|
Caroline Houchin MA, Translation Coordinator, and Diane Wild MSc, BSc, Director, Oxford Outcomes Ltd., Oxford, UK, and
Sheryl Pease MBA, Pfizer, Inc., Ann Arbor, MI, USA
Same Language, Different Country: Do All Roads Lead
to Rome?
Due to increasing interest in assessing Patient Reported Outcomes in
multi-national clinical trials, it is necessary to have cross-culturally valid
instruments in order to pool data across countries. With the changing of
boundaries and the fluidity of populations, choosing a language to correspond
to a study site is an increasingly pertinent issue, and can be complicated.
While one country can host several languages, certain languages are
spoken in multiple countries, for example English, French, Spanish, Russian,
and Arabic.
Determining Appropriate Methodology
Those involved in conducting multinational studies have sought to determine
the methodology that should be employed for translating a questionnaire for
particular groups of countries that share a language. However, there is no
consensus on this issue, and certainly an absence of systematic evidence to
support any single methodology over another. Thus, decisions are currently
based on the ad hoc accumulation of experience and discussion by those
involved in the planning of multinational trials and those involved in the translation
and linguistic validation of PROs.
Despite the absence of guidance on the best strategy for same language/
different country translations, we have chosen to outline the relevant issues
and to describe and critique current translation methodologies. It is hoped
that this paper will be helpful in assisting with the decision making process
in this area.
Potential Methodology
Four key methodologies exist for the development of a measure in one language
for use in different countries. The first and simplest of these is to
develop a single translation in a single country, to be used worldwide, on the
assumption that because of its “simplicity”, the translated measure will be
understood in other countries where the same language is spoken. The second
methodology is to create a single version in one country which is
reviewed early in the translation process in order to assess its comprehension
in another target country. A third, alternative approach is to determine
from the outset that a separate translation will be produced for each country,
with translators working independently. Finally, it is also possible to establish
from the outset that representatives from all countries involved will work
together to a produce a single, universal language version to be used in multiple
countries.
The methodologies are described below in the form of responses to frequently
asked questions in this area.
Is It Possible To Create A Single Version In One Country And Use It
Worldwide?
There are advantages as well as disadvantages to this methodology, many of
which are highlighted from the opposite perspective in subsequent sections.
The key benefits are time and money: with only one translation needed, the
translation can be used throughout the world as soon as it is completed.
However, there are significant risks when using one language version that
has not been checked for its suitability in other countries, including using language
that is unfamiliar to respondents, brand names with which respondents
cannot identify, and inappropriate idioms. One way of overcoming
some of these may be to pilot test the translation in the appropriate patient
groups in each of the countries where it is needed. Though more costly and
time-consuming, should any changes be required as a result of the testing,
the end result would be country-specific versions.
Would A Better Alternative Be To Develop A Single Translation In One
Country And Review It To Assess Its Comprehension In Another Target
Country?
Preferably performed early in the translation process, one or two in-country
consultants (native speakers of the target language, residing in the country
in question and preferably with expertise in health outcomes and related
medical research as well as in translation and/or translation management)
can review the translation, and pilot testing in the appropriate patient group
would usually be conducted. If a language version already exists this method
will be quicker, but this may not be possible (see point b below). Based on
the review the following actions can be taken:
a. Where no issues arise, there is no need for a separate version to be developed.
b. If issues do arise, they can be fed into the translation process with the aim
of producing a single harmonised version of the measure. This may not be
possible if the original version has already been used in trials or as the basis
for other translations. In these cases it is necessary to create separate language
versions, as in point c below.
c. Where significant irreconcilable differences arise, separate language versions
can be developed for each country.
The in-country review of an existing language version will therefore result in
language versions that are appropriate for use in different countries, whether
this means separate language versions or one single version for use in multiple
countries.
Again, several advantages and disadvantages are associated with this
method. For example, checking that the vocabulary used is appropriate for
each country ensures that the translation is more understandable and familiar
to respondents, which may make them more comfortable completing the
measure. This may especially be the case with older patients who are used
to a more traditional language with fewer foreign loan words.
The inclusion of some items may have been due to their particular importance
in the culture within which the measure was developed. These items
may be less relevant in other cultures, where there may be certain related
questions of greater importance that have not been included. Where this is
the case, and if the developer can be consulted, decentering can occur. With
the insertion of additional phrases or words into the translation for the new
country, the original measure may consequently also be altered, if deemed
acceptable and necessary by the developer. This can only happen if the
measure has not already been translated in its original form On a non-linguistic level, there are other aspects which should be considered
in terms of cultural relevance. For example, with the use of the metric system
in some countries and the imperial system in others, cultural adaptation
is required. When culturally adapting an item which includes a unit of measurement,
the employment of a sensible degree of accuracy is recommended,
for example 100 yards is better adapted to 100 metres than the more precise
91 metres. Similarly, product brand names may not be known in all
countries. Substituting a cultural equivalent in these cases is a necessity.
Another non-linguistic issue is the choice of response scales; an additional
explanatory sentence or an example may be necessary in some countries to
aid comprehension. Finally, the content of the measure might be deemed
unsuitable for particular countries. For example, a questionnaire with items
about sexual behaviour or alcohol consumption may cause offence in some
cultures.
There are several disadvantages associated with developing a translation in
one country and reviewing it for use in another country. It can be difficult to
determine whether the changes suggested during a review of an existing
translation are essential or rather the preferences of those doing the reviewing.
It may also be a challenge for alternative, culturally sensitive wording to
be found. In addition, logistically, using just one language version rather than
numerous versions is more straightforward.
To demonstrate the variety in the types of changes necessary when reviewing
an existing version, some examples are as follows. The Catalan for Spain
translation of “education” (“Educació“) in the Polytrauma-Outcome Chart
had to be changed to estudis for use with Catalan speakers in France. This
is a simple vocabulary change.
A grammatical change was necessary when reviewing the Brazilian
Portuguese translation of the same measure, for use in Portugal. In Brazil, the
pronoun “you” (você) needs to be inserted in the phrase “have you become...”
but pronouns are rarely used in this way in Portuguese for Portugal; instead the
verb ending is relied upon to indicate the subject of the verb.
Cultural changes can also be necessary: in the Parkinson's Disease Quality
of Life Questionnaire (PDQ-39) one question asks the respondent whether
they “lacked support in the ways you need from your spouse or partner?”.
There was no problem translating this into Russian for use in Russia, but
when the translation was reviewed for Ukraine it was necessary to remove
“or partner” as Russian-speaking Ukrainians still mean “business partner”
by this term. Another cultural change, this time in the Disabilities of the Arm,
Shoulder and Hand (DASH) questionnaire, was to change “10lbs” to “5kg”
in South African English. Lastly, in the Mini Mental State Examination
(MMSE) respondents are asked “What is the county [we are in now]?” This
was translated as Bundesland in German for use in Germany but when
revised for Switzerland the term had to be replaced by kanton, as Switzerland
is divided into cantons.
Should Separate Versions Be Produced For Each Country?
The advantages and disadvantages of this approach are similar to those of
the previous approach. However, one clear disadvantage with this process
over the previous method is that creating separate language versions from
the outset leads to a high risk of different translations, simply due to translator
styles. It cannot be determined if one version would have been appropriate
for use in both or all countries in question, or if all the differences between
versions were absolutely necessary.
One advantage of this approach, however, is that the translators are not
restricted by vocabulary and terms that may not be optimal for their country.
For countries that share the same language but are different culturally, geographically
or linguistically, this is also a benefit.
Is It Feasible To Create A Universal Language Version For Use In All
Countries?
This approach involves discussion at every stage between representatives of
all countries involved, and also pilot testing with a relevant patient group in
each of the target populations.
The production of a single harmonised version to be used in more than one
country can be time consuming. Agreement is required at all points between
consultants and reaching consensus might be difficult. This differs according
to the countries and languages involved. One example of a problem
encountered when trying to create a unified Spanish translation for use in
both Spain and Mexico was in an Interactive Voice Response System (IVRS)
bowel study script. When respondents were asked to “enter your PIN,” each
in-country consultant used a different term for “enter” and could not accept
the other country's word, as this had a different meaning in their own country.
In the same measure, translations for AM and PM had to differ because
while people living in Mexico use these terms, they are not frequently used
or understood in Spain. This latter country used “morning” and
“afternoon/evening” instead. In the Satisfaction with Oral Anti-Diabetic
Agents Scale (SOADAS) a compromise was reached between France and
Canada as numbers under ten are spelt out as words in Canada, whereas in
France these are left as numerals. The final unified translation included the
word followed by the figure in brackets. As we can see, when creating a unified
translation, the final version may end up being a compromise: “acceptable”
and not “ideal” for any of the countries concerned. It may also lack the
richness of idioms that is often present in a language version for use in just
one country. The principal advantage of this method, though, is that if only
one version exists this will be simpler for those involved in running multinational
trials.
Decision-making Criteria
Having explored the key methods for producing translations of a measure
into one language for use in multiple countries, it is necessary to consider
the following issues:
A. How much time and financial resources are available and what is the attitude
to risk and error margin?
B. Developer requirements must be adhered to. Many developers do not
advocate a specific strategy with regard to same language/different country
translations; however there are some notable exceptions, including the
EuroQol and the EORTC group. Thus it is necessary to take account of:
• The developer's translation requirements.
• Previously employed translation strategies of the measure in question.
C. The measure itself should be examined in order to determine the presence
of characteristics that are likely to result in country-specific language
differences. Such characteristics include:
• References to country-specific health care or education systems;
• References to product names;
• Complicated vocabulary that will present challenges to the respondent;
• Response scales that can present cognitive challenges to the respondent;
• Areas of enquiry that may offend the respondent; and
• Technical details; for example ePRO scripts may include terminology that
is unfamiliar in certain countries.
D. Thinking about the languages and countries in question is also important.
The strategy to be employed will differ according to which language and
which countries are being considered as certain countries and languages
share more linguistically and culturally than others. Languages that are commonly
translated for multiple countries include: English, Spanish, French,
Portuguese, Chinese, Russian, Arabic, and Malay.
E. Finally there is a need to determine whether a language version is already
in existence.
Conclusion
There are no hard and fast rules relating to the method that should be
employed when translating a measure into one language for use in multiple
countries. Guidelines can be followed which take into account the above criteria,
but with the high number of factors involved in deciding on the best
translation methodology to use, a number of options may still be available.
The complexity and characteristics of the questionnaire might well be as
important to the decision-making process as the language and countries
involved. It may be beneficial to conduct more research into the methods
described above, including comparing psychometric results of translations
for use in the same countries but obtained by different methods. However,
should conclusions be reached as a result of this research, it is still important
to consider future projects on a case-by-case basis, due to differing
characteristics of individual measures? Considering the highlighted pros and
cons of each method should nonetheless facilitate the decision-making
process.
|
