In the world of subjective tests, it can be difficult to investigate when things go wrong or when analysts have a different view of an appearance test.

For example:

  • The titrimetric test has an endpoint of blue-green, according to USP. The new analyst obtains an OOS result. The experienced analyst says the endpoint color should be more blue than green, and the new analyst titrated to a more green than blue color, which led to OOS. How do you investigate this or prove that the experienced analyst is right? It is equally possible that the experienced analyst has been wrong all along, and the material has been consistently OOS until new analyst came along.
  • The test procedure for an identity test says, “the precipitate is light blue.” An argument breaks out in the lab, with analysts claiming to see precipitate colors ranging from white to grey to blue.
  • The specification for appearance of the product includes the words “yellow-orange in color.” Is yellow acceptable? What if an analyst reports the color as orange? Is it OOS?

Once an investigation is opened, the investigator must consider a wide range of possible alternatives, including but not limited to; whether the specification was inaccurate to begin with, there are inconsistent viewer perceptions, and the history of the product (has the precipitate always been closer to white?). Is this an issue or not? How can I prove it? Do I really have to reject a batch when the precipitate color is in question?

When developing tests where the end result is subjective, it is important to recognize the risk and take steps to mitigate that risk.

In the case of the examples above, the risk of initiating investigations due to subjectivity of the test can be mitigated by the implementation of a colorimeter or a reference book of color chips, when developing a specification whenever subjectivity in determination of color might be a factor. The use of a specific color standard (or a range of color standards) mitigates the subjectivity risk.

The same reasoning can be applied to other subjective tests where one implements a standard that can be used as a comparator. It may seem like more trouble in the development of the test and specification, but it can save a lot of investigation time down the road.