Clinical Scoring Model Research Articles

Predictors of pathologic complete response after neoadjuvant chemoradiotherapy for locally advanced rectal cancer

Objective To evaluate the potential influencing factors associated with pathologic complete response (pCR) after neoadjuvant chemoradiotherapy for locally advanced rectal cancer (LARC). Methods A retrospective analysis was performed on the clinical data 265 patients with stage Ⅱ and Ⅲ(the 7th version of AJCC) rectal cancer admitted to our hospital from 2011 to 2013. All patients underwent neoadjuvant concurrent chemoradiotherapy (CCRT) followed by surgery with/or without induction chemotherapy during the interval between the complete of CCRT and surgery. The predictors associated with pCR were analyzed by univariate and multivariate logistic regression analyses. With the use of the independent predictive variables for pCR from multivariate analysis, a clinical risk score model was established according to the following criteria: no-risk group (0 factor); low-risk group (1 factor); high-risk group (2 factors). Results Among these 265 patients, 50(18.9%) achieved pCR. The univariate analysis showed that carcinoembryonic antigen (CEA) level before CCRT (P=0.017), T stage before CCRT (P=0.001), interval between complete of CCRT and surgery (P=0.000), and the maximum tumor thickness before CCRT (P=0.040) were significantly associated with pCR. The multivariate analysis showed that pre-CCRT CEA level (P=0.021 or 0.446) and interval between the complete of CCRT and surgery (P=0.000 or 3.774) were significant predictors of pCR. When stratifying for smoking status, only low pre-CCRT CEA level was significantly associated with pCR in the non-smoking patients (P=0.044). For the prediction of pCR by the clinical risk score model, the sensitivity was 0.805, the specificity was 0.460, the area under the receiver operating curve was 0.690(95% CI=0.613-0.767), the positive predictive value was 35.49%, the negative predictive value was 86.5%, and the predictive accuracy was 73.9%. Conclusions For locally advanced rectal cancer, pCR can be achieved in some patients after neoadjuvant therapy. Low pre-CCRT CEA level and long interval time between CCRT and surgery are independent factors associated with pCR, and only low pre-CCRT CEA level is an associated factor in the group of nonsmokers. The clinical risk score model based on pre-CCRT CEA level>5 ng/ml and time interval from CCRT completion to surgery ≤8 weeks can be used to predict pCR after neoadjuvant chemoradiotherapy for LARC. Key words: Rectal neoplasms/neoadjuvant therapy; Pathologic complete response; Factors analysis

Abstract 15903: An Improvement of Score Models in the Prediction of Deep Vein Thrombosis of Upper and Lower Limbs in Hospitalized Patients

Introduction:Deep Vein Thrombosis (DVT) of both lower and upper limbs is increasingly being recognized as a clinical entity in patients with chronic diseases hospitalized for several comorbidities. The occurrence of DVT in sites other than lower limbs is an emerging clinical presentation in a setting of a Department of Medicine. The relationship between score models, based on risk factors, symptoms and clinical signs of lower limbs DVT, and circulating biomarkers is not well studied. Hypothesis:We hypothesized to give an additional value to score models using circulating biomarkers in the prediction of DVT in patients hospitalized in Internal Medicine Department. Methods:We prospectively recruited 178 consecutive patients who presented a D-dimer value >500 ng/ml. In each patient Wells modified score for DVT, Hamilton score, Kahn score and St. André Hospital score were applied. We measured serum levels of albumin (%) and protein S (%). Diagnosis of DVT was made by Extended Compression Ultra-Sonography (ECUS) standardized procedure. All patients underwent also an angiographic-CT-scan Results:In 85 (48%) patients the diagnosis of DVT was confirmed. Sites of DVT were as follows: Lower limbs venous thrombosis n=71 (84%); Upper limb venous thrombosis n=8 (9%); portal vein thrombosis n=6 (7%). Patients with DVT had significantly lower albumin levels as compared with controls (mean ± SD 47.6 ± 7.9 vs 55.4 ± 4.6 %, respectively, P<0.001) and lower Protein S (60.7 ± 26.3 vs. 87.6 ± 11.9 %, P<0.001). In this setting, the use of clinical risk scores based on symptoms and clinical signs showed low accuracy (AUC 0.690). We gave an additional value to these scores by testing in combination biomarkers and two of the best items of all clinical score models used to predict DVT: we evaluated a ROC curve analysis including value of Albumin (%), Protein S (%), previous DVT and cancer obtaining a higher accuracy (AUC 0.901, Positive Likelihood Ratio (LR+) 12.97 and Negative Likelihood Ratio (LR-) 0.17). Conclusions:This study demonstrates that biomarkers improve the prediction of clinical scores in the diagnosis of DVT of all sites in hospitalized patients.

Diastolic myocardial dysfunction by tissue Doppler imaging predicts mortality in patients with cerebral infarction.

Several clinical prediction score models have been investigated for predicting mortality in patients with cerebral infarction. However, none of these include echocardiographic measures. Our objective was to evaluate the prognostic value of tissue Doppler imaging (TDI) of the myocardium in patients with cerebral infarction. Two hundred forty-four patients with cerebral infarction and subsequent echocardiographic examination in sinus rhythm were identified. Using TDI in three apical projections, longitudinal mitral annular velocities were obtained in six segments. Cox regression models, C-statistics and reclassification analysis were performed for global and segmental e'. During a median follow-up of 3 years 42 patients died. Patients who died had significantly impaired systolic and diastolic function (determined by LVEF and E/e'). The risk of dying increased with decreasing global e', being approximately 13 times higher for patients in the lowest tertile compared to patients in the highest tertile (HR 13.4 [3.2;56.3], p < 0.001). Patients with significantly impaired global e' showed increased mortality after multivariable adjustment for: LVEF, E/e', age, gender, heart failure, chronic obstructive pulmonary disease, prior cerebral infarction, ischemic heart disease, cancer, hypertension, hypercholesterolemia, carotid stenosis, mitral regurgitation, liver disease and thromboembolisms (HR 1.9 [1.1;3.2]), per 1 cm/s decrease, p < 0.05). Similar pattern was seen in segmental analyses of the e'. In contrast to e', no conventional echocardiographic parameters remained independent predictors of mortality after multivariable adjustment. Diastolic myocardial dysfunction determined as e' by TDI is a significant predictor of mortality in patients with cerebral infarction. Applying this parameter can aid the prognostic assessment after cerebral infarction.

Score model for predicting acute-on-chronic liver failure risk in chronic hepatitis B.

To establish a clinical scoring model to predict risk of acute-on-chronic liver failure (ACLF) in chronic hepatitis B (CHB) patients. This was a retrospective study of 1457 patients hospitalized for CHB between October 2008 and October 2013 at the Beijing Ditan Hospital, Capital Medical University, China. The patients were divided into two groups: severe acute exacerbation (SAE) group (n = 382) and non-SAE group (n = 1075). The SAE group was classified as the high-risk group based on the higher incidence of ACLF in this group than in the non-SAE group (13.6% vs 0.4%). Two-thirds of SAE patients were randomly assigned to risk-model derivation and the other one-third to model validation. Univariate risk factors associated with the outcome were entered into a multivariate logistic regression model for screening independent risk factors. Each variable was assigned an integer value based on the regression coefficients, and the final score was the sum of these values in the derivation set. Model discrimination and calibration were assessed using area under the receiver operating characteristic curve and the Hosmer-Lemeshow test. The risk prediction scoring model included the following four factors: age ≥ 40 years, total bilirubin ≥ 171 μmol/L, prothrombin activity 40%-60%, and hepatitis B virus DNA > 10(7) copies/mL. The sum risk score ranged from 0 to 7; 0-3 identified patients with lower risk of ACLF, whereas 4-7 identified patients with higher risk. The Kaplan-Meier analysis showed the cumulative risk for ACLF and ACLF-related death in the two risk groups (0-3 and 4-7 scores) of the primary cohort over 56 d, and log-rank test revealed a significant difference (2.0% vs 33.8% and 0.8% vs 9.4%, respectively; both P < 0.0001). In the derivation and validation data sets, the model had good discrimination (C index = 0.857, 95% confidence interval: 0.800-0.913 and C index = 0.889, 95% confidence interval: 0.820-0.957, respectively) and calibration demonstrated by the Hosmer-Lemeshow test (χ (2) = 4.516, P = 0.808 and χ (2) = 1.959, P = 0.923, respectively). Using the scoring model, clinicians can easily identify patients (total score ≥ 4) at high risk of ACLF and ACLF-related death early during SAE.

A Rapid Heparin Antibody Detection Assay Performs Better Than 4T’s Score in Predicting HIT Diagnosis: A Single Community Medical Center Retrospective Analysis

Introduction: The diagnosis of heparin induced thrombocytopenia (HIT) and thrombosis (HITT) is challenging due to poor availability of the gold standard serotonin releasing assay (SRA) and suboptimal positive predictive value from clinical scoring models such as 4T score. A common algorithm used for diagnosing HIT is: 4T’s pretest probability score estimation in cases suspected of HIT; followed by HIT antibody test in the intermediate to high risk groups; followed by confirmation with SRA test in HIT antibody positive patients. Since 2011, a Particle Immune-Filtration Assay (PIFA) Heparin/Platelet Factor 4 Rapid Assay (HPF4-RA) (Akers Bioscience, Inc, Thorofare, NJ) became available in our medical center and test results were available on the same day. We observed that HPF4-RA test was being routinely ordered along with SRA test at the same time. We performed this retrospective analysis to evaluate and compare the predictive performance for SRA positive HIT diagnosis using 4T score or HPF4-RA. We applied a regression analysis model, to calculate area under receiver operating characteristics (ROC) curve.Methods: A list of all consecutive patients who had HIT antibody test and/or SRA test performed between January 2010 and June 2013 was obtained, which consisted of 402 patients. Patients with duplication of tests were deleted from analysis. 283 patients had results reported for both HPF4-RA (positive in n=42, negative in n=241) and SRA tests (positive in n=16 and negative in n=267); and these results were used for calculation of HPF4-RA prediction model. Two patients had HPF4-RA negative result but SRA positive test result. 4T’s scores were calculated for 125 patients, consisting of all HPF4-RA positive patients (n=42), and patients randomly selected from the total HPF4-RA negative pool (n=83). Electronic medical records were reviewed for temporal trend of platelet counts, diagnosis, medication use, Doppler tests and competing causes of thrombocytopenia. Persons calculating the 4T’s score were blinded to the laboratory test results.Results: Stratification of the patients with 4T’s score analysis (n=125) revealed that the distribution of SRA positive patients (n=16) was 31.3% (n=5) in low risk, 31.3% (n=5) in intermediate risk, and 37.5% (n=6) in high risk groups; while the distribution of SRA negative patients (n=109) was 45.9% (n=50) in low risk, 43.1% (n=47) in intermediate risk and 11.0% (n=12) in high risk groups. The area under receiver operating characteristics (ROC) curve for 4T score as a continuous variable to predict SRA positive HIT was 0.659 (95% CI 0.516 - 0.802; p = 0.041), and the area under ROC curve for HPF4-RA to predict SRA positive HIT was 0.818 ( 95% CI 0.712 - 0.924; p = 0.00) (Figure 1). HPF4-RA test also showed better overall prediction parameters for HIT as shown in Table 1. A combination of HIT HPF4-RA positive result and a 4T score ≥ 4 did not increase the area under ROC curve for prediction of SRA positive HIT.Abstract 1457. Table1:Predictive performance of 4T’s score and HPF4-RA for HIT (defined by positive SRA)Sensitivity (95% CI) Specificity (95% CI) PPV (95% CI) NPV (95% CI) Number of patients (%)4T’s score ≤ 3 (Low Risk)0.31 (0.11 – 0.59)0.72 (0.64 - 0.79)0.11 (0.03 - 0.23)0.91 (0.84 - 0.95)56 (44.8)4T’s score ≥ 4 (Intermediate and High Risk)0.69 (0.41-0.89) 0.39 (0.29 - 0.48) 0.14 (0.72 - 0.24) 0.89 (0.77 - 0.96) 69 (55.2) 4T’s score ≥ 6 (High Risk)0.37 (0.15-0.65) 0.82 (0.74 - 0.89) 0.24 (0.09 - 0.45) 0.90 (0.82 - 0.95) 17 (13.6) HPF4-RA Test0.88 (0.62-0.98) 0.86 (0.81- 0.90) 0.26 (0.16 - 0.41) 0.99 (0.96 - 0.99) 283 PPV: Positive Predictive Value. NPV: Negative Predictive Value [Display omitted] Conclusions: Both 4T’s score and HPF4-RA testing predict SRA positive HIT more than chance; however HPF4-RA testing predicts SRA positive HIT better than 4T’s scores with higher sensitivity, specificity and NPV. This result challenges the pretesting algorithm for selecting patients for confirmatory SRA testing to diagnose HIT. Instead of using 4T’s score as a screening tool for selecting patients for HPF4 antibody testing; rapid HPF4 antibody assays when available, should be considered as upfront screening tool and positive results considered for confirmatory SRA testing for diagnosis of HIT. Further studies are warranted to confirm this data. DisclosuresNo relevant conflicts of interest to declare.

A combined ultrasound and clinical scoring model for the prediction of peripartum complications in pregnancies complicated by placenta previa

ObjectivesTo generate a combined ultrasound and clinical model predictive for peripartum complications in pregnancies complicated by placenta previa. Study designThis study included 110 singleton pregnant women with placenta previa delivered by cesarean section (CS) from July 2011 to November 2013. We prospectively collected ultrasound and clinical data before CS and observed the occurrence of blood transfusion, uterine artery embolization and cesarean hysterectomy. We formulated a scoring model including type of previa (0: partials, 2: totalis), lacunae (0: none, 1: 1–3, 2: 4–6, 3: whole), uteroplacental hypervascularity (0: normal, 1: moderate, 2: severe), multiparity (0: no, 1: yes), history of CS (0: none, 1: once, 2: ≥ twice) and history of placenta previa (0: no, 1: yes) to predict the risk of peripartum complications. ResultsIn our study population, the risk of perioperative transfusion, uterine artery embolization, and cesarean hysterectomy were 26.4, 1.8 and 6.4%, respectively. The type of previa, lacunae, uteroplacental hypervascularity, parity, history of CS, and history of placenta previa were associated with complications in univariable analysis. However, no factor was independently predictive for any complication in exact logistic regression analysis. Using the scoring model, we found that total score significantly correlated with perioperative transfusion, cesarean hysterectomy and composite complication (p<0.0001, Cochrane Armitage test). Notably, all patients with total score ≥7 needed cesarean hysterectomy. When total score was ≥6, three fourths of patients needed blood transfusion. ConclusionsThis combined scoring model may provide useful information for prediction of peripartum complications in women with placenta previa.

External Validation Of The HIT Expert Probability (HEP) Score, a Novel Pre-Test Probability Model For Heparin-Induced Thrombocytopenia Based On Broad Expert Opinion

IntroductionThe diagnosis of heparin-induced thrombocytopenia (HIT) is challenging due to the overlap of its clinical features with other entities, variable specificity of ELISA-based tests, poor availability of functional tests such as the serotonin release assay (SRA) and suboptimal positive predictive values of pre-test probability clinical scoring models, such as the 4T and Lillo-Le Louėt models. Improved diagnostic tools are desirable to guide early therapeutic decisions in patients with suspected HIT and to reduce inappropriate use of alternative anticoagulants. The HIT Expert Probability (HEP) Score, a pre-test probability clinical scoring model for HIT based on broad expert opinion, has recently been proposed to improve the diagnosis of HIT compared to the 4T score. We sought to externally and prospectively validate the HEP score. MethodsPre-test probability of HIT was prospectively assessed for 51 consecutive patients referred to the Cleveland Clinic Vascular Medicine Consultative Service for evaluation of possible HIT between August 1, 2012 and February 1, 2013. Electronic medical records were reviewed to obtain clinical, biochemical (including IgG specific HIT ELISA and SRA tests) and imaging data, and two Vascular Medicine fellows independently applied the 4T and HEP scores for each patient. Two independent HIT expert adjudicators rendered a diagnosis of HIT likely or unlikely. Both scorers and adjudicators were blinded to the Results of HIT laboratory testing, consultative service inputs and scores generated by other raters. ResultsThe median 4T and HEP scores were 4.5 [range 3.0, 6.0] and 5 [range 3.0, 8.5], respectively. Interobserver agreement was lower for the HEP Score compared to 4T score, intraclass correlation coefficient: 0.50 (95% CI 0.16–0.85) vs. 0.69 (0.54–0.86). There were no significant differences between area under receiver-operating characteristics curves of 4T and HEP score against HIT ELISA testing (0.62 vs. 0.58, P = 0.62), SRA (0.74 vs. 0.73, P = 0.97), expert panel diagnosis (0.71 vs. 0.79, P = 0.20), concordant HIT ELISA and SRA (0.76 vs. 0.72, P = 0.65), and positive SRA or positive HIT ELSA with HIT likely expert panel diagnosis (0.73 vs. 0.75, P = 0.80). The HEP score model was 100% sensitive and 17% specific for determining the presence of HIT as defined by positive SRA or positive HIT ELISA with HIT likely expert panel diagnosis compared to 4T score (sensitivity, 94%; specificity 38%) (Figure 1, Table 1). [Display omitted] ConclusionThe HEP and 4T scores are excellent screening pre-test probability models for HIT, however, in this prospective validation study, test characteristics for the diagnosis of HIT based on confirmatory laboratory testing and expert opinion are similar. Given the complexity of the HEP scoring model compared to that of the 4T model, further validation of the HEP score is warranted prior to widespread acceptance. Disclosures:Kim:Philips US: Consultancy; GE: Research Funding.

Evaluation of Hypertrophic Pyloric Stenosis by Pediatric Emergency Physician Sonography

The objective was to evaluate the accuracy of pediatric emergency physician (EP) sonography for infants with suspected hypertrophic pyloric stenosis (HPS). This was a prospective observational pilot study in an urban academic pediatric emergency department (PED). Patients were selected if the treating physician ordered an ultrasound (US) in the department of radiology for the evaluation of suspected HPS. Sixty-seven patients were enrolled from August 2009 through April 2012. When identifying the pylorus, pediatric EPs correctly identified all 10 positive cases, with a sensitivity of 100% (95% confidence interval [CI] = 62% to 100%) and specificity of 100% (95% CI = 92% to 100%). There was no statistical difference between the measurements obtained by pediatric EPs and radiology staff for pyloric muscle width or length (p = 0.5 and p = 0.79, respectively). Trained pediatric EPs can accurately assess the pylorus with US in the evaluation of HPS with good specificity.

Clinical Prediction of Pathological Complete Response After Preoperative Chemoradiotherapy for Rectal Cancer

The clinical pretreatment factors that accurately predict response to chemoradiation in rectal cancer are not currently known. The aim of this study is to evaluate the clinical factors associated with a pathological complete response after preoperative chemoradiotherapy for rectal cancer. This study is a retrospective review of prospectively collected data. This study was conducted at a tertiary care hospital/referral center in South Korea. From December 2000 to September 2011, a total of 391 consecutive patients with rectal cancer who underwent neoadjuvant chemoradiotherapy followed by radical surgery were identified. The treatment consisted of concurrent chemoradiation, which included preoperative 5-fluorouracil-based chemotherapy and pelvic radiation (median, 5040 cGy); this was followed 8 weeks later (median, 57 days) by surgery with curative intent. The primary outcome measured was the clinicopathological comparison between pathological complete response (n = 57, 14.6%) and non-pathological complete response (n = 334, 85.4%) groups. The pathological complete response groups had a higher percentage of noncircumferential tumors, nonmacroscopic ulceration, well differentiation, small tumor diameter, early clinical T stage, early clinical N stage, or low levels of pretreatment CEA than the non-pathological complete response group. In multivariate regression analysis, independent predictors of a higher pathological complete response rate were noncircumferentiality (p = 0.007; OR, 3.214), nonmacroscopic ulceration (p = 0.002; OR, 6.702), and low pretreatment CEA level (p = 0.004; OR, 2.656). Significant differences in the pathological complete response rate existed among the 4 risk stratification groups (p < 0.001). For the prediction of pathological complete response by the clinical risk score model, the sensitivity was 64.1% and the specificity was 73.7% (area under the curve, 0.706; p < 0.001). This study was limited because it was a single-institution study with a small sample size. Pretreatment clinical variables, including tumor circumferentiality, macroscopic ulceration, and CEA level, may be important determinants in achieving a pathological complete response.

Comparison of diagnostic performance of the heparin-induced thrombocytopenia expert probability and the 4Ts score in screening for heparin-induced thrombocytopenia

Heparin-induced thrombocytopenia (HIT) is a serious immunological complication of heparin administration. Diagnosis of HIT is challenging, especially in critically ill patients. The clinical scoring model for predicting HIT is helpful for guiding clinical decision. We analysed data of patients who underwent the heparin-induced platelet aggregation (HPA) test from 2006 to 2010 and compared diagnostic performance of the novel model HIT expert probability ('HEP'), which has been validated in a population mainly comprising surgical patients first, by the previously published model '4Ts' score. Clinical courses of the patients were also reviewed to ensure that HPA test results were accurate. There were 47 suspected HIT patients. The majority was from medical (70.2%) and/or critical care (61.7%) units. Ten (21.3%) yielded positive HPA. Among positive HPA patients, eight were medical patients. The HEP score ranged from -3 to 13, whereas the 4Ts score ranged from 3 to 7 in positive HPA patients. Both HEP and 4Ts scores were significantly higher in positive HPA than in negative HPA patients (5.35 vs. 1.81, P=0.010 and 4.85 vs. 3.32, P=0.001, respectively). The HEP score did not display better diagnostic performance than the 4Ts score, with receiver operating characteristic (ROC) area under curve of 0.72 and 0.79 (P=0.31), respectively. The HEP score did not show better diagnostic performance than the 4Ts score for predicting HIT in our population. A large prospective validation in different sets of patients is warranted.

Development and validation of a clinical trial feasibility assessment (CTFA) tool and evaluation of the NCI clinical trial evaluation and scoring (NCI-CTES) model.

e16549 Background: The ability to predict accrual success of a specific NCI sponsored clinical trial (CT) in the setting of a Community Clinical Oncology Program (CCOP) is typically based on the experience of the local principal investigator (PI) and research staff. This study was designed to develop an objective method to predict accrual success based on CT and institution specific factors. Funded by ARRA grant #3U10CA35412-26S1. Methods: 27 potential barriers to successful CT accrual were identified by Marshfield Clinic CCOP and validated by Eastern Cooperative Oncology Group (ECOG) PI and research staff (Cronbach alpha 0.922). This 27 item CTFA tool and the NCI-CTES model were prospectively applied to 32 consecutive NCI sponsored CT which opened at Marshfield Clinic CCOP (2009-11). Incorporating PI and research staff input each CT was scored (CTFA range 27-135; NCI-CTES 0-22). These scores were correlated with accrual outcome (Wilcoxon`s rank-sum test). Statistical analysis included CTFA item content validity testing via item-test correlation and CTFA/NCI-CTES internal consistency reliability testing via Cronbach alpha. Results: 28/32 CT were available for analysis; 4 CT were excluded (never opened to accrual). Duration of time open to accrual ranged from 1-27 months (mean=13.5). 21/28 CT achieved accrual ranging from 1-50 subjects (1-5 accrual=13; 5-10=4; &gt;10=4). Median time to complete the 27 item CTFA tool was 16 minutes and 8 minutes for the NCI-CTES model. After eliminating 6/27 CTFA items due to low content validity (&lt;0.2), the remaining 21 item final CTFA tool indicated very high internal consistency reliability (Cronbach alpha 0.872). Wilcoxon`s rank-sum test revealed that the differences in NCI-CTES and CTFA scores between the groups with &lt;=0.25 and &gt;0.25 subject accrual per month were significant (CTFA P-value=0.0418; NCI-CTES P-value=0.0032). Conclusions: Application of the CTFA tool and the NCI-CTES model proved to be feasible and predictive of accrual success of NCI sponsored clinical trials at Marshfield Clinic CCOP.

Host immune gene polymorphisms were associated with the prognosis of non‐small‐cell lung cancer in Chinese

Laboratory-based studies showed that host immune genes could influence the prognosis of non-small-cell lung cancer (NSCLC). Therefore, genetic polymorphisms in host immune genes may serve as predictors for NSCLC clinical outcome. To test the hypothesis that functional single nucleotide polymorphisms (SNPs) in host immune genes are associated with the prognosis of NSCLC, we systematically performed a genotyping analysis for a total of 178 SNPs from 52 immune genes in a prospective case cohort of 568 NSCLC patients. Among the 178 SNPs, 24 were significantly associated with NSCLC prognosis in different genetic models and four of them were remained in the final predictive model after multivariate stepwise Cox regression, including IL-5R rs11713419 (5'-untranslated region, 5'-UTR) (P = 0.001), IL23R rs6682925 (5'-flanking region, 5'-FR) (P = 0.017), TLR1 rs5743551 (5'-FR) (P = 0.02) and TLR3 rs3775291 (Leu412Phe) (P = 0.01). We then put the above four SNPs together, and found that the risk of death was significantly increased by 124% (HR = 2.24, 95% CI: 1.33-3.75) for the patients carrying "1" unfavorable locus and by 175% (HR = 2.75, 95% CI: 1.67-4.51) for those carrying "2-4" unfavorable loci. The risk score model and time-dependent ROC analyses further support the four SNPs and clinical risk score model. The area under curve (AUC) at year 5 increased from 0.484 to 0.831 after combining the four SNPs risk score with clinical risk score. These findings indicate that potentially functional polymorphisms in immune genes may serve as prognostic markers of clinical outcome of NSCLC.

Usefulness of Multiple Biomarkers for the Prediction of Significant Fibrosis in Chronic Hepatitis B

It has not been explicitly addressed whether new biomarkers, in addition to readily available clinical data, make a contribution to the prediction of liver fibrosis. A total of 209 patients with chronic hepatitis B who underwent liver biopsy were recruited. The Clinical Score (CS) model using only routine clinical data and the Biomarker Score (BS) model using 7 putative biomarkers were derived from the derivation set (n=105), and these models were applied to a separate patients group (n=104) to investigate whether the addition of BS improved the diagnostic accuracy in predicting significant fibrosis beyond an assessment based solely on CS. The most informative biomarkers for predicting significant fibrosis were hyaluronic acid and matrix metalloproteinase-2. The BS was an independent predictive factor for significant fibrosis even after accounting for CS in both sets. Among the derivation set, the incorporation of the BS into the CS model did not significantly increase the receiver operating characteristic area, with only a small improvement of about 2% (P=0.11). Similarly, in the validation set, a combined model with CS and BS showed no superior diagnostic accuracy over the CS model alone, with an improvement of approximately 2% [0.83 (0.75 to 0.92) vs. 0.81 (0.70 to 0.91); P=0.16]. The simultaneous addition of several biomarkers adds only modestly to clinical predictive factors for risk assessment of individual patients. These results highlight the need for the models to be validated in another cohort with a broader distribution of fibrosis severity.

The HIT Expert Probability (HEP) Score: a novel pre-test probability model for heparin-induced thrombocytopenia based on broad expert opinion.

The diagnosis of heparin-induced thrombocytopenia (HIT) is challenging. Over-diagnosis and over-treatment are common. To develop a pre-test clinical scoring model for HIT based on broad expert opinion that may be useful in guiding clinical decisions regarding therapy. A pre-test model, the HIT Expert Probability (HEP) Score, was constructed based on the opinions of 26 HIT experts. Fifty patients referred to a reference laboratory for HIT testing comprised the validation cohort. Two hematology trainees scored each patient using the HEP Score and a previously published clinical scoring system (4 T's). A panel of three independent experts adjudicated the 50 patients and rendered a diagnosis of HIT likely or unlikely. All subjects underwent HIT laboratory testing with a polyspecific HIT ELISA and serotonin release assay (SRA). The HEP Score exhibited significantly greater interobserver agreement [intraclass correlation coefficient: 0.88 (95% CI 0.80-0.93) vs. 0.71 (0.54-0.83)], correlation with the results of HIT laboratory testing and concordance with the diagnosis of the expert panel (area under receiver-operating curve: 0.91 vs. 0.74, P = 0.017) than the 4 T's. The model was 100% sensitive and 60% specific for determining the presence of HIT as defined by the expert panel and would have allowed for a 41% reduction in the number of patients receiving a direct thrombin inhibitor (DTI). The HEP Score is the first pre-test clinical scoring model for HIT based on broad expert opinion, exhibited favorable operating characteristics and may permit clinicians to confidently reduce use of alternative anticoagulants. Prospective multicenter validation is warranted.

Multicentric evaluation of a clinical and prognostic scoring system predictive of survival after resection of intrahepatic cholangiocarcinomas

To develop a clinical and prognostic scoring system predictive of survival after resection of intrahepatic cholangiocarcinomas (ICC). Two hundred and one consecutive ICC patients (83 from Essen, Germany, 54 from New York, USA and 64 from Chiba, Japan). The scoring systems were developed utilizing the data set from Essen University and then applied to the data sets from Mount Sinai Medical Center and Chiba University for validation. Eighteen potential prognostic factors were evaluated. Statistical analysis included multivariable regression analyses with the Cox proportional hazard model, power analysis, internal validation with structural equation modelling bootstrapping and external validation. The prognostic scoring model was based mainly in pathological and demographical variables, whereas the clinical scoring model was based mainly in radiological and demographical variables. Gender (P=0.0086), UICC stage (P=0.0140) and R-class (P=0.0016) were predictive of survival for the prognostic scoring model, while gender (P=0.0023), CA 19-9 levels (P=0.0153) and macrovascular invasion (P=0.0067) were predictive of survival for the clinical scoring model. Prognostic points were assigned as follows: female:male=1:2 points, UICC (I-II):UICC (III-IV)=1:2 points and R0:R1=1:2 points. Clinical points were allocated as follows: female:male=1:2 points, CA 19-9 (<100 U/ml):CA 19-9 (> or =100 U/ml)=1:2 points and no macrovascular invasion:macrovascular invasion=1:2 points. Prognostic groups with 3-4, 5 and 6 points (P=0.000001) and clinical groups with 3-4 and 5-6 points (P=0.0103) achieved statistically significant difference. We propose a clinical and prognostic scoring system predictive of long-term survival after surgical resections for ICC.

Predicting and preventing acute kidney injury after cardiac surgery

Acute kidney injury (AKI) after cardiac surgery is associated with significant morbidity and mortality. Despite the proliferation of predictive clinical scoring models of renal risk after cardiac surgery, limitations in preventing AKI through the use of pharmacological agents remain. Here we review the evolution of predictive models of renal risk after cardiac surgery, and highlight the important gains made in preventing its occurrence. Simple risk indices predicting AKI after cardiac surgery have been developed and can now be readily applied clinically. However, studies focusing on preventing AKI after surgery have yet to demonstrate any consistent renoprotective effect. Clinical scoring systems predicting AKI risk after cardiac surgery are available and should be employed in the preoperative assessment. Elucidation of beneficial preventive strategies of AKI after cardiac surgery requires ongoing research.

A Simple and Safe Method of Ruling Out Pulmonary Embolism in Postoperative Plastic Surgery Patients

Sir: Pulmonary embolism is one of the most feared complications of plastic surgery, and as authors have stated, preventive measures are the best way to minimize this potentially fatal complication. However, even with all the possible prophylactic measures, pulmonary embolism may be unavoidable and necessitates prompt diagnosis and aggressive treatment. A plastic surgeon is usually not the one who treats the pulmonary embolism, but he or she should be the one who diagnoses it first to get appropriate help in a timely manner to prevent a deadly outcome. In this communication, we share a simple, noninvasive, bedside algorithm that we have been using in our clinical practice to diagnose pulmonary embolism in postoperative plastic surgery patients. The diagnosis of pulmonary embolism is challenging in the postoperative period. If it is unrecognized or left untreated, it may result in a mortality rate as high as 30 percent in hospitalized patients; this rate can be decreased to 8 percent with early diagnosis and proper treatment.1,2 Thus, it is imperative to establish a bedside diagnostic algorithm to diagnose and treat this deadly complication. Wells et al. established a simplified clinical scoring model,3 which was derived from their original study,4 to eliminate some of the disadvantages of the complex scoring systems. In their simplified model, they gave points to various clinical signs, symptoms, and risk factors, and categorized the patients into two groups: pulmonary embolism unlikely and pulmonary embolism likely (Table 1). This is a relatively simple scoring system that can be applied at the bedside to any patient with suspected pulmonary embolism. This simplified model is particularly helpful when combined with D-dimer blood levels of patients with suspected pulmonary embolism.Table 1: Simplified Clinical Scoring Model for Diagnosis of Pulmonary EmbolismD-dimer is a degradation product of a cross-linked fibrin blood clot. The elevated D-dimer levels are typically seen with acute venous thromboembolism, but it is not a specific test, because elevated levels may also be seen with recent surgery, malignancy, pregnancy, infections, and in the older population. A pulmonary-embolism-unlikely result from the simplified clinical scoring model (Table 1) and a negative D-dimer value can safely exclude the diagnosis of pulmonary embolism without the need for further diagnostic studies with a 2 percent false-negative rate, which is similar to standard pulmonary angiography.5 This approach is particularly helpful in free flap patients, because transport of these patients to nuclear medicine or computed tomography suites is cumbersome, and when both simplified clinical scoring and D-dimer levels are negative, an unnecessary and potentially dangerous trip may be avoided. If pulmonary embolism is likely based on the simplified clinical scoring system and/or the D-dimer level is positive, further diagnostic studies are usually needed to rule in or rule out the diagnosis of pulmonary embolism. The most commonly used diagnostic studies are spiral computed tomography and ventilation-perfusion nuclear scanning. Both studies have advantages and disadvantages, and either one is acceptable as a first-line diagnostic study, which is usually determined by the institution policy, availability, and physician’s preference. The simplified clinical scoring system (Table 1) and blood D-dimer levels are readily available, can be performed at the bedside, and can safely exclude the pulmonary embolism when both are negative. If either one is positive, more sensitive and specific diagnostic modalities, such as spiral computed tomography and nuclear scanning, should be considered to rule in or rule out the diagnosis of pulmonary embolism. Adil Ceydeli, M.D., M.S. Jack Yu, M.D., D.M.D., M.S. Division of Plastic and Reconstructive Surgery Medical College of Georgia Atlanta, Ga.

Intérêt du score de probabilité clinique dans le diagnostic de thrombose veineuse profonde en gériatrie

There is a lack of scientific data concerning the incidence, clinical signs, risk factors and diagnostic management of deep vein thrombosis (DVT) in the elderly. We carried out a prospective case-control study in a geriatric setting. We included one in-patient without clinical signs of deep vein thrombosis for each hospitalised patient with clinically suspected DVT. Clinical signs, risk factors of DVT and a compression ultrasonography of the proximal and distal leg veins were performed in all patients at admission. Also, the clinical probability was assessed by means of the clinical score model described by Wells. There were 102 patients aged over 85 included during the 5 months period of the study. Fifty-one patients have had a clinical suspicion of DVT and 51 patients were clinically asymptomatic. There were 43% symptomatic DVT and 12% asymptomatic DVT. Clinical signs and symptoms of DVT were neither sensitive nor specific. Risk factors associated most frequently with DVT were: immobilisation (prevalence 64%), paralysis of the legs (22%), varicose veins (25%), deshydratation (28,5%), infections (18%). A significant relationship between a risk factor and DVT was found only for immobilisation (P<0,05) and deshydratation (P<0,02). The prevalence of DVT increased with the clinical probability score: 11.4% for the low score, 27.6% for the moderate score and 55% for the high score. Incidence of DVT is high among hospitalised elderly patients, especially for the asymptomatic ones. Clinical signs alone do not reliably predict DVT. Clinical probability score could be useful to improve diagnostic management of DVT in this population.

The importance of severity of illness adjustment in predicting adverse outcomes in the Medicare population

The importance of using risk-adjusted mortality rates to measure quality of care is well-established. However, mortality rates may be an insensitive measure of quality for surgical patients since death is a relatively rare outcome. This study used Medicare files to identify, through chart abstraction, clinical postoperative complications of four surgical procedures ( n = 8126) that could serve as measures of quality. Disease-specific severity of illness models using a moderate number of clinical variables and admission MedisGroups score models computed from approximately 250 clinical variables were compared in predicting postoperative adverse events. Initial differences between the two models disappeared upon cross-validation. Validated R-squareds and C statistics from models using half the data were generally positive, suggesting that these models had real, although modest, predictive power. We have shown that severity of illness on admission plays a role in predicting adverse events of surgery. Risk-adjusted outcomes may potentially be useful in screening for quality shortfalls.