AUC Measurement Research Articles

Perbandingan Performa Metode Pohon Model Logistik dan Random Forest pada Pengklasifikasian Data

Multicollinearity and missing data are two common problems in big data. Missing data could decrease the prediction accuracy. Logistic model tree (LMT) is used to handle multicollinearity because multicollinearity does not affect the decision tree. Random forest can be used to decrease variance in prediction case. This study aimed to study the comparison of two methods, LMT and random forest, in multicollinearity and missing data in various cases using simulation study and real data as dataset. Evaluation model is based on classification accuracy and AUC measurement. The result stated that random forest had better performance if the multicollinearity level is moderate. LMT with omitted missing data is proven to have better performance for big data and when a high percentage of missing data occurred, and the multicollinearity level is severe. The next step is analysed real data with different sample size. The result stated that random forest have better performance. Omitted missing data have better performance in classification “breast cancer” data which consist 0,3 % missing data.

Integrative identification of immune-related key genes in atrial fibrillation using weighted gene coexpression network analysis and machine learning

BackgroundThe immune system significantly participates in the pathologic process of atrial fibrillation (AF). However, the molecular mechanisms underlying this participation are not completely explained. The current research aimed to identify critical genes and immune cells that participate in the pathologic process of AF.MethodsCIBERSORT was utilized to reveal the immune cell infiltration pattern in AF patients. Meanwhile, weighted gene coexpression network analysis (WGCNA) was utilized to identify meaningful modules that were significantly correlated with AF. The characteristic genes correlated with AF were identified by the least absolute shrinkage and selection operator (LASSO) logistic regression and support vector machine recursive feature elimination (SVM-RFE) algorithm.ResultsIn comparison to sinus rhythm (SR) individuals, we observed that fewer activated mast cells and regulatory T cells (Tregs), as well as more gamma delta T cells, resting mast cells, and M2 macrophages, were infiltrated in AF patients. Three significant modules (pink, red, and magenta) were identified to be significantly associated with AF. Gene enrichment analysis showed that all 717 genes were associated with immunity- or inflammation-related pathways and biological processes. Four hub genes (GALNT16, HTR2B, BEX2, and RAB8A) were revealed to be significantly correlated with AF by the SVM-RFE algorithm and LASSO logistic regression. qRT–PCR results suggested that compared to the SR subjects, AF patients exhibited significantly reduced BEX2 and GALNT16 expression, as well as dramatically elevated HTR2B expression. The AUC measurement showed that the diagnostic efficiency of BEX2, HTR2B, and GALNT16 in the training set was 0.836, 0.883, and 0.893, respectively, and 0.858, 0.861, and 0.915, respectively, in the validation set.ConclusionsThree novel genes, BEX2, HTR2B, and GALNT16, were identified by WGCNA combined with machine learning, which provides potential new therapeutic targets for the early diagnosis and prevention of AF.

Motor imagery classification in Brain computer interface (BCI) based on EEG signal by using machine learning technique

This paper focuses on classification of motor imagery in Brain Computer Interface (BCI) by using classifiers from machine learning technique. The BCI system consists of two main steps which are feature extraction and classification. The Fast Fourier Transform (FFT) features is extracted from the electroencephalography (EEG) signals to transform the signals into frequency domain. Due to the high dimensionality of data resulting from the feature extraction stage, the Linear Discriminant Analysis (LDA) is used to minimize the number of dimension by finding the feature subspace that optimizes class separability. Five classifiers: Support Vector Machine (SVM), K-Nearest Neighbors (KNN), Naïve Bayes, Decision Tree and Logistic Regression are used in the study. The performance was tested by using Dataset 1 from BCI Competition IV which consists of imaginary hand and foot movement EEG data. As a result, SVM, Logistic Regression and Naïve Bayes classifier achieved the highest accuracy with 89.09% in AUC measurement.

Transdermal Rivastigmine Delivery for Alzheimer Disease: Amenability of Exposure Predictions of Rivastigmine and Metabolite, NAP226-90, by Linear Regression Model Using Limited Samples.

Although an optimized delivery of rivastigmine for management of Alzhiemer disease (AD) is provided by the transdermal patch, it is critical to establish a limited sampling strategy for the measurement of exposure of rivastigmine/NAP226-90. The relationship Cmax versus AUC0-24h for rivastigmine/NAP226-90 was established by regression models. The derived regression equations enabled the prediction AUC0-24h for rivastigmine and NAP226-90. Models were evaluated using statistical criteria. Mixed model was used to predict AUC0-24h for rivastigmine/NAP226-90 from time points such as 8 (C8h), 12 (C12h), and 18 (C18h) hours. Excellent correlation was established for between Cmax and AUC0-24h for rivastigmine and NAP226-90. AUC0-24h predictions of either rivastigmine or NAP226-90 were within 0.8- to 1.25-fold difference. The RMSE in the AUC0-24h predictions ranged from 17.6% to 21.9%, and the R for prediction were greater than 0.96 for both rivastigmine and NAP226-90. Use of mixed model for C8h, C12h, and C18h resulted in AUC0-24h within 1.5-fold difference for rivastigmine or NAP226-90. Cmax of rivastigmine and NAP226-90 was highly correlated with the corresponding AUC0-24h values confirming the role of a time point closer to Cmax for an effective AUC measurement of rivastigmine or the metabolite.

Kinetic Analysis of Membrane Potential Dye Response to NaV1.7 Channel Activation Identifies Antagonists with Pharmacological Selectivity against NaV1.5

The NaV1.7 voltage-gated sodium channel is a highly valued target for the treatment of neuropathic pain due to its expression in pain-sensing neurons and human genetic mutations in the gene encoding NaV1.7, resulting in either loss-of-function (e.g., congenital analgesia) or gain-of-function (e.g., paroxysmal extreme pain disorder) pain phenotypes. We exploited existing technologies in a novel manner to identify selective antagonists of NaV1.7. A full-deck high-throughput screen was developed for both NaV1.7 and cardiac NaV1.5 channels using a cell-based membrane potential dye FLIPR assay. In assay development, known local anesthetic site inhibitors produced a decrease in maximal response; however, a subset of compounds exhibited a concentration-dependent delay in the onset of the response with little change in the peak of the response at any concentration. Therefore, two methods of analysis were employed for the screen: one to measure peak response and another to measure area under the curve, which would capture the delay-to-onset phenotype. Although a number of compounds were identified by a selective reduction in peak response in NaV1.7 relative to 1.5, the AUC measurement and a subsequent refinement of this measurement were able to differentiate compounds with NaV1.7 pharmacological selectivity over NaV1.5 as confirmed in electrophysiology.

P045 : DONOR-SPECIFIC ANTIBODY CAN BE MITIGATED BY INCREASING ANTIMETABOLITE DOSING ALONE

Aim The alloantibody response in solid organ transplantation is felt to be durable and long lasting. The durability of this response is appreciated in pretransplant patients where alloantibody can only be removed by plasmapheresis in combination with bortezomib, a proteasome inhibitor indicated for patients with relapsing multiple myeloma. We hereby report a cohort of post-transplant patients with donor specific antibody that has been successfully mitigated using increased antimetabolite alone. Methods Since October 2008, annual screening and testing for renal dysfunction has identified 238 renal transplant patients with donor-specific antibody, as characterized by Luminex single antigen bead interrogation; 114 patients demonstrated clinically significant DSA, with SFI > 100,000. A retrospective analysis of this group was performed to identify those patients who were treated with increased antimetabolite therapy only and who demonstrated significant decrease in DSA to SFI Results Of the 114 patients with clinically significant DSA, 17 (15%) were noted to have significantly decreased SFI with intensified medical therapy only. MPA AUC was assessed after change in medical regimen. Mean values and ranges are summarized: Decrease in DSA levels. Initial Posttreatment SFI 154,684 (99,679–402,188) 29,022 (0–51,644) Creatinine (mg/dl) 1.80 (0.9–4.0) 1.80 (1.0–6.0) MPA AUC – 80.5 (30–133) MMF Dosage (mg/d) 1125 (500–3000) 1700 (1000–3000) Analysis of HLA loci demonstrated DSA at A(2 patients), DR (3 patients) and DQ (12 patients). Conclusions Increased medical immunosuppressive therapy alone may be sufficient to decrease clinically significant DSA levels and to ensure stable renal function. This is often achieved through early detection of DSA formation and optimization of antimetabolite dosage using MPA AUC measurement. Class II HLA loci are most frequently associated with DSA formation, particularly at the DQ locus, suggesting that more frequent monitoring of patients with known Class II mismatches may be beneficial. H.C. Yang: Speaker’s Bureau; Company/Organization; Novartis.

Prognostic implications of creatine kinase–MB measurements in ST-segment elevation myocardial infarction patients treated with primary percutaneous coronary intervention

Peak creatine kinase (CK)-MB concentration is related to reperfusion success and clinical outcomes after fibrinolytic therapy for acute myocardial infarction. However, prognostic implications of CK-MB measurements after primary percutaneous coronary intervention (PCI), which provides more predictable and consistent reperfusion, are unknown. We pooled 2,042 primary PCI-treated ST-segment elevation myocardial infarction (STEMI) patients from 3 trials with serial core laboratory-determined CK-MB measurements; 1,799 patients (88.1%) who survived to 36 hours and had ≥4 CK-MB measurements were studied. Cox regression modeling was performed to quantify the association between peak CK-MB concentration (and area under the time-concentration curve [AUC]) and mortality at 6 months, and death or congestive heart failure at 90 days. The median (25th-75th percentiles) peak CK-MB concentration and AUC measurement through 36 hours were 239 (109-429) ng/mL and 4,263 (2,081-7,124) ng/(mL h), respectively. By multivariable analysis, peak CK-MB concentration and AUC measurement were independently associated with 6-month mortality (adjusted hazard ratio [HR] 1.15, 95% CI 1.05-1.25, per 100-ng/mL increase, P = .002; and adjusted HR 1.09, 95% CI 1.03-1.14, per 1,000-ng/[mL h] increase, P < .001, respectively) and 90-day death or congestive heart failure (adjusted HR 1.26, 95% CI 1.18-1.34, P < .001; and adjusted HR 1.15, 95% CI 1.11-1.19, P < .001, respectively). Peak CK-MB concentration and AUC measurement are independent predictors of 3- to 6-month cardiovascular outcomes in primary PCI-treated STEMI patients. Our findings guide application of these measurements as efficacy end points in early-phase studies evaluating new therapies for STEMI.

Modeling 5-FU AUC-dose relationship to develop a PK dosing algorithm.

547 Background: 5-fluorouracil (5-FU) is an ideal candidate for pharmacokinetic (PK) dosing. 5-FU is administered based on standard BSA dosing. BSA administration results in highly variable PK and exposure, and potential underdosing in about 50% of patients. Patients with higher 5-FU AUC levels derive greater clinical benefit but with increased risk of toxicity. An immunoassay (OnDose) that measures 5-FU concentration in plasma and reports a computed AUC in mg·h/L has been developed to aid optimization of therapy using PK dosing. The results of an analysis to model the 5-FU AUC-dose change relationship from our clinical database are presented. Methods: A set of 585 sequential patients from our database receiving FOLFOX6 regimen with 5-FU continuous infusion (CI) over 41-48 hours, with or without bevacizumab, for the adjuvant and metastatic treatment of CRC was used for the preliminary analyses. A subset restricted to include only patients that had at least two consecutive cycles tested, received 1,600-3,600 mg/m2 at the initial test cycle, and had blood sample collection times ≥ 18 hours was used to conduct regression modeling of change in AUC versus change in dose. Outliers such as AUC values &lt; 5 mg·h/L and &gt; 50 mg·h/L were excluded from the analysis. Results: Our analysis corroborates previous findings that large variability in 5-FU PK measurements exists during the early part of the infusion cycle. The following simple regression model with an R2=0.51 developed over N=307 cycle-pair observations characterizes the AUC-dose relationship as: change in AUC (mg·h/L) = 0.02063 * dose change (mg/m2). Conclusions: The above model suggests that dose changes ranging from 145 to 720 mg/m2 would be sufficient to adjust the AUC to a potential therapeutic threshold of &gt;20 mg·h/L for most patients. Assuming a 2,400 mg/m2 baseline dose, a maximum 720 mg/m2 dose increase would correspond to a 30% increase, but would likely be needed only by 10% of the patient population with low AUC readings. Our data from the US clinical setting indicate that it is possible to optimize 5-FU therapy by adjusting the dose based on AUC measurement. [Table: see text] [Table: see text]

Availability of Drug Information on Bioequivalence of Generic Products -Findings of Graduate Interns at a University Pharmacy-

Access to drug information (DI) needed to evaluate generic product bioequivalence was studied to identify problems with the current status of DI availability and encourage proper use of DI. Ten items were chosen from among the stock of branded products at the University Pharmacy, and five corresponding generics were selected for each item. Conditions of access to information on pharmacokinetic tests and dissolution tests were rated and the assigned ratings compared. In the case of pharmacokinetic parameters obtainable from makers of generic drugs, we also performed Welch's t-test to compare the difference between values reported for branded and generic products. From the standpoint of individual tests, the pharmacokinetic tests yielded higher scores on the whole than did the dissolution tests, and low scores were obtained for the half-life of blood drug concentration (T1/2). We observed a tendency for the adequacy of information to depend more upon the drug item itself than upon the nature of the test. The percentage of tests allowing for comparison with branded products varied from 0%-75% (average 49%). Parameter by parameter, the range of variation was from 35% of Tmax to 63% of Cmax. Factors precluding comparison included insufficient data on branded products, mismatch in assayed chemical species between branded and generic, mismatch between final sampling time in AUC(t) measurement, dosage inconsistency, and insufficient data on generic products. DI should be provided in a manner that facilitates comparison of information supplied by generic drug makers with data released by makers of branded products.

Evaluation of AUC&lt;sub&gt;0—4&lt;/sub&gt; Predictive Methods for Cyclosporine in Kidney Transplant Patients

Cyclosporine (CyA) is the most commonly used immunosuppressive agent in patients who undergo kidney transplantation. Dosage adjustment of CyA is usually based on trough levels. Recently, trough levels have been replacing the area under the concentration-time curve during the first 4 h after CyA administration (AUC(0-4)). The aim of this study was to compare the predictive values obtained using three different methods of AUC(0-4) monitoring. AUC(0-4) was calculated from 0 to 4 h in early and stable renal transplant patients using the trapezoidal rule. The predicted AUC(0-4) was calculated using three different methods: the multiple regression equation reported by Uchida et al.; Bayesian estimation for modified population pharmacokinetic parameters reported by Yoshida et al.; and modified population pharmacokinetic parameters reported by Cremers et al. The predicted AUC(0-4) was assessed on the basis of predictive bias, precision, and correlation coefficient. The predicted AUC(0-4) values obtained using three methods through measurement of three blood samples showed small differences in predictive bias, precision, and correlation coefficient. In the prediction of AUC(0-4) measurement of one blood sample from stable renal transplant patients, the performance of the regression equation reported by Uchida depended on sampling time. On the other hand, the performance of Bayesian estimation with modified pharmacokinetic parameters reported by Yoshida through measurement of one blood sample, which is not dependent on sampling time, showed a small difference in the correlation coefficient. The prediction of AUC(0-4) using a regression equation required accurate sampling time. In this study, the prediction of AUC(0-4) using Bayesian estimation did not require accurate sampling time in the AUC(0-4) monitoring of CyA. Thus Bayesian estimation is assumed to be clinically useful in the dosage adjustment of CyA.

Long-term changes in mycophenolic acid exposure in combination with tacrolimus and corticosteroids are dose dependent and not reflected by trough plasma concentration: a prospective study in 100 de novo renal allograft recipients.

Tacrolimus and cyclosporine A have different effects on exposure to concomitantly administered mycophenolate mofetil (MMF), measured as the mycophenolic acid (MPA) dose interval area under the plasma concentration versus time curve (AUC0-12 h) or the plasma MPA predose concentration (C0). This has led to recommendations in using a 50% lower dose of MMF in combination with tacrolimus compared to cyclosporin A. At present, no long-term data are available regarding the pharmacokinetics (PK) of different dosages of MMF in combination with tacrolimus and the clinical variables that influence the dose-exposure relationship of MPA. A prospective 12-month pharmacokinetic study was performed in 100 de novo renal transplant recipients treated with two different MMF dosages (1 g/day vs. 2 g/day) in combination with tacrolimus and corticosteroids. MPA AUC data were collected 7 days, 6 weeks, and 3 and 12 months posttransplantation, and model-independent PK parameters were calculated. Clinical variables that could possibly influence MPA PK were evaluated. The MPA AUC0-12 h significantly increased toward 6 weeks (p < 0.05) but only in the 2-g MMF dosing group. The MPA AUC0-12 h in the 1-g MMF group reached its nadir at 3 months, while in the 2-g MMF group, it remained elevated until 3 months, returning to baseline values by 12 months. This differential evolution in exposure was not only inadequately reflected by the corresponding MPA C0 concentrations, but the MPA C0 concentrations also were not significantly different between the two dosing groups at early postgrafting (day 7) and at 12 months. Using multiple stepwise regression analysis, C0 (r = 0.51, p < 0.0001) and end-of-dose interval MPA plasma concentration (C12: r2 = 0.61, p < 0.0001) were found to poorly predict MPA AUC0-12 h, while MPA plasma concentrations at 4 hours (C4: r2 = 0.85, p < 0.0001) and 6 hours postdosing (C6: r2 = 0.83, p < 0.0001) were superior but hampered by a large prediction bias and imprecision. An abbreviated 2-hour AUC measurement (r2 = 0.78), using three sampling points (C0, C40 [MPA plasma concentration 40 min postdosing], C2), provided the best compromise between a monitoring tool that is theoretically ideal and practically feasible. MPA pharmacokinetics were not influenced by recipient age, gender, and body weight or by serum albumin concentrations, allograft function, or corticosteroid or tacrolimus dose. Mild hepatic dysfunction early after grafting did result in significantly reduced MPA exposure (MPA AUC0-12 h, p = 0.01 and C0, p = 0.03). In this study, it was demonstrated for the first time that the dynamics of long-term MPA pharmacokinetics in combination with tacrolimus differ according to the daily MMF dose and that this effect is not adequately reflected by MPA trough concentrations. Using the latter as a routine measure for therapeutic drug monitoring might mislead clinicians into drawing wrong conclusions in terms of relating questions of efficacy or toxicity to MPA exposure.

Evaluation of prediction of tacrolimus area under the curve by trough concentrations

The area under the plasma concentration-time curve (AUC) correlates best with clinical outcomes. However conventional methods for measuring AUC require many blood samples. Therefore AUC measurement must be simplified using a limited sampling strategy. Traditionally trough blood concentrations are measured routinely for approximation of the AUC. The question is: are trough concentrations optimal for prediction of tacrolimus AUC? This was a preliminary study to evaluate traditional trough concentration for prediction of tacrolimus AUC. Blood concentration curves over 12 h of eight stable renal transplant patients were determined. Mean (range) patient characteristics were: age 50.9 (32.7–67.8) years, gender male/female 7/1, time after transplantation 209 (151–864) days, medication tacrolimus and prednisolone (n = 8), mycophenolate (n = 1). Tacrolimus concentrations were measured with MEIA immunoassay (Abbott). Blood samples were collected just before, 0.25, 0.5, 0.75, 1, 2, 3, 4, 5, 7.5, 9 and 12 h after tacrolimus administration. Predictability of the time point measurements was calculated with multiple (stepwise) regression analysis (SPSS 8.0). Prediction algorithms, found with multiple stepwise regression, are demonstrated in Table 1. Table 1 The sample times with the best predictive algorithms were respectively C5, the combination of C5 and C2 and the combination of C5 and C2 and C1. The predictability of the algorithms based on trough levels, C0 and C12, was much lower. Predictability (r2) of the other isolated sampling time points was respectively, C0.25, 0.879; C0.5, 0.007; C0.75, 0.006; C1, 0.502; C2, 0.801; C3, 0.806; C4, 0.863; C7.5, 0.868; C9, 0.879. As can be seen the predictive power of the trough concentrations was lowest of all sampling time points with exception of C0.5, C0.75 and C1. Measuring trough concentrations has several important advantages such as experience with interpretation and logistic simplicity. However better predictability can be obtained with other sample time points, especially C5. Moreover analytical error is smaller at higher concentrations. For prediction of AUC the gain of addition of more time points to C5 seems small. In comparison with most other sampling times predictability of trough concentrations is poor. More extensive investigations are necessary to determine the optimal sampling strategy.

The area under pH curve: A single-figure parameter representative of esophageal acid exposure

This investigation aims at providing simpler methods for reading esophageal extended pH-metering tracings. Because the currently used parameters only quantify frequency and duration of acid exposure, we have measured the area between the pH curve and the pH 4 line (area under curve [AUC]) in an attempt to include in the evaluation the severity of acid exposure as well. We compared 20 control and 63 children with gastroesophageal reflux (GER). Extended pH metering curves were read according to currently accepted methods. In addition, we measured planimetrically total 24-hour AUC, AUC during daytime, AUC during sleep time, and total AUC excluding 2-hour postprandial periods. In spite of the very significant differences found between means of controls and GER patients for all variables considered, values overlapped widely. Receiver-operating characteristic (ROC) analysis indicated the best threshold values for differentiation of controls and GER patients and tested the diagnostic efficiency for each variable. We found that 24-hour AUC was the best comprehensive value (sensitivity = 1, specificity = 1) and that there was probably no use for fractional evaluation during selected periods of time. The optimum 24-hour AUC threshold value in the present conditions of measurement (paper speed, 0.25 cm/min; 1 pH unit, 2.5 cm) was 20 cm 2. In order to facilitate comparison we propose using a pH-surface unit (pHSU) equal to recording speed (in cm/min) multiplied by vertical shift per 1 pH unit (in centimeters) (in this case, 0.25 · 2.5 = 0.625 cm 2). A threshold value of 20 cm 2 thus becomes 32 pHSU. A similar transformation will be adequate for any other recording conditions. We suggest that AUC measurement should be incorporated into the current pH-metering software.

Correlation between free platinum AUC and total platinum measurement 24 h after i.v. bolus injection of cisplatin in humans.

The plasma kinetics of platinum after i.v. bolus administration of cisplatin was determined for 17 patients with advanced cancer. Statistical analysis of individual values revealed a high correlation between the area under the plasma concentration-time curve (AUC) of free platinum (unbound to proteins) and the concentration of platinum bound to plasma proteins 24 h after drug administration (Cp24). A similar correlation was found between the peak plasma values of ultrafiltrable platinum (Cp0) and Cp24. When studied in the same patient, increases in free platinum AUC and Cp0 were also found to result in increased Cp24. It is suggested that a single measurement of plasma platinum concentration 24 h after i.v. infusion of cisplatin could be a simple method either of detecting patients with extreme values of AUC and Cp0 or of studying the evolution of these parameters during multiple courses of treatment, although it cannot be used to give precise values for AUC and Cp0.