Why Regulators Are Questioning Paper-Only Risk Assessments

What is a paper-only risk assessment?

In short: a paper-only assessment predicts. It does not measure.

A paper-only (theoretical) risk assessment is a document that uses scientific reasoning, literature searches, knowledge of the molecule’s chemistry and expert opinion to conclude that a risk is low. It is commonly used to argue that a test is unnecessary and can be omitted from a specification, for example because an impurity is purged by later process steps or cannot form under the process conditions.

Under ICH Q9(R1), quality risk management is of strategic importance to an effective pharmaceutical quality system. The guideline itself, revised in 2023, puts new emphasis on reducing subjectivity in quality risk management [1]. That is the core weakness of a purely theoretical assessment: its conclusion is only as good as the assumptions behind it.

Why regulators are pushing back

In short: too many “low risk” conclusions needed data to back them up.

For years, theoretical assessments drew few questions from reviewers. That changed with the nitrosamine crisis that began in 2018, and with the move to risk-based control of elemental impurities under ICH Q3D [3] [4]. For nitrosamines, confirmatory testing under the regulators’ call for review found impurities in products whose first risk evaluation had not predicted them. For elemental impurities, Q3D replaced blanket heavy-metals testing with a risk assessment that reviewers expect to be backed by data.

Reviewers now cross-examine the logic of a theoretical assessment and increasingly ask for confirmatory laboratory data. When the data is missing, the justification may be rejected. The reason is not distrust of science. It is experience: too many theoretical conclusions failed to predict what happened in real batches.

The gap between QbD on paper and real processes

Quality by Design rests on a sound idea: if you understand a process completely, you can predict the quality of the product. In practice, a “perfect” process exists only on paper. Real manufacturing involves factors a theoretical assessment rarely captures:

• equipment ageing and wear
• variation in raw material quality
• a change of raw material or solvent supplier
• environmental conditions such as temperature and humidity

Any of these can change the impurity profile of the final product. So when an assessment concludes “not expected to be present” without considering them, the conclusion is open to challenge.

Computational versus experimental gaps

In short: in silico tools are a starting point for the assessment, not the evidence that closes it.

Much of the industry still relies on computational models more than on experimental results. (Q)SAR models used to predict the mutagenicity of potential impurities are valuable, and ICH M7(R2) accepts them when two complementary methodologies (one expert rule-based, one statistical-based) are applied, with expert review where warranted [2]. But they have limits:

• no model is infallible: a negative prediction is a classification, not an experimental result
• an Ames test can overturn a negative prediction, and high-resolution mass spectrometry (HRMS) can reveal an impurity the assessment never listed
• they predict a property from the structure alone; they say nothing about whether the impurity forms, survives or is released in the real product

The matrix is where many predictions break down. Interactions with excipients, packaging materials and process solvents can create, release or mask impurities in ways a model does not capture.

For a practical view of how this plays out, see ICH M7 in Real Life: Why Mutagenicity Assessment Is Still Misunderstood.

Example: leachables in an injectable bag

Hypothetical example

The nitrosamine lesson

In short: several real root causes were found only after the first round of nitrosamine assessments.

Nitrosamines are the clearest case of theoretical assessments failing. Early assessments focused mainly on nitrosating agents and amines in the synthesis. Investigations by industry and regulators showed many more root causes, as described in the FDA’s Control of Nitrosamine Impurities in Human Drugs guidance and the EMA’s nitrosamine guidance for marketing authorisation holders [4] [5]:

• recovered or contaminated solvents, reagents and catalysts
• manufacturing lines and equipment shared between products
• trace nitrites in excipients reacting with amines in the drug substance, including in the finished product
• packaging materials, such as nitrocellulose-based lidding foil

Several of these root causes were added to the regulators’ lists only after the first call for review in 2019. Confirmatory testing is now an expected part of nitrosamine risk evaluation. We cover where companies still go wrong in Nitrosamine Control in APIs and Drug Products: Where Most Companies Still Fail.

Why QC must back risk assessments with data

Because regulators increasingly ask for evidence behind theoretical conclusions, QC laboratories have a growing role: generating the data that supports each risk-based decision. In practice that means:

1. Spiking and purge studies. If the claim is that an impurity is purged in an earlier step, spike it at a known level and measure how much survives. This turns a predicted purge factor into a measured one. Veeprho’s purge factor assessment service supports this work.
2. Sensitive, validated methods. Establish the LOD and LOQ and show they sit well below the acceptable limit. Trace impurities typically need LC-MS/MS, GC-MS/MS or, for elemental impurities, ICP-MS or ICP-OES.
3. Qualified reference standards. Spiking studies and method validation need a well-characterised standard of the impurity itself, supplied with a certificate of analysis. Browse the impurity standards catalogue and the nitroso impurities range.
4. Extractables and leachables studies with the actual product. Supplier data on the container is a starting point, not the answer. Test the product in its container, following USP general chapters 1663 and 1664 [6].
5. Batch data. Confirm the result on several representative batches, not a single one.

The goal is to show that “absent” really means “below the level of concern”, measured with a method able to see it.

If an impurity cannot be synthesised for use as a standard, that itself needs documented evidence. See nitrosamine synthesis failure reports.

What a defensible answer looks like

A risk assessment is no longer a box-ticking document written to avoid testing. It is a technical record of what is known about the product, and why.

When a regulator asks why a test is not performed, the answer cannot be “because we don’t think the impurity is there.” It needs to sound like this:

A measured purge factor, a validated LOQ and batch data answer the reviewer’s question. An argument on its own invites the next one.