Results from this retrospective observational study of patients with ATTR-CM in the USA using data from the Komodo Healthcare Map demonstrate that many tafamidis-treated patients with ATTR-CM still experience adverse disease outcomes in real-world practice. While both clinical trial and real-world evidence have shown benefits of tafamidis treatment, this analysis of a large, diverse US cohort provides insight into rates of hospitalization and disease outcomes that warrant further examination. Within 6 months of initiating tafamidis, the cumulative incidence of first CVH was 22%, increasing to 35% by 12 months. Similarly, the 12-month cumulative incidence of OWHF with oral diuretic intensification was 33% and mortality occurred in approximately 6% within the first year of therapy. Within 6 months of initiating tafamidis, the cumulative incidence of the composite endpoint event (CVH, OWHF, or death) was 37% and 53% within 12 months; median time to first composite endpoint event was 10.4 months (315 days). Across these three key outcomes and their composite endpoint, stratification by ATTR-CM type revealed no notable differences. As expected, patients with elevated NT-proBNP (> 3000 pg/mL, or BNP > 600 pg/mL) at baseline demonstrated markedly worse outcomes than those with low NT-proBNP in the descriptive analysis, with the cumulative incidence of hospitalizations nearly twice as high and mortality rates reaching 9.7% at 12 months. This finding underscores the prognostic value of NT-proBNP/BNP in the risk stratification of patients with ATTR-CM. Additionally, patients with baseline arrhythmias demonstrated increased hazard for CVH and the composite endpoint, which aligns with general understanding that this comorbidity is associated with poorer outcomes. Regarding baseline medications, SGLT2 use was associated with reduced hazard for CVH, while MRA use was associated with increased hazard, likely reflecting residual confounding by disease severity, as MRA use in real-world settings is often concentrated in patients with more advanced HF [20, 21]. However, it is important to note that in a real-world study examining outcomes of MRAs in elderly patients with HF with reduced ejection fraction and moderately impaired renal function, after adjusting for worsening renal function, MRA use did not increase the overall risk of all-cause mortality [22]. Our analysis was not able to accurately capture potassium dosing and blood levels, which may have affected outcomes.

The demographic characteristics observed in this cohort align with established patterns in ATTR-CM epidemiology. Patients with ATTRv-CM have a variable age of onset (30‒80 years) that is dependent on the mutation, while ATTRwt-CM typically presents around 75 years of age [3]. In two retrospective studies of patients diagnosed with ATTR-CM, approximately 72‒80% of Black or African American patients had ATTRv-CM [23, 24]. The results of our study are consistent, as we report that patients with ATTRv-CM were younger than patients with ATTRwt-CM (mean 72.3 years vs. 77.8 years) and that a larger proportion of patients with ATTRv-CM were Black or African American compared with ATTRwt-CM (55.0% vs. 27.5%). These demographic differences between ATTR-CM subtypes underscore the importance of genetic testing and tailored screening approaches across diverse populations. When we compared this real-world patient population to those of randomized controlled trials (RCTs) for ATTR-CM [8, 25, 26], patients in the Komodo database were similarly aged; however, the database included more women (20% vs. 8‒10%) and Black or African American patients (30% vs. 5‒15%) compared with the RCTs (Supplementary Material Table S5). This aligns with a lack of representation of women and minorities in RCTs, as observed in other publications [27, 28], and highlights a benefit of using real-world data to evaluate groups typically underrepresented. A direct comparison of outcomes between RCTs and the current analysis is not feasible given differences in study design and endpoint selection, and the lack of published data.

Non-invasive diagnosis of ATTR-CM has led to improved recognition at earlier stages, with improvements in survival in the placebo arms of amyloid trials over the past decade [11]. In this context, recent studies reporting real-world outcomes for tafamidis-treated patients with ATTR-CM provide valuable benchmarks for comparison. Previous observational studies have reported survival rates in tafamidis-treated patients of 89.3% at 12 months [13], 72.5–84.4% at 30 months [11, 29], and 76.8% at 42 months [11]; the current analysis found 93.8% survival at 12 months. While differences in study design, patient populations, and the later period of data capture in the present analysis may explain some variation [11, 13, 29], focusing primarily on overall survival may underestimate the complete burden of disease in patients treated with tafamidis.

The patterns of other disease outcomes observed in the current analysis support the generalizability of prior findings in tafamidis-treated patients to settings beyond specialized amyloidosis centers, while revealing potentially higher event rates in broader real-world populations. In a multicenter study from Italy of 683 patients treated with tafamidis, the composite endpoint of death or HF hospitalization requiring intravenous diuretics occurred in 9% of patients with ATTR-CM within 12 months of tafamidis initiation [30]; this analysis was limited to patients with New York Heart Association functional class I and II symptoms, as the Italian Medicines Agency restricts tafamidis use to this subgroup. In a single-center analysis of 238 tafamidis-treated patients with ATTR-CM who were followed for 12 months, 21% had oral diuretic intensification and 11% had NT-proBNP level increases of > 700 pg/mL and > 30% from baseline [14], compared with approximately 30% experiencing OWHF with oral diuretic intensification and 15% having the same threshold of NT-proBNP increases in the present analysis. Similarly, Zeldin et al. reported in a single-center retrospective analysis of 303 tafamidis-treated patients with ATTR-CM (mean follow-up of 3.3 years) that 11.6% experienced hospitalization for HF and 47.9% experienced worsening HF with outpatient diuretic intensification (defined as any initiation or increase in total daily dose of loop diuretic lasting > 30 days) [31]. Despite a shorter follow-up period of 12 months in the current study, the cumulative incidences of first HF-related hospitalizations and OWHF with oral diuretic intensification at 12 months were 30% and 33%, respectively. These findings confirm that worsening cardiac disease despite tafamidis treatment occurs across healthcare settings, while suggesting that patients in broader real-world populations may experience some disease outcomes at higher rates than those treated in specialized care environments.

Even with the treatment benefits of tafamidis in ATTR-CM, these findings highlight the importance of recognizing the risk of adverse disease outcomes such as CVH and worsening HF, particularly those with elevated NT-proBNP/BNP at baseline. Our analysis confirmed that certain clinical factors—including high baseline NT-proBNP, older age, presence of HF, chronic kidney disease, and other comorbidities—significantly increase the hazard for adverse outcomes across multiple endpoints. These identified risk factors provide prognostic information that may inform cardiac disease monitoring in tafamidis-treated patients, though their practical utility in guiding management strategies requires prospective validation. It should be noted that while earlier identification and treatment of patients can lead to improved outcomes [5, 33, 34], this is especially important for patients with high-risk disease. Moreover, while the current analysis cannot determine whether alternative therapeutic approaches may benefit patients receiving tafamidis, it underscores the need for additional studies in this area.

The primary strength of this study is its use of a large, nationally representative real-world database with tokenization to laboratory data, allowing examination of tafamidis outcomes across a diverse ATTR-CM population. Notably, 20% of patients in this analysis were female, which exceeds the proportion of female patients in pivotal clinical trials; women comprised only 8.7% in the ATTR-ACT tafamidis arm and 8.8% in the ATTRibute-CM acoramidis arm (Supplementary Material Table S5) [8, 25]. The sex distribution of this cohort—with male predominance aligning with known ATTR-CM epidemiology but including a larger proportion of female patients—may better reflect real-world practice and enhance generalizability to both sexes.

Several limitations warrant consideration when interpreting these findings. First, this analysis shares limitations intrinsic to medical claims database studies and naturally differs from RCTs in the type of data captured and reported [32]. The study did not account for disease duration before tafamidis initiation, which likely varies among patients and could influence observed outcomes, as newly diagnosed patients may respond differently than those with longer duration or advanced stage of the disease. Additionally, the study lacked clinical staging, functional assessments, and quality of life measurements that would provide a more comprehensive understanding of patient outcomes beyond hospitalization rates and biomarker data.

Data representation limitations include an overrepresentation of patients from the Northeast region (42.8%) with fewer from the West (9.9%), likely because large insurance providers prevalent in Western states (such as Kaiser Permanente) are not captured within the Komodo database. Patients may have had multiple insurers or received tafamidis through patient assistance programs, potentially resulting in misclassification of some treated patients as untreated. Furthermore, the Komodo Healthcare Map includes only individuals who have interacted with the medical system or have insurance coverage, potentially limiting representativeness of the entire ATTR-CM population. This study was conducted during a period that included the COVID-19 pandemic, which substantially disrupted routine healthcare delivery and utilization. During this time, when healthcare encounters were delayed or deferred, the initiation of new tafamidis use may have been reduced during the early months of the pandemic. As a result, some time-to-event outcomes may have been longer, and fewer patients may have initiated tafamidis during the pandemic years. However, because the current analysis spans both prepandemic and postpandemic years, it is expected that most healthcare encounters and patients initiating tafamidis were captured in this analysis.

Methodologic limitations include the lack of genomic data and the small proportion of patients with laboratory data available, restricting NT-proBNP stratification analyses and the ability to classify disease severity at baseline. Additionally, more patients ≥ 80 years of age had elevated NT-proBNP rather than low NT-proBNP; increased CVHs and mortality are inherently higher in older patients [29], potentially confounding the observed associations between NT-proBNP levels and outcomes. The study also relied on ICD-10-CM coding algorithms to classify ATTR-CM subtypes, which may have resulted in misclassification bias. The high proportion of CVH and HF-related hospitalization relative to all-cause hospitalization reflects the use of ICD-10-CM diagnosis codes in any position rather than the primary position only. This approach, while sensitive for capturing cardiovascular events, may overestimate hospitalizations directly caused by cardiovascular disease, as CV diagnoses once established are routinely coded on subsequent admissions, regardless of the primary indication for admission.