Exposure-response modeling of average daily pain score, and dizziness and somnolence, for mirogabalin (DS-5565) in patients with diabetic peripheral neuropathic pain.

Mirogabalin (DS-5565) is an α2δ-1 ligand being developed for pain associated with diabetic peripheral neuropathy, fibromyalgia, and postherpetic neuralgia. Nonlinear mixed-effects analyses were performed on average daily pain and on the incidence of the adverse events dizziness and somnolence. These models were used to predict the dose of mirogabalin equivalent to pregabalin and the probability of meaningful reduction in pain compared with placebo and pregabalin. In addition, regimen effects were evaluated for reductions of adverse events. Mirogabalin was estimated to be 17-fold more potent than pregabalin. The effectiveness of 150 mg pregabalin, dosed twice daily, attenuated by week 5. Therefore, the estimated mechanism-based equivalent dose (ED) of 17.7 mg mirogabalin was higher than that predicted to achieve comparable pain reduction. If attenuation of the pregabalin effect is real, mirogabalin doses lower than the ED could yield comparable pain reduction, albeit with less differentiation in pain from placebo. The incidence rate of dizziness and somnolence decreased over time. Twice-daily dosing of mirogabalin was predicted to yield a lower incidence rate of dizziness than once-daily dosing; thus, titration of dosages should reduce adverse event rates. These model results were used to influence phase 3 dosing selection.

Modeling dropout from adverse event data: impact of dosing regimens across pregabalin trials in the treatment of generalized anxiety disorder.

Dizziness represents a major determinant of dropout in the treatment of generalized anxiety disorder with pregabalin. Titration (dose escalation) regimens based on clinical judgment were implemented to mitigate this adverse event and reduce patient dropout across clinical trials. Dropout is an important treatment failure endpoint, which can be analyzed using time-to-event models that incorporate daily dosing or other time-varying information. A parametric discrete-time dropout model with daily dizziness severity score as a covariate afforded a systematic, model-based assessment of titration dosing strategies, with model predictions evaluated against corresponding nonparametric estimates. A Gompertz hazard function adequately described the decreasing dropout hazard over time for individuals with severe or moderate dizziness and a lower, constant hazard for individuals reporting no dizziness or mild dizziness. Predictive performance of the model was adequate based on external validation with an independent trial and other goodness-of-fit criteria. Prospective simulations highlight the utility of this approach in reducing dropout based on examination of untested titration scenarios for future generalized anxiety disorder or other trials.

Population pharmacokinetic-pharmacodynamic modeling of biological agents: when modeling meets reality.

The pharmacokinetics (PK) and pharmacodynamics (PD) of many biological agents (biologics) have inherent complexities requiring specialized approaches to develop reliable, unbiased models. Three cases are covered: preponderance of zero values, nonresponder subpopulations, and adaptive dosing. Engineered biologics exhibit high affinity for target receptors. Biologics can saturate receptors, abolishing free receptor levels for protracted periods. Consequently, the distribution of observations can be heavy at, and near, the boundary. A 2-part model (ie, a truncated δ log-normal distribution) may be appropriate. Mixture models identify subpopulations based on bimodal or multimodal distributions of η values. With biologics, PD may be compromised because of lack of receptors, or the PD may be affected because of other events resulting in erratic excursions. Nonresponders exhibit a random walk-around placebo trajectory, resulting in high residual variability. The distributions of etas are often badly skewed or polymodal. An indescribable mixture model separates subjects who are nonresponders, providing diagnostic pharmacologic information on the drug. Many biologics use PD-based adaptive dosing. During model development, data used for model development include adaptive dosing. For simulation, adaptive dosing must be implemented. Failure to account for dose adjustments results in biased or inflated prediction intervals because subjects in the simulated data undergo inappropriate dose adjustments.

Joint modeling of dizziness, drowsiness, and dropout associated with pregabalin and placebo treatment of generalized anxiety disorder.

Dizziness and drowsiness are cited as being predictors of dropout from clinical trials for the medicine pregabalin. These adverse events are typically recorded daily on a four point ordinal scale (0 = none, 1 = mild, 2 = moderate, 3 = severe), with most subjects never reporting either adverse event. We modeled the dizziness, drowsiness, and dropout associated with pregabalin use in generalized anxiety disorder using piecewise Weibull distributions for the time to first non-zero dizziness or drowsiness score, after which the dizziness or drowsiness was modeled with ordinal regression with a Markovian element. Dropout was modeled with a Weibull distribution. Platykurtosis was encountered in the estimated random effects distributions for the ordinal regression models and was addressed with dynamic John-Draper transformations. The only identified predictor for the time to first non-zero dizziness or drowsiness score was daily titrated dose. Predictors for dropout included creatinine clearance and maximum daily adverse event score. Tolerance to adverse events over time was modeled by including a non-stationary component for the dizziness ordinal Markov regression while the piecewise Weibull distributions allowed a change point in the median time to first non-zero dizziness or drowsiness score.

Hepatic absorbed radiation dosimetry during I-131 metaiodobenzylguanidine (MIBG) therapy for refractory neuroblastoma.

PURPOSE: To compare the prediction of therapeutic hepatic radiation-absorbed dose rates from tracer imaging plus a linearity assumption to estimation based on intra-therapy imaging in (131)I metaiodobenzylguanidine (MIBG) therapy of refractory neuroblastoma. MATERIALS AND METHODS: Conjugate-view images of the liver were obtained before therapy for seven patients at seven times after a tracer infusion of (131)I MIBG and at three times after the therapy infusion. Measured liver activities were converted to dose-rate estimates. Three statistical models of the rates assuming double exponential dependences on time were examined. One of the three models allowed for a multiplicative correction to the therapeutic late-phase dose-rate amplitude. Results from that model: (1) the tracer prediction of the late-phase absorbed-dose-rate amplitude was a factor of 1.75 times the intra-therapy-estimated value, and (2) the difference between tracer prediction of the radiation-absorbed dose and intra-therapy estimation of it was statistically significant, and (3) the liver radiation-absorbed dose did not reach 30 Gy. CONCLUSIONS: A statistical modeling analysis finds that the radiation-absorbed dose after therapy appears to be lower than that which is predicted from the linear scaling with administered activity of the tracer radiation-absorbed dose. Hepatocyte toxicity is the most likely reason but it is not high enough to produce clinically observable results.

An alternative method for population pharmacokinetic data analysis under noncompliance.

Noncompliance presents a persistent problem while analyzing PK data from outpatient clinical studies. Ignoring dose omission or making uninformed assumptions about patient drug intake history can prove detrimental to the objectives of the analysis (e.g. determining the PK model parameters or identifying covariates) and ultimately compromise the interpretation of the data. In order to overcome this problem, an alternative method of handling noncompliant data is evaluated in this report. The proposed approach is based on the principle of superposition and works by separating the estimation of the elimination rate from the model based steady-state PK concentration. Simulations implementing this method under different scenarios of noncompliance demonstrate that it performs better than the conventional method of analyzing population PK data when compared on the basis of bias and imprecision in parameter estimation and power (and type I error) for covariate detection. Overall, the new method exhibits great potential to address the issue of uncertain/unreliable dosing histories frequently encountered in clinical trials.

Mixed effects modeling of weight change associated with placebo and pregabalin administration.

OBJECTIVE: To characterize change from baseline weight over time for pregabalin and placebo administration. METHODS: Asymptotic fraction of baseline weight was modeled with a nonmixture model and a mixture model as a function of baseline weight, exposure, time, covariate effects, and subject-specific random effects. Model fit was assessed using standard diagnostic plots. Predictive performance was assessed using both data similar to the original data, and open-label data. RESULTS: The nonmixture model indicated that a typical patient (baseline weight 82 kg) receiving placebo or 300 mg/day pregabalin approached an asymptotic fractional change from baseline weight of [mean (95% prediction interval for typical individual)] 0.7% (-5.5% to 7.4%) or 2.5% (-3.8% to 9.1%), respectively, with a half-life of 17 days. Substantial between-subject variability is observed, with some drug-treated subjects remaining weight neutral or losing weight, at all levels of exposure. Structural fixed effects parameters for the two submodels (mixture model) were in close agreement with each other and with those for the nonmixture model. The mixture model described two subpopulations differing in interindividual variability. No significant interindividual-varying covariates influencing the mixture probabilities were identified other than exposure. Both models had adequate fit; both models performed well during external validation. Predictive performance (nonmixture model) was adequate to ~900 days. CONCLUSIONS: The weight of a typical 82-kg patient receiving placebo or pregabalin (300 mg/day) approaches an asymptotic fractional change from baseline weight of 0.7%, or 2.5%, respectively, with a half-life of 17 days. Substantial between-subject variability remains unexplained.

Evaluation of mixture modeling with count data using NONMEM.

Mixture modeling within the context of pharmacokinetic (PK)/pharmacodynamic (PD) mixed effects modeling is a useful tool to explore a population for the presence of two or more subpopulations, not explained by evaluated covariates. At present, statistical tests for the existence of mixed populations have not been developed. Therefore, a simulation study was undertaken to evaluate mixture modeling with NONMEM and explore the following questions. First, what is the probability of concluding that a mixed population exists when there truly is not a mixture (false positive significance level)? Second, what is the probability of concluding that a mixed population (two subpopulations) exists when there is truly a mixed population (power), and how well can the mixture be estimated, both in terms of the population parameters and the individual subjects classification. Seizure count data were simulated using a Poisson distribution such that each subject's count could decrease from its baseline value, as a function of dose via an Emax model. The dosing design for the simulation was based on a trial with the investigational anti-epileptic drug pregabalin. Four hundred and forty seven subjects received pregabalin as add on therapy for partial seizures, each with a baseline seizure count and up to three subsequent seizure counts. For the mixtures, the two subpopulations were simulated to differ in their Emax values and relative proportions. One subpopulation always had its Emax set to unity (Emax hi), allowing the count to approach zero with increasing dose. The other subpopulation was allowed to vary in its Emax value (Emax lo = 0.75, 0.5, 0.25, and 0) and in its relative proportion (pr) of the population (pr = 0.05, 0.10, 0.25, and 0.50) giving a total of 4.4 = 16 different mixtures explored. Three hundred data sets were simulated for each scenario and estimations performed using NONMEM. Metrics used information about the parameter estimates, their standard errors (SE), the difference between minimum objective function (MOF) values for mixture and non-mixture models (MOF (delta)), the proportion of subjects classified correctly, and the estimated conditional probabilities of a subject being simulated as having Emax lo (Emax hi) given that they were estimated as having Emax lo (Emax hi) and being estimated as having Emax lo (Emax hi) given that they were simulated as having Emax lo (Emax hi). The false positive significance level was approximately 0.04 (using all 300 runs) or 0.078 (using only those runs with a successful covariance step), when there was no mixture. When simulating mixed data and for those characterizations with successful estimation and covariance steps, the median (range) percentage of 95% confidence intervals containing the true values for the parameters defining the mixture were 94% (89-96%), 89.5% (58-96%), and 95% (92-97%) for pr, Emax lo, and Emax hi, respectively. The median value of the estimated parameters pr, Emax lo (excluding the case when Emax lo was simulated to equal 0) and Emax hi within a scenario were within +/- 28% of the true values. The median proportion of subjects classified correctly ranged from 0.59 to 0.96. In conclusion, when no mixture was present the false positive probability was less than 0.078 and when mixtures were present they were characterized with varying degrees of success, depending on the nature of the mixture. When the difference between subpopulations was greater (as Emax lo approached zero or pr approached 0.5) the mixtures became easier to characterize.

Exposure-response analysis of pregabalin add-on treatment of patients with refractory partial seizures.

OBJECTIVE: Our objectives were to describe the exposure-response relationship of pregabalin add-on treatment for refractory partial seizures after multiple dosing in patients and to identify the factors that influence this relationship. METHODS: A mixed-effects model was used to characterize the relationship between monthly seizure frequency over a 3-month period and pregabalin daily dose (0, 50, 150, 300, and 600 mg) as add-on treatment in 3 double-blind, parallel-group studies in patients with refractory partial seizures (N = 1042). Seizure frequency was modeled as a Poisson process expressed as a function of baseline seizures, drug treatment, placebo effect, and subject-specific random effects. The model included a parameter that partitioned the population into subpopulations with respect to response. RESULTS: Seventy-five percent of patients showed an asymptotic decrease in seizure frequency with increasing doses of pregabalin, whereas 25% did not demonstrate a significant decrease in seizure frequency from baseline. In patients who demonstrated a dose-related decrease in seizure frequency from baseline, the maximal percentage of seizure reduction from baseline was 100% for women and 80% for men, with a 186-mg daily dose decreasing seizures on average to 50% of maximum. Age, race, and menopausal status did not significantly affect seizure frequency. CONCLUSION: Pregabalin add-on treatment demonstrates a dose-response relationship in 3 out of 4 patients with refractory partial seizures. A dose of 186 mg pregabalin daily is expected to decrease the seizure rate by 50% of maximum from baseline. Age, race, and menopausal status of women did not affect the dose-response relationship.

A two-part mixture model for longitudinal adverse event severity data.

We fit a mixed effects logistic regression model to longitudinal adverse event (AE) severity data (four-point ordered categorical response) to describe the dose-AE severity response for an investigational drug. The distribution of the predicted interindividual random effects (Bayes predictions) was extremely bimodal. This extreme bimodality indicated that biased parameter estimates and poor predictive performance were likely. The distribution's primary mode was composed of patients that did not experience an AE. Moreover, the Bayes predictions of these non-AE patients were nearly degenerative, i.e., the predictions were nearly identical. To resolve this extreme bimodality we propose using a two-part mixture modeling approach. The first part models the incidence of AE's, and the second part models the severity grade given the patient had an AE. Unconditional probability predictions are calculated by mixing the incidence and severity model probability predictions. We also report results of simulation studies, which assess the predictive and statistical (bias and precision) performance of our approach.