This retrospective cohort investigation was conducted using the TriNetX Research Network (TriNetX, LLC, Cambridge, MA, USA), a federated electronic health record database. The TriNetX database has been widely utilized and cited in numerous peer-reviewed observational studies across diverse clinical domains (20–22). In this study, we sought to determine whether VDD is associated with the development of new-onset heart failure in adults with OSA, thereby clarifying whether a low vitamin D status identifies a subgroup of OSA patients at particularly elevated cardiovascular risk. The research protocol was approved by the Institutional Review Board of Chi Mei Medical Center, which waived the requirement for informed consent because only de-identified data were used.

The study population comprised adults with a previous diagnosis of OSA. OSA was identified using the International Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM) code G47.33, with at least one documented diagnosis recorded prior to or on the index date in the TriNetX database. Patients were classified into two cohorts based on their serum 25-hydroxyvitamin D (25 [OH] D) levels. The exposure cohort consisted of patients with vitamin D deficiency (VDD), defined as a serum 25(OH)D level of < 20 ng/mL. The control cohort comprised patients with sufficient vitamin D status, operationalized as a serum 25(OH)D concentration ≥30 ng/mL, and without any documented measurement in the deficient range during the baseline period. To ensure persistent vitamin D status throughout the observation period, patients in the deficiency cohort were required to have at least one additional measurement confirming levels below 20 ng/mL within 3 months to 5 years after the index date, with no subsequent measurements reaching 20 ng/mL or above. Similarly, individuals assigned to the control cohort, defined as levels ≥30 ng/mL, were required to consistently exhibit 25(OH)D concentrations ≥30 ng/mL with no recorded values below 30 ng/mL during the defined assessment window.

The index date for each participant was designated as the earliest vitamin D measurement that fulfilled the eligibility criteria for the respective cohort. 25(OH)D levels were obtained from laboratory records available in the TriNetX database. As TriNetX aggregates de-identified electronic health record data from multiple healthcare organizations, detailed information regarding assay methods, laboratory platforms, or calibration procedures used to measure 25(OH)D was not available. Therefore, vitamin D status in this study was defined based on recorded serum 25(OH)D values rather than specific analytical techniques. To ensure adequate follow-up, patients were required to have at least one documented healthcare visit within 3 months to 5 years of the index date.

Patients were excluded if they were diagnosed with primary or secondary pulmonary hypertension at baseline or within 3 months after the index date. To minimize the influence of acute illness on vitamin D levels, patients with acute kidney failure, sepsis, COVID-19, or ICU admission were excluded. We excluded individuals with advanced chronic kidney disease because severe renal impairment profoundly alters the vitamin D metabolism. Additionally, patients with pre-existing conditions that could confound the relationship between vitamin D status and cardiovascular outcomes, including heart failure, pulmonary embolism, chronic pulmonary embolism, congenital malformations of the circulatory system, and human immunodeficiency virus infection, were excluded.

Baseline covariates included demographic characteristics (age, sex, and race), lifestyle factors (e.g., nicotine dependence), cardiovascular and metabolic comorbidities (e.g., ischemic heart disease and hypertension), chronic organ dysfunction (e.g., chronic kidney disease), anemia, neoplasms, and long-term steroid use. In addition, the use of continuous positive airway pressure (CPAP) for OSA and cardiometabolic medications (e.g., glucagon-like peptide-1 receptor agonists, sodium–glucose cotransporter 2 inhibitors) as well as laboratory markers (e.g., albumin, hemoglobin A1c) were incorporated into the propensity score model (Supplementary Table 1). The use of CPAP therapy was identified based on relevant prescription and procedure records within the database. Detailed information on CPAP adherence, including nightly usage duration or compliance metrics, was not available on the TriNetX platform; therefore, CPAP exposure in this study reflects documented treatment initiation rather than verified long-term adherence. Propensity score matching was performed in a 1:1 ratio between the VDD and control cohorts based on all listed characteristics. One-to-one greedy nearest-neighbor matching without replacement was performed using a caliper width of 0.1 standard deviations of the logit of the propensity score. Covariate balance between cohorts was evaluated using standardized mean differences, with values < 0.1 indicating adequate balance.

The primary outcome was incident heart failure, while secondary outcomes included secondary pulmonary hypertension, primary pulmonary hypertension, pulmonary embolism, and all-cause mortality at the 5-year follow-up. Outcomes were assessed during follow-up windows beginning 90 days after the index date and extending to 3 and 5 years, allowing for a latency period between exposure and outcome development while enabling the evaluation of both medium-term and long-term associations.

To evaluate the potential dose-response relationships between vitamin D status and adverse outcomes, an additional analysis was conducted to compare patients with vitamin D insufficiency against the control cohort. Vitamin D insufficiency was defined as serum 25(OH)D levels between 20 and 29.9 ng/mL. This intermediate-exposure cohort was then compared with patients who consistently had sufficient vitamin D levels by applying the same matching approach and outcome evaluation strategy over a 5-year follow-up period. This study aimed to determine whether a graded association exists between declining vitamin D levels and the risk of heart failure.

Patients were stratified by sex (male vs. female), age (18–50 years vs. >50 years), and the presence or absence of hypertension, obesity, cancer, hyperlipidemia, and diabetes mellitus. Within each stratum, new propensity score matching was carried out to achieve covariate balance between the vitamin D–deficient and control cohorts. Effect modification was evaluated by introducing interaction terms, and p-values for these interactions were reported to indicate whether the relationship between VDD and heart failure differed across subgroups.

Kaplan–Meier survival analysis was performed to estimate the cumulative incidence of each outcome across predefined follow-up periods, with time zero defined as the index date. Patients were censored at the time of the last available follow-up, death (for non-mortality outcomes), or the end of the observation window, whichever occurred first. Between-group differences in the survival distribution were evaluated using the log-rank test. Hazard ratios (HRs) with corresponding 95% confidence intervals (CIs) were calculated using Cox proportional hazard regression models. The proportional hazards assumption was assessed by inspecting Schoenfeld residuals and testing for time-dependent effects. All time-to-event analyses were conducted on propensity score-matched cohorts to reduce confounding by baseline imbalances. A two-sided p-value of less than 0.05 was considered statistically significant.

Using TriNetX, we identified 43,492 adults with OSA who had VDD and 136,108 adults with sufficient vitamin D levels that satisfied all predefined inclusion and exclusion criteria (Figure 1). After applying propensity score matching in a 1:1 ratio, 36,497 patients remained in each cohort for analysis. Before matching, there were notable differences between cohorts in age (46.9 ± 17.0 vs. 59.5 ± 14.3 years), racial distribution (54.6% vs. 81.6% White), and several comorbidities (e.g., hypertension, hyperlipidemia, nicotine dependence) (Table 1). Following propensity score matching, all baseline characteristics achieved excellent balance. The matched cohorts were comparable in mean age (50.1 ± 15.6 vs. 50.0 ± 15.8 years), sex distribution (52.8% vs. 52.6% female), obesity (BMI ≥ 30 kg/m²) prevalence (61.9% vs. 62.4%), and rates of hypertension, diabetes mellitus, chronic kidney disease, and other relevant comorbidities. Medication use and CPAP therapy were also well balanced between the groups.

During a mean follow-up of approximately 4 years (1,420 vs. 1,465 days in the VDD and control cohorts, respectively), VDD was significantly associated with higher rates of incident heart failure than the control group (7.7% vs. 5.6%; HR 1.45; 95% CI 1.37–1.53; p < 0.001; Table 2). All-cause mortality was also significantly higher among patients with VDD (HR 1.76; 95% CI 1.64–1.89; p < 0.001). With respect to other secondary outcomes, VDD demonstrated significant associations with secondary pulmonary hypertension (HR 1.25; 95% CI 1.13–1.38; p < 0.001) and pulmonary embolism (HR 1.31; 95% CI 1.17–1.47; p < 0.001). However, no statistically significant association was observed between VDD and primary pulmonary hypertension (HR 1.22; 95% CI 0.84–1.79; p = 0.300).

At the 3-year follow-up, the overall pattern of results remained unchanged (Table 3). VDD continued to be associated with higher incidences of heart failure (HR 1.50; 95% CI 1.40–1.61; p < 0.001) and all-cause mortality (HR 1.89; 95% CI 1.73–2.06; p < 0.001), as shown in Table 3. Elevated risks were also noted for secondary pulmonary hypertension (HR 1.22; 95% CI 1.09–1.38; p < 0.001) and pulmonary embolism (HR 1.41; 95% CI 1.23–1.61; p < 0.001), both remaining statistically significant at this time point. Primary pulmonary hypertension did not show a significant association with VDD at three-year follow-up (HR 1.15; 95% CI 0.74–1.79; p = 0.541).

To evaluate potential dose-response relationships, we compared patients with vitamin D insufficiency (25(OH)D 20–29.9 ng/mL) against those with sufficient vitamin D levels (Table 4). After propensity score matching (n = 51,582 per group), vitamin D insufficiency was associated with modestly elevated risks of heart failure (HR 1.15; 95% CI 1.09–1.21; p < 0.001), all-cause mortality (HR 1.36; 95% CI 1.28–1.44; p < 0.001), and pulmonary embolism (HR 1.15; 95% CI 1.04–1.27; p = 0.008) at five-year follow-up. Notably, the HRs in the insufficiency group were lower than those in the deficiency group, indicating a stepwise increase in cardiopulmonary risk as the vitamin D concentration declined.

In a priori–defined subgroup analyses, the relationship between VDD and new-onset heart failure over the 5-year observation period was generally consistent across the various clinical strata (Figure 2). The association remained significant regardless of sex, age category, and the presence or absence of hypertension, cancer, hyperlipidemia, or diabetes mellitus, with no significant interaction effects observed for these variables. However, a significant interaction was identified for obesity status (p for interaction = 0.028), with obese patients showing a stronger association between VDD and heart failure (HR 1.55; 95% CI 1.44–1.67) than non-obese patients (HR 1.36; 95% CI 1.24–1.49).