Serum albumin in patients undergoing transcatheter aortic valve replacement: A meta-analysis

value of serum albumin level in patients undergoing transcatheter aortic valve replacement. A literature search of PubMed, Embase, ScienceDirect, Web of Science, SciELO, BIOSIS, Wanfang, and CNKI databases was conducted. Articles published between January 2000 and December 2017 reporting on the prognostic value of low levels of serum albumin in patients undergoing transcatheter aortic valve replacement were an-alyzed according to Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines. 11 studies including 6456 patients met inclusion criteria for meta-analysis. A lower serum albumin level was associated with a lower survival rate at follow-up in patients who un-derwent transcatheter aortic valve replacement. A subgroup analysis of eight studies reporting adjusted hazard ratios indicated that low serum albumin was independently correlated with increased post-operative mortality. The hazard ratio of mortality risk associated with each 1 g/dL increment in serum albumin level was 0.46, suggesting a potential dose--response relationship between increased serum albumin level and increased survival rate in patients undergoing transcatheter aortic valve replacement. This meta-analysis provides strong evidence for the utility of serum albumin as a prognostic marker in aortic stenosis patients undergoing transcatheter aortic valve replacement, with low serum albumin levels (2.5-3.5 g/dL) suggesting poor prognosis.

Transcatheter aortic valve replacement is becoming a more common therapeutic option for the treatment of aortic stenosis in patients at high risk for invasive surgery, but detecting which patients will benefit clinically can be challenging. Hypoalbuminemia is a useful prognostic marker for chronic inflammation in this population. We carried out a systematic review and meta-analysis of studies evaluating the prognostic value of serum albumin level in patients undergoing transcatheter aortic valve replacement. A literature search of PubMed, Embase, ScienceDirect, Web of Science, SciELO, BIOSIS, Wanfang, and CNKI databases was conducted. Articles published between January 2000 and December 2017 reporting on the prognostic value of low levels of serum albumin in patients undergoing transcatheter aortic valve replacement were analyzed according to Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines. 11 studies including 6456 patients met inclusion criteria for metaanalysis. A lower serum albumin level was associated with a lower survival rate at follow-up in patients who underwent transcatheter aortic valve replacement. A subgroup analysis of eight studies reporting adjusted hazard ratios indicated that low serum albumin was independently correlated with increased post-operative mortality. The hazard ratio of mortality risk associated with each 1 g/dL increment in serum albumin level was 0.46, suggesting a potential dose--response relationship between increased serum albumin level and increased survival rate in patients undergoing transcatheter aortic valve replacement. This meta-analysis provides strong evidence for the utility of serum albumin as a prognostic marker in aortic stenosis patients undergoing transcatheter aortic valve re-

Introduction
Aortic stenosis (AS) is widespread, affecting approximately 3% of people aged over 75 years, its prevalence increasing with age (Joseph et al., 2017). AS is associated with a high mortality risk without aortic valve replacement (AVR), which has traditionally been performed via open surgery (Vahanian and Otto, 2010). Transcatheter AVR (TAVR) is a less invasive procedure that is increasingly being used to treat AS (Joseph et al., 2017). However, recent real world study estimates one-year mortality following AVR to be around 20% for intermediate to high risk patients (Brennan et al., 2017). To improve outcomes, accurate prognostication and early intensive management of high-risk patients is important (Vahanian and Otto, 2010). However, current risk assessment methods, (e.g., The Society for Thoracic Surgery Predictive Risk of Mortality (STS PROM) (O'Brien et al., 2009) or the Logistic European System for Cardiac Operative Risk Evaluation (Logistic EuroSCORE) (O'Brien et al., 2009), which are designed to predict surgical mortality for symptomatic AS patients planning to undergo AVR), have limitations tied to their weighting of chronological age and medical diagnoses without a comprehensive evaluation of the biological status of an elderly patients (Mc-Nallan et al., 2013;Von Haehling et al., 2013). Moreover, such risk models are tailored towards surgery rather than to TAVR, the latter commonly carried out in elderly patients with cardiovascular disease suffering from frailty (McNallan et al., 2013;Von Haehling et al., 2013).
An ideal approach to risk stratification is to identify a priori the factors related to a lower survival rate, thereby allowing clinicians to select the optimal treatment protocol (Bhattacharyya et al., 2012;Vahanian and Otto, 2010). Few novel and definitive risk factors in AS have emerged in recent literature (Afilalo et al., 2014). Notably, however, there is some evidence suggesting that serum albumin concentrations of less than 3.5 g/dL represent a novel risk factor indicating frailty and poor prognosis in these patients (Kappetein et al., 2012). Serum albumin, synthesised by the liver, is the most abundant serum protein in humans (Levitt and Levitt, 2016). Normal serum albumin is in the range of 3.5 to 5.0 g/dL, and hypoalbuminemia is usually defined as less than 3.5 g/dL (Artigas et al., 2016). Serum albumin level is a relevant prognostic factor in critically and seriously ill patients (Artigas et al., 2016;Levitt and Levitt, 2016). It is an important predictor of mortality in cardiovascular diseases (Findik et al., 2016;Grodin et al., 2016;Plakht et al., 2017), and elderly patients with AS often display hypoalbuminemia associ-ated with malnutrition and chronic inflammation (Bogdan et al., 2016;Koifman et al., 2015;Yamamoto et al., 2017). Studies have been undertaken to assess whether there is any association between serum albumin levels and prognosis in AS patients undergoing TAVR, but results to date have been conflicting (Bogdan et al., 2016;Green et al., 2015Green et al., , 2012Koifman et al., 2015;Yamamoto et al., 2017). We therefore conducted a meta-analysis to evaluate the prognostic value of preoperative serum albumin levels in patients undergoing TAVR.

Search strategy
The meta-analysis adheres to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (Moher et al., 2009) (Fig. 1). PubMed, Embase, ScienceDirect, Web of Science, SciELO, BIOSIS, Wanfang, and CNKI databases were analysed to detect studies that considered the prognostic value of serum albumin in patients undergoing TAVR. The final literature search was carried out on December 31, 2017. When data Figure 1. A flow chart represents the process whereby studies were assessed for eligibility against the inclusion and exclusion criteria set out in the methods section of this article. The number of studies is the bottom of the flowchart represents that of the selected studies that were ultimately considered eligible for inclusion in this meta-analysis.
were referenced in multiple publications from the same research group, the most recent study or that including the greatest number of patients was considered in the analysis. The reference lists of relevant articles were cross-matched with the search results. The primary endpoint was all-cause mortality rate after TAVI, either reported in the short (≤ 30 days) or long term (> 30 days), while the secondary outcomes were stroke, myocardial infarction, vascular complications, bleeding, pacemaker implantation, revision surgery and endocarditis. Cause of death was categorised as either cardiovascular and non-cardiovascular mortality. Serum albumin was sought as a secondary outcome.

Selection criteria
The inclusion criteria were: (a) prospective cohort or retrospective cohort studies; (b) patients diagnosed with AS and treated with TAVR; (c) preoperative serum or plasma albumin concentrations measured and analysed for associations with prognosis; and (d) risk estimates of the prognostic role of albumin, such as hazard ratio (HR) or relative risk (RR) with 95% confidence intervals (CIs), reported. The exclusion criteria were: (a) studies not meeting all the inclusion criteria; (b) letters, reviews, or case reports; and (c) studies on cells or experimental animals. In the case of multiple studies with overlapping data, only the study with the greatest sample size was included.

Data extraction and quality assessment
Two authors independently extracted data from each included study, and any disagreements between the two were settled by discussion. If HRs from univariate and multivariate analyses were provided, only those from the latter were used. Study quality was evaluated using the Newcastle-Ottawa scale, which assigns studies scores ranging from 0 to 9 points (Stang, 2010). Scores of 7 or more indicated high study quality, while those of 6 or less indicated low quality (Stang, 2010).

Statistical analysis
To assess the prognostic value of albumin concentration in TAVR patients, study-specific HRs with 95% CIs were pooled; given the probability of increased inter-observation variance, a random-effect meta-analysis (DerSimonian-Laird method) was applied (DerSimonian and Laird, 1986). A pooled HR of greater than 1 suggested worse prognosis in patients with low serum albumin concentrations. Inter-study heterogeneity was assessed via Cochran's Q test and the results complemented the I 2 method (Higgins et al., 2002). I 2 > 50% suggested a high degree of inter-study heterogeneity (Grant and Hunter, 2006). Effect modifiers or confounders were also assessed via meta-regression. Publication bias was assessed using funnel plots, the Begg test, and the Egger test (DerSimonian and Laird, 1986). Data analyses were carried out in STATA (Stata Corporation, College Station, TX, USA). Statistical tests were two-sided, and P values lower than 0.05 were deemed statistically significant.
The main features of these studies are shown in Table 1. In total, data pertaining to 6456 TAVR patients, who both underwent TAVR and whose serum albumin was measured prior to surgery, were available for analysis. Of the eleven studies, five were prospective cohort studies, and the other six were retrospective cohort studies from USA, Spain, Japan and Israel. Articles were published between 2000 and 2017. These articles reported on risk estimates of the prognostic value of low albumin concentration in AS patients, risk estimates of the mortality associated with each 1 g/dL increase in albumin concentration, and risk estimates. The average patient age was 70-86 years, and 29-83% of patients were male. Malnutrition and wasting were evaluated via serum albumin on the day prior to TAVR in all studies.
There was no heterogeneity among the four studies reporting  risk estimates of mortality associated with each 1 g/dL increment in albumin concentration (I 2 = 0%). The HR of the mortality risk associated with each 1 g/dL increment in serum albumin level was 1.14 (95% CI: 1.01-1.29; P = 0.04), suggesting a potential doseresponse relationship between increased serum albumin level and decreased risk of mortality in patients undergoing TAVR (Fig. 5). After excluding low-quality studies, the HR of mortality risk associated with each 1 g/dL increment in serum albumin was 1.09 (95% CI: 1.03-1.16; P = 0.003; Table 2, Fig. 7).

Publication bias
There was a possible risk of publication bias among the studies reporting risk estimates of the prognostic value of low serum albumin concentrations in AS patients (Fig. 4), with Egger and Begg P values of 0.04 and 1.00, respectively. There was no risk of publication bias among the studies reporting risk estimates of mortality associated with each 1 g/dL increment in albumin concentration, with Egger and Begg P values of 0.58 and 0.84, respectively (Fig. 8).

Discussion
This meta-analysis of studies examining the prognostic importance of serum albumin levels in AS patients undergoing TAVR shows that a low preoperative serum albumin level is a significant predictor of post-procedural mortality. This indicates that measuring serum albumin levels prior to TAVR along with other preoperative factors may help to stratify patients. Meta-analysis of adjusted HRs indicated that low serum albumin was independently associated with increased mortality risk during follow-up. Furthermore, the HR of mortality risk associated with each 1 g/dL increment in serum albumin level was 0.46 (P < 0.0001), indicating a potential dose-response relationship between increased serum albumin level and decreased mortality risk. This meta-analysis provides the strongest evidence to date supporting the utility of serum albumin as a prognostic biomarker in patients undergoing TAVR, with low serum albumin levels indicating a poor prognosis.
Serum albumin levels are indicators of the severity of malnutrition and inflammation in many disease states. For instance, hypoalbuminemia significantly leads to poor prognosis in cancer, chronic kidney disease and sepsis (Cabrerizo et al., 2015). Several studies have assessed the prognostic role of serum albumin in cardiovascular diseases (Kurtul et al., 2016;Plakht et al., 2017;Su et al., 2012). In Plakht et al. (2017), low serum albumin concentration was significantly associated with long-term all-cause mortality after acute myocardial infarction (Plakht et al., 2017); Kurtul et al. (2016) found that serum albumin concentration was inversely associated with mortality risk in acute coronary syndrome patients (Kurtul et al., 2016); and Su et al. (2012) showed that serum albumin was a significant prognostic factor in patients with heart failure (Su et al., 2012). The findings from the present meta-analysis consolidate the notion that serum albumin is a useful prognostic biomarker in patients undergoing TAVR, thereby further supporting the evidence of its prognostic role in cardiovascular diseases. Our findings show that decreased serum albumin levels in TAVR patients are related not only to deterioration in nutritional status that adversely affects myocardial recovery after correction of AS, but also to poor functional outcomes. Therefore, in acute care, a current focus is on restarting enteral nutrition immediately after extubation, with patient-specific nutrition.
Low serum albumin concentration implies a dysfunctional liver synthesis, which can be caused by heart failure, cirrhosis and malignancies, amongst other conditions. Albumin helps control serum electrolyte levels and confers antioxidant effects (Rothschild et al., 1988). Serum albumin provides a simple but objective clinical risk stratification in patients with severe AS assessed for TAVR to make appropriate decisions regarding treatment options for individual patients.

Limitations
Differences in study design, sampling method, inclusion criteria and patient characteristics, and adjusted confounding factors, could result in a high degree of heterogeneity between studies. Furthermore, some studies were retrospective cohort studies, where biases are more likely than in prospective study. More prospective studies with larger sample sizes are needed to provide further evidence of the utility of serum albumin level as a prognostic biomarker in patients undergoing TAVR. A third limitation was that in some studies, the follow-up duration was less than one year, so the long-term prognostic role of serum albumin in AS patients needs further study. The number of included studies, especially in the subgroup analyses, was limited. In addition, the systematic review protocol was not registered in PROSPERO (the international Figure 5. A forest plot represents the statistical effect of low serum albumin concentration on the risk of mortality, following evaluation via a multivariate Cox regression model. Figure 6. A forest plot represents the statistical effect of categorical serum albumin concentration on the overall risk of mortality, following appropriate statistical evaluation. Figure 7. A forest plot represents the statistical effect of continuous serum albumin concentration on the overall risk of mortality, following appropriate statistical evaluation. Figure 8. A forest plot represents the statistical effect of serum albumin concentration on the overall risk of mortality, following appropriate statistical evaluation, excluding studies whose quality was deemed low. prospective register of systematic reviews), and it may have benefited from such a preliminary peer-review process.
In conclusion, this meta-analysis strongly supports the utility of serum albumin as a prognostic biomarker in AS patients undergoing TAVR, with a low serum albumin level (2.5-3.5 g/dL) sug-gesting a poorer prognosis. Serum albumin is a convenient and economical prognostic factor, which can be easily monitored for risk stratification in AS patients. More prospective studies with large sample sizes are needed to provide further evidence of the utility of serum albumin as a prognostic biomarker in patients un-dergoing TAVR.