RWE guideline causal inference part1 (ENG)

We had our first meeting of the Pseudo Lab’s RWE (Real World Evidence) guideline study. To generate real world evidence from real world data, we need to understand the concept of causal inference.

We studied comparability and randomization, which are among the most important concepts in causal inference. Roughly speaking, comparability between two groups is established when the groups are similar. The state of the treatment group before receiving treatment should be similar to the state of the control group before receiving treatment (e.g., placebo). Also, both groups should respond similarly to treatment.

If these conditions are met, we can say two groups were similar. Since one group receives treatment while the other doesn’t, comparison becomes possible. But how is this possible? The answer lies in randomization. When subjects are randomly assigned to two groups, their basic characteristics become similar.

However, in reality, characteristics might be unbalanced when sample sizes are small. There might also be important variables (such as recruiting hospitals). Therefore, randomization is sometimes done after stratifying such information.

Please check the link below for detailed information.

https://causalinferencelab.github.io/Bridging-Causal-Inference-and-Real-World-Evidence-A-Study-of-FDA-and-Other-Regulatory-Guidelines/docs/causal_inference_part1.html

Source: Original Korean article | English translation by Claude 3.5 Sonnet (New)