PARP抑制劑在實體腫瘤中的療效和安全性：系統性回顧、統合分析和臨床試驗計畫書

Abstract

前言： 癌症目前為全球第二大死因，實體腫瘤佔癌症相關死亡的絕大多數。傳統治療多侷限於手術、放射治療與非選擇性化學治療，各自存在明顯限制。PARP抑制劑作為一類具前景的標靶藥物，透過「合成致死」機轉，在具有DNA修復缺陷的腫瘤細胞中誘導凋亡。本論文由兩個環環相扣的研究組成，分別評估PARP抑制劑在實體腫瘤中的跨腫瘤療效與血液學安全性，以及在子宮內膜癌中的特異性表現，最終發展出前瞻性臨床試驗計畫書（HOPE-EC）。 方法： 依循PRISMA 2020準則，於PubMed、Web of Science、Embase及Cochrane CENTRAL進行系統性文獻搜尋。Study 1以隨機效應逆變異數模型，統合十一種預先設定實體腫瘤的無惡化存活期（PFS）與總存活期（OS）風險比（HR），並以隨機效應Mantel-Haenszel模型統合各PARP抑制劑治療組合類別的血液學不良事件風險比（RR）。Study 2則針對四項子宮內膜癌隨機試驗，依合併治療類別與DNA修復基因變異狀態進行深入整合分析，並萃取PARP抑制劑單一藥物相較於化療在DNA修復基因變異次族群中的效果量。 結果： Study 1（跨腫瘤療效與血液學毒性分析）： 共納入十一種預先設定實體腫瘤、48項隨機對照試驗，結果呈現三層次療效梯度：攝護腺癌（PFS HR 0.59；OS HR 0.83）與胃癌（PFS HR 0.82；OS HR 0.67）在PFS與OS皆展現一致的顯著效益；卵巢癌（PFS HR 0.48；OS HR 0.90）、胰臟癌（PFS HR 0.53；OS HR 0.91）與黑色素瘤（PFS HR 0.78，p=0.035；OS HR 0.89）展現穩定的PFS效益，但OS效益未獲確認；子宮內膜癌（PFS HR 0.76，p=0.074；OS HR 0.79，p=0.185）、肺癌、神經膠質瘤、泌尿上皮癌、乳癌（PFS HR 0.81；OS HR 0.95）與大腸直腸癌（PFS HR 1.46；OS HR 1.26，兩者點估計皆偏向對照組）則在兩項指標上皆未達統計顯著，惟子宮內膜癌在四項獨立設計試驗中呈現一致方向性的PFS訊號。安全性方面，於21項具可萃取血液學資料的試驗中，PARP抑制劑單一藥物治療相較於安慰劑或非鉑類化療，皆呈現顯著較高的第三級以上血小板低下風險（RR 6.98–10.34）與貧血風險（RR 7.85–15.48）；但相較於非鉑類化療，其嗜中性球低下風險反而顯著較低（RR 0.59），顯示血小板低下與貧血、而非嗜中性球低下與白血球低下，才是可直接歸因於PARP抑制劑本身的毒性。 Study 2（子宮內膜癌整合分析）： 於四項隨機試驗中，療效依合併治療策略而分歧：單一藥物PARP抑制劑維持治療內部結果不一致（合併PFS HR 0.88，I²=76.1%）；PARP抑制劑併用抗血管新生藥物呈現有利但不精確的訊號（HR 0.70）；PARP抑制劑併用免疫檢查點抑制劑與化療則展現合併治療中最一致的效益（合併PFS HR 0.65）。以生物標記（同源重組修復缺陷／基因體不穩定）定義之次族群，在各試驗間呈現最具重現性的訊號（合併HR 0.50，I²=14.3%）。由於現有子宮內膜癌試驗中並無單純PARP抑制劑單一藥物相較於化療、且限定於此生物標記族群的乾淨對照，本研究改以Study 1的跨腫瘤整合分析（3.1.4節）萃取此比較，納入六個比較，得出PARP抑制劑單一藥物相較於化療之合併PFS風險比為0.70（95%信賴區間0.57–0.87；I²=62.7%；p=0.0015），並以此效果量作為HOPE-EC計畫書樣本數計算之依據。 結論： 本論文為PARP抑制劑於各實體腫瘤中建立明確的療效層級，並闡明其血液學毒性具藥物與治療組合特異性，而非整體藥物類別的一致現象；同時指出子宮內膜癌現存的關鍵證據缺口：目前尚無前瞻性試驗，針對DNA修復基因變異之病人族群，於首線鉑類化療失敗後測試PARP抑制劑單一藥物作為二線治療之角色。 未來展望與臨床試驗規劃： 承接此證據缺口，本論文進一步發展出一項第三期、隨機分派、開放標籤、主動對照之臨床試驗計畫書——HOPE-EC。該試驗收案對象為曾接受首線cisplatin為基礎之化療併用免疫檢查點抑制劑（ICI）治療後失敗、復發或轉移之子宮內膜癌病人，依DNA修復基因變異狀態篩選，評估rucaparib單一藥物治療，對照研究者選擇之化療。此計畫書納入生物標記分層設計、以BICR評估之PFS作為主要療效指標並以OS為關鍵次要指標，樣本數計算依Study 1（3.1.4節）所統合出的PARP抑制劑單一藥物相較於化療效果量而定，並依Study 1毒性訊號制定血液學監測架構。

Introduction: Cancer remains the second leading cause of death worldwide, with solid tumors accounting for the majority of cancer-related mortality. Treatment has traditionally been limited to surgery, radiotherapy, and non-selective chemotherapy, each with substantial limitations. PARP inhibitors (PARPis) represent a promising class of targeted agents that exploit synthetic lethality in tumors with DNA damage repair deficiency, inducing apoptosis in susceptible tumor cells. This dissertation comprises two linked studies evaluating the cross-tumor efficacy and hematologic safety of PARPis, and their endometrial-cancer-specific performance, culminating in the design of a prospective clinical trial protocol (HOPE-EC). Method: A systematic search of PubMed, Web of Science, Embase, and Cochrane CENTRAL was conducted following PRISMA 2020 guidelines. Study 1 pooled progression-free survival (PFS) and overall survival (OS) hazard ratios (HRs) across all eleven pre-specified solid tumor types using random-effects inverse-variance models, and synthesized hematologic adverse event risk ratios (RRs) across PARPi regimen categories using random-effects Mantel-Haenszel models. Study 2 performed an in-depth synthesis of four randomized trials in endometrial cancer, stratified by combination class and DNA-repair-gene mutation status, isolating the effect size of PARPi monotherapy versus chemotherapy in the DNA-repair-mutant subgroup. Result: Study 1 (Cross-tumor efficacy and hematologic toxicity): 48 randomized trials across all eleven pre-specified solid tumor types were included. A three-tier efficacy gradient emerged: prostate (PFS HR 0.59; OS HR 0.83) and gastric cancer (PFS HR 0.82; OS HR 0.67) showed concordant benefit on both endpoints; ovarian (PFS HR 0.48; OS HR 0.90), pancreatic cancer (PFS HR 0.53; OS HR 0.91), and melanoma (PFS HR 0.78, p=0.035; OS HR 0.89) showed a robust PFS benefit without a confirmed OS advantage; and endometrial (PFS HR 0.76, p=0.074; OS HR 0.79, p=0.185), lung, glioma, urothelial, breast (PFS HR 0.81; OS HR 0.95, neither significant once the full current trial set is pooled), and colorectal cancer (PFS HR 1.46; OS HR 1.26, both numerically favoring control) showed no statistically significant benefit on either endpoint, although endometrial cancer displayed a consistent directional PFS signal across four independently designed trials. On the safety side, across 21 trials with extractable hematologic data, PARPi monotherapy carried a significantly higher relative risk of grade ≥3 thrombocytopenia (RR 6.98-10.34 versus non-platinum chemotherapy and placebo, respectively) and anemia (RR 7.85-15.48 versus the same comparators) than either comparator, but a significantly lower relative risk of grade ≥3 neutropenia than non-platinum chemotherapy (RR 0.59), indicating that thrombocytopenia and anemia, rather than neutropenia and leukopenia, are the toxicities most directly attributable to PARP inhibition itself. Study 2 (Endometrial cancer synthesis): Across four randomized trials, efficacy diverged by combination strategy: single-agent PARPi maintenance was internally discordant (pooled PFS HR 0.88, I²=76.1%); PARPi plus anti-angiogenic therapy showed a favorable but imprecise signal (HR 0.70); and PARPi plus immune-checkpoint-inhibitor plus chemotherapy showed the most consistent benefit (pooled PFS HR 0.65) among combination regimens. Due to lack of evidences of a clean PARPi-monotherapy-versus-chemotherapy contrast within this biomarker-defined in endometrial cancer population, the cross-tumor synthesis in Study 1 (Section 3.1.4) was used instead to isolate this specific comparison, yielding a pooled PFS hazard ratio of 0.70 (95% CI 0.57-0.87; I²=62.7%; p=0.0015) across six treatment comparisons that provided the effect-size estimate used for the HOPE-EC sample-size calculation. Conclusion: This dissertation establishes a clear efficacy hierarchy for PARPis across solid tumors, delineates a hematologic toxicity profile that is agent- and regimen-specific rather than uniform across the drug class, and identifies a critical evidence gap in endometrial cancer: no prospective trial has yet tested PARP inhibitor monotherapy, in a DNA-repair-mutant-selected population, as second-line therapy after failure of first-line platinum-based chemotherapy.