# Translation approaches to support systemic anti-cancer therapy consent for individuals with limited English proficiency

## TL;DR

This study compares machine and professional translations of cancer treatment information and finds that bilingual consent forms improve understanding for non-English speakers.

## Contribution

The study introduces a novel evaluation of bilingual consent forms in improving comprehension for patients with limited English proficiency.

## Key findings

- Bilingual English–Bengali consent forms were associated with higher odds of understanding treatment intent.
- Machine translation introduced more errors than professional translation in medical information booklets.
- Most participants, regardless of translation type, did not comprehend crucial information from translated booklets.

## Abstract

Limited English proficiency (LEP) is associated with poor cancer outcomes, and written information about systemic anticancer treatment (SACT) is often provided in English only. Unsupervised machine translation of medical information is common, but its effectiveness and accuracy are unclear. This study aimed to (a) compare professional and machine translation of written SACT information and (b) investigate whether providing a bilingual SACT consent form altered comprehension of key information during an interpreted SACT consent consultation.

This randomised study included healthy, Bengali- or Sylheti-speaking adults with LEP across London, UK. Participants were randomised twice. First, they were allocated to either a machine translation or professional (human) translation of an English language SACT information booklet. Second, 1–2 weeks later, participants underwent a simulated SACT consent consultation with a doctor and interpreter, and were allocated to either a conventional English-only SACT consent form or a bilingual English–Bengali consent form.

After reading the translated booklet, 19/121 participants (15.7%) met the primary outcome of understanding treatment intent, with no difference by translation type (multivariate OR = 0.99, p = 0.99). The machine-translated booklet contained 11 meaning-changing errors, compared to 1 in the professional translation. Of 91 participants completing the consent consultation, randomisation to the bilingual translated SACT form was associated with higher odds of understanding treatment intent (multivariate OR = 3.73, p = 0.01).

In this study, a bilingual English–Bengali consent form improved comprehension amongst LEP participants and may hold value in other language-discordant settings. Unsupervised machine translation of a patient information booklet introduced more errors than professional translation, but most individuals receiving either translation type did not comprehend crucial information; other information formats should be explored.

University College London Research Ethics Committee approved this study as a low-risk application on 28 October 2024 at 15:15, reference number: AH/2024/52-6625/003. The statistical analysis plan of the study was registered in the Open Science Framework, https://osf.io/axg5d.

The online version contains supplementary material available at 10.1007/s00520-026-10464-w.