search

cdisc-org / ddf-core-poc

This repository will contain the results from the Proof of Concept project.
MIT License
0 stars 1 forks source link

Define rule DDF99992: Estimand and Endpoint and Analysis population - AnalysisPopulation must have one Estimand. #78

Closed rbakercdisc1 closed 5 months ago

rbakercdisc1 commented 10 months ago

CHK0097: Proposed rule:

This is regarding the Entities of Estimand, Endpoint and Analysis population -

Each AnalysisPopulation must have one Estimand. There is a one-to-one relationship between Endpoint and Estimand. Each Estimand nust have one or more IntercurrentEvent, but must have at least one.

Resources on Estimands Understanding estimands https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8112325/ FDA - Comment from Estimands in Oncology Working Group (ASA Biopharmaceutical Section SWG and EFSPI SIG) https://www.regulations.gov/comment/FDA-2020-D-2307-0027 What is an estimand in a clinical trial: The PIONEER 1 example | Research Square https://www.researchsquare.com/article/rs-6260/v1 Examples PSI EIWG Webinar: PIONEERing estimands in Clinical Research (psiweb.org) https://psiweb.org/vod/item/psi-eiwg-webinar-pioneering-estimands-in-clinical-research PSI EIWG Webinar: Estimands in Oncology - How and Why (psiweb.org) https://nam12.safelinks.protection.outlook.com/?url=https%3A%2F%2Fpsiweb.org%2Fvod%2Fitem%2Fpsi-eiwg-webinar-estimands-in-oncology---how-and-why&data=04%7C01%7Ccconnolly%40cdisc.org%7C5840fc0507ad43129ba708d9738bafa5%7C078244a1de674c9e9088986d7f110a37%7C0%7C0%7C637667866062814284%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C1000&sdata=Zrx9dskZhoeZzuXsC4rU4qubXIMXoCaLynRcTuf2KnQ%3D&reserved=0

DianeWold commented 9 months ago

I think it's more than an estimand has an associated analysis population. More than one estimand could have the same analysis population.

rbakercdisc1 commented 9 months ago

How about this?

Each analysis population condition must have at least one estimand.

DianeWold commented 9 months ago

The cardinality of the relationship in the model is 1:1. I think the cardinality is directional, so this is saying that each Estimand has one related AnalysisPopulation. I believe that a different Estimand could be related to the same AnalysisPopulation. So one could check that there is no "free-floating" AnalysisPopulation that doesn't "come from" an Estimand. For the set of AnalysisPopulations for a StudyDesign, each must have a relationship to an Estimand.

DianeWold commented 9 months ago

Created DDF99992. I believe it is checking cardinalities correctly. The description is, "Within a StudyDesign, each Estimand is related to exactly one AnalysisPopulation and each AnalysisPopulation is related to at least one Estimand."

This rule may be split into two parts at the creation stage.

rbakercdisc1 commented 9 months ago

Reviewed. Meets the definition of done.

ASL-rmarshall commented 8 months ago

For USDM v2.6 (through v2.9), I think this is just a cardinality check because every instance of the AnalysisPopulation class is defined within the context of an instance of the Estimand class. Even if you did have a single conceptual analysis population that is referenced by more than one estimand definition, the analysis population would have to be newly defined (with a different id value) for each of the estimands.