Developing and evaluating per-protocol adherence definitions in a large pragmatic clinical trial: insights from the Diuretic Comparison Project.

OBJECTIVES

Pragmatic clinical trials leveraging real-world data, such as the Diuretic Comparison Project (DCP), present unique challenges in measuring treatment adherence compared to traditional randomized controlled trials. In DCP, medications were managed through routine VA care and monitored via pharmacy claims, requiring tailored methods to assess adherence beyond standard measures like medication possession ratio (MPR). To develop a data-driven approach to define protocol treatment adherence for pragmatic comparative effectiveness trials.

METHODS

Structured EHR pharmacy dispense records guided categorization of treatment changes for two diuretics - chlorthalidone (CTD) and hydrochlorothiazide (HCTZ). We examined fill intervals, total drug supply, filling patterns among others to identify prescribing characteristics informing medication adherence definitions and time censoring for per-protocol analysis.

RESULTS

Nonadherence was defined as having a gap (>90-days pause in assigned treatment), switch (receiving the other study medication), or discontinuation (without CTD/HCTZ for >90-days). 5,476 patients met nonadherence criteria - 2,851 (52 %) in CTD and 2,625 (48 %) in HCTZ. Switching occurred earlier in CTD patients, gaps were more common in HCTZ, but discontinuation was similar between groups.

CONCLUSIONS

Defining adherence requires complex decision-making hybrid framework entailing both clinical and statistical expertise. Pragmatic studies should allow flexibility to generate data-driven, clinically-sensitive measures that appropriately describe trial compliance.