Helena Geys

Validation of surrogate end points in multiple randomized clinical trials with failure time end points

Before a surrogate end point can replace a final (true) end point in the evaluation of an experimental treatment, it must be formally ‘validated’. The validation will typically require large numbers of observations. It is therefore useful to consider situations in which data are available from several randomized experiments. For two normally distributed end points Buyse and co-workers suggested a new definition of validity in terms of the quality of both trial level and individual level associations between the surrogate and true end points. This paper extends this approach to the important case of two failure time end points, using bivariate survival modelling. The method is illustrated by using two actual sets of data from cancer clinical trials.

Statistical challenges in the evaluation of surrogate endpoints in randomized trials.

The validation of surrogate endpoints has been studied by Prentice, who presented a definition as well as a set of criteria that are equivalent if the surrogate and true endpoints are binary. Freedman et al. supplemented these criteria with the so-called proportion explained. Buyse and Molenberghs proposed to replace the proportion explained by two quantities: (1). the relative effect, linking the effect of treatment on both endpoints, and (2). the adjusted association, an individual-level measure of agreement between both endpoints. In a multiunit setting, these quantities can be generalized to a trial-level measure of surrogacy and an individual-level measure of surrogacy. In this paper, we argue that such a multiunit approach should be adopted because it overcomes difficulties that necessarily surround validation efforts based on a single trial. These difficulties are highlighted.

Topics in modelling of clustered data

INTRODUCTION Correlated Data Settings Developmental Toxicity Studies Complex Surveys Other Relevant Settings Reading Guide MOTIVATING EXAMPLES National Toxicology Program Data Heatshock Studies Belgian Health Interview Survey POPS Data Low-Iron Rat Teratology Data The Wisconsin Diabetes Study Congenital Ophthalmic Defects A Developmental Toxicology Study ISSUES IN MODELING CLUSTERED DATA Choosing a Model Family Joint Continuous and Discrete Outcomes Likelihood Misspecification and Alternative Methods Risk Assessment MODEL FAMILIES Marginal Models Conditional Models Cluster-Specific Models GENERALIZED ESTIMATING EQUATIONS General Theory Clustered Binary Data PSEUDO-LIKELIHOOD ESTIMATION Pseudo-Likelihood: Definition and Asymptotic Properties Relative Efficiency of PL versus ML Pseudo-Likelihood and Generalized Estimating Equations PSEUDO-LIKELIHOOD INFERENCE Test Statistics Simulation Results Illustration: EG Data FLEXIBLE POLYNOMIAL MODELS Fractional Polynomial Models Local Polynomial Models Other Flexible Polynomial Methods and Extensions ASSESSING THE FIT OF A MODEL A Hosmer-Lemeshow Approach for Likelihood Based Models Order Selection Tests Data-Driven Tests in Multiple Regression Testing Goodness of Fit QUANTITATIVE RISK ASSESSMENT Expressing Risks Analysis of NTP Data Asymptotic Study Concluding Remarks MODEL MISSPECIFICATION Implications of Misspecification on Dase Effect Assessment A Robust Bootstrap Procedure Implications of Misspecification on Safe Dose Determination A Profile Score Approach EXACT DOSE-RESPONSE INFERENCE Exact Nonparametric Dose-Response Inference Simulation Study Concluding Remarks INDIVIDUAL LEVEL COVARIATES Cluster-Specific Models Population-Averaged Models Efficiency of Modeling Approaches Analysis of Heatshock Data Continuous Outcomes Concluding Remarks COMBINED CONTINUOUS AND DISCRETE OUTCOMES Models for Bivariate Data of a Mixed Nature Application to Quantitative Risk Assessment Discussion MULTILEVEL MODELING OF COMPLEX SURVEY DATA Multilevel Models Application to the HIS Concluding Remarks APPENDIX: BAHADUR PARAMETER SPACE REFERENCES INDEX

Pseudolikelihood modeling of multivariate outcomes in developmental toxicology

Abstract The primary goal of this article is to determine benchmark doses based on the ethylene glycol study, which comprises data from a developmental toxicity study in mice. Because the data involve a vector of malformation indicators, a flexible model for multivariate clustered data is required. An exponential family model is considered and pseudolikelihood-based inferential tools are proposed, hence avoiding excessive computational requirements.

A pairwise likelihood approach to estimation in multilevel probit models

A pairwise likelihood (PL) estimation procedure is examined in multilevel models with binary responses and probit link. The PL is obtained as the product of bivariate likelihoods for within-cluster pairs of observations. The resulting estimator still enjoys desirable asymptotic properties such as consistency and asymptotic normality. Therefore, with this approach a compromise between computational burden and loss of efficiency is sought. A simulation study was conducted to compare PL with second-order penalized quasi-likelihood (PQL2) and maximum (marginal) likelihood (ML) estimation methods. The loss of efficiency of the PL estimator is found to be generally moderate. Also, PL tends to show more robustness against convergence problems than PQL2.

Modulation of group II metabotropic glutamate receptor (mGlu2) elicits common changes in rat and mice sleep-wake architecture.

Compiling pharmacological evidence implicates metabotropic glutamate mGlu(2) receptors in the regulation of emotional states and suggests positive modulators as a novel therapeutic approach of Anxiety/Depression and Schizophrenia. Here, we investigated subcutaneous effects of the metabotropic glutamate mGlu(2/3) agonist (LY354740) on sleep-wake architecture in rat. To confirm the specific effects on rapid eye movement (REM) sleep were mediated via metabotropic glutamate mGlu(2) receptors, we characterized the sleep-wake cycles in metabotropic glutamate mGlu(2) receptor deficient mice (mGlu(2)R(-/-)) and their arousal response to LY354740. We furthermore examined effects on sleep behavior in rats of the positive allosteric modulator, biphenyl-indanone A (BINA) alone and in combination with LY354740 at sub-effective doses. LY354740 (1, 3 and 10 mg/kg) dose-dependently suppressed REM sleep and prolonged its onset latency. Metabotropic glutamate mGlu(2)R(-/-) and their wild type (WT) littermates exhibited similar spontaneous sleep-wake phenotype, while LY354740 (10 mg/kg) significantly affected REM sleep variables in WT but not in the mutant. In rats, BINA (1, 3, 10, 20, 40 mg/kg) dose-dependently suppressed REM sleep, lengthened its onset latency and slightly enhanced passive waking. Additionally, combined treatment elicited a synergistic action on REM sleep variables. Our findings show common changes of REM sleep variables following modulation of metabotropic glutamate mGlu(2) receptor and support an active role of this receptor in the regulation of REM sleep. The synergistic action of BINA on LY354740s effects on sleep pattern implies that positive modulators would tune the endogenous glutamate tone suggesting potential benefit in the treatment of psychiatric disorders, in which REM sleep overdrive is manifested.

Recommendations on the statistical analysis of the Comet assay

In 2010, the Statisticians in the Pharmaceutical Industry (PSI) Toxicology Special Interest Group met to discuss the design and analysis of the Comet assay. The Comet assay is one potential component of the package of safety studies required by regulatory bodies. As these studies usually involve a three-way nested experimental design and as the distribution of the measured response is usually either lognormal or lognormal plus a point mass at zero, the analysis is not straightforward. This has led to many different types of analysis being proposed in the literature, with several different methods applied within the pharmaceutical industry itself. This article summarises the PSI Toxicology Groups discussions and recommendations around these issues.

Two latent variable risk assessment approaches for mixed continuous and discrete outcomes from developmental toxicity data

Measurements of both continuous and discrete outcomes are encountered in many statistical problems. Here we consider the particular context of teratology studies, where quantitative risk asessment is aimed at determining the effect of dose on the probability that an individual fetus is malformed or of low birth weight, both being important measures of teratogenicity. We will introduce two different joint marginal mean models for outcomes of a mixed nature. First, we propose the Plackett-Dale approach, where for each binary outcome it is assumed that there exists an underlying glatent variable. The latent malformation outcomes are then assumed to follow a Plackett distribution. The second approach we consider is a probit approach. Here it is assumed that there exists an underlying continuous variable for each binary outcome, so the joint distribution for weight and malformation can be assumed to follow a multivariate normal distribution. In both cases, specification of the full distribution will be avoided using pseudolikelihood and generalized estimating equations methodology, respectively. Quantitative risk assessment is illustrated using data from two developmental toxicology experiments.

Validation of a longitudinally measured surrogate marker for a time-to-event endpoint

The objective of this paper is to extend the surrogate endpoint validation methodology proposed by Buyse et al. (2000) to the case of a longitudinally measured surrogate marker when the endpoint of interest is time to some key clinical event. A joint model for longitudinal and event time data is required. To this end, the model formulation of Henderson et al. (2000) is adopted. The methodology is applied to a set of two randomized clinical trials in advanced prostate cancer to evaluate the usefulness of prostate-specific antigen (PSA) level as a surrogate for survival.

Investigating the criterion validity of psychiatric symptom scales using surrogate marker validation methodology.

This work investigates whether techniques that are generally used for the validation of surrogate markers in clinical trials can be applied in the validation of psychiatric health measurements (often scales) and more generally to investigate relationships between treatment effects on different measurements. However, the categorical nature of some scales makes these techniques inapplicable in the way they were originally defined. In this work, we show a possible extension of this methodology to the setting in which one of the scales is an ordinal categorical variable. When psychiatric health measurements are either developed or used in a new population, reliability and validity must be investigated. Reliability, more specifically internal consistency, test–retest reliability, and inter-rater reliability, is focused on the reproducibility of the measurement. Validity is defined as the degree to which the scale measures what it purports to measure. This can be performed through the analysis of content, construct, and criterion validity. We argue that recent methodology, in particular developed to study surrogate endpoints, can be used to examine criterion validity, concurrent validity, and predictive validity. In concurrent validity, we correlate the measurement with a criterion measure, both of which are given at the same time. In predictive validity, the criterion will not be available to some point in time in the future. The surrogate methods were applied on pooled data from five trials in schizophrenia.