制药行业调查:AI在设备确认中的应用(114名参与者结果)
制药行业调查:AI在设备确认中的应用
一项针对114名制药及API行业参与者的调查显示,46%的受访者在设备确认中不使用AI,22%表示每天使用,比例高于2024年调查中83%未在GMP环境使用AI的结果。设备确认仍以纸质方式为主(48%),52%采用混合方式,仅9%使用AI工具,但软件工具使用较2024年明显增加。58%的参与者强烈不同意AI几年后会消失,57%认为AI不会取代专家知识。
Currently, one of the 'hottest topics' in the GMP environment is the use of Artificial Intelligence (AI). Is it also possible to use AI in the field of equipment qualification? We already surveyed this in 2024. Now two years have passed. What has changed? What is the current view of the use of AI today?
目前,在GMP环境中,“人工智能(AI)的应用”已成为最热门的话题之一。那么,AI是否也能应用于设备验证领域?我们已于2024年对此进行过一次调查。如今两年已过,情况发生了哪些变化?当前对于人工智能的应用又持怎样的看法?
114 participants took part in the survey this time. Not every participant answered every single question, and multiple answers were possible in some of the questions. Rounding also means that the totals do not always add up to 100%.
本次调查共有114名参与者。并非每位参与者都回答了所有问题,部分问题允许多选,因此数据经过四舍五入处理,总和未必恰好为100%。
The results are presented below.
调查结果如下:
The majority of the 114 participants came from the pharmaceutical industry or from active pharmaceutical ingredient (API) manufacturers (46%), and 20% of participants work in engineering/consultancy. 13% of participants work with biologics. 55% of respondents came from companies with more than 500 employees. The remainder is divided as follows: participants from companies with 101-500 employees (18%), 12% from companies with 51-100 employees and 16% from companies with 1-50 employees.
114名参与者中,大多数来自制药行业或原料药(API)生产企业(占46%),20%的参与者就职于工程或咨询公司,13%从事生物制品相关工作。55%的受访者来自员工人数超过500人的企业。其余人员分布如下:员工人数在101至500人之间的企业占18%,51至100人之间的占12%,1至50人之间的占16%。
The first question on AI aimed to establish what participants' previous AI training looked like. The majority (40%) had more than 20 hours of training, 24% had 5-20 hours and 36% had 1-5 hours.
关于AI的第一个问题旨在了解参与者此前接受过多少AI培训。大多数(40%)的参与者接受了超过20小时的培训,24%接受了5至20小时,36%接受了1至5小时。
The next question targeted the use of AI directly within equipment qualification: how often do you use AI during equipment qualification activities? Here, 46% stated that they do not use AI. However, 22% of participants reported daily use. Other responses were once a week (18%) or once a day (14%). In 2024, 83% of participants generally did not use AI in the GMP environment.
下一个问题是关于在设备验证中直接使用AI的情况:您在设备验证活动中使用AI的频率如何?其中,46%的受访者表示未使用AI。然而,有22%的参与者表示每天都会使用。其他回答包括每周一次(18%)或每天一次(14%)。在2024年,83%的参与者通常在GMP环境中未使用AI。
And in which areas is AI used? The largest group was 'other', which is discussed in more detail below. 25% of responses related to writing risk assessments, 15% use AI to analyse data, 9% to write qualification reports and 7% to write qualification protocols. 5% use AI to write user requirements. And what falls under 'Other'? The responses were very varied. The most frequently mentioned were
AI在哪些领域被使用?其中最大的类别是“其他”,以下将对此进行更详细的说明。25%的受访者表示涉及撰写风险评估,15%用于数据分析,9%用于编写确认报告,7%用于编写确认方案。5%使用AI来撰写URS。那么“其他”包括哪些内容?回答非常多样,最常提及的是:
all of the above-mentioned activities (9%)
上述所有活动(9%)
research on regulatory requirements and guidelines (7%)
研究监管要求和指南(7%)
The next question (multiple answers possible) concerned how equipment qualification is generally carried out. 48% of participants stated that they do this purely on paper, 26% use software-based tools and 52% work in a 'hybrid' way. Just 9% reported using AI tools. In 2024, over 60% still worked purely on paper, and software use stood at 9% at that time.
下一个问题(可能有多个答案)涉及设备确认的普遍实施方式。48%的参与者表示他们仅通过纸质方式完成,26%使用基于软件的工具,52%采用“混合”方式工作,仅有9%的人报告使用AI工具。2024年,仍有超过60%的参与者仍完全依赖纸质流程,而当时软件使用率仅为9%。
And what does the outlook for the future look like?
那么未来的发展前景如何?
We asked participants what they expect. The responses could be grouped into four categories: 'strongly disagree', 'disagree', 'agree' and 'fully agree'. Because multiple answers could be selected, a somewhat mixed picture emerges in each case.
我们询问了参与者他们的预期。回答可以分为四个类别:“强烈不同意”、“不同意”、“同意”和“完全同意”。由于可选择多个答案,每个类别呈现出相对复杂的情况。
However, some answers are fairly clear. Asked whether AI will have disappeared again in a few years' time, 58% answered 'strongly disagree'. The answers to the question of whether AI will take over most GxP activities, except for release, were also interesting. Here the answers were almost evenly split at 50:50, with 45% disagreeing and 40% agreeing. Yet despite AI, expert training retains a high value: 57% of participants do not believe that AI will replace expert knowledge. And 60% agreed with the statement that guidelines will become even more important. High agreement (51%) was also received for the question of whether the quality of qualification activities will increase through AI because human errors are eliminated.
然而,一些答案相当明确。当被问及几年后AI是否会再次消失时,58%的人回答“强烈不同意”。关于AI是否会接管除发布之外的大多数GxP活动的问题,答案同样引人关注。这一问题的答案几乎平分秋色,45%的人表示不同意,40%的人表示同意。尽管有AI的存在,专家培训仍具有重要价值:57%的参与者认为AI不会取代专家知识。此外,60%的人认同指南的重要性将进一步提升的观点。对于AI是否能因消除人为错误而提高资质认证活动的质量,也有高达51%的人表示高度认同。
Conclusion: The use of AI within equipment qualification is still cautious, but higher than in the 2024 survey. Those who do use it, however, do so very extensively. Equipment qualification is still heavily paper-based, but the use of software-based tools has increased markedly since 2024.
结论:在设备确认中使用AI仍较为谨慎,但比2024年的调查有所提高。然而,那些实际使用AI的人也进行了广泛的应用。设备资质认证仍然以纸质方式为主,但从2024年以来,基于软件的工具使用已显著增加。
原文来自ECA网站,点击“阅读原文”可查看原文。
来源:GMP干货 · mp.weixin.qq.com