The world of translations is evolving, and translations for clinical trials are no exception. Technology keeps evolving, enabling companies to incorporate machine translation into their processes. Recent advances include incorporation of Natural Language Processing (NPL) and transformer-based machine learning techniques such as Google’s new Bidirectional Encoder Representations from Transformers (BERT) algorithm, which can consider the full context of a word by looking at the words that come before and after it, and therefore improve both search queries and machine translation.
Many translation and linguistic validation providers were quick to make use of these new technologies to streamline their processes, especially in areas involving a large number of documents, such as clinical trials. The growing globalization of clinical research demands for translation and linguistic validation in many target languages, some of which may be difficult for providers to find expert native speakers.
Mitigating the Risks of Translations
In a previous blog, we discussed the importance of clinical outcome assessments (COAs) and their paperless version, the electronic COAs (eCOAs). In a nutshell, the use of eCOAs in clinical trials is a reliable and cost-effective way to retrieve clinical data, helping increase patient compliance and ensuring the quality of the retrieved data.
Since the objective of COAs/eCOAs is to accurately capture patient data, it becomes apparent that patients are (or should be) the main focus during the preparation of these documents, including its translation.
Computer-assisted translation (CAT) has many benefits, including increasing productivity, ensuring consistency between translations of different documents, and the ability to create a translation memory, which allows translators to reuse existing strings of text which have been previously translated. But despite their advantages, relying solely on machines to do all the work of a human translators may come at a steep cost.
When CAT becomes the cornerstone of COA/eCOA translations, there are mistakes incorporated during the process which are not detectable until a human enters into the scene. While traditional translation processes have a team of translators and reviewers checking the documents for accuracy and consistency, machine-translated documents can go all the way to the cognitive debrief stage until mistakes are noticed. Even worse, some glaring errors may be only caught when the documents reach the patients’ hands and they comment that something is not quite correct in their language.
In these cases, the intention behind the COA/eCOA message is muted by the translation processes, and the resulting data is compromised.
Bringing Back Humans to Translations
By involving patients in the linguistic validation study, we can assure that their views are respected, through a faithful reporting of their feedback and its incorporation in the translated COA/eCOA.
No one likes to admit they made a mistake but carrying on is folly. When patients themselves state the obvious lack in meaning that undermines data integrity, when outcome collection is lost in unreliable translations, when linguistic validation becomes a tick in a box of the bureaucratic to-do list…
This is when we say enough. It’s time to unmute patients’ voices.
Linguistic validation matters. We choose to put the patient before the process, no matter how tight the deadline. We see the linguistic validation as an essential step, not only to avoid errors, but to ensure the translated text has the fluency of a document written by a native speaker.
The way we see it, a linguistic validation study should be conducted just like a research study, offering real-world evidence and legitimate feedback from patients. Using this approach, we can demonstrate the patients’ understanding and correct interpretation about the items of the instrument being tested and provide evidence of the conceptual equivalence of the COA/eCOA content in the different target languages. Linguistic validation can offer insights into distinct aspects of usability, especially for eCOAs.
By involving patients in the linguistic validation study, we can assure that their views are respected, through a faithful reporting of their feedback and its incorporation in the translated COA/eCOA. This adds a renewed focus on patient centricity and placing patient well-being at the core of all clinical initiatives. By leveraging linguistic validation as an essential facet of research, we provide a better understanding of the patient experience.
Machine translation can do part of the work, but it is time to bring humans back to the translation table. Only then can we unlock the true effectiveness of the reporting instruments and give back the voice to patients.
CQ fluency’s linguistic validation processes are based on self-correcting iterations that deliver psychometric equivalence. CQ fluency has over 1,500 linguists in our global pool with subject matter expertise in Life Sciences who are experts at minimizing linguistic validation iterations.
Contact us today to ensure your message is heard.