A reading list begins with papers you can see. An evidence assessment also asks which completed studies might be missing from that list.
- Distinguish a registered study, a results entry and a journal paper.
- Record the search date: publication status can change.
- Missing evidence raises uncertainty without revealing the missing result.
- Sample
- 585 large randomised trials
- Size threshold
- At least 500 planned or actual participants
- Registry cohort
- Closed to recruitment before 2009
- Final publication search
- November 2012
Original publication: 29 Oct 2013 · The date above refers to this brief.
Where results were found in this historical cohort
Number of trials; mutually exclusive categories, total 585
What the audit could see
Jones and colleagues searched for publications from large registered trials. They counted peer-reviewed reports with outcome data and checked the registry when no publication was found. The 171 trials without an identified journal paper included 38 with registry results. Searching only journals would therefore miss an additional route to evidence.
Source 1 ↗Why missing results can change a synthesis
Cochrane distinguishes missing whole studies from missing results within known studies. Bias can arise when availability depends on the result. The direction and size of that bias are not revealed by the number of absent papers alone. A funnel-plot pattern is also not a diagnosis of publication bias; differences between studies can contribute.
Source 2 ↗Build a small evidence ledger
Original exercise: take one research question and create a row for each unique trial identifier. Add separate fields for protocol, results registry, journal article and correction. If two papers share a trial identifier, inspect whether they report the same participants at different times. Keep both reports attached to one study until the relationship is clear. A blank field means not found in your search, not confirmed nonexistent.
How to describe an unsuccessful search
Write a reproducible note: the identifiers and terms used, the sources searched, the date, and the result. For example, “No results entry located in these two sources on this date” is a bounded observation. “The investigators hid the result” adds a claim about intent that the search cannot establish. Preserve unresolved possibilities for a later check.
What would move the question forward?
Our proposal is to follow a clearly defined contemporary cohort over time and separate late reporting, non-reporting and registry-only reporting. Measure completeness as well as presence. Do not use this older cohort’s percentage as a substitute for a new measurement. A better future evidence map would let readers see both the documents located and the gaps still unresolved.
Three different gaps to keep separate
| Evidence | What it contributes | What remains open |
|---|---|---|
| Registration without results | Confirms a study record can be located | Does not reveal the outcome |
| Registry results without a paper | May add otherwise missed numerical evidence | Methods and completeness still need checking |
| Paper without a planned outcome | Some findings are available | A particular result may still be missing |
Original reading map. Presence in a database is not a quality rating.
Your questions, answered
Are these figures the current publication rate?
No. They describe the historical cohort and search window shown above. A current estimate would require a new search of a defined cohort.
Is a trial registry just a list of titles?
A registration record and a results record serve different purposes. Check which information is actually present rather than relying on the existence of a registry link.
Can missing studies be assumed to show no benefit?
No. The unknown results stay unknown. Sensitivity analyses can examine assumptions, but assumptions must be labelled as such.
Should I count every paper as a separate trial?
No. Link reports by their study identifiers and participant descriptions. One trial may produce several papers; combining them carelessly can count the same evidence twice.
Limits of this interpretation
- Historical registry cohort; no claim about every trial or today’s reporting rate.
- Registry enrollment could be planned or actual, so participant totals require care.
- This is a reading aid, not a new systematic search or reanalysis.
Sources & transparency
- Jones et al. (2013): Non-publication of large randomized clinical trials
Indexed primary full-text methods and results: registry selection, publication search and results availability. Direct publisher retrieval was intermittent. · Accessed 27 Sep 2026
DOI: 10.1136/bmj.f6104 - Cochrane Handbook, chapter 13: missing evidence in a meta-analysis
Public HTML: missing studies, missing results and interpretation of reporting bias. Methodological context, not a new trial. · Accessed 27 Sep 2026
Prepared with AI using the source passages described below. No human editorial or clinical review has been completed. Reading exercises and future research questions are our interpretation, not additional study findings. This is an educational case study about research reporting, not a treatment recommendation.
Source check: AI source check — primary-source passages and reported numbers; 27 September 2026
Clinical review: Not applicable to this educational guide
Suggest a correction