moneymonday
指數/外匯
明報APP

專家觀點>徐家健

人工智能對藥物研發有正面作用

【Presented by:WAW Creation】幾個月前,我在本欄問:衛生領域的人工智能可取代健康教育嗎?回應世界衛生組織發布的《Regulatory Considerations on Artificial Intelligence for Health》。我的答案是:新世紀衛生領域的人工智能不容易取代傳統的健康教育,原因是涉及選擇行為,健康的生活習慣始終只有透過教育才能有效地培養出來。

Presented by:WAW Creation


 

然而,人工智能技術在衛生領域的應用,至少包括(1)疾病檢測與診斷(AI in disease detection and diagnosis);(2)個人化疾病治療(personalized disease treatment);(3)醫療影像(AI in medical imaging);(4)提升臨牀試驗效率(clinical trial efficiency);(5)加速藥物研發(accelerated drug development)。今天我想討論的,是人工智能究竟如何加速藥物研發?

FDA對新藥有極嚴格測試

救命還是殺人?這是一項致命的藥物安全監管。1962年,美國食品及藥物管理局聲稱為了加強對消費者的保護,大大增加了新藥物在推出市場前的各種測試要求。芝大老師鮑士民(Sam Peltzman)的經典研究《An Evaluation of Consumer Protection Legislation: The 1962 Drug Amendments》發現,這項管制不但數以年計地延誤了不少能救人一命的藥物推出市場,在藥物生產成本及風險上升下,藥廠更放棄了一些藥物的研究。過度嚴謹的藥物安全法例,對市民的整體健康影響得不償失。

作為全國醫療負擔最重的國家,美國要到半個世紀後,藥業管制才有機會在特朗普政府的管治下略為放鬆。當時,我的博士論文導師莫里根(Casey Mulligan)正為特朗普政府擔任白宮經濟委員會首席經濟師。莫里根先在他的回憶錄《You're Hired!: Untold Successes and Failures of a Populist President》分享了「放鬆藥業管制導致藥物價格下降」的觀察,之後還發表了《Peltzman Revisited: Quantifying 21st-Century Opportunity Costs of Food and Drug Administration Regulation》一文:

放鬆藥業管制導致藥物降價

Peltzman’s work is revisited in light of two recent opportunities to quantitatively assess trade-offs in drug regulation. First, reduced regulatory barriers to drug manufacturing associated with the 2017 reauthorization of generic-drug user fee amendments were followed by more entry and lower prices for prescription drugs. A simple, versatile industry model and historical data on entry indicate that easing restrictions on generics discourages innovation, but this cost is more than offset by benefits from enhanced competition, especially after 2016.

新冠疫情加快批核工作

首先量化放鬆藥業管制透過增強市場競爭為消費者帶來藥物價格下降的好處,再分析新冠疫情疫苗加快批核的正面作用:

Second, accelerated vaccine approval in 2020 had unprecedented net benefits as it improved health and changed the trajectory of the wider economy. Evidence suggests that cost-benefit analysis of Food and Drug Administration (FDA) regulation is incomplete without accounting for substitution toward potentially unsafe and ineffective treatments that are outside FDA jurisdiction and heavily utilized before FDA approval. Moreover, the policy processes initiating the regulatory changes show an influence of Peltzman’s findings.

AI加速藥物研發

種種證據顯示,在平衡藥物安全與鼓勵藥物創新之間,美國食品藥物管理局的管制仍然是過於嚴格的。過於嚴格的監管,成本最終還是要轉嫁到消費者身上。一隻新藥的審批,一般須時12至15年,投資金額是數以十億美元計,而成功率卻是萬中無一。於是,人工智能加速藥物研發,我認為是前途無限的。儘管早前上市失敗,總部設於紐約和香港的英科智能(Insilico Medicine)是個非常值得參考的例子。

徐家健

Orientis首席經濟師

https://www.facebook.com/economics3.0