Impact of frailty on outcomes in atrial fibrillation and HFpEF: data from the TOPCAT trial

Abstract

Background

Atrial fibrillation (AF) and heart failure with preserved ejection fraction (HFpEF) often coexist, worsening each other’s progression and contributing to poor outcomes. Frailty, a syndrome characterized by vulnerability to stressors, is highly prevalent in AF and HFpEF patients and has been associated with adverse outcomes such as stroke, hospitalization, and mortality. However, the specific prognostic implications of varying degrees of frailty in AF comorbid HFpEF patients remain unclear. This study investigates the relationship between frailty, measured using a deficit accumulation frailty index (FI), and adverse outcomes, particularly stroke, in this patient population.

Methods

This post hoc analysis of the TOPCAT randomized control trial included 721 patients with AF and HFpEF from the Americas classified into three groups based on FI: Group 1 (FI < 0.3), Group 2 (FI 0.3–0.4), and Group 3 (FI ≥ 0.4). The primary outcome was stroke, and secondary outcomes included heart failure hospitalization, cardiovascular death, and all-cause mortality. Cox proportional hazards models and Kaplan–Meier analyses were used to assess the association between frailty status and outcomes. A dose–response relationship was evaluated using restricted cubic splines.

Results

97.8% AF comorbid HFpEF patients were diagnosed frailty in this cohort. During a mean follow-up of 36 ± 19 months, stroke incidence was significantly higher in Groups 2 and 3 compared with Group 1 (adjusted HR for Group 2: 5.01 [95% CI, 2.00–12.53]; P = 0.001 and Group 3: 6.35 [95% CI, 2.26–17.86]; P < 0.001). A linear dose–response relationship between FI and stroke risk was observed. Higher frailty was also associated with increased cardiovascular and all-cause mortality but not significantly with heart failure hospitalization.

Background

Atrial fibrillation (AF) is the most common sustained cardiac tachyarrhythmia, affecting nearly 59.7million people worldwide [1]. The prevalence of AF is even higher among patients with heart failure, particularly among those heart failure with preserved ejection fraction (HFpEF), ranging from one-third to one-half of all patients [2–4]. Also, AF is commonly comorbid with AF, occurring in 21% of AF patients [5]. The pathological changes that occurred in patients with AF or HFpEF were complicated. Each of these two diseases could cause cardiac structural and functional changes, which always worsen each other in a vicious cycle. Patients with heart failure (HF) and AF have worse outcomes than patients with HF without AF or with AF without HF [5]. As a result, comorbidity of HFpEF and AF is associated with an increased risk of poor prognosis [6, 7].

Frailty, defined as a complex clinical syndrome characterized by vulnerability to stressors, is associated with increased susceptibility to adverse outcomes such as death, major cardiovascular events, hospitalization, falls, or fractures. Frail patients tend to have more comorbidities, require more therapeutic agents, and are more likely to experience adverse events [8]. Therefore, screening and diagnosing frailty is a significant and challenging task in clinical care. The most utilized definitions of frailty include the Frailty Phenotype and Frailty Index (FI). The later approach, which isbased on the cumulative deficit model, could content more information of the patients clinical status and have a wider spread of use in large retrospective dataset [9]. Previous studies have shown that frailty is related to poor outcomes such as stroke, heart failure, cardiovascular death Etc. in AF patients [10]. However, over 90% patients suffering frailty in AF comorbid HFpEF patients, and the prognosis of these frail patients are in the blind spot of previous study. To illustrate, we are not able to predict the adverse outcome or guiding clinical treatment based on “more frail” or “less frail” in all these frail patients. Current research have not prove the dosage-dependent relationship between frailty index and prognosis among those frailty patients in AF and HFpEF patients. Our study aims to investigate the relationship between frailty and prognosis in AF comorbid HFpEF patients.

Method

Population and group

The data for this study were obtained from the Treatment of Preserved Cardiac Function Heart Failure with an Aldosterone Antagonist (TOPCAT) trial (TOPCAT ClinicalTrials.gov number, NCT00094302.) the design of which has been previously published. Briefly, the TOPCAT study was a multicenter, double-blind, placebo-controlled trial that recruited 3,445 patients from 233 centers across the Americas, Russia, and Georgia. The inclusion criteria were as follows: 1. Symptomatic HF patients aged ≥ 50 years old. 2. LVEF ≥ 45%, combined with either hospitalization for HF within 12 months prior to inclusion or elevated natriuretic peptide levels (brain natriuretic peptide [BNP] ≥ 100 pg/mL or N-terminal pro-BNP [NT-proBNP] ≥ 360 pg/mL) within 60 days prior to inclusion. 3. Patients should have controlled systolic blood pressure [BP] < 140 mmHg (or ≤ 160 mmHg if the patient was taking three or more medications to control blood pressure) and a serum potassium level < 5.0 mmol/L. The main exclusion criteria were a life expectancy < 3 years and an estimated glomerular filtration rate (eGFR) < 30 ml/min/1.73 m2 or serum creatinine ≥ 2.5 mg/dL. Our study was a post-hoc analysis of the TOPCAT trial. Consistent with previous approaches, we excluded patients enrolled in Russia and Georgia due to significant differences in patient characteristics, outcomes, and adherence to study drug [11–13]. Based on all the inclusion and exclusion criteria of the TOPCAT trial, only participants with > 30 s recorded AF were included in our study.

Frailty index

Drawing on the insights gleaned from Sanders' study, a standard deficit accumulation approach was modified to adapt to the characteristics of our research population, proving robust in assessing the frail status of HFpEF patients [14]. Our approach entailed constructing a frailty index deficits model comprising 35 items, including 15 based on signs and symptoms from the Kansas City Cardiomyopathy Questionnaire, alongside laboratory abnormalities and self-reported comorbidities (Supplementary table S1). The frailty index is computed by summing the total deficits present and dividing by the number of deficits measured, with a frailty index exceeding 0.21 indicative of frailty [15].

Group

According to the baseline FI, the enrolled participants were divided into three groups: Group 1, FI < 0.3; Group 2, FI 0.3–0.4; Group 3, FI ≥ 0.4

Outcome

The primary outcome of the study was stroke. While the secondary outcomes included heart failure hospitalization, cardiovascular death, and all-cause mortality.

Statistical analysis

Continuous variables were expressed as means with standard deviations, while categorical variables were presented as counts and percentages. Differences between groups of continuous variables were analyzed using variance analysis or nonparametric tests, and the Fisher exact test was employed for categorical variables. The Kaplan–Meier estimator calculated the time to event-rate of primary and secondary endpoints, with log-rank tests comparing endpoint incidences. Three Cox proportional hazards models were utilized to estimate hazard ratios (HR) and quantify the relationship between baseline frailty status and outcomes. Model 1 remained unadjusted, while model 2 was adjusted for age, sex, race, and body mass index. Model 3 was further adjusted for age, race, body mass index, baseline systolic blood pressure (SBP), previous coronary artery disease (CAD), hypertension, estimated glomerular filtration rate (eGFR), and concomitant use of beta blockers(β blocker), calcium channel blocker (CCB), diuretic, and antiplatelet drugs. A dose–response relationship between frailty index and primary endpoint was assessed using a generalized additive model and fitting smooth curve (restricted cubic splines). All analyses were performed using data from the Americas (United States, Canada, Brazil, and Argentina) of the TOPCAT trial, with significance set at P < 0.05. STATA 17 (Stata Corp. College Station, TX, USA) and R (Version 4.3.2) were utilized for all analyses.

Results

Baseline

A total of 721 participants (35.1% women; average age, 67.9 ± 9.3 years) were enrolled in our analysis with a mean follow-up of 36 ± 19 months. The patients in group 1 had a mean age of 74.8 years, with 47.5% being women, and 92.4% identifying as white. Group 2 patients had a mean age of 74.8 years as well, with 49.8% being women, and 83.9% identifying as white. Patients in group 3 had a mean age of 72.8 years, with 43.7% being women, and 84.2% identifying as white. Briefly, patients in the higher frailty index group had a higher BMI and New York Heart Association Classification (NYHA class), were more likely to comorbid CAD and hypertension, had higher SBP and lower eGFR levels compared with lower FI groups. The comorbidity rates of diabetes and heart failure were similar between these groups.

The baseline utilization rate of diuretics, β blockers, CCBs, and antiplatelets was higher in the higher frailty index group. The baseline characteristics are shown in Table 1.

Table 1.

Baseline characteristic of the study cohort

	Group1 (score < 0.3)	Group2 (score: 0.3–0.4)	Group3 (score > 0.4)	P value
Age (years, mean ± SD)	74.8 ± 8.5	74.8 ± 8.6	72.8 ± 8.6	0.013
Gender				0.366
Female (%)	47.5	49.8	43.7
White Race(%)	92.4	83.9	84.2	0.014
BMI (Kg/m2, mean ± SD)	30.1 ± 5.6	32.7 ± 7.4	35.6 ± 8.3	< 0.001
NYHA class (%)				< 0.001
Ⅰ	79.8	66.7	51.8
Ⅱ	20.2	33.3	48.2
CHA2DS2-VASc score (mean ± SD)	4.0 ± 1.1	4.5 ± 1.1	5.0 ± 1.0	< 0.001
HAS-BLED score (mean ± SD)	2.0 ± 0.8	2.2 ± 0.7	2.4 ± 0.7	0.868
EF (mm, mean ± SD)	57.9 ± 7.3	57.3 ± 7.2	57.3 ± 7.6	0.613
Diabetes (%)	7.1	29.2	66.0	0.093
Hypertension (%)	77.6	89.4	95.6	< 0.001
CAD (%)	19.7	34.0	67.6	< 0.001
Heart Failure (%)	6.8	34.0	67.6	0.068
BNP	347 ± 300	463 ± 565	434 ± 458	0.267
NT-proBNP	1589 ± 1312	2275 ± 2636	1558 ± 1265	0.086
Systolic BP (mmHg, mean ± SD)	120 ± 12	123 ± 14	129 ± 16	< 0.001
Diastolic BP (mmHg, mean ± SD)	72 ± 10	70 ± 11	70 ± 11	0.093
Heart Rate(b.p.m, mean ± SD)	69 ± 10	70 ± 11	68 ± 11	0.322
eGFR (ml/min/1.73m2, mean ± SD)	68.1 ± 19.1	62.6 ± 18.0	57.1 ± 18.2	< 0.001
Antiarrhythmic drugs	18.6	22.0	23.1	0.524
Anticoagulant (%)	86.3	81.5	81.8	0.337
Antiplatelet	43.7	55.0	68.0	< 0.001
Diuretic	92.4	95.5	99.2	0.001
ACEI	50.8	52.2	60.7	0.066
ARB	25.1	32.7	33.6	0.127
β blocker	89.6	82.1	89.5	< 0.019
CCB	31.2	46.7	62.4	< 0.001
digoxin	32.2	30.6	30.8	0.926
spironolactone	52.5	54.0	55.5	0.824

BMI indicates body mass index, NYHA class, New York Heart Association Classification, EF ejection fraction, CAD coronary artery diseases, BNP brain natriuretic peptide, NT-proBNP N-terminal pro-brain natriuretic peptide, BP blood pressure, eGFR estimated glomerular filtration rate, ACEI Angiotensin-converting enzyme inhibitors, ARB angiotensin 2 receptor blockers, β blocker Beta blockers, CCB Calcium channel blockers

Primary outcome

70 (9.71%) primary outcome events occurred during the follow-up period. The Kaplan–Meier curve of the primary outcome is shown in Fig. 1. The cumulative incidence of the composite primary outcome was significantly higher in group 2 and group 3 than group 1 (group 2: unadjusted HR = 4.00; 95% confidence interval [CI]: 1.68–9.51; p = 0.002, group 3: unadjusted HR = 4.05; 95% CI: 1.68–9.77; p = 0.002).

Fig. 1.

Kaplan–Meier estimates of primary endpoint analysis

We constructed 3 Cox models to evaluate the relationship between the frailty status (group 1 as reference) and the primary outcome. There was a statistical difference in the risk of stroke between group 2 and group 3 versus group 1 (HR, 4.00 [95% CI, 1.68–9.51], P = 0.002; and HR, 4.05 [95% CI, 1.68–9.77], P = 0.002, respectively). Even after fully adjusting potential confounders in model 3, this difference was still consistent (HR, 5.01 [95% CI, 2.00–12.53]; P = 0.001; and HR, 6.35 [95% CI, 2.26–17.86]; P < 0.001, respectively). We also used the frailty index as a continuous variable to assess the relationship between the frailty index and stroke. The fitting smoothing curve in Fig. 2 showed a linear increase in HR.

Fig. 2.

Restricted cubic curve of the dose–response relationship between frailty status and outcomes.Deepred line represents references for HRs, and light red line represent 95% CI. Adjusted for all covariates in model 3. HR indicates hazard ratio

Secondary outcomes

There was a statistical difference in the risk of cardiovascular mortality and all-cause mortality between group 2 and group 3 compared with group 1 in univariable and multivariable analysis, as shown in Table 2. The difference was not statistically significant between group 2 and group 1 in HF hospitalization.

Table 2.

Primary and Secondary outcomes

	Model 1					Model 2				Model 3
	Group2 (score: 0.3–0.4) HR (95% CI)	p	Group3 (score > 0.4) HR (95% CI)	p	Group2 (score: 0.3–0.4) HR (95% CI)	p	Group3 (score > 0.4) HR (95% CI)	p	Group2 (score: 0.3–0.4) HR (95% CI)	p	Group3 (score > 0.4) HR (95% CI)	p
stroke	4.00 (1.68, 9.51)	0.002	4.05 (1.68, 9.77)	0.002	4.14 (1.73, 9.91)	0.001	4.27 (1.74, 10.45)	0.001	5.01 (2.00, 12.53)	0.001	6.35 (2.26, 17.86)	< 0.001
HF hospitalization	1.03 (0.68, 1.57)	0.883	2.08 (1.40, 3.08)	< 0.001	0.99 (0.64, 1.50)	0.919	1.99 (1.32, 2.99)	0.001	0.96 (0.62, 1.50)	0.871	1.79 (1.10, 2.93)	0.019
All-cause mortality	1.75 (1.15, 2.65)	0.009	1.72 (1.12, 2.65)	0.013	1.80 (1.18, 2.73)	0.006	1.88 (1.21, 2.92)	0.005	2.26 (1.45, 3.51)	< 0.001	3.13 (1.86, 5.29)	< 0.001
Cardiovascular death	1.54 (0.91, 2.60)	0.110	1.64 (0.96, 2.82)	0.072	1.66 (0.97, 2.82)	0.063	1.90 (1.09, 3.30)	0.023	2.01 (1.15, 3.51)	0.014	2.84 (1.47, 5.49)	0.002

Model 1 remained unadjusted, while model 2 was adjusted for age, sex, race, and body mass index. Model 3 was further adjusted for age, race, body mass index, baseline systolic blood pressure (SBP), previous coronary artery disease (CVDs), hypertension, estimated glomerular filtration rate (eGFR), and concomitant use of β blocker, calcium channel blocker (CCB), diuretic, and antiplatelet drugs

Discussion

The result of our study showed that patients with AF and HFpEF had a high median FI score. 97.8% AF comorbid HFpEF patients were diagnosed frailty using a deficit accumulation frailty index approach. In addition, the risk of stroke, HF hospitalization, all-cause mortality and Cardiovascular death was significantly higher in these two frailty group with higher frailty index score than the lowest FI groups among the AF with HFpEF patients. The prognosis of these two higher FI frailty groups was similar. Frailty evaluation might be helpful for stroke prevention, such as anticoagulation therapy in patients with AF and HFpEF. Our findings showed a clear dose–response relationship between FI and stroke risk, as well as increased cardiovascular and all-cause mortality.

Previous research has consistently highlighted the high risk of frailty among patients with atrial fibrillation (AF). For instance, data from the Framingham Heart Study Offspring cohort revealed that 44% of all AF patients were prefrail, with an additional 6% classified as frail [16]. Similarly, a study utilizing Belgian nationwide data found that 28.2% of AF patients diagnosed as frailty [17]. In Japan, among 12,585 AF patients, 31.5% were identified as mildly frail, 34.8% as moderately frail, and 25.9% as severely frail [18]. Moreover, a post hoc analysis of 20,867 participants in the ENGAGE AF-TIMI 48 trial demonstrated that 19.6% of the study population presented with frailty [19]. Subsequent systematic reviews and meta-analyses corroborated these findings, reporting a pooled prevalence of frailty ranging from 39.7% to 89.5% among AF patients across various studies [10]. Notably, among the total participants in meta-analyzed studies, which encompassed 9420 individuals, 2803 were diagnosed with AF, with 1517 classified as frail and 1286 as either pre-frail or robust [20]. In our study, the prevalence of frailty is even higher than these studies. The main reason may be considered as the vulnerability to stressors causing by HFpEF. Earlier research showed that the HFpEF patients are more likely to be frailty., The prevalence of frailty in our cohort (97.8%) was notably higher than in previous studies, possibly due to the added vulnerability conferred by HFpEF, but it still could not be rule out the possibility of the threshold of frailty index or cohort selection.

The prognosis of frailty AF patients is poorer than non-frailty AF patients. A systematic review showed that frailty was associated with increased stroke incidence, all-cause mortality, symptom severity, and length of hospital stay. This finding was in consensus with the results of our study [21].

Stroke is considered as the main adverse outcome of AF progression. Prevention of stroke has the priority in the treatment of atrial fibrillation. The guidelines of Europe and America recommended using CHA2DS2-VASc score for stroke risk assessment [22, 23]. However, during clinical practice, in some specific group of patients, the method of assessing the risk of stroke in atrial fibrillation still needs further research. Frailty has been linked to adverse outcomes in AF patients, and our results are consistent with previous studies that found frailty to be associated with higher stroke risk and mortality. Importantly, the high frailty prevalence observed in our cohort suggests that frailty could be an important factor in refining risk stratification in AF and HFpEF patients. Drawing useful lessons of DAPT and PRECISE-DAPT scores, future studies are needed to assess whether frailty could be integrated with existing risk scores, such as CHA2DS2-VASc, to improve clinical decision-making, particularly in frail patients at risk for both stroke and bleeding [24]. Previous studies reported the frailty AF patients had a higher risk of stroke. Among 5,070 AF patients treated with catheter ablation, 38.6% were deemed frail with a Hospital Frailty Risk Score > 5, including 8.3% at high risk, leading to statistically different mortality rates (up to 630 days) between each frailty groups [25]. The risk of bleeding increased with higher eFI levels, although this association was less pronounced for ischemic stroke/transient ischemic attack [18]. Among individuals with AF eligible for oral anticoagulation therapy (OAC), frailty was linked to higher risks of death, gastrointestinal bleeding, falls, and stroke in women [26]. Post hoc analysis of 20,867 participants in the ENGAGE AF-TIMI 48 trial revealed that, on average over the follow-up period, the risk of stroke or systemic embolism increased by 37% and major bleeding by 42% with each 0.1 increase in the frailty index [19]. Frail AF patients exhibited significantly elevated risks of all-cause death, ischemic stroke, and bleeding compared to robust individuals. Patients with known AF were enrolled in a prospective cohort study in Switzerland, of the 2369 included patients, pre-frailty and frailty were associated with a higher risk of unplanned hospitalization, all-cause mortality, and bleeding [27]. Frailty, but not pre-frailty was associated with a higher risk of stroke. A meta-analysis included 89 996 adults with AF and CHA2DS2-Vasc score of > _2 indicated that frailty was associated with higher risk of all-cause mortality and major bleeding [28]. A meta-analysis included 30,883 patients, in whom with AF, frailty was associated with increased stroke incidence, all-cause mortality, symptom severity and length of hospital stay [21]. These studies suggest that frailty is an effective method to complement CHA2DS2-Vasc score in assessing stroke risk in patients with atrial fibrillation. Owing to the extremely high prevalence of the frailty risk in AF comorbid HFpEF patients, dividing patients into frail and non-frail categories makes it difficult to effectively assess a patient's stroke risk. So, this study using an explicit frailty score method aiming to assess the FI-dependent relationship between frailty and stroke in AF comorbid HFpEF patients. The anticoagulation therapy might be modified in AF and HFpEF patients, especially in these patients with high FI score. According to the result of this study, we think that FI assessment could be used as a routine supplement to the CHA2DS2-VASc score to assess the risk of stroke in patients with AF comorbid HFpEF. Future studies are needed to assess whether frailty could be integrated with existing risk scores, such as CHA2DS2-VASc, to improve clinical decision-making, particularly in frail patients at risk for both stroke and bleeding. On the other hand, whether the AF and HFpEF patients with high FI score required an intensive anticoagulation therapy, such as left atrial appendage closure implantation, still need further research. High quality randomized controlled trials are required to focus on comparing the beneficial of different anticoagulation therapy in patients with AF and HFpEF. However, the stroke events in the BioLINCC TOPCAT dataset used here are not consistently coded for ischemic versus hemorrhagic subtype in a manner accessible to our analysis; therefore, we cannot determine whether the observed association between higher FI and increased stroke risk is driven primarily by ischemic (thromboembolic) events, hemorrhagic events, or both. Additionally, although baseline OAC prescription rates were similar across FI groups (≈81–86%), longitudinal adherence, dose changes, or time-varying anticoagulant exposure data necessary to robustly test FI and OAC interactions were not available for our analysis. Future studies with access to adjudicated stroke subtype and time-varying anticoagulation data are required to find out this issue.

In our cohort hypertension prevalence rose markedly across FI strata (Table 1), consistent with published literature showing high rates of hypertension among frail older adults. The recent ESH Working Group review highlights that frailty and functional status should be assessed when managing hypertension in older individuals because frailty may modify both the benefits and harms of antihypertensive therapy and because uncontrolled hypertension remains a leading contributor to cerebrovascular event [29]. We therefore emphasize that the high burden of hypertension in frail AF + HFpEF patients likely contributes to the observed association between higher FI and stroke and underscores the need for integrated management strategies addressing both vascular risk and frailty.

HF hospitalization was another concern of AF comorbid HFpEF patients. Why frailty was not significantly associated with HF hospitalization in our cohort. Prior evidence supports that frailty predicts rehospitalization in HF populations; however, our cohort differs in several important respects. First, the prevalence of frailty in our AF + HFpEF sample was extremely high (≈97.8%), which compresses between-group differences and reduces power to detect incremental associations with HF hospitalization. Second, HF hospitalization is highly dependent on short-term clinical instability (e.g., decompensation due to volume overload, arrhythmia, or infection), local hospitalization thresholds, and prior HF treatment intensity; an FI measured at baseline may better capture chronic vulnerability and long-term risk (e.g., mortality, stroke) than short-term HF admission triggers. Third, sample size and event counts for HF hospitalization within strata were modest, which may limit statistical power. These considerations may explain why, despite robust associations with stroke and mortality, FI was not independently associated with HF hospitalization in our adjusted models. Previous HFpEF studies report heterogeneity in frailty–rehospitalization associations that likely reflect differences in frailty measurement, cohort selection, and follow-up [30].

Limitation

There are some limitations: First, this study, which was a post hoc analysis of a randomized controlled trial study, had the limitations and biases inherent to an observational retrospective study. There may be some unknown confounders, which may influence the accuracy of the study results. Second, the frailty index was obtained only at baseline, while dynamic frailty evaluation might be more convincing. Third, small sample size is another limitation which may limit the statistical power for some subgroup analyses. We believe that larger multi-center clinical trials are needed to better solve this problem. Fourth, this study cohort only restricted to the Americas subgroup of TOPCAT which may limited broader applicability, especially to Eastern European populations where patient characteristics maybe different.

Conclusion

Frailty is a significant predictor of adverse outcomes in AF and HFpEF patients, wih higher frailty scores correlating with increased stroke risk and mortality. Frailty evaluation using a continuous FI score may refine risk stratification and guide therapeutic strategies, including anticoagulation therapy.

Authors’ contributions

L.Z. and C.-Y.L. contributed equally to this work. L.Z. and C.-Y.L. conceived the study, collected clinical data, and wrote the initial draft of the manuscript. Z.-J.Z. and L.-H.H. assisted with data analysis and interpretation. L.-Z.G. and S.-N.L. contributed to patient recruitment and procedural supervision. D.-Y.L. designed the study, provided critical revisions, and supervised the entire project. All authors reviewed and approved the final manuscript. C.-S.M and J.-Z.D supervised the entire project.

Funding

None.

Data availability

The data of the TOPCAT trial is available via reasonable request to the National Heart, Lung and Blood Institution, https://biolincc.nhlbi.nih.gov.

Declarations

Ethics approval and consent to participate

The TOPCAT trial was approved by the ethics committee and an informed consent form was signed by all participant. This paper was used TOPCAT clinical data obtained from the National Heart, Lung, and Blood Institute’s Biological Specimen and Data Repository Information Coordinating Center (BioLINCC, Calverton, Maryland). TOPCAT conformed to the principles outlined in the Declaration of Helsinki and was approved by institutional review boards at all sites (8). The present authors’ analysis was approved by the Beijing Anzhen Hospital Review Board and by BioLINCC.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.