Individualizing drug dosage by using a random intercept linear model

The introduction of clozapine at the Nathan Kline Institute in New York and its long-term consequences

Nathan S. Kline was a pioneer in psychopharmacology in the United States (US). In 1952, Kline started a research unit at Rockland State Hospital, New York. Kline brought clozapine from Switzerland since it was not yet available in the US. At Rockland State Hospital, George Simpson had conducted antipsychotic trials and had developed scales to assess movement disorders. In 1974, Simpson published the first US clozapine trial. In 1978, he published on 1) the effect of clozapine on tardive dyskinesia and 2) high plasma clozapine concentrations in two patients with seizures. His experience of clozapine withdrawal symptoms in his first 2 trials led in the future to more articles in this area. In Philadelphia, Simpson designed a double-blind randomized clinical trial (RCT) with 3 doses (100, 300 and 600 mg/day) which was published in 1999. From the 50 patients started on the RCT, 47 provided repeated plasma clozapine concentrations every other week of the RCT. This rich database of plasma clozapine concentrations under controlled conditions has contributed to many of the advances in clozapine pharmacokinetics in the last 5 years including: 1) obesity can be associated with clozapine poor metabolism (PM) status, 2) a clozapine ultrarapid metabolizer (UM) with a minimum therapeutic dose of 1591 mg/day, 3) a case of clozapine intoxication dropped from the RCT due to pneumonia, 4) cases of increased plasma concentrations during clozapine-induced fever, 5) the possibility that African-Americans may need higher clozapine doses than those of European ancestry, and 6) three indices of non-adherence.

Linear Mixed Effect Models for Rehabilitation Research

ABSTRACT

The growing emphasis on evidence-based methods in rehabilitation medicine calls for increase in the sophistication of study design and analytic methods across the discipline. To properly evaluate new treatment options, a physiatrist needs to be able to separate treatment effects from parallel changes that occur over time and variations that may be due to subject demographics. Simple t tests may not be appropriate where observations may vary randomly across different institutions participating in a multicenter trial, or the same rehabilitation course may lead to different outcomes because of various factors. In the analysis of any rehabilitation program, these random variations must be accounted for to receive accurate results. In this short review, we focus in one of the most common approaches that are appropriate to account for these variations, namely, linear mixed effect models.

Measuring Individual Benefits of Medical Treatments Using Longitudinal Hospital Data with Non-Ignorable Missing Responses Caused by Patient Discharge: Application to the Study of Benefits of Pain Management Post Spinal Fusion

Electronic health records (EHR) provide valuable resources for longitudinal studies and understanding risk factors associated with poor clinical outcomes. However, they may not contain complete follow-ups, and the missing data may not be at random since hospital discharge may depend in part on expected but unrecorded clinical outcomes that occur after patient discharge. These non-ignorable missing data requires appropriate analysis methods. Here, we are interested in measuring and analyzing individual treatment benefits of medical treatments in patients recorded in EHR databases. We present a method for predicting individual benefits that handles non-ignorable missingness due to hospital discharge. The longitudinal clinical outcome of interest is modeled simultaneously with the hospital length of stay using a joint mixed-effects model, and individual benefits are predicted through a frequentist approach: the empirical Bayesian approach. We illustrate our approach by assessing individual pain management benefits to patients who underwent spinal fusion surgery. By calculating sample percentiles of empirical Bayes predictors of individual benefits, we examine the evolution of individual benefits over time. We additionally compare these percentiles with percentiles calculated with a Monte Carlo approach. We showed that empirical Bayes predictors of individual benefits do not only allow examining benefits in specific patients but also reflect overall population trends reliably.

Using population crossover trials to improve the decision process regarding treatment individualization in N-of-1 trials

Healthcare researchers are showing renewed interest in the utilization of N-of-1 clinical trials for the individualization of pharmacological treatments. Here, we propose a frequentist approach to conducting treatment individualization in N-of-1 trials that we call "partial empirical Bayes." We infer the most beneficial treatment for the patient from combining the information provided by a previously conducted population crossover trial with individual patient data. We propose a method for estimating an optimal number of treatment cycles and investigate the statistical conditions under which N-of-1 trials are more beneficial than traditional clinical approaches. We represent the patient population with a random-coefficients linear model and calculate estimators of posttreatment individual disease severities. We show the estimators' consistency under the most common N-of-1 designs and examine their prediction errors and performance with small numbers of patient's responses. We demonstrate by simulating new patients that our approach is equivalent or superior to both the common clinical practice of recommending the on-average best treatment for all patients and the common individualization method that simply compares average responses to the tested treatments. We conclude that some situations exist in which individualization with N-of-1 trials is highly beneficial while other situations exist in which individualization may be unfruitful.

European Whites May Need Lower Minimum Therapeutic Clozapine Doses Than Those Customarily Proposed

PURPOSE/BACKGROUND

A nomogram from a British naturalistic study proposed that the clozapine dosing needed to reach a serum concentration of 350 ng/mL ranged from 265 mg/d (female nonsmokers) to 525 mg/d (male smokers). Some European reviews have used these dosing recommendations, which seem greater than what we found in an Italian White sample ranging from 245 mg/d (female nonsmokers) to 299 mg/d (male smokers). Five other published samples of European Whites were added to the Italian sample to estimate clozapine doses recommended for reaching 350 ng/mL.

METHODS/PROCEDURES

Average clozapine metabolizers were obtained by eliminating outliers with confounding variables: (1) psychiatric inducers and inhibitors; (2) doses less than 100 mg/d; and (3) when possible, patients with inflammation, obesity, or using oral contraceptives. The study included 1363 average metabolizer European Whites: the Italian sample and 5 new samples. Mean averages that reached serum concentration levels of 350 ng/mL were calculated after stratification by sex and smoking status in each sample. Then, weighted mean averages were obtained by combining the 6 samples.

FINDINGS/RESULTS

The estimated weighted mean clozapine dosages ranged from 236 to 368 mg/d (236 mg/d in 218 female nonsmokers, 256 mg/d in 340 male nonsmokers, 357 mg/d in 269 female smokers, and 368 mg/d in 546 male smokers).

IMPLICATIONS/CONCLUSIONS

Our recommended dosages are less than those recommended in Europe. Future studies in European Whites need to replicate these recommended doses for average metabolizer patients after sex and smoking stratification and further explore clozapine dosing for those with relevant clinical confounders.

A graphical approach to assess the goodness-of-fit of random-effects linear models when the goal is to measure individual benefits of medical treatments in severely ill patients

Background

Two-dimensional personalized medicine (2-PM) models are tools for measuring individual benefits of medical treatments for chronic diseases which have potential applications in personalized medicine. These models assume normality for the distribution of random effects. It is necessary to examine the appropriateness of this assumption. Here, we propose a graphical approach to assessing the goodness-of-fit of 2-PM models with continuous responses.

Methods

We propose benefit quantile-quantile (BQQ) plots which compare the empirical quantiles of individual benefits from a patient sample predicted through an empirical Bayes (EB) approach versus the quantiles of the theoretical distribution of individual benefits derived from the assumption of normality for the random effects. We examine the performance of the approach by conducting a simulation study that compared 2-PM models with non-normal distributions for the random effects versus models with comparable normal distributions. Cramer-von Mises discrepancies were used to quantify the performance of the approach. The approach was illustrated with data from a clinical trial of imipramine for patients with depression.

Results

Simulations showed that BQQ plots were able to capture deviations from the normality assumption for the random effects and did not show any asymmetric deviations from the y = x line when the random effects were normally distributed. For the depression data, the points of the BQQ plot were scattered around closely to the y = x line, without presenting any asymmetric deviations. This implied the adequacy of the normality assumption for the random effects and the goodness-of-fit of the 2-PM model for the imipramine data.

Conclusion

BQQ plots are sensitive to violations of the normality assumption for the random effects, suggesting that the approach is a useful tool for examining the goodness-of-fit of random-effects linear models when the goal is to measure individual treatment benefits.

Individualizing the dosage of Methylphenidate in children with attention deficit hyperactivity disorder

BACKGROUND

Attention deficit hyperactivity disorder (ADHD) is one of the most common childhood mental health disorders. Stimulant drugs as the most commonly used treatment and first-line therapy for ADHD have side effects. One of the newest approaches to select the best choices and optimize dosages of medications is personalized medicine.

METHODS

This historical cohort study was carried out on the data taken from the period of 2008 to 2015. Eligible subjects were included in the study randomly. We used mixed-effects logistic regression models to personalize the dosage of Methylphenidate (MPH) in ADHD. The patients' heterogeneity was considered using subject-specific random effects, which are treated as the realizations of a stochastic process. To recommend a personalized dosage for a new patient, a two-step procedure was proposed. In the first step, we obtained estimates for population parameters. In the second step, the dosage of the drug for a new patient was updated at each follow-up.

RESULTS

Of the 221 children enrolled in the study, 169 (76.5%) were male and 52 (23.5%) were females. The overall mean age at the beginning of the study is 82.5 (± 26.5) months. In multivariable mixed logit model, three variables (severity of ADHD, time duration receiving MPH, and dosage of MPH) had a significant relationship with improvement. Based on this model the personalized dosage of MPH was obtained.

CONCLUSIONS

To determine the dosage of MPH for a new patient, the more the severity of baseline is, the more of an initial dose is required. To recommend the dose in the next times, first, the estimation of random coefficient should be updated. The optimum dose increased when the severity of ADHD increased. Also, the results show that the optimum dose of MPH as one proceeds through the period of treatment will decreased.

Subgroup identification via homogeneity pursuit for dense longitudinal/spatial data

In the clinical trial community, it is usually not easy to find a treatment that benefits all patients since the reaction to treatment may differ substantially across different patient subgroups. The heterogeneity of treatment effect plays an essential role in personalized medicine. To facilitate the development of tailored therapies and improve the treatment efficacy, it is important to identify subgroups that exhibit different treatment effects. We consider a very general framework for subgroup identification via the homogeneity pursuit methods usually employed in econometric time series analysis. The change point detection algorithm in our procedure is most suitable for analyzing dense longitudinal or spatial data which are quite common for biomedical studies these days. We demonstrate that our proposed method is fast and accurate through extensive numerical studies. In particular, our method is illustrated by analyzing a diffusion tensor imaging data set.

The Effect of Body Weight Changes on Total Plasma Clozapine Concentrations Determined by Applying a Statistical Model to the Data From a Double-Blind Trial

PURPOSE/BACKGROUND

Some therapeutic drug monitoring studies suggest that increased weight is associated with small increases in clozapine concentrations. The goal of this study was to reanalyze a US double-blind study using a sophisticated statistical model to test whether weight gains from baseline or increases in percentage of body fat from baseline, computed from a published equation, are associated with increased total plasma clozapine concentrations after controlling for the effects of smoking and sex.

METHODS/PROCEDURES

Using data from a multidosage randomized double-blind US clozapine trial previously published, a random intercept linear model of steady-state total plasma clozapine concentrations was fitted to 424 concentrations from 47 patients.

FINDINGS/RESULTS

After adjusting for sex and smoking, (1) a 1-kg gain in body weight during clozapine treatment was significantly associated with a 1.4% increase in total plasma clozapine concentrations (95% confidence interval = 0.55 to 2.3) and (2) a 1-point increase in percentage of body fat during clozapine treatment was significantly associated with a 5.4% increase in total clozapine concentration (2.5 to 8.3) in females and 1.4% (-1.1 to 4.0) in males.

IMPLICATIONS/CONCLUSIONS

As hypothesized, weight increases during clozapine treatment, which probably reflect increases in fat tissue, were associated with increases in total plasma concentrations. Pending further replication in other samples, it seems likely that clozapine may deposit in body fat and that this may decrease clozapine clearance. This change may be small in most patients but may be clinically relevant in females with major gains in body fat.

Estimating individual benefits of medical or behavioral treatments in severely ill patients

There is a need for statistical methods appropriate for the analysis of clinical trials from a personalized-medicine viewpoint as opposed to the common statistical practice that simply examines average treatment effects. This article proposes an approach to quantifying, reporting and analyzing individual benefits of medical or behavioral treatments to severely ill patients with chronic conditions, using data from clinical trials. The approach is a new development of a published framework for measuring the severity of a chronic disease and the benefits treatments provide to individuals, which utilizes regression models with random coefficients. Here, a patient is considered to be severely ill if the patient’s basal severity is close to one. This allows the derivation of a very flexible family of probability distributions of individual benefits that depend on treatment duration and the covariates included in the regression model. Our approach may enrich the statistical analysis of clinical trials of severely ill patients because it allows investigating the probability distribution of individual benefits in the patient population and the variables that influence it, and we can also measure the benefits achieved in specific patients including new patients. We illustrate our approach using data from a clinical trial of the anti-depressant imipramine.

The Art of Pharmacotherapy: Reflections on Pharmacophobia

PURPOSE/BACKGROUND

This commentary deals with the neglected issue of the art of psychopharmacology by recounting the authors' journeys.

METHODS/PROCEDURES

First, a model of medical science situated within the history of medicine is described including (1) a limitation of the mathematical model of science, (2) the distinction between mechanistic science and mathematical science, (3) how this distinction is applied to medicine, and (4) how this distinction is applied to explain pharmacology to psychiatrists. Second, the neglected art of psychopharmacology is addressed by explaining (1) where the art of psychopharmacotherapy was hiding in the first author's psychopharmacology research, (2) how the Health Belief Model was applied to the art of medicine, (3) how the second author became interested in the Health Belief Model, and (4) his studies introducing the Health Belief Model in psychopharmacology. The authors' collaboration led to: (1) study of the effect of pharmacophobia on poor adherence and (2) reflection on the limits of the art of psychopharmacology.

FINDINGS/RESULTS

Low adherence was found in 45% (116/258) of psychiatric patients with pharmacophobia versus 22% (149/682) in those with no pharmacophobia, providing an odds ratio of 2.9 (95% confidence interval, 2.2-4.0) and an adjusted odds ratio of 2.5 (95% confidence interval, 1.8-3.5) after adjusting for other variables contributing to poor adherence.

IMPLICATIONS/CONCLUSIONS

Different cognitive patterns in different patients may contribute to poor adherence. Specific interventions targeting these varying cognitive styles may be needed in different patients to improve drug adherence.

Measuring the individual benefit of a medical or behavioral treatment using generalized linear mixed-effects models

We propose statistical definitions of the individual benefit of a medical or behavioral treatment and of the severity of a chronic illness. These definitions are used to develop a graphical method that can be used by statisticians and clinicians in the data analysis of clinical trials from the perspective of personalized medicine. The method focuses on assessing and comparing individual effects of treatments rather than average effects and can be used with continuous and discrete responses, including dichotomous and count responses. The method is based on new developments in generalized linear mixed-effects models, which are introduced in this article. To illustrate, analyses of data from the Sequenced Treatment Alternatives to Relieve Depression clinical trial of sequences of treatments for depression and data from a clinical trial of respiratory treatments are presented. The estimation of individual benefits is also explained. Copyright © 2016 John Wiley & Sons, Ltd.

Determining causal exposure-response relationships with randomized concentration-controlled trials

Determining causal effects in exposure-response relationships is an important but also a challenging task since confounding factors that affect both drug exposure and response often exist and lead to confounding biases in causal effect estimation. Randomized concentration control (RCC) trials are designed to eliminate or to reduce the confounding bias. However, statistical issues in the design and analysis of these trials have not been examined closely in the literature. Analysis of dose-exposure relationship may also be affected by confounding factors if they affect dose adjustments. We examined these issues and suggest methodological and practical solutions. In particular, we proposed using instrumental variables (IV) for the estimation of causal effects in both exposure-response and dose-exposure relationships. We also examined the impacts of confounded treatment heterogeneity on the IV estimate for RCC trials. We illustrated these approaches with a trial design scenario showing the importance of considering multiple practical factors that may alter the performance of the IV estimate. The performance of the IV estimates was examined by simulations for a wide range of scenarios. The results showed clear advantages for the IV estimates over routine estimates. Some situations in which the IV estimates may fail were also identified.

Can valproic acid be an inducer of clozapine metabolism?

INTRODUCTION

Prior clozapine studies indicated no effects, mild inhibition or induction of valproic acid (VPA) on clozapine metabolism. The hypotheses that (i) VPA is a net inducer of clozapine metabolism, and (ii) smoking modifies this inductive effect were tested in a therapeutic drug monitoring study.

METHODS

After excluding strong inhibitors and inducers, 353 steady-state total clozapine (clozapine plus norclozapine) concentrations provided by 151 patients were analyzed using a random intercept linear model.

RESULTS

VPA appeared to be an inducer of clozapine metabolism since total plasma clozapine concentrations in subjects taking VPA were significantly lower (27% lower; 95% confidence interval, 14-39%) after controlling for confounding variables including smoking (35% lower, 28-56%).

DISCUSSION

Prospective studies are needed to definitively establish that VPA may (i) be an inducer of clozapine metabolism when induction prevails over competitive inhibition, and (ii) be an inducer even in smokers who are under the influence of smoking inductive effects on clozapine metabolism.

Drug dosage individualization based on a random-effects linear model

This article investigates drug dosage individualization when the patient population can be described with a random-effects linear model of a continuous pharmacokinetic or pharmacodynamic response. Specifically, we show through both decision-theoretic arguments and simulations that a published clinical algorithm may produce better individualized dosages than some traditional methods of therapeutic drug monitoring. Since empirical evidence suggests that the linear model may adequately describe drugs and patient populations, and linear models are easier to handle than the nonlinear models traditionally used in population pharmacokinetics, our results highlight the potential applicability of linear mixed models to dosage computations and personalized medicine.

Role of Statistical Random-Effects Linear Models in Personalized Medicine

Some empirical studies and recent developments in pharmacokinetic theory suggest that statistical random-effects linear models are valuable tools that allow describing simultaneously patient populations as a whole and patients as individuals. This remarkable characteristic indicates that these models may be useful in the development of personalized medicine, which aims at finding treatment regimes that are appropriate for particular patients, not just appropriate for the average patient. In fact, published developments show that random-effects linear models may provide a solid theoretical framework for drug dosage individualization in chronic diseases. In particular, individualized dosages computed with these models by means of an empirical Bayesian approach may produce better results than dosages computed with some methods routinely used in therapeutic drug monitoring. This is further supported by published empirical and theoretical findings that show that random effects linear models may provide accurate representations of phase III and IV steady-state pharmacokinetic data, and may be useful for dosage computations. These models have applications in the design of clinical algorithms for drug dosage individualization in chronic diseases; in the computation of dose correction factors; computation of the minimum number of blood samples from a patient that are necessary for calculating an optimal individualized drug dosage in therapeutic drug monitoring; measure of the clinical importance of clinical, demographic, environmental or genetic covariates; study of drug-drug interactions in clinical settings; the implementation of computational tools for web-site-based evidence farming; design of pharmacogenomic studies; and in the development of a pharmacological theory of dosage individualization.

Estimating the effects of co-medications on plasma olanzapine concentrations by using a mixed model

The purpose of this study was to estimate the effect sizes of drug interactions on plasma olanzapine concentrations while adjusting for potentially confounding factors such as smoking. The estimation was performed by using a mixed model, data from a series of previously published studies of lamotrigine, oxcarbazepine, topiramate, and mirtazapine, and unpublished data from patients under clinical therapeutic drug monitoring (TDM). The total sample included 163 adult patients (age>or=18 years) who provided both steady-state plasma olanzapine concentrations and smoking information. They provided a total of 360 olanzapine concentrations (1 to 11 measures per patient). Smoking and concomitant carbamazepine or lamotrigine use were found to have significant effects on median plasma olanzapine concentrations. The effects of lamotrigine on plasma olanzapine concentrations were modified by smoking. After adjusting for olanzapine dose and carbamazepine intake, plasma olanzapine concentrations were 10% lower in non-smokers who were taking lamotrigine than in non-smokers who were not taking lamotrigine; olanzapine concentrations were 35% higher in smokers who were taking lamotrigine than in smokers who were not taking lamotrigine; olanzapine concentrations were 41% lower in smokers who were not taking lamotrigine than in non-smokers who were not taking lamotrigine; and olanzapine concentrations were 11% lower in smokers who were taking lamotrigine than in non-smokers who were taking lamotrigine. After adjusting for olanzapine dose and taking carbamazepine, the correction factor comparing smokers taking lamotrigine versus non-smokers who were not taking lamotrigine was 1.3. Gender, age, and concomitant use of mirtazapine, valproic acid, lamotrigine, topiramate, lorazepam, citalopram or oxcarbazepine did not have significant effects on olanzapine concentrations. The main limitation of this clinical design is the unavoidable substantial "noise" that characterizes (uncontrolled) clinical environments, which may make it difficult to detect the effects of some variables. Other limitations were the small sample size of some drug sub-samples and the lack of testing for plasma olanzapine metabolites.