影響首次腦中風病患日後死亡的因素─中風復發乃唯一替代終點？

Abstract

背景 關於腦中風死亡的替代終點，包括腦中風復發及腦血管疾病的生物標記，幾乎沒有文獻探討及證明其適用性。 目標 確認主要危險因子（含中風復發及生物標記）對首次腦中風病患長期預後的影響；評估為何復發及其他生物性指標為腦血管疾病死亡的合適替代終點，以傳統Cox 回歸模式及多階段隨機模式驗證。 方法 研究族群為一進行臨床隨機試驗的466位首次腦中風病患，該研究目的在評估aspirin是否減少中風復發，並收集與腦中風長期預後相關的資料。本研究以替代終點的「治療效益被解釋率(proportion of treatment effect explained, PTE)」觀點來闡述復發狀態的貢獻，採用Cox 迴歸模式檢定（復發狀態包括：時間相關與非時間相關）。迴歸分析併入多階段隨機模式中，以評估腦中風復發、生物因素及治療三者作為替代終點的地位。 結果 以多變數分析發現：影響首次腦中風之長期預後有四個預測因子，含復發、治療組、總膽固醇、及血糖。調整復發效應之後，以PTE法分析，發現後三者為影響長期預後的獨立因子。以多階段模式分析，復發為腦血管疾病死亡強而有力的替代終點；而治療、總膽固醇、及血糖為另外三個替代終點。 結論 本研究核可及驗明「腦血管疾病死亡」的替代終點有：復發，與其他三因子：血糖、總膽固醇、治療組別（在無復發變項存在時）。 關鍵字 腦中風死亡;腦中風復發;替代終點;多階段馬可夫模式

Background Surrogate endpoints of cerebrovascular death, including stroke recurrence and biological factors of cerebrovascular disease for long-term prognosis, were barely address and qualified. Aim To identify significant prognostic factors (including recurrence and biological markers) responsible for long-term prognosis of first ischemic stroke, and to evaluate where recurrence and other biological markers are good surrogate endpoints for long-term prognosis by using conventional Cox regression model and also multi-state stochastic models. Method A total of 466 patients with first stroke that were the study population designed for a randomized controlled trial on the comparison of treatment efficacy using aspirin in reducing recurrence was used. A series of predictors for long-term prognosis of death from cerebrovascular were collected. Cox regression model with or without time-dependent covariate of recurrence was used to model the proportion of treatment effect explained (PTE) for surrogate endpoint. The multi-state stochastic model with the incorporation of regressions was used to assess the role of surrogate endpoint on recurrence, other biological factors and treatment. Result By using multivariate analysis, we identified four predictors for long-term prognosis of first ischemic stroke, including recurrence, treatment, total cholesterol, and serum glucose. By using PTE method, we found the latter 3 factors were independent factors for long-term prognosis after adjustment for recurrence. The multi-state model found recurrence is a strong surrogate endpoint for long-term prognosis of cerebrovascular death. In addition, treatment, total cholesterol and glucose were three another surrogate endpoints. Conclusion Surrogate endpoints with recurrence and other factors without recurrence were qualified and identified for long-term prognosis of first ischemic stroke. Keywords Cerebrovascular death, multi-state Markov model, stroke recurrence, surrogate endpoint