Refining the 10-Year Prediction of Left Ventricular Systolic Dysfunction in Long-Term Survivors of Childhood Cancer

Background

In childhood cancer survivors (CCS) at risk for heart failure, echocardiographic surveillance recommendations are currently based on anthracyclines and chest-directed radiotherapy dose. Whether the ejection fraction (EF) measured at an initial surveillance echocardiogram can refine these recommendations is unknown.

Objectives

The purpose of this study was to assess the added predictive value of EF at >5 years after cancer diagnosis to anthracyclines and chest-directed radiotherapy dose in CCS, for the development of left ventricular dysfunction with an ejection fraction <40% (LVD40).

Methods

Echocardiographic surveillance was performed in 299 CCS from the Emma Children’s Hospital in the Netherlands. Cox regression models were built including cardiotoxic cancer treatment exposures with and without EF to estimate the probability of LVD40 at 10-year follow-up. Calibration, discrimination, and reclassification were assessed. Results were externally validated in 218 CCS.

Results

Cumulative incidences of LVD40 at 10-year follow-up were 3.7% and 3.6% in the derivation and validation cohort, respectively. The addition of EF resulted in an integrated area under the curve increase from 0.74 to 0.87 in the derivation cohort and from 0.72 to 0.86 in the validation cohort (likelihood ratio p < 0.001). Reclassification of CCS without LVD40 improved significantly (noncase continuous net reclassification improvement 0.50; 95% confidence interval [CI]: 0.40 to 0.60). A predicted LVD40 probability ≤3%, representing 75% of the CCS, had a negative predictive value of 99% (95% CI: 98% to 100%) for LVD40 within 10 years. However, patients with midrange EF (40% to 49%) at initial screening had an incidence of LVD40 of 11% and a 7.81-fold (95% CI: 2.07- to 29.50-fold) increased risk of LV40 at follow-up.

Conclusions

In CCS, an initial surveillance EF, in addition to anthracyclines and chest-directed radiotherapy dose, improves the 10-year prediction for LVD40. Through this strategy, both the identification of low-risk survivors in whom the surveillance frequency may be reduced and a group of survivors at increased risk of LVD40 could be identified.

A New Model for Caries Risk Prediction in Teenagers Using a Machine Learning Algorithm Based on Environmental and Genetic Factors

Dental caries is a multifactorial disease that can be caused by interactions between genetic and environmental risk factors. Despite the availability of caries risk assessment tools, caries risk prediction models incorporating new factors, such as human genetic markers, have not yet been reported. The aim of this study was to construct a new model for caries risk prediction in teenagers, based on environmental and genetic factors, using a machine learning algorithm. We performed a prospective longitudinal study of 1,055 teenagers (710 teenagers for cohort 1 and 345 teenagers for cohort 2) aged 13 years, of whom 953 (633 teenagers for cohort 1 and 320 teenagers for cohort 2) were followed for 21 months. All participants completed an oral health questionnaire, an oral examination, biological (salivary and cariostate) tests, and single nucleotide polymorphism sequencing analysis. We constructed a caries risk prediction model based on these data using a random forest with an AUC of 0.78 in cohort 1 (training cohort). We further verified the discrimination and calibration abilities of this caries risk prediction model using cohort 2. The AUC of the caries risk prediction model in cohort 2 (testing cohort) was 0.73, indicating high discrimination ability. Risk stratification revealed that our caries risk prediction model could accurately identify individuals at high and very high caries risk but underestimated risks for individuals at low and very low caries risk. Thus, our caries risk prediction model has the potential for use as a powerful community-level tool to identify individuals at high caries risk.

Construction of a Risk Prediction Model for Subsequent Bloodstream Infection in Intestinal Carriers of Carbapenem-Resistant Enterobacteriaceae: A Retrospective Study in Hematology Department and Intensive Care Unit

BACKGROUND

To establish a risk prediction model for carbapenem-resistant Enterobacteriaceae (CRE) bloodstream infection (BSI) in intestinal carriers.

METHODS

CRE screenings were performed every two weeks in hematology department and intensive care unit (ICU). Patients with positive CRE rectal swab screening were identified using electronic medical records from 15 May 2018 to 31 December 2019. Intestinal carriers who developed CRE BSI were compared with those who did not develop CRE infection. A 1:1 matched case-control study was conducted. The control group was selected by stratified random sampling based on the department to ensure that all the departments were represented. Univariate logistic analysis, multivariate logistic analysis and stepwise regression analysis were carried on a variety of patient factors and microbial factors.

RESULTS

A total of 42 cases were included. Multivariate analysis showed that gastrointestinal injury (OR 86.819, 95% CI 2.584-2916.592, P=0.013), tigecycline exposure (OR 14.991, 95% CI 1.816-123.737, P=0.012) and carbapenem resistance score (OR 11.236, 95% CI 1.811-69.700, P=0.009) were independent risk factors for CRE BSI in intestinal carriers (P<0.050). They were included in the Logistic regression model to predict BSI. According to receiver operating characteristic (ROC) curve analysis, the cut-off value of the model was 0.722, and the sensitivity, specificity and area under the curve (AUC) were 90.5%, 85.7% and 0.921, respectively.

CONCLUSION

The risk prediction model based on gastrointestinal injury, tigecycline exposure and carbapenem resistance score of colonizing strain can effectively predict CRE BSI in patients with CRE colonization. Early CRE screening and detection for inpatients in key departments may promote early warning and reduce the risk of nosocomial infection of CRE.

Incidence and Associated Risk Factors for Lactic Acidosis Induced by Linezolid Therapy in a Case-Control Study in Patients Older Than 85 Years

Background: Serum lactic acid is considered a prognostic indicator in critically ill patients. However, studies on linezolid-induced lactic acidosis (LILA) are still limited. Individuals older than 85 years old (very elderly) have limited capacity for organ compensation, and LILA data from these patients are lacking. In this study, we evaluated the risk factors for LILA in patients older than 85 years and established a risk prediction model for geriatric practice.

Methods: In this retrospective cohort study, blood gas analysis data and arterial lactate levels were monitored in patients older than 85 years during the use of teicoplanin or linezolid. After propensity score matching analyses, we compared the incidence of lactic acidosis between the teicoplanin and linezolid therapy groups and identified the risk factors of LILA.

Results: The incidence of lactic acidosis was found to be much lower in the group receiving teicoplanin than those receiving linezolid therapy (0 vs. 35.7%; p < 0.0001). A duration of linezolid therapy ≥ 9 days [odds ratio (OR), 3.541; 95% confidence interval (CI), 1.161–10.793; p = 0.026], an arterial blood glucose level ≥ 8 mmol/L (OR, 4.548; 95% CI, 1.507–13.725; p = 0.007), and a high sequential organ failure assessment score (OR, 1.429; 95% CI, 1.213–1.685; p < 0.0001) were risk factors for LILA. The constructed risk model could be used to predict LILA (area under the curve, 0.849; specificity, 65.1%; sensitivity, 91.4%, with a negative predictive value of 93.2% and a positive predictive value of 59.3%).

Conclusions: LILA can occur in patients older than 85 years after a relatively shorter duration of linezolid therapy. Therefore, close monitoring of blood gas and arterial lactate levels during linezolid therapy in the very elderly population is necessary.

Prediction of post-vaccination Guillain-Barré syndrome using data from a passive surveillance system

Purpose

Severe adverse events (AEs), such as Guillain‐Barré syndrome (GBS) occur rarely after influenza vaccination. We identify highly associated AEs with GBS and develop prediction models for GBS using the US Vaccine Adverse Event Reporting System (VAERS) reports following trivalent influenza vaccination (FLU3).

Methods

This study analyzed 80 059 reports from the US VAERS between 1990 and 2017. Several AEs were identified as highly associated with GBS and were used to develop the prediction model. Some common and mild AEs that were suspected to be underreported when GBS occurred simultaneously were removed from the final model. The analyses were validated using European influenza vaccine AEs data from EudraVigilance.

Results

Of the 80 059 reports, 1185 (1.5%) were annotated as GBS related. Twenty‐four AEs were identified as having strong association with GBS. The full prediction model, using age, sex, and all 24 AEs achieved an area under the receiver operating characteristic (ROC) curve (AUC) of 85.4% (90% CI: [83.8%, 86.9%]). After excluding the nine (e.g., pruritus, rash, injection site pain) likely underreported AEs, the final AUC became 77.5% (90% CI: [75.5%, 79.6%]). Two hundred and one (0.25%) reports were predicted as of high risk of GBS (predicted probability >25%) and 84 actually developed GBS.

Conclusion

The prediction performance demonstrated the potential of developing risk‐prediction models utilizing the VAERS cohort. Excluding the likely underreported AEs sacrificed some prediction power but made the model more interpretable and feasible. The high absolute risk of even a small number of AE combinations suggests the promise of GBS prediction within the VAERS dataset.

Monitoring of perfusion quality and prediction of donor heart function during ex-vivo machine perfusion by myocardial microcirculation versus surrogate parameters

Currently, lactate (Lac) is used to evaluate machine perfusion (MP) of hearts, donated after circulatory death (DCD). We hypothesize that monitoring of myocardial microcirculation (mLDP) by Laser-Doppler-Perfusion is superior to Lac to evaluate perfusion and predict contractility. In a pig model, DCD-hearts were perfused 4 hours followed by reperfusion and left ventricular contractility measurement. Lac and mLDP were measured every 30 min in successfully (N = 9) and unsuccessfully (N = 7) maintained hearts. Successfully maintained hearts showed decreasing Lac (5.6 to 2.8 mmol/L) and slightly downregulated (92%) mLDP. In unsuccessfully maintained hearts Lac first decreased (5.1 to 3.8 mmol/L) followed by increase and mLDP dropped to 39%. In a single-variable regression only mLDP showed a significant r² for systolic (0.514, p = 0.045) and diastolic (0.501, p = 0.049) parameters. The combination of mLDP and Lac (r² = 0.876, p = 0.005) showed best results. mLDP seems to be superior to Lac to show perfusion disorders and predict DCD-heart contractility.

Development and validation of nomograms for predicting blood loss in placenta previa with placenta increta or percreta

Background

To develop the risk prediction model of intraoperative massive blood loss in placenta previa with placenta increta or percreta.

Methods

This study included 260 patients, of whom 179 were allocated to the development group and 81 to the validation group. Univariate and multivariate logistic regression analyses were used to identify characteristics that were associated with massive blood loss (≥2,500 mL) during cesarean section. A nomogram was constructed based on regression coefficients. Receiver-operating characteristic curve, calibration curve, and decision curve analyses were applied to assess the discrimination, calibration, and performance of the model.

Results

Two models were constructed. The preoperative feature model (model A) consisted of vascular lacunae within the placenta and hypervascularity of the uterine-placental margin, uterine serosa-bladder wall interface, and cervix. The preoperative and surgical feature model (model B) consisted of an emergency cesarean section, no preoperative balloon placement of the abdominal aorta, and the previously mentioned four ultrasound signs. Model B had better discrimination than model A (area under the curve: development group: 0.839 vs. 0.732; validation group: 0.829 vs. 0.736). Model B showed a higher area under the decision curve than model A in both the training and validation groups.

Conclusions

The preoperative and surgical feature model for placenta previa with placenta increta or percreta can improve the early identification and management of patients who are at high risk of intraoperative massive blood loss.

A male-ABCD algorithm for hepatocellular carcinoma risk prediction in HBsAg carriers

Objective

Hepatocellular carcinoma (HCC) development among hepatitis B surface antigen (HBsAg) carriers shows gender disparity, influenced by underlying liver diseases that display variations in laboratory tests. We aimed to construct a risk-stratified HCC prediction model for HBsAg-positive male adults.

Methods

HBsAg-positive males of 35−69 years old (N=6,153) were included from a multi-center population-based liver cancer screening study. Randomly, three centers were set as training, the other three centers as validation. Within 2 years since initiation, we administrated at least two rounds of HCC screening using B-ultrasonography and α-fetoprotein (AFP). We used logistic regression models to determine potential risk factors, built and examined the operating characteristics of a point-based algorithm for HCC risk prediction.

Results

With 2 years of follow-up, 302 HCC cases were diagnosed. A male-ABCD algorithm was constructed including participant’s age, blood levels of GGT (γ-glutamyl-transpeptidase), counts of platelets, white cells, concentration of DCP (des-γ-carboxy-prothrombin) and AFP, with scores ranging from 0 to 18.3. The area under receiver operating characteristic was 0.91 (0.90−0.93), larger than existing models. At 1.5 points of risk score, 26.10% of the participants in training cohort and 14.94% in validation cohort were recognized at low risk, with sensitivity of identifying HCC remained 100%. At 2.5 points, 46.51% of the participants in training cohort and 33.68% in validation cohort were recognized at low risk with 99.06% and 97.78% of sensitivity, respectively. At 4.5 points, only 20.86% of participants in training cohort and 23.73% in validation cohort were recognized at high risk, with positive prediction value of 22.85% and 12.35%, respectively.

Conclusions

Male-ABCD algorithm identified individual’s risk for HCC occurrence within short term for their HCC precision surveillance.

Identification of risk factors of developing pressure injuries among immobile patient, and a risk prediction model establishment: A protocol for systematic review

BACKGROUD

Pressure injuries (PIs) bring a considerable physical and mental burden on immobile patients, and have put families and government under tremendous pressure to cover the cost of treatment. Therefore, this protocol proposes to identify risk factors of developing PIs in immobile patients from systematic reviews (SRs) and clinical practice guidelines (CPGs), in order to establish a risk prediction model for developing PIs and identify individual risk factors that can be modified to aid prevention.

METHODS

Electronic databases and specific databases for CPGs and SRs will be searched. Study selection and data collection will be performed independently by two reviewers. All included SRs and CPGs will be subject to critical appraisal. RevMan 5.3 will be used to calculate the pooled odds ratio (ORP) after appraising the quality of eligible studies, and the risk predictive model will be established using logistic regression model. A narrative synthesis, evidence summary table, and Sankey diagram will also be performed.

RESULTS

The results of this study will be submitted to a peer-reviewed journal for publication.

CONCLUSION

This systematic review will provide a risk prediction model of PI developing.

INSPLAY REGISTRATION NUMBER

INPLASY2020100097.

A Risk Prediction Model for Evaluating the Disease Progression of COVID-19 Pneumonia

Background and Objective: The epidemic of coronavirus disease 2019 (COVID-19) pneumonia caused by infection with severe acute respiratory syndrome coronavirus 2 (SARS-CoV2) has expanded from China throughout the world. This study aims to estimate the risk of disease progression of patients who have been confirmed with COVID-19. Methods: Meta-analysis was performed in existing literatures to identify risk factors associated with COVID-19 pneumonia progression. Patients with COVID-19 pneumonia were admitted to hospitals in Wuhan or Hangzhou were retrospectively enrolled. The risk prediction model and nomogram were developed from Wuhan cohort through logistic regression algorithm, and then validated in Hangzhou and Yinchuan cohorts. Results: A total of 270 patients admitted to hospital between Dec 30, 2019, and Mar 30, 2020, were retrospectively enrolled (Table 1). The development cohort (Wuhan cohort) included 87 (43%) men and 115 (57%) women, and the median age was 53 years old. Hangzhou validation cohort included 20 (48%) men and 22 (52%) women, and the median age was 59 years old. Yinchuan validation cohort included 12 (46%) men and 14 (54%) women, and the median age was 44 years old. The meta-analysis along with univariate logistic analysis in development cohort have shown that age, fever, diabetes, hypertension, CREA, BUN, CK, LDH, and neutrophil count were significantly associated with disease progression of COVID-19 pneumonia. The model and nomogram derived from development cohort show good performance in both development and validation cohorts. Conclusion: The severe COVID-19 pneumonia is associated with various types of risk factors including age, fever, comorbidities, and some laboratory examination indexes. The model integrated with these factors can help to evaluate the disease progression of COVID-19 pneumonia.

Establishment of risk prediction model and risk score for in-hospital mortality in patients with AECOPD

OBJECTIVE

Risk stratification for patients with acute exacerbation of chronic obstructive pulmonary disease (AECOPD) may help clinicians choose appropriate treatments and improve the quality of care.

METHODS

A total of 695 patients hospitalized with AECOPD from January 2015 to December 2017 were considered. They were assigned to a death and a survival cohort. The independent prognostic factors were determined by multivariate logistic regression analysis. Meanwhile, we also compared the new scale with three other scores and tested the new scale internally and externally.

RESULTS

A new risk score was created, made up of six independent variables: age, D-dimer, albumin, cardiac troponin I, partial pressure of carbon dioxide and oxygenation index. The area under the receiver operator characteristic curve (AUROC) for the model was 0.929, and the other three CURB-65, DECAF and BAP-65 models were 0.718, 0.922 and 0.708. The Cohen's kappa coefficient between the new scale and DECAF was calculated to be 0.648, suggesting that there is a substantial consistency between the two. In the internal and external validation cohorts, 490 and 500 patients were recruited with a total mortality rate of 5.15%. The AUROC for in-hospital mortality was 0.937 in the internal cohort and 0.914 in external cohort, which was significantly better than the scores for CURB-65 and BAP-65, but it was not significantly different from the DECAF.

CONCLUSIONS

The new scale may help to stratify the risk of in-hospital mortality of AECOPD. The DECAF performed as well as the new instrument, and it appears to be valid in Chinese patients.

Predictors of Residential Care Admission in Community-Dwelling Older People With Dementia

OBJECTIVES

The objectives of this study were to identify variables associated with dementia and entry into aged residential care (ARC) and derive and validate a risk prediction model for dementia and entry into ARC.

DESIGN

This was an observational study of prospectively collected Home Care International Residential Assessment Instrument (interRAI-HC) assessment data.

SETTING AND PARTICIPANTS

Participants included all people age ≥65 years who had completed an interRAI-HC assessment between July 1, 2012 and June 30, 2018. Exclusion criteria included death or entry into ARC within 30 days of assessment and not living at home at the time of the assessment.

MEASURES

InterRAI data from 94,202 older New Zealanders were evaluated for presence or absence of dementia. A multivariable competing-risks model for entry into ARC with death as the competing event was used to estimate subdistribution hazard ratios (SHR).

RESULTS

In total, there were 18,672 (19.8%) persons with dementia (PWD). PWD were almost twice as likely to enter ARC as persons without dementia [42.8% vs 25.3%; difference 17.5% (95% confidence interval 16.7%‒18.2%)]. PWD at highest risk of entering ARC were those where there was a desire to live elsewhere (SHR 1.44), depression (indicated, SHR 1.15), poor cognitive performance (Cognitive Performance Scale minimal SHR 1.32 and severe plus SHR 1.91), and wandering (SHR 1.19). Factors associated with reduced risks of PWD entering ARC were living with a child or relative, alcohol consumption, and comorbidities.

CONCLUSIONS AND IMPLICATIONS

A desire to live elsewhere, social isolation, independent activities of daily living, and depression were independently associated with entry into ARC. Supporting caregivers may improve outcomes for PWD that delay entry into ARC. Future revisions of the interRAI questionnaire could provide more insight on this matter.

Developing a prediction model for the self-evaluation of erectile dysfunction in an adult male population

The purpose of this study was to develop an erectile dysfunction (ED) risk assessment tool suitable for the general population. Based on an ED network survey of the general adult male population in China from October to November 2019, a total of 612 cases with a mean age of 31 years (interquartile range: 28-37) with valid data were collected: 357 cases were assigned to the training set and 255 to the validation set. The ED risk prediction model was established by multifactor logistic regression analysis, and nomograms were constructed for visualisation. In the validation set, a receiver operating characteristic curve, calibration curve analysis and decision curve analysis were used to evaluate the discrimination, calibration and clinical usefulness of the ED risk prediction model. Based on multivariate logistic regression, education, smoking, chronic diseases, feelings about one's spouse, frequency of sexual intercourse, masturbation and self-reported sexual satisfaction were selected as predictors to develop the ED prediction model. The model had good discrimination, calibration and clinical applicability. The ED risk prediction model developed in this study can effectively predict ED risk in the general population.

Inclusion of a gene-environment interaction between alcohol consumption and the aldehyde dehydrogenase 2 genotype in a risk prediction model for upper aerodigestive tract cancer in Japanese men

The well-known gene-environment interaction between alcohol consumption and aldehyde dehydrogenase 2 (ALDH2) genotype in upper aerodigestive tract cancer risk may improve our ability to identify high-risk subjects. Here, we developed and validated risk prediction models for this cancer in Japanese men and evaluated whether adding the gene-environment interaction to the model improved the predictive performance. We developed two case-cohort datasets in the Japan Public Health Center-based Prospective Study: one from subjects in the baseline survey for model development (108 cases and 4049 subcohort subjects) and the second from subjects in the 5-year follow-up survey for model validation (31 cases and 1527 subcohort subjects). We developed an environmental model including age, smoking status, and alcohol consumption, and a gene-environment interaction model including age, smoking status, and the combination of alcohol consumption and the ALDH2 genotype. We found a statistically significant gene-environment interaction for alcohol consumption and the ALDH2 genotype. The c-index for the gene-environment interaction model (0.71) was slightly higher than that for the environmental model (0.67). The values of integrated discrimination improvement and net reclassification improvement for the gene-environment interaction model were also slightly higher than those for the environmental model. Goodness-of-fit tests suggested that the models were well calibrated. Results from external model validation by the 5-year follow-up survey were consistent with those from the model development by the baseline survey. The addition of a gene-environment interaction to a lifestyle-based model might improve the performance to estimate the probability of developing upper aerodigestive tract cancer for Japanese men.

Risk Prediction Scoring System to Predict the Postsurgical Outcomes of Biliary Atresia

Aim

To find out association between liver function, liver histopathology and outcomes of biliary atresia (BA) following Kasai Portoenterostomy (KPE).

Materials and Methods

This is a retrospective study of children who underwent KPE at a single institute by single surgeon. The patient records analyzed and data of complete blood counts, liver function tests, coagulation profile and histopathology reports collected. The outcomes recorded as alive and jaundice free, alive but jaundiced, and deceased. Statistical analysis done using SPSS 23.

Observations

Total of 148 children operated during January 2000 to December 2018. Of these, 26 matched inclusion criteria. The parameters assessed were percentage of direct bilirubin, ratios of Aspartate transaminase (AST) to Alanine transaminase (ALT); Gamma glutamyl transferase (GGT) to AST; GGT to ALT and Aspartate transaminase to platelet ratio index (APRi). Among histopathology reports, fibrosis grade and bile ductular size noted. Among 26, 16 alive and ten are deceased. Among 16 alive, all are jaundice free. Of the parameters, ratio of AST to ALT, APRi and grade of fibrosis found statistically significant and further analysis showed if AST to ALT ratio < 2.1, APRi < 1.8 and grade of fibrosis < four, irrespective of age at surgery, had 96.2 % probability of successful KPE. Based on these observations, a scoring system and risk prediction model constructed based on Receiver operating characteristic (ROC) curves which are first in BA management.

Results and Conclusion

Although numbers are sufficient for statistical analysis, we further intend to validate the scoring system in a prospective trial. BA children can be subjected to risk prediction model and KPE performed in those who have a score less than seven and offered to those with score between eight and 16 out of 20.

Key Message

The scoring system and risk prediction model can guide in the management and post-operative follow up of children with biliary atresia.

Risk Prediction Model Development for Late On-Set Breast Cancer Screening in Low- and Middle-Income Societies: A Model Study for North Cyprus

Background: Early detection of breast cancer alters the prognosis and tools that can predict the risk for breast cancer in women will have a significant impact on healthcare systems in low- and middle-income regions, such as North Cyprus. Objective: In this study, we developed a simple breast cancer risk model for the women of North Cyprus. Methods: Data from 655 women, consisting of 318 breast cancer cases and 337 hospital-based controls, was used to develop and internally validate the model, external validation was carried out using, 653 women consisting of 126 cases and 527 controls. Data were obtained from medical records and interviews after informed consent. Results: A model was derived that consisted of age ≥50 years and <50 years and the presence and absence of >1 first-degree relatives (FDR) with breast cancer. From internal and external validations the model’s AUCs were, 0.66 (95% CI = 0.62–0.70) and 0.69 (95% CI = 0.63–0.74) respectively. Conclusions: A unique model for risk prediction of breast cancer was developed to aid in identifying high-risk women from North Cyprus that can benefit from mammogram screening. Further study on a large scale that includes environmental risk factors is warranted.

Breast cancer risk based on adapted IBIS prediction model in Slovenian women aged 40-49 years - could it be better?

Background The aim of the study was to assess the proportion of women that would be classified as at above-average risk of breast cancer based on the 10 year-risk prediction of the Slovenian breast cancer incidence rate (S-IBIS) program in two presumably above-average breast cancer risk populations in age group 40-49 years: (i) women referred for any reason to diagnostic breast centres and (ii) women who were diagnosed with breast cancer aged 40-49 years. Breast cancer is the commonest female cancer in Slovenia, with an incidence rate below European average. The Tyrer-Cuzick breast cancer risk assessment algorithm was recently adapted to S-IBIS. In Slovenia a tailored mammographic screening for women at above average risk in age group 40-49 years is considered in the future. S-IBIS is a possible tool to select population at above-average risk of breast cancer for tailored screening. Patients and methods In 357 healthy women aged 40-49 years referred for any reason to diagnostic breast centres and in 367 female breast cancer patients aged 40-49 years at time of diagnosis 10-years breast cancer risk was calculated using the S-IBIS software. The proportion of women classified as above-average risk of breast cancer was calculated for each subgroup of the study population. Results 48.7% of women in the Breast centre group and 39.2% of patients in the breast cancer group had above-average 10-year breast cancer risk. Positive family history of breast cancer was more prevalent in the Breast centre group (p < 0.05). Conclusions Inclusion of additional risk factors into the S-IBIS is warranted in the populations with breast cancer incidence below European average to reliably stratify women into breast cancer risk groups.

A mid-pregnancy risk prediction model for gestational diabetes mellitus based on the maternal status in combination with ultrasound and serological findings

Although previous studies have proposed predictive models of gestational diabetes mellitus (GDM) based on maternal status, they do not always provide reliable results. The present study aimed to create a novel model that included ultrasound data of maternal fat distribution and serum inflammatory factors. The clinical data of 1,158 pregnant women treated at Tangshan Gongren Hospital and eight other flagship hospitals in Tangshan, including the First Hospital of Tangshan Gongren Hospital group, Ninth Hospital of Tangshan Gongren Hospital group, Tangshan Gongren Hospital group rehabilitation hospital, Tangshan railway central hospital, Tangshan Gongren Hospital group Fengnan hospital, Tangshan Gongren Hospital group Qianan Yanshan hospital, Tangshan Gongren Hospital group Qianxi Kangli hospital and Tangshan Gongren Hospital group Jidong Sub-hospital, were analyzed following the division of subjects into GDM and non-GDM groups according to their diagnostic results at 24-28 weeks of pregnancy. Univariate analysis was performed to investigate the significance of the maternal clinical parameters for GDM diagnosis and a GDM prediction model was established using stepwise regression analysis. The predictive value of the model was evaluated using a Homer-Lemeshow goodness-of-fit test and a receiver operating characteristic curve (ROC). The model demonstrated that age, pre-pregnancy body mass index, a family history of diabetes mellitus, polycystic ovary syndrome, a history of GDM, high systolic pressures, glycosylated hemoglobin levels, triglyceride levels, total cholesterol levels, low-density lipoprotein cholesterol levels, serum hypersensitive C-reactive protein, increased subcutaneous fat thickness and visceral fat thickness were all correlated with an increased GDM risk (all P<0.01). The area under the curve value was 0.911 (95% CI, 0.893-0.930). Overall, the results indicated that the current model, which included ultrasound and serological data, may be a more effective predictor of GDM compared with other single predictor models. In conclusion, the present study developed a tool to determine the risk of GDM in pregnant women during the second trimester. This prediction model, based on various risk factors, demonstrated a high predictive value for the GDM occurrence in pregnant women in China and may prove useful in guiding future clinical practice.

Validation of Risk Prediction Models to Detect Asymptomatic Carotid Stenosis

Background Significant asymptomatic carotid stenosis (ACS) is associated with higher risk of strokes. While the prevalence of moderate and severe ACS is low in the general population, prediction models may allow identification of individuals at increased risk, thereby enabling targeted screening. We identified established prediction models for ACS and externally validated them in a large screening population. Methods and Results Prediction models for prevalent cases with ≥50% ACS were identified in a systematic review (975 studies reviewed and 6 prediction models identified [3 for moderate and 3 for severe ACS]) and then validated using data from 596 469 individuals who attended commercial vascular screening clinics in the United States and United Kingdom. We assessed discrimination and calibration. In the validation cohort, 11 178 (1.87%) participants had ≥50% ACS and 2033 (0.34%) had ≥70% ACS. The best model included age, sex, smoking, hypertension, hypercholesterolemia, diabetes mellitus, vascular and cerebrovascular disease, measured blood pressure, and blood lipids. The area under the receiver operating characteristic curve for this model was 0.75 (95% CI, 0.74-0.75) for ≥50% ACS and 0.78 (95% CI, 0.77-0.79) for ≥70% ACS. The prevalence of ≥50% ACS in the highest decile of risk was 6.51%, and 1.42% for ≥70% ACS. Targeted screening of the 10% highest risk identified 35% of cases with ≥50% ACS and 42% of cases with ≥70% ACS. Conclusions Individuals at high risk of significant ACS can be selected reliably using a prediction model. The best-performing prediction models identified over one third of all cases by targeted screening of individuals in the highest decile of risk only.

Risk prediction model for lung cancer incorporating metabolic markers: Development and internal validation in a Chinese population

BACKGROUND

Low-dose computed tomography screening has been proved to reduce lung cancer mortality, however, the issues of high false-positive rate and overdiagnosis remain unsolved. Risk prediction models for lung cancer that could accurately identify high-risk populations may help to increase efficiency. We thus sought to develop a risk prediction model for lung cancer incorporating epidemiological and metabolic markers in a Chinese population.

METHODS

During 2006 and 2015, a total of 122 497 people were observed prospectively for lung cancer incidence with the total person-years of 976 663. Stepwise multivariable-adjusted logistic regressions with P_entry  = .15 and P_stay  = .20 were conducted to select the candidate variables including demographics and metabolic markers such as high-sensitivity C-reactive protein (hsCRP) and low-density lipoprotein cholesterol (LDL-C) into the prediction model. We used the C-statistic to evaluate discrimination, and Hosmer-Lemeshow tests for calibration. Tenfold cross-validation was conducted for internal validation to assess the model's stability.

RESULTS

A total of 984 lung cancer cases were identified during the follow-up. The epidemiological model including age, gender, smoking status, alcohol intake status, coal dust exposure status, and body mass index generated a C-statistic of 0.731. The full model additionally included hsCRP and LDL-C showed significantly better discrimination (C-statistic = 0.735, P = .033). In stratified analysis, the full model showed better predictive power in terms of C-statistic in younger participants (<50 years, 0.709), females (0.726), and former or current smokers (0.742). The model calibrated well across the deciles of predicted risk in both the overall population (P_HL  = .689) and all subgroups.

CONCLUSIONS

We developed and internally validated an easy-to-use risk prediction model for lung cancer among the Chinese population that could provide guidance for screening and surveillance.

Systematic review: risk prediction models for recurrence of hepatocellular carcinoma after liver transplantation

Recurrence of hepatocellular carcinoma (HCC) after liver transplantation (LT) is a significant clinical problem associated with poor surgical outcomes. This study aims to summarize the current evidence on risk prediction models of HCC recurrence after LT. PubMed and EMBASE were searched to May 25, 2019, for relevant articles. Studies originally designed to develop or validate a risk prediction model for HCC recurrence after LT were included. Two independent authors summarized the study characteristics and evaluated the risk of bias and applicability concerns in the included studies. From 26 included studies, 18 original risk prediction models were determined, but only five models were externally validated. The average number of predictors involved in the construction of risk models was three. The most frequently employed predictors were alpha-fetoprotein, tumor size, vascular invasion, tumor number, tumor differentiation, and neutrophil-lymphocyte ratio. Most studies showed good discriminatory performance (AUC >0.75). The overall quality of the included studies was generally low. Most of the original models lacked the highly recommended external and prospective validation in diverse populations. The AFP model was the well-validated preoperative risk model that can stratify patients into high- and low-risk groups.

Developing risk models for multicenter data using standard logistic regression produced suboptimal predictions: A simulation study

Although multicenter data are common, many prediction model studies ignore this during model development. The objective of this study is to evaluate the predictive performance of regression methods for developing clinical risk prediction models using multicenter data, and provide guidelines for practice. We compared the predictive performance of standard logistic regression, generalized estimating equations, random intercept logistic regression, and fixed effects logistic regression. First, we presented a case study on the diagnosis of ovarian cancer. Subsequently, a simulation study investigated the performance of the different models as a function of the amount of clustering, development sample size, distribution of center‐specific intercepts, the presence of a center‐predictor interaction, and the presence of a dependency between center effects and predictors. The results showed that when sample sizes were sufficiently large, conditional models yielded calibrated predictions, whereas marginal models yielded miscalibrated predictions. Small sample sizes led to overfitting and unreliable predictions. This miscalibration was worse with more heavily clustered data. Calibration of random intercept logistic regression was better than that of standard logistic regression even when center‐specific intercepts were not normally distributed, a center‐predictor interaction was present, center effects and predictors were dependent, or when the model was applied in a new center. Therefore, to make reliable predictions in a specific center, we recommend random intercept logistic regression.

'Reduced' HUNT model outperforms NLST and NELSON study criteria in predicting lung cancer in the Danish screening trial

Hypothesis

We hypothesise that the validated HUNT Lung Cancer Risk Model would perform better than the NLST (USA) and the NELSON (Dutch‐Belgian) criteria in the Danish Lung Cancer Screening Trial (DLCST).

Methods

The DLCST measured only five out of the seven variables included in validated HUNT Lung Cancer Model. Therefore a ‘Reduced’ model was retrained in the Norwegian HUNT2-cohort using the same statistical methodology as in the original HUNT model but based only on age, pack years, smoking intensity, quit time and body mass index (BMI), adjusted for sex. The model was applied on the DLCST-cohort and contrasted against the NLST and NELSON criteria.

Results

Among the 4051 smokers in the DLCST with 10 years follow-up, median age was 57.6, BMI 24.75, pack years 33.8, cigarettes per day 20 and most were current smokers. For the same number of individuals selected for screening, the performance of the ‘Reduced’ HUNT was increased in all metrics compared with both the NLST and the NELSON criteria. In addition, to achieve the same sensitivity, one would need to screen fewer people by the ‘Reduced’ HUNT model versus using either the NLST or the NELSON criteria (709 vs 918, p=1.02e-11 and 1317 vs 1668, p=2.2e-16, respectively).

Conclusions

The ‘Reduced’ HUNT model is superior in predicting lung cancer to both the NLST and NELSON criteria in a cost-effective way. This study supports the use of the HUNT Lung Cancer Model for selection based on risk ranking rather than age, pack year and quit time cut-off values. When we know how to rank personal risk, it will be up to the medical community and lawmakers to decide which risk threshold will be set for screening.

Individual and joint performance of DNA methylation profiles, genetic risk score and environmental risk scores for predicting breast cancer risk

DNA methylation patterns in the blood, genetic risk scores (GRSs), and environmental risk factors can potentially improve breast cancer (BC) risk prediction. We assessed the individual and joint predictive performance of methylation, GRS, and environmental risk factors for BC incidence in a prospective cohort study. In a cohort of 5462 women aged 50–75 from Germany, 101 BC cases were identified during 14 years of follow‐up and were compared to 263 BC‐free controls in a nested case–control design. Three previously suggested methylation risk scores (MRSs) based on methylation of 423, 248, and 131 cytosine‐phosphate‐guanine (CpG) loci, and a GRS based on the risk alleles from 269 recently identified single nucleotide polymorphisms were constructed. Additionally, multiple previously proposed environmental risk scores (ERSs) were built based on environmental variables. Areas under the receiver operating characteristic curves (AUCs) were estimated for evaluating BC risk prediction performance. MRS and ERS showed limited accuracy in predicting BC incidence, with AUCs ranging from 0.52 to 0.56 and from 0.52 to 0.59, respectively. The GRS predicted BC incidence with a higher accuracy (AUC = 0.61). Adjusted odds ratios per standard deviation increase (95% confidence interval) were 1.07 (0.84–1.36) and 1.40 (1.09–1.80) for the best performing MRS and ERS, respectively, and 1.48 (1.16–1.90) for the GRS. A full risk model combining the MRS, GRS, and ERS predicted BC incidence with the highest accuracy (AUC = 0.64) and might be useful for identifying high‐risk populations for BC screening.

Integrated genomic and methylation profile analysis to identify candidate tumor marker genes in patients with colorectal cancer

Aberrant genomic expression and methylation serve important roles in cancer development. Integrated analysis of genetic and methylation profiles may identify potential tumor marker genes for colorectal cancer (CRC) prediction. In the current study, DNA methylation and mRNA expression profiles associated with CRC were downloaded from The Cancer Genome Atlas database. Differentially expressed mRNAs and methylated genes between tumor samples and adjacent healthy tissues were identified. Candidate tumor marker genes and prognostic clinical factors were screened according to univariable and multivariable Cox regression analysis. A total of 218 DEGs with aberrant methylation levels were screened from tumor samples. A risk prediction model was constructed based on identified genes and clinical factors. Randomization tests were used to evaluate the performance of the prediction model, including area under the curve (AUC) calculation and cross-validation. Cox regression analysis revealed that eight genes and six prognostic clinical factors were significantly associated with survival outcomes. Functional and pathway enrichment analysis revealed that the eight genes were mainly involved in ‘cell adhesion’, ‘fatty acid metabolism’ and ‘cytokine receptor interaction’ pathways. After combining six clinical factors with eight genes, the accuracy of risk prediction model has been increased intensively. The P-values representing the association between risk grouping and prognosis decreased from 0.009 to 0.001 and the AUC increased from 0.992 to 0.999, indicating that the comprehensive risk prediction model exhibited a good performance for disease prognosis prediction. The current study integrated genomic and methylation profiles and identified eight tumor marker genes in CRC. These candidate genes may improve the prediction accuracy of CRC prognosis.

Excess weight loss is a vital strategy for controlling hypertension among multi-ethnic population in northwest China: A cross-sectional analysis

Hypertension is a common global health problem including China. This study aimed to assess the prevalence and awareness of hypertension, and evaluate risk factors associated with hypertension among multi-ethnic population in northwest China using a random sampling cross-sectional data.A cross-sectional survey was conducted between 2014 and 2015 as part of a nationwide survey using stratified four-stage random sampling in Xinjiang. Hypertension was defined as mean systolic blood pressure (SBP) and/or diastolic blood pressure (DBP) ≥140/90 mm Hg and/or taking anti-hypertensive medication. In addition, the prevalence of hypertension (SBP ≥ 130 or DBP ≥ 80 mm Hg) was also estimated according to the 2017 American College of Cardiology (ACC)/American Heart Association (AHA) High Blood Pressure Guideline. Awareness of hypertension was based on self-report. An optimized risk score model was used to assess the risk and determine the predictive power of risk factors on hypertension.Totally 6722 subjects aged ≥18 years were enrolled and prevalence of hypertension was 24.3%, while the prevalence of hypertension based on the 2017 ACC/AHA guideline was approximately twice as high as that based on 2010 Chinese guideline (37.6%). Among individuals with hypertension, 55.5% were aware of their condition. Six potential factors were estimated to be associated with increased risk of hypertension including age, ethnicity, marital status, body mass index (BMI), waistline circumference, and comorbidity. In the analyses of calculated risk score, BMI ≥ 28.0 corresponded to the highest risk score of 23 points. The area under the receiver operation curve for the multivariable prediction model was 0.803 (95%CI: 0.789-0.813).There is a considerable prevalence of hypertension among Xinjiang adults, northwest China; awareness of hypertension is low. Excess weight loss may be a vital strategy for controlling hypertension, particularly if accompanied with other preventive measures in this region.

Understanding the relation between Zika virus infection during pregnancy and adverse fetal, infant and child outcomes: a protocol for a systematic review and individual participant data meta-analysis of longitudinal studies of pregnant women and their infants and children

INTRODUCTION

Zika virus (ZIKV) infection during pregnancy is a known cause of microcephaly and other congenital and developmental anomalies. In the absence of a ZIKV vaccine or prophylactics, principal investigators (PIs) and international leaders in ZIKV research have formed the ZIKV Individual Participant Data (IPD) Consortium to identify, collect and synthesise IPD from longitudinal studies of pregnant women that measure ZIKV infection during pregnancy and fetal, infant or child outcomes.

METHODS AND ANALYSIS

We will identify eligible studies through the ZIKV IPD Consortium membership and a systematic review and invite study PIs to participate in the IPD meta-analysis (IPD-MA). We will use the combined dataset to estimate the relative and absolute risk of congenital Zika syndrome (CZS), including microcephaly and late symptomatic congenital infections; identify and explore sources of heterogeneity in those estimates and develop and validate a risk prediction model to identify the pregnancies at the highest risk of CZS or adverse developmental outcomes. The variable accuracy of diagnostic assays and differences in exposure and outcome definitions means that included studies will have a higher level of systematic variability, a component of measurement error, than an IPD-MA of studies of an established pathogen. We will use expert testimony, existing internal and external diagnostic accuracy validation studies and laboratory external quality assessments to inform the distribution of measurement error in our models. We will apply both Bayesian and frequentist methods to directly account for these and other sources of uncertainty.

ETHICS AND DISSEMINATION

The IPD-MA was deemed exempt from ethical review. We will convene a group of patient advocates to evaluate the ethical implications and utility of the risk stratification tool. Findings from these analyses will be shared via national and international conferences and through publication in open access, peer-reviewed journals.

TRIAL REGISTRATION NUMBER

PROSPERO International prospective register of systematic reviews (CRD42017068915).

Prediction of colorectal cancer risk among adults in a lower middle-income country

Background

Globally, colorectal cancer (CRC) is ranked as the third most common cancer in men and the second in women. Use of a simple, validated risk prediction tool will offer a low-cost mechanism to identify the high-risk individuals for CRC. This will increase efficient use of limited resources and early identification of patients. The aim of our study was to develop and validate a risk prediction model for developing CRC for Sri Lankan adults.

Methods

The risk predictors were based on the risk factors identified through a logistic regression model along with expert opinion. A case control design utilizing 65 CRC new cases and 65 hospital controls aged 30 years or more was used to assess the criterion validity and reliability of the model. The information was obtained using an interviewer administered questionnaire based on the risk prediction model.

Results

The developed model consisted of eight predictors with an area under the curve (AUC) of 0.849 (95% CI: 0.8 to 0.9, P<0.001). It has a sensitivity of 76.9%, specificity of 83.1%, positive predictive value (PPV) of 82.0%, negative predictive value (NPV) of 79.3%. Positive and negative likelihood ratios are 4.6 and 0.3. Test re-test reliability revealed a Kappa coefficient of 0.88.

Conclusions

The model developed to predict the risk of CRC among adults aged 30 years and above was proven to be valid and reliable and it is an effective tool to be used as the first step to identify the high-risk population who should be referred for colonoscopy examination.

Construction of a flow chart-like risk prediction model of ganciclovir-induced neutropaenia including severity grade: A data mining approach using decision tree

WHAT IS KNOWN AND OBJECTIVE

Haematological toxicities such as neutropaenia are a common side effect of ganciclovir (GCV); however, risk factors for GCV-induced neutropaenia have not been well established. Decision tree (DT) analysis is a typical technique of data mining consisting of a flow chart-like framework that shows various outcomes from a series of decisions. By following the flow chart, users can estimate combinations of risk factors that may increase the probability of certain events. In our previous study, we demonstrated the usefulness of this approach in the evaluation of adverse drug reactions. Therefore, we aimed to construct a risk prediction model of GCV-induced neutropaenia including severity grade.

METHODS

We performed a retrospective study at the Hokkaido University Hospital and enrolled patients who received GCV between April 2008 and March 2018. Neutropaenia was defined as an absolute neutrophil count (ANC) <1500 cells/mm³ and a decrease to <75% relative to baseline. We classified the patients who developed neutropaenia in three groups (Grades 2-4) based on the National Cancer Institute-Common Terminology Criteria for Adverse Events. Data collection was achieved through the retrieval of medical records. We employed a chi-squared automatic interaction detection algorithm to construct the DT model and compared the accuracies to the logistic regression model (a conventional statistical method) to evaluate the established model.

RESULTS AND DISCUSSION

In total, 396 adult patients were included in the study; 61 (15.4%) developed neutropaenia. Three predictive factors (hematopoietic stem cell transplantation, baseline ANC <3854 cells/mm³ and duration of therapy ≥15 days) were extracted using the DT analysis to produce five subgroups, the incidence of neutropaenia ranged between 1.7% and 52.8%. In each subgroup, patients who developed neutropaenia were categorized based on the severity. The accuracies of each model were the same (84.6%), which indicated precision.

WHAT IS NEW AND CONCLUSION

We successfully built a risk prediction model of GCV-induced neutropaenia including severity grade. This model is expected to assist decision-making in the clinical setting.

Prediction of the 1-Year Risk of Incident Lung Cancer: Prospective Study Using Electronic Health Records from the State of Maine

BACKGROUND

Lung cancer is the leading cause of cancer death worldwide. Early detection of individuals at risk of lung cancer is critical to reduce the mortality rate.

OBJECTIVE

The aim of this study was to develop and validate a prospective risk prediction model to identify patients at risk of new incident lung cancer within the next 1 year in the general population.

METHODS

Data from individual patient electronic health records (EHRs) were extracted from the Maine Health Information Exchange network. The study population consisted of patients with at least one EHR between April 1, 2016, and March 31, 2018, who had no history of lung cancer. A retrospective cohort (N=873,598) and a prospective cohort (N=836,659) were formed for model construction and validation. An Extreme Gradient Boosting (XGBoost) algorithm was adopted to build the model. It assigned a score to each individual to quantify the probability of a new incident lung cancer diagnosis from October 1, 2016, to September 31, 2017. The model was trained with the clinical profile in the retrospective cohort from the preceding 6 months and validated with the prospective cohort to predict the risk of incident lung cancer from April 1, 2017, to March 31, 2018.

RESULTS

The model had an area under the curve (AUC) of 0.881 (95% CI 0.873-0.889) in the prospective cohort. Two thresholds of 0.0045 and 0.01 were applied to the predictive scores to stratify the population into low-, medium-, and high-risk categories. The incidence of lung cancer in the high-risk category (579/53,922, 1.07%) was 7.7 times higher than that in the overall cohort (1167/836,659, 0.14%). Age, a history of pulmonary diseases and other chronic diseases, medications for mental disorders, and social disparities were found to be associated with new incident lung cancer.

CONCLUSIONS

We retrospectively developed and prospectively validated an accurate risk prediction model of new incident lung cancer occurring in the next 1 year. Through statistical learning from the statewide EHR data in the preceding 6 months, our model was able to identify statewide high-risk patients, which will benefit the population health through establishment of preventive interventions or more intensive surveillance.

Probability of cancer in lung nodules using sequential volumetric screening up to 12 months: the UKLS trial

BACKGROUND

Estimation of the clinical probability of malignancy in patients with pulmonary nodules will facilitate early diagnosis, determine optimum patient management strategies and reduce overall costs.

METHODS

Data from the UK Lung Cancer Screening trial were analysed. Multivariable logistic regression models were used to identify independent predictors and to develop a parsimonious model to estimate the probability of lung cancer in lung nodules detected at baseline and at 3-month and 12-month repeat screening.

RESULTS

Of 1994 participants who underwent CT scan, 1013 participants had a total of 5063 lung nodules and 52 (2.6%) of the participants developed lung cancer during a median follow-up of 4 years. Covariates that predict lung cancer in our model included female gender, asthma, bronchitis, asbestos exposure, history of cancer, early and late onset of family history of lung cancer, smoking duration, FVC, nodule type (pure ground-glass and part-solid) and volume as measured by semiautomated volumetry. The final model incorporating all predictors had excellent discrimination: area under the receiver operating characteristic curve (AUC 0.885, 95% CI 0.880 to 0.889). Internal validation suggested that the model will discriminate well when applied to new data (optimism-corrected AUC 0.882, 95% CI 0.848 to 0.907). The risk model had a good calibration (goodness-of-fit χ[8] 8.13, p=0.42).

CONCLUSIONS

Our model may be used in estimating the probability of lung cancer in nodules detected at baseline and at 3 months and 12 months from baseline, allowing more efficient stratification of follow-up in population-based lung cancer screening programmes.

TRIAL REGISTRATION NUMBER

78513845.

Predicting survival in out-of-hospital cardiac arrest patients undergoing targeted temperature management: The Polish Hypothermia Registry Risk Score

BACKGROUND

Prompt reperfusion and post-resuscitation care, including targeted temperature management (TTM), improve survival in out-of-hospital cardiac arrest (OHCA) patients. To predict inhospital mortality in OHCA patients treated with TTM, the Polish Hypothermia Registry Risk Score (PHR-RS) was developed. The use of dedicated risk stratification tools may support treatment decisions.

METHODS

Three hundred seventy-six OHCA patients who underwent TTM between 2012 and 2016 were retrospectively analysed and whose data were collected in the Polish Hypothermia Registry. A multivariate logistic regression model identified a set of predictors of in-hospital mortality that were used to develop a dedicated risk prediction model, which was tested for accuracy.

RESULTS

The mean age of the studied population was 59.2 ± 12.9 years. 80% of patients were male, 73.8% had shockable rhythms, and mean time from cardiac arrest (CA) to cardiopulmonary resuscitation (CPR) was 7.2 ± 8.6 min. The inputs for PHR-RS were patient age and score according to the Mild Therapeutic Hypothermia (MTH) Scale. Criteria for the MTH score consisted of time from CA to CPR above 10 min, time from CA to the return of spontaneous circulation above 20 min, in-hospital CA, unwitnessed CA, and non-shockable rhythm, each counted as 1 point. The predictive value of PHR-RS was expressed as an area under the curve of 0.74.

CONCLUSIONS

PHR-RS is one of the simplest and easiest models to use and enables a reliable prediction of in-hospital mortality in OHCA patients treated with TTM.

Genetically elevated circulating homocysteine concentrations increase the risk of diabetic kidney disease in Chinese diabetic patients

Diabetic kidney disease (DKD) is a devastating and frequent complication of diabetes mellitus. Here, we first adopted methylenetetrahytrofolate reductase (MTHFR) gene C677T polymorphism as an instrument to infer the possible causal relevance between circulating homocysteine and DKD risk in a Chinese population and next attempted to build a risk prediction model for DKD. This is a hospital‐based case‐control association study. Total 1107 study participants were diagnosed with type 2 diabetes mellitus, including 547 patients with newly diagnosed and histologically confirmed DKD. MTHFR gene C677T polymorphism was determined using the TaqMan method. Carriers of 677TT genotype (14.55 μmol/L) had significantly higher homocysteine concentrations than carriers of 677CT genotype (12.88 μmol/L) (P < 0.001). Carriers of 677TT genotype had a 1.57‐fold increased risk of DKD (odds ratio: 1.57, 95% CI: 1.21‐2.05, P = 0.001) relative to carriers of 677CT genotype after adjusting for confounders. Mendelian randomization analysis revealed that the odds ratio for DKD relative to diabetes mellitus per 5 μmol/L increment of circulating homocysteine concentrations was 3.86 (95% confidence interval: 1.21‐2.05, P < 0.001). In the Logistic regression analysis, hypertension, homocysteine and triglyceride were significantly associated with an increased risk of DKD and they constituted a risk prediction model with good test performance and discriminatory capacity. Taken together, our findings provide evidence that elevated circulating homocysteine concentrations were causally associated with an increased risk of DKD in Chinese diabetic patients.

Construction of a risk prediction model of vancomycin-associated nephrotoxicity to be used at the time of initial therapeutic drug monitoring: A data mining analysis using a decision tree model

OBJECTIVES

In our previous study, we built a risk prediction model of vancomycin (VCM)-associated nephrotoxicity using decision tree (DT) analysis. However, this has several limitations in clinical applications. Our objective here was to construct a clinically applicable risk prediction model to be used at the time of initial therapeutic drug monitoring (TDM), in patients with uncomplicated infections.

METHOD

A retrospective study was conducted at Hokkaido University Hospital. Subjects that had received VCM were extracted between November 2011 and April 2017. Nephrotoxicity was defined as an increase in serum creatinine of 0.5 mg/dL or 50% or higher from baseline. The additional inclusion criteria in this study were as follows: (1) the target trough level of VCM was set to 10 to 15 mg/L, and (2) the duration of therapy was 7 to 14 days. Patients were assumed to have uncomplicated infections. Risk factors for nephrotoxicity were evaluated, which could be extracted at the initial TDM. In the DT analysis, a chi-squared automatic interaction detection algorithm was constructed.

RESULTS

A total of 402 patients were enrolled, and 56 (13.9%) patients developed nephrotoxicity. In the DT analysis, concomitant medications (furosemide, piperacillin-tazobactam, and vasopressor drugs) and an initial VCM trough concentration ≥ 15.0 mg/L were extracted as predictive variables by which patients were divided into six subgroups. The incidence of nephrotoxicity was 5.2% to 70.0%, with subgroups classified as low to high risk of nephrotoxicity. The accuracy of DT model was favourable (87.1%).

CONCLUSION

We propose that the DT model built in this study is applicable to clinical practice.

Development and validation of prediction models to estimate risk of primary total hip and knee replacements using data from the UK: two prospective open cohorts using the UK Clinical Practice Research Datalink

Objectives

The ability to efficiently and accurately predict future risk of primary total hip and knee replacement (THR/TKR) in earlier stages of osteoarthritis (OA) has potentially important applications. We aimed to develop and validate two models to estimate an individual’s risk of primary THR and TKR in patients newly presenting to primary care.

Methods

We identified two cohorts of patients aged ≥40 years newly consulting hip pain/OA and knee pain/OA in the Clinical Practice Research Datalink. Candidate predictors were identified by systematic review, novel hypothesis-free ‘Record-Wide Association Study’ with replication, and panel consensus. Cox proportional hazards models accounting for competing risk of death were applied to derive risk algorithms for THR and TKR. Internal–external cross-validation (IECV) was then applied over geographical regions to validate two models.

Results

45 predictors for THR and 53 for TKR were identified, reviewed and selected by the panel. 301 052 and 416 030 patients newly consulting between 1992 and 2015 were identified in the hip and knee cohorts, respectively (median follow-up 6 years). The resultant model C-statistics is 0.73 (0.72, 0.73) and 0.79 (0.78, 0.79) for THR (with 20 predictors) and TKR model (with 24 predictors), respectively. The IECV C-statistics ranged between 0.70–0.74 (THR model) and 0.76–0.82 (TKR model); the IECV calibration slope ranged between 0.93–1.07 (THR model) and 0.92–1.12 (TKR model).

Conclusions

Two prediction models with good discrimination and calibration that estimate individuals’ risk of THR and TKR have been developed and validated in large-scale, nationally representative data, and are readily automated in electronic patient records.

Risk prediction of post-hepatectomy liver failure in patients with perihilar cholangiocarcinoma

BACKGROUND AND AIM

In most patients with perihilar cholangiocarcinoma (PHCC), major hepatectomy and extrahepatic bile duct resection are needed for surgical radicality, and a high risk of hepatic insufficiency exists. This study aims to develop a prediction model for post-hepatectomy liver failure (PHLF) in patients with PHCC.

METHODS

A total of 143 patients who underwent major liver resection and extrahepatic bile duct resection for PHCC between October 2001 and December 2013 were included. Clinically relevant PHLF was defined as liver failure corresponding to grade B or C of the International Study Group of Liver Surgery criteria. Multivariate logistic regression was used to develop the PHLF risk model. Model performance was evaluated internally using the area under the curve analysis (discrimination) after 1000 bootstrap resampling and the Hosmer-Lemeshow goodness-of-fit test (calibration).

RESULTS

Post-hepatectomy liver failure occurred in 43.4% of patients (n = 62). In multivariate analysis, PHLF was significantly associated with future liver remnant ratio (odds ratio [OR] per 10% = 0.68, 95% confidence interval [CI] 0.51-0.88), intraoperative blood loss (OR per 1 L = 1.82, 95% CI 1.11-3.17), and preoperative prothrombin time > 1.20 (OR = 3.22, 95% CI 1.15-9.97). The PHLF risk score model showed good discrimination (area under the curve = 0.708, 95% CI 0.623-0.793) and calibration (P = 0.227).

CONCLUSIONS

The risk model proposed in this study accurately predicted PHLF in patients with PHCC. This offers surgeons a practical guide to quantitative risk assessment of hepatic insufficiency and aids decision-making in surgical treatment and perioperative management.

Validation of a risk prediction model for Barrett's esophagus in an Australian population

Background

Esophageal adenocarcinoma is a disease that has a high mortality rate, the only known precursor being Barrett's esophagus (BE). While screening for BE is not cost-effective at the population level, targeted screening might be beneficial. We have developed a risk prediction model to identify people with BE, and here we present the external validation of this model.

Materials and methods

A cohort study was undertaken to validate a risk prediction model for BE. Individuals with endoscopy and histopathology proven BE completed a questionnaire containing variables previously identified as risk factors for this condition. Their responses were combined with data from a population sample for analysis. Risk scores were derived for each participant. Overall performance of the risk prediction model in terms of calibration and discrimination was assessed.

Results

Scores from 95 individuals with BE and 636 individuals from the general population were analyzed. The Brier score was 0.118, suggesting reasonable overall performance. The area under the receiver operating characteristic was 0.83 (95% CI 0.78-0.87). The Hosmer-Lemeshow statistic was p=0.14. Minimizing false positives and false negatives, the model achieved a sensitivity of 74% and a specificity of 73%.

Conclusion

This study has validated a risk prediction model for BE that has a higher sensitivity than previous models.

Combining ctnnb1 genetic variability with epidemiologic factors to predict lung cancer susceptibility

OBJECTIVE

Early detection and diagnosis of lung cancer remain challenging but would improve patient prognosis. The goal of this study is to develop a model to estimate the risk of lung cancer for a given individual.

METHODS

We conducted a case-control study to develop a predictive model to identify individuals at high risk for lung cancer. Clinical data from 500 lung cancer patients and 500 population-based age- and gender-matched controls were used to develop and evaluate the model. Associations between environmental variants together with single nucleotide polymorphisms (SNPs) of beta-catenin (ctnnb1) and lung cancer risk were analyzed using a logistic regression model. The predictive accuracy of the model was determined by calculating the area under the receiver operating characteristic (ROC) curve.

RESULTS

Prior diagnosis of chronic obstructive pulmonary disease (COPD), pulmonary tuberculosis, family history of cancer, and smoking are lung cancer risk factors. The area under the curve (AUC) was 0.740, and the sensitivity, specificity, and Youden index were 0.718, 0.660, and 0.378, respectively.

CONCLUSION

Our risk prediction model for lung cancer is useful for distinguishing high-risk individuals.

A Novel Risk prediction Model for Patients with Combined Hepatocellular-Cholangiocarcinoma

Backgrounds: Regarding the difficulty of CHC diagnosis and potential adverse outcomes or misuse of clinical therapies, an increasing number of patients have undergone liver transplantation, transcatheter arterial chemoembolization (TACE) or other treatments.

Objective: To construct a convenient and reliable risk prediction model for identifying high-risk individuals with combined hepatocellular-cholangiocarcinoma (CHC).

Methods: 3369 patients who underwent surgical resection for liver cancer at Zhongshan Hospital were enrolled in this study. The epidemiological and clinical characteristics of the patients were collected at the time of tumor diagnosis. Variables (P <0.25 in the univariate analyses) were evaluated using backward stepwise method. A receiver operating characteristic (ROC) curve was used to assess model discrimination. Calibration was performed using the Hosmer-Lemeshow test and a calibration curve. Internal validation was performed using a bootstrapping approach.

Results: Among the entire study population, 250 patients (7.42%) were pathologically defined with CHC. Age, HBcAb, red blood cells (RBC), blood urea nitrogen (BUN), AFP, CEA and portal vein tumor thrombus (PVTT) were included in the final risk prediction model (area under the curve, 0.69; 95% confidence interval, 0.51-0.77). Bootstrapping validation presented negligible optimism. When the risk threshold of the prediction model was set at 20%, 2.73% of the patients diagnosed with liver cancer would be diagnosed definitely, which could identify CHC patients with 12.40% sensitivity, 98.04% specificity, and a positive predictive value of 33.70%.

Conclusions: Herein, the study established a risk prediction model which incorporates the clinical risk predictors and CT/MRI-presented PVTT status that could be adopted to facilitate the diagnosis of CHC patients preoperatively.

Development of a risk prediction model for lung cancer: The Japan Public Health Center-based Prospective Study

Although the impact of tobacco consumption on the occurrence of lung cancer is well‐established, risk estimation could be improved by risk prediction models that consider various smoking habits, such as quantity, duration, and time since quitting. We constructed a risk prediction model using a population of 59 161 individuals from the Japan Public Health Center (JPHC) Study Cohort II. A parametric survival model was used to assess the impact of age, gender, and smoking‐related factors (cumulative smoking intensity measured in pack‐years, age at initiation, and time since cessation). Ten‐year cumulative probability of lung cancer occurrence estimates were calculated with consideration of the competing risk of death from other causes. Finally, the model was externally validated using 47 501 individuals from JPHC Study Cohort I. A total of 1210 cases of lung cancer occurred during 986 408 person‐years of follow‐up. We found a dose‐dependent effect of tobacco consumption with hazard ratios for current smokers ranging from 3.78 (2.00‐7.16) for cumulative consumption ≤15 pack‐years to 15.80 (9.67‐25.79) for >75 pack‐years. Risk decreased with time since cessation. Ten‐year cumulative probability of lung cancer occurrence estimates ranged from 0.04% to 11.14% in men and 0.07% to 6.55% in women. The model showed good predictive performance regarding discrimination (cross‐validated c‐index = 0.793) and calibration (cross‐validated χ² = 6.60; P‐value = .58). The model still showed good discrimination in the external validation population (c‐index = 0.772). In conclusion, we developed a prediction model to estimate the probability of developing lung cancer based on age, gender, and tobacco consumption. This model appears useful in encouraging high‐risk individuals to quit smoking and undergo increased surveillance.

Clinical utility of the revised cardiac risk index in older Chinese patients with known coronary artery disease

Objectives

The revised Cardiac Risk Index (RCRI) is the most widely used risk prediction tool for postoperative cardiac adverse events. We aim to explore the predictive ability of the RCRI in older Chinese patients with coronary artery disease (CAD) undergoing noncardiac surgery, which has not been previously evaluated.

Methods

We performed a multicenter, prospective study. We enrolled a total of 1,202 patients, aged >60 years, with a history of CAD who underwent noncardiac surgery. Perioperative data were extracted from an electronic database. The primary end point was defined as an occurrence of a postoperative major cardiac event (PoMCE) within 30 days. Logistic regression analysis was performed to evaluate the performance of the RCRI. A modified RCRI was created and compared with the original RCRI with regard to its ability to predict postoperative cardiac events.

Results

Of the enrolled patients, 4.3% experienced PoMCE. Most components of the RCRI were not predictive of postoperative cardiac events with the exception of insulin-dependent diabetes mellitus (odds ratio =2.38, 95% CI: 1.11-5.11; P=0.03). The RCRI performed no better than chance (area under the curve =0.53; 95% CI: 0.45-0.61) in identifying patients' cardiac risk. The modified score had a higher discriminatory ability toward PoMCE (c index, 0.69 versus 0.53; P<0.01).

Conclusion

The original RCRI shows poor predictive ability in Chinese patients with CAD undergoing noncardiac surgery.

Validation of PAGE-B model in Asian chronic hepatitis B patients receiving entecavir or tenofovir

BACKGROUND & AIMS

A new hepatocellular carcinoma risk prediction model, PAGE-B, which includes age, gender and platelet count as constituent variables, has recently been proposed in Caucasian chronic hepatitis B patients. We validated PAGE-B model and compared its accuracy with that of conventional risk prediction models in Asian chronic hepatitis B patients.

METHODS

Chronic hepatitis B patients treated with entecavir or tenofovir were consecutively recruited. The performance of PAGE-B and three conventional risk prediction models (CU-HCC, GAG-HCC and REACH-B) were analysed.

RESULTS

A total of 1092 chronic hepatitis B patients (668 men, 61.2%) were selected between August 2006 and January 2015. The mean age was 48 years. During the follow-up period (median, 43.6 months), 36 (3.3%) patients developed hepatocellular carcinoma. Older age (hazard ratio [HR]=1.077), male gender (HR=3.676) and lower platelet count (HR=0.984) were independent predictors of hepatocellular carcinoma development. The PAGE-B showed similar area under receiver operating characteristic curves (AUROCs) to GAG-HCC and CU-HCC at 3 years (0.777 vs 0.793 and 0.743, respectively; all P>.05) and 5 years (0.799 vs 0.803 and 0.744, respectively; all P>.05), whereas the AUROCs of PAGE-B were significantly higher than those of the REACH-B (0.602 at 3 years and 0.572 at 5 years, P<.05).

CONCLUSIONS

Our study demonstrated that PAGE-B is applicable to Asian chronic hepatitis B patients receiving ETV or TDF therapy. The PAGE-B showed similar predictive performance to GAG-HCC and CU-HCC.

Development of a risk prediction model for Barrett's esophagus in an Australian population

Esophageal adenocarcinoma has poor 5-year survival rates. Increased survival might be achieved with earlier treatment, but requires earlier identification of the precursor, Barrett's esophagus. Population screening is not cost effective, this may be improved by targeted screening directed at individuals more likely to have Barrett's esophagus. To develop a risk prediction tool for Barrett's esophagus, this study compared individuals with Barrett's esophagus against population controls. Participants completed a questionnaire comprising 35 questions addressing medical history, symptom history, lifestyle factors, anthropomorphic measures, and demographic details. Statistical analysis addressed differences between cases and controls, and entailed initial variable selection, checking of model assumptions, and establishing calibration and discrimination. The area under the curve (AUC) was used to assess overall accuracy. One hundred and twenty individuals with Barrett's esophagus and 235 population controls completed the questionnaire. Significant differences were identified for age, gender, reflux history, family reflux history, history of hypertension, alcoholic drinks per week, and body mass index. These were used to develop a risk prediction model. The AUC was 0.82 (95% CI 0.78-0.87). Good calibration between predicted and observed risk was noted (Hosmer-Lemeshow test P = 0.67). At the point minimizing false positives and false negatives, the model achieved a sensitivity of 84.96% and a specificity of 66%. A well-calibrated risk prediction model with good discrimination has been developed to identify patients with Barrett's esophagus. The model needs to be externally validated before consideration for clinical practice.

Application of Concordance Probability Estimate to Predict Conversion from Mild Cognitive Impairment to Alzheimer's Disease

Subjects with mild cognitive impairment (MCI) have a substantially increased risk of developing dementia due to Alzheimer's disease (AD). Identifying MCI subjects who have high progression risk to AD is important in clinical management. Existing risk prediction models of AD among MCI subjects generally use either the AUC or Harrell's C-statistic to evaluate predictive accuracy. AUC is aimed at binary outcome and Harrell's C-statistic depends on the unknown censoring distribution. Gönen & Heller's K-index, also known as concordance probability estimate (CPE), is another measure of overall predictive accuracy for Cox proportional hazards (PH) models, which does not depend on censoring distribution. As a comprehensive example, using Alzheimer's Disease Neuroimaging Initiative (ADNI) dataset, we built a Cox PH model to predict the conversion from MCI to AD where the prognostic accuracy was evaluated using K-index.

A risk prediction model for post-stroke depression in Chinese stroke survivors based on clinical and socio-psychological features

Background

Post-stroke depression (PSD) is a frequent complication that worsens rehabilitation outcomes and patient quality of life. This study developed a risk prediction model for PSD based on patient clinical and socio-psychology features for the early detection of high risk PSD patients.

Results

Risk predictors included a history of brain cerebral infarction (odds ratio [OR], 3.84; 95% confidence interval [CI], 2.22-6.70; P < 0.0001) and four socio-psychological factors including Eysenck Personality Questionnaire with Neuroticism/Stability (OR, 1.18; 95% CI, 1.12-1.20; P < 0.0001), life event scale (OR, 0.99; 95% CI, 0.98-0.99; P = 0.0007), 20 items Toronto Alexithymia Scale (OR, 1.06; 95% CI, 1.02-1.10; P = 0.002) and Social Support Rating Scale (OR, 0.91; 95% CI, 0.87-0.90; P < 0.001) in the logistic model. In addition, 11 rules were generated in the tree model. The areas under the curve of the ROC and the accuracy for the tree model were 0.85 and 0.86, respectively.

Methods

This study recruited 562 stroke patients in China who were assessed for demographic data, medical history, vascular risk factors, functional status post-stroke, and socio-psychological factors. Multivariate backward logistic regression was used to extract risk factors for depression in 1-month after stroke. We converted the logistic model to a visible tree model using the decision tree method. Receiver operating characteristic (ROC) was used to evaluate the performance of the model.

Conclusion

This study provided an effective risk model for PSD and indicated that the socio-psychological factors were important risk factors of PSD.

Systematic Review of Health Economic Impact Evaluations of Risk Prediction Models: Stop Developing, Start Evaluating

BACKGROUND

Although health economic evaluations (HEEs) are increasingly common for therapeutic interventions, they appear to be rare for the use of risk prediction models (PMs).

OBJECTIVES

To evaluate the current state of HEEs of PMs by performing a comprehensive systematic review.

METHODS

Four databases were searched for HEEs of PM-based strategies. Two reviewers independently selected eligible articles. A checklist was compiled to score items focusing on general characteristics of HEEs of PMs, model characteristics and quality of HEEs, evidence on PMs typically used in the HEEs, and the specific challenges in performing HEEs of PMs.

RESULTS

After screening 791 abstracts, 171 full texts, and reference checking, 40 eligible HEEs evaluating 60 PMs were identified. In these HEEs, PM strategies were compared with current practice (n = 32; 80%), to other stratification methods for patient management (n = 19; 48%), to an extended PM (n = 9; 23%), or to alternative PMs (n = 5; 13%). The PMs guided decisions on treatment (n = 42; 70%), further testing (n = 18; 30%), or treatment prioritization (n = 4; 7%). For 36 (60%) PMs, only a single decision threshold was evaluated. Costs of risk prediction were ignored for 28 (46%) PMs. Uncertainty in outcomes was assessed using probabilistic sensitivity analyses in 22 (55%) HEEs.

CONCLUSIONS

Despite the huge number of PMs in the medical literature, HEE of PMs remains rare. In addition, we observed great variety in their quality and methodology, which may complicate interpretation of HEE results and implementation of PMs in practice. Guidance on HEE of PMs could encourage and standardize their application and enhance methodological quality, thereby improving adequate use of PM strategies.

A Comprehensive Model of Colorectal Cancer by Risk Factor Status and Subsite Using Data From the Nurses' Health Study

We expanded and updated our colon cancer risk model to evaluate colorectal cancer (CRC) and whether subsite-specific risk models are warranted. Using data from 1980-2010 for 90,286 women enrolled in the Nurses' Health Study, we performed competing-risks regression and tests for subsite heterogeneity (proximal colon: n = 821; distal colon: n = 521; rectum: n = 376). Risk factors for CRC were consistent with those in our colon cancer model. Processed meat consumption was associated with a higher risk of distal (hazard ratio (HR) = 1.45; P = 0.02) but not proximal (HR = 0.95; P = 0.72) colon cancer. Smoking was associated with both colon (HR = 1.21) and rectal (HR = 1.27) cancer and was more strongly associated with proximal (HR = 1.31) than with distal (HR = 1.04) colon cancer (P = 0.029). We observed a significant trend of cancer risk for smoking in subsites from the cecum (HR = 1.41) to the proximal colon (excluding the cecum; HR = 1.27) to the distal colon (HR = 1.04; P for trend = 0.040). The C statistics for colorectal (C = 0.607), colon (C = 0.603), and rectal (C = 0.639) cancer were similar, although C was slightly higher for rectal cancer. Despite evidence for site-specific differences for several risk factors, overall our findings support the application of risk prediction models for colon cancer to CRC.

A prospective and retrospective analysis of smoking behavior changes in ever smokers with high risk for lung cancer from New Mexico and Pennsylvania

Cigarette smoking is the leading preventable cause of death worldwide. The aim of this study is to conduct a prospective and retrospective analysis of smoking behavior changes in the Lovelace Smokers Cohort (LSC) and the Pittsburgh Lung Screening Study cohort (PLuSS). Area under the curve (AUC) for risk models predicting relapse based on demographic, smoking, and relevant clinical variables was 0.93 and 0.79 in LSC and PLuSS, respectively. The models for making a quit attempt had limited prediction ability in both cohorts (AUC≤0.62). We identified an ethnic disparity in adverse smoking behavior change that Hispanic smokers were less likely to make a quit attempt and were more likely to relapse after a quit attempt compared to non-Hispanic Whites. SNPs at 15q25 and 11p14 loci were associated with risk for smoking relapse in the LSC. Rs6495308 at 15q25 has a large difference in minor allele frequency between non-Hispanic Whites and Hispanics (0.46 versus 0.23, P<0.0001) and was associated with risk for ever relapse at same magnitude between the two ethnic groups (OR=1.36, 95% CI=1.10 to 1.67 versus 1.59, 95% CI=1.00 to 2.53, P=0.81). In summary, the risk prediction model established in LSC and PLuSS provided an excellent to outstanding distinguishing for abstainers who will or will not relapse. The ethnic disparity in adverse smoking behavior between Hispanics and non-Hispanic Whites may be at least partially explained by the sequence variants at 15q25 locus that contains multiple nicotine acetylcholine receptors.

Prognostic indices for early mortality in ischaemic stroke - meta-analysis

OBJECTIVES

Several models have been developed to predict mortality in ischaemic stroke. We aimed to evaluate systematically the performance of published stroke prognostic scores.

METHODS

We searched MEDLINE and EMBASE in February 2014 for prognostic models (published between 2003 and 2014) used in predicting early mortality (<6 months) after ischaemic stroke. We evaluated discriminant ability of the tools through meta-analysis of the area under the curve receiver operating characteristic curve (AUROC) or c-statistic. We evaluated the following components of study validity: collection of prognostic variables, neuroimaging, treatment pathways and missing data.

RESULTS

We identified 18 articles (involving 163 240 patients) reporting on the performance of prognostic models for mortality in ischaemic stroke, with 15 articles providing AUC for meta-analysis. Most studies were either retrospective, or post hoc analyses of prospectively collected data; all but three reported validation data. The iSCORE had the largest number of validation cohorts (five) within our systematic review and showed good performance in four different countries, pooled AUC 0.84 (95% CI 0.82-0.87). We identified other potentially useful prognostic tools that have yet to be as extensively validated as iSCORE - these include SOAR (2 studies, pooled AUC 0.79, 95% CI 0.78-0.80), GWTG (2 studies, pooled AUC 0.72, 95% CI 0.72-0.72) and PLAN (1 study, pooled AUC 0.85, 95% CI 0.84-0.87).

CONCLUSIONS

Our meta-analysis has identified and summarized the performance of several prognostic scores with modest to good predictive accuracy for early mortality in ischaemic stroke, with the iSCORE having the broadest evidence base.