REGULATORY FRAMEWORK FOR DYNAMIC DOSING ALGORITHMS IN BIOLOGICAL MEDICINES IN JAPAN

1. INTRODUCTION

Dynamic dosing algorithms are computational systems that recommend or modify the dosage of biological medicines according to changing patient characteristics, laboratory results, pharmacokinetic measurements, therapeutic responses and adverse reactions. They are increasingly relevant to monoclonal antibodies, therapeutic proteins, vaccines, blood-derived products and other biologics.

These systems may use artificial intelligence (AI), machine learning, population pharmacokinetics, pharmacodynamic modelling and model-informed precision dosing (MIPD). Their purpose is to optimise therapeutic effectiveness while reducing toxicity and unnecessary drug exposure.

Japan does not have one independent statute exclusively regulating dynamic dosing algorithms. Instead, their regulation operates through pharmaceutical legislation, medical-device regulation, clinical-trial requirements, pharmacovigilance obligations and professional medical standards.

2. PRINCIPAL LEGISLATIVE FRAMEWORK

A. Pharmaceuticals and Medical Devices Act (PMD Act)

The Act on Securing Quality, Efficacy and Safety of Products Including Pharmaceuticals and Medical Devices is the principal legislation governing biological medicines and associated medical software.

Its regulatory framework requires:

Appropriate marketing authorisation for biological medicines.

Scientific evidence supporting efficacy and safety.

Appropriate manufacturing and quality-management systems.

Compliance with approved dosage and administration conditions.

Post-marketing safety monitoring.

Regulatory assessment of software qualifying as a medical device.

Where a dosing algorithm is incorporated into a regulated medicine's authorised use, its performance and clinical implications must be adequately supported by evidence.

B. Medical Device Software Regulation

Under the PMD Act, standalone software intended for medical diagnosis, treatment or prevention may qualify as Software as a Medical Device (SaMD), depending on its intended purpose and risk.

Consequently, an algorithm that independently calculates or recommends biologic doses may require medical-device regulatory assessment if it meets the statutory definition.

The distinction is important:

Software performing administrative calculations may fall outside medical-device regulation.

Software providing clinical dosing recommendations may qualify as SaMD.

Software controlling treatment delivery may present additional regulatory risks.

PMDA evaluates applicable medical-device approval and certification requirements.

C. Good Clinical Practice (GCP)

Clinical trials evaluating dynamic dosing systems must comply with applicable Japanese GCP requirements.

Sponsors should demonstrate that algorithm-guided dosing is scientifically justified, clinically appropriate and supported by reliable clinical data.

3. VALIDATION AND CLINICAL EVIDENCE REQUIREMENTS

Dynamic dosing algorithms require rigorous validation before clinical implementation.

Important requirements include:

Clinical validation: Evidence that algorithm-generated recommendations are clinically appropriate.

Analytical validation: Accuracy and reliability of laboratory measurements used by the algorithm.

Pharmacokinetic validation: Assessment of drug concentration and exposure relationships.

Population validation: Evaluation across relevant age groups, body weights and patient populations.

Safety validation: Assessment of toxicity, immunogenicity and adverse reactions.

Algorithm validation: Testing of model assumptions, reproducibility and prediction accuracy.

Japan's experience with Model-Informed Drug Development (MIDD) demonstrates the growing regulatory importance of modelling in dosage selection. A survey of 322 Japanese medical-product approvals between January 2020 and March 2022 found that modelling analyses were used in approximately 47.8% of cases, and contributed substantially to dosage selection in approximately 24.2% of cases.

Monoclonal antibodies and orphan medicines were among the categories in which modelling was frequently used.

4. DOSAGE MODIFICATION AND PRESCRIBING RESPONSIBILITIES

A dynamic dosing algorithm must operate within a clinically justified framework.

The manufacturer should establish:

Permitted dosage ranges.

Relevant patient eligibility criteria.

Laboratory and biomarker requirements.

Conditions requiring dose reduction or interruption.

Procedures for identifying contraindications.

Monitoring intervals.

Emergency intervention procedures.

Conditions requiring physician override.

The algorithm should not automatically override authorised prescribing information without an appropriate regulatory and clinical justification.

Physicians remain responsible for evaluating the patient's overall clinical condition and determining whether an algorithm-generated recommendation is suitable.

5. CASE LAWS AND JUDICIAL PRECEDENTS

There is no established Japanese Supreme Court judgment specifically addressing AI-based dynamic dosing algorithms for biologics. However, the following six judicial authorities provide relevant principles.

A. Supreme Court Judgment, 23 January 1996

Authority: Direct authority on pharmaceutical prescribing precautions; analogous to algorithmic dosing.

The Supreme Court considered the relationship between pharmaceutical package-insert precautions and physicians' duties of care.

It held that failure to follow precautions in pharmaceutical instructions may support a finding of negligence unless there is a particular reasonable justification.

Relevance: Algorithm-generated dosing recommendations cannot automatically displace approved pharmaceutical instructions and safety precautions.

B. Supreme Court Judgment, 8 November 2002

Authority: Direct authority on medication-related medical negligence; analogous to dynamic dosing.

The Court examined a case involving continued medication administration despite symptoms potentially indicating serious adverse reactions.

It emphasised the importance of considering medical knowledge available at the relevant time and the physician's duty to respond appropriately to warning signs.

Relevance: Dynamic dosing systems must incorporate appropriate adverse-event indicators and escalation procedures.

C. Supreme Court HIV Blood-Product Case, 3 March 2008

Case No. 2005 (A) 947.

Authority: Direct authority on biological-product safety responsibility; analogous to dosing algorithms.

The Supreme Court upheld criminal negligence liability concerning the use of HIV-contaminated unheated blood products.

The decision demonstrates that persons exercising relevant regulatory responsibilities may face criminal liability where foreseeable risks are not appropriately addressed.

Relevance: Developers, manufacturers and regulators must take foreseeable biologic safety risks seriously.

D. Hepatitis C Contaminated Blood-Product Litigation, Tokyo District Court, 23 March 2007

Authority: Direct authority on biologic-product safety; analogous to algorithmic dosing.

The Court addressed responsibilities associated with contaminated blood-derived pharmaceutical products and inadequate warnings concerning known risks.

Relevance: Manufacturers must provide appropriate safety information and respond to emerging evidence of biologic-product risks.

E. Supreme Court Judgment, 8 November 2002, Medication Adverse-Reaction Litigation

Authority: Direct authority on adverse-reaction monitoring; analogous to algorithmic safety monitoring.

The Court considered whether physicians should have anticipated and responded to symptoms potentially indicating Stevens-Johnson syndrome.

It recognised the importance of assessing emerging clinical evidence rather than relying exclusively on the absence of fully developed symptoms.

Relevance: Algorithmic systems should identify clinically significant warning signs and support timely medical intervention.

F. Lamictal Dosage Litigation, Japanese District Court

Authority: Analogous authority on dosage compliance.

Japanese litigation concerning excessive lamotrigine prescribing considered the importance of following prescribed dosage escalation requirements intended to reduce serious skin reactions.

The decision illustrates the legal significance of established dosage precautions and the need for reasonable clinical justification when departing from them.

Relevance: Dynamic dosing algorithms must not recommend unjustified dosage escalation or disregard recognised safety limitations.

6. DATA GOVERNANCE AND ALGORITHM ACCOUNTABILITY

Dynamic dosing systems process sensitive patient information, including genetic data, laboratory results, medical histories and treatment responses.

The Act on the Protection of Personal Information (APPI) imposes requirements concerning the handling of personal information and specially cared-for personal information.

Manufacturers and healthcare institutions should implement:

Appropriate data-use permissions.

Access restrictions and cybersecurity controls.

Data accuracy and integrity procedures.

Algorithm version management.

Audit trails documenting dose recommendations.

Human oversight and correction mechanisms.

Procedures for reporting algorithm-related safety incidents.

Where algorithms are updated after approval, the regulatory significance of the changes must be assessed. Changes affecting intended use, clinical performance or safety may require additional regulatory procedures.

7. REGULATORY ENFORCEMENT AND COMPLIANCE

Non-compliance may result in regulatory investigations, corrective orders, product-related restrictions, suspension of relevant authorisations and civil or criminal liability where statutory requirements are violated.

Biologic manufacturers should maintain a documented compliance programme covering clinical validation, approved dosage conditions, software lifecycle management, adverse-event reporting, physician training and post-marketing surveillance.

8. CONCLUSION

Japan regulates dynamic dosing algorithms in biological medicines through an integrated pharmaceutical and medical-device regulatory framework. The PMD Act provides the principal foundation, while GCP, pharmacovigilance requirements, APPI and professional medical standards establish additional safeguards.

The central regulatory principle is that algorithmic innovation must not compromise patient safety, scientific reliability or physician accountability. Dynamic dosing recommendations must be clinically validated, appropriately documented and consistent with authorised pharmaceutical conditions or supported by the necessary regulatory procedures.

9. PRINCIPAL LEGAL AND REGULATORY REFERENCES

Pharmaceuticals and Medical Devices Act, Japan.

Act on the Protection of Personal Information.

Japanese Good Clinical Practice requirements.

Japanese Good Vigilance Practice requirements.

Japanese Good Post-Marketing Study Practice requirements.

PMDA, Software as a Medical Device regulatory framework.

ICH E6, Good Clinical Practice.

ICH E8, General Considerations for Clinical Studies.

ICH E17, Multi-Regional Clinical Trials.

WHO, Model-Informed Drug Development and pharmacovigilance guidance.