Statistical Analysis
Providing Insight into Data
HEOR analysis of the RCT data
1
We analyze both RCT data and real-world observational data, including surveys, claims databases and registries.
​
Our goal is to fully exploit their HEOR potential, and explore additional treatment effects.
​
We analyze Phase 2 and Phase 3 FDA-approved STDN ADAM datasets in SAS.
What for?
Efficient use of existing resources, resulting in deep insight in own clinical data (40-80 new data tables).
Allows for data exploration and finding of additional treatment effects
Psychometric performance of PRO or clinical endpoints
2
What for?
-
To better understand the items and the total score of a PRO
-
To identify the most responsive items
-
To explore treatment effects on items
-
To validate the PRO in the study population
​​​
Determining the MCID of a PRO or clinical endpoint
3
What for?
-
To calculate the minimum amount of change in a (primary or secondary endpoint) PRO that represents a true change in the patient’s health condition.
-
Results are used to interpret study results, recalculate treatment effect on a secondary endpoint, and calculate the proportion of responders.
Determining the minimal important difference of EQ-5D-5L utility values in CIDP patients using data from a large clinical trial. To be submitted
Estimating the Minimal Clinically Important Difference for the Myasthenia Gravis Quality of Life revised scale (MG-QOL15r). Submitted Qual Life Res
Mixed methods and textual analysis
4
What for?
We analyze qualitative data in NVIVO, to understand what patients find important:
-
Which symptoms are most or least burdensome?
-
Are there symptoms not included in the clinical study programme?
-
Which symptoms would they choose to alleviate first?
-
What are the most experienced symptoms?
It is important to integrate the patient's voice in the selection of endpoints.
Other statistical analysis techniques
We use SAS/STATA, R, Python or VBA, and a variety of analysis techniques including, but not limited to:​​
-
Survival analysis
-
Longitudinal analysis, with missing data methods
-
Resource utilization and cost analysis
-
Time-series analysis
-
Multilevel analysis
-
Analysis of discrete choice experiments
-
Factor analysis
-
Clustering analysis
-
Propensity score adjustment for non-randomization