NT-proBNP during and after primary PCI for improved scheduling of early hospital discharge

Primary percutaneous coronary intervention (PPCI) substantially improved outcome in patients with ST-elevation myocardial infarction (STEMI), which has been paralleled by a decrease in hospital length of stay [1‐3]. In addition several studies have demonstrated that most low-risk PPCI patients are eligible candidates for early discharge [4‐11].

The Zwolle Risk Score (ZRS) is a simple tool designed to identify PPCI patients who can be safely discharged within 72 h, based on their 30-day mortality risk [11]. In patients with a score ≤3 (low risk) the mortality rate was 0.1% at 2 d and 0.2% between 2–10 d and early discharge between 48–72 h could be applied in slightly more than 60% of patients (Fig. 1). Recently, we demonstrated that the predictive accuracy of the ZRS was improved by adding baseline NT-proBNP [12]. In this report we studied the value of NT-proBNP measurement at different timepoints over and beyond the ZRS.

Briefly, enrolment was from June 2006 to November 2007. Eligible patients were aged 21–85 years with symptoms of acute myocardial infarction of more than 30 min but less than 24 h and ST-segment elevation of more than 1 mV in two adjacent electrocardiograph (ECG) leads. Exclusion criteria were severe renal dysfunction (glomerular filtration rate <30 ml/min or serum creatinine >200 mmol/l (>2.5 mg/dl)), therapy-resistant cardiogenic shock (systolic blood pressure ≤80 mm Hg for >30 min), persistent severe hypertension (systolic blood pressure >180 mm Hg or diastolic blood pressure >110 mm Hg), or a contraindication to anticoagulation, or increased risk of bleeding. Also, patients with left bundle branch block, pregnant and/or breastfeeding women, and patients with a life expectancy of less than one year were excluded. Written informed consent for participation in both the On-TIME 2 study and future data analysis was obtained from each patient. All local ethics committees involved approved the study protocol.

Patients were categorised by percentiles of NT-proBNP levels. We assessed the predictive value of the ZRS, serial NT-proBNP levels and the absolute change of NT-proBNP levels from baseline to 18–24 h (delta NT-proBNP) for the primary endpoint.

We computed and plotted receiver operating characteristic (ROC) curves. A total of 74 patients had missing ZRS values and 36 patients did not have baseline NT-proBNP levels. Three patients had missing values of both ZRS and baseline NT-pro BNP. A multiple missing value imputation procedure was applied in which the values of the NT-proBNP levels and the values of the elements of the ZRS were imputed in case they were missing.

We computed the net reclassification index (NRI) and the integrated discrimination improvement (IDI) to evaluate models with an extra predictor relative to a model with the ZRS alone [15]. The IDI combines the increase in mortality probability for those experiencing an event plus the decrease in mortality probability for those not experiencing an event. The NRI quantifies the net proportion of subjects being correctly reclassified to low or high mortality probabilities when adding a predictor to the model using a cut-off point of 30%.

To analyse models using a linear combination of ZRS and NT-proBNP, we calculated weighted scores for each as follows: (β1 * ZRS) + (β2 * ln NT-proBNP), where β1 and β2 denote β‑coefficients for the ZRS and log NT-proBNP obtained from the multivariate Cox regression model. We used ROC curves to define optimal cut-off values in estimating mortality outcomes. We calculated the pooled areas under the curve (AUC) and 95% confidence intervals by averaging over the 10 AUCs from the imputed datasets using Rubin’s rule [16]. The cut-off values and scores for defining possible early discharge were obtained from the mean scores across the 10 datasets, which gave the highest specificity in combination with a 100% sensitivity for the primary endpoint (e. g. zero mortality 30 d after PPCI). A two-sided p-value of ≤0.05 was considered to be statistically significant. The analyses were performed in SPSS version 22 and SAS version 9.3.

A total of 861 patients underwent PPCI. Complete follow-up was present in 845 patients. Baseline characteristics are shown in Table 1. The mean age was 62 years and 76% were male. Most patients had low Killip class with a post PPCI ejection fraction >40%. Mean hospital stay was 4.89 d.

Baseline NT-proBNP values ranged from 9–33,927 pg/ml (mean 599 ± 1883 pg/ml) with a median of 138 pg/ml. At 18–24 h and 72–96 h, NT-proBNP was determined in 786 and 768 patients respectively, and ranged from 45–20,653 pg/ml (mean 2089 ± 2738 pg/ml, median 1283 pg/ml) and 17–24,119 pg/ml (mean 1618 ± 2556 pg/ml, median 860 pg/ml) respectively.

The primary endpoint was reached in 21 patients. In univariate and multivariate analysis, ZRS and NT-proBNP measured at the different timepoints, but not delta NT-proBNP, were strongly associated with 30-day mortality (Table 2).

Our study demonstrates that NT-proBNP, withdrawn at different timepoints after PPCI, improves prognostication of the established ZRS and consequently may improve the scheduling of early hospital discharge.

Although the clinical outcomes in STEMI patients treated by PPCI have been considerably improved, optimising hospital length of stay in these patients is still challenging.

Recent reports make clear that large groups of patients can be discharged safely at or within 48 h [6, 9, 10], while the present guidelines recommend discharge after a minimum of 3 d [1, 2]. In addition, a lot of international variation in hospital admission duration among low-risk PPCI patients still exists [3, 17‐19]. Assuming that the number of low-risk PPCI patients worldwide will increase because of improvement of treatment, there is simultaneously a need for reduction of health care costs. This highlights the necessity for easy tools for risk stratification, in which ideally large proportions of patients could be discharged early. For this purpose, De Luca et al. developed the ZRS. Important variables in this risk model are the success of coronary reperfusion and the occurrence of heart failure. These and other early adverse post PPCI events can also be predicted by NT-proBNP [20‐22].

Within this context, NT-proBNP can be considered to be an overlapping variable. The incremental value of adding NT-proBNP to established risk factors may be explained by the fact that biomarker measurement provides more accurate information, without observer variability.

The prognostic significance of the serial changes of plasma NT-proBNP in STEMI patients has previously been studied [23‐25]. It was shown that NT-proBNP increases markedly within 24 h. At this timepoint a maximum value is reached after successful reperfusion and then usually decreases, while unsuccessful reperfusion is associated with increased NT-proBNP values at 48–96 h [23, 25].

Consequently, the change in NT-proBNP concentration could also be predictive. In one report, the difference between NT-proBNP at 72 h and baseline NT-proBNP was superior to baseline NT-proBNP, but similar to NT-proBNP at 72 h in predicting short-term adverse cardiac events [25]. Interestingly, in our study the change in NT-proBNP was less predictive and cut-off values were difficult to define, since some patients had negative changes in NT-proBNP (or higher values at baseline). However, measurement at two different timepoints is more laborious and would not be useful within the context of early discharge.

Based on our decision rule at 48 h after PPCI (ZRS < 2 or NT-proBNP < 2500 pg/ml at 18–24 h), a proportion as large as 75% of patients could be discharged early. Therefore NT-proBNP measurement at 18–24 h post PPCI may optimise length of hospital stay in low-risk patients. Furthermore, this risk strategy can also be applied to patients in non-intervention hospitals, since many patients are repatriated in the first 24 h. Especially this group tends to be kept in hospital for a longer period of time [26].