Huge congrats to all the doctoral students and supervisors involved! This is a step forward for transparency in clinical research. @cburgwinkel.bsky.social @jifanw.bsky.social @kahin-tai.bsky.social @heldleonhard.bsky.social 🎉
Check out the paper here:
🔗 link.springer.com/article/10.1...
Reproducibility isn't just a buzzword. Our case study highlights:
1. The immense value of systematic reproducibility checks.
2. The urgent need for researchers to fix reporting gaps.
3. Why "Data Sharing" must be more than just a checkbox—it needs to be functional.
The good news? The original conclusions of the oncology trial were confirmed.
The "real" news? The path to getting there was bumpy. We encountered:
🚩 Incomplete protocols
🚩 Broken links in documentation
🚩 Data sharing hurdles that slowed down the reproduction process.
In July 2025, 11 PhD students, divided into 5 teams, took on a challenge: Reproduce a large-scale clinical trial on oncology.
Supervised by Prof. Dr. Ulrich Mansmann, Prof. Dr. Ulrich Sax, and Prof. Dr. Leonhard Held, we went deep into the raw data to see if the original conclusions held up.
Big news! 🚀 We are thrilled to announce that our first joint publication by 11 SHARE-CTD doctoral students is officially out in @springernature.com
"Supporting the Reanalysis and Reuse of Clinical Trial Data: a case study"
This paper is the fruit of our very first Datathon in Göttingen. 🧵
Yesterday was the last day of our @sharectd.bsky.social Datathon 2 in Utrecth. Over the 1 week, 3 independent teams worked on the same research question: “Does esketamine have an impact on suicidality?”, using the same individual patient data shared by Johnson & Johnson on the The YODA repository.
In a traditional RCT, we recruit an intervention group and a control group simultaneously.
In External Controlled Trials, the control group is sourced externally from:
1️⃣ Real-World Data / Electronic Health Records.
2️⃣ Data from previous clinical trials (Historical Controls).
3️⃣ Patient registries.
Why skip the internal control?
✅ Ethical Concerns: In rare or terminal diseases, giving a placebo can be unethical if an intervention shows early promise.
✅ Feasibility: nearly impossible to find data in an extremely small population.
✅ Efficiency: speed up the development of life-saving therapies
Beware of BIAS = our biggest enemy!
Since there’s no randomisation, we face "Confounding by Time." Is the patient doing better because of the new drug, or because the "Standard of Care" in 2026 is better than the historical data from 2015?
The takeaways:
1. ECT requires even more rigorous data cleaning.
2. Data Quality is King: Garbage in, garbage out. If your external source is messy, your trial is invalid.
3. Regulatory Alignment: Always consult with FDA/EMA/local regulators early if you plan to use an ECT for drug registration.
The world of clinical trials is evolving from rigid to "Pragmatic & Adaptive." ECT proves that past data can save future lives.
Have you ever worked with External Controls or Synthetic Arms? Let’s discuss below! 👇
#PeerOnSky #ClinicalTrials #ExternalControlledTrials #RealWorldEvidence
Hi #ScienceSky people, back with #PeerOnSky series! After discussing reporting bias, let’s dive into a topic that’s currently reshaping regulatory landscapes and drug development: External Controlled Trials (ECT). Ever heard of "Synthetic Controls" or "Historical Controls"? Let’s break it down!
This paper is fully aligned with the themes of our @sharectd.bsky.social doctoral network featuring stellar PhD students (such as @kahin-tai.bsky.social and @giulia-varvar.bsky.social with whom we have published several nice papers on data sharing).
a) www.jclinepi.com/article/S089...
@kahin-tai.bsky.social's excellent talk (delivered as part of the LORIER Webinar Series on Reproducibility and Meta-Research, in partnership with @sharectd.bsky.social, ResToRes Research integrity in biomedical research, OSIRIS, the @reproductible-fr.bsky.social and Embrace) is now online.
Over the coming days, three independent teams (@sharectd.bsky.social PhD students + clinical & statistical experts) will each conduct their own IPD meta-analysis to answer the same research question.
👉 Can esketamine reduce suicidality?
Stay tuned... I’ll share updates during the week. And @sharectd.bsky.social website is here : www.sharectd.eu
How do we fix this?
✅ Preregistration is King: Stick to your original protocol.
✅ SAP Transparency: Publish your Statistical Analysis Plan before unblinding.
✅ Report Everything: If it was a primary outcome in your protocol, it MUST be in your paper, regardless of the p-value.
The message for us is clear: Negative results are not "failed" results.
When we delete non-significant data, we:
1. Distort systematic reviews/meta-analyses.
2. Waste resources on "new" studies that have actually failed before.
3. Betray the trust of the participants who volunteered.
But here's the kicker: The omission was significantly more likely if the results were not statistically significant (p > 0.05)
The interviews revealed that the reasons aren't always malicious. It's often: the "it's not interesting" excuse; lack of clinical importance (as perceived by the author); inconsistency between the protocol and the final analysis.
Smyth (2010) interviewed trialists and found a troubling pattern. ORB isn't just a myth; it's a systematic issue where researchers tend to report "good" results while burying non-significant ones.
The study found that about one-third of trials had at least one primary outcome changed or omitted. 🚩
Long time no see, #ScienceSky people! 👋 After a short hiatus, we're reliving the #PeerOnSky series.
A "silent" issue in clinical trials: Outcome Reporting Bias (ORB). Let's revisit a classic but vital paper by Smyth et al. (2010) that gets into the heads of trialists. www.bmj.com/content/342/...
#PeerOnSky Article 2! "Meta-analysis of individual participant data: rationale, conduct, and reporting" by Riley et al. (2010). IPD boosts power for subgroups and adjusts for confounders. Read: doi.org/10.1136/bmj.... What challenges do you face in IPD meta-analyses? #OpenScience #MetaAnalysis
A must read work from our network. 👍
🔬 #PeerOnSky starts! “Clinical Trial Participants’ Views of the Risks and Benefits of Data Sharing” by Mello et al. (2018). They found participants value data sharing for science, but worry about privacy breaches. A 2018 perspective still relevant today! Does this reflect current views? #DataEthics
🎉 Hello all! Welcome to #PeerOnSky! It stands as 'Peer discussion on the Sky'. We will dive into cutting-edge research, sparking discussion on BlueSky using the articles discussed within our network. We’ll kick off with a 2018 gem on clinical trial data sharing. #OpenScience #Research #JournalClub
a must read article for the week!
BBQ under the rain
class dynamics
Gottingen City tour!
Group presentation
We worked heavily on coding and biostats exercises, and balanced it out with 'rainy' BBQ evening, City tours, and German/Italian feasts. Datathon1 had it all! 😍Big thanks to the organisers and all participants! #hackathon #OpenScience #ClinicalResearch #biostatistics
🌧️ Just wrapped Datathon1 at UMC Göttingen, from 07 to 11 July! Part of SHARE-CTD, this in-person event brought doctoral students, PIs, and lecturers together for clinical trials, biostats, and data mgmt. Check the Vibes! #SHARECTD #Datathon #Research #biostatistics
The paper @mrcctu.bsky.social offers 8 principles: proactively share results, give participants a choice, make it timely & accessible, involve patients in planning, and more. It’s a call to honor participants’ roles in research. So important for trust! #PatientVoice #OpenScience What do you think?