Misrepresentation and overinterpretation of results in clinical studies evaluating biomarkers in ovarian cancer: a systematic review of ‘spin’

Session: 

Oral session: Diagnostic test accuracy review and prognostic methods (2)

Date: 

Tuesday 18 September 2018 - 14:50 to 15:10

Location: 

All authors in correct order:

Ghannad M1, Olsen M1, Boutron I2, Bossuyt P1
1 Amsterdam Public Health Research Institute, University of Amsterdam, The Netherlands
2 Centre de Recherche Épidémiologie et Statistique Sorbonne Paris Cité (CRESS-UMR1153), Université Paris Descartes, France
Presenting author and contact person

Presenting author:

Mona Ghannad

Contact person:

Abstract text
Background: Research in cancer biomarkers (BM) has expanded in recent years. However, despite the large number of publications, very few BMs have been successfully implemented in the clinic. BM discovery studies may suffer from weak study designs, and incomplete or biased reporting, rendering them vulnerable to exaggerated interpretation of BM performance. Spin is a way of reporting, conscious or unconscious, that makes the study findings appear more favourable than results justify. Spin relies on misrepresentation and overinterpretation of results.

Objectives: To document and classify spin, as well as facilitators of spin, in recent clinical studies evaluating performance of BMs in ovarian cancer.

Methods: We searched PubMed systematically for all studies published in 2015. Studies eligible for inclusion reported the clinical performance of diagnostic, prognostic, or predictive BMs. We excluded reviews, animal studies, and cell line studies. All studies were screened by two reviewers. To document and characterize spin, we developed a set of scoring criteria with two reviewers. We extracted information on spin (i.e. misrepresentation and misinterpretation of results in title, abstract or main text conclusion), and practices that facilitate spin.

Results: Our search strategy retreived 1026 potentially eligible articles in total; 326 studies met all eligibility criteria, of which the first 200 studies, when ranked according to publication date, were included in our analysis. Only one-third (60; 30%) were free of spin, one-third (65; 32.5%) contained one type of spin, and another third (75; 38%) contained two or more forms of spin. The most frequent forms of spin identified were (Table 1):
1) other purposes of BM claimed but not investigated (65; 32.5%);
2) mismatch between intended aim and conclusion (57; 28.5%); and
3) incorrect presentation of results (40; 20%).
Frequently observed facilitators of spin were (Table 2):
1) not stating sample size calculations (200; 100%);
2) not mentioning potential harms (200; 100%); and
3) not prespecifying a positivity threshold for a continuous BM (84 of 164 studies; 51.2%).

Conclusions: Reports of studies evaluating the clinical performance of BMs in ovarian cancer frequently have spin. Misrepresentation of study findings may lead to serious implications for patients and healthcare providers.

Attachments: 

Relevance to patients and consumers: 

Spin or misrepresentation of study findings may misinform healthcare providers, potentially leading to serious implications for patients. It has been shown that spin in articles may indeed hinder the ability of readers to confidently appraise results. Scientific articles with spin are also more frequently misrepresented in press releases and news. Biomarkers hold great promise for early detection, diagnosis and treatment of cancer patients. Yet that promise can only be fulfilled with strong evaluations of the performance of markers, complete reporting of the study design and conduct, and a fair and balanced interpretation of research.