Developing and Validating a Survival Prediction Model for NSCLC Patients Through Distributed Learning Across 3 Countries

This study aimed to create a tailored prediction model of hepatocellular carcinoma (HCC)-specific survival after transplantation based on pre-transplant parameters. Data collected from June 2006 to July 2018 were used as a derivation dataset and analyzed to create an HCC-specific survival prediction model by combining significant risk factors. Separate data were collected from January 2014 to June 2018 for validation. The prediction model was validated internally and externally. The data were divided into three groups based on risk scores derived from the hazard ratio. A combination of patient demographic, laboratory, radiological data, and tumor-specific characteristics that showed a good prediction of HCC-specific death at a specific time (t) were chosen. Internal and external validations with Uno’s C-index were 0.79 and 0.75 (95% confidence interval (CI) 0.65–0.86), respectively. The predicted survival after liver transplantation for HCC (SALT) at a time “t” was calculated using the formula: [1 − (HCC-specific death(t’))] × 100. The 5-year HCC-specific death and recurrence rates in the low-risk group were 2% and 5%; the intermediate-risk group was 12% and 14%, and in the high-risk group were 71% and 82%. Our HCC-specific survival predictor named “SALT calculator” could provide accurate information about expected survival tailored for patients undergoing transplantation for HCC.

Download Full-text

Machine learning‐based individualized survival prediction model for total knee replacement in osteoarthritis: Data from the Osteoarthritis Initiative

Arthritis Care & Research ◽

10.1002/acr.24601 ◽

2021 ◽

Author(s):

Afshin Jamshidi ◽

Jean‐Pierre Pelletier ◽

Aurelie Labbe ◽

François Abram ◽

Johanne Martel‐Pelletier ◽

...

Keyword(s):

Machine Learning ◽

Total Knee Replacement ◽

Prediction Model ◽

Knee Replacement ◽

Survival Prediction ◽

Osteoarthritis Initiative ◽

Total Knee

Download Full-text

A machine learning‐based survival prediction model of high grade glioma by integration of clinical and dose‐volume histogram parameters

Cancer Medicine ◽

10.1002/cam4.3838 ◽

2021 ◽

Vol 10 (8) ◽

pp. 2774-2786

Author(s):

Haiyan Chen ◽

Chao Li ◽

Lin Zheng ◽

Wei Lu ◽

Yanlin Li ◽

...

Keyword(s):

Machine Learning ◽

Prediction Model ◽

High Grade Glioma ◽

Survival Prediction ◽

High Grade ◽

Dose Volume Histogram ◽

Dose Volume

Download Full-text

Abstract 127: Machine Learning Techniques to Predict Cardiac Re-Arrest in Out-Of-Hospital Setting

Circulation ◽

10.1161/circ.140.suppl_2.127 ◽

2019 ◽

Vol 140 (Suppl_2) ◽

Author(s):

Andoni Elola ◽

Elisabete Aramendi ◽

Unai Irusta ◽

Naroa Amezaga ◽

Jon Urteaga ◽

...

Keyword(s):

Cardiac Arrest ◽

Prediction Model ◽

Prehospital Care ◽

Hospital Setting ◽

Area Under The Curve ◽

Interquartile Range ◽

Machine Learning Techniques ◽

Survival Prediction ◽

Poincaré Plot ◽

Rr Intervals

Background: Re-arrest occurs when a cardiac arrest patient being treated by the emergency medical services experiences another cardiac arrest after return of spontaneous circulation (ROSC).The incidence of re-arrest is high, close to 40% in out-of-hospital cardiac arrest (OHCA), and it is associated with lower survival. Prediction of re-arrest could improve prehospital care. The aim of this study was to develop a re-arrest prediction model based on heart rate variability (HRV) features. Materials and methods: OHCA cases treated by Dallas-FortWorth Center of Resuscitation Research were analyzed. Patients with at least two minutes of ROSC were included. Re-arrest was ascertained by the presence of life-threatening ECG and/or presence of chest compressions within 12 minutes after ROSC. Eighteen HRV characteristics for 1 min and 2 min intervals after ROSC were computed. Features were fed into a Random Forest (RF) classifier with 100 trees to predict re-arrest cases. Ten-fold cross-validation with 30 repetitions was applied to train the model and assess the performance in terms of area under the curve (AUC). Results: Inclusion criteria were met by 98 patients, 41 of which suffered re-arrest. The median time (interquartile range) to re-arrest from ROSC onset was 5 (3-7) min. The re-arrest prediction model showed a median AUC of 0.71 and 0.75 for 1 and 2 min post ROSC intervals, respectively. The most important HRV features in the RF predictor were the SD1/SD2 ratio (where SD1 and SD2 are the dispersions of points both perpendicular and parallel to the line-of-identity in the Poincaré plot), SD2, the interquartile range of the RR intervals, peak frequency in the high frequency band (0.15-0.4 Hz) and coefficient of variation of RR intervals (the ratio between the mean and standard deviation of RR intervals). Conclusions: HRV metrics predict re-arrest in OHCA. Further studies with larger datasets are needed to better understand re-arrest dynamics and confirm conclusions.

Download Full-text

Abstract 205: Resuscitation Using eCPR to Predict Survival After In-Hospital Cardiac Arrest (Rescue-IHCA Score) Survival Prediction Model--An Analysis of the American Heart Association Get With the Guidelines--Resuscitation Registry

Circulation ◽

10.1161/circ.140.suppl_2.205 ◽

2019 ◽

Vol 140 (Suppl_2) ◽

Author(s):

Joseph E Tonna ◽

Lance B Becker ◽

Saket Girotra ◽

Craig Selzman ◽

Ravi R Thiagarajan ◽

...

Keyword(s):

Sensitivity Analysis ◽

Prediction Model ◽

Heart Association ◽

P Value ◽

Survival Prediction ◽

Model Parameters ◽

Extracorporeal Cardiopulmonary Resuscitation ◽

Initial Rhythm ◽

Get With The Guidelines ◽

Resuscitation Registry

Background: To guide extracorporeal cardiopulmonary resuscitation (eCPR) use, a generalizable survival prediction model is needed. Methods: We identified patients≥18 years with IHCA who received eCPR (January 2000-December 2017) in the AHA Get With The Guidelines—Resuscitation registry to build a survival model. We categorized admission CPC into ‘good’ (CPC 1) vs other. We singly imputed variables with ≥15% missing (admission CPC [20%], duration of event [15%]). Variables associated with death (p-value ≤0.1) were retained and initial rhythm was forced into the model. We used firth penalized logistic regression to estimate model parameters. To test the imputation effect, we performed a sensitivity analysis excluding CPC. We performed a Kaplan Meier survival analysis stratified by resuscitation duration (0 to ≤15, 15 to ≤30, 30 to ≤60, ≥60 min). Results: Of 1,082 patients who underwent eCPR, 963 were included in the model ( Table 1 ). Area Under the Receiving Operating Characteristic (AUROC) = 0.81 (95% CI [0.78 to 0.83]). Associations with death included: nighttime eCPR use; non-white race; patients with prior renal insufficiency, preceding hypoperfusion, and congestive heart failure. Initial rhythm was not associated with death. Every 10 minutes of resuscitation was associated with 12% increased odds of death. Shorter resuscitation duration was strongly associated with hospital survival ( Figure 1 ). The AUROC was unchanged (0.81 [95% CI 0.78 - 0.84]) after sensitivity analysis excluding CPC. Conclusions: In this preliminary registry analysis, survival after eCPR for IHCA was estimated by patient and arrest characteristics. Our findings require validation.

Download Full-text

Enhancing the Lasso Approach for Developing a Survival Prediction Model Based on Gene Expression Data

Computational and Mathematical Methods in Medicine ◽

10.1155/2015/259474 ◽

2015 ◽

Vol 2015 ◽

pp. 1-7 ◽

Cited By ~ 2

Author(s):

Shuhei Kaneko ◽

Akihiro Hirakawa ◽

Chikuma Hamada

Keyword(s):

Gene Expression ◽

Prediction Model ◽

Gene Expression Data ◽

Prediction Models ◽

Mixture Distribution ◽

Tuning Parameter ◽

Survival Prediction ◽

Expression Data ◽

True Positive ◽

False Negatives

In the past decade, researchers in oncology have sought to develop survival prediction models using gene expression data. The least absolute shrinkage and selection operator (lasso) has been widely used to select genes that truly correlated with a patient’s survival. The lasso selects genes for prediction by shrinking a large number of coefficients of the candidate genes towards zero based on a tuning parameter that is often determined by a cross-validation (CV). However, this method can pass over (or fail to identify) true positive genes (i.e., it identifies false negatives) in certain instances, because the lasso tends to favor the development of a simple prediction model. Here, we attempt to monitor the identification of false negatives by developing a method for estimating the number of true positive (TP) genes for a series of values of a tuning parameter that assumes a mixture distribution for the lasso estimates. Using our developed method, we performed a simulation study to examine its precision in estimating the number of TP genes. Additionally, we applied our method to a real gene expression dataset and found that it was able to identify genes correlated with survival that a CV method was unable to detect.

Download Full-text

Model-based optimization of subgroup weights for survival analysis

Bioinformatics ◽

10.1093/bioinformatics/btz361 ◽

2019 ◽

Vol 35 (14) ◽

pp. i484-i491

Author(s):

Jakob Richter ◽

Katrin Madjar ◽

Jörg Rahnenführer

Keyword(s):

Survival Analysis ◽

Prediction Model ◽

Model Building ◽

Prediction Models ◽

Survival Prediction ◽

Good Prediction ◽

Similar Data ◽

Weighted Likelihood ◽

Model Based ◽

Specific Subgroup

AbstractMotivationTo obtain a reliable prediction model for a specific cancer subgroup or cohort is often difficult due to limited sample size and, in survival analysis, due to potentially high censoring rates. Sometimes similar data from other patient subgroups are available, e.g. from other clinical centers. Simple pooling of all subgroups can decrease the variance of the predicted parameters of the prediction models, but also increase the bias due to heterogeneity between the cohorts. A promising compromise is to identify those subgroups with a similar relationship between covariates and target variable and then include only these for model building.ResultsWe propose a subgroup-based weighted likelihood approach for survival prediction with high-dimensional genetic covariates. When predicting survival for a specific subgroup, for every other subgroup an individual weight determines the strength with which its observations enter into model building. MBO (model-based optimization) can be used to quickly find a good prediction model in the presence of a large number of hyperparameters. We use MBO to identify the best model for survival prediction of a specific subgroup by optimizing the weights for additional subgroups for a Cox model. The approach is evaluated on a set of lung cancer cohorts with gene expression measurements. The resulting models have competitive prediction quality, and they reflect the similarity of the corresponding cancer subgroups, with both weights close to 0 and close to 1 and medium weights.Availability and implementationmlrMBO is implemented as an R-package and is freely available at http://github.com/mlr-org/mlrMBO.

Download Full-text

External Validation of the Dutch SOURCE Survival Prediction Model in Belgian Metastatic Oesophageal and Gastric Cancer Patients

Cancers ◽

10.3390/cancers12040834 ◽

2020 ◽

Vol 12 (4) ◽

pp. 834

Author(s):

J.J. van Kleef ◽

H.G. van den Boorn ◽

R.H.A. Verhoeven ◽

K. Vanschoenbeek ◽

A. Abu-Hanna ◽

...

Keyword(s):

Gastric Cancer ◽

Prediction Model ◽

Oesophageal Cancer ◽

Prediction Error ◽

Prediction Models ◽

Mean Difference ◽

Survival Prediction ◽

Cancer Model ◽

The Mean ◽

Mean Prediction Error

The SOURCE prediction model predicts individualised survival conditional on various treatments for patients with metastatic oesophageal or gastric cancer. The aim of this study was to validate SOURCE in an external cohort from the Belgian Cancer Registry. Data of Belgian patients diagnosed with metastatic disease between 2004 and 2014 were extracted (n = 4097). Model calibration and discrimination (c-indices) were determined. A total of 2514 patients with oesophageal cancer and 1583 patients with gastric cancer with a median survival of 7.7 and 5.4 months, respectively, were included. The oesophageal cancer model showed poor calibration (intercept: 0.30, slope: 0.42) with an absolute mean prediction error of 14.6%. The mean difference between predicted and observed survival was −2.6%. The concordance index (c-index) of the oesophageal model was 0.64. The gastric cancer model showed good calibration (intercept: 0.02, slope: 0.91) with an absolute mean prediction error of 2.5%. The mean difference between predicted and observed survival was 2.0%. The c-index of the gastric cancer model was 0.66. The SOURCE gastric cancer model was well calibrated and had a similar performance in the Belgian cohort compared with the Dutch internal validation. However, the oesophageal cancer model had not. Our findings underscore the importance of evaluating the performance of prediction models in other populations.

Download Full-text

Prediction of EGFR-TKI efficacy in non-small cell lung cancer patients by metabolomics and genomics.

Journal of Clinical Oncology ◽

10.1200/jco.2019.37.15_suppl.e20627 ◽

2019 ◽

Vol 37 (15_suppl) ◽

pp. e20627-e20627

Author(s):

Lei Zhang ◽

Rongrong Luo ◽

Lin Wang ◽

Jiarui Yao ◽

Di Wu ◽

...

Keyword(s):

Lung Cancer ◽

Prediction Model ◽

Small Cell Lung Cancer ◽

Somatic Mutations ◽

Small Cell ◽

Joint Analysis ◽

Small Cell Lung ◽

Ffpe Samples ◽

Nsclc Patients ◽

Egfr Tki

e20627 Background: Metabolites and somatic mutations involved in EGFR-TKI efficacy remains unclear in non-small cell lung cancer (NSCLC) patients with EGFR sensitizing mutation (EGFRsm+). Here we performed a joint analysis of metabolomics and genomics data to identify metabolites and somatic mutations as biomarkers for EGFR-TKI efficacy. Methods: Metabolomic profiling of plasma samples (n = 43) from NSCLC patients with EGFRsm+, consisting of cohort A (n = 30) and B (n = 13), was conducted using UPLC or rapid separation LC-MS/MS. The 13 matched FFPE samples in cohort B were also used in the targeted sequencing below. FFPE samples (n = 18) from NSCLC patients with EGFRsm+ were subjected to targeted sequencing. According to progression free survival (PFS), all patients were assigned a status of poor (PFS≤42 weeks) and good responders (PFS > 42 weeks). A joint analysis of metabolomics and genomics data was adopted to identify biomarkers for EGFR-TKI efficacy. Results: The partial least squares discrimination analysis mothod was performed to establish a prediction model responsible for separation of good and poor responders in cohort A, comprising 27 metabolites with variable importance in projection score (VIP) > 1.5. Based on the prediction model, the ROC analysis demonstrated the sensitivity of 0.8, the specificity of 0.75, and the area under the ROC curve (AUC) of 0.7 in cohort B. The Welch’s t test method identified 15 significant metabolites ( P < 0.05) in cohort A. With the criteria of VIP > 1.5 and P < 0.05, four metabolites, 3-Methyl-L-Histidine, LysoPE(18:2(9Z,12Z)/0:0), Histamine, and SM(d18:1/16:0), were detected as potential biomarkers. To further validate them, associations of these metabolites and somatic mutations were explored in 13 patients with both metabolomics and genomics data available using the Welch’s t test. The results revealed patients with either CTCF R415X or PTK2B G491X had significantly lower Histamine level compared with those without either mutation (both P < 0.05), and significantly increased level of SM(d18:1/16:0) was observed in patients with either GATA2 P250A or MAGI1 S763X (both P < 0.05). Intriguingly, worse PFS was showed in patients with any mutation of GATA2 P250A ( P = 0.02), CTCF R415X ( P = 0.002), PTK2B G491X ( P = 0.002), and MAGI1 S763X ( P = 0.0007). Conclusions: Our joint analysis identified two plasma metabolites and four somatic mutations as biomarkers for EGFR-TKI efficacy. The present findings may provide insights into molecular mechanisms of EGFR-TKI efficacy. Further validation in prospective studies was warranted.

Download Full-text