Chapter 05 / Understanding Trial Designs Through Real Project Work / Paid beta preview

Chapter 5: Observational Studies and Real-World Evidence

How observational studies, real-world data, and real-world evidence become project work. ## 5.1 When the Trial Does Not Assign the Treatment The first time Lauren Brooks heard a senior leader say, "We can just do real-world evidence," she wrote the phrase in her notebook and circled the word just. That was the right instinct. In clinical development, just is often where the risk is hiding. The meeting was about REGISTRY-RA-10, a prospective rheumatoid arthritis registry designed to follow patients receiving routine care across community rheumatology practices and academic clinics. Rheumatoid arthritis is a chronic inflammatory disease that can change slowly, flare unpredictably, and affect function, pain, fatigue, work, and quality of life in ways that do not always fit neatly into a single clinic visit. The proposed registry would collect physician assessments, laboratory values, medication exposure, disease activity measures, patient-reported outcomes, safety events, treatment changes, and longitudinal follow-up over several years. The sponsor had several questions in mind. Medical wanted to understand treatment patterns and outcomes in patients who looked different from the narrower population enrolled in a prior randomized [ICH E9] trial. Safety wanted better visibility into infections and other events that might emerge with longer exposure. Health economics and outcomes research, often called HEOR, wanted evidence that could support payer discussions about persistence, switching, resource use, and patient function. Regulatory wanted to know whether any part of the registry might someday support a postmarketing commitment, a safety assessment, or a carefully limited labeling discussion. Those are all legitimate questions. They are not all the same question. That distinction matters before a single site is activated. An observational study is a study in which participants are observed and outcomes are assessed, but the investigator does not assign the participant to a particular treatment or intervention as part of the study. A patient may receive a drug, biologic, device, procedure, behavioral intervention, or no intervention, but that exposure comes from routine care or real-life circumstances rather than assignment by the research protocol. ClinicalTrials.gov [ClinicalTrials.gov] distinguishes this from an interventional study, or clinical trial, where participants are assigned prospectively to an intervention according to a protocol. That simple distinction carries a great deal of project work. In CARDIA-301, the cardiovascular trial from Chapter 1, the protocol assigned treatment and controlled many parts of the patient journey. In DERM-450, the dermatology study from Chapter 2, standardized images had to be collected because the trial design required them. In VAX-PED-102, the pediatric vaccine study from Chapter 3, randomization, blinding [ICH E9], dosing, and safety follow-up were protocol-driven obligations. In RESP-640, the respiratory trial from Chapter 4, supply and randomization controls mattered because the protocol assigned product under a blinded design. REGISTRY-RA-10 was different. The registry did not tell the rheumatologist which medicine to prescribe. It did not randomize patients to therapy A or therapy B. It did not require the physician to switch a patient at a fixed time. It did not create a placebo arm. It observed what happened in care that was already being delivered, under a study plan that defined what data would be collected, when, from whom, and for what purpose. That does not make the work easy. It makes the work different. A randomized [ICH E9] trial protects interpretation through design. Randomization helps balance known and unknown factors between treatment groups. Blinding [ICH E9] can reduce the influence of expectations. Protocol-defined visits and assessments can make data more consistent. Eligibility criteria define the population tightly. That structure is expensive, burdensome, and sometimes less reflective of routine care, but it exists for a reason. An observational study usually gives up some of that control. In exchange, it may see patients and care patterns that a conventional trial misses. It may include older patients, patients with comorbidities, patients treated in community settings, patients who switch therapies, patients who stop and restart treatment, and patients who live with the disease outside the tidy rhythm of trial visits. That is the opportunity. It is also the danger. Real-world does not mean automatically true. Routine care data can be incomplete, inconsistent, biased, delayed, coded for billing rather than research, and shaped by physician preference, patient access, insurance coverage, geography, disease severity, and many other forces. A large database can be impressively wrong if the study question is weak, the variables are missing, or the comparison is unfair. Daniel Liang told Lauren this after the first REGISTRY-RA-10 design meeting. "Do not let anyone treat this registry as a cheaper randomized [ICH E9] trial," Daniel Liang said. "It is not a discount version of something else. It is its own evidence machine. The question is whether we are building the right machine for the decision we want to support." That became the chapter's starting point. Real-world data, or RWD, are data relating to patient health status or health care delivery that are routinely collected from sources such as electronic health records, medical claims, registries, patient-generated data, and digital health technologies. Real-world evidence, or RWE, is clinical evidence about the use, benefits, or risks of a medical product derived from analysis of RWD. FDA uses those

...