To meet the challenges caused by the Covid-19 pandemic, RWS integrated machine translation (MT) technology into its workflow.
Pharmacovigilance refers to the practice of collecting, detecting, assessing, monitoring and preventing adverse effects (AEs) caused by new drug substances. AEs can be caused by using medicinal products within or outside the terms of their marketing authorisations (such as off-label therapeutic use, overdose, misuse, abuse or medication errors), as well as from occupational exposure. They are recognised by the World Health Organisation (WHO) alongside a number of other issues that pharmacovigilance must check for, including interactions between medicines, abuse and misuse of medicines, counterfeit medicines, medication errors, and suboptimal efficacy.
Before drugs are authorised, evidence of their safety and efficacy is limited to results from clinical trials, so it is essential that pharmacovigilance continues after a therapy has been approved. After AEs are recorded, they must be translated and put back out to doctors and scientists for processing.
Adapting pharmacovigilance to deal with surging case reports
The following is adapted from a presentation given by Dave Paré and Silvio Scozzari at the DIA 2022 conference
Among the many effects the Covid-19 pandemic has had on the pharmaceutical industry is a substantial increase in the number of AEs being reported from drug trials, due to the fast-tracked development, approval and marketing of vaccines for the SARS-CoV-2 virus. Many of these were reported in different languages and file formats, and the numbers of AEs being reported from a single trial often varied greatly depending on how many patients were enrolled in the trial.
For multinational translation company RWS, processing vastly scaled-up numbers of AEs required the company to redesign its workflow in order to stay ahead of the curve. Prior to the pandemic, RWS’ safety programmes involved translating case reports into a variety of languages, while the company could expect varying numbers of case reports depending on the indications that new products were designed for. In addition, the company was used to tailoring its approach to suit the needs of its clients. As such, RWS felt confident that it had the basics to support what it describes as a ‘tidal wave’ of incoming case reports. There were, however, aspects of its translation programme that needed to be revised.
The process of actioning case reports is typically split into six stages: case receipt, triage (deciding whether a case should be classified as serious, non-serious or a non-event), data entry, quality review, medical review, and submission. Non-serious cases should ideally be processed within 48 hours of being received, but for serious cases, this timeframe is reduced to 24 hours. Therefore, any delays in translation (which happens at the first stage) can prolong the whole process and push out the subsequent stages, meaning that cases may not be actioned to deadline. With RWS receiving high numbers of case reports each day, and many of them needing translating from other languages and file formats, the company turned to new methods to enhance its translation process and meet case deadlines.
Machine translation as a time-saver during pharmacovigilance
To meet these challenges, RWS integrated machine translation (MT) technology into its workflow in combination with conventional human translation. MT uses artificial intelligence (AI) to replace words in one language with their equivalent in another language. Although this saves time at the translation stage when processing case reports, it can easily lead to mistranslations, since many MT applications cannot recognise whole phrases or their closest translations in the target language. Since it is essential to ensure new drug products do not cause serious adverse effects, many drug companies feel that the risk of mistranslation outweighs any potential benefits offered by MT.
In view of this risk, RWS used its safety-specific MT engines to translate cases before having them checked and edited as necessary by human translators. The company tested the applicability and usefulness of MT for translations by carrying out productivity assessments ‒ this involved running scenarios of typical manual translation against the company’s standard translation engines, then transferring this to the company’s highly-trained translation engines – in the event, these performed very well for certain types of documents. This all had to be done in the background while RWS was processing the waves of new requests being received each day.
Based on results from the productivity testing, RWS took a hybrid approach that involved applying MT strategically to certain cases where it was expected to deliver reliable results. This had the effect of reducing triage times by 25-40% and allowing the company to scale its capacity to match client demand.
Several of RWS’ customers had already made use of MT during the triage phase of safety programmes, as it allowed safety specialists to quickly assess incoming case reports before they were sent to RWS for translation. The technology can also prove beneficial later on in the safety programme lifecycle, when it comes to translating AE and special adverse effect (SAE) case reports, although this depends on how the content is configured upstream, such as the file format and the layout of the files.
As an MT platform, RWS’ Language Weaver Cloud system acted as an instant and secure alternative to Google Translate for its teams. Contract research organisations’ (CROs) safety teams could use the platform to upload cases in any file format to get a good gist translation, which enabled the company to determine if a case should be classed as serious or non-serious. These were then routed to the appropriate translation workflow to be given a certified translation. RWS’ teams also used Language Weaver Cloud to translate article abstracts, most of them from medical journals, to assess whether the content seemed relevant; any articles that were deemed relevant were then purchased and submitted for certified translations.
Redesigning the translation process to increase efficiency
The increase in case reports brought about by the Covid-19 pandemic impacted RWS’ dedicated safety translation team in several ways. Not only did the team and its capabilities expand, but its ability to scale its operations has been permanently altered.
One of the first areas that the company looked at was its traditional translation process. For translations that are certified by the ISO, several steps are required to ensure ballot certification. To maintain these quality standards while working to a greatly reduced timeframe due to the higher numbers of cases, RWS had to remove as many administrative steps as possible from its process. The company integrated into its translation management systems a built-in hybrid process, in which team members could select on-demand machine translation and post-editing or a traditional translate, edit and proofread process based on set criteria. This was all driven through an expedited triage process to address the greatly increased numbers of cases.
“RWS is the largest language service provider and we happen to employ the largest number of full-time in-house linguists of any language service provider, which gave us control and flexibility when we needed it most.”Dave Pare
RWS’ resource management team also had to develop new methods to ensure the availability of linguistic resources and ‘any language, anytime’ capabilities. The multinational team works in coordination to cover all time zones under a ‘follow the sun’ methodology, so it was necessary not only to reduce or eliminate downtime for its own qualification processes, but to reduce any time zone variations needed to cover for various languages in various regions.
Processing more case reports while meeting accuracy standards
By adopting MT practices in conjunction with human translation, RWS has been able to drastically increase the numbers of case reports it can process without compromising accuracy. The company has cemented its position as a reputable partner for CROs, with several major players in the race to develop vaccines and treatments for Covid-19 trusting it with critical projects. It has translated more than 6.3 million words of safety report content with a 99.94 quality rating, and handled more than 44,000 minutes of over-the-phone interpretation (OPI) and 1,500 calls. In addition, RWS has delivered more than 300 language proficiency tests and over 900 hours of interpreting for CRO safety specialist teams, as well as reducing translation delivery timelines by up to 60%.