The linguistic validation of Clinical Outcome Assessments (COAs) is meant to ensure that translated versions are conceptually equivalent to the original. This means that items, responses, screen prompts and instructions must be understood in the same way regardless of the language version: the Krio version of a pain score being used in clinical centres in Freetown, Sierra Leone, needs to be understood in the same way as its counterpart version in Amharic, Korean, Guaraní and Romansch.

This is all very sound, but the process assumes that language is the primary and only barrier to comprehension. It does not consider how people from diverse cultures process information differently. This is because of:

·       Different reading directions, such as left-to-right, right-to-left, top-to-bottom

·       Different scripts, such as Latin, Arabic, Cyrillic, Armenian, Georgian, Asian scripts

·       Cultural differences in reading practices and the cognitive processing of texts

·       Font readability issues specific to certain language and scripts.

This is a problem in the localisation of COA because, not only does text have to be conceptually faithful to the source, but the entire ‘look and feel’ of the instrument must be replicated in all target language versions. It is an expectation in clinical practice and regulation to use assessments with the same semantics, the same typography, font size, boxes and general presentation as the source – essentially, a Western document - regardless of the target version. This fails to take into consideration any aspect of local differences in processing information or the cognitive burden that a Western document imposes on, for instance, the average Korean reader. It should come as little surprise, then, that this can have a direct effect on variations in data from country to country.

Why is this not taken into account at a regulatory level? Why is this not a central consideration in the Linguistic Validation process? Instead of only focusing on words and phrases, it should consider the cultural applicability of the entire document. This is something that cognitive debriefing cannot do on its own. This would entail a shift from what is understood in a COA to how the document is used. Besides, it is no surprise that little consideration is given to this in Linguistic Validation because issues like cultural differences in cognitive processing and of the cognitive burden when reading texts is little understood, under-researched and is only just starting to receive attention in academic circles. It does not seem to be filtering through to the commercial milieu.

This set of articles looks at some of the ways that cultural differences in cognitive processing may affect COAs.

Different reading directions change how information is processed

It is obvious to state that not all languages are read in the same direction:

·       Left-to-Right (LTR): English, Spanish, German, French

·       Right-to-Left (RTL): Arabic, Hebrew, Persian, Urdu

·       Top-to-bottom, right-to-left: Traditional Chinese, vertical writing formats.

The problem is that many COAs are designed with left-to-right formats. This causes usability problems, chiefly navigation, in RTL or vertical-script languages. Key problems include:

·       If RTL readers are expected to a read a shared-stem question format that was initially designed to be left-aligned, they may struggle to follow the logical flow of that section of questions

·       Many digitally formatted eCOAs often assume a LTR logic, which can create usability issues for RTL readers. Is this ever considered during the iterative design-test-design phase (if there is any)?

·       Then there is the ‘real estate’ issue. Target versions need to ‘look and feel’ like the source and this often includes how much space the target version can use. However, the reality is that different scripts have different spacing and formatting needs to maintain readability.

When is any of this ever discovered in the Linguistic Validation stage? Hardly ever. This means that no value is placed on how RTL readers may struggle to follow an LTR structure without necessary layout and reformatting adjustments. Consequently, these difficulties carry on into the live clinical phases and affect data accordingly.

The next article focuses on how different scripts carry different cognitive loads on their readers.

Thank you for reading,

Mark Gibson

Leeds, United Kingdom, March 2025