Epidemiology and publication of discontinued randomized trials—the DISCO study

Article type
Authors
Kasenda B1, von Elm E2, You J3, Blümle A4, Tomonaga Y5, Saccilotto R1, Amstutz A1, Bengough T2, Meerpohl JJ4, Stegert M1, Briel M1
1Basel Institute for Clinical Epidemiology and Biostatistics, University Hospital Basel, Switzerland
2Cochrane Switzerland, CHUV and University of Lausanne, Switzerland
3Department of Clinical Epidemiology & Biostatistics, McMaster University, Canada
4German Cochrane Center, Freiburg, Germany
5Institute of Social and Preventive Medicine, Zurich, Switzerland
Abstract
Background: Discontinuation of randomized clinical trials (RCTs) has ethical implications: Participants consent on the premise of contributing to new medical knowledge, non-publication of discontinued RCTs compromises systematic reviews, and precious resources are wasted. Little is known about the epidemiology and publication history of discontinued RCTs, especially those discontinued due to poor recruitment.

Objectives: To (i) estimate the risk of trial discontinuation, (ii) identify risk factors for discontinuation due to poor recruitment, and (iii) identify risk factors for non-publication in journals.

Methods: We established a multicentre cohort of RCTs based on protocols approved by six research ethics committees (REC) from 2000 to 2003 in Switzerland, Germany, and Canada. From included RCT protocols we extracted data on study design and planned recruitment. We determined completion status of RCTs using REC files, identified publications, and surveys of trialists. We investigated factors associated with discontinuation due to poor recruitment and full publication using logistic regression.

Results: We included 1080 RCT protocols; 956 (88.5%) enrolled patients or participants at risk, and 124 (11.5%) included healthy volunteers only. The latter were excluded from the present analysis. 52 RCTs (5.4%) were never started, 10 (1.0%) are still ongoing. Of the remaining 894 RCTs, 248 (27.7%) were discontinued; reasons thereof are summarized in the table. In multivariable analysis, industry-initiation was the strongest factor preventing discontinuation due to poor recruitment (adjusted odds ratio [OR] 0.24, 95% CI 0.14–0.40; p < 0.001). 532 (59.5%) RCTs were published in journals including 115 discontinued RCTs (46% of 248). Trial discontinuation was strongly associated with non-publication (adjusted OR 2.94, 95% CI, 2.12–4.10; p < 0.001).

Conclusions: Discontinued RCTs are common, in particular if they are investigator-initiated. Data from discontinued RCTs are frequently not published and therefore compromise systematic reviews and meta-analyses.