This in vitro experimental study was conducted at Konkuk University Medical Center (KUMC), Seoul, Korea, from July 2020 to January 2021. The study protocol was designed in accordance with the Declaration of Helsinki and was approved by the Institutional Review Board of KUMC. This study used anonymized samples and required neither additional sampling nor therapeutic intervention. Therefore, the requirement of obtaining written informed consent from the enrolled individuals was waived.

The study population consisted of four subsets: samples from the universal sample bank of the American Association for Clinical Chemistry (AACC) [19], samples from KUMC healthy individuals, samples from KUMC HF patients, and samples from the Barcelona ambulatory HF cohort subsets (Barcelona samples) [20]. The AACC and KUMC healthy individual samples were used for cut-off validation, the samples from KUMC HF patients for assay comparison, and the Barcelona samples for the equivalence of prognosis prediction. The AACC samples were a part of the full sample set (n = 800) that the assay manufacturer (Boditech Med Inc.) purchased from the AACC sample bank (invoice #1084290) with a purchase request (inclusion criteria: mixed males and females in all-age ranges). After the internal use for assay development and validation, the manufacturer provided remaining samples randomly with associated information. The manufacturer also purchased 300 samples (invoice # FV20/0035) from the Barcelona ambulatory HF cohort subsets. After excluding 48 samples obtained from patients who died for reasons other than cardiovascular (CV) events, the manufacturer provided the remaining 252 samples with associated information. For the KUMC samples (n = 123 from healthy individuals and n = 206 from patients with HF), we used samples that were leftover after routine laboratory testing. The characteristics of the study population are summarized in Table 1 (Ref. [19, 20]).

The samples were stored at –70 °C and thawed at 37 °C for measuring sST2 levels. The sST2 levels measured using Presage ST2 were considered as reference, and those measured using AFIAS ST2 and ichroma ST2 were compared with the reference.

Presage ST2 is an ELISA comprising a ready-to-use 96-well microtiter plate coated with mouse monoclonal anti-human sST2 antibodies; spectrophotometric absorbance is measured at 450 nm with a microtiter well reader. The assay was performed using a Gemini automated microplate processor (Stratec Biomedical Systems, Birkenfeld, Germany) [11, 12]. It takes 3 h to measure sST2 levels using Presage ST2, and its hands-on time is 30 min. The measurable range of Presage ST2 was 3.1–200 ng/mL, and its coefficient of variation (%) was less than 10.0%. The limit of detection (LoD) and limit of quantification (LoQ) were 1.8 ng/mL and 2.4 ng/mL, respectively.

AFIAS ST2 is a fluorescent sandwich immunoassay for the automatic quantitative determination of the sST2 antigen. AFIAS ST2 can measure sST2 levels in various samples, such as whole blood, serum, and plasma collected in lithium heparin or EDTA vacutainers. For this study, 100 $\mu$L of serum was dispensed into the sample well of the cartridge containing the test strip. After loading the cartridge into the AFIAS-6 system (Boditech Med Inc.), all procedures, from loading the detection buffer into the cartridge to obtaining test results, were automated. Briefly, a fluorescence-labeled antibody conjugate in the detection buffer binds to the antigen in the sample to form antibody-antigen complexes. The complexes migrate onto a nitrocellulose membrane and are captured by antibodies on the test line of the strip. More antigens in the sample form more antigen-antibody complexes, leading to stronger fluorescence intensity [21]. It takes 12 min to obtain sST2 levels using AFIAS ST2, without hands-on time. The manufacturer-claimed measurable range was 3.1–200 ng/mL, and its coefficient of variation (%) was less than 5.0%. The LoD and LoQ were 2.8 ng/mL and 3.1 ng/mL, respectively.

The ichroma ST2 is a manual-type assay with the same principle as that of AFIAS ST2. It can also be used to measure sST2 levels in various samples, such as whole blood, serum, and plasma collected in lithium heparin or EDTA vacutainers. For this study, 150 $\mu$L of diluent and 75 $\mu$L of serum mixtures were transferred to a detector tube containing a fluorescence-labeled antibody conjugate, and 75 $\mu$L of the mixture was loaded into the ichroma cartridge manually. After 12 min, the test results were displayed on the screen of the ichroma II reader (Boditech Med Inc.). It takes 12 min to obtain sST2 levels using ichroma ST2, and its hands-on time is one minute. The manufacturer-claimed measurable range was 3.1–200 ng/mL, and its coefficient of variation (%) was less than 5.4%. The LoD and LoQ were 2.8 ng/mL and 3.1 ng/mL, respectively.

On performing the Shapiro–Wilk test, all data exhibited a non-parametric distribution [22]. Therefore, the data are expressed as medians with interquartile ranges (IQR) for the continuous variables and as numbers with proportions for the categorical and binary variables. Using the Reed method and generalized extreme studentized deviate technique, three outliers were identified and excluded from the analysis (n = 1 in KUMC healthy individuals; n = 2 in KUMC HF patients) [23, 24].

We validated the clinical cut-off (35 ng/mL) according to the CLSI guideline EP28-A3C [25]. The 95th percentile upper reference limit (URL) with a 90% confidence interval (CI) was calculated for the AFIAS ST2 and ichroma ST2 results.

We compared the results of the three assays using Passing–Bablok regression and Bland–Altman plots, according to the CLSI guideline EP09C-ED3 [24]. The correlation coefficients (r) were interpreted as follows: 0.30, negligible; 0.30–0.49, low; 0.50–0.69, moderate; 0.70–0.89, high; and $\ge$0.90, very high correlations [26]. On the Bland–Altman plot, the mean difference and $\pm$1.96 standard deviations (SD) were interpreted informally to visualize the mean difference. Weighted kappa ($\kappa$) values with 95% CI were used to calculate the degree of agreement using the clinical cut-off and were interpreted as follows: 0.20, poor; 0.21–0.40, fair; 0.41–0.60, moderate; 0.61–0.80, good; and $>$0.81, very good [27].

We assessed the assay equivalence for predicting CV death at a cut-off of 35 ng/mL using areas under the curve (AUC) in receiver operating characteristic (ROC) curves, according to the CLSI guideline EP24-A2 [28]. The sensitivity, specificity, Youden index, positive likelihood ratio, and negative likelihood ratio were calculated.

Our sample size fulfilled the minimum requirement recommended by the CLSI guidelines (120 observations for cut-off validation and 100 samples for assay comparison) [24, 25]. For the prognosis prediction equivalence assessment, the sample size was thought to have approximately 95% power (1-$\beta$) to detect a difference between the two assays with a 0.05 two-tailed significance level [28]. All statistical analyses were conducted using MedCalc Statistical Software (version 20.027; MedCalc Software Ltd, Ostend, Belgium). Rounding rules were applied to summary statistics, and a two-tailed p-value less than 0.05 was considered statistically significant [29].

Table 2 shows the clinical cut-off validation of Presage ST2, AFIAS ST2, and ichroma ST2 results. In both sample subsets, the 95th percentile URLs of AFIAS ST2 and ichroma ST2 were within the clinical cut-off of 35 ng/mL established using Presage ST2. There was no visible trend according to the assay or the origin of the samples.

For the KUMC HF samples, both AFIAS ST2 and ichroma ST2 results were highly correlated with Presage ST2 results (r = 0.82 and 0.81, respectively); however, the former showed a positive proportional bias (+21% and +17%, respectively). The mean differences of AFIAS ST2 and ichroma ST2 with Presage ST2 were –4.8 ng/mL and –3.7 ng/mL, respectively (Fig. 1). The results of AFIAS ST2 and ichroma ST2 showed strong agreement with those of Presage ST2 ($\kappa$ = 0.84 and 0.82, Table 3). The ST2 levels measured using AFIAS ST2 and ichroma ST2 were higher than those measured using Presage ST2.

Fig. 1.

Comparison of AFIAS ST2, ichroma ST2, and Presage ST2 assays using KUMC HF samples (n = 206). Correlation between (A) AFIAS ST2 vs. Presage ST2 and (B) ichroma ST2 vs. Presage ST2 using Passing–Bablok regression analysis. Differences between (C) AFIAS ST2 vs. Presage ST2 and (D) ichroma ST2 vs. Presage ST2 using Bland–Altman plots. All p-values were 0.001. Solid line, Passing–Bablok regression or mean difference; dashed line, 95% CI or $\pm$ 1.96 SD. Abbreviations: CI, confidence interval; HF, heart failure; KUMC, Konkuk University Medical Center; n, number; SD, standard deviation; ST2, suppression of tumorigenicity 2.

At a cut-off of 35 ng/mL, Presage ST2, AFIAS ST2, and ichroma ST2 could predict CV death in the Barcelona ambulatory HF cohort subsets; all three assays showed comparable AUCs, and there were no statistically significant differences across the results of the three assays (Fig. 2). In all three assays, the specificity for CV death was $>$85%.

Fig. 2.

ROC curve analyses of AFIAS ST2, ichroma ST2, and Presage ST2 assay results to predict CV death in the Barcelona ambulatory HF cohort subsets at a cut-off of 35 ng/mL (n = 252). Abbreviations: AUC, area under the curve; CI, confidence interval; CV, cardiovascular; HF, heart failure; n, number; ROC, receiver operating characteristic; ST2, suppression of tumorigenicity 2; +LR, positive likelihood ratio; -LR, negative likelihood ratio.

This study had several limitations. First, we compared the three sST2 assays using 206 samples from patients with HF. Due to the relatively small sample size and lack of serial samples from the same patients, in-depth analyses were not conducted considering HF classification and the number of recurrent hospitalizations. Second, we analyzed 252 samples from the Barcelona ambulatory HF cohort to predict CV death. Further large-scale studies are needed to determine the prognosis of HF among patients with elevated sST2 level in the Korean population. Third, we focused on validation of the clinical cut-off, comparison of analytical performances, and assessment of equivalence for prognosis prediction among the three assays. Due to limited clinical and laboratory information on the AACC sample bank and Barcelona ambulatory HF cohort, we could not include other demographic and baseline data, such as body mass index, waist circumference, and lipid profiles. Further, we could not analyze the prognostic significance considering the sST2 levels using the three ST2 assays and HF classification by left ventricular ejection fraction according to the AHA/ACC/HFSA guidelines [1]. Further studies are needed to explore prognosis through these ST2 assays, depending on the HF classification according to the AHA/ACC/HFSA guidelines. Finally, we did not analyze the turnaround time (TAT) or hands-on time for each POC assay. Based on the manufacturer-claimed TAT, we determined that AFIAS ST2 and ichroma ST2 are more suitable for clinical practice than ELISA.

Considering the increasing prevalence of preclinical stages of HF, early diagnosis and personalized POC strategies are required for HF management [6]. AFIAS and ichroma ST2 are newly launched IVD POC assays and could be an easy-to-use option for sST2 level measurement. Using these POC assays, clinicians can make immediate clinical decisions when treating HF patients, which may decrease the overall medical burden. Further research is required regarding the relationship between decreased TAT, a shorter hospitalization period, and medical cost reduction in the real world.