BACKGROUND: The correct assessment of drug exposure is crucial to the validity of safety research. A case control study using CPRD GOLD data (2005-2011) to assess the risk of sudden cardiac death related to oral domperidone (DOM) use required a detailed assessment of daily dose and duration.
OBJECTIVES: To assess dose and duration of exposure to DOM, using all available data in the CPRD prescription records, free-text field, and a questionnaire sent to CPRD physicians
METHODS: Numeric daily dose (NDD) in CPRD GOLD prescription records is derived by an algorithm using free text instructions. We examined all recorded NDD values for DOM prescriptions, compared them with the dosage instructions in the free text, and classified them as (1) reasonable NDD present or (2) NDD value missing or insufficient information. We sent a questionnaire to CPRD physicians of patients with DOM prescriptions closest to the index date without an NDD recorded (n=47) and, for verification, for a sample of patients with NDD recorded (n=75). Only CPRD physicians willing to participate in surveys were included.
RESULTS: Of the 336,782 DOM prescriptions, 75% had NDD values; 71% of NDD values were 3 (indicating tablets/day), and 4% had values =10 suggesting dose in mg. Of 119 physician questionnaires sent, 84 (71%) were received, 37 of 44 (84%) for prescriptions without NDD recorded in the CPRD GOLD and 47 of 75 (63%) for prescriptions with known NDD. Physicians provided dose information for 34 of 37 (91%) prescriptions without NDD values and for 44 of 47 (94%) with known NDD values. Dose category as derived from CPRD GOLD data was the same as that derived from physician questionnaires for 91% (95% CI, 79%-97%) of the prescriptions. Using all sources of information, we were able to estimate average daily dose for 241 of 246 subjects (98%) exposed to DOM at the index date. Using only NDD values recorded in CPRD GOLD would have resulted in missing information for 54 DOM-exposed subjects (22%).
CONCLUSIONS: Supplemental information can be useful in estimating dose in CPRD GOLD when recorded information is not complete.
