新診斷急性骨髓性白血病患者早期死亡預測因子分析

Abstract

急性骨髓性白血病為造血幹細胞異常增生分化及基因變異累積所導致之血液癌症。隨著醫學研究的演進及抗癌藥物和骨髓移植之發展, 急性骨髓性白血病已由六十年前的不治之症變成現今可治癒之癌症。然而,仍有部分病人是在初期診斷治療時,就面臨死亡。過去已有諸多文獻分析過早期死亡率相關危險因子,包括:年長者、複雜的染色體核型變化、較差之日常體能狀態、較差之腎功能、共病、感染症、診斷時較高之周邊白血球、及較高之骨髓芽細胞數目。然而,其分析大多屬於歐美國家的數據,且未納入基因突變及基因表達之分析。因此,仍有相當大的研究空間。 本研究納入1,056名新診斷非M3型急性骨髓性白血病成人病患，所有病患皆接受高強度誘導化學治療。早期死亡定義為急性骨髓性白血病初次診斷後八週內之死亡。我們使用Cox比例風險模型建立治療前預後風險評分;透過核糖核酸定序資料進行基因表達及基因集合富集分析;採用CIBERSORTx分析免疫細胞族群分布。 本研究中位追蹤時間為78.3個月，八週早期死亡率為7.8% (人數=82)。多變項回歸分析發現，2017年後診斷的病患,其早期死亡風險顯著降低。我們建立了兩個治療前早期死亡預後風險評分模型：第一個模型針對全部病患，其預測因子包含年齡、診斷年份、白蛋白濃度及DNMT3A突變；第二個模型針對2017年後診斷的病患，其預測因子包含白蛋白濃度、TP53、IDH2及NRAS突變。兩模型把病人分成三組(高風險、中風險、低風險) ，Harrell's C-指數分別為0.72及0.81。基因集合富集分析及免疫細胞族群分析結果顯示，早期死亡組患者的MYC及E2F路徑增強，免疫相關路徑活性降低，且CD8 T細胞數量減少。 根據研究建立的早期死亡預後風險評分模型，能有效預測患者接受高強度誘導化學治療後的早期死亡風險。另外，研究發現早期死亡組患者呈現MYC及E2F路徑增強，且骨髓微環境具有免疫抑制特性。結合這些風險評估工具與生物學標記，可協助臨床醫師為新診斷急性骨髓性白血病患者制定個人化治療策略。

Background: Despite significant advances in acute myeloid leukemia (AML) treatment, early mortality remains a clinical challenge, with approximately 3–17% and 11–24% of patients dying within four and eight weeks of diagnosis, respectively. This study aims to identify molecular markers and pathways associated with early mortality and to develop pretreatment prognostic risk scores for early mortality prediction. Methods: We included 1,056 adults with newly diagnosed non-M3 AML who received intensive induction chemotherapy. Patients were recruited from National Taiwan University Hospital between 1994 and 2023. Early mortality was defined as death occurring within eight weeks of initial AML diagnosis. Cox proportional hazard models were employed to establish the pretreatment prognostic risk scores. Gene expression and gene set enrichment analyses (GSEA) were performed using bulk RNA sequencing data. CIBERSORTx was utilized to analyze immune cell populations. Results: The median age of the entire cohort was 51 years, and 54.5% were male. The median follow-up time was 78.3 months (95% CI: 31.7-142.9). The 8-week early mortality rate was 7.8% (n = 82). Multivariable Cox regression showed significantly lower early mortality risk for patients diagnosed after 2017 (HR: 0.51, P=0.029). We established two pretreatment prognostic risk scores for early mortality: one for the entire cohort and another for patients diagnosed after 2017 (Post-2017 subgroup). The entire cohort risk score included age, diagnostic year, albumin level, and DNMT3A mutation; the Post-2017 subgroup included albumin level, TP53, IDH2, and NRAS mutations. Harrell's C-indices for the prediction models were 0.72 (95% CI 0.66-0.78) and 0.81 (95% CI: 0.71–0.92), respectively. GSEA and immune cell population evaluation revealed positive enrichment in MYC and E2F pathways, negative enrichment in immune-related pathways, and decreased numbers of CD8 T cells in the early mortality patients. Conclusions: This study established pretreatment prognostic risk scores combining clinical and genetic parameters for early mortality prediction in patients undergoing intensive induction chemotherapy. The identified biological features in early mortality patients, including MYC pathway overexpression and immunosuppressive microenvironment, may enable more precise risk stratification and personalized therapy.