To minimize transcription errors, data entry should be as direct as possible from the source document, eliminating the need for manual calculations by site personnel. Allowing staff to enter the lab value and select the corresponding unit as recorded in the source (Option B) achieves this. This method transfers the responsibility of unit conversion to a centralized, programmable, and verifiable system within the clinical database. This systematic approach is significantly less error-prone than relying on manual conversions performed at multiple international sites, which can introduce inconsistencies and calculation mistakes.

A. Deriving units from the value's magnitude is speculative and unreliable, as different units can have overlapping numerical ranges, leading to incorrect assumptions.
C. Forcing data entry into a standard unit requires site staff to perform manual conversions if the source unit differs, which is a primary source of error.
D. This is functionally the same as option C; mandating manual conversion by site staff before data entry significantly increases the risk of calculation errors.

1. Prokscha, S. (2011). Practical Guide to Clinical Data Management (3rd ed.). CRC Press. In Chapter 5, "Case Report Form Design," it is emphasized that CRFs should be designed to be clear and easy to complete. A key principle stated is, "Avoid asking the site to perform calculations, such as calculating a patient’s age from the date of birth or converting pounds to kilograms" (p. 68). This directly supports capturing the original value and unit without on-site conversion.

2. Society for Clinical Data Management (SCDM). (2023). Good Clinical Data Management Practices (GCDMP). The principles of quality data collection emphasize capturing data accurately as it appears in the source. Section 4.6, "CRF/eCRF Design," highlights the need for "clear and unambiguous" forms. Forcing on-site conversion (Options C and D) introduces ambiguity and potential for error, while allowing entry of the original value and unit (Option B) is a clearer, more direct method of transcription.

3. Clinical Data Interchange Standards Consortium (CDISC). (2023). Study Data Tabulation Model Implementation Guide for Human Clinical Trials (SDTMIG) v4.0. The SDTM model design inherently supports this best practice. It specifies variables for the original result and unit (--ORRES, --ORRESU) and separate variables for the standardized result and unit (--STRESC, --STRESU). This structure is built on the premise that original data is collected first and then standardized centrally (p. 118, Section 4.1.5.3, Standard Test Results).