Some Disadvantages of Multicentre Randomised Trials 229
Some Potential Disadvantages of PMAs 231
Basic Steps in a PMA 232
A prospective meta-analysis (PMA) conducts single-site randomised trials in conjunction with an active, contemporaneous approach to a meta-analysis. This contrasts with the common approach of investigators designing, conducting, and publishing a trial without any further plan. Frequently, they passively wait for the results to be combined with other similar randomised trials in a traditional, retrospective meta-analysis as part of a systematic review. Alternatively, they may eventually plan a traditional meta-analysis that includes their trial. However, in a PMA, an investigator designs a meta-analysis while planning and conducting a component trial that will contribute to the PMA.
A prospective meta-analysis (PMA) circumvents some of the challenges to traditional, retrospective meta-analyses. A PMA is a meta-analysis of randomised trials that were recognised and determined to be appropriate for the meta-analysis before the findings of any of those trials are known. Unaware of the findings of individual trials, investigators stipulate hypotheses a priori and establish prospective application of study selection criteria. Moreover, PMAs allow a priori declarations of planned analyses, including any subgroup analyses, before the findings of the individual component trials are known. Thus PMAs skirt potential problems in interpretation related to post hoc, data-dependent analyses in traditional meta-analyses.
In addition, all of the sites contributing to a PMA must agree on the intervention and outcomes. Thus problems are averted compared with retrospective meta-analyses with discrepancies in interventions being combined and disparities in outcomes being measured. Without the collective determinations of the PMA collaborative process, the independent sites might generate irreconcilable data that could not be combined in a scientifically acceptable meta-analysis.
PMAs are commonly implemented by a collaborative group, and they usually collect and analyse individual patient data (IPD). Unlike traditional multicentre trials, they allow variation in the protocols of the included trials, while maximising power in the preplanned meta-analyses. A PMA would usually require fewer resources than a multicentre randomised trial while cultivating opportunities for research participation. Thus PMAs may foster increased research capacity through collaboration.
Moreover, PMAs expand opportunities for authorship of peer-reviewed publications. With a PMA, contributing institutions can publish their own trial results. Indeed, the PMA methodology enables several investigators at each site to qualify for authorship of their site report.
PMAs present a host of efficiencies and opportunities. Clinical researchers need to be familiar with the concepts behind a PMA and, when appropriate, capitalise on its advantages. We encourage interested readers to consider the approach.
In planning a randomised trial, investigators frequently discover that they have low power with their anticipated sample size. They lack a sufficient number of potentially available participants at their institution. In these situations, some methodologists and ethicists will advise those investigators to abandon such an apparent ‘low-power’ trial. Some ethics review boards (ERBs) deem low-power trials as unethical. However, we take a different view as described in Chapter 11 . We believe, with three caveats, that these so-called underpowered trials can be acceptable because they can ultimately be combined in a traditional, retrospective meta-analysis. Traditional meta-analysis is not, however, the only option for addressing low power.
Multicentre Randomised Controlled Trial
Another solution to low power is a multicentre randomised controlled trial (MCRCT). Essentially, it replicates a single-site randomised controlled trial; extra sites add numbers of participants and thus statistical power. This usually involves a collective protocol, standard eligibility criteria, standard interventions, shared randomisation procedures, common data-collection forms, shared data management, standard operational procedures, collective data analysis, and centralised oversight. We deliberately avoid discussion of issues of the design and planning of MCRCTs in this book for several reasons. First, good descriptions are found elsewhere. Second, novice investigators may not have the resources necessary to initiate an MCRCT. They may have the opportunity to respond to a funder’s solicitation, but the design and planning will likely have been done by that funding source. Third, some potential disadvantages make MCRCTs less attractive to single-site investigators who are interested in collaborative trials.
A PMA is a planned, active approach to a meta-analysis. It conducts single-site randomised trials in conjunction with an active, contemporaneous approach to a meta-analysis. This contrasts with the common approach of investigators designing, conducting, and publishing a trial without any further plan. Frequently, they passively wait for their results to be combined with other similar randomised trials in a traditional, retrospective meta-analysis as part of a systematic review. However, in a PMA, investigators design a meta-analysis while planning and conducting a component trial that will contribute to the PMA.
This chapter provides a primer on PMAs and familiarises readers with the concept. We discuss some of the disadvantages of MCRCTs, which remain the main alternative scientific study design to a PMA. Next, we review the advantages and disadvantages of PMAs vis-à-vis MCRCTs, retrospective meta-analyses, and single-site RCTs. Finally, we describe the basic steps in a PMA and encourage readers to consider this approach. We will not provide comprehensive, detailed methods for planning and conducting a PMA.
Some Disadvantages of Multicentre Randomised Trials
Well-conducted, multicentre randomised trials are the gold standard for evidence-based medicine. Yet they can be expensive, cumbersome, and unrewarding for individual investigators. Disadvantages can emerge in the design, conduct, and reporting of an MCRCT.
The need for a uniform data-collection form, or case report form (CRF), in an MCRCT can pose logistical challenges. In a multicentre trial, one centre may consent to inclusion of another centre’s desired data-collection items if that other centre consents to its own desired data-collection items. When many centres are involved, these multiple quid pro quo negotiations frequently yield a long and unwieldy CRF. These negotiations may delay implementation, increase the cost of the trial, and potentially decrease the quality of the data gathered. Moreover, long CRFs unnecessarily burden participants, which discourages participant retention.
The requirement for approvals by multiple institutional ERBs in an MCRCT creates more hurdles. Usually an ERB for one centre will not allow the trial to begin at that centre until all the other ERBs involved in the multicentre trial also approve the trial at their respective centres. Thus an individual centre may have all its own approvals and be ready to begin enrolling, but be prohibited from doing so because it awaits ERB approvals from all the contributing centres. This consequently delays commencement and increases the cost of the MCRCT due to ‘treading water’ or idling trial costs. Also, the delays can frustrate the research staff. The slowest ERB becomes the ‘rate-limiting enzyme’ for starting the trial.
The most industrious and efficient investigators can suffer further disadvantages in an MCRCT. If investigators at an individual centre proficiently and promptly complete their enrolment and follow-up, their rewards will be delayed. With other centres lagging, analysis and manuscript activities must wait for those other centres to complete their data-collection activities. Thus costs are increased, and time to publication is delayed while waiting on the slowest centre to finish.
Investigators may be less invested in an MCRCT than in a trial done only at their institution. First, they probably are not as intimately involved in the analysis and manuscript writing as they would have been had they conducted a single-site RCT. Second, in multicentre trials, authorship is often limited to the principal investigators at each site. Investigators may only be identified in group authorship in the final manuscript (often represented by an asterisk and each author’s name in small font at the end of the paper). Group authorship does not carry the same weight as having one’s name in the byline. Although it certainly counts as being a valid author, many find it a bit anticlimactic after the long, hard, arduous tasks of recruitment, data collection, and follow-up. Indeed, some hard-working investigators involved in MCRCTs are frustrated by the lack of recognition.
When the resources are available and investigators accept the challenges, we certainly encourage the conduct of large multicentre randomised trials. However, in many circumstances, PMAs may be a less costly, more efficient, faster, and more satisfying approach to gathering evidence ( Panel 21.1 ).
With a PMA, individual contributing sites can have a more concise and focused data-collection form. All the sites in a PMA must agree upon the intervention groups and the measurement of the primary and secondary outcomes, along with perhaps some baseline measurements. Other than those uniform data requirements, individual sites are free to collect only information of interest to them. Unlike in an MCRCT, they do not have to collect data of interest to other sites. Thus their data collection becomes more specific, and their trial, in general, is more streamlined. That leads to lower costs, quicker completion, and potentially better data quality. The shorter data-collection form is beneficial to participants. Treating them responsibly by being parsimonious with data gathering will likely improve follow-up and retention of participants.
PMAs simplify the role of an institutional ethical review board (ERB). Instead of waiting for all the other ERBs to approve the trial, as in a multicentre trial, a contributing institution in a PMA simply secures approval from their institution’s ERB. That can remove exogenous obstacles, reduce costs, and hasten completion.
PMAs expedite time to publication for a contributing institution. A multicentre trial must wait for the slowest centre to finish their data collection before the team begins to analyse the results and write their report. However, a PMA is more expeditious. When a contributing institution completes its data collection, researchers can analyse their own data, write the manuscript, and then submit it for publication. Therefore results reach the medical literature more quickly from their single-site trial than if they had been part of a multicentre trial.
PMAs can expand opportunities for authorship of peer-reviewed publications. With a PMA, contributing institutions can publish their own results. Indeed, several investigators at each site conduct the trial and qualify for authorship of the individual site publication. Typically, with a PMA, an individual institution publishes their results in a journal and then also contributes, through group authorship, to the publication of the eventual PMA. This bestows byline authorship benefits to the contributing investigators at each site. That visible authorship at the front of the paper, immediately following the title, is an immensely satisfying reward for their efforts.
The individual trials contributing to a PMA may yield improved trial conduct as a byproduct of separate, individual trial authorship. Some investigators have speculated that this responsibility of authorship may improve study conduct because it ‘may optimize individual site protocol execution (proper administration of medications, maximizing participant retention, and collecting all data points) and motivate individual sites to conclude the project in a timely fashion’.
A PMA expands opportunities for research mentoring. The PMA may foster increased research capacity through collaboration. PMAs can involve both senior and novice researchers. Novice researchers from smaller sites can benefit from involvement in a PMA by gaining access to well-developed protocols and the opportunity to recruit smaller numbers of participants. Thus the PMA provides a vehicle for mentorship from senior researchers and allows multiple sites the opportunity to gain experience and scientific expertise in the conduct of randomised trials. Indeed, a primary aim of one PMA was to create the opportunity for mentorship to new investigators and to promote more ‘rapid and effective growth of the … research community’.
A PMA requires fewer resources than an MCRCT. When sites conduct their own trials, they develop protocols that best match their interests, the accessibility of trial participants, and the availability of institutional resources. This decentralisation lowers cost by avoiding centralised oversight. Furthermore, individual site investigators can seek smaller grants for their sites rather than attempting to pursue ‘a large, less readily obtainable grant for an unwieldy and costly multi-center trial’.