Statistical Analysis

Methodologies and Scope of Services

Syreon statistical analysis capabilities include analysis plan preparation, sample size estimation, randomization, statistical programming, multivariate stepwise regression, Cox proportional hazard modeling, and logistic regression analysis to examine the simultaneous role of multiple continuous or categorical variables in determining principal outcomes. Compound models may also be developed to incorporate both categorical and continuous variables in comparing treatment outcomes. Data is analyzed in SAS version 9.1 and results and tables can be compiled in accordance with in-house specifications or Sponsor requirements.

The analytical plan is professionally designed to meet the highest scientific and regulatory standards, and continuous analysis permits maximum data accuracy and review. Longitudinal data from scheduled visits and event driven data collection can be combined in a number of ways, based on actual visit dates or on scheduled visit IDs. They are then analysed using methods ranging from survival data methods (Kaplan-Meier curves, log-Rank tests, Cox proportional hazards models, time-varying covariate or coefficient models, etc.), to mixed-effects linear and non-linear regression models (simple repeated measures ANOVA, random coefficient regression models, nonlinear pharmacokinetic models with mixed effects, etc.), and generalized linear models with or without random effects (logistic regression models, GEE methods, quasi-likelihood methods, etc.). Recurrent event data structures, where a particular event can occur a number of times to the same subject, are of particular interest and expertise within the Syreon biometrics group as described in their recent methodological publications in the statistical literature (e.g., Journal of the American Statistical Association, Biometrics).

Continuous statistical review of clinical, pharmacokinetic, health quality and economic data throughout the study, identifies problems or trends, provides critical information for Data Review Boards and facilitates preparation for the final analysis.

Post-Marketing Evaluations – Unique Data Challenges

Syreon post-marketing studies involve substantial longitudinal data recording utilizing both scheduled visits and event-driven data collection. While conventional randomized clinical trials have defined and scheduled visits during the study duration, naturalistic studies are driven by principles of clinical care and often require different analytical strategies, with greater requirement for statistical methods to control for biases introduced by pre-existing differences between patient subgroups.

Many Syreon studies incorporate resource utilization data, offering the possibility for within study economic assessments of cost-effectiveness or cost-utility. Some of these data require the development of bootstrap resampling algorithms to characterize the precision of incremental cost-effectiveness estimates.