seqDesign - Simulation and Group-Sequential Monitoring of Randomized Treatment Efficacy Trials with Time-to-Event Endpoints

A broad spectrum of both event-driven and fixed follow-up preventive vaccine efficacy trial designs, including designs of Gilbert, Grove et al. (2011, Statistical Communications in Infectious Diseases), are implemented, with application generally to individual-randomized clinical trials with multiple active treatment groups and a shared control group, and a study endpoint that is a time-to-event endpoint subject to right-censoring. The design accommodates the following features: (1) the possibility that the efficacy of the treatment/vaccine groups may take time to accrue while the multiple treatment administrations/vaccinations are given, (2) hazard ratio and cumulative incidence-based treatment/vaccine efficacy parameters and multiple estimation/hypothesis testing procedures are available, (3) interim/group-sequential monitoring of each treatment group for potential harm, non-efficacy (lack of benefit), efficacy (benefit), and high efficacy, (3) arbitrary alpha spending functions for different monitoring outcomes, (4) arbitrary timing of interim looks, separate for each monitoring outcome, in terms of either event accrual or calendar time, (5) flexible analysis cohort characterization (intention-to-treat vs. per-protocol/as-treated; counting only events for analysis that occur after a specific point in study time), and (6) division of the trial into two stages of time periods where each treatment is first evaluated for efficacy in the first stage of follow-up, and, if and only if it shows significant treatment efficacy in stage one, it is evaluated for longer-term durability of efficacy in stage two. The package produces plots and tables describing operating characteristics of a specified design including a description of monitoring boundaries on multiple scales for the different outcomes; event accrual since trial initiation; probabilities of stopping early for potential harm, non-efficacy, etc.; an unconditional power for intention-to-treat and per-protocol analyses; calendar time to crossing a monitoring boundary or reaching the target number of endpoints if no boundary is crossed; trial duration; unconditional power for comparing treatment efficacies; and the distribution of the number of endpoints within an arbitrary study time interval (e.g., events occurring after the treatments/vaccinations are given), useful as input parameters for the design of studies of the association of biomarkers with a clinical outcome (surrogate endpoint problem). The code can be used for a single active treatment versus control design and for a single-stage design.

Last updated 2 years ago

4.60 score 2 stars 7 scripts 174 downloads

CoRpower - Power Calculations for Assessing Correlates of Risk in Clinical Efficacy Trials

Calculates power for assessment of intermediate biomarker responses as correlates of risk in the active treatment group in clinical efficacy trials, as described in Gilbert, Janes, and Huang, Power/Sample Size Calculations for Assessing Correlates of Risk in Clinical Efficacy Trials (2016, Statistics in Medicine). The methods differ from past approaches by accounting for the level of clinical treatment efficacy overall and in biomarker response subgroups, which enables the correlates of risk results to be interpreted in terms of potential correlates of efficacy/protection. The methods also account for inter-individual variability of the observed biomarker response that is not biologically relevant (e.g., due to technical measurement error of the laboratory assay used to measure the biomarker response), which is important because power to detect a specified correlate of risk effect size is heavily affected by the biomarker's measurement error. The methods can be used for a general binary clinical endpoint model with a univariate dichotomous, trichotomous, or continuous biomarker response measured in active treatment recipients at a fixed timepoint after randomization, with either case-cohort Bernoulli sampling or case-control without-replacement sampling of the biomarker (a baseline biomarker is handled as a trivial special case). In a specified two-group trial design, the computeN() function can initially be used for calculating additional requisite design parameters pertaining to the target population of active treatment recipients observed to be at risk at the biomarker sampling timepoint. Subsequently, the power calculation employs an inverse probability weighted logistic regression model fitted by the tps() function in the 'osDesign' package. Power results as well as the relationship between the correlate of risk effect size and treatment efficacy can be visualized using various plotting functions. To link power calculations for detecting a correlate of risk and a correlate of treatment efficacy, a baseline immunogenicity predictor (BIP) can be simulated according to a specified classification rule (for dichotomous or trichotomous BIPs) or correlation with the biomarker response (for continuous BIPs), then outputted along with biomarker response data under assignment to treatment, and clinical endpoint data for both treatment and placebo groups.

Last updated 4 years ago

4.15 score 14 scripts 206 downloads

sievePH - Sieve Analysis Methods for Proportional Hazards Models

Implements a suite of semiparametric and nonparametric kernel-smoothed estimation and testing procedures for continuous mark-specific stratified hazard ratio (treatment/placebo) models in a randomized treatment efficacy trial with a time-to-event endpoint. Semiparametric methods, allowing multivariate marks, are described in Juraska M and Gilbert PB (2013), Mark-specific hazard ratio model with multivariate continuous marks: an application to vaccine efficacy. Biometrics 69(2):328-337 <doi:10.1111/biom.12016>, and in Juraska M and Gilbert PB (2016), Mark-specific hazard ratio model with missing multivariate marks. Lifetime Data Analysis 22(4):606-25 <doi:10.1007/s10985-015-9353-9>. Nonparametric kernel-smoothed methods, allowing univariate marks only, are described in Sun Y and Gilbert PB (2012), Estimation of stratified mark‐specific proportional hazards models with missing marks. Scandinavian Journal of Statistics}, 39(1):34-52 <doi:10.1111/j.1467-9469.2011.00746.x>, and in Gilbert PB and Sun Y (2015), Inferences on relative failure rates in stratified mark-specific proportional hazards models with missing marks, with application to human immunodeficiency virus vaccine efficacy trials. Journal of the Royal Statistical Society Series C: Applied Statistics, 64(1):49-73 <doi:10.1111/rssc.12067>. Both semiparametric and nonparametric approaches consider two scenarios: (1) the mark is fully observed in all subjects who experience the event of interest, and (2) the mark is subject to missingness-at-random in subjects who experience the event of interest. For models with missing marks, estimators are implemented based on (i) inverse probability weighting (IPW) of complete cases (for the semiparametric framework), and (ii) augmentation of the IPW estimating functions by leveraging correlations between the mark and auxiliary data to 'impute' the augmentation term for subjects with missing marks (for both the semiparametric and nonparametric framework). The augmented IPW estimators are doubly robust and recommended for use with incomplete mark data. The semiparametric methods make two key assumptions: (i) the time-to-event is assumed to be conditionally independent of the mark given treatment, and (ii) the weight function in the semiparametric density ratio/biased sampling model is assumed to be exponential. Diagnostic testing procedures for evaluating validity of both assumptions are implemented. Summary and plotting functions are provided for estimation and inferential results.

Last updated 5 months ago

4.00 score 8 scripts 170 downloads

speff2trial - Semiparametric Efficient Estimation for a Two-Sample Treatment Effect

Performs estimation and testing of the treatment effect in a 2-group randomized clinical trial with a quantitative, dichotomous, or right-censored time-to-event endpoint. The method improves efficiency by leveraging baseline predictors of the endpoint. The inverse probability weighting technique of Robins, Rotnitzky, and Zhao (JASA, 1994) is used to provide unbiased estimation when the endpoint is missing at random.

Last updated 3 years ago

3.72 score 2 stars 52 scripts 307 downloads

futility - Interim Analysis of Operational Futility in Randomized Trials with Time-to-Event Endpoints and Fixed Follow-Up

Randomized clinical trials commonly follow participants for a time-to-event efficacy endpoint for a fixed period of time. Consequently, at the time when the last enrolled participant completes their follow-up, the number of observed endpoints is a random variable. Assuming data collected through an interim timepoint, simulation-based estimation and inferential procedures in the standard right-censored failure time analysis framework are conducted for the distribution of the number of endpoints--in total as well as by treatment arm--at the end of the follow-up period. The future (i.e., yet unobserved) enrollment, endpoint, and dropout times are generated according to mechanisms specified in the simTrial() function in the 'seqDesign' package. A Bayesian model for the endpoint rate, offering the option to specify a robust mixture prior distribution, is used for generating future data (see the vignette for details). Inference can be restricted to participants who received treatment according to the protocol and are observed to be at risk for the endpoint at a specified timepoint. Plotting functions are provided for graphical display of results.

Last updated 3 years ago

3.70 score 1 stars 6 scripts 189 downloads

pssmooth - Flexible and Efficient Evaluation of Principal Surrogates/Treatment Effect Modifiers

Implements estimation and testing procedures for evaluating an intermediate biomarker response as a principal surrogate of a clinical response to treatment (i.e., principal stratification effect modification analysis), as described in Juraska M, Huang Y, and Gilbert PB (2020), Inference on treatment effect modification by biomarker response in a three-phase sampling design, Biostatistics, 21(3): 545-560 <doi:10.1093/biostatistics/kxy074>. The methods avoid the restrictive 'placebo structural risk' modeling assumption common to past methods and further improve robustness by the use of nonparametric kernel smoothing for biomarker density estimation. A randomized controlled two-group clinical efficacy trial is assumed with an ordered categorical or continuous univariate biomarker response measured at a fixed timepoint post-randomization and with a univariate baseline surrogate measure allowed to be observed in only a subset of trial participants with an observed biomarker response (see the flexible three-phase sampling design in the paper for details). Bootstrap-based procedures are available for pointwise and simultaneous confidence intervals and testing of four relevant hypotheses. Summary and plotting functions are provided for estimation results.

Last updated 4 years ago

2.70 score 9 scripts 171 downloads