Geometry of the randomized evidence for treatments of pulmonary hypertension

Abstract

Objective: We studied the entire agenda of randomized clinical trials in pulmonary hypertension (PH) using sociological methods. We explored the geometry of the PH network to interpret the evidence on multiple competing treatments for the same indication.

Design: We searched MEDLINE, Embase and Cochrane Library Databases for published studies. We queried clinicaltrials.gov and WHO International Clinical Trials Registry platform for non-published studies.

Results: We found 75 randomized trials (41 published [n = 4136 participants] and 34 registered unpublished [planned n = 3470 participants]). Of the published randomized studies, all used placebo as the comparator arm except for two nonindustry-sponsored comparisons between phosphodiestearase-5 (PDE-5) inhibitors and endothelin receptor antagonists (ERA), and one study comparing two different regimens of treprostinil. Similarly, only five unpublished/ongoing trials used an active PH treatment as comparator (PDE-5 inhibitors versus ERA (n = 3), different doses of sildenafil (n = 1) and two formulations of epoprostenol (n = 1). Of the 75 trials, 47 were sponsored by the manufacturer of the tested active product(s), and only two trials were sponsored by two companies comparing their products.

Conclusions: The relative merits of different treatment options are not directly known, as there are very few head-to-head comparisons. A limited number of ongoing studies are using active FDA-approved PH-treatments for comparison. This lack of information can be overcome by carefully designing comparative effectiveness trials.

Keywords: Pulmonary hypertension; Treatment.

Figure 1

Flowchart of eligible published randomized trials. Some of the eligible studies from different sources overlapped, and thus, the included studies are less than the sum of the eligible ones by source.

Figure 2

Trials network of randomized studies on FDA-approved PH medications. Each regimen is shown by a circle and the same color is used for treatments in the same class. An autoloop is generated when different doses of the same medication are compared. The thickness of the lines either in the autoloops or connecting the nodes is proportional to the number of studies that have compared the regimens. Trep, treprostinil; Epo, epoprostenol; Trep Inh, inhaled treprostinil; Ilo Inh, inhaled iloprost; Amb, ambrisentan; Bos, Bosentan; Tad, Tadalafil; Sild, sildenafil. Panel A: published studies in all PH groups, Panel B: published studies in PH group I, Panel C: both published and unpublished studies in all PH groups, Panel D: both published and unpublished studies in PH group I.

Figure 3

Network of companies manufacturing/commercializing PH-specific medications. Panel A: published trials only, considering the common backbone interventions, panel B: published and unpublished studies, considering also the common backbone interventions, panel C: published trials only, without considering the common backbone interventions, and panel D: published and unpublished studies, without considering the common backbone interventions. Only industry-sponsored trials are shown in all panels. The active medication versus placebo contributes to an autoloop around the company name. If backbone medication is used, an autoloop is generated if both the active and backbone medication belong to the same company; however, if products are from different companies, these two nodes were linked. If background therapy belongs to two different companies then half-a-point is given to each link. The thickness of the lines either in the autoloops or connecting the nodes is proportional to the number of sponsored studies. Iloprost is co-marketed by Bayer Schering Pharma AG in Europe and Actelion in the USA. Ambrisentan is co-marketed by GSK in Europe and Gilead in the USA. Lung LLC is a subsidiary of United Therapeutics. Cotherix was acquired by Actelion. UT, united therapeutics.