What extrapolation means

Extrapolation is not a shortcut around clinical evidence; it is the logical consequence of what a biosimilar development program sets out to prove. An innovator program proves that a molecule works in a disease, one indication at a time. A biosimilar program proves that a molecule is the same active substance as one whose efficacy is already established. Once sameness is shown with sufficient rigor, the reference product's clinical history becomes the biosimilar's clinical history, and the indications follow.

The comparative clinical study in a biosimilar program is therefore designed differently from a pivotal efficacy trial. Its purpose is to detect any clinically meaningful difference, so it is run in the population and with the endpoint most sensitive to such a difference, not necessarily in the largest indication. The result then supports all indications for which the scientific justification holds.

Three boundaries define the concept. It applies only to indications already approved for the reference product in the same jurisdiction. It must be justified for each indication, or group of indications sharing a mechanism, in the application. And it can be withheld for legal reasons, such as orphan or pediatric exclusivity, even when the science supports it.

Scientific basis: mechanism of action, PK, immunogenicity, totality of evidence

The justification for extrapolation rests on the totality of evidence, and the FDA's 2015 guidance, Scientific Considerations in Demonstrating Biosimilarity to a Reference Product, lists the factors a sponsor must address for each indication:

• Mechanism of action in each condition of use: target receptor, binding and dose–response, signaling pattern, structure–target relationship, and target location and expression.
• Pharmacokinetics and biodistribution in different patient populations, including whether exposure is expected to differ between indications.
• Immunogenicity in different patient populations, since immune status and concomitant therapy vary by disease.
• Differences in expected toxicities across conditions of use.
• Any other factor that could affect safety or efficacy in the extrapolated indication.

The analytical package carries most of the weight. If primary structure, higher-order structure, post-translational modifications, binding to target and to Fc receptors, and functional potency are all shown to be highly similar, then the biological activity underlying every indication is the same, and the clinical study serves to confirm rather than establish similarity. This is why the stepwise approach places analytical similarity first; residual uncertainty, not a fixed list of trials, determines what clinical data are needed.

Where a reference product acts through more than one mechanism, the justification must cover each. A monoclonal antibody that blocks a receptor and also engages antibody-dependent cellular cytotoxicity (ADCC) needs similarity data on both the Fab-mediated and the Fc-mediated functions, because one indication may depend more on ADCC than another.

FDA and EMA positions

The FDA and the EMA reach the same place by slightly different routes; the table shows where they align and where emphasis differs.

In both systems, extending indications is the default outcome of a successful biosimilar application rather than an exception that must be separately petitioned, provided the justification is in the dossier. Taiwan's TFDA and Japan's PMDA apply the same comparability principles.

Worked example: a trastuzumab biosimilar approved across early breast cancer, metastatic breast cancer and gastric cancer

Trastuzumab illustrates extrapolation well because its three main indications share one target but differ in disease biology and treatment setting. The reference product is licensed for HER2-positive early breast cancer, HER2-positive metastatic breast cancer and HER2-positive metastatic gastric cancer. A biosimilar sponsor typically runs its comparative clinical study in one breast cancer setting and seeks to extrapolate to the other two indications.

The justification follows the factor list:

1. Mechanism: in all three indications trastuzumab binds the extracellular domain of HER2, inhibits HER2-mediated signaling and recruits ADCC through Fc–FcγRIIIa engagement. Similarity in HER2 binding, Fc receptor binding and ADCC potency therefore supports every indication.
2. Sensitive population: HER2-positive breast cancer, measured by pathological complete response in the neoadjuvant setting or overall response rate in the metastatic setting, is accepted as a sensitive model because the treatment effect of trastuzumab is large and measurable within months.
3. Pharmacokinetics: comparative PK shows the same exposure; gastric cancer patients may clear the drug differently, but the difference applies equally to reference and biosimilar.
4. Immunogenicity: trastuzumab has low immunogenicity, and comparative anti-drug antibody data from the clinical study cover the breast cancer population; the gastric population is not expected to alter the comparison.
5. Residual uncertainty: the glycan profile that drives ADCC is the attribute most scrutinized, since afucosylation differences could change Fc-mediated activity. A sponsor that shows afucosylated glycan content and ADCC potency within the reference range closes this question.

The result is a label covering all three indications on the strength of one comparative study.

When extrapolation is refused

Regulators refuse or restrict extrapolated indications when one of the factors in the justification cannot be closed with the available data, and the historical cases follow a small number of patterns:

• Different mechanism or different mechanism weighting: an anti-TNF antibody acts through soluble TNF neutralization in rheumatoid arthritis but may depend more on membrane TNF binding and Fc-mediated effects in inflammatory bowel disease. One national regulator initially declined to extend an infliximab biosimilar to IBD indications in 2014 for this reason and later reversed the decision once additional data were available.
• Route-dependent immunogenicity: for erythropoietins, subcutaneous administration carries a higher risk of anti-drug antibodies and pure red cell aplasia than intravenous administration, so extrapolating from the intravenous route to the subcutaneous route has required dedicated immunogenicity data.
• Insufficiently sensitive clinical model: if the comparative study was run in a population where the treatment effect is small or slow, the study may not support extrapolating to indications with a different benefit–risk profile.
• Analytical differences in a mechanism-relevant attribute: a shift in glycosylation, charge variants or Fc receptor binding outside the reference product's range reopens the question for any indication that depends on that attribute.
• Legal exclusivity: an indication protected by orphan drug exclusivity is excluded from the biosimilar label until the exclusivity expires, regardless of the science.

The common thread is that refusal is attribute-specific and reversible; additional analytical, functional or clinical data have in most cases resolved the uncertainty.

What it means for sponsors and CDMOs

Because biosimilar efficacy extrapolation is decided largely on analytical similarity, the manufacturing and analytical program determines how many indications a biosimilar can reach, long before the clinical study reads out. Three implications follow for sponsors and their CDMOs:

• Target the reference range, not just the specification. Quality attributes linked to mechanism, especially glycosylation, charge variants and Fc receptor binding, should be steered during cell line and process development to sit inside the reference product's observed range across multiple lots.
• Build the functional assay panel early. Binding by SPR, Fc receptor binding, ADCC and complement-dependent cytotoxicity assays, and cell-based potency must be in place before the similarity exercise, since the justification for each indication cites them.
• Protect similarity through scale-up and site transfer. Any process change after the similarity exercise must be shown comparable under ICH Q5E, or the similarity argument built on the earlier data weakens.

For a CDMO, cell line development, process development and analytical characterization are therefore not separate services in a biosimilar program; they are the evidence base for extrapolated indications, read by regulators indication by indication.