Assessing the prediction of type 2 diabetes risk using polygenic and clinical risk scores in South Asian study populations

Background

Genome-wide polygenic risk scores (PRS) have shown high specificity and sensitivity in predicting type 2 diabetes (T2D) risk in Europeans. However, the PRS-driven information and its clinical significance in non-Europeans are underrepresented. We examined the predictive efficacy and transferability of PRS models using variant information derived from genome-wide studies of Asian Indians (AIs) (PRSAI) and Europeans (PRSEU) using 13,974 AI individuals.

Methods

Weighted PRS models were constructed and analyzed on 4602 individuals from the Asian Indian Diabetes Heart Study/Sikh Diabetes Study (AIDHS/SDS) as discovery/training and test/validation datasets. The results were further replicated in 9372 South Asian individuals from UK Biobank (UKBB). We also assessed the performance of each PRS model by combining data of the clinical risk score (CRS).

Results

Both genetic models (PRSAI and PRSEU) successfully predicted the T2D risk. However, the PRSAI revealed 13.2% odds ratio (OR) 1.80 [95% confidence interval (CI) 1.63-1.97; p = 1.6 × 10-152] and 12.2% OR 1.38 (95% CI 1.30-1.46; p = 7.1 × 10-237) superior performance in AIDHS/SDS and UKBB validation sets, respectively. Comparing individuals of extreme PRS (ninth decile) with the average PRS (fifth decile), PRSAI showed about two-fold OR 20.73 (95% CI 10.27-41.83; p = 2.7 × 10-17) and 1.4-fold OR 3.19 (95% CI 2.51-4.06; p = 4.8 × 10-21) higher predictability to identify subgroups with higher genetic risk than the PRSEU. Combining PRS and CRS improved the area under the curve from 0.74 to 0.79 in PRSAI and 0.72 to 0.75 in PRSEU.

Conclusion

Our data suggest the need for extending genetic and clinical studies in varied ethnic groups to exploit the full clinical potential of PRS as a risk prediction tool in diverse study populations.

Predicting multi-vascular diseases in patients with coronary artery disease

Background: Because of its systemic nature, the occurrence of atherosclerosis in the coronary arteries can also indicate a risk for other vascular diseases.  However, screening program targeted for all patients with coronary artery disease (CAD) is highly ineffective and no studies have assessed the risk factors for developing multi-vascular diseases in general. This study constructed a predictive model and scoring system to enable targeted screening for multi-vascular diseases in CAD patients. Methods: This cross-sectional study includes patients with CAD, as diagnosed during coronary angiography or percutaneous coronary intervention from March 2021 to December 2021. Coronary artery stenosis (CAS) and abdominal aortic aneurysm (AAA) were diagnosed using Doppler ultrasound while peripheral artery disease (PAD) was diagnosed based on ABI score. Multivariate logistic regression was conducted to construct the predictive model and risk scores. Validation was conducted using ROC analysis and Hosmer-Lemeshow test. Results: Multivariate analysis showed that ages of >60 years (OR [95% CI] = 1.579 [1.153-2.164]), diabetes mellitus (OR = 1.412 [1.036-1.924]), cerebrovascular disease (OR = 3.656 [2.326-5.747]), and CAD3VD (OR = 1.960 [1.250-3.073]) increased the odds for multi-vascular disease. The model demonstrated good predictive capability (AUC = 0.659) and was well-calibrated (Hosmer-Lemeshow p = 0.379). Targeted screening for high-risk patients reduced the number needed to screen (NNS) from 6 in the general population to 3 and has a high specificity of 96.5% Conclusions: Targeted screening using clinical risk scores was able to decrease NNS with good predictive capability and high specificity.

Predictive Factors for the Survival Outcomes of Preoperative Chemotherapy in Patients with Resectable and Borderline Resectable Colorectal Cancer with Liver Metastasis

BACKGROUND

Preoperative chemotherapy increases resectability in borderline resectable colorectal liver metastasis (CRLM) patients who undergo curative liver surgery. Most clinical risk scores and other predictive factors for survival have been extensively studied in patients who undergo upfront liver surgery. However, predictive factors of CRLM patients who received preoperative chemotherapy remains controversial.

METHODS

CRLM patients who received preoperative systemic therapy followed by curative liver surgery at our institution between 1/2012 and 12/2018 were included. This study aimed to investigate factors that predicted the outcomes of preoperative systemic treatment, optimal dose/duration, and toxicity in patients with CRLM.

OUTCOMES

Ninety-eight patients were eligible for analysis. Most patients received oxaliplatin-based chemotherapy (72.7%), while 15.9% received both oxaliplatin and irinotecan. Biologic agents were administered in 48.9% of patients. Overall, chemotherapy-induced liver injury was observed in 38.5%. The median disease-free survival (DFS) and overall survival (OS) were 8.7 months and 3.6 years, respectively. Baseline, pre-surgery, and increased Fong scores after preoperative chemotherapy were significantly associated with DFS and OS. In multivariate analysis, a high Fong score at baseline (p=0.018) was significantly associated with shorter DFS, whereas male sex (p=0.040) and liver surgery (p=0.044) were related to longer OS.

CONCLUSION

In our study, Fong clinical risk scores, female sex, and liver surgery as a part of liver-directed therapy were independent prognostic factors for survival in CRLM patients who received preoperative chemotherapy. These clinical factors should be considered as an option to guide physicians' decisions in selecting patients with CRLM who may benefit most from curative liver-directed therapy.

Clinical Risk Score for Prediction of Urgency in Carotid Cavernous Sinus Fistulas

Purpose

To develop a clinical risk score for the prediction of urgency in patients with carotid cavernous sinus fistulas (CCFs) and test for the discriminative ability of the diagnostic prediction.

Methods

The medical charts of 60 patients with CCFs were retrospectively reviewed. The clinical characteristics of direct and dural CCFs were analyzed by logistic regression. The clinical risk score was developed from the coefficient in the multivariable regression model and used to predict direct CCFs which were more urgent than the dural type. The score prediction was reported as an area under the receiver operating characteristic (AuROC) curve and 95% confidence interval (95% CI).

Results

In a univariable analysis, the clinical characteristics which increased the risk of direct CCFs were age, gender, trauma, underlying diseases, visual acuity (VA) at presentation, bruit, chemosis, and dilated retinal vessels. However, in multivariable analysis, the significant predictors were limited to age, trauma, bruit, underlying diseases and logMAR VA. Regression coefficient of each predictor was converted to a risk score and summation of scores from these predictors for each patient was calculated. The total risk score predicted the urgent direct CCFs correctly with AuROC of 97.77% (95% CI; 93.57, 100).

Conclusion

The clinical risk score for the prediction of urgent direct CCFs has been developed and used in the patients with CCFs in our setting. The discriminative ability of the score prediction is high. This simple clinical risk score may help clinicians suspect direct CCFs and urgently refer the patients to have prompt angiography and treatment.

Prediction of Deep Myometrial Infiltration, Clinical Risk Category, Histological Type, and Lymphovascular Space Invasion in Women with Endometrial Cancer Based on Clinical and T2-Weighted MRI Radiomic Features

PURPOSE

To predict deep myometrial infiltration (DMI), clinical risk category, histological type, and lymphovascular space invasion (LVSI) in women with endometrial cancer using machine learning classification methods based on clinical and image signatures from T2-weighted MR images.

METHODS

A training dataset containing 413 patients and an independent testing dataset consisting of 82 cases were employed in this retrospective study. Manual segmentation of the whole tumor volume on sagittal T2-weighted MRI was performed. Clinical and radiomic features were extracted to predict: (i) DMI of endometrial cancer patients, (ii) endometrial cancer clinical high-risk level, (iii) histological subtype of tumor, and (iv) presence of LVSI. A classification model with different automatically selected hyperparameter values was created. The area under the curve (AUC) of the receiver operating characteristic (ROC) curve, F1 score, average recall, and average precision were calculated to evaluate different models.

RESULTS

Based on the independent external testing dataset, the AUCs for DMI, high-risk endometrial cancer, endometrial histological type, and LVSI classification were 0.79, 0.82, 0.91, and 0.85, respectively. The corresponding 95% confidence intervals (CI) of the AUCs were [0.69, 0.89], [0.75, 0.91], [0.83, 0.97], and [0.77, 0.93], respectively.

CONCLUSION

It is possible to classify endometrial cancer DMI, risk, histology type, and LVSI using different machine learning methods.

Integrating polygenic and clinical risks to improve stroke risk stratification in prospective Chinese cohorts

The utility of the polygenic risk score (PRS) to identify individuals at higher risk of stroke beyond clinical risk remains unclear, and we clarified this using Chinese population-based prospective cohorts. Cox proportional hazards models were used to estimate the 10-year risk, and Fine and Gray's models were used for hazard ratios (HRs), their 95% confidence intervals (CIs), and the lifetime risk according to PRS and clinical risk categories. A total of 41,006 individuals aged 30-75 years with a mean follow-up of 9.0 years were included. Comparing the top versus bottom 5% of the PRS, the HR was 3.01 (95%CI 2.03-4.45) in the total population, and similar findings were observed within clinical risk strata. Marked gradients in the 10-year and lifetime risk across PRS categories were also found within clinical risk categories. Notably, among individuals with intermediate clinical risk, the 10-year risk for those in the top 5% of the PRS (7.3%, 95%CI 7.1%-7.5%) reached the threshold of high clinical risk (⩾7.0%) for initiating preventive treatment, and this effect of the PRS on refining risk stratification was evident for ischemic stroke. Even among those in the top 10% and 20% of the PRS, the 10-year risk would also exceed this level when aged ⩾50 and ⩾60 years, respectively. Overall, the combination of the PRS with the clinical risk score improved the risk stratification within clinical risk strata and distinguished actual high-risk individuals with intermediate clinical risk.

Development of a clinical risk score to predict death in patients with COVID-19

OBJECTIVE

To build a clinical risk score to aid risk stratification among hospitalised COVID-19 patients.

METHODS

The score was built using data of 417 consecutive COVID-19 in patients from Kuwait. Risk factors for COVID-19 mortality were identified by multivariate logistic regressions and assigned weighted points proportional to their beta coefficient values. A final score was obtained for each patient and tested against death to calculate an Receiver-operating characteristic curve. Youden's index was used to determine the cut-off value for death prediction risk. The score was internally validated using another COVID-19 Kuwaiti-patient cohort of 923 patients. External validation was carried out using 178 patients from the Italian CoViDiab cohort.

RESULTS

Deceased COVID-19 patients more likely showed glucose levels of 7.0-11.1 mmol/L (34.4%, p < 0.0001) or >11.1 mmol/L (44.3%, p < 0.0001), and comorbidities such as diabetes and hypertension compared to those who survived (39.3% vs. 20.4% [p = 0.0027] and 45.9% vs. 26.6% [p = 0.0036], respectively). The risk factors for in-hospital mortality in the final model were gender, nationality, asthma, and glucose categories (<5.0, 5.5-6.9, 7.0-11.1, or 11.1 > mmol/L). A score of ≥5.5 points predicted death with 75% sensitivity and 86.3% specificity (area under the curve (AUC) 0.901). Internal validation resulted in an AUC of 0.826, and external validation showed an AUC of 0.687.

CONCLUSION

This clinical risk score was built with easy-to-collect data and had good probability of predicting in-hospital death among COVID-19 patients.

Developing the Total Health Profile, a Generalizable Unified Set of Multimorbidity Risk Scores Derived From Machine Learning for Broad Patient Populations: Retrospective Cohort Study

BACKGROUND

Multimorbidity clinical risk scores allow clinicians to quickly assess their patients' health for decision making, often for recommendation to care management programs. However, these scores are limited by several issues: existing multimorbidity scores (1) are generally limited to one data group (eg, diagnoses, labs) and may be missing vital information, (2) are usually limited to specific demographic groups (eg, age), and (3) do not formally provide any granularity in the form of more nuanced multimorbidity risk scores to direct clinician attention.

OBJECTIVE

Using diagnosis, lab, prescription, procedure, and demographic data from electronic health records (EHRs), we developed a physiologically diverse and generalizable set of multimorbidity risk scores.

METHODS

Using EHR data from a nationwide cohort of patients, we developed the total health profile, a set of six integrated risk scores reflecting five distinct organ systems and overall health. We selected the occurrence of an inpatient hospital visitation over a 2-year follow-up window, attributable to specific organ systems, as our risk endpoint. Using a physician-curated set of features, we trained six machine learning models on 794,294 patients to predict the calibrated probability of the aforementioned endpoint, producing risk scores for heart, lung, neuro, kidney, and digestive functions and a sixth score for combined risk. We evaluated the scores using a held-out test cohort of 198,574 patients.

RESULTS

Study patients closely matched national census averages, with a median age of 41 years, a median income of $66,829, and racial averages by zip code of 73.8% White, 5.9% Asian, and 11.9% African American. All models were well calibrated and demonstrated strong performance with areas under the receiver operating curve (AUROCs) of 0.83 for the total health score (THS), 0.89 for heart, 0.86 for lung, 0.84 for neuro, 0.90 for kidney, and 0.83 for digestive functions. There was consistent performance of this scoring system across sexes, diverse patient ages, and zip code income levels. Each model learned to generate predictions by focusing on appropriate clinically relevant patient features, such as heart-related hospitalizations and chronic hypertension diagnosis for the heart model. The THS outperformed the other commonly used multimorbidity scoring systems, specifically the Charlson Comorbidity Index (CCI) and the Elixhauser Comorbidity Index (ECI) overall (AUROCs: THS=0.823, CCI=0.735, ECI=0.649) as well as for every age, sex, and income bracket. Performance improvements were most pronounced for middle-aged and lower-income subgroups. Ablation tests using only diagnosis, prescription, social determinants of health, and lab feature groups, while retaining procedure-related features, showed that the combination of feature groups has the best predictive performance, though only marginally better than the diagnosis-only model on at-risk groups.

CONCLUSIONS

Massive retrospective EHR data sets have made it possible to use machine learning to build practical multimorbidity risk scores that are highly predictive, personalizable, intuitive to explain, and generalizable across diverse patient populations.

Histological tumor response to neoadjuvant chemotherapy correlates to Immunoscore in colorectal cancer liver metastases patients

BACKGROUND

This study investigated the correlation between tumor regression grade (TRG) score and Immunoscore, and prognostic values of TRG score and a risk score in colorectal cancer liver metastases (CRLMs) patients after neoadjuvant chemotherapy.

METHODS

Patients undergoing neoadjuvant chemotherapy from 2014 to 2019 were selected. TRG score and Immunoscore were evaluated in 200 CRLMs. A risk score combining TRG score, Immunoscore, and clinical risk score (CRS) was defined and divided patients into the low-, medium-, and high-risk groups. Differences in relapse-free survival (RFS) and overall survival (OS) between groups were determined.

RESULTS

The densities of CD3⁺ and CD8⁺ immune cells were higher in TRG1-3 group than in TRG4-5 group, and the ratio of high Immunoscores was higher in TRG1-3 group than in TRG4-5 group (60.0% vs. 15.8%, p < 0.001). Patients in TRG1-3 group had significantly longer RFS and OS than those in TRG4-5 group. The low-risk group shows a significantly higher 2-year RFS and 5-year OS rate than the medium- and high-risk group (RFS: 59.9%, 36.2%, and 6.4%, p < 0.001; OS: 82.0%, 41.0%, and 16.9%, p < 0.001).

CONCLUSION

TRG score may be proposed to evaluate the prognosis of neoadjuvant chemotherapy and may be used for predicting the postoperative survival of CRLMs.

Significance and clinical impact of routinely tested urinary ethyl glucuronide after liver transplantation - development of a risk score

BACKGROUND & AIMS

Alcohol abuse after liver transplantation can seriously impact graft and patient survival. However, to date, there is no defined standard procedure to identify patients consuming alcohol after liver transplantation. The aim of this study was to analyze the diagnostic value and clinical impact of routinely measured urinary ethyl glucuronide (uEtG) - a metabolite of ethanol - in patients after liver transplantation.

METHODS

Data of 362 consecutive patients after liver transplantation who visited the University Hospital of Tuebingen for outpatient follow-up were analyzed.

RESULTS

48 patients (13%) displayed positive uEtG results. The uEtG positive group contained significantly more patients with pre transplant alcoholic liver disease. However, two thirds of the uEtG positive patients had no history of pre transplant alcoholic liver disease. Several clinical parameters were significantly associated with positive uEtG. In order to enable a more cost-effective application of uEtG in the future, a clinical risk score was developed (specificity 0.95).

CONCLUSIONS

Routine testing for uEtG reveals a considerable percentage of patients practicing alcohol intake after liver transplantation. Application of our proposed risk score could help focusing uEtG testing on patients at risk.

Performance of two prognostic scores that incorporate genetic information to predict long-term outcomes following resection of colorectal cancer liver metastases: An external validation of the MD Anderson and JHH-MSK scores

INTRODUCTION

Two novel clinical risk scores (CRS) that incorporate KRAS mutation status were developed: modified CRS (mCRS) and GAME score. However, they have not been tested in large national and international cohorts. The aim of this study was to validate the prognostic discrimination utility and determine the clinical usefulness of the two novel CRS.

METHODS

Patients undergoing hepatectomy for CRLM (2000-2018) in 10 centers were included. The discriminatory abilities of mCRS, GAME, and Fong CRS were evaluated using Harrell's C-index and Akaike's Information Criterion.

RESULTS

In the entire cohort, the C-index of the GAME score (0.61) was significantly higher than those of Fong score (0.57) and mCRS (0.54), while the C-Index of mCRS was significantly lower than that of Fong score. When we compared the models in the various geographical regions, the C-index of GAME score was significantly higher than that of mCRS in North America, Europe, and South America. The AIC of Fong score, mCRS, and GAME score were 14 405, 14 447, and 14 319, respectively.

CONCLUSION

In conclusion, using the largest and most heterogenous population of CRLM patients with known KRAS status, this independent, external validation demonstrated that the GAME score outperforms both the traditional Fong score and mCRS.

A new clinical risk score for predicting the prevalence of low-voltage areas in patients undergoing atrial fibrillation ablation

INTRODUCTION

Although the presence of left atrial low-voltage areas (LVAs) is strongly associated with the recurrence of atrial fibrillation (AF) after ablation, few methods are available to classify the prevalence of LVAs. The purpose of this study was to establish a risk score for predicting the prevalence of LVAs in patients undergoing ablation for AF.

METHODS

We enrolled 1004 consecutive patients who underwent initial ablation for AF (age, 68 ± 10 years old; female, 346 (34%); persistent AF, 513 (51%)). LVAs were deemed present when the voltage map after pulmonary vein isolation demonstrated low-voltage areas with a peak-to-peak bipolar voltage of <0.5 mV covering ≥5 cm² of the left atrium.

RESULTS

LVAs were present in 206 (21%) patients. The SPEED score was obtained as the total number of independent predictors as identified on multivariate analysis, namely female sex (odds ratio [OR], 3.4 [95% confidence interval {CI} 2.2-5.2], p < .01), persistent AF (OR, 1.8 [95% CI, 1.1-3.0], p = .02), age ≥ 70 years (OR, 2.3 [95% CI, 1.5-3.4], p < .01), elevated brain natriuretic peptide ≥100 pg/ml or N-terminal probrain natriuretic peptide ≥400 pg/ml (OR, 1.7 [95% CI, 1.02-2.8], p = .04), and diabetes mellitus (OR, 1.8 [95% CI, 1.1-2.8], p = .02). LVAs were more frequent in patients with a higher SPEED score, and prevalence increased with each additional SPEED score point (OR, 2.4 [95% CI, 2.0-2.8], p < .01).

CONCLUSION

The SPEED score accurately predicts the prevalence of LVAs in patients undergoing ablation for AF.

Population-Based Screening or Targeted Screening Based on Initial Clinical Risk Assessment for Atrial Fibrillation: A Report from the Huawei Heart Study

Background: A general-population approach has been advocated to improve the screening of patients with atrial fibrillation (AF). A more pragmatic alternative may be targeted screening of patients at high risk of developing AF. We assess the value of a simple clinical risk score, C₂HEST (C2, coronary artery disease/chronic obstructive pulmonary disease; COPD (1 point each); H, hypertension; E, elderly (age ≥75, doubled); S, systolic heart failure; HF (doubled); T, hyperthyroidism)); to facilitate population screening and detection of incident AF in the general population, in a prespecified ancillary analysis of the Huawei Heart Study. Methods: The Huawei Heart Study investigated general population screening for AF, identified using photoplethysmography (PPG)-based HUAWEI smart devices. We compared the value of a general population approach to a target screening approach between 26 October 2018 and 20 November 2019. Results: There were 644,124 individuals (mean age ± standard deviation, SD 34 ± 11; female 15.9%) who monitored their pulse rhythm using smart devices, among which 209,274 individuals (mean age 34 years, SD11; 10.6% female) completed the questionnaire on cardiovascular risk factors, with 739 detecting AF. Of these, 31.4% (n = 65,810) subjects reported palpitations. The median (interquartile range, IQR) duration to first detected AF was 11 (1–46), 6 (1–49), and 4 (1–24) in the population with low, intermediate, and high C₂HEST score category, respectively (p = 0.03). Detected AF events rates increased with increasing C₂HEST score points, stratified by age (p for trend, p < 0.001). Hazard ratios of the components of the C₂HEST score for detected AF were between 1.31 and 2.75. A combination of symptomatic palpitations and C₂HEST score increased prediction of AF detection, compared to using C₂HEST score alone (c-indexes 0.72 vs. 0.76, Delong test, p < 0.001). Conclusions: The C₂HEST score, especially when combined with symptoms, could facilitate a targeted population-based screening and preventive strategy for AF.

A boosting inspired personalized threshold method for sepsis screening

Sepsis is one of the biggest risks to patient safety, with a natural mortality rate between 25% and 50%. It is difficult to diagnose, and no validated standard for diagnosis currently exists. A commonly used scoring criteria is the quick sequential organ failure assessment (qSOFA). It demonstrates very low specificity in ICU populations, however. We develop a method to personalize thresholds in qSOFA that incorporates easily to measure patient baseline characteristics. We compare the personalized threshold method to qSOFA, five previously published methods that obtain an optimal constant threshold for a single biomarker, and to the machine learning algorithms based on logistic regression and AdaBoosting using patient data in the MIMIC-III database. The personalized threshold method achieves higher accuracy than qSOFA and the five published methods and has comparable performance to machine learning methods. Personalized thresholds, however, are much easier to adopt in real-life monitoring than machine learning methods as they are computed once for a patient and used in the same way as qSOFA, whereas the machine learning methods are hard to implement and interpret.

Different clinical risk scores for prediction of early mortality after liver resection for hepatocellular carcinoma: which is the best?

BACKGROUND

Prediction of early mortality after hepatectomies for hepatocellular carcinoma is essential to identify high-risk patients and to decrease the operative mortality rate. Several post-operative clinical risk scores were developed recently to predict mortality post-hepatectomy; however, which one is the best remains undefined. Therefore, the aim of this study was to evaluate the performance of the different post-operative clinical risk scores in predicting early mortality after hepatectomies.

METHODS

A total of 240 patients who underwent liver resection for hepatocellular carcinoma at our hospital between June 2011 and July 2016 were retrospectively reviewed. Post-operative clinical risk scores including 50-50 criteria, peak bilirubin >7 mg/dL, model for end-stage liver disease (MELD), risk assessment for early mortality and Hyder scores were evaluated for their performance in predicting early mortality after hepatic resection using the receiver operating characteristic (ROC) curve.

RESULTS

The 90-day mortality rate after hepatic resection was around 2.5%. The 50-50 criteria and peak bilirubin >7 mg/dL were weak predictors of early mortality with low sensitivity (area under the ROC curve: 0.65, 0.66, respectively), whereas, Hyder, risk assessment for early mortality, and post-operative MELD were good predictors of early mortality (area under the ROC curve: 0.89, 0.91 and 0.88, respectively). Moreover, MELD score on post-operative day 3 was an independent risk factor for 90-day mortality with an odds ratio of 1.4 (95% confidence interval 1.06-1.81, P = 0.02).

CONCLUSIONS

Post-operative clinical risk scores, especially MELD, were capable of predicting early mortality after liver resection and should be used to identify high-risk patients and provide them with more intensive medical care.

Predicting the risk of apparent treatment-resistant hypertension: a longitudinal, cohort study in an urban hypertension referral clinic

Apparent treatment-resistant hypertension (aTRH) is associated with higher prevalence of secondary hypertension, greater risk for adverse pressure-related clinical outcomes, and influences diagnostic and therapeutic decision-making. We previously showed that cross-sectional prevalence estimates of aTRH are lower than its true prevalence as patients with uncontrolled hypertension undergoing intensification/optimization of therapy will, over time, increasingly satisfy diagnostic criteria for aTRH. aTRH was assessed in an urban referral hypertension clinic using a 140/90 mm Hg goal blood pressure target in 745 patients with uncontrolled blood pressure, who were predominately African-American (86%) and female (65%). Analyses were stratified according to existing prescription of diuretic at initial visit. Risk for aTRH was estimated using logistic regression with patient characteristics at index visit as predictors. Among those prescribed diuretics, 84/363 developed aTRH; the risk score discriminated well (area under the receiver operating curve = 0.77, bootstrapped 95% CI [0.71, 0.81]). In patients not prescribed a diuretic, 44/382 developed aTRH, and the risk score showed a significantly better discriminative ability (area under the receiver operating curve = 0.82 [0.76, 0.87]; P < .001). In the diuretic and nondiuretic cohorts, 145/363 and 290/382 of patients had estimated risks for development of aTRH <15%. Of these low-risk patients, 139/145 and 278/290 did not develop aTRH (negative predictive value, diuretics - 0.94 [0.91, 0.98], no diuretics - 0.95 [0.93, 0.97]). We created a novel clinical score that discriminates well between those who will and will not develop aTRH, especially among those without existing diuretic prescriptions. Irrespective of baseline diuretic treatment status, a low-risk score had very high negative predictive value.

Development of a clinical risk score for pain and function following total knee arthroplasty: results from the TRIO study

Objectives

The aim was to develop and validate a simple clinical prediction model, based on easily collected preoperative information, to identify patients at high risk of pain and functional disability 6 months after total knee arthroplasty (TKA).

Methods

This was a multicentre cohort study of patients from nine centres across the UK, who were undergoing a primary TKA for OA. Information on sociodemographic, psychosocial, clinical and quality-of-life measures were collected at recruitment. The primary outcome measure for this analysis was the Oxford knee score (OKS), measured 6 months postoperatively by postal questionnaire. Multivariable logistic regression was used to develop the model. Model performance (discrimination and calibration) and internal validity were assessed, and a simple clinical risk score was developed.

Results

Seven hundred and twenty-one participants (mean age 68.3 years; 53% female) provided data for the present analysis, and 14% had a poor outcome at 6 months. Key predictors were poor clinical status, widespread body pain, high expectation of postoperative pain and lack of active coping. The developed model based on these variables demonstrated good discrimination. At the optimal cut-off, the final model had a sensitivity of 83%, specificity of 61% and positive likelihood ratio of 2.11. Excellent agreement was found between observed and predicted outcomes, and there was no evidence of overfitting in the model.

Conclusion

We have developed and validated a clinical prediction model that can be used to identify patients at high risk of a poor outcome after TKA. This clinical risk score may be an aid to shared decision-making between patient and clinician.

Validation of clinical prognostic scores for patients treated with curative-intent for recurrent colorectal liver metastases

BACKGROUND

Scoring systems were developed to stratify patients with colorectal liver metastases considered for liver resection into different risk groups. Such scores have never been evaluated in recurrent liver metastases. The aim of this study was to evaluate whether these scores are applicable to patients with recurrent colorectal liver metastases and treated with curative intent.

METHODS

We retrospectively analyzed data from 375 consecutive patients who underwent liver surgery for colorectal liver metastases between June 2010 and August 2015. Seventy-three patients developed liver-limited recurrence treated with curative intent. The predictive value of 6 scores (Fong, Sofocleous, Nagashima, Nordlinger, Konopke, and the Basingstoke index) was assessed in this set of patients.

RESULTS

Median follow-up was 36.2 months. Overall survival and progression-free survival were 33.6 and 5.6 months, respectively. When scores were applied for OS, none showed a significant stratification between patients, although Nagashima's score showed a significant difference in overall survival between patients from the low-risk group and those from the intermediate- and high-risk groups (40.8 vs 30.5 months, P = 0.039). For PFS, only Fong's score showed a statistically significant stratification (6.6 vs 4.7 months, P = 0.027).

CONCLUSION

Scoring systems are of limited-value in stratifying patients operated on for recurrent colorectal liver metastases.

Survival prediction in patients with resectable colorectal liver metastases: Clinical risk scores and tumor response to chemotherapy

Clinical risk scores and response to pre-operative chemotherapy are prognostic factors of colorectal liver metastases. The aim of the present study was to evaluate the effectiveness of combining these factors to predict patient survival and to select patients for curative therapy. The study included 189 patients who underwent hepatectomy following neo-adjuvant chemotherapy, for initially resectable colorectal liver metastases, between January 2005 and December 2015. Patients were stratified into four sub-groups: A1-2, low clinical risk scores with/without a response to pre-operative chemotherapy; and B1-2, high clinical risk scores with or without a response to pre-operative chemotherapy. Treatment and survival data were analysed. Survival was significantly longer in patients with low clinical risk scores and a response to pre-operative chemotherapy; these factors were confirmed as independent prognostic factors by multivariate analysis. Combining clinical risk score and chemotherapy response classification, patient survival was significantly longer for groups A1-2/B1 compared with for group B2, in which only 10.2% of patients were alive after 5 years. Of those with no response to first-line chemotherapy, survival was significantly longer in patients who responded to second-line chemotherapy. A combined clinical risk score and chemotherapy response classification may aid in identifying suitable candidates for potentially curative therapy.

Prediction of Hepatocellular Carcinoma Recurrence Beyond Milan Criteria After Resection: Validation of a Clinical Risk Score in an International Cohort

OBJECTIVE

This study aims to validate a previously reported recurrence clinical risk score (CRS).

SUMMARY OF BACKGROUND DATA

Salvage transplantation after hepatocellular carcinoma (HCC) resection is limited to patients who recur within Milan criteria (MC). Predicting recurrence patterns may guide treatment recommendations.

METHODS

An international, multicenter cohort of R0 resected HCC patients were categorized by MC status at presentation. CRS was calculated by assigning 1 point each for initial disease beyond MC, multinodularity, and microvascular invasion. Recurrence incidences were estimated using competing risks methodology, and conditional recurrence probabilities were estimated using the Bayes theorem.

RESULTS

From 1992 to 2015, 1023 patients were identified, of whom 613 (60%) recurred at a median follow-up of 50 months. CRS was well validated in that all 3 factors remained independent predictors of recurrence beyond MC (hazard ratio 1.5-2.1, all P < 0.001) and accurately stratified recurrence risk beyond MC, ranging from 19% (CRS 0) to 67% (CRS 3) at 5 years. Among patients with CRS 0, no other factors were significantly associated with recurrence beyond MC. The majority recurred within 2 years. After 2 years of recurrence-free survival, the cumulative risk of recurrence beyond MC within the next 5 years for all patients was 14%. This risk was 12% for patients with initial disease within MC and 17% for patients with initial disease beyond MC.

CONCLUSIONS

CRS accurately predicted HCC recurrence beyond MC in this international validation. Although the risk of recurrence beyond MC decreased over time, it never reached zero.

Can a novel clinical risk score improve pneumonia prediction in acute stroke care? A UK multicenter cohort study

Background

Pneumonia frequently complicates stroke and has a major impact on outcome. We derived and internally validated a simple clinical risk score for predicting stroke‐associated pneumonia (SAP), and compared the performance with an existing score (A²DS²).

Methods and Results

We extracted data for patients with ischemic stroke or intracerebral hemorrhage from the Sentinel Stroke National Audit Programme multicenter UK registry. The data were randomly allocated into derivation (n=11 551) and validation (n=11 648) samples. A multivariable logistic regression model was fitted to the derivation data to predict SAP in the first 7 days of admission. The characteristics of the score were evaluated using receiver operating characteristics (discrimination) and by plotting predicted versus observed SAP frequency in deciles of risk (calibration). Prevalence of SAP was 6.7% overall. The final 22‐point score (ISAN: prestroke Independence [modified Rankin scale], Sex, Age, National Institutes of Health Stroke Scale) exhibited good discrimination in the ischemic stroke derivation (C‐statistic 0.79; 95% CI 0.77 to 0.81) and validation (C‐statistic 0.78; 95% CI 0.76 to 0.80) samples. It was well calibrated in ischemic stroke and was further classified into meaningful risk groups (low 0 to 5, medium 6 to 10, high 11 to 14, and very high ≥15) associated with SAP frequencies of 1.6%, 4.9%, 12.6%, and 26.4%, respectively, in the validation sample. Discrimination for both scores was similar, although they performed less well in the intracerebral hemorrhage patients with an apparent ceiling effect.

Conclusions

The ISAN score is a simple tool for predicting SAP in clinical practice. External validation is required in ischemic and hemorrhagic stroke cohorts.