Sterility testing systems

The compendial sterility test in USP <71> and Ph. Eur. 2.6.1 passes a sample of the batch through a membrane of 0.45 micrometre pore size, rinses the membrane to remove residual antimicrobial product, transfers it into two nutrient media — one at 30–35 °C for bacteria including anaerobes, one at 20–25 °C for fungi — and looks for visible turbidity after fourteen days. Everything interesting about the method is in what that procedure can and cannot establish.

It is a sampling test, and the arithmetic is unforgiving

Sterility is a property of every unit in a batch; the test examines twenty of them, and destroys those twenty in the process. If a fraction of units in the lot are contaminated, the probability of catching at least one follows directly from the binomial: at a contamination rate of one unit in a thousand, testing twenty gives roughly a two percent chance of a positive result. A batch contaminated at that level will pass the test about forty-nine times in fifty.

This is why the regulatory framework treats the finished-product test as a confirmatory check and not as the source of sterility assurance. That assurance comes from the process: a validated terminal sterilisation cycle delivering a sterility assurance level of one in a million, or for a product that cannot be terminally sterilised, aseptic process validation, media fills, environmental monitoring and a documented contamination control strategy of the kind EU GMP Annex 1 requires. The sterility test detects gross failure. It cannot certify a batch that the process did not already make sterile.

Why fourteen days

The endpoint is visible turbidity, which requires the culture to reach on the order of a hundred million cells per millilitre. A healthy organism from a lush environment gets there quickly, but organisms recovered from a pharmaceutical process are not healthy: they have been starved, dried, and sublethally injured by sanitisers or heat. Injured cells repair before they divide, and the lag phase that repair imposes can run for days. The fourteen-day incubation is an allowance for that lag, and for genuinely slow-growing organisms such as Propionibacterium species and some moulds.

Two consequences follow honestly. The media are general-purpose, and no medium supports everything — organisms with unusual nutritional or atmospheric requirements, and cells in a viable-but-non-culturable state, are not counted. Operationally, “sterile” means “produced no growth in these two media under these two conditions in fourteen days”. Second, the test is performed by a human handling open containers, so a positive result is as likely to be laboratory contamination as product contamination — which is why the test itself is now run inside a barrier isolator, and why every positive triggers an investigation rather than an automatic rejection.

What rapid methods substitute for growth

Alternative methods, addressed by Ph. Eur. 5.1.6, detect something that appears before turbidity: adenosine triphosphate by bioluminescence, carbon dioxide production or oxygen consumption in the headspace, viable cells by solid-phase cytometry, or nucleic acid by amplification. Each buys time by changing the observable, and each inherits the limits of the new one — respirometric methods still require growth, only less of it; nucleic acid amplification cannot by itself distinguish a live organism from residual DNA of a dead one. Validation must therefore demonstrate an equivalent limit of detection, ideally down to a single organism, across a representative spectrum of species.

The pressure to make this work is not economic. An autologous cell therapy may have a shelf life measured in hours, and cannot wait fourteen days for a result that would arrive after the patient has been dosed.