Recommendations for assessing translatability.
Development of clinical outcomes assessments (COAs) for use in multinational clinical trials requires thorough assessment of readability and translatability of the source. A highly translatable source will facilitate the achievement of conceptual equivalence during linguistic validation, resulting in valid data pooling across languages. Several issues should be considered in an effort to eliminate translation difficulty during harmonization and cognitive debriefing. Instrument developers should collaborate with the language service provider (LSP), and complete a translatability assessment prior to linguistic validation.
When developing COAs for multinational trials, several key steps can serve to ensure ease of translatability and achievement of conceptual equivalence across languages. The LSP should review a draft of the COA to identify any potential translatability and readability concerns. This "face validity" review serves to pinpoint problematic phrasing and concepts and to make suggestions for revision prior to translation. Common problems identified may include conceptual overlap in response sets, unbalanced response sets, ambiguous key concepts, idiomatic items for which direct equivalents may not exist across languages, incomplete instructions and double-barreled questions in which a single question seeks to measure multiple constructs. Identifying and addressing such issues early in development saves time and unnecessary expense. The LSP may also make recommendations for minimizing respondent burden. A Flesch-Kincaid readability test can be applied and revisions suggested to improve comprehension by a broad target population. A recent COA face validity assessment required that the questionnaire be taken from grade level 17 to grade level 8. Successful revision was accomplished by simplifying instructions and terminology, defining medical terms, and providing examples of complex concepts.
Following the face validity review and instrument revision, a translatability assessment should be completed. This review is carried out by linguists within desired language groups to identify translatability concerns within and across languages and to recommend source revisions to improve translatability. An important outcome of this assessment is to identify items that are so culturally-specific that they would potentially yield differing interpretations across languages. According to the "ISPOR Patient-Reported Outcomes Translation and Linguistic Validation Good Research Practices Task Force Report,"1 in order to pool data from multi-country studies, cognitive equivalence of PRO measurement concepts should be established. A translatability assessment will identify idiomatic phrasing, ambiguous concepts, and other areas of a COA with the potential for problematic translation. Though idiomatic phrases such as put yourself down and step on a few toes can create difficulty in finding equivalent concepts when translated, equivalence is achievable through completion of linguistic validation but may require more rounds of revision during harmonization.2
Recent research has revealed ways in which the translatability of COAs can be improved. A popular concept found in patient questionnaires, bother, has recently been shown to hinder data validity. The ambiguous nature of bother results in multiple interpretations across languages. For example, German respondents interpret the concept as something that is a nuisance, but the same concept is interpreted as nervousness or anxiety to Russian speaking respondents. These discrepancies present a problem for data pooling across languages. Alternatively, it is recommended that the concepts for measurement be specifically defined. For example, instead of asking a patient how much psoriasis bothers him or her, it would be best to ask how painful or embarrassing one's psoriasis has been. Using these alternatives increases the likelihood that what is intended for measurement is actually being measured because the variation in interpretation of the concepts is much less likely. Analysis of the translatability of bother has also shown that more specific concepts, such as pain and embarrassment, will yield greater conceptual equivalency when translated. Replacement of ambiguous concepts with more specific concepts will better measure what researchers are intending to measure.3
Research exploring the translatability of response sets was conducted by Gawlicki, et al. 2012 to determine optimal response sets for patient questionnaires for which a high degree of translatability is vital. A popular response set measuring symptom severity or frequency is not at all, a little bit, somewhat, quite a bit, and a great deal. In observed translations, the ambiguities in this response set will cause conceptual overlap when translated, as reported by linguists participating in the study. Furthermore, the concept of abit is often omitted as it does not exist in many languages. A response set observed to have a high degree of translatability and no issues with conceptual overlap when translated is not at all, rarely, sometimes, often, and always. These terms render this response set more specific and generally will result in a more sound translation. Results from the study showed that a maximum of five response options will help prevent concept overlap in translation. If more than five response options in a set are desired, either a numeric rating scale or a visual analog scale is preferred.4
Creating a COA for use in multinational trials requires careful review and consideration during the instrument development phase. The aforementioned steps are recommended in order to yield the sound translations necessary for later pooling of data.
1. D. Wild, et al., "Multinational Trial—Recommendations on the Translations Required, Approaches to Using the Same Language in Different Countries, and the Approaches to Support Pooling the Data," Value in Health, 12 (4) 430-440 (2009).
2. M. Gawlicki, B. Brandt, S. McKown, M. Talbert, "Achievement of Cultural Equivalency when Translating a Patient-Reported Outcomes Instrument Containing English-US Idioms and Colloquialisms," poster presented at the ISPOR 17th Annual Meeting; (June 2-6, 2012).
3. M. Gawlicki, B. Brandt, S. McKown, M. Talbert, "Application of the Bother Concept Across Cultures," poster presented at the ISOQOL 18th Annual Conference (October 26-29, 2011).
4. M. Gawlicki, B. Brandt, S. McKown, M. Talbert, "Translatability of Response Sets Used in Patient Reported Outcomes and Best Practices for Development," poster to be presented at the ISPOR 15th Annual European Congress (Nov 3-7, 2012).
Barbara A. Brandt, MA and Matthew Talbert, MA are Survey Research Analysts at Corporate Translations, Inc., 77 Hartland Street, East Hartford, CT, www.corptransinc.com