利用醫院電子病歷建立心臟衰竭病人使用sacubitril/valsartan死亡或心臟移植之風險評估工具

Utilizing Electronic Health Records for Risk Assessment of Death or Heart Transplantation in Patients with Heart Failure Treated with Sacubitril/Valsartan

鄭宇揚(Yu-Yang Cheng)

指導教授 ： 徐莞曾

國立臺灣大學/醫學院/臨床藥學研究所/碩士(2022年)

https://doi.org/10.6342/NTU202204068

摘要

研究背景
臨床試驗顯示sacubitril/valsartan對心臟衰竭合併左心室射出分率低下 (heart failure with reduced ejection fraction, HFrEF) 的病人有降低死亡率與逆轉心臟重塑的臨床效益，但臨床使用sacubitril/valsartan時可能面臨病人腎功能下降或心臟超音波 (ultrasound cardiography, UCG) 量測之重塑指標改善有限等情境。目前較無實證探討前述問題對於預後的影響，因此本研究希望納入腎功能下降與心臟重塑變化等相關變項，建立風險評估模型，協助臨床醫療人員進行風險估計與決策。
研究方法
本回溯性世代追蹤研究以臺大醫院HFrEF病人為研究對象，透過臺大醫院電子病歷系統收集其可能與預後相關的臨床變項，主要研究終點為全因性死亡與接受心臟移植之複合事件，亦評估啟用sacubitril/valsartan後短期內心臟重塑變化與發生心血管相關住院的相關性。
研究第一部份透過集群分析 (cluster analysis) 的方式，根據HFrEF病人啟用sacubitril/valsartan後6個月的UCG量測數值進行分群 (clustering)，加權後以Kaplan-Meier方法估計不同群間的風險比值 (hazard ratio)。
研究第二部份建立多變項time-dependent Cox’s regression model (Cox’s model) 探討多個變項與預後的相關性。透過逐步選取法 (stepwise selection) 進行變項篩選以獲得最終模型，同時於過程當中搭配penalized spline評估連續變項與預後是否符合線性關係，並產生用以分層的臨界值。
研究結果
根據收案條件初步篩選291位HFrEF病人，追蹤期間中位數為33.6個月。集群分析顯示病人可分為「Cluster 01」與「Cluster 02」兩個次族群，當中Cluster 01的病人心臟衰竭病程較長且啟用sacubitril/valsartan前最近一次N端B型利鈉利尿胜肽原 (N-terminal pro B-type natriuretic peptide) 的中位數較高。加權後分析顯示Cluster 01族群後續有較高的心血管相關住院發生率，但是兩個次族群間並未觀察到主要複合事件的顯著差異。
多變項Cox’s model顯示與預後負相關之變項包含末期腎臟病透析、使用非類固醇類消炎止痛藥之用藥史、事件發生前9個月內血液尿素氮 (blood urea nitrogen, BUN) 上升幅度，以及事件發生前3個月內腎絲球過濾率 (estimated glomerular filtration rate, eGFR) 降低幅度等。與預後正相關之變項包含追蹤期間平均左心室射出分率、事件發生前1個月之sacubitril/valsartan日劑量與追蹤期間的血紅素 (hemoglobin, Hb) 數值等。本研究建構之預後模型亦呈現於互動式網頁當中，臨床醫療人員可以根據病人的臨床狀況估算預後做為參考。
結論
本研究利用Cox’s model呈現心臟重塑指標、血紅素、sacubitril/valsartan每日劑量、以eGFR與BUN為基礎的腎臟功能變化指標與預後的相關性。結果也強調HFrEF病人使用sacubitril/valsartan時，除了定期監測病人UCG，應同時評估Hb、eGFR與BUN等腎臟功能相關參數。希望透過密切關注病人的臨床變化並據此權衡病人潛在風險與效益，能幫助病人接受適當的sacubitril/valsartan耐受劑量。

關鍵字

心臟衰竭 ； 左心室射出分率 ； 血液尿素氮 ； 腎絲球過濾率 ； 血紅素 ； sacubitril/valsartan

並列摘要

Background
Sacubitril/valsartan is considered an effective therapy for patients with heart failure with reduced ejection fraction (HFrEF). Several clinical trials have indicated that sacubitril/valsartan can help reduce mortality and reverse cardiac remodeling. However, renal function decline and blunt changes in systolic cardiac function are the two most common problems associated with the real-world use of sacubitril/valsartan. Because information on how these unfavorable responses affect patient prognosis is limited, identifying individuals at an increased risk of such situations and recommending more specific dosage adjustments based on patients’ responses are complex. Therefore, in this study, we aim to develop a risk assessment model to aid clinicians with risk estimation and decision-making.
Methods
In this retrospective cohort study, we analyzed data from patients with HFrEF undergoing sacubitril/valsartan therapy at National Taiwan University Hospital. Potential clinical covariates reported in the patients’ electronic health records were considered. The primary endpoint of the study was defined as a composite of all-cause mortality and heart transplantation. In addition, the association between short-term changes in ultrasound cardiography (UCG) parameters after the initiation of sacubitril/valsartan and cardiovascular-related hospitalization (CVH) was assessed.
In the first part of the study, the echocardiographic phenotypes were identified using k-means clustering based on UCG parameters after 6 months. The hazard ratios adjusted using the inverse probability of the weight of both the primary endpoint and CVH were then estimated using the Kaplan–Meier method.
In the second part, multivariate analysis was conducted by fitting Cox’s proportional hazards models with time-dependent covariates, called the Cox’s model, to estimate the adjusted effects of clinical covariates. The stepwise variable selection procedure was applied to obtain the final model. The penalized spline smoothing method was used to determine the linearity of and yield optimal cut-off values for the continuous covariates during the variable selection procedure.
Results
A total of 291 patients with HFrEF with a median follow-up of 33.6 months met the eligibility criteria and were included in the study. Two echocardiographic phenotypes were identified (“Cluster 01” and “Cluster 02”) using the k-means clustering method. Compared with the patients in Cluster 02, the patients included in Cluster 01had a longer heart failure duration and a higher median N-terminal pro-B-type natriuretic peptide level at baseline. After weighting was performed for other unbalanced covariates except for the UCG parameters, the Cluster 01 group was discovered to have an increased rate of CVH. However, no considerable differences were observed in the primary composite events.
In the multivariable Cox’s model, dialysis for end-stage renal disease, non-steroid anti-inflammatory drug usage at baseline, an increased blood urea nitrogen (BUN) level within 9 months, and a decreased estimated glomerular filtration rate (eGFR) within 3 months were found to be significantly associated with an increased risk of composite events. In addition, the average left ventricular ejection fraction, hemoglobin level, and the daily dose of sacubitril/valsartan were found to have prognostic benefits. An interactive website was developed to allow clinicians to access and predict prognoses depending on the clinical condition of patients.
Conclusion
The significant findings of the Cox’s model revealed target values for measuring UCG parameters, dose-dependent responses of sacubitril/valsartan, and, most importantly, eGFR-based and BUN-based clinical criteria for evaluating renal function changes in patients with HFrEF. The results emphasized the importance of routinely carefully examining the entire renal function panel rather than the creatinine levels alone and of assessing cardiac responses during the sacubitril/valsartan treatment period. By closely monitoring the clinical status of patients and identifying potential risks accordingly, clinicians can determine optimal sacubitril/valsartan doses for patients with HFrEF in clinical practice.

並列關鍵字

heart failure ； left ventricular ejection fraction ； blood urea nitrogen ； estimated glomerular filtration rate ； hemoglobin ； sacubitril/valsartan

參考文獻

1. Hasenfuss G, Mann DL. Chapter 47: Pathophysiology of Heart Failure. In: Libby P, Bonow RO, et al, eds. Braunwald's Heart Disease: a Textbook of Cardiovascular Medicine. 12 ed. Elsevier/Saunders; 2022:913:chap 47.

2. Januzzi JL, Mann DL. Chapter 48: Approach to the Patient with Heart Failure. In: Libby P, Bonow RO, et al, eds. Braunwald's Heart Disease: a Textbook of Cardiovascular Medicine. 12 ed. Elsevier/Saunders; 2022:933:chap 48.

3. Virani SS, Alonso A, et al. Heart Disease and Stroke Statistics - 2021 Update: A Report from the American Heart Association. Circulation. 2021;143(8):e254.

4. Vos T, Lim SS, et al. Global Burden of 369 Diseases and Injuries in 204 Countries and Territories, 1990–2019: A Systematic Analysis for the Global Burden of Disease Study 2019. The Lancet. 2020;396(10258):1204.

5. Farre N, Vela E, et al. Real World Heart Failure Epidemiology and Outcome: A Population-Based Analysis of 88,195 Patients. PLoS One. 2017;12(2):e0172745.