Economic evaluation of NT-proBNP testing plus standard clinical assessment versus standard clinical assessment alone for the prevention of heart failure in older adults with type 2 diabetes

Abstract

Introduction

N-terminal pro-B-type natriuretic peptide (NT-proBNP) testing for early heart failure (HF) detection in asymptomatic older adults with diabetes in the USA remains limited despite guideline recommendations. This study evaluated the cost-effectiveness of incorporating NT-proBNP testing into routine clinical assessments for adults aged ≥65 years from a Medicare perspective.

Research design and methods

A five-state Markov model with an embedded decision tree compared NT-proBNP testing plus standard clinical assessments with standard assessments alone in adults aged ≥65 years with type 2 diabetes, with or without other HF risk factors. Model inputs (ie, baseline event rates, treatment effects, utilities, and costs) were derived from published literature and publicly available Medicare data. Outcomes, including direct medical costs (ie, diagnostic testing, cardioprotective treatments, HF-related hospitalization, and chronic/advanced HF treatment), life-years (LYs), and quality-adjusted life-years (QALYs), were calculated over a lifetime horizon. Sensitivity and scenario analyses were performed.

Results

Compared with standard clinical assessment alone, NT-proBNP testing prolonged survival (6.27 vs 4.99 LYs) and improved quality of life (3.45 vs 2.84 QALYs). Although lifetime HF hospitalizations were similar (0.43 vs 0.43), LY-adjusted hospitalizations were lower with NT-proBNP testing (0.29 vs 0.43). Lifetime costs were higher with NT-proBNP testing (US$117 010 vs US$92 283), reflecting longer survival and greater use of appropriate cardioprotective treatments. NT-proBNP was cost-effective, with an incremental cost per QALY gained of US$40 883. Results were robust, with the model most sensitive to changes in diabetes management costs, mortality among high-risk individuals, and cardioprotective drug effectiveness.

Conclusions

Annual NT-proBNP testing in older adults with type 2 diabetes is cost-effective for early HF detection in a Medicare setting, supporting broader adoption of guideline-recommended testing in this high-risk population.

WHAT IS ALREADY KNOWN ON THIS TOPIC.

WHAT THIS STUDY ADDS

• Adding NT-proBNP testing to standard annual clinical assessments in asymptomatic older adults with type 2 diabetes is clinically beneficial and cost-effective from a US Medicare perspective.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• Findings support guideline-recommended use of NT-proBNP testing for early HF detection in US clinical practice for older adults with type 2 diabetes and provide evidence to inform broader payer coverage for routine assessment in this high-risk population.

Introduction

Cardiovascular complications remain a leading cause of morbidity and mortality among individuals with type 2 diabetes.¹ Heart failure (HF) is often the first manifestation of cardiovascular disease in this population, with a prevalence of approximately 10%–30% and an incidence rate ~2.5 times higher than in demographically comparable persons without diabetes.^2–4 HF risk increases with age and other diabetes-related complications, such as diabetic kidney disease.^{5 6} Individuals with HF and diabetes account for up to 44% of all HF-related hospitalizations,⁷ and HF prognosis is poor, with outcomes worsening with advancing age.⁸ Taken together, these factors, along with the high per-patient treatment costs and healthcare resource utilization in this population,⁹ highlight the substantial and growing burden of comorbid HF and type 2 diabetes on healthcare systems.

Although early detection of HF may improve patient outcomes and reduce long-term healthcare costs by enabling earlier initiation of appropriate cardioprotective therapies,¹⁰ the signs and symptoms of HF are often clinically silent or non-specific in its early stages, particularly among older adults with multiple comorbidities.¹¹ Nearly 80% of people receive their initial HF diagnosis during a hospital admission,¹² when disease is more advanced and treatment options are limited. Biomarkers such as N-terminal pro-B-type natriuretic peptide (NT-proBNP) can improve subclinical HF detection; elevated NT-proBNP concentrations, even in seemingly stable patients, signify cardiac stress in HF that precedes symptom onset.¹⁰ In this context, NT-proBNP is a stronger predictor of cardiac complications in type 2 diabetes than other established biomarkers (including albuminuria),¹³ with elevated NT-proBNP values associated with increased risk of unplanned cardiovascular hospitalization or mortality.¹⁴

Despite strong evidence supporting the use of NT-proBNP for early HF detection, risk stratification, and prognostic assessment in persons with type 2 diabetes—as well as explicit recommendations from the American Diabetes Association (ADA) and American College of Cardiology/American Heart Association (ACC/AHA)^{2 6 15}—uptake in the USA remains limited,¹⁶ and payer coverage is often restricted. Given that early HF detection can reduce future healthcare utilization and costs through earlier cardioprotective therapy initiation,¹⁰ closing this NT-proBNP usage gap could improve both health and economic outcomes. Although economic evaluations conducted across several European countries suggest that long-term cost savings from earlier HF detection can outweigh the upfront costs of testing, follow-up, and treatment,^{17 18} studies specifically evaluating the cost-effectiveness of routine NT-proBNP testing among older adults with type 2 diabetes in the USA are non-existent, creating uncertainty for both clinicians and payers.

In this analysis, we estimated the cost-effectiveness of NT-proBNP testing plus standard clinical assessments compared with standard clinical assessments alone in adults aged ≥65 years with type 2 diabetes, with or without other HF risk factors, on Medicare.

Research design and methods

A decision-analytic model was developed from a US Medicare perspective to assess the lifetime cost-effectiveness of adding annual NT-proBNP testing to standard clinical assessments, using real-world data to estimate costs, clinical outcomes, and uncertainty through scenario and sensitivity analyses.

Model population

Adults aged ≥65 years with type 2 diabetes, with or without other HF risk factors (eg, hypertension, obesity, and coronary artery disease), and no clinical signs or symptoms of HF were included in the model.

HF classification

Individuals were classified using ACC/AHA criteria for HF stage,¹⁵ ranging from asymptomatic patients at risk but without HF (stage A) and progressing to those with pre-HF but no symptoms (stage B), symptomatic HF (stage C), and, finally, advanced HF (stage D). As diabetes is an established risk factor for HF under ACC/AHA staging criteria, all individuals entering the model were classified as having at least stage A HF.

Comparator interventions

The analysis compared two diagnostic strategies for the detection and management of HF as part of routine assessment: standard clinical assessment and NT-proBNP testing plus standard clinical assessment. Under standard clinical assessment, individuals received annual clinical evaluations without NT-proBNP testing, where HF management was based solely on clinical examination. Echocardiography was performed only if individuals presented with clinical signs or symptoms of HF. In the NT-proBNP-guided assessment strategy, individuals underwent NT-proBNP testing as part of their annual clinical visit, with those showing elevated NT-proBNP levels (>125 pg/mL) proceeding to confirmatory echocardiography to guide management, per clinical guidelines.^{6 15}

Model structure

A decision-analytic model in which risk assessment was modeled via a Markov structure with an embedded decision tree was developed using Microsoft Excel (V.2605). The model used an annual cycle and a lifetime horizon, which was deemed long enough to capture the full benefits and risks associated with NT-proBNP testing, as recommended by cost-effectiveness guidelines.¹⁹ The Markov structure captured transitions over time between five distinct, mutually exclusive health states (figure 1): No event, limited cardioprotective (LCP) treatment, representing individuals with stage A or B HF on standard or low-intensity cardioprotective therapy; No event, intensified cardioprotective (ICP) treatment, representing individuals at high risk for HF (stage B) receiving high-intensity cardioprotective therapy; HF hospitalization, for individuals experiencing an HF hospitalization event; Chronic/advanced HF, for individuals with established chronic or advanced HF (stage C or D); and All-cause death, an absorbing state to which individuals could transition from any other health state. The embedded decision tree specifically modeled the annual clinical assessment process, determining whether individuals remained in the No event, LCP treatment state or progressed to the No event, ICP treatment state on the basis of NT-proBNP results and/or standard assessment outcomes.

Figure 1. HF model structure. Note: Cost accruals for state transitions included risk assessment costs (NT-proBNP testing, standard clinical assessments, echocardiograms, and cardiologist visits for echocardiograms), annual costs for diabetes treatment in individuals without HF, costs for HF hospitalization events, and annual costs for treating chronic or advanced HF. ^aIndividuals identified as at high risk for HF hospitalization who have their risk confirmed by echocardiography are placed on ICP treatment. In the model, these individuals transition to the No event, ICP treatment health state, where their risk of HF hospitalization and death is reduced due to appropriate treatment. ^bIndividuals at high risk for HF hospitalization but with a test result that denoted NT-proBNP ≤125 pg/mL do not undergo echocardiographic confirmation and are inappropriately categorized as low risk. They remain on limited cardioprotective treatment and are therefore undertreated, resulting in a higher risk of HF hospitalization and death. ^cIndividuals at low risk for HF hospitalization but with a test result denoting NT-proBNP >125 pg/mL undergo an unnecessary echocardiogram and remain on LCP treatment. ^dIndividuals at low risk for HF hospitalization with a test result denoting NT-proBNP ≤125 pg/mL remain on LCP treatment. ^eIndividuals with subclinical HF are asymptomatic and cannot be identified as high risk through standard clinical assessment alone (assumed detection sensitivity, 0%); therefore, they remain on LCP treatment until symptomatic HF or HF hospitalization occurs, after which they transition directly to Chronic/advanced HF rather than to No event, ICP treatment state. HF, heart failure; ICP, intensified cardioprotective; LCP, limited cardioprotective; NT-proBNP, N-terminal pro–B-type natriuretic peptide.

Adults with type 2 diabetes and no clinical signs or symptoms of HF, but who may have other HF risk factors, all enter the model in the No event, LCP treatment health state and are assessed annually for risk of HF events and progression to Chronic/advanced HF. Under standard clinical assessment, those without symptoms remain in the No event, LCP treatment health state. Individuals who develop signs or symptoms of HF, as determined through standard clinical examination, transition to the Chronic/advanced HF health state. Individuals who experience an HF event requiring hospitalization move to the HF hospitalization health state. Under standard clinical assessment plus NT-proBNP, individuals are classified on the basis of their NT-proBNP test result as high risk (>125 pg/mL) or low risk (≤125 pg/mL) for HF. Individuals identified as high risk undergo a confirmatory echocardiogram to verify elevated HF risk. Those with confirmed high risk (ie, true positives) who have not yet experienced an HF hospitalization event initiate ICP treatment and move to the No event, ICP treatment health state.

Individuals with an NT-proBNP test result >125 pg/mL whose echocardiogram does not indicate elevated HF risk (ie, false positives) remain in the No event, LCP treatment health state, incurring the costs of unnecessary echocardiography. Individuals with an NT-proBNP test result ≤125 pg/mL are classified as low risk and, regardless of whether this classification is accurate, are assessed similarly to those receiving standard clinical assessment alone and remain in the No event, LCP treatment health state. Those inappropriately categorized by NT-proBNP as low risk (ie, false negatives) remain on LCP treatment and are therefore undertreated.

From both the No event, LCP treatment and No event, ICP treatment health states, individuals have a chance of progressing to either HF hospitalization or Chronic/advanced HF. After an HF hospitalization, individuals remain in the HF hospitalization state for the remainder of the year. Individuals in this health state incur the costs and utilities associated with HF hospitalization as well as the costs and utilities of an individual classified as having chronic/advanced HF. These individuals may also incur subsequent HF hospitalizations within the same year. After this period, they transition to the Chronic/advanced HF health state, where they remain until death and accrue the resource use, costs, and utilities associated with chronic/advanced HF, including subsequent HF hospitalizations. Individuals could move to the All-cause death state at any time.

Model parameters were derived from published literature and publicly available Medicare data and are described in the following sections.

Model inputs and adjustments

Natural history

The natural history of clinical events is summarized in table 1. The prevalence of NT-proBNP levels >125 pg/mL in individuals with type 2 diabetes was taken from a pooled analysis of seven epidemiological cohort studies.²⁰ Each year, the incidence of individuals at high risk (ie, NT-proBNP >125 pg/mL) increases. The resulting annual increase in individuals with NT-proBNP >125 pg/mL over subsequent model cycles was estimated from a post hoc analysis of the Canagliflozin Cardiovascular Assessment Study (CANVAS) trial,²¹ which included participants with type 2 diabetes with and without prior HF, similar to this model. Annual HF-related hospitalizations for both low- and high-risk individuals were also derived from the CANVAS trial.²¹

Table 1. Model inputs for clinical parameters.

Parameter	Estimate (95% CI)	Source
Percentage of patients with type 2 diabetes at high risk of HF
Initial year	60.9% (48.7% to 73.1%)	Patel et al20
Increase in high-risk patients in subsequent years	2.3% (1.9% to 2.8%)	Januzzi et al21
Annual probability of having an HF hospitalization
Low-risk patients	0.6% (0.5% to 0.7%)	Rådholm et al34
High-risk patients	2.8% (2.2% to 3.4%)	Rådholm et al34
90-day HF readmission	14.9% (14.7% to 15.0%)	Khan et al35
Annual probability of progression to chronic/advanced HF
Low-risk patients	0.0024 (0.0019 to 0.0029)	Young et al22
High-risk patients	0.0143 (0.0115 to 0.0172)	Young et al22
Distribution among chronic and advanced HF
Chronic HF (stage C)	90.2% (79.6% to 97.2%)	Ammar et al36
Advanced HF (stage D)	9.8%	Calculated
Annual hospitalization rate (HR)
Chronic HF (stage C)*	0.84 (0.80 to 0.88)	Shakoor et al37 Dunlay et al38
Advanced HF (stage D)	2.91 (2.78 to 3.06)	Dunlay et al38
Treatment effect on reduction in HF hospitalization (HR)
SGLT2i	0.51 (0.33 to 0.78)	Rådholm et al34
Intensification of baseline cardioprotective treatments†	0.51 (0.33 to 0.78)	Rådholm et al34
Mortality (HR)
Diabetes versus no diabetes	1.68 (1.52 to 1.87)	Li et al39
High risk for HF versus low risk for HF	2.53 (1.78 to 3.61)	Januzzi et al21
Treatment effect due to addition of SGLT2i versus control	0.70 (0.51 to 0.96)	Rådholm et al34
Treatment effect due to intensification of non-SGLT2i cardioprotective treatments versus control*	0.70 (0.51 to 0.96)	Rådholm et al34
HF hospitalization (inpatient mortality)	3.6% (2.9% to 4.3%)	Akintoye et al40
Annual mortality for patients with chronic HF	4.7% (3.9% to 5.6%)	Kalogeropoulos et al41
Annual mortality for patients with advanced HF	49.9% (46.5% to 53.0%)	Dunlay et al38
NT-proBNP test
Sensitivity	90.0% (72.0% to 100.0%)	Elecsys proBNP II PI24
Specificity	93.0% (74.4% to 100.0%)	Elecsys proBNP II PI24

^*The HR for the annual hospitalization rate in chronic HF (stage C) was estimated by proportionally scaling the advanced HF hospitalization rate from Dunlay et al using the ratio of 1-year hospitalization rates for chronic versus advanced HF reported by Shakoor et al.

^†Non-SGLT2i treatments included treatment for hypertension with ACE inhibitors, angiotensin II receptor blockers, renin-angiotensin system inhibition with angiotensin receptor-neprilysin inhibitors, and beta-blockers. Treatment effects for non-SGLT2i were assumed to be the same as treatment effects for SGLT2i from Rådholm et al.

HF, heart failure; HR, Hazard ratio; NT-proBNP, N-terminal pro–B-type natriuretic peptide; PI, prescribing information; SGLT2i, sodium-glucose cotransporter-2 inhibitor.

Progression to Chronic/advanced HF was dependent on whether an individual was classified as higher or lower risk for HF and was based on a US general population study, where patients with stage A or B HF were followed up over an average of 4 years to determine progression to stage B or C/D.²² To align with the source study, the Chronic/advanced HF health state combined stage C (chronic HF) and stage D (advanced HF) disease; however, costs, utilities, and mortality rates were modeled separately for each stage. Individuals with chronic/advanced HF may experience additional HF hospitalizations, which are associated with higher costs and reduced quality of life. Age-specific all-cause mortality was obtained from US National Vital Statistics²³ and adjusted using published HRs to account for the impacts of type 2 diabetes, risk for HF, HF hospitalization, and the effects of ICP treatment (table 1).

Diagnostic testing

Individuals correctly identified by NT-proBNP testing as high risk for HF initiate ICP treatment, reducing the probability of HF hospitalization and all-cause mortality. Conversely, individuals incorrectly classified as low risk for HF remain undertreated on LCP treatment and face a correspondingly higher risk of HF events. Sensitivity, reflecting the proportion of high-risk individuals (NT-proBNP >125 pg/mL) correctly identified, and specificity, reflecting the proportion of low-risk individuals (NT-proBNP ≤125 pg/mL) correctly classified, were set at 90.0% and 93.0%, respectively, based on Elecsys proBNP II product information.²⁴ Echocardiograms were initiated when the NT-proBNP test result denoted >125 pg/mL, an HF hospitalization event occurred, or an individual transitioned directly to Chronic/advanced HF.

Treatment effect

Cardioprotective treatment was categorized as either LCP or ICP treatment, differing by the extent of therapy received (online supplemental table S1). Treatments included sodium-glucose cotransporter-2 inhibitors (SGLT2i), mineralocorticoid receptor antagonists (MRA), beta-blockers, renin-angiotensin system inhibition with angiotensin receptor-neprilysin inhibitors (ARNi), angiotensin II receptor blockers (ARB), and angiotensin-converting enzyme inhibitors (ACEi). Glucagon-like peptide-1 receptor agonists (GLP1-RA) were excluded, as they are currently not approved for use as a cardioprotective treatment in HF.

To better reflect real-world practice, the proportion of individuals receiving LCP or ICP treatment was estimated from two real-world studies rather than assuming universal guideline-concordant care. One study analyzed Medicare beneficiaries with type 2 diabetes and HF to assess the use of cardioprotective treatments before and after HF hospitalization,²⁵ and the other examined US commercial healthcare claims to characterize SGLT2i and ACEi/ARB use among individuals with and without diabetes and HF or atherosclerotic cardiovascular disease.²⁶ As the baseline transition probabilities already reflect the effects of receiving LCP treatment, no additional LCP treatment effect on the rate of progression to Chronic/advanced HF was applied. Treatment effects for ICP treatment were assumed to be nonadditive when used in combination and were applied sequentially in the following order: SGLT2i, then intensified baseline treatments (eg, ACEi, ARB, MRA, and ARNi). Following HF hospitalization, individuals were assumed to have already progressed to Chronic/advanced HF; therefore, ICP treatment was not assumed to modify the rate of progression. The percentage of individuals receiving each type of treatment was used to estimate the overall impact of cardioprotective treatment on HF-related hospitalization and death.

Costs

Annual prescription costs for cardioprotective treatment were estimated using Red Book pricing (2024) combined with the proportion of individuals receiving each type of treatment (online supplemental table S1). No discounts, copays, or rebates were applied. Costs associated with the treatment of diabetes in the absence of HF, HF hospitalizations, and chronic or advanced HF treatment were derived from published literature and data from the Centers for Medicare and Medicaid Services and the US Bureau of Labor Statistics (online supplemental table S2). Other costs included HF risk assessment costs for NT-proBNP, standard clinical assessment, and echocardiograms. Standard clinical assessment was assumed to be the cost of a physician visit. All costs were calculated in 2024 US dollars, and medical consumer price indices were used to inflate to 2024 values, where appropriate. Beyond the first year of the model, costs were discounted at 3.0% per annum.¹⁹

Utilities

Health state-specific utilities and utility decrements were used to capture the impact of HF and related events on individuals’ health-related quality of life (online supplemental table S2). A utility value of 1.0 represents perfect health, whereas a value of 0.0 represents death. A multiplicative approach was used to combine health state utilities with age-based utility weights (online supplemental table S3). Health outcomes were discounted at 3.0% per annum.¹⁹

Model verification and validation

Model structure, assumptions, and outputs were reviewed by two practicing clinicians to assess clinical plausibility. Selected model inputs and outputs were also informally compared with published literature to assess whether model behavior was consistent with external evidence. Model calculations were independently verified by a health economist who was not involved in model development. Input parameter values were checked against source documents to confirm accurate data extraction and implementation, and model functionality was tested across a range of input values to ensure calculations performed as expected.

Analyses

Cost-effectiveness analysis

Medical costs, life-years (LYs), and quality-adjusted life-years (QALYs) were calculated across a lifetime horizon. We measured the cost-effectiveness using incremental cost-effectiveness ratios (ICERs), expressed as incremental cost per QALY gained, and calculated by dividing the difference in costs by the difference in QALYs for standard clinical assessment with and without NT-proBNP testing. ICER was compared with a willingness-to-pay (WTP) threshold of US$50 000, a commonly cited benchmark in US cost-effectiveness analyses.²⁷

Scenario analyses

A scenario analysis was conducted to evaluate the effects of ICP treatment using ACEi, ARB, MRA, and beta-blockers and was based on treatment regimens and outcomes reported in PONTIAC I.²⁸ In this study, patients with type 2 diabetes and elevated NT-proBNP (>125 pg/mL) but without established cardiac disease received individualized visits for initiation and uptitration to maximum tolerated dosages of renin-angiotensin system antagonists and beta-blockers. The treatment effect for ICP treatment in this scenario analysis was based on the HR for HF hospitalization of 0.14 reported in the ‘intensified’ arm of PONTIAC I.²⁸

A second scenario analysis was performed to evaluate the impact of expanding the study population to include adults of all ages with type 2 diabetes from a commercial perspective (mean age, 50 years), reflecting a broader population of individuals who may benefit from earlier HF identification. Age-specific mortality, the probability of 90-day HF rehospitalization (15.5%), and inpatient mortality during HF hospitalization (1.4%) were adjusted accordingly for this cohort. Costs related to diabetes treatment, HF hospitalization, and the treatment of chronic/advanced HF for the commercial cohort are presented in online supplemental table S4.

Sensitivity analyses

Deterministic sensitivity analyses were conducted by varying key parameters—including costs, treatment benefits, relative risks, and transition probabilities—within plausible ranges defined by SEs, 95% CIs, or ±20% of baseline values to determine their relative impact on the results.

Probabilistic sensitivity analyses were performed using a second-order Monte Carlo simulation with 10 000 iterations to reflect uncertainty across all model parameters simultaneously. In each simulation, parameters were randomly sampled from appropriate statistical distributions: beta distributions for prevalence, utilities, and probabilities; gamma distributions for costs, treatment effects, and mortality impacts; and normal distributions for hospitalization rates. Results were presented using a scatter plot, showing how incremental costs and health benefits varied across simulations relative to the WTP threshold, and a cost-effectiveness acceptability curve, showing the probability that NT-proBNP testing was cost-effective compared with standard clinical assessment for a given WTP threshold.

Data and resource availability

Data sources for this study are publicly available. The model and the health economic analysis plan that guided its development are not publicly available.

Results

Base-case analysis

Individuals assessed with NT-proBNP testing plus standard clinical assessment versus those receiving standard clinical assessment alone gained more LYs (6.27 vs 4.99) and QALYs (3.45 vs 2.84) (figure 2A). The increase in longevity led to an increased number of echocardiograms (1.24 vs 0.44), but HF hospitalizations (0.43 vs 0.43) were comparable. When assuming equal LYs between groups (4.99 years), individuals receiving NT-proBNP assessment incurred more echocardiograms (1.01 vs 0.44) to confirm test results but experienced fewer HF hospitalizations (figure 2B; 0.29 vs 0.43), resulting in lower total costs (US$91 125 vs US$92 283). However, over the lifetime horizon, NT-proBNP testing resulted in higher total costs (figure 2C; US$117 010 vs US$92 283), driven by longer survival and greater time accruing healthcare costs. Overall, NT-proBNP testing was cost-effective at the US$50 000 WTP threshold, with an ICER of US$40 883. All lifetime health outcomes and medical costs for the base-case analysis are presented in online supplemental table S5.

Figure 2. (A) LYs and QALYs, (B) average number of hospitalizations, and (C) lifetime medical costs accrued by adults ≥65 years receiving NT-proBNP plus standard clinical assessment and standard clinical assessment alone. (D) Scatter plot showing the incremental costs and incremental QALYs for adults ≥65 years across 10 000 simulations. ICER, incremental cost-effectiveness ratio; LY, life year; NT-proBNP, N-terminal pro-B-type natriuretic peptide; QALY, quality-adjusted life-year; WTP, willingness-to-pay.

Scenario analysis

When assuming an improved treatment effect for intensified baseline cardioprotective treatment (HR=0.14), outcomes for NT-proBNP testing in the scenario analysis remained comparable to those found in the base case (online supplemental table S6). Minimal changes in LYs (6.28 vs 4.99), QALYs (3.45 vs 2.84), echocardiograms (1.22 vs 0.44), and HF hospitalizations (0.41 vs 0.43) decreased the ICER to US$40 193 compared with US$40 883 in the base case. This minor impact reflects the assumption that treatment effects from SGLT2is, which are used by most individuals when intensifying cardioprotective treatment, override those of baseline cardioprotective therapies and are not additive, limiting the incremental impact of intensifying baseline treatment.

When a commercial population with a younger average age of 50 years was considered, scenario results were broadly consistent with those in the base case: NT-proBNP testing compared with standard clinical assessment alone led to longer survival (19.87 vs 16.89 LYs), greater QALYs (9.63 vs 8.54), and more echocardiograms (5.08 vs 4.10) (online supplemental table S7). Unlike the base case, NT-proBNP testing in the scenario analysis resulted in fewer HF-related hospitalizations (3.76 vs 4.06), despite individuals living longer. Although lifetime costs were higher with NT-proBNP testing than with standard clinical assessment alone (US$369 259 vs US$360 866), the smaller incremental difference relative to the base case (US$8393 vs US$24 727) resulted in a substantially lower ICER of US$7743—well below the US$50 000 WTP threshold—indicating cost-effectiveness in this younger commercial population. Assuming equivalent LYs (16.89), NT-proBNP testing compared with standard clinical assessment alone was cost-saving, resulting in fewer HF-related hospitalizations (2.88 vs 4.06) and lower total costs (US$316 914 vs US$360 866).

One-way sensitivity analysis

The ICER was most sensitive to variations in the annual cost of diabetes management among individuals with type two diabetes without HF, the increased mortality among individuals at high risk for HF, and the effect of SGLT2is on decreasing mortality. In all cases, NT-proBNP testing remained cost-effective (ICER <US$50 000 WTP threshold) when parameters were varied within their plausible ranges. The tornado plot for the one-way sensitivity analysis is presented in figure 3.

Figure 3. One-way sensitivity analysis results: base-case analysis. Note: For each parameter, the label lists the base-case value followed by the minimum and maximum values used in the one-way sensitivity analysis. HF, heart failure; ICER, incremental cost-effectiveness ratio; NT-proBNP, N-terminal pro-B-type natriuretic peptide; SGLT2i, sodium-glucose cotransporter-2 inhibitor.

Probabilistic sensitivity analysis

NT-proBNP testing was cost-effective (ICER <US$50 000 WTP threshold) in 84.8% of model simulations (online supplemental file S1). The probability of cost-effectiveness exceeded >99.9% at WTP thresholds of US$100 000 or US$150 000. The mean ICER across all simulations was US$42 492, closely aligning with the base-case value of US$40 883 (figure 2D).

Discussion

This study found that early detection of asymptomatic HF through systematic NT-proBNP testing of older adults with type 2 diabetes is a cost-effective approach that prolongs survival, increases quality of life, and may lead to fewer LY-adjusted HF-related hospitalizations. Importantly, NT-proBNP testing remained cost-effective even when parameters, such as diabetes management costs and cardioprotective therapy efficacy, were varied within plausible ranges, supporting confidence in the reliability of the results for clinical policy decision-making. Scenario analyses further showed that the benefits of NT-proBNP testing extend to adults with type 2 diabetes across all ages, although younger adults represent a smaller proportion of type 2 diabetes and HF cases. Overall, our findings suggest that incorporating NT-proBNP testing into the routine assessment of older adults with type 2 diabetes and no symptoms of HF is cost-effective for early HF detection, supporting current guideline recommendations and broader adoption in US clinical practice.

Despite strong clinical evidence and explicit recommendations from the ADA^{2 6} and ACC/AHA,¹⁵ uptake of NT-proBNP-leveraged early detection of HF risk in the USA remains limited, leaving a large percentage of individuals underdiagnosed. Although in one study nearly half of individuals with type 2 diabetes who underwent NT-proBNP testing in clinical practice met subclinical HF thresholds, fewer than 5% of those eligible were tested.¹⁶ Clinician hesitancy—driven largely by concerns around false positive result burden or uncertainty around diagnostic thresholds in certain populations (eg, older or renally impaired patients)²⁹—combined with limited US-based economic evidence, likely contributes to restricted payer coverage and underuse of NT-proBNP in practice.³⁰

Importantly, by facilitating early initiation of appropriate cardioprotective treatment, NT-proBNP testing may reduce LY-adjusted HF hospitalizations, which are among the most resource-intensive and costly components of HF management. Cardioprotective treatment costs themselves represented only a modest proportion of total lifetime costs in this analysis, and as several of these agents, including ARNi and SGLT2i, transition to generic formulations, their lower acquisition costs are expected to reduce overall treatment expenditures. These reductions, combined with the savings associated with fewer HF-related events (ie, hospitalization and progression to chronic/advanced HF) achieved through earlier cardioprotective treatment initiation, may further improve the cost-effectiveness of NT-proBNP-guided care by lowering the incremental cost associated with starting these therapies. However, it should be noted that although therapies such as SGLT2i and GLP1-RA are strongly recommended in older adults with type 2 diabetes to reduce HF risk, large gaps remain in their application.³¹

Our findings are consistent with previous economic evaluations conducted outside the USA showing that NT-proBNP testing is cost-effective for HF detection and management within local health systems.¹⁷ For example, Moertl et al³² used a Markov model with a 20-year horizon to demonstrate that NT-proBNP testing and risk stratification to guide HF management was a cost-saving strategy for patients recently discharged after HF hospitalization compared with usual care by a primary physician or a multidisciplinary care approach involving scheduled HF nurse visits and specialist follow-up in both Canadian and Austrian healthcare systems. NT-proBNP-guided care improved survival and QALYs while lowering costs, achieving cost-effectiveness probabilities of 92% and 93% at their respective WTP thresholds. Similarly, Walter et al¹⁸ showed that NT-proBNP-guided screening improved clinical outcomes and was highly cost-effective or cost-saving in Austrian and Swiss outpatients aged ≥60 years with undetected HF, with or without type 2 diabetes. Specifically, using a lifetime Markov model, they reported ICERs of €3218 (Austria) and ₣6616 (Switzerland), and although NT-proBNP-guided screening was more costly in individuals with type 2 diabetes than in those without diabetes, it was still cost-effective relative to standard clinical assessment alone.

Our study expands on existing evidence to show that NT-proBNP testing for early HF risk detection in asymptomatic older adults with type 2 diabetes in the USA is cost-effective. These findings contrast with a recent perspective by Herman and Kuo,³⁰ which raised concerns about the ADA’s NT-proBNP testing recommendation in asymptomatic adults with diabetes,⁶ citing insufficient clinical evidence and potentially unjustifiable costs. By accounting for long-term HF prevention benefits and costs, our analysis shows that NT-proBNP-guided early HF detection provides clear economic value in this population.

This study has several strengths, including providing a comprehensive evaluation of NT-proBNP testing in older adults in the USA with type 2 diabetes receiving Medicare, a population at high risk for HF and its associated costs. The model used a lifetime horizon, incorporated a range of cost inputs, and applied scenario and sensitivity analyses to confirm robustness. Model parameters and sources used real-world evidence, where available, and were validated by two clinicians. However, certain limitations should be acknowledged. First, the model itself is not publicly available, which may limit independent replication; however, detailed reporting of model structure, assumptions, inputs, and data sources was provided to maximize transparency. Although clinical studies have shown that GLP1-RAs have cardioprotective benefit for patients with type 2 diabetes, we excluded them from the analysis because they are currently not approved as a cardioprotective treatment in HF. Including these therapies would reduce HF-related events but increase treatment costs, which could impact cost-effectiveness estimates.

Treatment effects for the prevention of HF hospitalization for individuals receiving intensified baseline treatments (eg, ACEi, ARB, ARNi, and MRA) and SGLT2i were assumed to be the same,³³ non-additive, and applied sequentially. In practice, these therapies are often used in combination, and treatment effects may interact synergistically or antagonistically. Prescribing patterns also vary between patients. These differences could influence both costs and health outcomes relative to the model estimates. By assuming non-additive effects, our approach likely underestimates the clinical benefit of combination therapy and avoids overestimating the cost-effectiveness of NT-proBNP testing.

A further limitation is that discount rates were not varied in sensitivity analyses; given the lifetime horizon of the model, alternative discount rates could affect the magnitude of discounted costs and the resulting ICER. Finally, because the model used a Medicare perspective, findings may not be fully generalizable to younger populations or healthcare settings with different treatment patterns and resource utilization. Scenario analyses extending to adults of all ages showed similar or improved clinical outcomes and lower ICERs compared with those in the base case, although the exact costs a commercial payer would pay are unknown. Although these results suggest that NT-proBNP testing remains cost-effective, absolute costs should be interpreted cautiously.

Conclusion

This study provides a comprehensive economic evaluation of routine NT-proBNP testing in a Medicare population of asymptomatic older adults with type 2 diabetes, addressing an important gap in US-focused evidence. Routine NT-proBNP testing in this high-risk population represents a cost-effective strategy for early HF detection within the framework of US Medicare costs, offering the potential to improve clinical outcomes by enabling timely initiation of cardioprotective treatment. Overall, these findings reinforce guideline-recommended use of NT-proBNP testing in US clinical practice and support broader payer coverage for routine assessment in this older adult population.

Data availability statement

No data are available.