Spin of information and inconsistency between abstract and full text in RCTs investigating upper limb rehabilitation after stroke: An overview study

Article type: Research Article

Authors: Tosatto, Diego^{a; 1} | Bonacina, Daniele^{a; 1} | Signori, Alessio^b | Pellicciari, Leonardo^c | Cecchi, Francesca^{c; d} | Cornaggia, Cesare Maria^e | Piscitelli, Daniele^{e; f; g; *}

Affiliations: [a] Istituti Clinici Zucchi - Gruppo San Donato, Carate Brianza (MB), Italy | [b] Department of Health Sciences, University of Genoa, Genoa, Italy | [c] IRCCS Fondazione Don Carlo Gnocchi, Florence, Italy | [d] Department of Experimental and Clinical Medicine, University of Florence, Florence, Italy | [e] School of Medicine and Surgery, University of Milano Bicocca, Milano, Italy | [f] School of Physical and Occupational Therapy, McGill University, Montreal, Canada | [g] Department of Kinesiology, University of Connecticut, Storrs, CT, USA

Correspondence: [*] Corresponding author: Daniele Piscitelli, PT, PhD., School of Medicine and Surgery, University of Milano Bicocca via Cadore 48, 20900 Monza (MB), Italy. Tel.: +39 03 6298 6446; E-mail: [email protected].

Note: [1] These authors contributed equally to this work.

Abstract: Background:Researchers may be tempted to favorably distort the interpretation of their findings when reporting the abstract (i.e., spin). Spin bias overemphasizes the beneficial effects of the intervention compared with the results shown in the full text. Objective:To assess the occurrence of spin bias and incompleteness in reporting abstracts in post-stroke upper limb (UL) rehabilitation randomized clinical trials (RCTs). Methods:A sample of 120 post-stroke UL rehabilitation RCTs (indexed in PEDro database), published in English between 2012 and 2020, was included. The completeness of reporting and spin were assessed using the Consolidated Standards of Reporting Trials for Abstracts (CONSORT-A) and the spin checklist. The relationship between CONSORT-A and spin checklist scores with RCT and journal characteristics was assessed. Results:CONSORT-A and spin checklist scored 5.3±2.4 (max 15-points, higher scores indicating better reporting) and 5.5±2.0 (max 7-points, higher scores indicating presence of spin), respectively; Significant differences were detected between abstract and full-text scores in the CONSORT-A checklist (p < 0.01) and the spin checklist (p < 0.01). Items of the CONSORT-A checklist in the abstracts and full text showed a fair agreement (k = 0.31), while a moderate agreement (k = 0.59) for the spin checklist was detected. Completeness of abstract was associated (R2 = 0.46) with journal Impact Factor (p < 0.01), CONSORT Guideline endorsement (p = 0.04), and abstract word number (p = 0.02). A lower spin was associated with a higher journal Impact Factor (p = 0.01) and CONSORT Guideline endorsement (p = 0.01). Conclusions:Post-stroke UL rehabilitation RCTs abstracts were largely incomplete showing spin. Authors, reviewers, publishers, and stakeholders should be aware of this phenomenon. Publishers should consider allowing more words in abstracts to improve the completeness of reporting abstracts. Although we have investigated only stroke rehabilitation, our results suggest that health care professionals of all disciplines should avoid clinical decision-making based solely upon abstracts.

Keywords: Data accuracy, stroke, abstracting and indexing, neurological rehabilitation, spin bias, randomized controlled trials

DOI: 10.3233/RNN-211247

Journal: Restorative Neurology and Neuroscience, vol. 40, no. 3, pp. 195-207, 2022