January 1, 1970 (Vol. , No. )
Zachary N. N. Russ Bioengineering graduate student UC Berkeley
Oxytocin is no simple molecule. It consists of 135 atoms (nine amino acids) totaling 1 kilodalton (kDa). It pushes up against the size limits of common techniques like GC-MS, but it remains squarely in the realm of synthetic drugs because it is commonly made via a synthetic chemistry pathway.
There’s little room for variation; the polypeptide is unadorned and it’s a fairly simple task to verify you have pure oxytocin in your tube, and that it’s the exact same molecule, in form and function, as the one used to pass clinical trials.
Get a little bigger, however, and we start to enter biologic territory. Insulin is a popular one, at 5.8 kDa and 51 amino acids. It’s small (for a protein) and has come from both animal and E. coli sources for decades. There’s even a set of biosimilar insulin guidelines from the EMA, so approval should be easy.
Biosimilar Insulin: Not a Piece of Cake
As of now, all of the insulins available on the EU market were approved using the standard pathway for new drugs. Only one company attempted to enter the EU insulin market through the biosimilar route, and it failed. Marvel Lifesciences applied in 2007, claiming their insulin formulations were similar to Lilly’s Humulin.
There were numerous problems with the application: Critical information was missing, ranging from tracking numbers on product lots to chemical analysis on lots used in the PK, PD, and safety trials. The trials themselves were too short, and poorer performance versus reference in PK and safety trials did not help their case.
Another company, Biocon, recently ended its year-old collaboration with Pfizer to bring biosimilar insulin to markets; Biocon will be left to fend for itself in this regard. I don’t anticipate them having too much trouble getting approval, but they may have to wait: New guidelines for insulin analogues haven’t been released yet (the commentary ended last year).
Even so, one wonders, why are trials necessary? Wouldn’t proof of identical composition be sufficient? Such proof is not forthcoming. Biologics are too large to examine every atom and their source processes too variable to be assured that a biosimilar is indeed identical. A small amount of protein background might be permissible with one host organism, but immunogenic with another. Host organisms can decorate their protein products differently, leaving different sugar chains or amino acid modifications. In some cases, different glycosylation patterns can be advantageous, producing a longer-lasting product, but these changes also mean that previous data on pharmacokinetics no longer applies and dosing regimens would have to be adjusted.
Finally, considerations such as protein aggregation and misfolding mean that changes in product packaging can have more severe results than they would with small molecule drugs. In 1998, the epoetin-alpha product Eprex was switched from human serum albumin buffer to polysorbate-80 and glycine. It is suspected that the subsequent increase in protein aggregation in transit/storage led to greater immunogenicity and caused a sudden rise in cases of antibody-mediated pure red cell aplasia, a serious and potentially lethal autoimmune disease. Post-market surveillance caught this, and the contraindication of subcutaneous administration (generally more immunogenic) prevented more cases.
Working with Unknowns
The idea of drug packaging influencing the drug’s behavior is nothing new; it’s how “extended release” forms of small molecules are developed. However, when combined with the subtle and numerous ways in which manufacturing processes influence the biologics they produce, it makes sense that the EMA would pursue a “one process = one product” approach. With both general guidelines and specific guidelines appropriate for particular classes of biologics, the EMA is properly handling the question of how to evaluate biosimilars.
But other questions remain: As assay technologies and expression systems continue to improve, will we ever see two different manufacturers with different processes make identical biologics? Can the biosimilar approval model be used for approving biologics with intentionally different kinetics? What level of similarity is worth pursuing? In the event that a biologic finds a new indication, how will the approval of that indication for approved biosimilars be handled?
It’s not necessary to know everything about a drug to declare it functionally identical; there’s little interest in the isotope profile of aspirin or the average molecular weight in polymer gelatin capsules. However, without a complete understanding of how form relates to function and the analytical tools to prove it, we won’t be able to set out a minimal list of analytical tests to show that two biologic products are functionally identical and not just similar.
The EMA’s approach of making class-specific biosimilar guidelines is well-suited to match the evolving understanding of these products. We’re about to see the first biosimilar mAb go up for approval, and there are already approved somatotropins and epoetins.
However, guidelines aren’t everything: the EMA guideline on recombinant insulin was one of the first … so where’s the biosimilar insulin? Lilly applied for FDA approval of its recombinant insulin 30 years ago this week, and three months later it was approved. In some ways, it was the first biosimilar, mimicking a biologic but made through a radically different process.
We’ve waited 30 years—perhaps now, with the insulin analogue patents soon to expire, we can see the first pharmaceutical product of genetic engineering come full circle.