Primer | Clinical Trials
- Introduction
- Research question and hypothesis
- Knowledge gap
- Primary and secondary endpoints
- Sampling
- Intervention and control group
- Patient preferences
- Randomization
- Stratification
- Blinding
- Essential elements of a clinical trial
*Note: Information on this page was adapted from the Practical Guide to Designing a Clinical Trial in Surgery, JAMA Surgery; 2022.
Introduction
A randomized clinical trial represents a controlled experiment. This controlled experiment takes place over a prolonged period of time, throughout the length of the study and allows the investigators to compare a new treatment to a group of patients and compare it to the standard of care. Patients are subsequently followed to assess the effect of the new treatment and its effect on outcomes over time.
Research question
Inception of a clinical trial starts by formulating a well thought out research question that is stated in the form of a hypothesis that will be tested.
Trials are best reserved for mature questions that are supported by existing literature and can address a well-defined knowledge gap. Issues of patient safety and ethics deserve a special consideration in clinical trials, since the uncertain nature of the treatments tested or, inversely, the risk of withholding such treatment due to study protocol requirements may both be associated with patient harm. For this reason, assessing patient and clinician equipoise for the proposed research question is important, as lack of such equipoise limits trial feasibility and adversely affects the implementation of results.
Knowledge gap
The research question is formulated to inform a well-defined literature gap, an area of knowledge deficit that when addressed is predicted to have a substantial impact on clinical practice or health policy. There should be ample evidence from preliminary analyses that the proposed trial is well positioned to address this gap, along with a plan on how trial findings will be disseminated to the scientific community.
Primary and secondary endpoints
All trials designate a clearly defined single primary endpoint. This is an outcome that directly addresses to the research question and serves as the anchor for calculating the sample size, and thus determining the length of the study and many of its logistical aspects.
The primary endpoint may be a single outcome, or one predefined composite outcome of multiple events, but not multiple single outcomes. The endpoint needs to be measured in a valid and objective way by masked outcome assessors. Continuous variables (such as hospital length of stay) can be used as primary endpoints and are conceptually attractive because they may reduce the required sample size. However, most of the clinically important primary endpoint candidate events in a clinical trial, such as mortality or the occurrence of postoperative complications, are binary.
Designating a composite primary endpoint has a dual benefit of a) serving as a global summary measure of all the outcomes it captures, and b) improving power and reducing the required sample size. Composite outcomes need to be defined with caution so that their individual events are of similar clinical significance, occur with similar frequency and change towards the same direction.
In addition to the primary, secondary outcomes should be defined a priori to better describe the various effects of treatments under comparison. Post-hoc analysis (analysis of outcomes not prespecified in the study analysis plan) may be purely exploratory and serve as hypothesis-generating to drive future research. Such analysis is subject to outcome reporting bias and type I and II errors, and if pointing to a different treatment effect compared to the primary endpoint may alter the main message of the trial.
Sampling
The included study sample should be representative of the target audience for whom the research question applies. The population of individuals in which inclusion and exclusion criteria are applied must be clearly defined in a way that minimizes selection bias and maximizes generalizability. The levels of attrition and accounting for drop-outs and cross-overs are also important considerations to mitigate selection bias.
Intervention and control group
Most randomized clinical trials consist of a single intervention, such as a new treatment that may be superior to conservative management. The control group receives either no treatment or the treatment that represents the existing standard of care. For instance, in the OVER trial, the intervention group received endovascular aneurysm repair, whereas the control group received open aneurysm repair that had been the standard type of repair of infrarenal abdominal aortic aneurysm for decades.
Patient preferences
Strong patient preferences for one of the treatment alternatives may affect feasibility and introduce bias, particularly in studies where patients are required to sustain an effortful role. Assessing patient equipoise early the planning stage of the study and informing patients on the clinical equipoise are key aspects for assuring patient engagement. Performing a patient preference clinical trial is a more sophisticated and complicated alternative, in which patients are provided upfront with all of the robustly researched and efficient treatment alternatives.
Randomization
Randomization is the process of randomly allocating subjects to the treatment arms, which optimizes baseline comparability of groups and minimizes confounding.
The process of randomization should assure allocation concealment. Sealed, opaque, tamperproof envelopes, or a central randomization facility (for multicenter trials) are methods typically used to assure that the patient allocation is truly random.
Simple randomization (each participant is equally likely to be randomized to either group) is preferred in most surgical trials, although other methods also exist. For example, in the Ladies trial evaluating surgical management of complicated acute diverticulitis, the pre-trial best evidence indicated that Hartmann’s procedure and resection with primary anastomosis were equal in terms of postoperative outcomes, while laparoscopic lavage was the new surgical treatment up for evaluation. For this reason, a three way 2:1:1 randomization was performed.
Stratification
Stratified blocked randomization is appropriate to minimize imbalance of an important predictor of the outcome among the study groups. For instance, having a recurrent hernia increases the likelihood of future recurrence after a surgical intervention. In a trial to compare two methods for hernia repair stratifying the patients in two strata (those with and those without prior recurrence) will help balance the groups for this important predictor of outcome. The stratification should ideally be a decision made a priori. Only a small number of baseline variables, typically up to three, can be balanced using this technique. Post-hoc subgroup analyses should be considered exploratory in nature.
The value of stratified randomization diminishes as the sample size increases, because random assignment will eventually assure similar distribution of baseline variables between the study groups. Varying the size of the block randomly (for instance define block size from 3 to 6) assures that the end of the block will not be anticipated, assuring unpredictability of treatment allocation.
Blinding
Blinding is a process that prevents study participants, investigators and personnel that adjudicates the outcomes from knowing which treatment was received. Blinding can be single (only the patient is blinded) or double (involves both the patient and the treating physician).
Blinding in a randomized trial is as important as randomization, since it greatly reduces the likelihood of bias during the follow up and outcome adjudication. Unfortunately, blinding in surgical trials is often problematic and sometimes outright impossible. For instance, in a trial comparing transfemoral vs. open aortic valve replacement it is impossible to blind either patients or surgeons as the operative approach between the procedures is substantially different. Therefore, a requirement for surgical trials is to ensure blinded outcome assessment.
Essential elements of a clinical trial
1) Define a research question and primary outcome.
2) Define inclusion and exclusion criteria.
3) Achieve baseline comparability of patient groups.
4) Implement appropriate blinding.
5) A priori, define and choose masked outcomes that can be measured validly, reliably, and objectively.
6) Calculate a sample size that accounts for expected attrition, drop-outs, and cross-overs and is based on the primary endpoint.
7) Manage and analyze the data appropriately.
8) Conclude only what is supported by the data results.
Contact us: