Cross-border segmented production splits a biologic's manufacturing across countries, typically bulk drug substance in one site and finished product in another. China moved from an effective ban to conditional approval between 2019 and 2026. Four dimensions decide whether a split process is safe: firm and product attributes, the technical chain, the holder's quality system, and supervision. The weakest link is comparability data, not the shipment.

Why the model moved from banned to conditional

For most of the past decade, Chinese rules did not discourage segmented production of biologics so much as exclude it. Contract manufacturing was allowed in principle, but biologics were carved out, and partial-process subcontracting fell outside the definition of contracted work. A product therefore had to be made end to end at one site.

The 2019 revision of the Drug Administration Law changed the legal basis by introducing the marketing authorization holder (MAH) system and letting holders contract manufacture. What followed was a series of narrow openings rather than one broad one. The order matters more than any single step.

Table 1. China's segmented production rule changes from 2019 to 2026

Practice moved before the text did. Insulin products reached the Chinese market with bulk substance made abroad and finished product filled domestically. COVID-19 vaccines used a pattern of centralized core processing with fill-finish distributed across several sites. Both predate the pilot, and both show what the model looks like once the text catches up.

How four regulators treat a product made in several countries

Four jurisdictions allow the model. Three principles run through all of them. The same standards apply to a split process as to a full one. Inspection frequency and method follow risk rather than a fixed calendar. Foreign sites are reached through mutual recognition arrangements rather than through separate national regimes.

Table 2. How four jurisdictions supervise a product made at more than one site

Where they differ is in where the extra attention goes. In the United States, the federal drug regulator decides between a paper review and an on-site inspection for pre-approval purposes, using the site's own history and information held by other authorities. Post-approval change inspections follow the same logic. For a product made at several sites, the emphasis shifts to four things: whether the quality agreement is complete and enforceable, whether the contractor understands that CGMP responsibility is its own, whether the holder actually supervises the contracted work and releases the product, and whether the sites are inspected together rather than one at a time.

The European Union splits the work by territory. A risk-based plan is set centrally and national authorities inspect sites on their own territory, while sites outside the union fall to the member state where the importer sits. When one product enters several member states, the regional agency coordinates the national actions. Japan allocates by site type rather than by risk score, and Brazil runs a risk-based model in which the federal government coordinates state authorities.

The four risk dimensions behind cross-border segmented production

A split process inherits every risk the holder system already carried and adds a set that appears only when a border sits between two stages. Keeping the two apart matters, because different instruments control them. Inherited risks answer to quality system discipline. Added risks answer to contract design and information architecture. The four dimensions below follow the quality risk management approach set out in ICH Q9(R1), applied to a process that crosses a border.

Table 3. The four risk dimensions used in this assessment

This framework is qualitative by design. It does not score products or rank sites, and no weights attach to the four dimensions. What it does is force a holder to answer a fixed set of questions at each stage before a batch moves, which is what makes it work as a checklist rather than a scorecard.

Where the added risk sits: existing exposure versus new exposure

Table 4. Existing risk and added risk, by dimension

Two rows deserve separate treatment. These are the ones a holder is most likely to misjudge.

The first is whether the holder runs a stage itself. A holder that makes bulk substance in-house and contracts only fill-finish holds direct, daily knowledge of the hardest part of the process. A holder that contracts out both stages holds a dossier and a set of audit reports. The framework scores the second higher on fundamental risk. Not because contractors are less competent, but because the holder's technical judgment thins out at exactly the point where a comparability question gets settled.

The second is approval status. A product already approved in the exporting market or in China carries a documented clinical and manufacturing history that a supervisor can check. A product approved in neither market has no such anchor, and the framework puts it at the top of the product risk scale. Light and temperature sensitivity sit underneath, since they decide how much of the quality margin the transport leg consumes.

Technical risk at each of the four manufacturing stages

The four stages of a cross-border segmented production chain are not equally demanding, and the framework asks a different question at each one.

Table 5. What each manufacturing stage has to prove

Transfer gets the most attention in the framework, and characterization is the reason. Moving a product from abroad into domestic production requires characterization data on representative batches made after the transfer, set against the pre-transfer product in a comparability analysis. Characterization here means molecular structure, biological activity, purity and impurity profile, studied well beyond what a release test covers. When bulk and finished product are split across sites, process parameters, material sources and storage conditions can all shift, so the comparability question widens rather than narrows. A thin package leaves the holder unable to say whether the split changed the product at all.

Two further requirements attach to the transfer step. Clinical studies run during development and post-marketing experience both belong in the assessment. Products built on novel adjuvants, adjuvant systems or special delivery systems, and products whose manufacturing differs from another approved mode, need bridging work where the assessment shows a gap.

Bulk production and fill-finish share four tests. Does the holder have segmented production experience, evidenced by a product already marketed in a PIC/S member market and by written operating procedures? Is the site inside the holder's group, and if not, does the holder have a dedicated contract management team and a complete quality agreement? Does the site itself have segmented production experience? Is the site, and this product at it, GMP compliant and inspected? The transport stage asks three different questions, and every one of them is new: cross-border and intercontinental shipping experience, cold chain capability, and capacity to absorb a loss.

What holders must put in place before the first shipment

Four control areas carry the load, each mapping to a dimension above.

Table 6. Four control areas for the holder

The resident personnel rule is where holders look for relief. Under a risk-based reading, a parent and subsidiary operating one quality system may be exempt from stationing staff at each other's sites. The exemption removes a person, not the interface. Someone still has to own the decision when the two sites disagree about a deviation.

Transport and data come second. They are also the most testable part of the arrangement. Validation has to cover container performance and temperature monitoring, run against worst-case scenarios rather than an average lane, and produce a record for every shipment. Alongside it, a holder needs a platform that records production data as it happens, traces product from raw material to finished pack, and gives a supervisor enough remote access to inspect without traveling.

The lifecycle system comes third. Split across several sites, pharmacovigilance, change control, annual reports, traceability and recall fail in a specific way: each site reports correctly and nobody aggregates the reports.

The fourth area concerns overseas holders. The domestic responsible person named by an overseas holder has to cooperate with inspections and carries joint liability. That only works if the holder keeps a live channel to each overseas site, with named contacts and an escalation route, instead of a shared mailbox.

What supervisors must do differently

Provincial authorities carry territorial supervision. The annual floor is one full-system GMP compliance inspection per domestic manufacturer. The focus list is specific: testing and release of bulk substance and excipients, bulk stability studies, execution of the finished product release standard, and cleaning validation where lines are shared. Findings carry deadlines, and serious breaches go into the safety credit file.

Table 7. Supervisor duties and their cadence

Overseas manufacturers fall to the national inspection body, with provincial authorities taking part. The focus there differs from territorial work: whether the process matches the registered standard, and whether the quality control system actually functions rather than merely exists.

Three supporting mechanisms matter as much as the inspections. A cross-region coordination arrangement with explicit duties turns separate inspections into one picture. An annual sampling plan gives a laboratory read that documents cannot provide. Key monitoring for a defined period after launch catches what clinical use reveals and trials do not.

The counterintuitive finding: the leg everyone watches is not the one that fails

Transport is the one stage the framework classifies as entirely new risk, and it is also the cheapest to verify. Container performance can be tested, temperature can be mapped, worst-case lanes can be run, and every shipment produces a record. A holder that invests in cold chain gets a defensible answer, and an inspector can check it without going inside the plant.

Characterization is the opposite case. It sits inside a transfer step that the framework treats as partly pre-existing, so it attracts less attention in a risk register and less budget in a project plan. No inspection samples it directly. The requirement is a comparability argument built from representative batches, and the evidence is a document a supervisor can read but cannot verify from the outside. This is where a split process goes wrong quietly, and it is the one failure that site inspection cannot correct.

The pattern repeats twice more. The resident personnel exemption looks like a reduction in cost and in regulatory burden, and it is also a transfer of interface risk onto procedures that nobody inspects until something breaks. Adding inspectors raises inspection frequency. It does not raise the depth of a comparability argument, and the added risks turn on that depth rather than on how often someone visits.

Limitations

This analysis is qualitative and built from a regulatory framework rather than from inspection outcomes, so it ranks dimensions in words and not in numbers. No scoring weights, failure frequencies or case counts are available, and none are invented here. The consolidated risk control table in the source material was not machine-readable, so the control and supervision tables above were rebuilt from the narrative sections: coverage is complete, wording is not reproduced. Rules in this area are still moving, with operational detail arriving through notices rather than through the regulations themselves. Nothing here is legal advice, and a holder should confirm current requirements with the regulator before designing a cross-border flow.

Bottom line

Cross-border segmented production of biological products is now a permitted model in China, and the pilot, the 2026 regulations and the implementing notice form one continuous opening rather than three separate events. The risk that decides whether a program succeeds is not the shipment but the comparability argument, together with the holder's ability to aggregate change across sites. A holder that builds one quality system and one data chain before the first batch moves will find the rest manageable.

FAQ

What is cross-border segmented production of biological products?

It is a manufacturing model in which the process for a biological product is divided into stages and completed at sites in different countries, usually with bulk drug substance made in one country and finished product filled and packaged in another. The stages can be run by the same corporate group or by separate contractors. The model differs from ordinary contract manufacturing because a border and a second regulatory jurisdiction sit between two consecutive process steps.

When did China start allowing segmented production of biologics?

China moved in stages rather than at once. The 2019 revision of the Drug Administration Law introduced the holder system and permitted contract manufacturing, ending the earlier position in which biologics could not be contracted at all. Post-approval change measures in 2021 accepted multi-site production, vaccine provisions in 2022 allowed segmented bulk or fill-finish for multi-valent products, and a national pilot opened in 2024. The revised implementing regulations effective in January 2026 made conditional liberalization binding.

Which products carry the highest risk under a split process?

Products approved in neither the exporting market nor China rank highest, because a supervisor has no prior clinical or manufacturing history to check against. Products already approved in one of the two markets rank lower. Underneath approval status sit traits that decide how much quality margin the transport leg consumes, chiefly light sensitivity and temperature sensitivity. Products built on novel adjuvants or special delivery systems add a further layer, because the transfer assessment may need bridging work.

How does ICH Q9(R1) shape a cross-border risk assessment?

ICH Q9(R1) supplies the method rather than the answer. It asks for risk to be assessed from the product and the process outward, with the formality of the assessment matched to the level of risk, and it treats gaps in knowledge as a risk in their own right. Applied to cross-border segmented production, it yields four dimensions: attributes of the firms and the product, the technical chain, the holder's own quality system, and supervision. Each is then read twice, once for risk that already existed and once for risk the border creates.

Why is characterization the hardest part of a cross-border transfer?

Characterization studies molecular structure, biological activity, purity and impurity profile far beyond what a routine release test covers, and it is the basis for arguing that a product made after transfer is comparable to the one made before. A split process shifts process parameters, material sources and storage conditions at the same time, so the comparison has to absorb several changes at once. A thin package leaves the holder without an answer, and no site inspection surfaces the gap.

How does the US regulator inspect a product made at several sites?

The federal drug regulator decides between a paper review and an on-site inspection using the site's own inspection history and information from other authorities, including outcomes shared under mutual recognition arrangements. Pre-approval and post-approval change inspections follow the same risk logic. For a multi-site product, attention goes to the quality agreement, the contractor's own CGMP duty, the holder's oversight and release role, and whether the sites are inspected together.

Who inspects an overseas bulk substance site in the European Union?

A risk-based inspection plan is set centrally, but the inspection itself is national. Sites inside the union are inspected by the authority of the member state where they sit. Sites outside the union fall to the member state where the importer is located. When the same product is imported into several member states, the regional medicines agency coordinates the national actions so that one site is not assessed twice in different ways.

What must a quality agreement cover in a segmented process?

It has to cover purchasing and acceptance, production, testing, storage, transport, release, retention samples, stability and pharmacovigilance, and it has to go beyond a list of duties. Transfer requirements, process validation, deviation handling and change control belong in the text, because those are the points where two quality systems hand work to each other. The agreement should also name the person who releases the batch. A complete agreement that nobody can enforce is treated as a risk in its own right.

Can a parent and subsidiary skip resident personnel at each other's sites?

Under a risk-based reading they can, provided both sit inside one quality management system and the interface between the two systems is shown to work. The exemption concerns stationing staff, not responsibility, and it does not remove the need for a named owner when the sites disagree about a deviation or a change. Treat the exemption as a shift in where control lives rather than as a reduction in control.

How often must a domestic manufacturer be inspected?

At least once a year, with a full-system GMP compliance inspection. The focus list is specific: testing and release of bulk substance and excipients, bulk stability studies, execution of the finished product release standard, and cleaning validation where lines are shared. Findings carry deadlines for correction, and serious breaches are recorded in the safety credit file. Products under cross-border segmented production also enter an annual sampling plan, with at least one batch per product tested each year.

What transport evidence does a cross-border bulk shipment need?

Validation that covers container performance and temperature monitoring, run against worst-case conditions rather than an average lane, plus a complete record for every shipment and a traceability chain from raw material to finished pack. The framework also looks at capability rather than paperwork: whether the holder already supplies cold chain products globally, whether a validated shipping process has a team behind it, and whether the holder can absorb a loss if a shipment fails.

What is the biggest supervision gap in cross-border segmented production?

The gap is not the absence of standards but the split of duties. Overseas inspection and territorial supervision sit with different bodies, so a product can be well inspected at each site and still be weakly supervised as a whole. Information exchange agreements, a shared platform and joint inspections close part of it. What remains is the comparability argument for cross-border segmented production, a document-based judgment that no inspection schedule can strengthen.