Poster Presentation Sydney Spinal Symposium 2022

Characteristics of baseline frequency data in spinal RCTs do not suggest widespread fraud (#108)

Manon M Levayer 1 , Gem R Chew 2 , Kyle A Sheldrick 3 , Ashish D Diwan 4 5
  1. School of Biomedical Engineering, The University of Sydney, Darlington, NSW, Australia
  2. Faculty of Medicine, University of New South Wales, Kensington, NSW, Australia
  3. St George and Sutherland Clinical School, Faculty of Medicine, University of New South Wales, Kensington, NSW, Australia
  4. Spine Labs, St George and Sutherland Clinical School, Faculty of Medicine, University of New South Wales, Kensington, NSW, Australia
  5. Spine Service, Department of Orthopaedic Surgery, St George Hospital Campus, Kogarah, NSW, Australia

Background

Fraud in clinical research is a reality. A recent scandal of non-genuine data in numerous randomised clinical trials from one site in the field of back pain has highlighted the risk of fraud impacting the assessment of treatment effects and consequently impacting clinical guidelines.

Aim

We investigated whether evidence of non-random baseline data in randomised studies for spine disorders was widespread or if the aforementioned case was an anomaly.

Methods

We performed a PubMed search to identify all papers published in 4 major spine journals (Spine, The Spine Journal, the Journal of Neurosurgery Spine, and the European Spine Journal) classified by the journal as RCTs between 2016 and 2020. We extracted all baseline covariates which were reported as frequencies and calculated p values using a Pearson Chi Square test of independence. Individual variable p-values were combined into a study wise p-value using the Stouffer method. We calculated the proportion of studies with a study wise p-value of <0.01, <0.05, >0.95 and >0.99. All studies with p values >0.99 or <0.01 were examined for explanations.

Results

Our literature search returned 228 studies, of which 169 were included. In total, we extracted 938 variables. After calculating all study wise p-values and looking at the proportions that presented a study wise p-value of <0.01, <0.05, >0.95 or >0.99, we observed only small differences between our results and the number of studies that would be expected to fall into these categories by random chance.

Conclusion

We conclude that the overall distribution of frequency baseline similarities between study arms in RCTs published in major spine journals is not suspicious. No evidence of systemic presentation of false data was found. We will examine continuous variables in future work.