Statistical analysis

Survival analysis

We conduct parametric and non-parametric survival analysis of time-to-event data, including extrapolation and proportional hazards/accelerated failure time testing. We also undertake more complex modelling where required, including: piecewise and spline-based methods, mixture/cure models, competing risks and time-shifted analysis such as assessment schedule matching. 

Adjustment for treatment switching

Treatment switching is common in clinical trials, whether incidental or by design. In such scenarios, a standard intention-to-treat analyses will not identify the true comparative effectiveness of the treatments under investigation. We have expertise in statistical methods to address this issue, including inverse-probability-of-censoring weighting (IPCW), two-stage estimation (TSE), and rank preserving structural failure time models (RPSFTM).

Additional analysis of clinical trial data

We can help unlock further insights from your trial data using additional analyses and combining it with other data sources where necessary. For example, there may be value in further subgroup analyses to investigate heterogeneity and identify patient groups in which the intervention is most effective. Other applications include developing risk prediction models, investigating subsequent treatment pathways, or analysing process outcomes to inform future trial planning. We work with you to identify research questions and advise on the most appropriate techniques.

Utility mapping and analysis

Economic evaluations typically require generic preference-based measures of health-related quality-of-life (HRQoL) such as the EQ-5D. We can map results from disease-specific HRQoL instruments to these measures,  identify the appropriate mapping algorithm and apply suitable tariffs for different countries. We can also conduct repeated measure analyses and provide time-dependent health-state utilities as inputs for cost-effective models.

Surrogate endpoint analysis

Surrogate endpoints such as biomarkers are typically observed earlier than the true outcome of interest that directly measures clinical benefit, and may be used in trials as a substitute to predict treatment effect, particularly if the outcome of interest is rare or requires long follow-up. We can conduct exploratory analysis of potential surrogate endpoints using empirical data, or perform surrogate endpoint validation using meta-analysis, depending on the level of evidence available.

Study design and analysis

We can advise on the design and analysis of clinical studies, including sample size calculations and randomization procedures, and prepare or review the statistical sections of protocols. We can help with the design and review of data collection forms, and draft statistical analysis plans.

We have extensive experience in the design and analysis of large retrospective/prospective cohort studies, case-control studies, cross-sectional surveys and routine datasets/registries. 

Joint modelling

Joint modelling is a rapidly developing area of biostatistical research that models longitudinal and time-to-event data simultaneously. We have experience in applying this novel approach as an alternative to traditional survival modelling to help improve survival predictions from limited follow-up data. Joint models can effectively adjust the estimation of biomarker profiles in the presence of informative drop-out, helping to reduce the potential bias introduced by missing data or measurement errors. They can also predict the time-to-event outcome for censored patients conditional upon their biomarker profile up to a specific time point, and incorporate multiple biomarkers into a single model. 

Statistical validation 

We offer statistical validation services based on your level of need, from review of program code and quality control checks to full independent parallel programming. We can also review statistical reports, checking for accuracy, consistency and interpretation, and conduct cross-checking with program outputs.