This article is more than 5 years old. Information may no longer be current.

Google Translate accurately translates clinical trial data

Google Translate can effectively and accurately translate non-English randomized clinical trials for systematic reviews, according to a research report published in the Annals of Internal Medicine.

“A critical marker of high-quality systematic reviews is the identification and inclusion of all relevant, important studies,” Jeffrey L. Jackson, MD, MPH, of the Medical College of Wisconsin, and colleagues wrote. “Up to 78% of systematic reviews have language restrictions; as a consequence, most reviews (93%) exclude at least one randomized controlled trial.”

Results from a previous study that assessed Google Translate recommended that researchers exercise caution when using it to translate non-English language studies. The service has been updated since then, prompting Jackson and colleagues to review its accuracy in translating data from randomized control trials for systematic reviews.

Researchers used PubMed to identify randomized control trials published from Jan. 1, 2000, to Dec. 15, 2018 in nine languages other than English (Chinese, French, German, Italian, Japanese, Korean, Romanian, Russian and Spanish).

Each trial included in the study was translated through Google Translate, and a researcher who could not speak any of the non-English languages abstracted data from the translation. A native-speaking physician abstracted data from the original version of the study. Abstracted variables included study characteristics, outcomes, adverse events and withdrawals.

Abstractors rated the quality and risk of bias for each study and were blinded to each other’s findings.

Researchers compared the accuracy of data taken from the translation and the original language version by calculating the percentage agreement between the two. Agreements greater than 80% were considered good agreements.

A total of 45 articles with 6,370 variables were included in the study, with an overall percentage agreement of 91% (5,791 variables). The percentage agreement for different languages included in the study ranged from 85% to 97%.

For quality ratings, the agreement was 96% for quality scores and 87% on risk of bias ratings. Only one disagreement in ratings was due to a difference in translation, and all other disagreements were based on differences of the interpretation of items in the quality standards.

“We conclude that Google Translate is a viable, accurate tool for translating non-English-language trials for the purpose of conducting systematic reviews,” Jackson and colleagues wrote.

“Given the increased ability to perform sensitivity analyses and the potential for more precise estimates and generalizability of results, we recommend maximizing the number of trials available by using Google Translate to include non–English-language trials in systematic reviews,” they continued. – by Erin Michael

Disclosures: The authors report no relevant financial disclosures.