Merck Manual

Please confirm that you are a health care professional

honeypot link

Clinical Trial Essentials


Byron J. Hoogwerf

, MD, Cleveland Clinic

Last review/revision Aug 2021 | Modified Sep 2022
View Patient Education

Clinical trials are research studies designed to assess the efficacy and safety of an intervention. The intervention is most often a drug but can also be a device, such as a pacemaker or stent, a surgical procedure, or a diagnostic tool, such as a blood test. Clinical trials typically involve patient volunteers, but sometimes they involve healthy subject volunteers. Thousands of clinical trials are conducted each year around the world and may take place at a variety of locations, including universities, hospitals, clinics, physicians’ private offices, and professional clinical research sites. In the beginning of 2021, there were over 374,000 studies registered with, a U.S. National Library of Medicine database of privately and publicly funded clinical studies conducted around the world. Several thousand patients or more typically participate in clinical trials for each new drug before it becomes available to the general public.

The International Council on Harmonisation (ICH), previously called the International Conference on Harmonisation, brings together the major regulatory authorities of the world to bring uniformity to the processes and documents required for new drug development. This uniformity helps allow a clinical trial done in one country or group of countries to be used in submission to regulatory authorities in other countries. Importantly, the ICH guidelines are safeguards to ensure that quality, safety, efficacy, and regulatory obligations to protect public health are met. However, for a variety of reasons, a drug or device approved in one country or region may not be approved in others.

Investigators, also called clinical researchers, are the people who conduct clinical trials. Investigators are usually physicians, but they may be other health care professionals, who are paid to conduct the trials by pharmaceutical, biotechnology, or medical device companies. Other sources of support may include personnel (eg, statisticians) and resources (eg, computer, information technology). Investigators may also conduct trials supported by professional organizations such as the American Diabetes Association or the American Heart Association. Supporting agencies are called sponsors Study sponsor Clinical trials are research studies designed to assess the efficacy and safety of an intervention. The intervention is most often a drug but can also be a device, such as a pacemaker or stent... read more . Investigators may also be supported by government agencies such as the U.S. Department of Veterans Affairs (VA) or the National Institutes of Health (NIH). The NIH often sends out requests for a research proposal for a trial concept and potential investigators respond, often competitively seeking support to do the proposed research, by describing their expertise in the area, ability to recruit research subjects, and ideas (eg, study designs) they propose to answer the research question. Sometimes small clinical trials called investigator-initiated trials (IITs) are conducted by individual physicians who have an academic interest in the disorder being investigated. These trials may be submitted to pharmaceutical companies, specialty societies, or the NIH. IITs sponsored by pharmaceutical companies are almost always limited to approved drugs and are not used for registration of a drug (eg, with the U.S. Food and Drug Administration [FDA]).

Clinical trials often involve multiple sites to facilitate recruitment of larger numbers of participants more quickly. Each research site has investigators who are responsible for all aspects of the conduct of the study at that site. Multicenter trials may involve many, sometimes more than a hundred, sites. Large multicenter trials may also be located in multiple countries.

Clinical trial protocol and data analysis

The clinical trial protocol is a critical document that dictates every detail of the conduct of the study and ensures that the study is conducted the same way at each site. The clinical trial protocol includes the following:

If the drug or device is approved by regulatory bodies for clinical use, the details included above determine the prescribing information (also called package insert or label) for approved use indications; the aim is to provide information that helps clinicians make evidence-based decisions.

Ideally, results can be extrapolated to populations broader than the one in the clinical trial. However, trial participants may not be the same as everyone who has the disease under investigation. Socioeconomic status, ability to drive, literacy, and proximity to the study site are issues that limit the applicability of trial results to everyone with the disease. In addition, inclusion and exclusion criteria may limit the population enrolled in the study. The best clinical trials often need to make significant efforts to recruit a racially diverse pool of participants. Diversity in clinical trials helps ensure that the study sample reflects real world diversity, which includes many ethnic and racial populations. In the US, racial and ethnic minorities make up almost 40% of the population. A sample that lacks diversity could miss some clinically important racial differences. For example, a deficiency of the enzyme G6PD is more common in men of African, Asian, or Mediterranean descent, and certain drugs can trigger hemolytic anemia in people with G6PD deficiency. Also, for some drugs,the race and genetic background of a person may influence the effectiveness of that drug. Evidence from clinical trials that include people from diverse backgrounds can more confidently be extrapolated to the wider population.

A statistical analysis plan (data analysis plan) defines how the data will be analyzed. This includes a calculation of the minimum number of patients required to detect a clinically meaningful effect (sample size, or power calculation). The statistical analysis plan also defines the primary endpoint and what, if any, secondary endpoints will be analyzed. Study endpoints are ideally a patient-centered outcome, such as mortality, morbidity, and symptom duration. Surrogate endpoints such as laboratory results and changes in imaging study findings are sometimes used. Trial participants are often assessed at baseline and again after treatment, and the change (if any) is then statistically analyzed for the whole group of patients in the study. The statistical analysis plan is included as part of the clinical trial protocol. In a randomized, blinded trial, the statistical analysis plan must be finalized before the blind is broken and the analysis is done. This ensures that study results cannot bias the way data are analyzed. Furthermore, the statistical analysis plan ensures against the possibility that various statistical methods were tried and only the most favorable analysis was reported. Endpoints in clinical trials are commonly analyzed by one of the following methods:

  • Continuous variables (eg, HbA1c, LDL-cholesterol, blood pressure, weight) are often compared using the Student's t-test or other parametric test (eg, analysis of variance) when data have a Gaussian (normal or parametric) distribution; a non-parametric comparison test is used when data have a non-Gaussian distribution.

  • Discrete variables (hospitalization, death, myocardial infarction) are compared by Chi-squared test or Fisher's exact test in which tables of number of events are compared for each treatment arm.

  • The time to discrete events are often compared by Kaplan-Meier (KM) curves.

  • When there are differences between the treatment arms, KM curves show the time at which the curves begin to separate and when the differences become statistically significant.

Study sponsor

Clinical trials are typically sponsored by a company with an interest in the development or better understanding of the intervention, such as a pharmaceutical company or a company that manufactures devices. The company that sponsors a study in development of an intervention (see phases I to III of Phases of Drug Development Drug Development Promising compounds can be identified by mass screening of hundreds or thousands of molecules for biologic activity. In other cases, knowledge of the specific molecular pathophysiology of various... read more ) typically develops the clinical trial protocol and data analysis plan in close discussion with the regulatory authority, such as the FDA as well as principal investigators. The performance and management of the study is often a major undertaking that includes the following tasks:

A company may do all of these tasks internally or may outsource some or all tasks. A small company may need to outsource tasks, but even a large company may decide to outsource. Some or all tasks may be outsourced to a clinical research organization (CRO) or an academic research organization (ARO), which are companies or academic centers that specialize in coordinating many aspects of clinical trials, including providing input into design, site selection, participant recruitment, and trial monitoring.

Clinical Trial Design

The clinical trial design is determined by the purpose and goal of the trial. Sometimes the design is influenced by the availability of resources. Clinical trials are typically done as prospective trials, meaning that the data are collected as they occur. Clinical trials generally have two types of endpoints: discrete (yes/no, also called dichotomous) and continuous. Discrete endpoints include events such as fractures, myocardial infarctions, or death. Continuous variables include events such as changes in laboratory values (eg, HbA1c, LDL-cholesterol), clinical biomarkers (eg, weight, blood pressure), or questionnaires (eg, pain scales, activities of daily living). The selection of endpoints is based on the clinical study question being addressed.

Prospective, double-blind, controlled trials

Clinical trials compare the intervention being investigated to another treatment, either a placebo or an active comparator. A placebo is an inert preparation (or sham procedure). In a blinded study for an oral drug, the placebo is made to look identical to the drug being studied. If that is not possible, both the study drug and the placebo may be placed in capsules that are identical. Even patients receiving placebo can have beneficial effects (called the placebo effect) or report adverse effects (called the nocebo effect). Without a placebo comparator, both positive and negative effects could be attributed to the study drug. An active comparator is often a commonly prescribed treatment that has a well-established efficacy and safety profile. Active comparators are also used to compare relative efficacies for the defined endpoint, safety (ie, adverse effects), and ease of administration (eg, pills versus injection, fewer injections).

Clinical trials are often double-blinded, meaning that both the patient and the investigator (and everyone involved in the trial who comes in contact with the patient) do not know if the patient is receiving the study drug or placebo. The trial is completed double-blinded to eliminate possible biases that may otherwise enter into the analysis of the trial. Without double-blind design, biases may occur. Bias may be conscious but can also be unconscious and cannot be avoided; therefore, double-blinding is the best way to assess a new intervention.

A randomized, double-blind, placebo-controlled trial is the most common study design used to demonstrate efficacy and safety of an intervention. In these trials, patients are randomly assigned to one of 2 or more groups, those that will receive the placebo and those that will receive the intervention being investigated. Sometimes more than one intervention is studied, such as different doses of a drug or more than one type of interventional procedure. The results for the groups are compared with regard to the primary endpoint. If the intervention group is statistically significantly superior to the placebo group, the difference is assumed to be a true population difference (eg, that it is not due to random variation). If the difference results in a clinically meaningful improvement, the intervention is considered efficacious. (Especially in large studies, there may be a statistically significant difference between intervention and placebo groups that may not be of great clinical significance.) Often 2 large, randomized, controlled trials are required to show efficacy before it is accepted by a regulatory agency (such as the FDA). Combining trial results also contributes to evaluation of the safety of the intervention, because important safety data, such as rare adverse events, may occur only infrequently. Therefore, establishing safety requires data on much larger numbers of patients. These trials are usually done as a parallel design such that the 2 groups receive either the study intervention or placebo treatment, and the results of these 2 arms are compared at the end of the trial for efficacy and safety. Often the improvement from baseline is the endpoint that is compared between the 2 groups, and the differences between the changes are also compared.

Sometimes a crossover design is used. In these trials, each group is randomly treated first with a study intervention or placebo in a blinded fashion. However in the crossover design, after the endpoint is assessed, each group then receives the other treatment for the same period of time, again in a blinded fashion. The endpoints are assessed by treatment assignment (whether the treatment was given in the first or the second period). This design has the advantage of allowing each participant to receive both active treatment and placebo, allowing comparison of their effects in the same individual and eliminating the variability between subjects (the largest source of random variability in clinical trials). As a result, the crossover design allows the trial to include fewer patients. However, a crossover design introduces a potentially problematic new variable, the sequence in which interventions are received, and the effect of one treatment on the other. If a treatment in the first period has demonstrable efficacy, then the baseline values in the second period are different for some participants. Such differences can be addressed by appropriate statistical methods. To minimize carryover effects from the first period, some crossover trials have an intervening "washout" period between the treatments. A washout period also helps to ensure that baseline values in the second treatment period are similar to the first.

A randomized, double-blind, active comparator-controlled trial compares the trial intervention to a comparator that is a known treatment for the disorder. Active comparator trials are performed in several situations, such as:

  • When the disorder is serious, perhaps life-threatening, and it would be unethical to treat some patients with placebo when there is a treatment that has established efficacy

  • When an experimental drug may have better efficacy than a currently available drug (both are added to the current standard of care)

  • When demonstrating that a new drug has efficacy that is "non-inferior" to a drug in the same class (common in pharmaceutical trials in which drugs in the same class from different companies may be compared; data often used in clinical, payer, and formulary discussions). The new drug may have advantages to the other drug other than efficacy (eg, ease of administration), so that if efficacies are equal, new drug may be preferred.

Active comparator trials may be double blinded. When active comparator trials are double blinded, they are sometimes called "double dummy," and each participant gets active drug A and placebo for B or active drug B and placebo for A. Several statistical approaches are used in each of the 3 common scenarios. These approaches may include the opportunity to first demonstrate noninferiority and then test for superiority. A statistical analysis plan must be specified before the start of the study.

Open-label, observational, and retrospective studies

If a clinical trial is not blinded, it is called open-label. Open-label studies are used in several situations in which placebos are not appropriate, such as:

  • When comparing surgical versus medical interventions or lifestyle versus medical interventions

  • When comparing different routes of administration for therapies (eg, oral versus injectable)

  • In active comparator trials when a pharmaceutical company wishes to compare its drug to another company's drug without having a suitable placebo for the other company's drug (eg, if some component of the placebo, such as a drug capsule, could have effects on efficacy or safety)

Open-label studies are subject to biases that may influence the results. These include investigator bias or participant bias, for example, when there is a preconceived idea that one therapy may be better than the other. Open-label studies may also be done to assess safety in a large number of patients, especially when a placebo controlled trial might be cost prohibitive. Sometimes open-label studies are the only approach to answer a clinical question.

An observational study is one in which there is no experimental intervention. Observational studies can be prospective or retrospective. They are often used when randomization may not be appropriate or may even be unethical, such as when studying the effects of smoking or alcohol use. Observational studies are also used when randomized trials are impractical because an independent variable cannot be easily controlled for, such as family history, education, or socioeconomic status. Observational studies can be used to study rare events that are unlikely to develop in sufficient numbers in controlled prospective trials. Observational studies are not used for pharmaceutical registration trials (ie, trials used as evidence for registration of a drug), but they may be used to assess pharmacovigilance (drug safety) by using large databases to compare adverse events between groups of people who have taken a drug and those who have not.

A cohort study is an observational study in which a group (called a cohort) of people at risk for a disorder are compared to a group of people who are similar except for being at risk for the disorder. In a prospective cohort study, people are followed over time to learn about the onset and progression of the disorder. Cohort studies include, for example, comparing people who are vaccinated against a certain disease to those who are not vaccinated, tracking outcomes such as subsequent infection, hospitalization, and/or mortality. Usually each person in a cohort study is "matched" to one or up to several controls who have similar age, gender, and clinical features except for the risk factors under observation.

Propensity score matching is a statistical technique that attempts to estimate the effect of having received an intervention. Propensity matching works by accounting for the factors (covariates) that predict receiving the treatment or intervention. Propensity scoring, which is often a more effective method than trying to find matched controls, reduces the bias due to confounding variables that could result from comparing unadjusted outcomes among people who received the treatment to those who did not. Propensity scoring can be particularly useful when potentially confounding variables cannot easily be matched or controlled for, but it cannot account for unmeasured or unsuspected variables (eg, clinical "gestalt").

A case-control study is another type of observational study in which patients with an outcome of interest are compared to "controls" who are as similar as possible except for having had the outcome of interest. Case-control studies are particularly useful when outcomes are rare and thus unlikely to occur in sufficient numbers in prospective, controlled clinical trials (which, due to practical limitations, cannot enroll very large numbers of patients).

Retrospective studies use already existing information, such as patients' clinical characteristics and how they were treated. Retrospective studies have the advantage of being an efficient way to obtain information quickly on large numbers of patients. The outcome of interest is defined, and then relevant variables (eg, predictive variables) are analyzed. Data sources for retrospective studies include chart reviews, national data registries such as the Centers for Disease Control and Prevention (CDC) National Health and Nutrition Examination Survey (NHANES) and the United Kingdom's primary care physician database Clinical Practice Research Datalink (CPRD), claims databases from large insurance companies, or other health care databases such as from the U.S. Department of Veterans Affairs (VA). However, absence of uniformity in data collection and recording is a limitation of retrospective studies. Often retrospective studies use matched controls to help control for variability in data acquisition. Propensity score matching is often used in retrospective studies.

More Information

The following English-language resources may be useful. Please note that THE MANUAL is not responsible for the content of these resources.

Listings of clinical trials are available at:

View Patient Education
NOTE: This is the Professional Version. CONSUMERS: View Consumer Version
quiz link

Test your knowledge

Take a Quiz!