How to Survive and Succeed in the Design and Execution of a Collaborative Randomized Controlled Trial

ABSTRACT

This article guides the design and execution of successful collaborative randomized controlled trials (RCTs). Aimed at researchers looking to impact clinical practice, it highlights the advantages of RCTs over observational studies in driving clinical advancements. The article details essential steps in trial development, starting with assembling a core team of a principal investigator, experienced collaborator, methodologist, patient representative, and coordinator. Critical strategies for defining impactful research questions, selecting adequate primary endpoints, and developing a robust protocol are discussed. Additional emphasis is placed on patient‐centered outcomes and data integrity, supported by carefully designed electronic case report forms. The article also covers strategies for recruiting collaborators, and managing ethical approvals across multiple centers. Ultimately, it offers insights from the author's experience, encouraging new researchers to overcome challenges in launching their first RCT and reinforcing the role of high‐quality, multicentric research in enhancing patient care and clinical outcomes.

1. Introduction: The Crucial Role of Collaborative Randomized Controlled Trials (RCTs)

If you are reading this article, you are probably passionate about clinical research. And if you feel that calling, you have a mission: to change clinical practice and to improve patient management. Observational studies serve, broadly speaking, to generate hypotheses. However, they keep clinical practice the same; they rarely allow us to move forward. They are a tool for designing experimental studies such as randomized controlled trials (RCTs). However, we have barriers to our first RCT. It may seem that it requires a lot of time and effort, and you cannot combine it with your clinical duties. Moreover, it is not enough for the design to be experimental; it must also be collaborative. It is almost impossible to draw firm conclusions with single‐center RCTs; you need collaborators: networks of enthusiastic colleagues to accompany you on this journey. Thus, designing your first multicenter trial takes work. But the truth is that, after all, every time we do something the first time, it seems complicated. My first clinical trial was straightforward conceptually; we compared in a single‐center RCT normal saline versus lactated Ringer's solution‐based fluid resuscitation in 40 patients with acute pancreatitis [1]. I finished the protocol in October 2011, and it was accepted in my favorite scientific journal, the UEG journal, in September 2017, after almost 6 years that included many doubts and challenges [2]. However, the next RCT we conducted, the WATERFALL trial, involved 249 patients from 18 hospitals in 4 countries and took only 3 years from protocol to publication [3]. RCTs have a learning curve, and by the time you do the first one and understand the tips and tricks of this noble art, everything is much easier; you will become an RCT veteran. This article aims to give you tips that will allow you to overcome the barriers to making your first RCTs (summarized in Figure 1), accelerate your learning curve, and pass on to you what I have learned after designing 6 RCTs, most of them collaborative. This is a special article commissioned by United European Gastroenterology (UEG) journal as a recipient of the UEG Research Prize 2024.

FIGURE 1.

Solitude: Do not embark on this adventure alone. Surround yourself with collaborators to design the study, including at least one colleague who is an expert in your field, a methodological expert, a patient representative, and a trial coordinator. Wrong question: The research question must be engaging, unanswered, and exciting to attract collaborators. Wrong primary endpoint: Make sure that your main endpoint is robust, that the scientific community will accept it as valuable, and that it is relevant from the patients' point of view. Wrong sample size calculations: Make an honest, evidence‐based sample size calculation without moving the data to obtain affordable sample sizes. If you move the data, the reviewers will detect it (and reject grants or publications), or your study will not have enough statistical power to see differences. Problems with funding: Clinical trials need financing, so make a great effort to apply for all available grants. Few collaborators: Use all possible channels to recruit collaborators. Email your colleagues, ask scientific societies to spread the word, and use social networks. Develop your leadership. Inefficient collaborators: Detect early collaborators who fill in the wrong data in the case report form and correct them by making them aware. Missing data is unacceptable in a clinical trial. If they persist in providing poor‐quality data, they should leave the study. Slow recruitment: react to slow recruitment, understand why patient acquisition has slowed, talk to collaborators, and, if necessary, recruit more centers or change the inclusion or exclusion criteria if you are at an early study stage.

1.1. Design Phase

1.1.1. Create Your Hard‐Core

The basic structure of my RCTs' hardcore is the principal investigator, the experienced collaborator, the methodologist, a patient representative, and the coordinator.

The principal investigator (you) leads the team and is the ultimate project manager, the final decision‐maker. Your role is pivotal in the success of the RCT, and your decisions will shape the direction of the study. This leadership position empowers you to steer the project toward its goals and ensures that the study is conducted with the highest standards of quality and ethics.

The experienced collaborator is someone with research experience in the field of the study. They provide a second opinion on the design and execution of the study, often spotting different problems, strengths, solutions, and shortcuts than the principal investigator. This creates a very good synergy and enhances the quality of the research. My experienced collaborator in my fluid resuscitation studies is James L. Buxbaum, a good friend with extensive experience in this topic.

No matter how much methodology you know, and even if you are a statistical superhero, you need a methodologist to help you because the methodology is the skeleton of the project, the basis, which must be perfect. The methodologist's expertise and guidance will ensure that the study is based on a robust and sound methodology, giving you and your team the confidence that the study is designed and executed with the highest scientific standards. The patients' representative will give the vision of this group, enriching the study in a significant way, see the following section. Finally, you need a coordinator to handle the tasks, problems and bureaucracy of the different centers. The coordinator may be a professional employed with the study funds or a PhD student. There are very successful models of clinical trials in which a PhD student has coordinating roles and responsibilities, as is the case in the Netherlands with the Dutch Pancreatitis Study Group and other research groups in that country that have conducted complex trials that have changed clinical practice.

In addition, you can bring in professionals from different specialties with different backgrounds and points of view, such as cardiologists, anesthesiologists, or nephrologists in fluid resuscitation studies, as examples.

1.1.2. Include Representatives of Patients or the General Population

It is increasingly important to have the support of patients themselves [4]. In our WATERLAND trial [5], an RCT comparing normal saline versus lactated Ringer's solution‐based fluid resuscitation in acute pancreatitis, we have the support of Carlos, a patient with recurrent pancreatitis. Carlos assessed the study protocol, will assess the data and results, and will participate in the final article drafting so that he can give his opinion: the point of view of the people we want to help. You can also rely on patient societies or other non‐scientific institutions.

1.1.3. Choose an Appropriate Research Question

The research question around which your RCT orbits is essential. It must be a relevant topic, unresolved by previous RCTs, that will improve how we manage our patients. If there are previous RCTs, yours must provide something substantially novel, either because of a more significant number of patients than previous ones, allowing you to explore more powerful endpoints, or because it is better designed, resolving methodological flaws in previous studies. The relevance and interest of the question you want to solve is essential in collaborative studies: it must be attractive to your collaborators. If we simplify everything as much as possible, the research question boils down to the RCT treatment arms and the main endpoint. Treatment arms should be clear and simple; avoid combinations of treatments, for example, don't compare aggressive fluid resuscitation associated with subcutaneous low molecular weight heparin versus. moderate fluid therapy plus oral aspirin in acute pancreatitis; we will not know where the benefit comes from if there is any.

1.1.4. Choose the Right Study Endpoints

Choosing your main study endpoint is essential and is one of the most important decisions you must make. There are three possible choices for your primary endpoint [6]:

1. “Hard” endpoints: variables important per se to the patient, those that measure morbidity or mortality. These are usually relatively infrequent endpoints, for example, mortality in acute pancreatitis (frequency 4%) or organ failure lasting for more than 48 h (7% of the patients, associated with maximum patient suffering, intensive care unit admission, invasive procedures, and a 50% chance of mortality) [7]. This leads to huge sample sizes, but the great importance per se of these variables means that demonstrating differences in them is of the utmost importance.

2. “Soft” endpoints, such as hospitalizations, symptomatology (e.g., patient‐reported outcomes [8]), and changes in selected physiologic measures (e.g., c‐reactive protein [1]) [6]. These are variables of lesser importance than the above. The less relevant they are for the patient, the worse: it is less attractive to demonstrate lower levels of C‐reactive protein or interleukin‐6 than improving symptoms or quality of life.

3. Composite study endpoints [6]: combining different hard endpoints to increase the frequency of events so that you don't need thousands of patients to see changes in the primary endpoint. For example, “major complications (new‐onset multiple‐organ failure or multiple systemic complications, perforation of a visceral organ or enterocutaneous fistula, or bleeding) or death” [9].

The primary endpoint determines the sample size and the relevance of your study; if you want to know more, you can read my clinical research blog post “The one and only Unicorn Primary Endpoint” [10].

As for the rest of the variables in the study, strike a good balance between having an adequate amount of patient data and going overboard with the number of total variables that make including a patient a nightmare. Very complex and cumbersome studies will have slow and difficult recruitment, which may slow or even stop the study from progressing.

1.1.5. Make a Near‐Perfect Protocol and Publish It

You have decided on the treatment arms and primary endpoint with a good balance between relevance and event frequency. Now, we calculate the sample size based on the available data. This is more important than you think. When you apply for a grant or submit your article for publication, the reviewer will carefully analyze how you have calculated the number of patients needed. As a grant reviewer, I have rejected projects because of poor sample calculation. If the calculation is incorrect, it can miss differences if there are fewer events than expected, or it can lead to recruiting more patients than necessary if there are more events than expected.

Make a protocol based on the SPIRIT guidelines [11]. Some journals have templates that make your work easier, for example, Trials [12]. Register the study in databases such as clinicaltrials.gov. In addition, a comprehensive statistical analysis plan (SAP) should be created based, for example, on the recommendations of Gamble et al. [13]. Some issues require special attention. If there are variables that may strongly influence the primary endpoint, do stratified randomization based on this: for example, in WATERFALL, the randomization was stratified by center, presence of baseline hypovolemia, and baseline inflammatory response syndrome [3]. Decide how you will handle missing values, whether you will do interim analysis, and how it impacts sample size; detail how you will analyze the data. Pay close attention to all ethical issues, including safety variables and what to do if safety problems are detected. Include this SAP as an appendix in the protocol and publish it in an open‐access format. You have the recently published protocol for the WATERLAND trial [5] as an example. Publishing the protocol and SAP guarantees that no cheating will occur later by changing the variables to be analyzed to look for significant results that were not planned. This opens the possibility of publishing in better journals, but more importantly, it increases the transparency of your study.

1.1.6. Invest a Lot of Effort in the Electronic Case Report Form (eCRF)

The eCRF is not only an online form for the different participating centers to enter data; it is a fundamental tool to avoid mistakes. With the eCRF, we give information to the collaborator about the definition of the variables; we put ranges in the variables to prevent mistakes when transcribing values so that if a result is not logical, for example, an age of 150 years or a potassium of 14, an alarm is triggered so that it is corrected. We can program logical alerts; for example, a patient with organ failure cannot have mild pancreatitis; if these data are entered, an alarm is triggered, explaining to the collaborator that they must correct the mistake. Alarms may also be raised if essential variables are missing. For quality data with as few errors as possible, invest a lot of effort in the eCRF. It is one of the most energy‐consuming tasks but very worthwhile. I use the RedCap platform, which is very powerful, cheap, and has a fast‐learning curve.

1.1.7. Look for Grants to Finance Your Study

Apply for grants at different levels: from national and international scientific societies, public grants from your country, etc. If you want advice on how to apply for grants, read my blog post, “Writing the perfect grant is like planting a tree” [14].

1.1.8. Be Prepared for Bureaucracy

The bureaucracy surrounding the approval and conduct of clinical trials is challenging. I recommend you read the article “Reducing Bureaucracy in Clinical Trials” by Ianiro, Ollivier, and Ricciardiello [15].

1.2. Phase of Recruiting Centers and Collaborators

In reality, phase 1 and 2 overlap, and the longer you take to recruit collaborators, the worse: it will delay permission from ethics committees, and some centers may not even be able to join if you delay too long; this happened to us in WATERLAND, where many European centers could not participate because of the high bureaucratic demands of the Clinical Trial Information System (CTIS). This body approves RCTs in Europe. Get collaborating centers to apply to their ethics committees as soon as possible!

1.2.1. Choose Your Collaborator Recruitment Strategy

There are two strategies for recruiting partners: I call them “open” and “closed.” The open strategy consists of publicly disseminating the study and inviting any center to join the project. In general, it requires relatively simple studies that do not require collaborators who are highly specialized in the topic to be investigated or do not need much research experience. WATERFALL and WATERLAND used open calls. To advertise my studies, I use my social networks (mainly the @DeMadaria account of X, formerly Twitter) and the social networks and newsletters of national and international gastroenterology, surgery, or pancreatology scientific associations.

Some studies, such as those requiring complex diagnostic and therapeutic techniques, are not susceptible to open calls. You may also not have the logistical capacity to coordinate more than a few centers. In such “closed calls,” it is better to pre‐select and contact the centers directly, ensuring that only a few highly specialized collaborators are involved.

1.2.2. Ensure That Your Collaborators Feel That the Trial is a Project That Belongs to the Whole Team

You must understand that this is not your project for which you are looking for collaborators to help you; it is a collaborative project that belongs to the whole team. Do a great effort to make collaborators feel essential to the project. An attractive name, a logo that provides a brand image, a language in the invitation and in every communication that makes evident the role of the collaborator as an essential part of the trial, clear rules of authorship, communication to encourage recruitment, messages of congratulations to good recruiters and of support and advice to slow recruiters, generosity in authorship, offering post hoc studies with the final database of the study. All this creates a cooperative project environment that makes collaborators feel part of the research effort. It is not always the centers that recruit the most that are academically the strongest. Especially congratulate collaborators from academically less powerful centers who are good recruiters; they can be the seed of new research groups.

1.2.3. Nurture and Train Your Collaborators

Given that you already have the relevant question, the perfect primary endpoint, a majestic protocol, and the collaborators, you now need to focus on ensuring that your partners effectively recruit patients and that the data is of the highest possible quality. Part of the work is already done, with a carefully crafted eCRF full of definitions, warnings, and alarms. Contributors need to feel part of an epic, engaging, collaborative project that will improve patient care and give them a valuable publication. The first thing is to make the rules of authorship clear. I include them in the study protocol; see WATERLAND as an example [5]. You must train your collaborators in the study; having a teleconference meeting to explain it, fill in fictitious eCRF records, and answer questions is good. You must respect and pamper your collaborators; they are the basis of your collaborative RCT. You must verbalize that there should be all the data in the eCRF and that the data entered should be checked and revised to avoid transcription mistakes or misconceptions.

1.3. Patient Recruitment Phase

1.3.1. Early Data Monitoring

Start monitoring data early. A source of problems and anxiety for the principal investigator and coordinator are (A) inefficient collaborators in entering data and (B) inefficient collaborators in recruiting patients.

In my case, I prefer slow recruiters to poor data entry quality: it is crucial to prevent contributors from causing many variables to have missing data or entering mistakes, which distort the associations we are trying to detect. It is prevalent for many sites to recruit few or no patients. In WATERLAND, only 37% of the centers recruited 15 or more patients, and 34% recruited 0 or 1 patient. Monitor the data early to detect collaborators that are inefficient in data entry, talk to them, educate them, and help them improve. If they persist, you may need to stop them from recruiting more patients to ensure data quality. The trial coordinator is central to this work.

1.3.2. React to Slow Recruitment

One of the biggest concerns in an RCT is slow recruitment. There are several possible causes. Sometimes, the fault lies with you: you have designed a very complex study, each patient takes too much time to work with, and collaborators throw the towel. In this case, there is little you can do. Another possibility is that the inclusion and exclusion criteria are very restrictive, so only a few patients can be recruited. In that case, if it seriously jeopardizes the conduct of the study, consider changing the requirements and opening them up, modifying the protocol, and communicating it transparently. In other cases, collaborators are demotivated; communication with them should be improved, highlighting the quality of the study, the relevance of the issue it addresses, and the potential quality of the publication. Sometimes, natural disasters affect RCTs, such as pandemics, in which new respiratory viruses spread out of control. Analyze the cause of the delay, talk to the collaborators, and react.

1.4. Analysis and Publication Phase

Thoroughly clean your database and look for inconsistencies, outliers, and missing data. Try to obtain this data to improve its quality. Never invent or delete data to favor your hypothesis. If your RCT is open‐label, make sure that the statistician's analysis of the data is blinded to the treatment arm (having made this clear in the SAP). If you conduct analyses not planned in the SAP, make it clear in the manuscript: describe them in a section called “post‐hoc analyses.” As a reviewer of RCT manuscripts, I study whether the author has followed their own SAP; if they haven't, it may mean a rejection from a major journal; better to make it clear from the start. Please don't force the data, don't run away from negative results, and don't try to make the data show something that doesn't exist to get a better journal; reviewers like me smell those fishing and spin maneuvers [14] from a distance. Any result from a clinical trial is valuable if it is derived from a robust methodology. Moreover, it doesn't matter if your expectations are met or your hypothesis is confirmed; only truth matters.

2. Conclusions

While conceptually simple, we have seen that answering a clinically relevant question using an RCT is complex and time‐consuming. However, as we gain experience, RCTs not only become more accessible but also profoundly addictive—a healthy addiction and one that helps patients. A well‐designed multicenter clinical trial with a relevant question, a solid primary endpoint, and a protocol published in open‐access with its accompanying SAP and which reaches the predicted sample size is usually published in high‐impact scientific journals, whether positive or negative. It also creates fruitful and lasting contacts and even friendships. With each RCT you do, your network of collaborators will become closer and more efficient, and you will grow together, producing quality science.

Conflicts of Interest

The author declares no conflicts of interest.

Data Availability Statement

Data sharing not applicable to this article as no datasets were generated or analyzed during the current study.