Silica Glossary

Control arm : refers to the group of patients participating in a clinical trial who do not receive the treatment being evaluated, in order to compare them with the group of patients who do receive it.

External control arm : control arm using any type of data not generated during the clinical trial, for example from healthcare or other studies (observational, previous clinical trials), or artificial data.

Historical control arm : external control arm constructed from multiple existing data sources (e.g., observational studies or previous clinical trials).

Simulated control arm : external control arm obtained by numerical simulation resulting from the execution of a model of the patient group, therefore composed of artificial data (see definition). NB : sometimes referred to as an artificial control arm.

Synthetic control arm : an external control arm constructed from multiple sources of assembled or simulated patient data, optimized for statistical power and proximity to the group of patients receiving the treatment under evaluation, allowing for the most informative comparison possible.(NB : Artificial control arm : ambiguous depending on the context, used to refer to either a simulated control arm or a synthetic control arm.)

Hybrid control arm : a control arm comprising both data generated during the clinical trial and supplemented by additional data of any type generated outside the trial.

Cohort : a group of people sharing a common characteristic (disease, treatment, exposure to a risk factor) and monitored over a period of time to observe changes in their health or the occurrence of certain events (recovery, complications, death, etc.). Cohort studies thus provide insight into the impact of these factors on patient health.

Artificial cohort : a cohort generated exclusively by calculation, algorithms, and automated processes.

Synthetic cohort : a cohort generated either by calculation, by aggregating previously collected cohort data, or by a combination of both.

Artificial data : patient data generated exclusively by computation, algorithms, and automated processes.

Augmented data : health data created or enriched by computation based on real health data.

Synthetic data : patient data generated either by calculation, algorithms, and automated processes, or by aggregating previously collected patient data, or by a combination of both.

Emulated data : almost synonymous with synthetic data, the meaning of the term may vary in the context of clinical studies.

• Sometimes synonymous with data obtained in silico
• Can also refer to the simulation of data from a randomized clinical trial, or an arm of that trial, based on observational data

In silico: refers to an experiment or intervention using numerical simulations (on a computer), as opposed to in vivo (in a living organism) or in vitro (in a laboratory). The term applies equally to a simple biological experiment or intervention, a patient, a complete clinical study, or a cohort of patients.

In silico clinical trial : a clinical trial conducted using numerical simulations to test a drug, medical device, or intervention under well-defined conditions, using models that have been verified and validated according to well-defined protocols. In silico clinical trials help to limit the need for animal and human (in vivo) experimentation and are typically used in two different contexts:

1. Planning and designing a real clinical trial. This is common practice in the pharmaceutical and medical device industries, helping to conduct studies that are as informative as possible at optimal cost and without risk to patients, as well as predicting and optimizing statistical power.
2. Substitution for a real clinical trial. In this still rare case, its use is generally limited to the control arm of the trial.

Simulated clinical trial : see in silico clinical trial.

Virtual clinical trial : may be synonymous with a simulated clinical trial or refer to a clinical trial synthesized entirely or partially from observational data.

Uncontrolled clinical trial : uncontrolled and therefore non-randomized interventional study, also known as a single-arm trial. However, an external control arm may be associated with these trials.

Clinical trial emulation : Trial emulation involves aligning the causal questions of the observational study with those of a controlled trial, mimicking the hypothetical randomized trial that one would like to conduct to answer a causal question. This innovative methodology applies rigor in the design and implementation of the emulated trial protocol, which, combined with appropriate statistical adjustment methods, allows for the understanding of biases inherent in observational studies, including the quantification of potential residual confounding bias.

Pragmatic trial : a randomized trial that relies on standard information gathering structures to monitor research (standard medical practice, simplified data collection).

Augmented randomized trial via external control : these trials, sometimes referred to as hybrid trials, combine data from patients included in the study with artificial data or data from databases, derived from healthcare or other research. The design, analysis, and statistical inference of the treatment effect are based on patient data from one randomization arm and the external control cohort.

Hybrid randomized trial : see augmented randomized trial via external control.

Observational study : Non-interventional study, which may use data collected in routine practice, such as healthcare data. In this case, such a study may include an experimental group and a control group, both derived from real-life data.

Synthetic Data Generation (SDG) : Computational methods for creating artificial or synthetic data. These include, for example, generative AI approaches, traditional statistical and Bayesian approaches, or simulations based on mechanistic models.

Digital twin : a model that can be used to perform simulations. It can be applied at the scale of a molecule, an organ, one or more physiological processes, an entire body, or a group.

Virtual patient : a model that can be used to perform simulations representing the specific characteristics of an individual, allowing, for example, the specificities of their anatomy, lifestyle, or other characteristics to be taken into account in order to support individualized clinical decision-making.

Virtual population : a set of virtual patients that allows a plausible subset of the population to be obtained by reflecting the inter-individual variability of patients.

REFERENCES

• US Food and Drug Administration:
  • General Glossary : https://www.fda.gov/science-research/artificial-intelligence-and-medical-products/fda-digital-health-and-artificial-intelligence-glossary-educational-resource
  • Additional glossary, page 26: Using Artificial Intelligence & Machine Learning in the Development of Drug and Biological Products,
• Avicenna Alliance Glossary : https://www.avicenna-alliance.com/glossary.html