Now that CENTRAL indexes clinical trials registry records, should I still search ClinicalTrials.gov?
Cochrane Central Register of Controlled Trials (CENTRAL) is an ongoing dream for systematic reviewers interested in finding all randomised controlled trials (RCTs) in one place. A one-stop-shop. As part of the efforts to be as inclusive as possible, CENTRAL started including records of randomised controlled trials (RCTs) from other sources such as MEDLINE, Embase, KoreaMed, CINAHL, ClinicalTrials.gov, and WHO ICTRP (World Health Organisation’s International Clinical Trials Registry Portal). Leave alone that the negotiation team must have done a brilliant job convincing the publishers/owners to contribute to CENTRAL, the outcome is not always what we expect.
The question if CENTRAL is enough has been brought out within the Cochrane Information Specialists’ community at least three times (1, 2, 3), and each time, we hesitate to replace ClinicalTrials.gov with CENTRAL.
Here I summarise why searching CENTRAL and ClinicalTrials.gov are very different and what to consider before making your final decision on searching CENTRAL or CENTRAL plus ClinicalTrials.gov.
1. Differences in Search Fields in ClinicalTrials.gov and CENTRAL
A classic bibliographic database has bibliographic or reference records with fields such as title, abstract, authors, publication year, journal name, volume, DOI, and page numbers. As a result, you can run searches in these fields.
Clinical trial registries contain study records, not reference records. As a result, you can search Healthcare Condition, Population, Interventions, Outcomes, Registration Date, Completion Date, and so on. Such study records may also contain the title, DOI, and summary. You can imagine how this will affect the search strategy:
If you search ClinicalTrials.gov directly, you will search Healthcare Condition and Interventions; however, if you search its content as indexed in CENTRAL, you will search title, abstract, and keywords.
2. The Gap between Publishing in ClinicalTrials.gov and Indexing in CENTRAL
It takes time for CENTRAL to get the new records from ClinicalTrials.gov, screen them to identify RCTs and then add them to CENTRAL. In addition, the CENTRAL on Cochrane Library usually is published monthly, and ClinicalTrials.gov is updated with new records or updates to the new records almost daily — I think they are off on weekends.
3. Irreproducibility of Same Results
In most cases, you cannot reproduce the same search results (in terms of number and content) using a translated search strategy across these two sources. Since the nature of records (Reference vs Study) and the available search fields varies between CENTRAL and ClinicalTrials.gov, one cannot replace the other until CENTRAL add the Study-Based Search feature based on PICOS or ClinicalTrials.gov adds a Reference-Based Search feature. The former is more likely in the far future.
4. Unique Retrieved Clinical Trial Registry Records
Interestingly and as a result of the abovementioned reasons, each source can give you some unique, relevant results that the other does not. Would I date to miss such unique records?
5. Inaccurate Description of Randomisation in ClinicalTrials.gov
Leave alone that the description of randomisation in ClinicalTrials.gov is not standard, and its content is not peer-reviewed, I have seen records that are not reporting an RCT, but they claim to be ‘randomised’ and vice versa. Some don’t report randomisation at all. Some records are a randomised or non-randomised extension of a randomised trial. Such inaccuracies could cause confusion in indexing them in CENTRAL, and it may not be possible to contact every trialist to make an accurate decision during the indexing process; however, if we search ClinicalTrials.gov for every systematic review of RCTs, we probably will keep the suspicious records to contact the trialists.
6. CENTRAL for RCTs, ClinicalTrials.gov for RCTs Plus!
CENTRAL tries to index RCTs or quasi-RCTs from ClinicalTrials.gov. ClinicalTrials.gov contains many other study designs, including cohorts, case-control studies, case series, non-randomised controlled studies, single-group interventional trials, and so on. So if your systematic review includes beyond RCTs, ClinicalTrials.gov will offer much more than CENTRAL.
7. Complex Search Strategy for CENTRAL, Simple and Auto-Mapping for ClinicalTrials.gov
Since CENTRAL is another classic bibliographic database, you can design and run search strategies using operators such as truncation and adjacency in long search strings. ClinicalTrials.gov needs a simpler search strategy with a limited number of characters for the search box. However, ClinicalTrials.gov has a hidden mapping function that retrieves synonyms of words or expands the search (automatic query expansion). This function usually saves space, and upon test, you can have a shorter search strategy (I’ll write about this).
8. Evidence to Support the Exclusion of ClinicalTrials.gov
There is no evidence to support searching CENTRAL is enough, probably for the above obvious reasons. While Australian colleagues published a recent guide in BMJ to guide the systematic reviewers in searching the ClinicalTrials.gov, a recent study from Australia on additional records found by searching clinical trials registers reported that “On average, there was no difference in the meta-analytic effect estimates before versus after adding the new trials.” Of course, each study should be critically appraised in context and considering the limitations. We need to consider that we do not search extra sources only to change the effect estimates — think about it until I write another blog post about this.
9. CENTRAL Indexes Some Contents from ClinicalTrials.gov Records
You might have noticed that CENTRAL only indexes some of the content from ClinicalTrials.gov records. As a result, the search results between the two sources will vary.
10. Livingness of ClincialTrials.gov Records
The clinical trials registry records are being updated frequently with valuable data and information, including the trial’s progress, availability of results, and automatically linked publications from PubMed. Regardless of where you find the trial record (CENTRAL, WHO ICTRP, Google, or ClinicalTrials.gov), you or the review team will have to check the record in ClinicalTrials.gov for the most recent updates and data.
11. MECIR Standards
Based on MECIR standards for Cochrane reviews, we should search the clinical trial registers alongside the CENTRAL. If Cochrane — the producer of CENTRAL — were sure that CENTRAL is enough, they would change the recommendation.
Conclusion
Although CENTRAL may contain all RCTs from ClinicalTrials.gov, we have to continue searching ClinicalTrials.gov alongside CENTRAL for the above 11 reasons.
CENTRAL however is specifically useful for providing the estimated number of trials per question. Many grant application teams have asked me to give a quick quote on how many trials they should expect to screen searching CENTRAL.
We all wish that one day we will search only one source for the systematic reviews and will not have to translate and run searches across the platforms. Let’s dream on.
While the above reasons could also apply to WHO ICTRP, I feel I need another blog post to tame that beast.
Note: One more thing, let’s stop referring to ClinicalTrials.gov as CT.gov. I know both URLs were working once, but CT.gov takes you to Connecticut’s Official State Website right now.
Question for you: What is the difference between Register and Registry? When to use each?
If you liked this blog post, please support me by pressing the green Follow button and signing up so I can write more. Email Subscription Dysfunctions. Thank you :D
