NT-proBNP by itself predicts mortality and CV outcomes in T2DM

Introduction and methods

B-type natriuretic peptides (BNP) are biomarkers for myocardial stress and well established predictors of heart failure (HF) outcomes [1,2]. BNPs are also, when incorporated in multivariable hazard models, a predictor of death and CV events in individuals with type 2 diabetes [3-6], especially in those who have a comorbidity of HF [7,8], chronic kidney disease (CKD) [9-11] or recent acute coronary syndrome [12,13]. But despite the improved discriminatory strength of this marker in risk models for T2DM, NT-proBNP has not been used for risk assessment in clinical practice.

The current study assessed the discriminatory ability of NT-proBNP alone in predicting the risk of death and CV outcomes in 5509 patients with T2DM and CVD and/or CKD using data from the Aliskiren in Type 2 Diabetes Using Cardiorenal Endpoints (ALTITUDE) trial.

The ALTITUDE trial was a randomized, double-blind, placebo-controlled, parallel group study in patients with type 2 diabetes. Individuals (≥35 years) received antidiabetic drugs or had a documented fasting plasma glucose of ≥126 mg/dL or 2-hour plasma glucose ≥200 mg/dL and were either on ACE-inhibitors or ARBs without any change in antihypertensive therapy for at least 4 weeks before randomization. Participants were also diagnosed with persistent macroalbuminuria (albumin-to-creatinine ratio ≥200 mg/g) and an eGFR ≥30 mL/min/1.73m², or microalbuminuria (albumin-to-creatinine ratio ≥20 mg/g) and/or a history of CV disease (MI, stroke, HF, or coronary artery disease) and a reduced kidney function (eGFR ≥30 mL/min/1.73 m²). Patients were subdivided into 10 deciles based upon risk prediction or NT-proBNP concentration levels (<36, 37-62, 63-92, 93-128, 129-176, 177-244, 245-348, 349-544, 545-1002, and >1003 pg/mL). The endpoints were all cause death and CV events defined as CV death, resuscitated cardiac arrest, nonfatal MI, nonfatal stroke, or unplanned HF hospitalization. Median follow-up was 2.6 years.

Three risk models were generated: 1) a multivariable base model with 20 important clinical variables, 2) an univariable NT-proBNP model, and 3) the multivariable base model combined with the NT proBNP model. The risk prediction performance of the NT-proBNP only model was compared to that of the base risk model and the base model plus NT-proBNP, using C-statistics.

Main results

• Baseline characteristics showed increased levels for NT-proBNP in deceased patients or patients who had developed a CV event. For all cause mortality, baseline NT-proBNP levels were 1267.9±2611.8 pg/mL in the deceased group (n=469) compared to 389.5±1091.9 pg/mL in the survivor group (n=5040) [P<0.001]. For CV outcomes, baseline NT-proBNP levels were 1126.1±2286.5 pg/mL in the group that experienced a CV event (n=768) compared to 357.1±1040.3 pg/mL in the group that had no CV event (n=4741) [P<0.001].
• Prediction of death in the cohort using the multivariable base risk model, resulted in a C-statistic of 0.744 (95% CI:0.722-0.767). Using the univariable NT-proBNP model gave a similar C-statistic value of 0.745 (95% CI:0.723-0.768). Combining the base risk model with NT-proBNP improved the prediction of death significantly, and resulted in a C-statistic value of 0.779 (95% CI:0.723-0.768, P<0.001 compared to base model).
• The mortality rates per 100 persons-year using the base model were 0.7 (95% CI:04-1.2) in low-risk patients (1st decile of risk) and 11.6 (95% CI:9.9-13.7) in high-risk patients (10th decile of risk). Similar results were obtained with NT-proBNP as a single variable in a risk model. The mortality rate using this model was 0.7 (95% CI:0.4-1.2) in patients in the lowest decile (NT-proBNP <36 pg/mL) and 11.6 (95% CI:9.9-13.6) in patients in the highest decile (NT-proBNP >1003 pg/mL).
• The prediction of an CV event gave similar results as prediction of death. The base risk model gave a C-statistic value of 0.731 (95% CI: 0.714-0.749) compared to 0.723 (95% CI:0.704-0.741) in the NT proBNP only model (P=0.37). Combining the two models improved the prediction ability significantly, resulting in a C-statistic of 0.763 (95% CI:0.746-0.78, P<0.001).
• The predicted CV incidence rates per 100 person- years by the risk base model were 0.9 (95% CI:0.5-1.5) in low-risk patients (1st decile of risk) and 19.2 (95% CI:16.8-22.0) in high-risk patients (10th decile of risk). The NT-proBNP single variable model predicted similar incidence rates compared to the base model. In low-risk individuals with a NT-proBNP concentration of <36 pg/mL an incidence rate of 1.3 (95% CI:0.5-1.5) was predicted, and 19.4 (95% CI:16.9-22.1) in those with an NT-proBNP <1003 pg/mL.

Conclusion

References

Find this article online at J Am Heart Assoc.