Preventing recurrence of endometriosis by means of long-acting progestogen therapy (PRE-EMPT): report of an internal pilot, multi-arm, randomised controlled trial incorporating flexible entry design and adaption of design based on feasibility of recruitment.

BACKGROUND: Endometriosis is associated with the growth of endometrium in ectopic sites mainly within the pelvis. This results in inflammation and scarring, causing pain and impaired quality of life. Endometriotic lesions can be excised or ablated surgically, but the risk of recurrence is high. A Heath Technology Assessment commissioning call in 2011 sought applications for trials aimed at evaluating long-term effectiveness of postoperative, long-acting, reversible contraceptives (LARCs) in preventing recurrence of endometriosis. A survey of gynaecologists indicated that there was no consensus about which LARC (Levonorgestrel Intrauterine System (LNG-IUS) or depot medroxyprogesterone acetate injection (DMPA)) or comparator (combined oral contraceptive pill (COCP) or no treatment) should be evaluated. Hence, we designed a 'flexible-entry' internal pilot to assess whether a four-arm trial was feasible including a possible design adaption based on pilot findings. METHODS: In this pilot, women could be randomised to two, three or four treatment options provided that one was a LARC and one was a non-LARC. An assessment of feasibility based on recruitment to these options and a revised substantive trial design was considered by an independent oversight committee. RESULTS: The study ran for 1 year from April 2014 and 77 women were randomised. Only 5 (6%) women accepted randomisation to all groups, with 63 (82%) having a LARC preference and 55 (71%) a non-LARC preference. Four-way and three-way designs were ruled out with a two-way LARC versus COCP design, stratified by prerandomisation choice of LARC and optional subrandomisation to LNG-IUS versus DMPA considered a feasible substantive study. CONCLUSIONS: Multi-arm studies are potentially efficient as they can answer multiple questions simultaneously but are difficult to recruit to if there are strong patient or clinician preferences. A flexible approach to randomisation in a pilot phase can be used to assess feasibility of such studies and modify a trial design based on chosen recruitment options, but trialists should consider carefully any practical arrangements should groups need to be dropped during a study.