Research Waste in ME/CFS

Objective To compare the prevalence of selective reporting in ME/CFS research areas: psychosocial versus cellular. Method A bias appraisal was conducted on three trials (1x psychosocial and 2x cellular) to compare risk of bias in study design, selection and measurement. The primary outcome compared evidence and justifications in resolving biases by proportions (%) and ORs (Odds Ratio); the secondary outcome determined the proportion (in %) of ME/CFS grants at risk of bias. Results NS (cellular study) was twice as likely to present evidence in resolving biases over PACE (psychosocial trial) (OR = 2.16; 65.6% vs 46. 9%), but this difference was not significant (p = 0.13). However, NS was five times more likely to justify biases over PACE (OR = 4.76; 46.9% vs 15. 6%) and this difference was significant (p = 0.0095; p < 0.05). PACE was weak in place (operational aspects 32%) and NS for data practices (37%). The proportion of grants were more biased in PACE (72%) than NS (28%) for evidence, and also more biased in PACE (86%) than NS (14%) for justifications. Conclusion Psychosocial trials on ME/CFS are more likely to engage in selective reporting indicative of research waste than cellular trials. Improvements to place may help reduce these biases, whereas cellular trials may benefit from adopting more translatable data methods. However, these findings are based on two trials. Further risk of bias appraisals are needed to determine the number of trials required to make robust these findings.

1. Background research communities. 23 As clinical trials become more complex, there is increasing concern selec-24 tive reporting is harder to detect, and unforeseen complexities may escalate 25 between the oversight bodies that monitor research integrity (eg. issues of 26 research misconduct) versus the autonomy which allow research communities 27 to freely conduct their own research. This review seeks to demonstrate these 28 complexities in Chronic Fatigue Syndrome (CFS). 29 2. ME/CFS 30 Myalgic Encephalomyelitis (ME) and CFS are not yet considered distinct 31 diagnoses, but have been in the past (White et. al., 2007). The time to 32 onset and the causes of these fatigue-like symptoms are confined to case 33 studies (low evidence), and is still debated among experts. Nevertheless, 34 both are diagnosed when there is an absence of fatigue-related disorders, and 35 the patient achieves a minimum threshold score for ME/CFS in at least one = 0·0013, p < 0·05) over 52 weeks (White et. al., 2011). ME/CFS support 48 groups have rejected this treatment regimen due to harms from post-exertion 49 malaise after GET, and no improvements after CBT. Biomolecular findings 50 further support these claims with evidence of cellular level harms detected af-51 ter GET (Cook et. al., 2017), and have proposed biomarkers that are unique 52 to ME/CFS (Fluge et. al., 2016). The consensus is that ME/CFS is a com-53 plex, multi-faceted disorder that requires a multi-disciplinary approach, and 54 aetiologies at multiple angles, such as: gut microbiota, hormonal, endocrine 55 and immune functions. However, psychosocial angles can also offer impor-  Search terms "myalgic encephalomyelitis", "cognitive", "behaviour", "gra-63 ded exercised therapy", "adaptive pacing therapy", "gene", "cell", "clinical-64 trials.gov" were automatically mined from PubMed using E-utilities on a that may choose to replicate this search strategy. This search strategy did not scope for treatment effects as done in systematic reviews, but on research 73 trends in selecting high impact trials for a bias appriasal. and Adaptive Pacing Therapy) to represent psychosocial trends (White et. 79 al., 2011); 2. A neural study (NS) on post-exertion malaise after GET to 80 represent cellular trends (Cook et. al., 2017). 3. A gut study (GS) on 81 profiling gut microbial differences in ME/CFS individuals (Giloteaux et. al.,82 2017) to also represent cellular trends. In table 1, biases were categorised 83 by: "study design", "selection" and "measurement." Each potential bias was 84 rated by the author with a plus (+) or a minus sign (-) to indicate whether 85 a study presented evidence (E) or a justification (J) for resolving a potential 86 bias. The first two columns E and J rated PACE (White et. al., 2011). The 87 next two columns E and J rated GS (Giloteaux et. al., 2017); followed by 88 ratings for NS (Cook et. al., 2017). The column "Neural Study" offered an 89 example of each rating from Cook and colleagues' (2017) paper. The column 90 "Potential Biases" defined these biases in public health terms. The far right 91 column with the letters "T", "P", "D", "R": Theory (theories and models 92 used in the trial); Place (operational conduct); Recruitment (participant 93 recruitment); Data (data practices) were collated to predict the areas of 94 strengths and weaknesses (selective reporting) in each trial.  Figure 2: Summary of bias ratings. "+" sign indicates evidence or justifications present in the trial, "-" sign indicates it was not present. Areas "T", "P", "D", "R" stand for Theory, Place, Data and Recruitment to represent the potential weak areas in the trial. ME/CFS research trends: psychosocial or cellular were compared using 98 a 2x2 contingency table to determine the strength of evidence (table 1) 99 and justifications (table 2). Columns E and J from the bias appraisal

119
The search strategy identified 1750 published articles for psychosocial al., 2011); 2. Cellular: Neural study (Cook et. al., 2017) four treatment arms all ending in 52 weeks but these endpoints were not 141 justified. NS also did not justify measuring neural endpoints in 3 days.

142
Hence both studies were potentially biased on whether treatment effects 143 were measured at an optimal level for a true effect. PACE registered 144 the trial (ISRCTN54285094), NS did not, however, it did specify the    . 4).

198
However, there was no significant difference between the two studies in pre-199 senting evidence to resolve biases (p = 0·13; p > 0·05).   al., 2015; Lian & Nettleton, 2015). However, the results also suggest, cellular 226 trials are also likely to engage in selective reporting, but its therapeutic ben-227 efit is difficult to assess, since no study as of yet have proposed a therapeutic 228 agent (eg. drug) exclusively designed and marketed for treating ME/CFS 229 (Collatz et. al., 2016). Brurberg and colleagues (2014)  is not feasible, then begin with pre-clinical models (eg. animal models) and I dedicate this to ME Awareness Week 2017. I would like to thank ME 262 groups for your passionate advocacy and for sharing your stories with me.

269
Supplementary material