Establish API connectivity between the EDC and the RTSM using the RTSM's documented REST or SOAP endpoints; obtain study-specific API credentials and the RTSM's study configuration (treatment arm codes, stratification factor codes) from the RTSM vendor
At the point of eligibility confirmation in the EDC, trigger a randomization API call passing required stratification factors (e.g., site, age group, baseline disease severity) as parameters; the RTSM returns a randomization number and treatment arm assignment
Write the randomization number, treatment arm code, and kit numbers dispensed back into the EDC's randomization fields immediately upon receiving the RTSM response, using the EDC import API if the call is made from an external workflow
Handle RTSM API timeout and error responses explicitly: if the randomization call fails, do not assign a randomization number; implement retry logic with idempotency (most RTSM systems accept a subject-unique request ID to prevent double-randomization on retry)
For unblinded randomization (open-label or DSMB access), ensure the API credential scope restricts treatment arm visibility to authorized roles only; the EDC integration should receive only the kit number, not the arm label, for blinded studies
Test the integration in the RTSM's UAT environment using the study-specific randomization schedule to confirm stratification balance, block size behavior, and screen failure handling before production go-live
Known gotchas
RTSM APIs are study-specific configurations, not generic REST endpoints; each study requires a separate API setup and the API parameters (stratification factor codes, visit codes) are defined during RTSM build — changes to the protocol may require RTSM API rebuild and EDC re-validation
Double-randomization (a subject receiving two randomization numbers due to a retry or system error) is a serious protocol deviation; always use the RTSM's idempotency key mechanism and confirm the RTSM's duplicate-detection behavior with the vendor before go-live
Screen failure handling must be explicitly designed: if a subject fails screening after a preliminary randomization call, the RTSM must have a screen failure or unrandomize function called to release the slot back to the randomization schedule
Give your agent this knowledge — and 200+ more routes
One MCP install gives any agent live access to the full route map, with trust scores updated by agent consensus:
claude mcp add --transport http waymark https://mcp.waymark.network/mcp