We are confronted with authors of primary RCTs who refuse to share details of their published data with us for a meta-analysis we are currently conducting. The authors want to be co-author of our meta-analysis only on the basis of their shared data. If we don't agree on having the as co-author in our paper they will not share their data. Sharing data only will not meet the minimum requirements for authorship according to http://www.icmje.org/ethical_1author.html. We cannot pool data if their standard deviations on their already published data are lacking. In my opinion this hampers the progress of science.
It's a tough situation - nobody wants to have hundreds of authors on a paper, especially when the majority only gave access to the data and were not involved in the analysis and reporting!
I wonder if it'd be acceptable to include a list of 'submissions/participant labs' in the acknowledgements? Or as part of the supplemental data?
It's clear that some sort of recognition is required in these cases, but the standard author format doesn't quite seem appropriate for meta-analyses and systematic reviews
I concur. An acknowledgement would be suitable. My co reviewers and I suggested that. I hope to find more support online to persuade them sharing data. Thanks for responding. Much appreciated !
Hi Spyridon, I also completely agree and probably would go one step further and evaluate their study out of the review because of missing data.
Thank you. It would be a shame if we have to exclude a study just because of the suborn authors. If their data would truly be worthless, okay. But it seems they have valid data only reported suboptimal and therefor useless for pooling in our meta-analysis. Now we can point them at this open and transparent discussion.
Rudolf I think there are two issues here:
1) accurate reporting of results (i.e. include the SDs and generally follow the CONSORT guidelines) - needed for two-stage meta-analysis
2) sharing primary data which costs money and the authors have gone through great effort to collect - needed for the better individual patient data (IPD) meta-analysis.
Regarding accurate reporting I think that's self explanatory. It's completely unacceptable for the authors to refuse to provide information they should have reported in the first place and of course there is no question about authorship here. I would consider writing to the journal in which they published and name and shame. Your piece will appear there for everyone to see and the authors should be embarrassed enough to respond and provide what you need. Unless they are retired they will not want to jeopardise their relationship with the journal either... At the end of the day this is bad research (under-reporting).
As for data, this is much more complicated. The standard IPD model invites all authors who contribute data to be co-authors. Of course they provide some input on the paper and are informed through the process. This is what we did with out TARDIS paper (see my profile). Although I agree this is not ideal, you must acknowledge that these people have spent time and put considerable effort in collecting the data and they need to be acknowledged in that (and I'm saying as a hippy data-for-all statistician). Realistically, if you don't offer authorship you will never get primary data. It gets more complicated when the data was collected through public funding but currently only a few organisations have the insight to actually own and redistribute the data. Eventually we will move towards that but I don't see the authorship model change any time soon. And at the end of the day, does it really matter if you have 20 authors on your paper, when people know that's how it works? And then again you do naming and shaming in your paper, people who never got back to you or refused to share (some are actually not allowed to share the data if funded by pharma etc) even though they were offered authorship - which places a question mark over their research since it implies they might have something to hide!
Dear Evangelos, tnx for your useful addition to the discussion. Naming and shaming will bring only negative energy. The thing is that although I have no problem with adding authors acknowledging their hard work many journals won't allow this number of authors. Perhaps researchgate can be a new way acknowledging authors who share data.
Thanks Rudolf. We need to move towards openness in order to advance science and people who make a living through science need to acknowledge that. If people ignore requests or refuse to provide clarifications, that should set off alarm bells - at least for me. Unfortunately, there are many stories of corruption in science and openness is the best way to address that. Naming people who refuse, is the least one can do and might eventually bring a collective change - I see it as positive energy...
I am still confused as to whether you mean sharing (primary) data or sharing results (statistical measures and estimates). Like I said the latter is a no-brainer when it comes to authorship.
I just need results with standard deviations to facilitate pooling in our MA. This is not an IPD MA. I hope they will read this discussion moving fwd science.
Well, I think that if there are vague basic statistics in the primary studies, the quality of those studies is questionable. I'd suggest strict exclusion/inclusion criteria for studies. I'd also expect, that such studies with unsufficient description of the sample are not multicentric trials with thousands of participants; therefore, exclusion of a study with low N may be rather beneficial.
- Badges
- Science topic