Spin and distortion in reporting
Spin makes research appear more favourable than its results warrant. It can arise through selective reporting, inappropriate interpretation or extrapolation, and pressure to present a positive story.
Forms of spin
Spin consists of reporting practices that distort interpretation and mislead readers towards a more favourable view of results. It can reflect ignorance of the scientific issue, unconscious bias or deliberate deception. Attempts to find something significant to report can produce positive conclusions from negative research.
A systematic review reported in 2017 included 35 reports covering clinical trials, observational studies, diagnostic-accuracy studies, systematic reviews and meta-analyses. Nineteen examined the practices researchers used, identifying four categories.
- Inappropriate interpretation given the study design.
- Inappropriate extrapolation or recommendations for clinical practice.
- Selective reporting.
- More robust or favourable presentation of data.
The review also classified spin itself in four categories.
- Reporting that distorts interpretation and creates misleading conclusions suggesting a more favourable result.
- Discordance between results and their interpretation, with interpretation more favourable than the results.
- Attribution of causality where the study design does not support it.
- Overinterpretation or inappropriate extrapolation of results.
Frequency and location
In 72 reports of randomised trials, spin appeared in titles, abstracts and several sections of the main text.
| Location | Reports with spin |
|---|---|
| Title | 13 |
| Abstract Results | 27 |
| Abstract Conclusions | 42 |
| Main-text Results | 21 |
| Main-text Discussion | 31 |
| Main-text Conclusions | 36 |
More than 40% had spin in at least two main-text sections. Across the 35-report systematic review, prevalence varied substantially: it reached 100% in the main text of ten implantable-cardioverter-defibrillator trials and was as low as 9.7% in abstracts of randomised trials of systemic lung-cancer therapy. Trials had the highest prevalence and greatest variability, while industry sponsorship was not significantly associated with spin.
In a cohort study reported in 2012, 40% of 70 trial articles contained spin in their abstract conclusions, alongside spin in 47% of related press releases and 51% of news articles. Abstract-conclusion spin was the only factor significantly associated with press-release spin in the multivariable analysis, with relative risk 5.6 (95% CI 2.8–11.1; p<0.001).
Incentives and selective analysis
Publication counts, journal prestige, grant competition and pressure for statistically significant findings can encourage researchers to present data as favourably as possible. Selectively reporting significant associations can conceal the large number of comparisons that generated them. For example, testing 200 correlations at a 0.05 cutoff under chance alone would be expected to produce ten significant findings; reporting them requires disclosing upfront that multiple correlations were examined. Changing the statistical test or excluding outliers can move a borderline result across the significance threshold. Findings should nevertheless be reported truthfully, accurately and completely.
Pressure on writers
Clients can pressure freelance writers to make data appear more positive than they are; writers should stand firm and report only what is true. Recognising biased data and distinguishing good science from bad are part of that responsibility. One freelance writer reports refusing such requests by explaining that EMWA's ethical guidelines prevent her from complying, and finding that clients usually accept this explanation.
Subject-matter experts may encourage inflated positive wording and omission of negative findings. Objections can lead to patronising responses or bullying, leaving difficult choices about ignoring the behaviour, reporting it to management or resigning. Writers should maintain objective, accurate and complete documents despite that pressure.
Review and prevention
Peer reviewers and editors should check that abstract and manuscript conclusions match the results, that causal wording fits the design and that findings are not overgeneralised. The 2017 review called for better identification and classification of spin and noted that its effect on decision-making remained unknown. Its further recommendations were to base clinical practice guidelines on systematic reviews and to reform academic incentives that reward positive conclusions, publication volume and media attention.