Background Serial serum B-type natriuretic peptide (BNP or NT-BNP) testing to guide the titration of drug therapy (known as BNP monitoring) may improve clinical outcomes in patients with heart failure (HF). We used primary care data from the Clinical Practice Research Datalink (CPRD) linked with secondary care data from Hospital Episode Statistics (HES) to determine whether BNP monitoring compared with symptom-guided monitoring (usual care) is associated with reduced risk of death and all-cause hospital admission.
Methods This is a population based cohort study. We identified incident HF patients in CPRD between April 2005 and April 2013 from Read codes. Our initial definition of HF included both specific HF diagnosis codes and symptom-based codes. Patients were classified as BNP monitored or not based on the rate of BNP testing; monitoring was defined as: ≥6 months of exposure time and ≥3 BNP tests carried out with ≥2 tests per year. All-cause mortality and hospital admissions were identified through linked Office for National Statistics (ONS) and HES data, respectively.
Results Our cohort included 40,125 patients with incident HF. Of these, 100 (0.3%) were BNP monitored, 5,839 (14.6%) were BNP tested but did not meet our monitoring definition, and 34,186 (85.2%) were never tested. The groups were comparable with respect to gender, age, co-morbidities and social deprivation index. Median survival for the BNP monitored, tested and never tested groups was 3.2 years, 5.4 years and 3.6 years, respectively, and median time to first all-cause hospitalisation was 0.7 years, 0.6 years and 0.4 years, respectively.
Conclusion The proportion of patients identified as BNP monitored was low. Patients who were BNP monitored had poor survival, indicating that they were sicker. We are currently investigating whether our definition of monitoring represents repeated diagnostic testing (e.g. on admission to hospital with acute HF exacerbation). We are also conducting survival analyses accounting for patient-level confounding. Analyses will be repeated for cohorts identified using different incidence algorithms. There were challenges in setting up and analysing the study: how to manage the large size of the dataset (90,000,000 rows of data in the clinical CPRD dataset alone); there is no standard Read code set that general practitioners use to record HF; data are stored in multiple places (e.g. BNP tests are in the clinical and test datasets) in different ways (e.g. test done/not done or test result) and inconsistently across patients; and medication doses are recorded inconsistently.