| 摘要: | 醫療器材產業市場的蓬勃發展，許多科技業、電子業都跨足醫療器材產業的領域。全球人口老化及非傳染性疾病(NCD)死亡人口比例逐年增加，以上原因將促使醫療器材產業有龐大的市場需求。醫療器材從設計到量產上市階段都有相關法規，而且上市前也有嚴謹的審查制度，但為何每年、每月都有召回事件呢？
基於美國FDA 2002年至2019年召回資料統計，總召回事件為38,180件，重覆事件扣除後為15,209件。對應分析可以在低維度空間找出各變數之間的相對關係，本研究以召回事件最多的前30間公司，扣除N/A機型，探討前十大召回原因，研究樣本為2,021件的召回事件，運用對應分析進行召回機型與召回原因的分析，以達下列二項目的：
(1) 美國FDA在醫療器材上市前透過嚴謹的審核機制，還是有召回的可能，是否在哪個環節出現問題以探討召回的可能性。
(2) 整合對應分析和集群分析法建構召回地圖，找出召回原因和醫療器材類別之間的關聯性，並比較不同期間的召回事件。
研究結果如下：
(1) 根據30間樣本公司的召回機型觀察到，90%以上甚至更高的比例皆透過510(k)上市前通知(PMN)得到許可上市。因為510(k)審查只從書面判定是否符合規範，故本研究認定仍有未查覺到的潛在風險。
(2) 2002年至2019年間的召回資料進行分析，可分為以下三集群：
①第一群為「包裝過程控制類」的失誤，小型拋棄式的醫療器材發生頻率較高。
②第二群可細分為「研發設計類」、「製造流程設計類」及「流程管制類」的失誤，分析系統的醫療器材及外科或植入式器材發生的頻率較高。
③第三群為「軟體設計」的失誤，判讀系統的醫療器材發生的頻率較高。
以2011年為分水嶺，在群組間比較，各機型於二個時期中除了小型拋棄式的醫療器材，其他機型並無明顯的變化差異；在群組內比較，各機型在二個時期皆有變化，其中軟體設計類由DIHD(免疫學和血液學器材)變為DHT1A(眼科器材)，讓我們最容易理解的是3C產品造成人們的眼部疾病增多，此類的器材需求也增加。
因此，由召回地圖探討各機型其召回原因的發生頻率和趨勢，擬定風險管理之
相關事宜，可以減少召回事件發生的可能。
;Based on the statistics of the US FDA′s recall data from 2002 to 2019, the total recall events were 38,180, and the repeated events were 15,209 after deduction. Correspondence analysis can find the relative relationship between variables in a low-dimensional space. This study selects the top 30 companies with the most recall events, deducts N/A models, and discusses the top ten recall causes. The research sample is 2,021 recall events. Correspondence analysis is used to analyze the recall models and recall causes to achieve the following two goals : (1) The US FDA still has the possibility of the recall through the rigorous review system before the medical device is placed on the market. Whether there is a problem in which link to explore the possibility of recall; (2) Integrate correspondence analysis and cluster analysis to construct a recall map, find out the correlation between the recall cause and the category of medical device, and compare the recall events in different periods.
The research results are as follows: (1) According to the recalled models of 30 sample companies, more than 90% or even higher proportions are permission for listing through 510(k). Because the 510(k) review only determines whether it meets the specifications in writing, this study believes that there are still undetected potential risks; (2) Analysis of recall data from 2002 to 2019 can be divided into the following three clusters: ① The first group is the "Packaging Process Control" mistakes, and the frequency of small disposable medical devices is relatively high; ② The second group can be subdivided into "R&D and Design", "Manufacturing Process Design" and "Process Control" mistakes, "medical device of analysis system" and "surgical or implantable device" occur more frequently; ③The third group is the mistake of "software design", and the "medical device of discrimination system" has a higher frequency.
Taking 2011 as the watershed, comparing between the groups, each model in addition to "small disposable medical device", there is no obvious difference between other models in the two periods; Within the group comparison, each model has changed in two periods. Among them, the "Software Design Category" has changed from DIHD (Immunology and Hematology Devices) to DHT1A (Ophthalmic Devices). The easiest thing for us to understand is that 3C products have caused an increase in people′s eye diseases, and the demand for such equipment has also increased.
Therefore, the recall map is used to discuss the frequency and trend of the causes for the recall of each model, and to formulate related issues of risk management, which can reduce the possibility of recall events. |
