Decision curve analysis to evaluate the clinical benefit of prediction models

Predicting complications of spine surgery: external validation of three models

BACKGROUND CONTEXT

Numerous prediction tools are available for estimating postoperative risk following spine surgery. External validation and comparison of these tools is critical prior to clinical use. No model for adverse events after spine surgery has undergone decision curve analysis.

PURPOSE

External validation, comparison, and decision curve analysis of 3 previously described models [SpineSage, Risk Assessment Tool (RAT), National Surgical Quality Improvement Program Risk Calculator (NSQIP)] for predicting 30-day postoperative complications after spine surgery STUDY DESIGN: Retrospective cohort study.

PATIENT SAMPLE

Three hundred fifteen patients who underwent spine surgery at a tertiary academic surgical center in New Zealand between January 2019 and April 2020.

OUTCOME MEASURES

As defined by each risk prediction tool and objectively using the Comprehensive Complication Index.

METHODS

We retrospectively reviewed risk of postoperative complication was calculated for each patient according to the 3 models. Overall model fit, calibration, discrimination, and decision curve analysis for each model were assessed in line with the transparent reporting of a multivariable prediction model for individual prognosis or diagnosis (TRIPOD) guidelines.

RESULTS

100 (35%) patients experienced complications. SpineSage and RAT were well calibrated, NSQIP systematically underestimated risk. Area under the curve was greatest for SpineSage (0.75) compared with the NSQIP (0.72) and the RAT (0.69). Decision curve analysis showed SpineSage resulted in greatest net benefit across all risk thresholds.

CONCLUSIONS

Of the models studied, SpineSage most accurately predicted risk and can be expected to perform better than a strategy of treating all patients if patient or surgeon deem complication risk >10% significant. NSQIP may not be suitable for the clinical use in our local population.

Multidimensional characteristics, prognostic role, and preoperative prediction of peritoneal sarcomatosis in retroperitoneal sarcoma

Background

Peritoneal sarcomatosis (PS) could occur in patients with retroperitoneal sarcomas (RPS). This study aimed to expand the understanding of PS on its characteristics and prognostic role, and develop a nomogram to predict its occurrence preoperatively.

Methods

Data of 211 consecutive patients with RPS who underwent surgical treatment between 2011 and 2019 was retrospectively reviewed. First, the clinicopathological characteristics of PS were summarized and analyzed. Second, the disease-specific survival (DSS) and recurrence-free survival (RFS) of patients were analyzed to evaluate the prognostic role of PS. Third, preoperative imaging, nearly the only way to detect PS preoperatively, was combined with other screened risk factors to develop a nomogram. The performance of the nomogram was assessed.

Results

Among the 211 patients, 49 (23.2%) patients had PS with an incidence of 13.0% in the primary patients and 35.4% in the recurrent patients. The highest incidence of PS occurred in dedifferentiated liposarcoma (25.3%) and undifferentiated pleomorphic sarcoma (25.0%). The diagnostic sensitivity of the preoperative imaging was 71.4% and its specificity was 92.6%. The maximum standardized uptake value (SUVmax) was elevated in patients with PS (P<0.001). IHC staining for liposarcoma revealed that the expression of VEGFR-2 was significantly higher in the PS group than that in the non-PS group (P = 0.008). Survival analysis (n =196) showed significantly worse DSS in the PS group than in non-PS group (median: 16.0 months vs. not reached, P < 0.001). In addition, PS was proven as one of the most significant prognostic predictors of both DSS and RFS by random survival forest algorithm. A nomogram to predict PS status was developed based on preoperative imaging combined with four risk factors including the presentation status (primary vs. recurrent), ascites, SUVmax, and tumor size. The nomogram significantly improved the diagnostic sensitivity compared to preoperative imaging alone (44/49, 89.8% vs. 35/49, 71.4%). The C-statistics of the nomogram was 0.932, and similar C-statistics (0.886) was achieved at internal cross-validation.

Conclusion

PS is a significant prognostic indicator for RPS, and it occurs more often in recurrent RPS and in RPS with higher malignant tendency. The proposed nomogram is effective to predict PS preoperatively.

18F-FDG PET/CT radiomics nomogram for predicting occult lymph node metastasis of non-small cell lung cancer

Purpose

To investigate the ability of a PET/CT-based radiomics nomogram to predict occult lymph node metastasis in patients with clinical stage N0 non-small cell lung cancer (NSCLC).

Materials and methods

This retrospective study included 228 patients with surgically confirmed NSCLC (training set, 159 patients; testing set, 69 patients). ITKsnap3.8.0 was used for image(CT and PET images) segmentation, AK version 3.2.0 was used for radiomics feature extraction, and Python3.7.0 was used for radiomics feature screening. A radiomics model for predicting occult lymph node metastasis was established using a logistic regression algorithm. A nomogram was constructed by combining radiomics scores with selected clinical predictors. Receiver operating characteristic (ROC) curves were used to verify the performance of the radiomics model and nomogram in the training and testing sets.

Results

The radiomics nomogram comprising six selected features achieved good prediction efficiency, including radiomics characteristics and tumor location information (central or peripheral), which demonstrated good calibration and discrimination ability in the training (area under the ROC curve [AUC] = 0.884, 95% confidence interval [CI]: 0.826-0.941) and testing (AUC = 0.881, 95% CI: 0.8031-0.959) sets. Clinical decision curves demonstrated that the nomogram was clinically useful.

Conclusion

The PET/CT-based radiomics nomogram is a noninvasive tool for predicting occult lymph node metastasis in NSCLC.

An application based on bioinformatics and machine learning for risk prediction of sepsis at first clinical presentation using transcriptomic data

Background: Linking genotypic changes to phenotypic traits based on machine learning methods has various challenges. In this study, we developed a workflow based on bioinformatics and machine learning methods using transcriptomic data for sepsis obtained at the first clinical presentation for predicting the risk of sepsis. By combining bioinformatics with machine learning methods, we have attempted to overcome current challenges in predicting disease risk using transcriptomic data.

Methods: High-throughput sequencing transcriptomic data processing and gene annotation were performed using R software. Machine learning models were constructed, and model performance was evaluated by machine learning methods in Python. The models were visualized and interpreted using the Shapley Additive explanation (SHAP) method.

Results: Based on the preset parameters and using recursive feature elimination implemented via machine learning, the top 10 optimal genes were screened for the establishment of the machine learning models. In a comparison of model performance, CatBoost was selected as the optimal model. We explored the significance of each gene in the model and the interaction between each gene through SHAP analysis.

Conclusion: The combination of CatBoost and SHAP may serve as the best-performing machine learning model for predicting transcriptomic and sepsis risks. The workflow outlined may provide a new approach and direction in exploring the mechanisms associated with genes and sepsis risk.

A novel lower extremity non-contact injury risk prediction model based on multimodal fusion and interpretable machine learning

The application of machine learning algorithms in studying injury assessment methods based on data analysis has recently provided a new research insight for sports injury prevention. However, the data used in these studies are primarily multi-source and multimodal (i.e., longitudinal repeated-measures data and cross-sectional data), resulting in the models not fully utilising the information in the data to reveal specific injury risk patterns. Therefore, this study proposed an injury risk prediction model based on a multi-modal strategy and machine learning algorithms to handle multi-source data better and predict injury risk. This study retrospectively analysed the routine monitoring data of sixteen young female basketball players. These data included training load, perceived well-being status, physiological response, physical performance and lower extremity non-contact injury registration. This study partitions the original dataset based on the frequency of data collection. Extreme gradient boosting (XGBoost) was used to construct unimodal submodels to obtain decision scores for each category of indicators. Ultimately, the decision scores from each submodel were fused using the random forest (RF) to generate a lower extremity non-contact injury risk prediction model at the decision-level. The 10-fold cross-validation results showed that the fusion model was effective in classifying non-injured (mean Precision: 0.9932, mean Recall: 0.9976, mean F2-score: 0.9967), minimal lower extremity non-contact injuries risk (mean Precision: 0.9317, mean Recall: 0.9167, mean F2-score: 0.9171), and mild lower extremity non-contact injuries risk (mean Precision: 0.9000, mean Recall: 0.9000, mean F2-score: 0.9000). The model performed significantly more optimal than the submodel. Comparing the fusion model proposed with a traditional data integration scheme, the average Precision and Recall improved by 8.2 and 20.3%, respectively. The decision curves analysis showed that the proposed fusion model provided a higher net benefit to athletes with potential lower extremity non-contact injury risk. The validity, feasibility and practicality of the proposed model have been confirmed. In addition, the shapley additive explanation (SHAP) and network visualisation revealed differences in lower extremity non-contact injury risk patterns across severity levels. The model proposed in this study provided a fresh perspective on injury prevention in future research.

Identification of tumor microenvironment-related signature for predicting prognosis and immunotherapy response in patients with bladder cancer

The tumor microenvironment (TME) not only provides fertile soil for tumor growth and development but also widely involves immune evasion as well as the resistance towards therapeutic response. Accumulating interest has been attracted from the biological function of TME to its effects on patient outcomes and treatment efficacy. However, the relationship between the TME-related gene expression profiles and the prognosis of bladder cancer (BLCA) remains unclear. The TME-related genes expression data of BLCA were collected from The Cancer Genome Atlas (TCGA) database. NFM algorithm was used to identify the distinct molecular pattern based on the significantly different TME-related genes. LASSO regression and Cox regression analyses were conducted to identify TME-related gene markers related to the prognosis of BLCA and to establish a prognostic model. The predictive efficacy of the risk model was verified through integrated bioinformatics analyses. Herein, 10 TME-related genes (PFKFB4, P4HB, OR2B6, OCIAD2, OAS1, KCNJ15, AHNAK, RAC3, EMP1, and PRKY) were identified to construct the prognostic model. The established risk scores were able to predict outcomes at 1, 3, and 5 years with greater accuracy than previously known models. Moreover, the risk score was closely associated with immune cell infiltration and the immunoregulatory genes including T cell exhaustion markers. Notably, the predictive power of the model in immunotherapy sensitivity was verified when it was applied to patients with metastatic urothelial carcinoma (mUC) undergoing immunotherapy. In conclusion, TME risk score can function as an independent prognostic biomarker and a predictor for evaluating immunotherapy response in BLCA patients, which provides recommendations for improving patients’ response to immunotherapy and promoting personalized tumor immunotherapy in the future.

Development of a machine learning model to predict lateral hinge fractures by analyzing patient factors before open wedge high tibial osteotomy

PURPOSE

Several methods have been developed to prevent lateral hinge fractures (LHFs), using only classic statistical models. Machine learning is under the spotlight because of its ability to analyze various weights and model nonlinear relationships. The purpose of this study was to create a machine learning model that predicts LHF with high predictive performance.

METHODS

Data were collected from a total of 439 knees with medial osteoarthritis (OA) treated with Medial open wedge high tibial osteotomy (MOW-HTO) from March 2014 to February 2020. The patient data included age, sex, height, and weight. Preoperative, determined, and modifiable factors were categorized using X-ray and CT data to create ensemble models with better predictive performance. Among the 57 ensemble models, which is the total number of possible combinations with six models, the model with the highest area under curve (AUC) or F1-score was selected as the final ensemble model. Gain feature importance analysis and the Shapley additive explanations (SHAP) feature explanation were performed on the best models.

RESULTS

The ensemble model with the highest AUC was a combination of a light gradient boosting machine (LGBM) and multilayer perceptron (MLP) (AUC = 0.992). The ensemble model with the highest F1-score was the model that combined logistic regression (LR) and MLP (F1-score = 0.765). Distance X was the most predictive feature in the results of both model interpretation analyses.

CONCLUSION

Two types of ensemble models, LGBM with MLP and LR with MLP, were developed as machine learning models to predict LHF with high predictive performance. Using these models, surgeons can identify important features to prevent LHF and establish strategies by adjusting modifiable factors.

STUDY DESIGN

Retrospective cohort study.

Development and Validation of a Machine Learning-Based Nomogram for Prediction of Ankylosing Spondylitis

Introduction

Ankylosing spondylitis (AS) is a chronic progressive inflammatory disease of the spine and its affiliated tissues. AS mainly affects the axial bone, sacroiliac joint, hip joint, spinal facet, and adjacent ligaments. We used machine learning (ML) methods to construct diagnostic models based on blood routine examination, liver function test, and kidney function test of patients with AS. This method will help clinicians enhance diagnostic efficiency and allow patients to receive systematic treatment as soon as possible.

Methods

We consecutively screened 348 patients with AS through complete blood routine examination, liver function test, and kidney function test at the First Affiliated Hospital of Guangxi Medical University according to the modified New York criteria (diagnostic criteria for AS). By using random sampling, the patients were randomly divided into training and validation cohorts. The training cohort included 258 patients with AS and 247 patients without AS, and the validation cohort included 90 patients with AS and 113 patients without AS. We used three ML methods (LASSO, random forest, and support vector machine recursive feature elimination) to screen feature variables and then took the intersection to obtain the prediction model. In addition, we used the prediction model on the validation cohort.

Results

Seven factors—erythrocyte sedimentation rate (ESR), red blood cell count (RBC), mean platelet volume (MPV), albumin (ALB), aspartate aminotransferase (AST), and creatinine (Cr)—were selected to construct a nomogram diagnostic model through ML. In the training cohort, the C value and area under the curve (AUC) value of this nomogram was 0.878 and 0.8779462, respectively. The C value and AUC value of the nomogram in the validation cohort was 0.823 and 0.8232055, respectively. Calibration curves in the training and validation cohorts showed satisfactory agreement between nomogram predictions and actual probabilities. The decision curve analysis showed that the nonadherence nomogram was clinically useful when intervention was decided at the nonadherence possibility threshold of 1%.

Conclusion

Our ML model can satisfactorily predict patients with AS. This nomogram can help orthopedic surgeons devise more personalized and rational clinical strategies.

Supplementary Information

The online version contains supplementary material available at 10.1007/s40744-022-00481-6.

AS is a chronic progressive inflammatory disease of the spine and its affiliated tissues. AS starts gradually, and its early symptoms are mild. Some hospitals lack HLA-B27 and related imaging instruments to assist in the diagnosis of AS. There are relatively few studies on liver function and kidney function of patients with AS. We used ML methods to construct diagnostic models. Our model can satisfactorily predict patients with AS. This diagnostic model can help orthopedic surgeons devise more personalized and rational clinical strategies.

Development and assessment of a risk prediction model for moderate-to-severe obstructive sleep apnea

Background

OSA is an independent risk factor for several systemic diseases. Compared with mild OSA, patients with moderate-to-severe OSA have more severe impairment in the function of all organs of the body. Due to the current limited medical condition, not every patient can be diagnosed and treated in time. To enable timely screening of patients with moderate-to-severe OSA, we selected easily accessible variables to establish a risk prediction model.

Method

We collected 492 patients who had polysomnography (PSG), and divided them into the disease-free mild OSA group (control group), and the moderate-to-severe OSA group according to the PSG results. Variables entering the model were identified by random forest plots, univariate analysis, multicollinearity test, and binary logistic regression method. Nomogram were created based on the binary logistic results, and the area under the ROC curve was used to evaluate the discriminative properties of the nomogram model. Bootstrap method was used to internally validate the nomogram model, and calibration curves were plotted after 1,000 replicate sampling of the original data, and the accuracy of the model was evaluated using the Hosmer-Lemeshow goodness-of-fit test. Finally, we performed decision curve analysis (DCA) of nomogram model, STOP-Bang questionnaire (SBQ), and NoSAS score to assess clinical utility.

Results

There are 6 variables entering the final prediction model, namely BMI, Hypertension, Morning dry mouth, Suffocating awake at night, Witnessed apnea, and ESS total score. The AUC of this prediction model was 0.976 (95% CI: 0.962–0.990). Hosmer-Lemeshow goodness-of-fit test χ² = 3.3222 (P = 0.1899 > 0.05), and the calibration curve was in general agreement with the ideal curve. The model has good consistency in predicting the actual occurrence of moderate-to-severe risk, and has good prediction accuracy. The DCA shows that the net benefit of the nomogram model is higher than that of SBQ and NoSAS, with has good clinical utility.

Conclusion

The prediction model obtained in this study has good predictive power for moderate-to-severe OSA and is superior to other prediction models and questionnaires. It can be applied to the community population for screening and to the clinic for prioritization of treatment.

Construction and validation of a nomogram for predicting prolonged air leak after minimally invasive pulmonary resection

Background

Prolonged air leak (PAL) remains one of the most frequent postoperative complications after pulmonary resection. This study aimed to develop a predictive nomogram to estimate the risk of PAL for individual patients after minimally invasive pulmonary resection.

Methods

Patients who underwent minimally invasive pulmonary resection for either benign or malignant lung tumors between January 2020 and December 2021 were included. All eligible patients were randomly assigned to the training cohort or validation cohort at a 3:1 ratio. Univariate and multivariate logistic regression were performed to identify independent risk factors. All independent risk factors were incorporated to establish a predictive model and nomogram, and a web-based dynamic nomogram was then built based on the logistic regression model. Nomogram discrimination was assessed using the receiver operating characteristic (ROC) curve. The calibration power was evaluated using the Hosmer-Lemeshow test and calibration curves. The nomogram was also evaluated for clinical utility by the decision curve analysis (DCA).

Results

A total of 2213 patients were finally enrolled in this study, among whom, 341 cases (15.4%) were confirmed to have PAL. The following eight independent risk factors were identified through logistic regression: age, body mass index (BMI), smoking history, percentage of the predicted value for forced expiratory volume in 1 second (FEV1% predicted), surgical procedure, surgical range, operation side, operation duration. The area under the ROC curve (AUC) was 0.7315 [95% confidence interval (CI): 0.6979–0.7651] for the training cohort and 0.7325 (95% CI: 0.6743–0.7906) for the validation cohort. The P values of the Hosmer-Lemeshow test were 0.388 and 0.577 for the training and validation cohorts, respectively, with well-fitted calibration curves. The DCA demonstrated that the nomogram was clinically useful. An operation interface on a web page (https://lirongyangql.shinyapps.io/PAL_DynNom/) was built to improve the clinical utility of the nomogram.

Conclusion

The nomogram achieved good predictive performance for PAL after minimally invasive pulmonary resection. Patients at high risk of PAL could be identified using this nomogram, and thus some preventive measures could be adopted in advance.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12957-022-02716-w.

Machine learning algorithms to predict major bleeding after isolated coronary artery bypass grafting

Objectives

Postoperative major bleeding is a common problem in patients undergoing cardiac surgery and is associated with poor outcomes. We evaluated the performance of machine learning (ML) methods to predict postoperative major bleeding.

Methods

A total of 1,045 patients who underwent isolated coronary artery bypass graft surgery (CABG) were enrolled. Their datasets were assigned randomly to training (70%) or a testing set (30%). The primary outcome was major bleeding defined as the universal definition of perioperative bleeding (UDPB) classes 3–4. We constructed a reference logistic regression (LR) model using known predictors. We also developed several modern ML algorithms. In the test set, we compared the area under the receiver operating characteristic curves (AUCs) of these ML algorithms with the reference LR model results, and the TRUST and WILL-BLEED risk score. Calibration analysis was undertaken using the calibration belt method.

Results

The prevalence of postoperative major bleeding was 7.1% (74/1,045). For major bleeds, the conditional inference random forest (CIRF) model showed the highest AUC [0.831 (0.732–0.930)], and the stochastic gradient boosting (SGBT) and random forest models demonstrated the next best results [0.820 (0.742–0.899) and 0.810 (0.719–0.902)]. The AUCs of all ML models were higher than [0.629 (0.517–0.641) and 0.557 (0.449–0.665)], as achieved by TRUST and WILL-BLEED, respectively.

Conclusion

ML methods successfully predicted major bleeding after cardiac surgery, with greater performance compared with previous scoring models. Modern ML models may enhance the identification of high-risk major bleeding subpopulations.

Predicting Homelessness Among U.S. Army Soldiers No Longer on Active Duty

INTRODUCTION

The ability to predict and prevent homelessness has been an elusive goal. The purpose of this study was to develop a prediction model that identified U.S. Army soldiers at high risk of becoming homeless after transitioning to civilian life based on information available before the time of this transition.

METHODS

The prospective cohort study consisted of observations from 16,589 soldiers who were separated or deactivated from service and who had previously participated in 1 of 3 baseline surveys of the Army Study to Assess Risk and Resilience in Servicemembers in 2011-2014. A machine learning model was developed in a 70% training sample and evaluated in the remaining 30% test sample to predict self-reported homelessness in 1 of 2 Longitudinal Study surveys administered in 2016-2018 and 2018-2019. Predictors included survey, administrative, and geospatial variables available before separation/deactivation. Analysis was conducted in November 2020-May 2021.

RESULTS

The 12-month prevalence of homelessness was 2.9% (SE=0.2%) in the total Longitudinal Study sample. The area under the receiver operating characteristic curve in the test sample was 0.78 (SE=0.02) for homelessness. The 4 highest ventiles (top 20%) of predicted risk included 61% of respondents with homelessness. Self-reported lifetime histories of depression, trauma of having a loved one murdered, and post-traumatic stress disorder were the 3 strongest predictors of homelessness.

CONCLUSIONS

A prediction model for homelessness can accurately target soldiers for preventive intervention before transition to civilian life.

Personalized Surgical Transfusion Risk Prediction Using Machine Learning to Guide Preoperative Type and Screen Orders

BACKGROUND

Accurate estimation of surgical transfusion risk is essential for efficient allocation of blood bank resources and for other aspects of anesthetic planning. This study hypothesized that a machine learning model incorporating both surgery- and patient-specific variables would outperform the traditional approach that uses only procedure-specific information, allowing for more efficient allocation of preoperative type and screen orders.

METHODS

The American College of Surgeons National Surgical Quality Improvement Program Participant Use File was used to train four machine learning models to predict the likelihood of red cell transfusion using surgery-specific and patient-specific variables. A baseline model using only procedure-specific information was created for comparison. The models were trained on surgical encounters that occurred at 722 hospitals in 2016 through 2018. The models were internally validated on surgical cases that occurred at 719 hospitals in 2019. Generalizability of the best-performing model was assessed by external validation on surgical cases occurring at a single institution in 2020.

RESULTS

Transfusion prevalence was 2.4% (73,313 of 3,049,617), 2.2% (23,205 of 1,076,441), and 6.7% (1,104 of 16,053) across the training, internal validation, and external validation cohorts, respectively. The gradient boosting machine outperformed the baseline model and was the best- performing model. At a fixed 96% sensitivity, this model had a positive predictive value of 0.06 and 0.21 and recommended type and screens for 36% and 30% of the patients in internal and external validation, respectively. By comparison, the baseline model at the same sensitivity had a positive predictive value of 0.04 and 0.144 and recommended type and screens for 57% and 45% of the patients in internal and external validation, respectively. The most important predictor variables were overall procedure-specific transfusion rate and preoperative hematocrit.

CONCLUSIONS

A personalized transfusion risk prediction model was created using both surgery- and patient-specific variables to guide preoperative type and screen orders and showed better performance compared to the traditional procedure-centric approach.

EDITOR’S PERSPECTIVE

A Nomogram Model for Evaluating the Risk of Lymph Node Metastasis in cT2-cT4N0M0 Gastric Cancer Population

BACKGROUND Neoadjuvant chemotherapy is an important treatment for advanced gastric cancer, but it has been unclear whether neoadjuvant chemotherapy is closely related to lymph node metastasis. Therefore, based on the disease characteristics of the cT2-cT4N0M0 gastric cancer population, this study established a nomogram prediction model of lymph node metastasis risk in this gastric cancer population to help clinicians optimize clinical decision-making. MATERIAL AND METHODS We analyzed the data of 336 patients with advanced gastric cancer with CT imaging stage of cT2-cT4N0M0 admitted to the Third Department of the Fourth Hospital of Hebei Medical University from 2015 to 2021. Combined with the results of univariate and multivariate logistic regression analysis, 7 indicators were selected to establish a nomogram prediction model. The calibration curves, ROC curves, and decision curves were drawn against the nomogram model using R language. RESULTS The results showed that the AUC value of the model and the external validation data set were 0.925 and 0.911, respectively. The P value of the Hosmer-Lemeshow test for the internal validation dataset was 0.082, and the P value of Hosmer-Lemeshow test for the external validation dataset was 0.076.The decision curve results showed that when the threshold probability was 0.1-0.9, this model could benefit patients by predicting the risk of lymph node metastasis in patients with advanced gastric cancer, and formulating appropriate treatment schemes accordingly. CONCLUSIONS This nomogram has shown good discrimination and fit, and can also be combined with imaging examination to screen the populations suitable for neoadjuvant chemotherapy, avoid the risk of misdiagnosis of N staging to the greatest extent, and to assist clinicians to optimize clinical decision-making.

Development and Validation of a Radiomics Model Based on 18F-FDG PET of Primary Gastric Cancer for Predicting Peritoneal Metastasis

Objectives

The aim of this study was to develop a preoperative positron emission tomography (PET)-based radiomics model for predicting peritoneal metastasis (PM) of gastric cancer (GC).

Methods

In this study, a total of 355 patients (109PM+, 246PM-) who underwent preoperative fluorine-18-fludeoxyglucose (¹⁸F-FDG) PET images were retrospectively analyzed. According to a 7:3 ratio, patients were randomly divided into a training set and a validation set. Radiomics features and metabolic parameters data were extracted from PET images. The radiomics features were selected by logistic regression after using maximum relevance and minimum redundancy (mRMR) and the least shrinkage and selection operator (LASSO) method. The radiomics models were based on the rest of these features. The performance of the models was determined by their discrimination, calibration, and clinical usefulness in the training and validation sets.

Results

After dimensionality reduction, 12 radiomics feature parameters were obtained to construct radiomics signatures. According to the results of the multivariate logistic regression analysis, only carbohydrate antigen 125 (CA125), maximum standardized uptake value (SUVmax), and the radiomics signature showed statistically significant differences between patients (P<0.05). A radiomics model was developed based on the logistic analyses with an AUC of 0.86 in the training cohort and 0.87 in the validation cohort. The clinical prediction model based on CA125 and SUVmax was 0.76 in the training set and 0.69 in the validation set. The comprehensive model, which contained a rad-score and the clinical factor (CA125) as well as the metabolic parameter (SUVmax), showed promising performance with an AUC of 0.90 in the training cohort and 0.88 in the validation cohort, respectively. The calibration curve showed the actual rate of the nomogram-predicted probability of peritoneal metastasis. Decision curve analysis (DCA) also demonstrated the good clinical utility of the radiomics nomogram.

Conclusions

The comprehensive model based on the rad-score and other factors (SUVmax, CA125) can provide a novel tool for predicting peritoneal metastasis of gastric cancer patients preoperatively.

Development and Validation of a Model to Predict Posttraumatic Stress Disorder and Major Depression After a Motor Vehicle Collision

IMPORTANCE

A substantial proportion of the 40 million people in the US who present to emergency departments (EDs) each year after traumatic events develop posttraumatic stress disorder (PTSD) or major depressive episode (MDE). Accurately identifying patients at high risk in the ED would facilitate the targeting of preventive interventions.

OBJECTIVES

To develop and validate a prediction tool based on ED reports after a motor vehicle collision to predict PTSD or MDE 3 months later.

DESIGN, SETTING, AND PARTICIPANTS

The Advancing Understanding of Recovery After Trauma (AURORA) study is a longitudinal study that examined adverse posttraumatic neuropsychiatric sequalae among patients who presented to 28 US urban EDs in the immediate aftermath of a traumatic experience. Enrollment began on September 25, 2017. The 1003 patients considered in this diagnostic/prognostic report completed 3-month assessments by January 31, 2020. Each patient received a baseline ED assessment along with follow-up self-report surveys 2 weeks, 8 weeks, and 3 months later. An ensemble machine learning method was used to predict 3-month PTSD or MDE from baseline information. Data analysis was performed from November 1, 2020, to May 31, 2021.

MAIN OUTCOMES AND MEASURES

The PTSD Checklist for DSM-5 was used to assess PTSD and the Patient Reported Outcomes Measurement Information System Depression Short-Form 8b to assess MDE.

RESULTS

A total of 1003 patients (median [interquartile range] age, 34.5 [24-43] years; 715 [weighted 67.9%] female; 100 [weighted 10.7%] Hispanic, 537 [weighted 52.7%] non-Hispanic Black, 324 [weighted 32.2%] non-Hispanic White, and 42 [weighted 4.4%] of non-Hispanic other race or ethnicity were included in this study. A total of 274 patients (weighted 26.6%) met criteria for 3-month PTSD or MDE. An ensemble machine learning model restricted to 30 predictors estimated in a training sample (patients from the Northeast or Midwest) had good prediction accuracy (mean [SE] area under the curve [AUC], 0.815 [0.031]) and calibration (mean [SE] integrated calibration index, 0.040 [0.002]; mean [SE] expected calibration error, 0.039 [0.002]) in an independent test sample (patients from the South). Patients in the top 30% of predicted risk accounted for 65% of all 3-month PTSD or MDE, with a mean (SE) positive predictive value of 58.2% (6.4%) among these patients at high risk. The model had good consistency across regions of the country in terms of both AUC (mean [SE], 0.789 [0.025] using the Northeast as the test sample and 0.809 [0.023] using the Midwest as the test sample) and calibration (mean [SE] integrated calibration index, 0.048 [0.003] using the Northeast as the test sample and 0.024 [0.001] using the Midwest as the test sample; mean [SE] expected calibration error, 0.034 [0.003] using the Northeast as the test sample and 0.025 [0.001] using the Midwest as the test sample). The most important predictors in terms of Shapley Additive Explanations values were symptoms of anxiety sensitivity and depressive disposition, psychological distress in the 30 days before motor vehicle collision, and peritraumatic psychosomatic symptoms.

CONCLUSIONS AND RELEVANCE

The results of this study suggest that a short set of questions feasible to administer in an ED can predict 3-month PTSD or MDE with good AUC, calibration, and geographic consistency. Patients at high risk can be identified in the ED for targeting if cost-effective preventive interventions are developed.