How Sticky Are Clinical Trial Interventions? Site-Level Clinical Trial Participation and Differential Post-Trial Use...

A provider's participation in a randomized clinical trial (RCT) may influence their use of the trial intervention outside of trial contexts. We explored the association between site-level participation in a trial evaluating a postradical prostatectomy (RP) genomic classifier (GC; Genomics in Michigan Impacting Observation or Radiation [G-MINOR], ClinicalTrials.gov identifier: NCT02783950) and use of post-RP GC after completion of the trial's enrollment window.

The Michigan Urological Surgery Improvement Collaborative (MUSIC) data registry, in which G-MINOR was embedded, was queried for G-MINOR-eligible patients outside of the trial context (nonparticipating sites, chronology). A logistic regression model compared time with a patient's receipt of post-RP GC testing at G-MINOR participating and nonparticipating sites, before and after the trial's enrollment window.

A total of 7,144 patients (5,822 at G-MINOR sites, 1,322 non-G-MINOR sites) met study inclusion criteria between October 2015 and October 2020. Post-RCT, GC testing peaked among G-MINOR sites at 0.122 tests per eligible patient-quarter; no testing was observed among nonparticipating sites. Adjusting for patient characteristics, an interaction term between site-level RCT participation and pre-/postenrollment was statistically significant (hazard ratio, 21.9 [95% CI, 3.57 to 134]; P < .001).

Site-level participation in the G-MINOR RCT was significantly associated with a differential change in post-RCT GC use, where trial sites showed a greater post-RCT increase compared with nontrial sites. Whether this is caused by trial participation or represents a pre-existing intention to adopt an intervention remains unknown. Implementation and deimplementation considerations should be included in trial design.