Statistical learning in preclinical drug proarrhythmic assessment.

Torsades de pointes (TdP) is an irregular heart rhythm characterized by faster beat rates and potentially could lead to sudden cardiac death. Much effort has been invested in understanding the drug-induced TdP in preclinical studies. However, a comprehensive statistical learning framework that can accurately predict the drug-induced TdP risk from preclinical data is still lacking. We proposed ordinal logistic regression and ordinal random forest models to predict low-, intermediate-, and high-risk drugs based on datasets generated from two experimental protocols. Leave-one-drug-out cross-validation, stratified bootstrap, and permutation predictor importance were applied to estimate and interpret the model performance under uncertainty. The potential outlier drugs identified by our models are consistent with their descriptions in the literature. Our method is accurate, interpretable, and thus useable as supplemental evidence in the drug safety assessment.

Repolarization studies using human stem cell-derived cardiomyocytes: Validation studies and best practice recommendations.

Human stem cell-derived cardiomyocytes (hSC-CMs) hold great promise as in vitro models to study the electrophysiological effects of novel drug candidates on human ventricular repolarization. Two recent large validation studies have demonstrated the ability of hSC-CMs to detect drug-induced delayed repolarization and "cellrhythmias" (interrupted repolarization or irregular spontaneous beating of myocytes) linked to Torsade-de-Pointes proarrhythmic risk. These (and other) studies have also revealed variability of electrophysiological responses attributable to differences in experimental approaches and experimenter, protocols, technology platforms used, and pharmacologic sensitivity of different human-derived models. Thus, when evaluating drug-induced repolarization effects, there is a need to consider 1) the advantages and disadvantages of different approaches, 2) the need for robust functional characterization of hSC-CM preparations to define "fit for purpose" applications, and 3) adopting standardized best practices to guide future studies with evolving hSC-CM preparations. Examples provided and suggested best practices are instructional in defining consistent, reproducible, and interpretable "fit for purpose" hSC-CM-based applications. Implementation of best practices should enhance the clinical translation of hSC-CM-based cell and tissue preparations in drug safety evaluations and support their growing role in regulatory filings.

Assessment of Proarrhythmic Potential of Drugs in Optogenetically Paced Induced Pluripotent Stem Cell-Derived Cardiomyocytes.

Cardiac side-effects are one of the major reasons for failure of drugs during preclinical development. Induced pluripotent stem cell-derived cardiomyocytes (iPSC-CMs) have been proposed as a model for predicting drug-induced arrhythmias under the Comprehensive in vitro Proarrhythmia Assay (CiPA) paradigm. Field potential duration (FPD) in spontaneously beating iPSC-CMs is commonly corrected for beating rate using formulas originally derived from the clinical QT-RR relationship that have not been thoroughly validated for use with iPSC-CMs. In this study, channelrhodopsin-2 was expressed in iPSC-CMs allowing for recordings in both spontaneously beating and optically paced (0.8, 1, and 1.5 Hz pacing rate) iPSC-CMs using a microelectrode array system (Maestro, Axion Biosystems). After optimizing the intensity (>1 mW/mm2), duration (15 ms) and frequency of the stimulating light pulses, we recorded iPSC-CMs' responses to 28 blinded CiPA compounds with clinically characterized risk of causing ventricular arrhythmia (Torsade de Pointes or TdP). Drug-induced FPD prolongation data along with drug-induced arrhythmia-like events were used to build a logistic regression model, separating high or intermediate TdP risk drugs from low-or-no TdP risk drugs. The area under the receiver operator characteristic curve for drug TdP risk prediction was identical for spontaneously beating and 0.8 Hz-paced iPSC-CMs (AUC = 0.96; 95% CI [0.9, 1]), while it was slightly lower for 1 and 1.5 Hz pacing (AUC = 0.88; 95% CI [0.76, 1] and 0.93; 95% CI [0.84, 1], respectively). In this study, optical pacing did not offer substantial improvement in proarrhythmic risk prediction when compared with nonpaced iPSC-CMs in the sample of 28 drugs.

Assay sensitivity in "Hybrid thorough QT/QTc (TQT)" study.

A concurrent positive control should be included in a thorough QTc clinical trial to validate the study according to ICH E14 guidance. Some pharmaceutical companies have started to use "hybrid TQT" study to meet ICH E14 regulatory requirements since the release of ICH E14 Q&A (R3). The "hybrid TQT" study includes the same treatment arms (therapeutic and/or supratherapeutic dose of investigational drug, placebo, and positive control) with sample size less than traditional TQT studies, but use concentration-QTc (C-QTc) analysis as primary analysis and assay sensitivity analysis. To better understand the statistical characteristics of assay sensitivity with a commonly used positive control - Moxifloxacin - in "hybrid TQT" studies, we examined the original and subsampled moxifloxacin and placebo data from more than a hundred of TQT studies submitted to FDA. The assay sensitivity results are quite consistent between classical E14 analysis and C-QTc analysis using the original datasets. Performance of assay sensitivity in "hybrid TQT" studies using subsampled data depends on number of moxifloxacin subjects, study design (crossover design and parallel design), and C-QTc model. The results presented here can aid the design of future "hybrid TQT" studies.

International Multisite Study of Human-Induced Pluripotent Stem Cell-Derived Cardiomyocytes for Drug Proarrhythmic Potential Assessment.

To assess the utility of human-induced pluripotent stem cell-derived cardiomyocytes (hiPSC-CMs) as an in vitro proarrhythmia model, we evaluated the concentration dependence and sources of variability of electrophysiologic responses to 28 drugs linked to low, intermediate, and high torsades de pointes (TdP) risk categories using two commercial cell lines and standardized protocols in a blinded multisite study using multielectrode array or voltage-sensing optical approaches. Logistical and ordinal linear regression models were constructed using drug responses as predictors and TdP risk categories as outcomes. Three of seven predictors (drug-induced arrhythmia-like events and prolongation of repolarization at either maximum tested or maximal clinical exposures) categorized drugs with reasonable accuracy (area under the curve values of receiver operator curves ∼0.8). hiPSC-CM line, test site, and platform had minimal influence on drug categorization. These results demonstrate the utility of hiPSC-CMs to detect drug-induced proarrhythmic effects as part of the evolving Comprehensive In Vitro Proarrhythmia Assay paradigm.

Evaluation of dependent variable, time effect, covariates, and covariation structure in concentration-QTc modeling: A simulation study.

The revised ICH E14 Question and Answer (R3) document issued in December 2015 enables pharmaceutical companies to use concentration-QTc (C-QTc) modeling as the primary analysis for assessing QTc prolongation risk of new drugs. A new approach by including the time effect into the current C-QTc model is introduced. Through a simulation study, we evaluated performances of different C-QTc modeling with different dependent variables, covariates, and covariance structures. This simulation study shows that C-QTc models with ΔQTc being dependent variable without time effect inflate false negative rate and that fitting C-QTc models with different dependent variables, covariates, and covariance structures impacts the control of false negative and false positive rates. Appropriate C-QTc modeling strategies with good control of false negative rate and false positive rate are recommended.

Cross-Site Reliability of Human Induced Pluripotent stem cell-derived Cardiomyocyte Based Safety Assays Using Microelectrode Arrays: Results from a Blinded CiPA Pilot Study.

Recent in vitro cardiac safety studies demonstrate the ability of human induced pluripotent stem cell-derived cardiomyocytes (hiPSC-CMs) to detect electrophysiologic effects of drugs. However, variability contributed by unique approaches, procedures, cell lines, and reagents across laboratories makes comparisons of results difficult, leading to uncertainty about the role of hiPSC-CMs in defining proarrhythmic risk in drug discovery and regulatory submissions. A blinded pilot study was conducted to evaluate the electrophysiologic effects of 8 well-characterized drugs on 4 cardiomyocyte lines using a standardized protocol across 3 microelectrode array platforms (18 individual studies). Drugs were selected to define assay sensitivity of prominent repolarizing currents (E-4031 for IKr, JNJ303 for IKs) and depolarizing currents (nifedipine for ICaL, mexiletine for INa) as well as drugs affecting multichannel block (flecainide, moxifloxacin, quinidine, and ranolazine). Inclusion criteria for final analysis was based on demonstrated sensitivity to IKr block (20% prolongation with E-4031) and L-type calcium current block (20% shortening with nifedipine). Despite differences in baseline characteristics across cardiomyocyte lines, multiple sites, and instrument platforms, 10 of 18 studies demonstrated adequate sensitivity to IKr block with E-4031 and ICaL block with nifedipine for inclusion in the final analysis. Concentration-dependent effects on repolarization were observed with this qualified data set consistent with known ionic mechanisms of single and multichannel blocking drugs. hiPSC-CMs can detect repolarization effects elicited by single and multichannel blocking drugs after defining pharmacologic sensitivity to IKr and ICaL block, supporting further validation efforts using hiPSC-CMs for cardiac safety studies.

Correction to: Scientific white paper on concentration-QTc modeling.

The original version of this article unfortunately contained an error in Equation 1 under the section "Pre-specified linear mixed effects model". The correct equation has given below.

Scientific white paper on concentration-QTc modeling.

The International Council for Harmonisation revised the E14 guideline through the questions and answers process to allow concentration-QTc (C-QTc) modeling to be used as the primary analysis for assessing the QTc interval prolongation risk of new drugs. A well-designed and conducted QTc assessment based on C-QTc modeling in early phase 1 studies can be an alternative approach to a thorough QT study for some drugs to reliably exclude clinically relevant QTc effects. This white paper provides recommendations on how to plan and conduct a definitive QTc assessment of a drug using C-QTc modeling in early phase clinical pharmacology and thorough QT studies. Topics included are: important study design features in a phase 1 study; modeling objectives and approach; exploratory plots; the pre-specified linear mixed effects model; general principles for model development and evaluation; and expectations for modeling analysis plans and reports. The recommendations are based on current best modeling practices, scientific literature and personal experiences of the authors. These recommendations are expected to evolve as their implementation during drug development provides additional data and with advances in analytical methodology.

Baseline correction in parallel thorough QT studies.

BACKGROUND: In parallel thorough QT (TQT) studies, it has been speculated that either baseline correction should be omitted, under the assumption that it only adds noise to the data, or a time-averaged baseline instead of a time-matched baseline correction should be considered in order to reduce the study variability. OBJECTIVE: This study characterized the assumptions and implications of different baseline correction approaches in parallel TQT studies submitted for regulatory review. DATA AND METHODS: 57 parallel TQT studies conducted between 2002 and 2009 in 5591 healthy volunteers were evaluated. Only moxifloxacin and placebo arms, including their baselines, were considered. The options of using no baseline correction, time-averaged baseline correction, and time-matched baseline correction were examined and compared. RESULTS: QTc values exhibited a diurnal pattern, with longer QTc intervals during sleep preserved when correcting for a time-averaged baseline. Post-dose and baseline QTc values were highly correlated (mean ρ = 0.80, range 0.56-0.98 and mean ρ = 0.79, range 0.50-0.96 in the placebo and moxifloxacin groups, respectively). The variability of raw QTc values was substantially larger than that of baseline-adjusted QTc values. The difference in the point estimate of QTc differences between moxifloxacin and placebo differed by up to ± 4 ms between the time-averaged and the time-matched baseline corrections. Statistical tests indicate that assumptions of time-averaged baseline and no baseline correction are not appropriate. CONCLUSIONS: Baseline correction in parallel TQT studies leads to more precise QTc estimates. Because of possible inaccuracy introduced by time-averaged baseline correction, the time-matched baseline correction appears to be preferable for a parallel TQT study to both reduce the intrinsic variability due to circadian patterns and obtain more accurate point estimates.

Validation of QT Interval Correction Methods When a Drug Changes Heart Rate.

The QT interval is correlated with heart rate; therefore, the QT interval is usually corrected by heart rate when drug-induced QT effect is studied. Currently, there are many correction methods that use either fixed or data-driven approaches. The effectiveness of correction methods depends on many factors and varies from study to study. Statistical validation and comparisons need to be performed to determine the most appropriate correction method for each study. We examined different validation methods and explored a new approach to use when the testing drug changes heart rate.

Statistical characteristics of moxifloxacin-induced QTc effect.

Moxifloxacin has been the most commonly used positive control in "thorough" QTc (TQT) studies. In a TQT study, the assay sensitivity is often considered to be established if the baseline corrected mean difference in QTc between moxifloxacin and placebo is greater than 5 ms in common practice at one or more prespecified time points and the observed moxifloxacin induced QTc effect over time follows the proper pharmacokinetics profile. To better understand the statistical characteristics of moxifloxacin-induced QTc prolongation and to provide guidance for future studies, 20 TQT studies that involved moxifloxacin have been evaluated. We study the QTc profile of the baseline adjusted mean difference in QTc between moxifloxacin and placebo over time. Zhang (2008) proposed that the moxifloxacin induced QTc effect can be evaluated between 1 and 4 h after a single dose (400 mg) administration near the time (T(max)) of peak concentration instead of all time points (typically 9-12 time points) at which QT was measured for the study drug evaluation. After evaluating 20 TQT studies, we confirm that the maximum moxifloxacin effect occurs in the time window between 1 and 4 h post dose. We also investigate the variability of the data as well as correlations between time points and between regimens. These findings and results can be used as a reference for future studies.

Comparison of sleep/wake behavior in CKD stages 4 to 5 and hemodialysis populations using wrist actigraphy.

BACKGROUND: Patients with kidney failure have more sleep symptoms than the general population, but the contribution to sleep symptoms of kidney failure versus its treatment with thrice-weekly hemodialysis has been unclear. We assessed the influence of hemodialysis on sleep/wake behavior by using wrist actigraphy and self-reported sleep quality compared with patients with chronic kidney disease (CKD) stages 4 to 5. STUDY DESIGN: Cross-sectional study. SETTING & PARTICIPANTS: Thirty-six patients with CKD stages 4 to 5 and 51 hemodialysis (HD) patients wore wrist actigraphs and completed sleep diaries for 2 weeks. PREDICTORS: Thrice-weekly HD versus CKD stages 4 to 5, unstable total sleep times (TSTs), early HD shift. OUTCOMES: Self-reported sleep quality and objective measures of sleep/wake behavior. MEASUREMENTS: Diaries, sleep questionnaires, and wrist actigraphy were performed. RESULTS: The group with CKD stages 4 to 5 had an average age of 51 years, 69% were men, 19% were African American, and average body mass index was 28.9 kg/m2. The HD group had an average age of 54 years, 60.8% were men, 49% were African American, and average body mass index was 27.5 kg/m2. Average TST was 66.8 minutes shorter and sleep efficiency was 5.2% lower in the HD group compared with the population with CKD stages 4 to 5. In the HD population, 28 individuals had a mean change in TST greater than 60 minutes between HD and non-HD nights, and this unstable sleep pattern was associated with daytime sleepiness. The early-HD group had TST 62 minutes (95% confidence interval, approximately 22 to 102) shorter than those with later HD shifts. No significant differences in sleep efficiency or fragmentation index were found between the early- and late-HD groups. LIMITATIONS: Study included those older than 18 years. CONCLUSIONS: Both patients with CKD stages 4 to 5 and HD patients have short and fragmented sleep. An early-morning HD shift was associated with shorter TST and greater variation in nightly TST. Additional trials of the possible beneficial effect of behavioral sleep interventions, more frequent HD, and later HD shifts on sleep patterns are needed.

Pulmonary embolism incidence is increasing with use of spiral computed tomography.

BACKGROUND: Pulmonary embolism causes significant morbidity in hospitalized patients, yet few studies have explored the impact of spiral computed tomography (CT) scanning on diagnosis and clinical outcome. METHODS: Incidence rates of pulmonary embolism, chest and spiral CT rates, D-dimer assay, anticoagulation, and in-hospital mortality were assessed on statewide pulmonary embolism discharge data (1997-2001) from the Pennsylvania Health Care Cost Containment Council. RESULTS: The incidence of pulmonary embolism increased from 47 to 63 per 100,000 patients from 1997 to 2001 (mean of 0.004% per year, P < .001). Mean pulmonary embolism incidence rates were higher for African American patients (0.031% per year higher than for white patients), patients aged 70 years or more (0.007% higher than for patients aged<70 years), and female patients (0.013% higher than for male patients) (all P < .001). Concomitantly, the proportion undergoing CT (including spiral) scans increased from 23.23% to 45.18% (odds ratio=1.30; P<.001), controlling for age, gender, race, and cancer, whereas rates for other procedures remained unchanged. By comparing 1999 and before with 2000 and after, there was a significant decrease in the 2 highest Atlas Severity of Illness categories (49.4%-37.7%) and a significant increase in the 3 lowest categories (50.6%-62.3%; P < .001). The risk of in-hospital deaths among patients with pulmonary embolism decreased in this period from 12.8% to 11.1% (P < .001). CONCLUSION: The incidence of pulmonary embolism is increasing with the increasing use of spiral CT scans, with a lower severity of illness and lower mortality, suggesting the increase is due to earlier diagnosis.

Sample size and power calculations based on generalized linear mixed models with correlated binary outcomes.

The generalized linear mixed model (GLIMMIX) provides a powerful technique to model correlated outcomes with different types of distributions. The model can now be easily implemented with SAS PROC GLIMMIX in version 9.1. For binary outcomes, linearization methods of penalized quasi-likelihood (PQL) or marginal quasi-likelihood (MQL) provide relatively accurate variance estimates for fixed effects. Using GLIMMIX based on these linearization methods, we derived formulas for power and sample size calculations for longitudinal designs with attrition over time. We found that the power and sample size estimates depend on the within-subject correlation and the size of random effects. In this article, we present tables of minimum sample sizes commonly used to test hypotheses for longitudinal studies. A simulation study was used to compare the results. We also provide a Web link to the SAS macro that we developed to compute power and sample sizes for correlated binary outcomes.

The 8-item Short-Form Health Survey and the physical comfort composite score of the quality of recovery 40-item scale provide the most responsive assessments of pain, physical function, and mental function during the first 4 days after ambulatory knee surgery with regional anesthesia.

BACKGROUND: We evaluated the validity and responsiveness of three instruments: the numeric rating scale (NRS) pain score, the 8-item Short-Form Health Survey (SF-8), and the 40-item Quality of Recovery from Anesthesia (QoR) Survey in 154 outpatients undergoing anterior cruciate ligament reconstruction (ACLR). The objective was to provide a robust psychometric basis for outcome survey selection for surgical outpatients undergoing regional anesthesia without general anesthesia. METHODS: Patients undergoing ACLR with a standardized spinal anesthesia plan were randomized to receive a perineural catheter with either placebo injection-infusion, or injection-infusion with levobupivacaine. Patients completed the NRS, SF-8, and QoR instruments for four postoperative days to evaluate pain, physical function, and mental function. RESULTS: Regarding pain, neither the NRS nor the QoR offered advantages over the SF-8. Regarding physical function, the QoR physical independence composite offered no advantage over the SF-8 physical component summary. The QoR physical comfort composite assessed short-term changes in treatment-related side effects, and thus provided information not covered by the SF-8. Regarding mental function, the SF-8 mental component summary and QoR emotional state composite showed little change over the four days, although the latter measure showed higher responsiveness to change. CONCLUSIONS: For ACLR outpatients receiving regional anesthesia, the SF-8 is sufficient to assess postoperative pain and physical function. Adding the QoR physical comfort composite will help assess short-term side effects.

Daily smoking patterns, their determinants, and implications for quitting.

In this article, the authors examine daily temporal patterns of smoking in relation to environmental restrictions on smoking and cessation outcomes. Time-series methods were used for analyzing cycles in 351 smokers who monitored their smoking in real time for 2 weeks. The waking day was divided into 8 "bins" of approximately 2 hr, cigarette counts were tallied for each bin, and temporal patterns of smoking and restriction were analyzed. Cluster analyses of smoking patterns by time of day resulted in 4 clusters: daily decline (n = 30; 9%), morning high (n = 43; 12%), flatline (n = 247; 70%), and daily dip-evening incline (n = 31; 9%). Clusters differed in baseline demographic, smoking, and psychosocial variables. Results suggest that smoking behavior can be characterized by regular patterns of smoking frequency during the waking day: Smoking in the flatline cluster was within +/-0.5 standard deviation at all times. For the other clusters, smoking was high in the morning (daily dip-evening incline: +1.7 standard deviations; morning high: +2.8 standard deviations; daily decline: +1.7 standard deviations); moderate (morning high: -0.8 standard deviations; daily decline: +0.3 standard deviations) or low (daily dip-evening incline: -1.0 standard deviations) midday; and high (daily dip-evening incline: +2.0 standard deviations), moderate (morning high: +0.5 standard deviations), or low (daily decline: -1.5 standard deviations) in the evening. Daily smoking patterns were related to environmental smoking restrictions, but the strength of this relationship differed among clusters and by time of day. Clusters differed in lapse risk.

Religious beliefs and practices are associated with better mental health in family caregivers of patients with dementia: findings from the REACH study.

OBJECTIVE: Providing care to a loved one with dementia and the death of that loved one are generally considered two of the most stressful human experiences. Each puts family caregivers at risk of psychologic morbidity. Although research has suggested that religious beliefs and practices are associated with better mental health, little is known about whether religion is associated with better mental health in family caregivers. Our objective, then, is to explore the relationship between religion and mental health in active and bereaved dementia caregivers. METHODS: A total of 1,229 caregivers of persons with moderate to severe dementia were recruited from six geographically diverse sites in the United States and followed prospectively for up to 18 months. Three measures of religion: 1) the frequency of attendance at religious services, meetings, and/or activities; 2) the frequency of prayer or meditation; and 3) the importance of religious faith/spirituality were collected. Mental health outcomes were caregiver depression (Center for Epidemiological Studies-Depression [CES-D] scale) and complicated grief (Inventory of Complicated Grief [ICG]). RESULTS: Religious beliefs and practices were important to the majority of caregivers. After controlling for significant covariates, the three measures of religion were associated with less depressive symptoms in current caregivers. Frequent attendance was also associated with less depression and complicated grief in the bereaved. CONCLUSIONS: Religious beliefs and practices, and religious attendance in particular, are associated with better mental health in family caregivers of persons with dementia.

Using trajectories from a bivariate growth curve as predictors in a Cox regression model.

An important research objective in most psychiatric clinical trials of maintenance treatment is to find predictors of recurrence of illness. In those trials, patients are first admitted into an open treatment period also called acute treatment. If they respond to the treatment and are considered to have stable remission from the illness, they enter the second phase of the trial where they are randomized into different arms of the 'maintenance treatments'. Often, more than one response variable is measured longitudinally in the acute treatment phase to monitor treatment responses. Trajectories of these response measures are believed to have predictive ability for recurrences in the maintenance phase of the trial. By using a bivariate growth curve from two such longitudinal measures, we developed a method to use the estimated trajectories of each subject in a Cox regression model to predict recurrence in the maintenance phase. To adjust for the parameter estimation errors, we applied a full likelihood approach based on the conditional expectations of the predictors. Simulation studies indicate that the estimation error corrected estimators for the Cox model parameters are less biased when compared to the naive regression estimators without accounting for these errors. The uniqueness of this method lies in estimating trajectories from bivariate unequally spaced longitudinal response measures. An illustrative example is provided with data from a maintenance treatment trial for major depression in an elderly population. Visual Fortran 90 programs were developed to implement the algorithm.

Preparedness for the death of a loved one and mental health in bereaved caregivers of patients with dementia: findings from the REACH study.

BACKGROUND: Although it has been suggested that family and friends who are prepared for the death of a loved one have less distress, the relationship between preparedness and bereavement mental health is inconclusive. OBJECTIVES: To determine the relationship between preparedness for the death and mental health in bereaved caregivers of dementia patients and explore predictors of preparedness. DESIGN: A prospective study of family caregivers of persons with dementia. Standardized assessment instruments and structured questions were used to collect data at study entry and at 6, 12, and 18 months. Multiple caregiving-related variables were collected. Bereaved caregivers reported whether they were "not at all" prepared or prepared for the death of their loved one. SUBJECTS: Two hundred twenty-two bereaved caregivers RESULTS: Twenty-three percent of caregivers were not prepared for the death. These caregivers had more depression, anxiety, and complicated grief symptoms. Black caregivers, caregivers with less education, those with less income, and those with more depressive symptoms prior to the death were more likely to perceive themselves as "not at all" prepared. In contrast, the amount of pain the care recipient was in prior to death was positively associated with preparedness. CONCLUSIONS: Despite providing high-intensity care, often for years, many bereaved caregivers perceived themselves as unprepared for the death. These caregivers had more depression, anxiety, and complicated grief symptoms. Future work should be directed to confirming these findings and determining how best to intervene with high-risk caregivers.