Response to Comment on Pop-Busui et al. Screening Natriuretic Peptide Levels Predicts Heart Failure and Death in Individuals With Type 1 and Type 2 Diabetes Without Known Heart Failure

We thank Fujisawa and Fujikawa (1) for their thoughtful letter regarding our study (2), and we welcome the opportunity to clarify several important points concerning our heart failure (HF) outcome ascertainment as well as issues related to coding practices and their impact on administrative claims–based diagnosis. Their central concern, as we understand it, is that an elevated screening natriuretic peptide (NP) result may itself prompt downstream evaluation for HF, and that some of the outpatient HF diagnoses we captured may therefore reflect workup triggered by the biomarker rather than incident clinical disease. This is an important methodological consideration that we anticipated and addressed in our original study design. We would like to highlight several prespecified features of our study design that mitigate the concerns raised by the authors about short-term coding bias following an elevated NP result.

As described in the RESEARCH DESIGN AND METHODS section, incident HF was defined as either one inpatient HF diagnosis or two outpatient HF diagnoses recorded at least 90 days apart, with the second diagnosis used as the incident HF date (2). This approach has been applied in prior large-scale epidemiologic work using Medicare and other administrative data sets (3,4), and it was selected because a single outpatient diagnosis can lack specificity and may reflect initial suspicion of HF without confirmation. The use of the second diagnosis date allows for confirmation in the outpatient setting and improves specificity. A transient evaluation cascade triggered by a single elevated NP would generate clustered diagnostic activity within a single clinical episode, not two confirmed HF diagnoses sustained across at least two calendar quarters. Patients whose HF diagnosis is reaffirmed by an independent clinician encounter months later are far more likely to have clinically meaningful, persistent disease.

To further address this issue, we report two prespecified sensitivity analyses in our study. First, we applied a more stringent definition of incident HF that additionally required the qualifying ICD-10 codes to be recorded in the primary position, with HF as the principal reason for the encounter rather than a secondary or rule-out consideration. Under this stricter definition, the magnitude of the associations was essentially unchanged. Second, we performed a landmarked analysis excluding individuals with HF diagnoses within 3 months of the index NP test without affecting the findings. Furthermore, the separation in the Kaplan-Meier curves by NP result category continued well beyond the period immediately following the NP test, suggesting that we are capturing true clinical events rather than transient administrative labeling. Finally, the graded, monotonic dose-response across NP categories provides further support for a biological rather than ascertainment-driven signal. A workup-prompted diagnosis labeling effect would be expected to produce a short-term increase in incident events clustered around the time of testing and a step change at the clinical decision threshold of NP. Instead, we observed sustained risk across the entire follow-up period and a continuous gradient extending across both moderately and markedly elevated strata of NP.

The authors also raise concerns that the criterion of two outpatient diagnoses could be influenced by the U.S. insurance environment, where payer mix has been shown to affect diagnostic coding intensity. While coding variation is an inherent limitation of any administrative data, it is unlikely that systemic coding practices or payer-driven behaviors would correlate so strongly with the physiological magnitude of a patient’s NP lab result as to substantially alter the graded, dose-response association we observed.

We appreciate the author’s recommendation regarding additional secondary analysis restricting the incident HF outcomes to inpatient diagnosis only. However, such a restriction would be methodologically inappropriate for a study evaluating prognostic implications of biomarker-based screening. In contemporary U.S. practice, ∼40% of all incident HF is now diagnosed in the outpatient setting, and this proportion has been rising steadily, from 36% in 2004 to 41% in 2013 (4). Restricting incident HF to inpatient diagnoses would therefore systematically exclude a substantial and growing share of true incident HF and would shift the captured outcome toward late-stage decompensated disease while excluding the earlier, ambulatory HF phenotype that screening is specifically designed to identify. HF diagnosed in the outpatient versus inpatient setting represents distinct disease severity stages rather than equivalent end points differing only in coding setting. This is reflected in the markedly different prognosis of the two settings, with 1-year mortality rate of 13.5% in the outpatient diagnosis versus ∼36% in the inpatient diagnosis incident HF cases (4).

While outpatient identification of incident HF reflects earlier-stage clinically recognized disease, it does not undermine the value of NP screening but rather reinforces it by facilitating early initiation of evidence-based therapies that can prevent disease progression and subsequent HF hospitalization. Together, the validated outcome definition, sensitivity analysis results, and pattern of results demonstrating sustained long-term divergence of the survival curves with a continuous dose-response across NP strata collectively argue against a workup-driven explanation for our findings.

Funding Statement

R.P.-B. has received research funding to her institutions from the National Institute of Diabetes and Digestive and Kidney Diseases (grants U01 DK119083 and R01-DK126837-01A1). A.P. has received research support from the National Institute on Minority Health and Disparities (grant R01MD017529), the National Heart, Lung, and Blood Institute (grant R21HL169708), and the American Heart Association.