The multiple facets of registry-based randomised controlled trials

Abstract

Traditional randomised controlled trials (RCTs) have long been considered the gold standard for evaluating clinical interventions. However, registry-based trials (RBTs) are emerging as a promising methodology, offering the potential for generating high-quality clinical evidence with greater external validity, reduced complexity, and lower costs. Despite being hailed as the “next disruptive technology in clinical research”, the adoption of RBTs has been slower than anticipated. This lag could be attributed, in part, to the lack of consistent definitions for RBTs, which are often conflated with registry-based RCTs (RRCTs). Our analysis of seven RRCT reviews revealed significant variability in how RRCTs are defined and reported across clinical trial registries, reflecting the absence of standardised descriptors and nomenclature when registering these trials on clinical trial registries. This ambiguity complicates accurate estimation of their implementation and impact, reflected in reviews reporting widely varying numbers of trials registered as RRCTs. While RRCTs represent a valuable approach for addressing critical research questions, for their full potential to be realised, clear and consistent definitions, along with consensus on standardised registration terms, are essential. Given this, we would propose a taxonomy for RRCTs based on how the registry is used to support a RCT. We also welcome the CONSORT extension for the reporting of RRCTs and advocate for a standardised approach on how they are registered on clinical trial registries. Achieving this would not only enhance the recognition of RRCTs but also their impact on clinical practice. We propose a two-tier RRCT classification based on the number of outcomes captured in the registry and its use for identification or recruitment of participants, and encourage ongoing discussion around RRCT taxonomy to help guide future research.

Background

Traditional randomised controlled trials (RCTs) are the gold standard for evaluating clinical interventions [1–4]. Registry-based trials (RBTs) are an emerging methodology showing potential to provide high quality clinical evidence with external validity, reduced complexity and costs [5]. Despite Lauer et al. [6] heralding RBTs as potentially the “next disruptive technology in clinical research”, 12 years ago, uptake has been slower than expected. This is partly due to inconsistent definitions of RBTs, often used interchangeably with registry randomised controlled trials (RRCTs), making their implementation difficult to ascertain. This is evidenced by a number of reviews that have yielded inconsistent findings in the number of RRCTs captured [1, 7, 8]. Contributing to this are the variable definitions used by the review authors and the lack of standardised descriptors when registering them on clinical trial registries [1, 2, 9]. This presents a need for consensus on a definition and taxonomy for RRCTs [8], and for RRCTs to be registered in accordance with the CONSORT extension guidelines [10].

Randomised controlled trials that involve a registry exhibit diverse trial designs and variable utilisation of the registry. In some cases, the registry is used for a limited or single purpose, variably capturing some elements of patient characteristics, disease detail, treatment received and/or outcomes [4, 11] or may be limited to the identification of participants [12]. At the other extreme, the majority of trial data is captured in the registry, which may also include randomisation [13, 14]. Therefore, RCTs that utilise a registry to some extent are usually captured under the broader umbrella of a RBT or may be referred to as a RRCT, registry-embedded RCT, registry trial, hybrid trial or pragmatic trial.

A tighter definition for what constitutes an RRCT has been suggested by Li et al., where the registry is used as a platform for participant recruitment and acquisition of outcome/endpoint data [15]. The extent to which the registry should be utilised is not clearly defined or quantified. The lack of consistency in how a registry is used to support a RCT, coupled with the absence of an agreed standard definition of an RRCT, could well explain the wide variability in the number of RRCTs captured in various reviews. Table 1 summarises the findings from seven reviews [1, 7–9, 16–18] captured through a Medline search for review articles using the terms RRCTs or RBTs. We also used relevant pre-existing references from our previous publications. Our search was not date defined. We only used reviews that articulated how the registry was used to support the RCT. Reviews that provided just general commentary on RRCTs but did not elaborate on the inclusion criteria of what constituted an RRCT and did not state how many RRCTs they identified were not used. Of particular note is the fact that three reviews [8, 16, 17] included post-hoc studies that also used a registry outcome for long-term follow-up. This would explain the larger number of RRCTs identified by them. Reviews that did not include post-hoc studies but included patient identification or recruitment and at least one trial outcome yielded fewer studies [1, 7, 9, 18].

Table 1.

RRCT reviews

Author	Inclusion criteria	Studies search timeline	Number of studies and nomenclature
Mathes et al. 2018 [16]	• Randomisation and • At least one outcome from registry or • Follow-up RRCTs	Inception to 2016	76 RRCTs
Karanatsios et al. 2020 [1]	• Clinical registry, and • Patient identification from registry, and • Randomisation, and • At least one outcome from registry	Inception to June 2018	17 RRCTs
McCord et al. 2022 [18]	• Patient Identification, and/or • At least one outcome from registry	2007–2018	47 RRCTs
Ahern et al. 2022 [9]	• Utilisation of a clinical registry for either: o Recruitment, and/or o baseline/outcome data	Not specified	20 RBTs in Australia
Baba et al. 2023 [17]	• Registry to assess eligibility • Patient recruitment from registry • and/or capture outcomes from registry, and/or • Follow-up RRCTS	2010–2021	90 RRCTs
Urban et al. 2025 [8]	• Randomisation, and • Patient registry, and • At least one outcome from registry, and/or • Follow-up RRCTs	Not specified	162 RRCTs (121 completed)
Shiely et al. 2024 [7]	• Clinical registry, and • Randomisation, and • Patient recruitment from registry, and • At least one outcome from registry	Studies up to 2020	24 RRCTs

The release and adoption of the CONSORT-ROUTINE as an extension of the CONSORT 2010 guidelines [19] should help improve how RRCTs are captured and reported and therefore make it easier for researchers to better appreciate the true representation of RRCTs underway or completed. It is envisaged, however, that the impact of the extended guidelines will not be immediate and it will take considerable time for the intended benefits of their implementation to be realised. Our search on the ANZCTR for Registry-Based Trials was challenging as most trials did not mention in what capacity and to what extent a registry supported the RCT, making it difficult to identify, let alone further classify RRCTs in accordance with our proposed taxonomy.

The variability used to describe trials utilising a registry has created challenges when trying to ascertain their implementation and impact. The lack of reporting in how a registry is used, coupled with the lack of a consistent definition for RRCTs, may have resulted in not all RRCTs being captured or misclassifying some RBTs as RRCTs [7–9, 17].

Furthermore, this complicates the development of a feedback loop within the research community, where shared lessons could improve the quality and impact of RRCTs. While trials like TASTE [20] and VALIDATE-SWEDEHEART [21] have guided best practice, with the TASTE trial in particular substantially decreasing the use of thrombus aspiration in Sweden from 39.8% to 11.8% [22], other relevant studies may be missed during literature searches or mislabelled due to inconsistent terminology. As a result, RRCTs appear to represent only a small fraction of trials registered in databases such as the Australian New Zealand Clinical Trials Registry (ANZCTR) and ClinicalTrials.gov [1, 2, 7–9], potentially underestimating their prevalence and impact on clinical practice. Whereby RRCTs are considered under the broader umbrella of RBTs, they offer unique efficiencies and attributes, such as cost savings and overall reduced resource requirements, that do not apply to conventional RBTs; therefore, justifying them being distinguished and reported on as a distinct subset of RBTs. Table 2 describes the key differences between RCTs, RBTs and RRCTs. RBTs that use a registry only for participant recruitment, or capture a minimal number of study outcomes or for post-trial follow-up may not realise the full benefits of an RRCT.

Table 2.

Characteristics of traditional RCTs, RBTs and RRCTs

	Traditional RCT	RBT	RRCT
Registry use	Not commonly used. When used predominantly for post RCT follow up	Any one or combination of patient identification or enrolment	Registry captures outcomes required for trial. Patients may or may not be identified and recruited from the registry. Patients must be enrolled in the registry
Research question	Commercially driven, or investigator initiated, or collaborative group trials	Public interest or commercially driven	Predominantly public interest
Eligibility criteria and generalisability of results	Commercially sponsored RCTs have narrow inclusion criteria Strong internal validity	Broad inclusion criteria Strong external validity but subject to use of registry	Broad inclusion criteria Strong external validity
Trial design	Patient randomisation	Patient randomisation	Patient randomisation
	Cluster randomisation	Cluster randomisation	Cluster randomisation
Intervention	Novel therapy against standard care or placebo	standard care or novel intervention	Usually standard of care. Potential for novel therapies if registry data capture is enhanced
Blinding	Patient and/or principal investigator or both	Blinding if registry used to support traditional RCT	Blinding not common
Endpoints	Specific protocol defined endpoints (adjudicated)	Depending on role of registry and trial funding. For commercially sponsored trials similar to traditional RCT	Preference for undisputed hard end points. Adjudication of soft endpoints not common
Sponsor	Commercial, collaborative group or investigator initiated	Government, academic not for profit entity, or commercial sponsor	Government, academic or not for profit entity

On this premise, we question the legitimacy of the general assumption that RBTs provide the same efficiencies and benefits that can be attributed to RRCTs. RCTs that utilise a registry in minimal capacity would still need to have in place the requisite infrastructure to accommodate the capture of trial-related outcome data. This would incur additional costs, slow trial set-up and necessitate greater resourcing. RBTs may encompass RCTs exploring non-standard of care therapies using a registry for patient identification only and therefore necessitating compliance against more rigorous regulatory and monitoring obligations on top of creating suitable systems for trial data capture. An example is the FANTrial (ACTRN12624000648527p), which is relying on the ANZ Fontan Registry to identify participants for an open-label RCT assessing the effect of perindopril erbumine on Fontan-Associated Nephropathy.

We therefore believe that attributing or inferring RRCT benefits under the broader umbrella of RBTs without further qualification can be misleading and could possibly confuse the reader, potentially undermining the uptake, representativeness, and impact of RRCTs. For RRCTs, the variability that exists in relation to the extent to which outcomes are being captured in the registry means that at the extremes RRCTs have data capture either minimally or fully embedded in a registry [8]. As can be appreciated, the continuum in the possible number of trial outcomes/elements captured in a registry will also impact on the potential efficiencies and cost savings that the registry can deliver and therefore should be a further point of distinction even when categorising RRCTs.

We propose that RCTs which use a registry in a quite limited capacity (e.g. patient recruitment only) are to be referred to as RBTs. We advocate that a RCT that is embedded within a registry with capture of study-related data elements/outcomes be referred to as an RRCT. Patients may or may not be identified from the registry. Patients, however, would need to be enrolled in the registry for the conduct of the trial. We propose the following two RRCT categories to allow for more precise registration of RRCTs and their subsequent identification in searches: RRCT Type 1: one or more study outcomes but not all are collected in the registry (i.e. may be collected manually in a Case Report Form), RRCT Type 2: all study outcomes are captured in the registry. Outcomes include both primary and secondary outcomes. Outcomes that can be readily or only captured by data linkage do not affect this classification and will be a subset of Type 1 or Type 2 RRCTs denoted by the letter H (Hybrid) as per Fig. 1. We support classifying RRCTs across these two categories as a subset of RBTs, which, in turn, are part of the broader category of pragmatic trials on the basis of the number of outcomes captured in the registry as this is more reflective of the cost effectiveness that is attributed to RRCTs. RRCTs that rely on additional means to capture essential study outcomes, other than via data linkage, would necessitate greater funding to accommodate this additional data capture. We also believe that the utilisation of a registry for post hoc studies should not fall under the broader umbrella of RBTs, and even more so RRCTs given that the registry did not support any aspects of the conduct of the RCT.

Fig. 1.

Illustration of the proposed relationship of RRCTs under the broader umbrella of pragmatic trials. RRCTs are considered to be further evolved from RBTs to the extent that the registry is used to support the trial. A guide for RBTs [19] identified four types based on how a registry was being utilised. Registry-based/registry-embedded/registry-nested trials—these trials utilise the infrastructure and database of the registry for many or all elements of the trial. Some of these trials would be classified as RRCTs. Registry-linked trials—these trials that obtain a dataset from one or more registries (either planned as part of the design phase for the trial, or added post-trial to obtain longer-term follow-up data). Registry-compared trials—these trials compare trial results/populations to a registry to determine compatibility/generalisability. Registry-emulated trials—these trials use data from a registry cohort to emulate a trial

Patient identification and/or recruitment would not be a prerequisite for the two categories. For studies where the registry is used to either identify or recruit participants from the registry, then these RRCTs would be identified by the following outline subset numbering, i.e. Type 1.1 or Type 2.1. Where the registry is not used for identifying or recruiting patients but for enrolment in the registry for the purpose of the study, then such RRCTs would be identified by the following numbering outline: Type 1.2 and Type 2.2. The randomisation module does not necessarily need to be embedded within the registry.

Not all data elements can necessarily be captured via the registry. Certain data elements that are not captured in the registry that can be accessed via data linkage would not impact on the RRCT Type 1 or 2 classification. To allow for the distinction of RRCTs that also utilise data linkages, we are proposing that they are delineated by the letter H to identify them as Hybrid RRCTs, a term used for such RRCTs by some trialists [23]. The hybrid identifier can be applied to any classification level of an RRCT (i.e. RRCT Type 1.1H, Type 1.2H) and would denote capture of additional outcomes via data linkage that are not captured in the primary registry.

We described two examples of RRCT that utilise a registry to varying degrees (Table 3). The TASTE trial [20] based on our proposed taxonomy is an RRCT Type 2.1H. It utilises the SWEDEHEART registry to identify, recruit, and randomise patients and captures all trial-related outcomes. The primary end point (mortality at 30 days) is obtained via data linkage to a national population registry. The per patient cost for enrolment was significantly reduced, amounting to approximately 10% of a conventional RCT [24], making it truly representative of the benefits that RRCTs can provide but which are incorrectly broadly inferred to also apply to all RBTs. The ALT-TRACC trial (ACTRN12618001480279) is an RRCT Type 2.2 based on our taxonomy comparing two standard of care treatments given at different intervals. Patients are not identified or recruited from the registry but enrolled into the registry and recruited into the study. All study-related outcomes are captured in the registry.

Table 3.

Various types of RRCTs

Study name	Trial description	USE of registry in RCT	Study Type ANZCTR or Clinicaltrial.gov	Classification	Benefits inferred
TASTE	RCT comparing thrombus aspiration followed by PCI with PCI only	Enrolled trial participants from the national comprehensive Swedish Coronary Angiography and Angioplasty Registry (SCAAR), (SWEDEHEART) registry. Outcomes collected in the registry. One study outcome collected from external registry	Intervention Two Standards Of Care (SOC) being tested	Registry Randomised Controlled Trial Type 2.1H (RRCT)	Trial data requirements are accommodated through the SWEDEHEART registry. This trial capitalised on all benefits afforded by the stricter definition of an RRCT
ALT-TRACC	Alternating oxaliplatin and irinotecan doublet schedules versus continuous doublet chemotherapy in previously untreated metastatic colorectal cancer:	Study outcomes collected in registry	Intervention. Registry based prospective randomised trial. Two SOC being tested	Registry Randomised Controlled Trial Type 2.2 (RRCT)	Patients recruited into registry concurrently with recruitment into study. Greatest benefit gained from study outcomes captured from registry

The way RRCTs are defined and registered on clinical trial registries presents a number of challenges. The non-standardised manner in how RRCTs are reported or registered means the search methodology to identify eligible studies can be quite laborious and necessitates the screening of titles and abstracts against the inclusion criteria [1, 7, 8]. Although the CONSORT-ROUTINE will help improve the quality of trials using cohort or routinely collected data, it will not substantially facilitate easier identification of RRCTs as the fundamental RRCT identifying elements are only required to be captured in the Methods section. To help address this limitation, Table 4 contains several recommendations. This should make RRCTs more readily searchable and provide a more accurate estimation of their implementation and impact.

Table 4.

Recommendations

RRCTs should be registered on trial registries in accordance with the CONSORT-ROUTINE guidelines and contain the following descriptors in their title and/or abstract:
• Name of primary registry used to support RCT
• Number of outcome measures captured in the registry
o (Type 1 or Type 2 RRCT)
• If participants identified or recruited from registry
o (Type 1.1, 1.2 or Type 2.1, 2.2 RRCT)
• If data linkage was used for some outcomes
o (Type 1.1H, 1.2H or Type 2.1H, 2.2H)

Limitations

As this is a commentary manuscript, we are presenting our perspectives based on the research that we have undertaken in this area over the years. As such, we run the risk that what we are proposing may be as a result of our selective interpretation or influenced by our own personal perspectives. We relied on our previous research and publications, and also conducted a Medline search to support our findings focussing on RRCT reviews only. In the absence of a more comprehensive scoping review, we may have missed articles that may have been relevant in informing this manuscript. Furthermore, as current registration of trials does not always adequately capture how and the extent a registry was used to support the trial, we may have misclassified trials as either RRCTs or not included some at all.

Conclusion

RBTs exhibit wide variation in how a registry is utilised to support an RCT. As such, it is at times difficult to attribute the benefits that RRCTs have to offer under the broader umbrella of RBTs, and thus RRCTs should be defined as a distinct subset of RBTs. We welcome the CONSORT extension for the reporting of RRCTs using cohorts and routinely collected data (CONSORT-ROUTINE) and advocate for a standardised approach on how they are registered on clinical trial registries like the ANZCTR. We propose a taxonomy that will aid in the identification of RRCTs and propose that the CONSORT-ROUTINE extension guidelines be extended to necessitate the capture of key RRCT identifiers that are currently only required in the methodology section. We advocate for our proposed taxonomy to also be a requirement in the title and/or abstract. RRCTs are a valuable trial method for addressing important research questions. Clear definitions and consensus on RRCT registration terms will improve recognition and impact on clinical practice. We welcome further discussion to refine RRCT taxonomy and to help guide future research.

Authors’ contributions

Karanatsios B. conceived the manuscript concept, which was further refined with P. Khic and Gibbs P. Karanatsios and Khic wrote the initial manuscript, and all authors were involved in critically reviewing it.

Funding

Not applicable.

Data availability

Not applicable.

Declarations

Ethics and approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

None of the authors have any competing interests to declare.