In search of an efficient strategy to monitor disease status of chronic heart failure outpatients: added value of blood biomarkers to clinical assessment

The Serial Biomarker Measurements and New Echocardiographic Techniques in Chronic Heart Failure Patients Result in Tailored Prediction of Prognosis (Bio-SHiFT) study was designed to investigate the hypothesis that temporal patterns of biomarkers involved in CHF are associated with prognosis. Bio-SHiFT is an ongoing prospective, observational study of stable outpatients with CHF, conducted in Erasmus MC, Rotterdam, the Netherlands and Noordwest Ziekenhuisgroep, Alkmaar, the Netherlands. Patients were recruited during their regular outpatient visits and were in a clinically stable condition. Patients were eligible for inclusion if aged 18 years or older, capable of understanding and signing informed consent, and if CHF (including HF with preserved ejection fraction) was diagnosed ≥3 months ago according to the guidelines of the European Society of Cardiology (ESC) [1, 4, 19]. Detailed inclusion and exclusion criteria are shown in Fig. 1. The study was approved by the medical ethics committees of the participating hospitals and was conducted in accordance with the Declaration of Helsinki. Written informed consent was obtained from all patients. The study is registered in ClinicalTrials.gov, number NCT01851538. Follow-up for this analysis lasted from October 2011 until November 2013.

At baseline, patients were evaluated by trained research physicians, who collected information on HF-related symptoms, including NYHA class [1, 4]. History of chronic renal failure was defined as a glomerular filtration rate less than 60 ml/min/1.73 m². Alcohol consumption was defined as drinking ≥1 alcoholic consumption per day. Electrocardiography and echocardiography were performed. Data were entered into electronic case report forms. Non-fasting blood and urine samples were collected.

Study follow-up visits were scheduled every 3 months (a window of ±1 month was allowed), for a maximum of 30 months. At each tri-monthly study visit, a short medical evaluation was performed, NYHA functional class was scored and blood and urine samples were collected.

Blood samples were collected at baseline and at each follow-up visit, and were processed and stored locally at a temperature of −80 °C within 2 h after blood collection. When applicable, samples were transported to the central laboratory (Erasmus MC, Rotterdam, the Netherlands) under controlled conditions (at a temperature of −80 °C), until batch analysis was performed. Thus, the biomarker measurements performed for this study did not lead to treatment adjustments.

Batch analysis of NT-proBNP, Hs-TnT, and CRP was performed in the Clinical Chemistry Laboratory of the Erasmus MC. Plasma NT-proBNP and Hs-TnT were analysed using electrochemiluminesence immunoassays (Roche Diagnostics, Elecsys 2010, Indianapolis, Indiana, USA). For NT-proBNP, concentrations were measured ranging from 0.6 to 4130 pmol/l. Coefficients of variation (CVs) were <5% at mean values ranging from 5.19–274 pmol/l. For Hs-TnT, concentrations were measured ranging from 3–10,000 ng/l. CVs were <5% at mean values ranging from 12.7–1,819 ng/l. CRP was measured using an immunoturbidimetric assay (Roche Hitachi 912 chemistry analyser, Basel, Switzerland). This system measures concentrations ranging from 0.3 to 350 mg/l, and CVs were <5% at mean values ranging from 0.84–284 mg/l.

During follow-up, endpoints were recorded in the electronic case report forms by trained research physicians, and associated hospital records and discharge letters were collected. Subsequently, a clinical event committee blinded to the biomarker results reviewed all collected information and adjudicated primary and secondary endpoints.

The primary endpoint comprised the composite of cardiac death, cardiac transplantation, left ventricular assist device implantation, and hospitalisation for the management of acute or worsened HF.

Cardiac death was defined as death from myocardial infarction (MI) or other ischaemic heart disease (ICD-10: I20–I25), death from other heart disease including HF (I30–I45 and I47–I52), sudden cardiac death (I46), sudden death undefined (R96) or unwitnessed or ill-described death (R98, R99). Hospitalisation for acute or worsened HF was defined as exacerbation of symptoms typical of HF, in combination with two of the following: BNP or NT-proBNP > 3x the upper limit of normal, signs of worsening HF, such as pulmonary rales, raised jugular venous pressure or peripheral oedema, increased dose or intravenous administration of diuretics, or administration of positive inotropic agents.

Statistical methods are described in detail in the supplemental text (online Electronic Supplementary Material). In brief, we used linear mixed models to assess the associations between serial biomarker measurements and repeated assessment of NYHA functional class. Associations between serial measurements of biomarkers and NYHA class, and occurrence of the primary endpoint, were examined by entering the serial measurements as time-varying covariates into extended Cox proportional hazards models. First, the models were adjusted for age, gender, systolic blood pressure and estimated glomerular filtration rate (eGFR; calculated using the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equation). Subsequently, all variables, i. e. NT-proBNP, Hs-TnT, CRP and NYHA functional class, were entered simultaneously into the models to investigate their independence. We calculated time-dependent C‑indices based on the extended Cox models. Analyses were performed with R Statistical Software and MedCalc.

A total of 263 patients were included from October 2011 to June 2013. Baseline characteristics are displayed in Tab. 1. Mean age was 67 years (SD ± 13), 72% were men, and 73% were in NYHA class I or II. Median duration of HF at inclusion was 4.6 years (IQR 1.7–9.9). Median baseline NT-proBNP was 137.3 pmol/l (IQR 51.9–272.9), Hs-TnT 18.0 ng/l (IQR 9.6–33.2) and CRP 2.2 mg/l (IQR 0.9–4.8).

During follow-up, we collected 921 blood samples and scored NYHA functional class 1,292 times. Of all follow-up visits, 1,135 took place before the occurrence of the primary endpoint. During these follow-up visits, 886 blood samples were drawn (median 3; IQR 2–5 per patient), and NYHA functional class was scored 1,114 times (median 4; IQR 2–6 per patient).

Repeatedly measured NT-proBNP and Hs-TnT showed strong associations with repeatedly assessed NYHA class, whereas CRP did not (NT-proBNP: β [95% CI]: 1.56 [1.17–2.06]ln(ng/l) increase per point increase in NYHA class, p = 0.002; HsTNT: β [95%CI]: 1.58 [1.21–2.07], p = 0.001; CRP: β [95% CI]: 1.22 [0.98–1.53], p = 0.076) (Tab. 1 of the online Electronic Supplementary Material).

The composite endpoint was reached by 41 patients (16%), during a median follow-up of 1.0 [0.6–1.4] years: 5 patients died from a cardiovascular cause, 35 patients were re-hospitalised for worsened HF and 1 patient underwent heart transplantation. Of the 35 patients reaching the primary endpoint because of re-hospitalisation for HF, 16 patients died eventually during further follow-up, of whom 12 patients died from a cardiovascular cause. Overall all-cause mortality was 21 (8.0%).

NT-proBNP, Hs-TnT, CRP, and NYHA class all displayed strong and positive associations with the primary endpoint (Tab. 2). After multivariable adjustment, NT-proBNP, CRP and NYHA class remained independently associated with the primary endpoint, while Hs-TnT lost statistical significance. Of all the other baseline characteristics, only age was independently associated with the primary endpoint.

Temporal evolutions of serial biomarker measurements and NYHA class are displayed in Fig. 2. Serially measured NT-proBNP, Hs-TnT, CRP, and NYHA class all displayed strong and positive associations with the primary endpoint (Tab. 2). After multivariable adjustment for all serially measured variables, NT-proBNP, CRP and NYHA class remained significantly associated with the primary endpoint.

The discriminative abilities of the models containing baseline measurements of NT-proBNP, Hs-TnT, CRP and NYHA class are shown in the online Electronic Supplementary Material, supplemental Tab. 2, and those of serial measurements in Tab. 3. All individual serially assessed C‑indices, except for Hs-TnT, were numerically higher than corresponding baseline C‑indices. Adding serial NT-proBNP to the model containing age, sex, systolic blood pressure and eGFR and NYHA class provided a substantial increase in C‑index from 0.76 (CI 0.66–0.86) to 0.84 (0.74–0.93), although this did not reach statistical significance (p = 0.26). Adding serial CRP instead to this same multivariable model resulted in a C-index of 0.82 (0.72–0.92), p = 0.40. Adding both serial NT-proBNP and CRP to the multivariable model only resulted in a slight improvement compared with addition of NT-proBNP only: 0.85 (CI 0.75–0.95), p = 0.20.

Extended Cox models with time-dependent covariates were used to analyse the effects of changes in NYHA class and temporal biomarker patterns on the primary endpoint, because these models are able to accommodate multiple time-varying covariates. However, time-dependent Cox models assume that biomarker levels do not change between measurements [21]. In reality, blood biomarkers are dynamic and continuously change over time, parallel to the condition of the patient. Therefore, we performed a sensitivity analysis by means of a joint modelling approach. Joint models combine a linear mixed-effects model for the serial biomarker measurements with a Cox proportional hazards model for the occurrence of the primary endpoint [21]. We estimated the individual biomarker trajectories and NYHA trajectories using separate joint models, then extracted the fitted trajectories from the joint models, and entered the extracted trajectories simultaneously into one extended Cox model. The results of this analysis were materially the same as those we described in the paper. Furthermore, the majority of the patient population was in NYHA class I or II, and thus at relatively low risk. The results and conclusions should be judged accordingly.