Cardiovascular Surgery, Cardiology, and Clinical Research, Taipei Tzuchi Hospital, New Taipei City, Taiwan;

Surgery, Tzuchi University College of Medicine, Hualian, Taiwan.

Health One Network Co., Ltd. New Taipei City Taiwan, (telephone: 886-2-22993688--x13; cellphone: 886-930142606); Critical Diagnostics.

Soluble ST2 (sST2) is a novel biomarker (Dieplinger et al. 2009) of fibrosis (Moore and Januzzi 2010) that reflects cardiac dysfunction such as remodeling or post-injury scaring (Sanada et al. 2007, Weir et al. 2010, Willems et al. 2013, Díez 2008). According to previous research, its trending has superior signal-to-noise and clinical relevance for cardiac prognosis to conventional biomarkers for heart failure such as NT-proBNP (Weinberg et al. 2002, Pascual-Figal et al. 2011a, Santhanakrishnan et al. 2012). For valvular heart disease patients and their outcomes, we would like to investigate the association and predictive value of ST2 in cardiac surgical non-coronary valve patients.

Conventional biomarkers for assessing heart failure are BNP / NT-proBNP. But they have various confounders like age, BMI, and renal disease (Mueller and Dieplinger 2013). When heart perceives stress, their serum levels begin to rise and their random variations are high. Thus their role of monitoring heart failure prognosis may be unsatisfactory for physicians who need timely responses (Willems et al. 2013, Bayes-Genis et al. 2011, Rehman et al. 2008).

Soluble ST2 can reflect the process of cardiac remodeling and myocardial post-injury fibrosis in heart failure (Rehman et al. 2008, Schmitz et al. 2005, Moore and Januzzi 2010), and so it has much less confounding than conventional BNP/NT-proBNP (Mueller and Dieplinger 2013). When ST2 serum levels climb, it means there is considerable myocardial damage (Broch et al. 2012, Lupón et al. 2013). After treatment we can check if ST2 begins to drop to assess whether the treatment works (Pascual-Figal et al. 2011a, Chen et al. 2013). Even in the asymptomatic patients (Wang et al. 2013), if the ST2 serum levels are higher than the cutoff 35 ng/ml, it means there is significant cardiac damage (Mueller and Dieplinger 2013, Januzzi 2013). We can monitor ST2 serum level to predict disease occurrence (Chen et al. 2013) and make treatment plans as early as possible (Boisot et al. 2008, Manzano-Fernández et al. 2012). In addition ACCF/AHA has proposed the 2013 Guideline for the Management of Heart Failure that includes ST2 in the recommendations (Yancy et al. 2013). ST2 and NT-proBNP can work together for better sensitivity and specificity (Pascual-Figal et al. 2011a, Bayes-Genis et al. 2011, Sabatine et al. 2008).

As above, there are also various prior research studies that demonstrated the association of sST2 with acute and chronic heart failure (Mueller et al. 2008, Shah et al. 2009, Wojtczak-Soska et al. 2013, Daniels et al. 2010, Bhardwaj and Januzzi 2010, Henry-Okafor et al. 2012, Manzano-Fernández et al. 2011, Shah and Januzzi 2010, Weinberg 2003).

To stratify the operation risk of valvular heart patients by preop sST2 and other known preop variables;

Currently we can implement ST2 in the following scenarios:

Health checkup and outpatient screening (Wang et al. 2013);

We will collect ST2, and other cardiac biomarkers at different time points of cardiac valve surgical patients and investigate their relationships with different outcome measures and also may compare the performance of these biomarkers on predicting various outcomes (Gaggin et al. 2014).

We plan to recruit 70 patients and measure three serum samples per case. There will be 210 laboratory tests. Due to budget limitation, we set the sample size to be 7, since we can only afford 210 laboratory tests. We set the alpha-level to be 0.05.

The outcome variables (dependent variables) will be the following and we will select more appropriate variables to present according to the interim analysis results. The primary endpoints include: hospital mortality, all-cause mortality, unexpected ICU return, and MACE. The secondary endpoints include: time to extubation, ICU time, and total LOS.

We will transform several variables to facilitate analysis. Renal function will be graded by GFR (glomerular filtration rate) as estimated by creatine-related formulae. Comorbidity such as diabetes or hypertension will be determined by clinical judgment. For biomarker analysis (sST2, NT-proBNP, troponin-T), in addition to absolute values, we will use relate values for better demonstrate the change from baseline and the trends. Relative serum level (RSL) is a dimensionless value defined as the ratio of serum level to the baseline (level divided by baseline).

For comparing scale or ordinal (quasi-scale) variables, we will use Mann-Whitney U test for two groups and Kruskall-Wallis test for three or more groups. For comparing nominal variables, we will use Fisher’s exact test. For comparing time-event variables, we will use log-rank test. For modeling predictions, we will choose appropriate statistical regression methods based upon dependent variables. For scale dependent variable, we will use linear or log-linear regression models depending the likelihood of normality violation. For binary dependent variables, we will use logistic regression model. For time-event variables, we will use Cox regression model. For evaluating biomarker performance on accuracy of classification, we will receiver operating characteristic (ROC) analysis. The truth variables will be one of the outcome variables and the classifier variables will be ordinal-ranged biomarker absolute level or RSL. The tentative optimal cutoff values for each biomarker may be proposed depending on the outcome variables or clinical scenarios.

The backbone for the results presentation will be: cohort background statistics, time trends of biomarkers, biomarker and predictors for the outcomes, and comparisons of biomarkers.

The statistical software we used will be Stata/MP 13.1 for Mac (64-bit Intel).

(1) ST2 and other cardiac biomarker lab assay;

(2) Statistical analysis of the study;

(3) Interpretation and implement of the study results.

(Table 2)