AI-generated research brief — verify at source
Bilingual Consent Forms Improve Cancer Treatment Comprehension in LEP Patients
Among Bengali- or Sylheti-speaking adults with limited English proficiency undergoing a simulated cancer treatment consent consultation, those given a bilingual English-Bengali consent form were nearly four times more likely to understand treatment intent than those given an English-only form (multivariate OR 3.73, p = 0.01). Despite this improvement, only 15.7% of participants correctly understood treatment intent after reading translated written information alone, regardless of whether translation was performed by a machine or a professional human translator.
What Was Studied
This study examined two distinct questions relevant to cancer care equity: whether professional human translation of patient information outperforms unsupervised machine translation in accuracy and comprehension, and whether providing a bilingual consent form during an interpreted systemic anticancer treatment (SACT) consent consultation improves understanding of key information among patients with limited English proficiency (LEP). Both questions address a documented gap — written cancer treatment materials are predominantly available in English only, yet LEP is consistently associated with worse cancer outcomes.
How It Was Studied
This was a randomised study conducted in London, UK, enrolling 121 healthy adults who were speakers of Bengali or Sylheti with limited English proficiency. Participants underwent two sequential randomisations. In the first phase, they were allocated to read either a machine-translated or a professionally (human) translated version of a written SACT patient information booklet. One to two weeks later, 91 of those participants returned for the second phase: a simulated SACT consent consultation involving a doctor and an interpreter, during which participants were randomised again to receive either a standard English-only consent form or a bilingual English-Bengali consent form. Comprehension of treatment intent served as the primary outcome, assessed after each phase.
What Was Observed
- Comprehension of treatment intent after reading the translated booklet was low overall — only 19 of 121 participants (15.7%) met the primary outcome — and there was effectively no difference between machine and professional translation on this measure (multivariate OR 0.99, p = 0.99), meaning the two approaches performed identically poorly in supporting understanding.
- Despite equivalent comprehension outcomes, translation quality differed markedly: the machine-translated booklet contained 11 meaning-changing errors, compared to only 1 in the professionally translated version. This distinction suggests that accuracy at the text level does not straightforwardly translate into measurable differences in patient comprehension under these study conditions.
- In the consent consultation phase, participants randomised to the bilingual English-Bengali consent form had approximately 3.7 times higher odds of understanding treatment intent compared to those receiving the English-only form (multivariate OR 3.73, p = 0.01), representing a statistically meaningful and clinically relevant improvement in a setting that more closely mirrors real-world clinical practice.
Why This Matters
These findings challenge the assumption that translating written patient information — regardless of method — is sufficient to bridge health literacy gaps for LEP populations in oncology. The near-universal failure to comprehend treatment intent from the booklet alone, across both translation types, indicates that text-based materials in their current format may be fundamentally inadequate for this purpose. The positive effect of the bilingual consent form within a supported, interpreted consultation setting points toward multicomponent, interaction-based approaches as a more productive direction for research and intervention design in language-discordant cancer care.
How to Read This Result
The study used a healthy volunteer sample in a simulated rather than real clinical setting, and was limited to two closely related language varieties (Bengali and Sylheti), which restricts direct generalisation to actual cancer patients or other LEP populations.
Limitations
The authors note that most participants across both translation arms failed to comprehend crucial treatment information, suggesting that written formats alone — regardless of translation quality — are insufficient and that alternative information modalities warrant investigation. The study also acknowledges that unsupervised machine translation introduced substantially more meaning-changing errors than professional translation, a finding with practical implications even though it did not produce a detectable difference in comprehension scores under these trial conditions.