External validation of the SORG 90-day and 1-year machine learning algorithms for survival in spinal metastatic disease

BACKGROUND CONTEXTPreoperative survival estimation in spinal metastatic disease helps determine the appropriateness of invasive management. The SORG ML 90-day and 1-year machine learning algorithms for survival in spinal metastatic disease were previously developed in a single institutional sample but remain to be externally validated.PURPOSEThe purpose of this study was to externally validate these algorithms in an independent population from another institution.STUDY DESIGN/SETTINGRetrospective study at a large, tertiary care center.PATIENT SAMPLEPatients 18 years or older who underwent surgery between 2003 and 2016.OUTCOME MEASURESNinety-day and 1-year mortality.METHODSBaseline characteristics of the validation cohort were compared to the developmental cohort for the SORG ML algorithms. Discrimination (c-statistic and receiver operating curve), calibration (calibration slope, intercept, calibration plot, and observed proportions by predicted risk groups), overall performance (Brier score), and decision curve analysis were used to assess the performance of the SORG ML algorithms in the validation cohort.RESULTSOverall, 176 patients underwent surgery for spinal metastatic disease, of which 44 (22.7%) experienced 90-day mortality and 99 (56.2%) experienced 1-year mortality. The validation cohort differed significantly from the developmental cohort on primary tumor histology, metastatic tumor burden, previous systemic therapy, overall comorbidity burden, and preoperative laboratory characteristics. Despite these differences, the SORG ML algorithms generalized well to the validation cohort on discrimination (c-statistic 0.75–0.81 for 90-day mortality and 0.77–0.78 for 1-year mortality), calibration, Brier score, and decision curve analysis.CONCLUSION and RELEVANCEInitial results from external validation of the SORG ML 90-day and 1-year algorithms for survival prediction in spinal metastatic disease suggest potential utility of these digital decision aids in clinical practice. Further studies are needed to validate or refute these algorithms in large patient samples from prospective, international, multi-institutional trials.     

Comparison of three different tools for prediction of seminal vesicle invasion at radical prostatectomy

Background

Statistical prediction tools are increasingly common, but there is considerable disagreement about how they should be evaluated. Three tools—Partin tables, the European Society for Urological Oncology (ESUO) criteria, and the Gallina nomogram—have been proposed for the prediction of seminal vesicle invasion (SVI) in patients with clinically localized prostate cancer who are candidates for a radical prostatectomy.

Objectives

Using different statistical methods, we aimed to determine which of these tools should be used to predict SVI.

Design, settings, and participants

The independent validation cohort consisted of 2584 patients treated surgically for clinically localized prostate cancer at four North American tertiary care centers between 2002 and 2007.

Interventions

Robot-assisted laparoscopic radical prostatectomy.

Outcome measurements and statistical analysis

Primary outcome was the presence of SVI. Traditional (area under the receiver operating characteristic [ROC] curve, calibration plots, the Brier score, sensitivity and specificity, positive and negative predictive value) and novel (decision curve analysis and predictiveness curves) statistical methods quantified the predictive abilities of the three models.

Results and limitations

Traditional statistical methods (ie, ROC plots and Brier scores) could not clearly determine which one of the three SVI prediction tools should be preferred. For example, ROC plots and Brier scores seemed biased against the binary decision tool (ESUO criteria) and gave discordant results for the continuous predictions of the Partin tables and the Gallina nomogram. The results of the calibration plots were discordant with those of the ROC plots. Conversely, the decision curve indicated that the Partin tables represent the best strategy for stratifying the risk of SVI, resulting in the highest net benefit within the whole range of threshold probabilities.

Conclusions

When predicting SVI, surgeons should prefer the Partin tables over the ESUO criteria and the Gallina nomogram because this tool provided the highest net benefit. In contrast to traditional statistical methods, decision curve analysis gave an unambiguous result applicable to both continuous and binary models, providing an insight into clinical utility.     

Performance of models predicting residual lymph node disease in melanoma patients following sentinel lymph node biopsy

BackgroundAmong melanoma patients with a tumor-positive sentinel node biopsy (SNB), approximately 20% harbor disease in non-sentinel nodes (nSN), as determined by a completion lymph node dissection (CLND). CLND lacks a survival benefit and has high morbidity. This study assesses predictive factors for nSN metastasis and validates five models predicting nSN metastasis.MethodsPatients with invasive melanoma were identified from the BC Cancer Agency (2005–2015). Clinicopathological data were collected from 296 patients who underwent a CLND after a positive SNB. Multivariate analysis was completed to assess predictive variables in the study population. Five models were externally validated using overall model performance (Brier score [calibration and discrimination]) and discrimination (area under the ROC curve [AUC]).ResultsSeventy-three patients had nSN metastasis at the time of CLND. The variable most predictive of nSN involvement was lymphovascular invasion (odds ratio [OR] 3.99; 95% confidence interval [CI] 1.67–9.54; p = 0.002). The highest discrimination was Lee et al. (2004) (AUC 0.68 [95% CI 0.61–0.75]), Rossi et al. (2018) (AUC 0.68 [95% CI 0.57–0.77]), and Bertolli et al. (2019) (AUC 0.68 [95% CI 0.60–0.75]). Rossi et al. (2018) had the lowest overall model performance (Brier score 0.44). Rossi et al. (2018) and Bertolli et al. (2019) had the ability to stratify patients to a risk of nSN involvement up to 99% and 95%, respectively.ConclusionBertolli et al. (2019) had amongst the highest overall model performance, was the most clinically meaningful and is recommended as the preferred model for predicting nSN metastasis.     

Evaluation of models that predict short-term outcome after traumatic brain injury

PRIMARY OBJECTIVE: This study aimed to identify models that predicted the short-term outcome after traumatic brain injury (TBI) from the literature and to evaluate their clinical significance. METHODS: Literatures from PubMED were reviewed. Regression coefficients and intercepts were extracted. A group of 229 cases was used for validation and the unfavourable rate was calculated to assess the validity of these models by the area under receiver operating. Characteristic curve (AUC), C-statistic and Brier score. MAIN RESULTS: In total, 13 studies of 18 different models were included. Data from the validation group were in accordance with the indicators of the studies reviewed. All models got an AUC value ranging from 0.644-0.890 except two (AUC value <0.6) and their Brier scores were near zero. However, the calibration of most studies was insufficient (p < 0.05). CONCLUSIONS: Most of the models included in this study have a good discriminatory power while lacking sufficient calibration. However, they all predict with relative accuracy at the level of individuals. Therefore, current models can be used to predict the survival rate of individual patients and may be useful to inform patients and relatives about the likelihood of a beneficial outcome.     

External validation of a predictive algorithm for in-hospital and 90-day mortality after spinal epidural abscess

BACKGROUND CONTEXTMortality in patients with spinal epidural abscess (SEA) remains high. Accurate prediction of patient-specific prognosis in SEA can improve patient counseling as well as guide management decisions. There are no externally validated studies predicting short-term mortality in patients with SEA.PURPOSEThe purpose of this study was to externally validate the Skeletal Oncology Research Group (SORG) stochastic gradient boosting algorithm for prediction of in-hospital and 90-day postdischarge mortality in SEA.STUDY DESIGN/SETTINGRetrospective, case-control study at a tertiary care academic medical center from 2003 to 2021.PATIENT SAMPLEAdult patients admitted for radiologically confirmed diagnosis of SEA who did not initiate treatment at an outside institution.OUTCOME MEASURESIn-hospital and 90-day postdischarge mortality.METHODSWe tested the SORG stochastic gradient boosting algorithm on an independent validation cohort. We assessed its performance with discrimination, calibration, decision curve analysis, and overall performance.RESULTSA total of 212 patients met inclusion criteria, with a short-term mortality rate of 10.4%. The area under the receiver operating characteristic curve (AUROC) of the SORG algorithm when tested on the full validation cohort was 0.82, the calibration intercept was -0.08, the calibration slope was 0.96, and the Brier score was 0.09.CONCLUSIONSWith a contemporaneous and geographically distinct independent cohort, we report successful external validation of a machine learning algorithm for prediction of in-hospital and 90-day postdischarge mortality in SEA.     

STAR人工踝关节数字器械库的建立和手术模拟

[目的]探讨数字模拟(Scandinavian total ankle replacement,STAR)人工踝关节置换术的可行性和方法.[方法]应用Mimics 10.01、Geomagic studio 10.0、PRO/E 2.0软件模拟建立三维踝关节、STAR常用手术器械库、STAR人工关节假体模型库,利用PRO/E 2.0软件的强大建模及装配功能,对STAR人工踝关节手术步骤进行逐步模拟.[结果]成功模拟STAR人工踝关节置换术手术步骤.[结论]数字模拟STAR人工踝关节置换术可行,有助于熟悉及掌握该手术,对STAR人工踝关节置换术的术前准备、术中都有指导意义.     

The kidney failure risk equation predicts kidney failure: Validation in an Australian cohort

Aims

Predicting progression to kidney failure for patients with chronic kidney disease is essential for patient and clinicians' management decisions, patient prognosis, and service planning. The Tangri et al Kidney Failure Risk Equation (KFRE) was developed to predict the outcome of kidney failure. The KFRE has not been independently validated in an Australian Cohort.

Methods

Using data linkage of the Tasmanian Chronic Kidney Disease study (CKD.TASlink) and the Australia and New Zealand Dialysis and Transplant Registry (ANZDATA), we externally validated the KFRE. We validated the 4, 6, and 8-variable KFRE at both 2 and 5 years. We assessed model fit (goodness of fit), discrimination (Harell's C statistic), and calibration (observed vs predicted survival).

Results

There were 18 170 in the cohort with 12 861 participants with 2 years and 8182 with 5 years outcomes. Of these 2607 people died and 285 progressed to kidney replacement therapy. The KFRE has excellent discrimination with C statistics of 0.96–0.98 at 2 years and 0.95–0.96 at 5 years. The calibration was adequate with well-performing Brier scores (0.004–0.01 at 2 years, 0.01–0.03 at 5 years) however the calibration curves, whilst adequate, indicate that predicted outcomes are systematically worse than observed.

Conclusion

This external validation study demonstrates the KFRE performs well in an Australian population and can be used by clinicians and service planners for individualised risk prediction.     

Impact of systemic Immune–inflammation Index on oncologic outcomes in patients treated with radical prostatectomy for clinically nonmetastatic prostate cancer

PurposeTo evaluate the predictive and prognostic value of the Systemic Immune–inflammation Index (SII) in a large cohort of patients treated with radical prostatectomy (RP) for clinically non–metastatic prostate cancer (PCa).MethodsWe retrospectively analyzed our multicenter database comprising 6,039 consecutive patients. The optimal preoperative SII cut–off value was assessed with the Youden index calculated on a time–dependent receiver operating characteristic (ROC) curve. Logistic regression and Cox regression analyses were used to investigate the association of SII with pathologic features and biochemical recurrence (BCR), respectively. The discriminatory ability of the models was evaluated by calculating the concordance-indices (C-Index). The clinical benefit of the implementation of SII in clinical decision making was assessed using decision curve analysis (DCA).ResultsPatients with high preoperative SII (≥ 620) were more likely to have adverse clinicopathologic features. On multivariable logistic regression analysis, high preoperative SII was independently associated with extracapsular extension (odds ratio [OR] 1.16, P = 0.041), non–organ confined disease (OR 1.18, P = 0.022), and upgrading at RP (OR 1.23, P < 0.001). We built two Cox regression models including preoperative and postoperative variables. In the preoperative multivariable model, high preoperative SII was associated with BCR (hazard ratio [HR] 1.34, 95% CI 1.15-1.55, P < 0.001). In the postoperative multivariable model, SII was not associated with BCR (P = 0.078). The addition of SII to established models did not improve their discriminatory ability nor did it increase the clinical net benefit on DCA.ConclusionIn men treated with RP for clinically nonmetastatic PCa, high preoperative SII was statistically associated with an increased risk of adverse pathologic features at RP as well as BCR. However, it did not improve the predictive accuracy and clinical value beyond that obtained by current predictive and prognostic models. SII together with a panel of complementary biomarkers is praised to help guide decision–making in clinically nonmetastatic PCa.     

External validation of the SORG machine learning algorithms for predicting 90-day and 1-year survival of patients with lung cancer-derived spine metastases: a recent bi-center cohort from China

BACKGROUND CONTEXTThe survival prediction of lung cancer-derived spinal metastases is often underestimated by several scores. The SORG machine learning (ML) algorithm is considered a promising tool to predict the risk of 90-day and 1-year mortality in patients with spinal metastases, but not been externally validated for lung cancer.PURPOSEThis study aimed to externally validate the SORG ML algorithms on lung cancer-derived spinal metastases patients from two large-volume, tertiary medical centers between 2018 and 2021.STUDY DESIGN/SETTINGRetrospective, cohort study.PATIENT SAMPLEPatients aged 18 years or older at two tertiary medical centers in China are treated surgically for spinal metastasis.OUTCOME MEASURESMortality within 90 days of surgery, mortality within 1 year of surgery.METHODSThe baseline characteristics were compared between the development cohort and our validation cohort. Discrimination (receiver operating curve), calibration (calibration plot, intercept, and slope), the overall performance (Brier score), and decision curve analysis was used to assess the overall performance of the SORG ML algorithms.RESULTSThis study included 150 patients with lung cancer-derived spinal metastases from two medical centers in China. Ninety-day and 1-year mortality rates were 12.9% (19/147) and 51.3% (60/117), respectively. Lung Cancer with targeted therapies had the lowest Hazard Ratio (HR=0.490), showing an optimal protecting factor. The AUC of the SORG ML algorithm for 90-day mortality prediction in lung cancer-derived spinal metastases is 0.714. While the AUC for 1-year mortality prediction is 0.832 (95CI%, 0.758–0.906). The algorithm for 1-year mortality was well-calibrated with an intercept of 0.13 and a calibration slope of 1.00. However, the 90-day mortality prediction was underestimated with an intercept of 0.60 and a slope of 0.37. The SORG ML algorithms for 1-year mortality showed a greater net benefit than the “treats all or no patients” strategies.CONCLUSIONSIn the latest cohort of lung cancer-derived spinal metastases in China, the SORG algorithms for predicting 1-year mortality performed well on external validation. However, 90-day mortality was underestimated. The algorithm should be further validated by single primary tumor-derived metastasis treated with the latest comprehensive treatment in diverse populations.     

Development of Machine Learning Algorithms to Predict Clinically Meaningful Improvement for the Patient-Reported Health State After Total Hip Arthroplasty

BackgroundFailure to achieve clinically significant outcome (CSO) improvement after total hip arthroplasty (THA) imposes a potential cost-to-risk imbalance in the context of bundle payment models. Patient perception of their health state is one component of such risk. The purpose of the current study is to develop machine learning algorithms to predict CSO for the patient-reported health state (PRHS) and build a clinical decision-making tool based on risk factors.MethodsA retrospective review of primary THA patients between 2014 and 2017 was performed. Variables considered for prediction included demographics, medical history, preoperative PRHS, and modified Harris Hip Score. The minimal clinically important difference (MCID) for the PRHS was calculated using a distribution-based method. Five supervised machine learning algorithms were developed and assessed by discrimination, calibration, Brier score, and decision curve analysis.ResultsOf 616 patients, a total of 407 (69.2%) achieved the MCID for the PRHS. The random forest algorithm achieved the best performance in the independent testing set not used for algorithm development (c-statistic 0.97, calibration intercept −0.05, calibration slope 1.45, Brier score 0.054). The most important factors for achieving the MCID were preoperative PRHS, preoperative opioid use, age, and body mass index. Individual patient-level explanations were provided for the algorithm predictions and the algorithms were incorporated into an open access digital application available here: https://sorg-apps.shinyapps.io/THA_PRHS_mcid/.ConclusionThe current study created a clinical decision-making tool based on partially modifiable risk factors for predicting CSO after THA. The tool demonstrates excellent discriminative capacity for identifying those at greatest risk for failing to achieve CSO in their current health state and may allow for preoperative health optimization.     

Predicting in-hospital mortality after traumatic brain injury: External validation of CRASH-basic and IMPACT-core in the national trauma data bank

BackgroundTraumatic brain injury (TBI) prognostic prediction models offer value to individualized treatment planning, systematic outcome assessments and clinical research design but require continuous external validation to ensure generalizability to different settings. The Corticosteroid Randomization After Significant Head Injury (CRASH) and International Mission on Prognosis and Analysis on Clinical Trials in TBI (IMPACT) models are widely available but lack robust assessments of performance in a current national sample of patients. The purpose of this study is to assess the performance of the CRASH-Basic and IMPACT-Core models in predicting in-hospital mortality using a nationwide retrospective cohort from the National Trauma Data Bank (NTDB).MethodsThe 2016 NTDB was used to analyze an adult cohort with moderate-severe TBI (Glasgow Coma Scale [GCS] ≤ 12, head Abbreviated Injury Scale of 2–6). Observed in-hospital mortality or discharge to hospice was compared to the CRASH-Basic and IMPACT-Core models’ predicted probability of 14-day or 6-month mortality, respectively. Performance measures included discrimination (area under the receiver operating characteristic curve [AUC]) and calibration (calibration plots and Brier scores). Further sensitivity analysis included patients with GCS ≤ 14 and considered patients discharged to hospice to be alive at 14-days.ResultsA total of 26,228 patients were included in this study. Both models demonstrated good ability in differentiating between patients who died and those who survived, with IMPACT demonstrating a marginally greater AUC (0.863; 95% CI: 0.858 – 0.867) than CRASH (0.858; 0.854 – 0.863); p < 0.001. On calibration, IMPACT overpredicted at lower scores and underpredicted at higher scores but had good calibration-in-the-large (indicating no systemic over/underprediction), while CRASH consistently underpredicted mortality. Brier scores were similar (0.152 for IMPACT, 0.162 for CRASH; p < 0.001). Both models showed slight improvement in performance when including patients with GCS ≤ 14.ConclusionBoth CRASH-Basic and IMPACT-Core accurately predict in-hospital mortality following moderate-severe TBI, and IMPACT-Core performs well beyond its original GCS cut-off of 12, indicating potential utility for mild TBI (GCS 13–15). By demonstrating validity in the NTDB, these models appear generalizable to new data and offer value to current practice in diverse settings as well as to large-scale research design.Introduction     

Artificial Intelligence to Identify Arthroplasty Implants From Radiographs of the Knee

BackgroundRevisions and reoperations for patients who have undergone total knee arthroplasty (TKA), unicompartmental knee arthroplasty (UKA), and distal femoral replacement (DFR) necessitates accurate identification of implant manufacturer and model. Failure risks delays in care, increased morbidity, and further financial burden. Deep learning permits automated image processing to mitigate the challenges behind expeditious, cost-effective preoperative planning. Our aim was to investigate whether a deep-learning algorithm could accurately identify the manufacturer and model of arthroplasty implants about the knee from plain radiographs.MethodsWe trained, validated, and externally tested a deep-learning algorithm to classify knee arthroplasty implants from one of 9 different implant models from retrospectively collected anterior-posterior (AP) plain radiographs from four sites in one quaternary referral health system. The performance was evaluated by calculating the area under the receiver-operating characteristic curve (AUC), sensitivity, specificity, and accuracy when compared with a reference standard of implant model from operative reports.ResultsThe training and validation data sets were comprised of 682 radiographs across 424 patients and included a wide range of TKAs from the four leading implant manufacturers. After 1000 training epochs by the deep-learning algorithm, the model discriminated nine implant models with an AUC of 0.99, accuracy 99%, sensitivity of 95%, and specificity of 99% in the external-testing data set of 74 radiographs.ConclusionsA deep learning algorithm using plain radiographs differentiated between 9 unique knee arthroplasty implants from four manufacturers with near-perfect accuracy. The iterative capability of the algorithm allows for scalable expansion of implant discriminations and represents an opportunity in delivering cost-effective care for revision arthroplasty.     

External validation of a predictive model of survival after cytoreductive nephrectomy for metastatic renal cell carcinoma

Introduction

Recent trials have emphasized the importance of a precise patient selection for cytoreductive nephrectomy (CN). In 2013, a nomogram was developed for pre- and postoperative prediction of the probability of death (PoD) after CN in patients with metastatic renal cell carcinoma. To date, the single-institutional nomogram which included mostly patients from the cytokine era has not been externally validated. Our objective is to validate the predictive model in contemporary patients in the targeted therapy era.

Methods

Multi-institutional European and North American data from patients who underwent CN between 2006 and 2013 were used for external validation. Variables evaluated included preoperative serum albumin and lactate dehydrogenase levels, intraoperative blood transfusions (yes/no) and postoperative pathologic stage (primary tumour and nodes). In addition, patient characteristics and MSKCC risk factors were collected. Using the original calibration indices and quantiles of the distribution of predictions, Kaplan–Meier estimates and calibration plots of observed versus predicted PoD were calculated. For the preoperative model a decision curve analysis (DCA) was performed.

Results

Of 1108 patients [median OS of 27 months (95% CI 24.6–29.4)], 536 and 469 patients had full data for the validation of the pre- and postoperative models, respectively. The AUC for the pre- and postoperative model was 0.68 (95% CI 0.62–0.74) and 0.73 (95% CI 0.68–0.78), respectively. In the DCA the preoperative model performs well within threshold survival probabilities of 20–50%. Most important limitation was the retrospective collection of this external validation dataset.

Conclusions

In this external validation, the pre- and postoperative nomograms predicting PoD following CN were well calibrated. Although performance of the preoperative nomogram was lower than in the internal validation, it retains the ability to predict early death after CN.

     

Design and internal validation of an obstetric early warning score: secondary analysis of the Intensive Care National Audit and Research Centre Case Mix Programme database

We designed and internally validated an aggregate weighted early warning scoring system specific to the obstetric population that has the potential for use in the ward environment. Direct obstetric admissions from the Intensive Care National Audit and Research Centre's Case Mix Programme Database were randomly allocated to model development (n = 2240) or validation (n = 2200) sets. Physiological variables collected during the first 24 h of critical care admission were analysed. Logistic regression analysis for mortality in the model development set was initially used to create a statistically based early warning score. The statistical score was then modified to create a clinically acceptable early warning score. Important features of this clinical obstetric early warning score are that the variables are weighted according to their statistical importance, a surrogate for the F_IO₂/P_aO₂ relationship is included, conscious level is assessed using a simplified alert/not alert variable, and the score, trigger thresholds and response are consistent with the new non‐obstetric National Early Warning Score system. The statistical and clinical early warning scores were internally validated using the validation set. The area under the receiver operating characteristic curve was 0.995 (95% CI 0.992–0.998) for the statistical score and 0.957 (95% CI 0.923–0.991) for the clinical score. Pre‐existing empirically designed early warning scores were also validated in the same way for comparison. The area under the receiver operating characteristic curve was 0.955 (95% CI 0.922–0.988) for Swanton et al.'s Modified Early Obstetric Warning System, 0.937 (95% CI 0.884–0.991) for the obstetric early warning score suggested in the 2003–2005 Report on Confidential Enquiries into Maternal Deaths in the UK, and 0.973 (95% CI 0.957–0.989) for the non‐obstetric National Early Warning Score. This highlights that the new clinical obstetric early warning score has an excellent ability to discriminate survivors from non‐survivors in this critical care data set. Further work is needed to validate our new clinical early warning score externally in the obstetric ward environment.     

Rapid identification of slow healing wounds

Chronic nonhealing wounds have a prevalence of 2% in the United States, and cost an estimated $50 billion annually. Accurate stratification of wounds for risk of slow healing may help guide treatment and referral decisions. We have applied modern machine learning methods and feature engineering to develop a predictive model for delayed wound healing that uses information collected during routine care in outpatient wound care centers. Patient and wound data was collected at 68 outpatient wound care centers operated by Healogics Inc. in 26 states between 2009 and 2013. The dataset included basic demographic information on 59,953 patients, as well as both quantitative and categorical information on 180,696 wounds. Wounds were split into training and test sets by randomly assigning patients to training and test sets. Wounds were considered delayed with respect to healing time if they took more than 15 weeks to heal after presentation at a wound care center. Eleven percent of wounds in this dataset met this criterion. Prognostic models were developed on training data available in the first week of care to predict delayed healing wounds. A held out subset of the training set was used for model selection, and the final model was evaluated on the test set to evaluate discriminative power and calibration. The model achieved an area under the curve of 0.842 (95% confidence interval 0.834–0.847) for the delayed healing outcome and a Brier reliability score of 0.00018. Early, accurate prediction of delayed healing wounds can improve patient care by allowing clinicians to increase the aggressiveness of intervention in patients most at risk.     

Development of Machine Learning Algorithms to Predict Patient Dissatisfaction After Primary Total Knee Arthroplasty

BackgroundPostoperative dissatisfaction after primary total knee arthroplasty (TKA) that requires additional care or readmission may impose a significant financial burden to healthcare systems. The purpose of the current study is to develop machine learning algorithms to predict dissatisfaction after TKA.MethodsA retrospective review of consecutive TKA patients between 2014 and 2016 from 1 large academic and 2 community hospitals was performed. Preoperative variables considered for prediction included demographics, medical history, flexion contracture, knee flexion, and outcome scores (patient-reported health state, Knee Society Score [KSS], and KSS-Function [KSS-F]). Recursive feature elimination was used to select features that optimized algorithm performance. Five supervised machine learning algorithms were developed by training with 10-fold cross-validation 3 times. These algorithms were subsequently applied to an independent testing set of patients and assessed by discrimination, calibration, Brier score, and decision curve analysis.ResultsOf 430 patients, a total of 40 (9.0%) were dissatisfied with their outcome after primary TKA at a minimum of 2 years postoperatively. The random forest algorithm achieved the best performance in the independent testing set not used for algorithm development (c-statistic: 0.77, calibration intercept: 0.087, calibration slope: 0.74, Brier score: 0.082). The most important factors for predicting dissatisfaction were age, number of medical comorbidities, presence of one or more drug allergies, preoperative patient-reported health state score, and preoperative KSS.ConclusionThe current study developed machine learning algorithms based on partially modifiable risk factors for predicting dissatisfaction after TKA. This model demonstrates good discriminative capacity for identifying those at greatest risk for dissatisfaction and may allow for preoperative health optimization.     

Development of a new nomogram for predicting the probability of a positive initial prostate biopsy in Japanese patients with serum PSA levels less than 10 ng/mL

Objectives:   Although several nomograms for prostate cancer detection have been developed for Western populations, the models constructed on Japanese data would be more useful for the Japanese population because of various differences between Western and Asian populations. We previously developed a model for predicting the probability of a positive initial prostate biopsy using clinical and laboratory data from Japanese males. In the present study, a predictive model for Japanese males with a prostate-specific antigen (PSA) < 10 ng/mL was developed to guide decision-making for prostate biopsies.
Methods:   The age, total PSA level, free to total PSA ratio, prostate volume, and the digital rectal examination findings of 1037 Japanese males with a PSA < 10 ng/mL undergoing initial prostate biopsy as part of individual screening were analyzed. For study validation, 20% of these data was randomly reserved. Logistic regression analysis estimated relative risk, 95% confidence intervals, and P -values.
Results:   Age and the independent predictors of a positive biopsy result (elevated PSA, decreased free to total PSA ratio, small prostate volume, and abnormal digital rectal examination findings) were used to develop a predictive nomogram. The area under the receiver operating characteristic curve was significantly higher for the model (73.0%) than for PSA alone (55.0%). If externally validated, the use of this nomogram could reduce unnecessary biopsies by 26% and overall prostate biopsies by 7.8%.
Conclusions:   This predictive nomogram could provide more precise risk-analysis information for individual Japanese patients with PSA levels less than 10 ng/mL and may help to identify patients who need a prostate biopsy.     

Establishment and validation of a novel predictive model to quantify the risk of bone metastasis in patients with prostate cancer

BackgroundPatients with prostate cancer (PCa) commonly suffer from bone metastasis during disease progression. This study aims to construct and validate a nomogram to quantify bone metastasis risk in patients with PCa.MethodsClinicopathological data of patients diagnosed with PCa between 2010 and 2015 were retrospectively retrieved from the Surveillance, Epidemiology, and End Results (SEER) database. Predictors for bone metastasis were identified by logistic regression analyses to establish a nomogram. The concordance index (c-index) and calibration plots were generated to assess the nomogram’s discrimination, and the area under the receiver operating characteristic curve (AUC) was used to compare the precision of the nomogram with routine staging systems. The nomogram’s clinical performance was evaluated by decision curve analysis (DCA) and clinical impact curves (CIC). Independent prognostic factors were identified by Cox regression analysis.ResultsA total of 168,414 eligible cases were randomly assigned to the training cohort or validation cohort at a ratio of 1:1. The nomogram, which was established based on independent factors, showed good accuracy, with c-indexes of 0.911 in the training set and 0.910 in the validation set. Calibration plots also approached 45 degrees. After other distant metastatic sites were included in the predictive model, the new nomogram displayed superior prediction performance. The AUCs and net benefit of the nomograms were both higher than those of other routine staging systems. Furthermore, bone metastasis prediction points were shown to be a new risk factor for overall survival.ConclusionsNovel validated nomograms can effectively predict the risk of bone metastasis in patients with PCa and help clinicians improve cancer management.     

预测T4a期胃癌腹膜转移危险因素的列线图模型的建立与验证

目的探究T_4a期胃癌发生腹膜转移的危险因素,以此为训练组构建列线图模型,并进行内验证及外验证。 方法回顾性分析2011年1月至2017年12月394例行胃切除术+D2淋巴结清扫术影像学诊断为T_4a期胃癌患者的临床资料,其中将2011年1月至2015年12月224例患者作为训练集,2016年1月至2017年12月170例患者作为验证集。根据术前影像学表现判断有无腹膜转移并且最终经病理资料证实,采用SPSS 24.0软件通过t检验、秩和检验或卡方检验对发生腹膜转移的危险因素进行统计学分析,经单因素和多因素Logistic回归筛选T_4a期胃癌患者发生腹膜转移的潜在危险因素,采用R软件（版本4.0.2）建立列线图模型。采用Bootstrap法进行内验证,采用ROC曲线评价模型的符合度并计算95%CI,绘制校准曲线评价模型的符合度。绘制DCA曲线评价模型的临床获益度。 结果224例训练集患者中,共37（16.5%）例患者发生腹膜转移,验证集患者中23（13.5%）例患者发生腹膜转移,多因素分析显示糖类抗原125（CA125）、腹水、术前白蛋白（ALB）和肿瘤分化程度是胃癌患者发生腹膜转移的独立危险因素。绘制ROC曲线结果提示,内部训练集AUC曲线下面积为0.783（95%CI：0.699-0.867）,外部验证集AUC曲线下面积为0.848（95%CI：0.763-0.932）;且以此建立的列线图模型具有良好的区分度、校准度和临床获益度。 结论基于4个独立危险因素的列线图模型对T_4a期胃癌患者腹膜转移的发生具有良好的区分度和校准度,可用于术前对T_4a期胃癌患者腹膜转移风险进行评估,具有一定的临床推广和参考价值。