Developing a Wellness AI Platform: Is a Clinical Trial Necessary?

AI Platforms and Patient Data: Why Do Clinical Trials Come Up?

The performance of an AI platform depends on the amount and quality of its training data. Especially for AI that directly affects patient health—such as health‑status assessment, prognosis prediction, or treatment recommendations—robust clinical data are essential to produce reliable results. Many companies try to collect real patient data during development, but because patient data are sensitive personal information, strict ethical and legal procedures are mandatory.

First, determine whether your work constitutes human‑subjects research. In Korea, any study that collects data from people or uses information that can identify an individual must receive review by an Institutional Review Board (IRB) in advance.

In other words, even if the AI platform is not categorized as an official “clinical trial,” obtaining patient consent to collect data requires IRB approval and procedures equivalent to clinical research. This is not about the “clinical trial” defined under the Medical Service Act; rather, it is a basic requirement to uphold research ethics and patient safety.

If you aim to secure sufficient clinical data, you will inevitably need to collaborate with hospitals or other clinical trial sites. To collect data from patients, you must prepare a standardized informed consent form (ICF), define data‑management methods, implement personal‑information safeguards, and pass institutional review. In short, even for a wellness AI platform, if people are involved, you need preparation comparable to a clinical trial. Developers often think “we just need consent,” but drafting the consent form and obtaining IRB approval require time, effort, and regulatory compliance.

Medical Device vs. Wellness Product: How Regulations Apply

Whether your AI platform qualifies as a medical device changes the legal procedures. Korea’s Ministry of Food and Drug Safety (MFDS) determines medical‑device status based on intended use and risk. Software intended for the diagnosis or treatment of disease is classified as a medical device, while personal health‑management or exercise‑assistance tools for everyday use are typically considered non‑medical, wellness products.

• Likely to be classified as a medical device: If the AI diagnoses a disease or directly influences treatment decisions, it is considered Software as a Medical Device (SaMD). Examples include apps that reference specific diseases (e.g., depression, diabetes) while claiming medical benefit, or algorithms that support clinical decision‑making. In such cases, you must complete MFDS medical‑device authorization and obtain approval to conduct clinical trials.

• May be excluded from medical‑device scope: If the AI performs low‑risk functions such as lifestyle management or exercise assistance instead of direct medical judgment, it may not be regulated as a medical device. For example, an app that collects/analyzes data for internal research without using the results in patient care—or a noncommercial wellness app—can be exempt from MFDS approval/notification, enabling relatively free development and release.

Infographic caption: Summary of MFDS criteria for personal wellness products. It contrasts cases that are classified as medical devices with those that are not, showing large differences in approval timelines and costs.

As the infographic suggests, medical‑device classification can require 1–4 years of preparation, and clinical trials may cost tens of billions of KRW. By contrast, if developed as a wellness product, release may be possible in about two months with costs reduced to tens of millions of KRW. This classification is therefore a pivotal strategic decision. The core question is usually: “Is it used to determine disease or for medical purposes?” The yes/no answer dictates the subsequent pathway.

Clinical‑Trial Procedures: Differences by Classification

Procedural requirements differ depending on medical‑device status, but the overarching flow is similar because both involve research with human participants. If the product is a medical device, you must obtain regulatory approval to run a clinical trial; if it is not, you will still follow ethics‑equivalent steps.

• If classified as a medical device: Obtain MFDS approval of the clinical trial plan (IND). Your study is treated as a formal clinical trial, requiring submission and approval of a protocol. Secure IRB approval at the hospital(s) conducting the trial. With patient consent, collect data and statistically evaluate the effectiveness and safety of the AI. After completing the trial, prepare a Clinical Study Report (CSR), submit an end‑of‑trial report to MFDS, and undergo MFDS review for product authorization/approval.

• If not classified as a medical device: You may omit the MFDS IND step; you can start the study without formal clinical‑trial plan approval, but IRB review remains mandatory. You can run small pilot studies, including at primary‑care clinics, to collect data.

The flow—patient consent, data collection/analysis, and results reporting—is similar. When the research ends, submit results (CSR) to the site’s IRB. This process verifies the AI’s effectiveness and provides evidence that supports product credibility. Even for consumer wellness products, an ethics‑grounded clinical‑research process is advisable.

In short: While regulatory involvement and timelines vary by classification, the core—collecting and analyzing patient data—does not. Avoiding medical‑device certification can reduce time and cost, but to demonstrate reliability and clinical value, you still need a structured study.

Many startups assume “we can move fast without clinical trials,” but collaboration with hospitals still requires IRB review, a protocol, and data‑management plans. This matters not only under Korean law but also for partnerships with global pharma and potential overseas registrations (e.g., FDA).

Conclusion & Recommendations

Even for wellness AI platforms, procedures equivalent to clinical trials are needed. If your product is classified as a medical device, obtain MFDS approval and conduct a formal trial; if not, collect and validate clinical data under IRB approval. Whenever patient data are involved, meeting patient‑safety and ethical standards is essential to launch a trustworthy AI product.

These regulatory procedures can feel unfamiliar and complex. A Korean CRO can help. A specialist CRO supports the entire process—from assessing medical‑device status and designing the research protocol, to preparing CRFs/ICFs, handling IRB submissions/approvals, managing multicenter data and statistics, and writing the final CSR.

By partnering with a CRO such as IntoInworld, you can optimize your development plan and strategy for Korea’s regulatory environment. For estimates, please reach us via our Contact page. We also recommend subscribing to our newsletter for the latest clinical‑trial insights and know‑how that Intoinworld shares with the industry. We promise tailored solutions and professional support for your innovative wellness‑AI journey.

Frequently Asked Questions (FAQ)

Q1. Do wellness‑purpose AI products also need clinical trials?

A1. They may not require a clinical trial in the full legal sense, but if your research uses patient data, you should follow equivalent procedures. Even for non‑medical, wellness products, any human‑subjects research requires IRB approval and ethical review. To demonstrate AI performance, obtain patient consent, collect data, and run statistical analyses to ensure reliability. While MFDS IND approval can be waived for non‑medical products, you still need clinical‑trial‑equivalent validation.

Q2. If it isn’t a medical device, can we ignore clinical‑trial regulations?

A2. Even without medical‑device classification, minimum ethical and scientific verification is necessary. You can skip MFDS authorization and formal clinical‑trial approval, but you still need IRB review. Adequate patient data are also essential to build trust for future clinical adoption. According to MFDS criteria, wellness products can reduce preparation time to about two months and lower costs significantly, but clinical effectiveness should still be demonstrated. In short, prepare evidence that supports safety and effectiveness.

Q3. How can a Korean CRO help?

A3. A Korean CRO professionally supports AI developers throughout clinical research. By working with a CRO experienced in approvals and study design, you gain: Early‑stage advice on medical‑device determination and strategy, Protocol drafting and preparation of essential documents (CRF, consent forms), IRB application/approval support to save time and streamline steps, Professional help with multicenter recruitment, data collection, and statistical analysis, and End‑to‑end support through CSR drafting and regulatory reporting.

With such one‑stop services, developers can focus on building the product while a trusted partner ensures compliance and clinical validation. Leveraging a Korea‑specialized CRO like IntoInworld can reduce the time and cost needed for a successful launch and global expansion of your wellness AI platform.