patient outcomes and treatment efficacy.¹⁵

Analysis of Risk Factors Associated with PONV in Patients with Liver Cancer

This study identified several clinically significant factors associated with an elevated risk of PONV in patients undergoing hepatic resection for liver cancer. The analysis revealed that female sex, age <60 years, a documented history of previous PONV, prolonged operative duration, and a hepatic portal vein occlusion time ≥15 minutes were independently correlated with a higher incidence of PONV. Notably, a history of previous PONV emerged as the most robust predictor among the evaluated variables.¹⁶ This finding aligns with established literature across surgical disciplines and underscores the role of patient-specific physiological predisposition, potentially mediated by genetic polymorphisms in neurotransmitter receptors or inherent sensitivity of the vestibular and chemoreceptor trigger zones.^17,18

The identification of hepatic portal vein occlusion time ≥15 minutes as a significant risk factor highlights a procedure-specific dimension in liver surgery. Prolonged portal triad clamping may induce splanchnic venous congestion and transient mesenteric hypoperfusion,¹⁹ leading to local ischemia-reperfusion injury and subsequent release of pro-inflammatory cytokines and serotonin-key mediators implicated in the pathogenesis of nausea and vomiting.⁷ Furthermore, this maneuver can exacerbate postoperative hepatic dysfunction, potentially altering the metabolism of anesthetic agents and opioids, thereby prolonging their emetogenic effects.²⁰

Longer operative duration was also significantly associated with increased PONV risk,²¹ likely reflecting the compounded effects of extended exposure to anesthetic and analgesic agents, greater surgical trauma, fluid shifts, and systemic inflammatory activation. Concurrently, the observed associations with female sex^22,23 and age <60 years²⁴ are consistent with the well-characterized influence of sex hormones on neuroendocrine pathways and age-related differences in pharmacodynamics and autonomic regulation.

From a clinical perspective, the practical application of this prediction model can be integrated into routine perioperative care pathways.²⁵ For patients identified as high-risk by the nomogram (eg, those with a predicted probability above the optimal cutoff value of 0.413), clinicians could consider a stepwise preventive strategy.²⁶ During the intraoperative phase, anesthesiologists might opt for total intravenous anesthesia with propofol instead of volatile anesthetics, which are known to have emetogenic properties.²⁷ Additionally, adopting opioid-sparing techniques-such as the use of non-steroidal anti-inflammatory drugs or regional analgesia-could reduce opioid-related PONV.²⁸ For prophylactic antiemesis, high-risk patients could receive combination therapy with agents from different pharmacological classes, including 5-HT₃receptor antagonists (eg, ondansetron), neurokinin-1 receptor antagonists (eg, aprepitant), and dexamethasone, which target different emetic pathways and have been shown to provide additive or synergistic effects.²⁴ In the postoperative period, enhanced monitoring and early administration of rescue antiemetics at the first sign of nausea could be implemented for high-risk patients, along with non-pharmacological measures such as acupoint stimulation (eg, P6 acupressure) or early mobilization and oral intake protocols as tolerated.²⁹ By stratifying patients based on their individual risk profiles, this model enables a more personalized and resource-efficient approach to PONV prevention, potentially reducing overall incidence and improving patient satisfaction and recovery outcomes.^25,26

Clinical Utility of Risk Prediction Models

This study is one of the earliest to establish a predictive model to assess the risk of PONV in patients with liver cancer. To visualize the model and facilitate the provision of risk probabilities to patients, a nomogram was developed. Model validation typically requires assessing the discrimination ability and accuracy of the model. The accuracy of the risk prediction model has a significant impact on the selection of preventive measures and the effectiveness of intervention. This study evaluated the calibration ability of the predictive model using the ROC curve and the Hosmer-Lemeshow test. The results showed that the area under the ROC curve was 0.717, the sensitivity was 69.9%, and the specificity was 62%. These results indicate that the model has good discrimination ability and accuracy, which is helpful in identifying individuals at high risk of PONV. During internal validation, the C value was 0.717, which was greater than 0.7 indicating a good model resolution. Additionally, the uncorrected and corrected calibration curves were consistent with the reference line in the classification calibration curve, indicating that the named graph has a high calibration ability. Therefore, this risk prediction model is applicable to postoperative patients with liver cancer and can provide personalized, high-precision, and quantifiable nausea and vomiting risk assessment for these patients with clinical applicability. Due to the influence of multiple factors on nausea and vomiting, the model developed in this study covers personal general conditions (age, gender), disease history (history of PONV), and surgical intraoperative conditions (time of hepatic portal vein occlusion during operation, duration of surgery). Medical professionals can use this model to assess the risk of accidental nausea and vomiting in liver cancer patients and implement timely preventive and nursing measures through early identification of high-risk groups, aiming to achieve early prevention and intervention and reduce the incidence of PONV. While the present study focuses on risk prediction rather than intervention efficacy, identifying high-risk patients through this model can nonetheless support clinical decision-making regarding the intensity of prophylactic antiemetic regimens and the selection of anesthetic techniques. Future studies are warranted to evaluate whether such model-guided preventive strategies ultimately translate into measurable improvements in patient outcomes.

Limitations

This study has several limitations that should be acknowledged. First, the study cohort was derived from only two tertiary medical centers in Shanghai, China. Therefore, the generalizability of the findings may be limited, as variations in patient demographics, surgical protocols, and perioperative care across different geographic and healthcare settings could affect the model’s external validity. Second, the sample size, while sufficient for model development, permitted only bootstrap internal validation. External validation in independent, multi-center cohorts is therefore essential to confirm the model’s robustness, calibration, and clinical applicability before it can be recommended for widespread use in the diverse population of patients undergoing liver cancer surgery. Future studies should focus on prospective, multi-institutional validation to refine the model and establish its general predictive performance.

Conclusion

This study has confirmed that the incidence of PONV in patients undergoing liver cancer surgery remains high (47.5%), underscoring the need for enhanced clinical awareness. The nomogram developed herein, incorporating five readily available predictors (age, gender, duration of surgery, time of hepatic portal vein occlusion during operation, and history of PONV), demonstrates acceptable discrimination (AUC=0.717) and good calibration, providing a practical tool for early risk stratification. From a clinical perspective, this model enables healthcare providers to identify high-risk patients preoperatively and guide individualized perioperative strategies—including optimized multimodal antiemetic prophylaxis, total intravenous anesthesia, and opioid-sparing analgesic techniques—to reduce PONV incidence and improve postoperative recovery. However, given that the model has undergone only internal validation, further external validation in independent, multi-center cohorts is essential before it can be recommended for widespread clinical implementation.

Funding Statement

This study has been supported by the Shanghai Municipal Health Commission (Grant no. 20224Y0288), Fudan University Fuxing Nursing Research Funding (Grant no. FNF202107) and the Youth Medical Talents – Nursing Program (Grant no. 2026-071). The funding sources had no role in the design of this study, its execution, data analyses, interpretation of the data, report writing, or the decision to submit the article for publication.

Abbreviations

PONV, Postoperative nausea and vomiting; AUC-ROC, Area under the receiver operating characteristic curve; C-index, Consistency index; BMI, Body mass index; DCA, Decision curve analysis.

Data Sharing Statement

The datasets generated and/or analyzed during the current study are available from the corresponding author, Xiao CHEN, upon reasonable request.

Ethics Approval and Informed Consent

All procedures performed in this study involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards. The study protocol was reviewed and approved by the Institutional Review Board of Zhongshan Hospital, Fudan University (approval number: B2020-074). Written informed consent was obtained from all individual participants included in the study.

Consent for Publication

Written informed consent for publication of identifying information/images (if applicable) was obtained from all participating individuals. The participants have reviewed and approved the final version of the manuscript prior to submission. Signed consent forms are available from the corresponding author upon request by the journal editorial office.

Author Contributions

All authors made a significant contribution to the work reported, whether that is in the conception, study design, execution, acquisition of data, analysis and interpretation, or in all these areas; took part in drafting, revising or critically reviewing the article; gave final approval of the version to be published; have agreed on the journal to which the article has been submitted; and agree to be accountable for all aspects of the work.

Disclosure

All authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this manuscript.