OOT Trends and Batch History: Reading a Supplier's Data

Every number in the file is inside specification. Assay at 99.1 percent against a 98.0 to 102.0 percent range, total impurities at 0.4 percent against a 1.0 percent limit, loss on drying at 0.6 percent against 2.0 percent. Nothing to reject. And yet the previous six batches read 99.6, 99.5, 99.7, 99.6, 99.4 and 99.5, and the seventh reads 99.1.

That is an out-of-trend signal. It does not make the batch unusable. It makes it worth a question, and the fact that almost nobody asks is why it is worth writing about.

Why a certificate cannot show this

A certificate of analysis reports one batch against a specification. It has no memory. Every statement on it is true of the material in front of the analyst and says nothing about whether that material resembles the last twenty batches or the first one.

Trend information lives in the batch history, which is a different document and is rarely offered. A buyer who never asks for it is relying on specification limits to detect process drift, and specification limits are set to catch failure, not to detect movement. By the time a limit is breached, the drift has usually been running for several batches.

What to ask for, and in what form

Ask for a tabulated history of the last ten batches, with batch identifiers removed if confidentiality requires it, covering a small fixed set of attributes. Ten is enough to see a range and a direction. Fewer than five is a sample, not a trend.

Attribute	What movement usually means
Assay	A step change often indicates an analytical cause; a drift often indicates a process cause
Total related substances	Drift upward points to purification or to starting-material quality
Individual specified impurity	A single impurity moving alone localises the cause
Loss on drying	Movement points to drying conditions or to packaging and storage
Optical rotation	Movement points to chiral purity, which is a different conversation

The value of the table is in reading the rows together. An assay step change with no impurity movement is almost always analytical. An impurity drift with a stable assay is almost always process. Two rows moving in opposite directions is worth a call.

Reading a step change against a drift

A step change is a jump between two consecutive batches that persists. A drift is a gradual movement across several. They have different causes and different implications for a buyer.

Step changes are usually traceable to a discrete event: a new analytical standard, a change of column supplier, a change of analytical site, a new starting-material lot. These are often benign, and often poorly communicated. If the supplier can name the event, the signal is explained.

Drifts are more interesting. A slow upward movement in a specified impurity over eight batches suggests a purification step losing efficiency, a column approaching the end of its life, or a starting material gradually shifting. None of these produces an OOS result quickly. All of them eventually do.

The question that separates a real answer from a polite one

Ask which batch the change began at, and what changed at that point. A supplier with a working trend process will answer with a batch number and an event. A supplier without one will answer with a general statement about normal variation.

Both answers are informative. The second is not a lie, it is a description of the system, and it tells you that the trend you are looking at was not being watched by anyone until you asked.

What to do with the finding

For most buyers the correct response to an OOT signal is to document the question and the answer, then continue. Rejecting a conforming batch on trend grounds alone is not usually defensible and damages the relationship for no material gain.

The exception is where the trend touches an attribute your own process is sensitive to. If your tableting line needs a narrow particle-size band, a drift in particle size matters more to you than the specification limit suggests. That is a reason to define your own tighter internal limit, not a reason to argue about the supplier's.

Why suppliers resist, and what to offer instead

Batch history feels like process disclosure. Suppliers worry that a customer will read a trend, infer a process detail, and take it to a competitor. That concern is real but manageable: aggregated results without batch identifiers, without process parameters, and without dates carry very little process information and quite a lot of assurance.

We provide history in that form when a qualification file requires it. It is a small document and it answers a question that a certificate cannot.

The habit worth building

Ask for ten batches at qualification, and then once a year thereafter. Two documents, twenty minutes each, and you will notice a drift before it becomes a deviation. That is a better use of a quality team's attention than re-reading certificates that all say the same thing.

The signal is never in the batch that failed. It is in the five batches that passed while something was moving.

How to build the request so it gets answered

Frame the request as a qualification input rather than as an audit. Ask for a table of results for a defined set of attributes over the last ten batches, with identifiers removed and no process parameters included. That framing removes the two objections that usually stop the request: commercial disclosure and process inference.

Put the request in the quality agreement rather than in an email, so that it recurs annually without a new negotiation. A clause that says trend data will be provided once a year on request is easy to accept at signature and easy to operate afterwards.

Three articles here cover the neighbouring questions. Batch-to-batch consistency covers what the numbers mean in aggregate. An impurity sitting at its limit covers the case where the trend has already arrived at the specification. And Retest period versus shelf life covers the dating consequence when a trend is stable enough to support an extension.

Where the trend expectation comes from

The expectation that a process is monitored rather than merely released comes from the stability and quality-system guidance published by the International Council for Harmonisation, and from inspection practice under the Pharmaceutical Inspection Co-operation Scheme. The European Medicines Agency publishes the related quality guidance that competent authorities apply at inspection.

None of these texts requires a customer to ask for trend data. They require the supplying organisation to have it. The customer's request is simply a way of finding out whether the requirement is being met.

What a control chart would show

Plotted on a control chart, most of what looks like noise in a batch history resolves into three patterns: ordinary variation inside the control limits, a run of points on one side of the mean, and a single point outside. The second pattern is the one that matters and the one a table of numbers hides.

Six consecutive results above the historical mean is a signal in any standard rule set, and it usually precedes a specification breach by several batches. You do not need software to see it. You need the history in one column and five minutes.

Why this is cheap to do

Two requests a year, twenty minutes each. Against that, the cost of discovering a drift through a rejected batch is a shipment, a deviation, a replacement and a conversation with your own customer. The asymmetry is large enough that the only real obstacle is remembering to ask.

Writing the findings into your own file

An out-of-trend observation that stays in an email has no effect on anything. Written into the qualification file, it becomes a baseline that a future reviewer can compare against, and a note that says what was asked and what was answered.

Three lines are enough: the attribute, the batch at which the pattern changed, and the explanation received. If the explanation was that normal variation applies, write that down too. It is an accurate description of the supplier's position, and it is the sentence you will want when the same attribute later moves outside its limit.

This is the difference between a quality file that records decisions and one that records documents. Both look similar in an audit. Only the first one helps you make the next decision.

When the trend justifies a tighter internal limit

Where an attribute matters to your own process, the response is not to argue about the supplier's specification. It is to set your own internal limit, tighter than the release limit, and to reject at goods-in against it.

An internal limit is defensible because it is yours, it reflects your process rather than the material in general, and it does not require the supplier to change anything. It also converts a trend observation into an operational control, which is the only form in which a trend observation does any work.

Frequently asked questions

What is the difference between OOS and OOT?

OOS means a result outside the specification limit. OOT means a result inside the limit but outside the pattern the process normally produces. An OOT result does not require rejection of the batch. It requires an explanation, because it is an early signal that something in the process has moved.

How many batches do I need before a trend means anything?

Enough to establish a range and a typical spread, which in practice means five to ten batches for a well-behaved attribute. Fewer than five data points can show a difference but cannot show whether it is meaningful. Ask for the last ten and read the last three carefully.

Which attributes are worth trending?

Assay, total related substances, the individual specified impurities, loss on drying and, where relevant, optical rotation. These five carry most of the process information. Particle size is worth trending if your downstream process is sensitive to it, but it varies more for reasons unrelated to quality.

What does a step change in assay usually indicate?

Often a change in the assay standard, the titration or chromatographic conditions, or a change of analytical site, rather than a change in the material. A step change in an impurity, by contrast, more often reflects a process change. Reading the two curves together separates analytical causes from process causes.

Should I reject a batch because it is out of trend?

Not on the trend alone. Rejection needs a specification failure or a documented risk to your process. An OOT result is grounds for a question, and a supplier that can explain it with reference to a known change is in a stronger position than one that cannot explain it at all.

Is trend data confidential?

The individual results are usually treated as confidential because they describe a process. Aggregated history, with batch identifiers removed, is commonly shared and is enough for a buyer to judge whether the process is stable. Asking for a summary is more productive than asking for the raw release records.