An observational study of missing data inquiry from randomized trial authors showed a poor response

Sofia Tsokani; Dimitrios Mavridis; Shahab Hajibandeh; Shahin Hajibandeh; George A Antoniou; Alexandros Andreou; Stavros A Antoniou

doi:10.1016/j.jclinepi.2019.11.014

An observational study of missing data inquiry from randomized trial authors showed a poor response

J Clin Epidemiol. 2020 Mar:119:19-25. doi: 10.1016/j.jclinepi.2019.11.014. Epub 2019 Nov 25.

Authors

Sofia Tsokani¹, Dimitrios Mavridis², Shahab Hajibandeh³, Shahin Hajibandeh⁴, George A Antoniou⁵, Alexandros Andreou⁶, Stavros A Antoniou⁷

Affiliations

¹ Department of Mathematics, School of Sciences, University of Ioannina, Ioannina, Greece.
² Department of Primary Education, School of Education, University of Ioannina, Ioannina, Greece; Paris Descartes University, Sorbonne Paris Cité, Faculté de Médecine, Paris, France.
³ Department of General Surgery, Royal Bolton Hospital, Bolton, UK.
⁴ Department of General Surgery, Stepping Hill Hospital, Stockport, UK.
⁵ Department of Vascular and Endovascular Surgery, The Royal Oldham Hospital, Pennine Acute Hospitals NHS Trust, Manchester, UK.
⁶ Upper GI Department, Castle Hill Hospital, Hull and East Yorkshire Hospitals NHS Foundation Trust, Hull, UK.
⁷ School of Medicine, European University Cyprus, Nicosia, Cyprus. Electronic address: stavros.antoniou@hotmail.com.

PMID: 31778777
DOI: 10.1016/j.jclinepi.2019.11.014

Abstract

Background and objectives: The quality of evidence from a meta-analysis might be affected by poor or inadequate reporting of individual randomized controlled trials (RCTs). The results of data request from authors have not been investigated to date. Aim of this study was to quantify these results and propose a bundle of actions for improvement of data retrieval.

Study design and setting: After completion of a feasibility trial, the authors of 116 RCTs selected for inclusion in 4 meta-analyses and network meta-analyses were contacted via email to check abstracted data from their studies and/or provide missing data. Univariable and multivariable analyses were performed to investigate the association of response rate with predefined characteristics, including time interval between publication of the RCT and the date of contact, role of the contact author as corresponding author, the number of participating centers, the source of contact email address and risk of bias assessments.

Results: The response rate to the primary email invitation was 31.1% (median response time 0 days, interquartile range [IQR]: 0-1). Of the authors who did not respond to the primary invitation, 31.2% responded to the reminder (median response time 1 day, IQR: 0-3). Despite an overall response rate of 41%, both confirmation of data accuracy and supplementation of missing data were provided for 9 RCTs only (7.8%). No association was found between lack of response and predefined characteristics on univariable and multivariable analyses.

Conclusion: Despite a fair response rate, contacting authors of RCTs to check abstracted data and complete missing items yielded a very low rate of task completion. These findings highlight the importance of endorsing a data-sharing culture among researchers. The role of alternative incentives is yet to be investigated.

Keywords: Data acquisition; Data sharing; Meta-analysis; Randomized controlled trial.

Publication types

Observational Study

MeSH terms

Authorship*
Data Accuracy*
Humans
Information Storage and Retrieval / methods*
Information Storage and Retrieval / statistics & numerical data*
Prospective Studies
Randomized Controlled Trials as Topic / statistics & numerical data*