BNP and troponin I measurements did not improve risk stratification of acute HF patients

Background

Patients with acute heart failure (AHF) are at high risk of HF re-hospitalization and death within 30 days of admission [1]. Accurate risk stratification at the acute phase could support physicians to select the most appropriate treatment strategies, which might lead to a better prognosis for these patients [2]. However, existing predictive models are of limited clinical use because they are not very accurate, especially for readmissions [3].

In this retrospective analysis of the Value of Endothelin Receptor Inhibition with Tezosentan in

Acute heart failure (VERITAS) studies [4], the value of adding BNP and troponin I to conventional prognostic markers was evaluated. For this purpose, multivariable prognostic models including baseline clinical data were created for:

a) the composite of death, in-hospital worsening HF, or HF re-hospitalization at 30 days,

b) death or HF re-hospitalization at 30 days,

c) death at 90 days.

Excluding biomarker data, the multivariable model included age, heart rate, respiratory rate, SBP, history of COPD, history of DM, history congestive HF, history of renal impairment, dyspnea severity by visual analogue scale (dyspnea VAS) at baseline, albumin, blood urea nitrogen (BUN), hemoglobin, and sodium.

The data of 1347 patients (93%) in the VERITAS studies were analyzed. They had been enrolled within 24 hours from admission for AHF, and were randomized to intravenous tezosentan, an endothelin antagonist, or placebo. BNP and troponins were measured retrospectively on stored plasma, because they had been measured in only 20% of patients when the study was ongoing. Moreover, troponin I, and not troponin T, was used for this analysis, because a higher number of troponin T values were missing, and troponin I and T values were highly correlated.

Note: the VERITAS program was discontinued prematurely because of the low probability of achieving a significant treatment effect.

Main results

• BNP values were elevated in most patients, with a median value of 422 pg/mL, and 5% of patients had values below the assay’s lower limit (41 pg/mL).
• Measurement of BNP did not add to the model and troponin I added little information, increasing the c-index to only 0.6595 (from 0.6528) without significant differences in c-indices between models.
• At 30 days, 150 patients had died or been re-hospitalized for worsening HF. The independent predictors for this outcome were age, heart rate, SBP, history of congestive HF, dyspnea VAS at baseline, albumin, BUN, hemoglobin, and sodium.
• The final model resulted in a c-statistic of 0.6855, which was not significantly improved by either biomarker.
• At 30 days 97 patients were re-hospitalized for worsening HF. The risk of worsening HF hospitalization was associated with similar baseline characteristics but, neither troponin I nor BNP was predictive of this outcome.
• At 90 days, 135 patients had died. Biomarkers showed a much stronger relationship to mortality than to composite outcomes on univariable analysis. The multivariable model identified age, heart rate, SBP, history of COPD, history of vascular disease, dyspnea VAS at baseline, white blood cell count, albumin, BUN, and sodium as significant predictors of mortality with an overall c-index of 0.7394. BNP did not provide further prognostic information. The addition of troponin I provided a small improvement in the c-index to 0.7461. The difference in c-indices between the two models was not statistically significant.

Conclusion

References

Find this article online at Eur J Heart Failure