Development and validation of a novel nomogram to predict cancer-specific survival in patients with uterine cervical adenocarcinoma

Integrating MRI-based radiomics and clinicopathological features for preoperative prognostication of early-stage cervical adenocarcinoma patients: in comparison to deep learning approach

OBJECTIVES

The roles of magnetic resonance imaging (MRI) -based radiomics approach and deep learning approach in cervical adenocarcinoma (AC) have not been explored. Herein, we aim to develop prognosis-predictive models based on MRI-radiomics and clinical features for AC patients.

METHODS

Clinical and pathological information from one hundred and ninety-seven patients with cervical AC was collected and analyzed. For each patient, 107 radiomics features were extracted from T2-weighted MRI images. Feature selection was performed using Spearman correlation and random forest (RF) algorithms, and predictive models were built using support vector machine (SVM) technique. Deep learning models were also trained with T2-weighted MRI images and clinicopathological features through Convolutional Neural Network (CNN). Kaplan-Meier curve was analyzed using significant features. In addition, information from another group of 56 AC patients was used for the independent validation.

RESULTS

A total of 107 radiomics features and 6 clinicopathological features (age, FIGO stage, differentiation, invasion depth, lymphovascular space invasion (LVSI), and lymph node metastasis (LNM) were included in the analysis. When predicting the 3-year, 4-year, and 5-year DFS, the model trained solely on radiomics features achieved AUC values of 0.659 (95%CI: 0.620-0.716), 0.791 (95%CI: 0.603-0.922), and 0.853 (95%CI: 0.745-0.912), respectively. However, the combined model, incorporating both radiomics and clinicopathological features, outperformed the radiomics model with AUC values of 0.934 (95%CI: 0.885-0.981), 0.937 (95%CI: 0.867-0.995), and 0.916 (95%CI: 0.857-0.970), respectively. For deep learning models, the MRI-based models achieved an AUC of 0.857, 0.777 and 0.828 for 3-year DFS, 4-year DFS and 5-year DFS prediction, respectively. And the combined deep learning models got a improved performance, the AUCs were 0.903. 0.862 and 0.969. In the independent test set, the combined model achieved an AUC of 0.873, 0.858 and 0.914 for 3-year DFS, 4-year DFS and 5-year DFS prediction, respectively.

CONCLUSIONS

We demonstrated the prognostic value of integrating MRI-based radiomics and clinicopathological features in cervical adenocarcinoma. Both radiomics and deep learning models showed improved predictive performance when combined with clinical data, emphasizing the importance of a multimodal approach in patient management.

Analysis of prognostic factors for cervical mucinous adenocarcinoma and establishment and validation a nomogram: a SEER-based study

Up to now, there are no relevant studies on prognostic factors of cervical mucinous adenocarcinoma. Therefore, we explored the prognostic factors for cervical mucinous adenocarcinoma, and established and validated the prognostic model using the SEER database. We selected the independent factors through univariate and multivariate analyses. LASSO regression analysis was conducted to identify potential risk factors. In conjunction with LASSO and multivariate analysis, the nomogram incorporated three variables, including age, tumour size, and AJCC stage for OS. The c-index was 0.794 and 0.831 in development and validated cohorts, indicating that this prediction model showed adequate discriminative ability in the development cohort. Besides, calibration curves showed good concordance for the development cohort, as well as the validation cohort. We constructed a first-of-its-kind nomogram to predict cervical mucinous adenocarcinomas OS and it showed better performance than AJCC and FIGO stages. Patients with cervical mucinous adenocarcinoma might benefit from using this model to develop tailored treatments.IMPACT STATEMENTWhat is already known on this subject? Cervical cancer has a variety of pathological types. The biological behaviour of each type is different, and the prognosis is quite different.What do the results of this study add? We analysed and explored the relevant factors affecting the prognosis of cervical mucinous adenocarcinoma.What are the implications of these findings for clinical practice and/or further research? Through the analysis of the SEER dataset, the prognostic factors affecting cervical mucinous adenocarcinoma were identified, and the first predictive model was created to predict the prognosis to help doctors develop individualised treatment plans and follow-up plans.

Development and validation of a nomogram to predict overall survival for cervical adenocarcinoma: A population-based study

This study aimed to develop and validate a nomogram for predicting the overall survival of cervical adenocarcinoma (CAC) patients using a large database comprising patients with different ethnicities. We enrolled primary CAC cases with complete clinicopathological and survival data from the Surveillance, Epidemiology, and End Results program during 2004 to 2015. For training set samples, this work applied the Cox regression model to obtain factors independently associated with patient prognosis, which could be incorporated in constructing the nomogram. Altogether 3096 qualified cases were enrolled, their survival ranged from 0 to 155 (median, 45.5) months. As revealed by multivariate regression, age, marital status, tumor size, grade, International Federation of Gynecology and Obstetrics (FIGO) classification, pelvic lymph node metastasis, surgery, and chemotherapy served as the factors to independently predict CAC (all P < .05). We later incorporated these factors for constructing the nomogram. According to the concordance index determined, this nomogram had superior discrimination over FIGO classification system (all P < .001). Based on calibration plot, the predicted value was consistent with actual measurement. As revealed by time-independent area under the curves, our constructed nomogram had superior 5-year overall survival over FIGO system. Additionally, according to decision curve analysis, our constructed nomogram showed high clinical usefulness as well as favorable discrimination. Our constructed nomogram attains favorable performances, indicating that it may be applied in predicting survival for CAC patients.

Establishment and validation of prognostic nomograms integrating histopathological features in patients with endocervical adenocarcinoma

AIMS

To develop and verify pathological models using pathological features basing on HE images to predict survival invasive endocervical adenocarcinoma (ECA) postoperatively.

METHODS

There are 289 ECA patients were classified into training and validation cohort. A histological signature was produced in 191 patients and verified in the validation groups. Histological models combining the histological features were built, proving the incremental value of our model to the traditional staging system for individualised prognosis estimation.

RESULTS

Our model included five chosen histological characteristics and was significantly related to overall survival (OS). Our model had AUC of 0.862 and 0.955, 0.891 and 0.801 in prognosticating 3-year and 5 year OS in the training and validation cohort, respectively. In training cohorts, our model had better performance for evaluation of OS (C-index: 0.832; 95% CI 0.751 to 0.913) than International Federation of Gynecology and Obstetrics (FIGO) staging system (C-index: 0.648; 95% CI 0.542 to 0.753) and treatment (C-index: 0.687; 95% CI 0.605 to 0.769), with advanced efficiency of the classification of survival outcomes. Furthermore, in both cohorts, a risk stratification system was built that was able to precisely stratify stage I and II ECA patients into high-risk and low-risk subpopulation with significantly different prognosis.

CONCLUSIONS

A nomogram with five histological signatures had better performance in OS prediction compared with traditional staging systems in ECAs, which might enable a step forward to precision medicine.

Development and validation of nomograms to recurrence and survival in patients with early-stage cervical adenocarcinoma

PURPOSE

Cervical adenocarcinoma is one of the most common types of cervical cancer and its incidence is increasing. The biological behavior and treatment outcomes of cervical adenocarcinoma (CA) differ from those of squamous cell carcinoma (SCC). We sought to develop a model to predict recurrence and cancer-specific survival (CSS) deaths in CA patients.

METHODS

131 patients were included in model development and internal validation, and patients from the SEER database (N = 1679) were used for external validation. Multivariable Cox proportional hazards regression analysis was used to select predictors of relapse-free survival (RFS) and CSS and to construct the model, which was presented as two nomograms. Internal validation of the nomograms was performed using the bootstrap resampling method.

RESULTS

Age, FIGO (International Federation of Gynecology and Obstetrics) stage, size of the tumor, lymph metastasis and depth of invasion were identified as independent prognostic factors for RFS, while age, FIGO stage, size of the tumor and number of positive LNs were identified as independent prognostic factors for CSS. The nomogram of the recurrence model predicted 2- and 5-year RFS, with optimism adjusted c-statistic of 75.41% and 74.49%. Another nomogram predicted the 2- and 5-year CSS with an optimism-adjusted c-statistic of 83.22% and 83.31% after internal validation; and 68.6% and 71.33% after external validation.

CONCLUSIONS

We developed and validated two effective nomograms based on static nomograms or online calculators that can help clinicians quantify the risk of relapse and death for patients with early-stage CA.

A Bayesian network predicting survival of cervical cancer patients-Based on surveillance, epidemiology, and end results

This study aimed to build a comprehensive model for predicting the overall survival (OS) of cervical cancer patients who received standard treatments and to build a series of new stages based on the International Federation of Gynecologists and Obstetricians (FIGO) stages for better such predictions. We collected the cervical cancer patients diagnosed since the year 2000 from the Surveillance, Epidemiology, and End Results (SEER) database. Cervical cancer patients who received radiotherapy or surgery were included. Log-rank tests and Cox regression were used to identify potential factors of OS. Bayesian networks (BNs) were built to predict 3- and 5-year survival. We also grouped the patients into new stages by clustering their 5-year survival probabilities based on FIGO stage, age, and tumor differentiation. Cox regression suggested black ethnicity, adenocarcinoma, and single status as risks for poorer prognosis, in addition to age and stage. A total of 43,749 and 39,333 cases were finally eligible for the 3- and 5-year BNs, respectively, with 11 variables included. Cluster analysis and Kaplan-Meier curves indicated that it was best to divide the patients into nine modified stages. The BNs had excellent performance, with area under the curve and maximum accuracy of 0.855 and 0.804 for 3-year survival, and 0.851 and 0.787 for 5-year survival, respectively. Thus, BNs are excellent candidates for predicting cervical cancer survival. It is necessary to consider age and tumor differentiation when estimating the prognosis of cervical cancer using FIGO stages.

Construction and validation of a predictive model for the risk of three-month-postoperative malnutrition in patients with gastric cancer: a retrospective case-control study

Background

This study analyzed both the influencing factors of malnutrition in patients with gastric cancer and established a multi-dimensional risk model to predict postoperative malnutrition three months after surgery.

Methods

The clinical data of gastric cancer patients hospitalized for the first time and receiving laparoscopic surgery in the general surgery department of our hospital were retrospectively analyzed through the hospital information system and divided into a training set and a validation set in the ratio of 7:3. Nutritional status was assessed using the Patient Generated Subjective Global Assessment scale and follow-up records three months after surgery. Patients were divided into a non-malnutrition group and a malnutrition group, and a risk prediction model was established and displayed in the form of a nomogram.

Results

A total of 344 patients were included, with 242 in the training and 102 in the validation set. Tumor node metastasis stage (TNM Stage, P=0.020), cardiac function grading (CFG, P=0.013), prealbumin (PAB, P<0.001), neutrophil-to-lymphocyte ratio (NLR, P=0.027), and enteral nutrition within 48 hours post-operation (EN 48 h post-op, P=0.025) were independent risk factors. We established a prediction model with the above variables and displayed it via a nomogram, then verified its effectiveness through internal and external verification. This revealed a C-index of 0.84 (95% CI: 0.79-0.89), and the area under curve (AUC) areas of 0.840 (training set) and 0.854 (validation set), which was better than the nutritional risk screening 2002 (NRS2002) scale. The calibration curve brier scores were 0.159 and 0.195, and the Hosmer-Lemeshow test chi-square values were 14.070 and 1.989 (P>0.05). The decision curve analysis (DCA) of the training set model indicated the clinical applicability was good and within the threshold probability range of 10%-85%, which was also better than NRS2002.

Conclusions

A clinical prediction model including multi-dimensional variables was established based on independent risk factors of malnutrition three months after gastrectomy in patients with gastric cancer. The model yields greater prediction accuracy of the risk of three-month-postoperative malnutrition in patients with gastric cancer, helps screen high-risk patients, formulates targeted nutritional prescriptions early, and improves the overall prognosis of patients.

Construction and validation of a novel nomogram to predict cancer-specific survival in patients with gastric adenocarcinoma

Background and aims

Adenocarcinoma is one of the most common pathological types of gastric cancer. The aims of this study were to develop and validate prognostic nomograms that could predict the probability of cancer-specific survival (CSS) for gastric adenocarcinoma (GAC) patients at 1, 3, and 5 years.

Methods

In total, 7747 patients with GAC diagnosed between 2010 and 2015, and 4591 patients diagnosed between 2004 and 2009 from the Surveillance, Epidemiology, and End Results (SEER) database were included in this study. The 7747 patients were used as a prognostic cohort to explore GAC-related prognostic risk factors. Moreover, the 4591 patients were used for external validation. The prognostic cohort was also divided into a training and internal validation sets for construction and internal validation of the nomogram. CSS predictors were screened using least absolute shrinkage and selection operator regression analysis. A prognostic model was built using Cox hazard regression analysis and provided as static and dynamic network-based nomograms.

Results

The primary site, tumor grade, surgery of the primary site, T stage, N stage, and M stage were determined to be independent prognostic factors for CSS and were subsequently included in construction of the nomogram. CSS was accurately estimated using the nomogram at 1, 3, and 5 years. The areas under the curve (AUCs) for the training group at 1, 3, and 5 years were 0.816, 0.853, and 0.863, respectively. Following internal validation, these values were 0.817, 0.851, and 0.861. Further, the AUC of the nomogram was much greater than that of American Joint Committee on Cancer (AJCC) or SEER staging. Moreover, the anticipated and actual CSS values were in good agreement based on decision curves and time-calibrated plots. Then, patients from the two subgroups were divided into high- and low-risk groups based on this nomogram. The survival rate of high-risk patients was considerably lower than that of low-risk patients, according to Kaplan-Meier (K-M) curves (p<0.0001).

Conclusions

A reliable and convenient nomogram in the form of a static nomogram or an online calculator was constructed and validated to assist physicians in quantifying the probability of CSS in GAC patients.

Conditional survival analysis and real-time prognosis prediction for cervical cancer patients below the age of 65 years

Background

Survival prediction for cervical cancer is usually based on its stage at diagnosis or a multivariate nomogram. However, few studies cared whether long-term survival improved after they survived for several years. Meanwhile, traditional survival analysis could not calculate this dynamic outcome. We aimed to assess the improvement of survival over time using conditional survival (CS) analysis and developed a novel conditional survival nomogram (CS-nomogram) to provide individualized and real-time prognostic information.

Methods

Cervical cancer patients were collected from the Surveillance, Epidemiology, and End Results (SEER) database. The Kaplan-Meier method estimated cancer-specific survival (CSS) and calculated the conditional CSS (C-CSS) at year y+x after giving x years of survival based on the formula C-CSS(y|x) =CSS(y+x)/CSS(x). y indicated the number of years of further survival under the condition that the patient was determined to have survived for x years. The study identified predictors by the least absolute shrinkage and selection operator (LASSO) regression and used multivariate Cox regression to demonstrate these predictors' effect on CSS and to develop a nomogram. Finally, the CSS possibilities predicted by the nomogram were brought into the C-CSS formula to create the CS-nomogram.

Results

A total of 18,511 patients aged <65 years with cervical cancer from 2004 to 2019 were included in this study. CS analysis revealed that the 15-year CSS increased year by year from the initial 72.6% to 77.8%, 84.5%, 88.8%, 91.5%, 93.5%, 94.8%, 95.7%, 96.4%, 97.3%, 98.0%, 98.5%, 99.1%, and 99.4% (after surviving for 1-13 years, respectively), and found that when survival exceeded 5-6 years, the risk of death from cervical cancer would be less than 5% in 10-15 years. The CS-nomogram constructed using tumor size, lymph node status, distant metastasis status, and histological grade showed strong predictive performance with a concordance index (C-index) of 0.805 and a stable area under the curve (AUC) between 0.795 and 0.816 over 15 years.

Conclusions

CS analysis in this study revealed the gradual improvement of CSS over time in long-term survived cervical cancer patients. We applied CS to the nomogram and developed a CS-nomogram successfully predicting individualized and real-time prognosis.

Establishment and validation of nomograms to predict survival probability of advanced malignant pleural mesothelioma based on the SEER database and a Chinese medical institution

Objective

The purpose of this study was to build nomograms for predicting the survival of individual advanced pleural mesothelioma (MPM) patients using the Surveillance, Epidemiology, and End Results (SEER) database.

Methods

The 1251 patients enrolled from the SEER database were randomized (in a 7:3 ratio) to a training cohort and an internal validation cohort. Eighty patients were enrolled from the Harbin Medical University Cancer Hospital as the external validation cohort. Nomograms were constructed from variables screened by univariate or multivariate Cox regression analyses and evaluated by consistency indices (C-index), calibration plots, and receiver operating characteristic (ROC) curves. Patients from the SEER database who received chemotherapy alone and chemoradiotherapy were statistically paired using propensity score matching of the two groups and performed subgroup analysis in the screened variables.

Results

The nomograms are well-structured and well-validated prognostic maps constructed from four variables: gender, histology, AJCC stage, and treatment. All individuals were allocated into high-risk versus low-risk groups based on the median risk score of the training cohort, with the high-risk group having worse OS and CSS in all three cohorts (P<0.05). The outcomes of the subgroup analysis indicated that the advanced MPM patients receiving chemotherapy with or without local radiotherapy do not affect OS or CSS.

Conclusion

The accurate nomograms to predict the survival of patients with advanced MPM were built and validated based on an analysis of the SEER database with an external validation cohort. The study suggests that the additional local radiotherapy to chemotherapy does not increase the survival benefit of patients.

Establishment of predictive nomogram and web-based survival risk calculator for malignant pleural mesothelioma: A SEER database analysis

Background

Malignant pleural mesothelioma (MPM) is an uncommon condition with limited available therapies and dismal prognoses. The purpose of this work was to create a multivariate clinical prognostic nomogram and a web-based survival risk calculator to forecast patients’ prognoses.

Methods

Using a randomization process, training and validation groups were created for a retrospective cohort study that examined the Surveillance, Epidemiology, and End Results (SEER) database from 2010 to 2015 for individuals diagnosed with MPM (7:3 ratio). Overall survival (OS) and cancer-specific survival (CSS) were the primary endpoints. Clinical traits linked to OS and CSS were identified using Least Absolute Shrinkage and Selection Operator (LASSO) Cox regression analysis, which was also utilized to develop nomogram survival models and online survival risk calculators. By charting the receiver operating characteristic (ROC), consistency index (C-index), calibration curve, and decision curve analysis (DCA), the model’s performance was assessed. The nomogram was used to classify patients into various risk categories, and the Kaplan-Meier method was used to examine each risk group’s survival rate.

Results

The prognostic model comprised a total of 1978 patients. For the total group, the median OS and CSS were 10 (9.4-10.5) and 11 (9.4-12.6) months, respectively. As independent factors for OS and CSS, age, gender, insurance, histology, T stage, M stage, surgery, and chemotherapy were chosen. The calibration graphs demonstrated good concordance. In the training and validation groups, the C-indices for OS and CSS were 0.729, 0.717, 0.711, and 0.721, respectively. Our nomogram produced a greater clinical net benefit than the AJCC 7th edition, according to DCA and ROC analysis. According to the cut-off values of 171 for OS and 189 for CSS of the total scores from our nomogram, patients were classified into two risk groups. The P-value &lt; 0.001 on the Kaplan-Meier plot revealed a significant difference in survival between the two patient groups.

Conclusions

Patient survival in MPM was correctly predicted by the risk evaluation model. This will support clinicians in the practice of individualized medicine.

Nomogram Predicts Overall Survival in Patients With Stage IV Thyroid Cancer (TC): A Population-Based Analysis From the SEER Database

Background

Stage IV Thyroid cancer (TC) has a relatively poor prognosis and lacks a precise and efficient instrument to forecast prognosis. Our study aimed to construct a nomogram for predicting the prognosis of patients with stage IV TC based on data from the SEER programme.

Methods

We enrolled patients diagnosed with TC from 2004 to 2015 in the study. Furthermore, the median survival time (MST) for the patients equalled 25 months. The patients were split into two groups: the training group and validation group. We used descriptive statistics to calculate demographic and clinical variables, Student’s t test was used to describe continuous variables, and the chi-square test was used to describe classified variables. We used the concordance index (C-index) to evaluate discrimination ability and calibration plots to evaluate calibration ability. The improvement of the nomogram compared with the AJCC TNM system was evaluated by the net weight classification index (NRI), comprehensive discriminant rate improvement (IDI) and decision curve analysis (DCA).

Results

There were 3501 patients contained within our cohort, and the median follow-up was 25 months [quartile range (IQR): 6-60] in the whole population, 25 months (IQR: 6-60) in the training cohort, and 25 months (IQR: 5-59) in the validation cohort. The C-index value of the training cohort equalled 0.86 (95% CI: 0.85-0.87), and the value of the validation cohort equalled 0.85 (95% CI: 0.84-0.86). The NRI values were as follows: training queue: 1.16 for three-year and 1.12 for five-year OS prediction; authentication group: 1.22 for three-year and 1.21 for five-year OS prediction. The IDI values were as follows: training cohort: 0.25 for three-year and 0.21 for five-year OS prediction; validation cohort: 0.27 for three-year and 0.21 for five-year OS prediction. The DCA diagram showed that the nomogram was superior in predicting the three-year and five-year trends.

Conclusions

Our nomogram can be used to forecast the survival of patients with stage IV TC.

Serum Calprotectin Level Is Independently Associated With Carotid Plaque Presence in Patients With Psoriatic Arthritis

Background

Whether calprotectin could play a role in augmenting cardiovascular (CV) risk in patients with psoriatic arthritis (PsA) remains uncertain. The aim of this study is to elucidate the association between serum calprotectin level and subclinical atherosclerosis in patient with PsA.

Method

Seventy-eight PsA patients (age: 52 ± 10 years, 41 [52.6%] male) without CV disease were recruited into this cross-sectional study. Carotid intima-media thickness (cIMT) and the presence of plaque were determined by high-resolution ultrasound. Calprotectin levels in serum were quantified by enzyme-linked immunosorbent assay. The variables associated with the presence of carotid plaque (CP) were selected from the least absolute shrinkage and selection operator (LASSO) regression analysis.

Results

29/78 (37.2%) of patient had carotid plaque (CP+ group). Serum calprotectin level was significantly higher in the CP+ group (CP− group: 564.6 [329.3–910.5] ng/ml; CP+ group: 721.3 [329.3–910.5] ng/ml, P = 0.005). Serum calprotectin level correlated with PsA disease duration (rho = 0.280, P = 0.013) and mean cIMT (rho = 0.249, P = 0.038). Using LASSO regression analysis, the levels of Ln-calprotectin (OR: 3.38, 95% CI [1.37, 9.47]; P = 0.026) and PsA disease duration (OR: 1.09, 95% CI [1.01, 1.18]; P = 0.013) were screened out from a total of 19 variables. The model in predicting the presence of CP was constructed by Ln-calprotectin and PsA disease duration with an area under the receiver-operating characteristic (ROC) curve of 0.744, (95 CI% [0.59, 0.80], P = 0.037).

Conclusion

Serum calprotectin level is associated with the presence of CP in PsA. Further studies are required to confirm whether this pathway is associated with CV events in PsA.