台灣癌症相對存活分析與生命損失

Abstract

截至2018年，癌症已連續36年蟬聯台灣十大死因第一位。隨著高齡化的趨勢，癌症所造成的社會負擔，已是台灣群體健康的重要議題。受限於死亡登記檔死因錯誤分類（Misclassification）以及垃圾死因碼（Garbage Code）的問題，目前進行的存活分析大多都是使用全死因死亡為終點事件的存活分析（Overall Survival）。然而，近期在全人口的癌症存活分析中，相對存活分析（Relative Survival）的概念已經越來越被討論。相較於傳統的存活分析方式，相對存活分析納入了競爭風險（Competing Risks）的概念，利用癌症造成的額外死亡風險率（Excess Mortality Hazard），將癌症病患的死亡分解成因癌症而死亡以及因其他非癌症因素而死亡兩部分。進而在假設癌症病患只會死於癌症，而不會死於其他競爭死因的假想情境中，推算出可用在不同族群、國家之間比較的兩個指標：淨存活率（NS）與設限平均淨存活時間（RMNST）; 以及真實情境中計算出來的兩個存活機率指標和兩個存活時間指標，分別為：癌症病患死於癌症的粗死亡機率（CPDcancer）、癌症病患死於非癌症因素的粗死亡機率（CPDother）、因癌症所造成的設限平均生命損失（NLYLcancer）以及因其他非癌症因素所造成的設限平均生命損失（NLYLother）。這四個指標提供了比以往更精確的癌症存活狀況，同時也進一步地描繪出因罹患癌症所造成的疾病預後狀況與生命損失。本研究藉由台灣癌症登記資料與死亡登記資料，納入1997年1月1日至2007年12月31日被診斷成癌症的病患，並追蹤至2017年12月31日。本研究藉由相對存活分析的各個指標，以性別、年齡、居住地都市化程度、收治醫療院所規模以及癌症期別作為分層變項，完整的描述台灣18種癌症的癌症病患因癌症與其他非癌症因素所造成的預後狀況與生命損失。

Cancer has been the leading cause of death for 37 years since 1982 in Taiwan. For evaluating the prognosis of cancer patients, population-based cancer survival analysis plays an important role. However, researchers are unable to distinguish the deaths due to cancer and others because the cause of death is unreliable, such as misclassified and garbage code. In contract to conventional survival analysis, relative survival analysis allows researchers to disentangle the impact on survival of the cancer from other causes of death by estimating excess mortality hazard of cancer even when the cause of death is unavailable or unreliable. In relative survival setting, net survival probability (NS) and restricted mean net survival time (RMNST) estimates the survival that cancer patients would experience if they could only die from the cancer under study. They are useful for comparing the prognosis of cancer between different populations or countries because both two measures do not depend on mortality from other causes. Crude probability of death (CPD) and number of life year lost (NLYL) are both calculate in the real situation and disentangle into two parts: Crude probability of death due to cancer (CPDcancer), crude probability of death due to other causes (CPDother), number of life year lost due to cancer (NLYLcancer) and number of life year lost due to other causes (NLYLother). These measures together present the whole picture about the death with probabilities and life lost after diagnose the cancer. We recruit patients diagnosed between 1997 and 2007 and followed up to December 31, 2017 from Taiwan population-based cancer registry. Include total 18 cancers. Estimation are stratified by age, sex, calendar year, development level of city, level of hospital volume and cancer stage. We finally provide eight measurements from both survival analysis and relative survival analysis. Each of them contributing differently to provide more comprehensive and detailed information to patients, clinicians, health policy makers.