Narcolepsy and risk of cardiovascular outcomes beyond stimulant use

Abstract

Study Objectives:

We aimed to assess associations between narcolepsy and cardiovascular disease risk while accounting for stimulant use in clinical practice.

Methods:

Using 2005–2021 MarketScan Commercial and Medicare Supplemental databases, we identified patients with newly diagnosed narcolepsy using International Classification of Diseases, Ninth or Tenth, Clinical Modification diagnosis codes, were matched at a ratio of 1:3 patients without narcolepsy or hypersomnia using propensity score matching based on baseline demographics, comorbidities, and medication use. Primary outcomes included time to first (1) composite cardiovascular disease event and (2) major adverse cardiac event. Multivariable Cox proportional hazards regression models were used to compare outcome risks between groups, following PS matching, adjusting for time-fixed and time-varying variables, including stimulant use. Individual outcomes were examined separately and analyses were stratified by age, sex, and selected comorbidities.

Results:

After PS matching, data from 134967 patients (mean [SD] age, 39.9 [16.8] years; 61.5% female) were included in final analyses. Following adjustment of baseline and time-varying covariates, patients with narcolepsy had increased risks of cardiovascular disease (adjusted hazard ratio, 1.89 [95% CI = 1.71 to 2.09]) and major adverse cardiac event (adjusted hazard ratio,1.95 [95% CI = 1.70 to 2.23]) compared with patients without narcolepsy. Results remained consistent across individual cardiovascular diseases and major adverse cardiac events. Subgroup analyses yielded similar findings.

Conclusions:

After adjusting for stimulant use, this cohort study found that patients with narcolepsy experienced increased risk of developing cardiovascular disease compared with patients without narcolepsy. This finding suggests that patients with narcolepsy may benefit from routine screening and monitoring for cardiovascular events.

Introduction

Narcolepsy, a chronic and lifelong disabling disease, presents with a range of symptoms, including excessive daytime sleepiness, sleep fragmentation, cataplexy (sudden muscle weakness triggered by emotions), and sleep hallucinations (hypnagogic or hypnopompic), alongside sleep paralysis. Narcolepsy is estimated to affect 25 to 50 individuals per 100 000 worldwide [1, 2], and the prevalence of narcolepsy in the United States was estimated to be 79.4 per 100 000 individuals between 2008 and 2010 [2]. Both types of narcolepsy—type 1 (NT1), which is caused by a deficiency of orexin (a neuropeptide that regulates sleep and wakefulness), and type 2 (NT2), which involves near-normal orexin levels—disrupt sleep and wake patterns, and recent research has unveiled a concerning association with cardiovascular disease (CVD) [1, 3–7].

A recent observational study, Cardiovascular Burden of Narcolepsy Disease (CV-BOND), showed an increased risk of myocardial infarction (MI), stroke, cardiac arrest, and heart failure (HF) among individuals with narcolepsy compared with controls. While the CV-BOND study is among the largest to date, the findings are limited due to an unbalanced control group (e.g. 34% of patients with narcolepsy had sleep apnea versus 2% in the matched control group). Additionally, the sensitivity analysis, based on narcolepsy type, revealed notable differences in baseline patient characteristics between the patients with NT2 and control groups, with NT1 analyses limited by small event sizes and wide confidence intervals (CIs) [8]. The study also did not account for medications used for narcolepsy management, including stimulants that may increase CVD risks.

Stimulants, such as methylphenidate and amphetamines, are believed to manage narcolepsy by enhancing the activity of neurotransmitters such as dopamine and norepinephrine, which are known to play a role in wakefulness [9, 10]. Although stimulants are effective in managing narcolepsy, these medications can also elevate heart rate and blood pressure [11, 12], potentially leading to increased CVD risks [13, 14]. There are mixed findings from large-scale studies regarding the association between stimulant use and CVD risks [15–19]. Still, little is known about whether narcolepsy increases CVD risks independent of stimulant use.

To address these questions and knowledge gaps, our study investigated the association between narcolepsy and CVD risk independent of other risk factors, such as comorbidities and medication use, particularly stimulants, by using a robust methodology assessing patients with employer-sponsored health plans in the US.

Methods

Study design and data source

This retrospective cohort study used data from IBM MarketScan Commercial Claims and Encounters and the Medicare Supplemental Database from January 1, 2005 to December 31, 2021, now maintained by Merative. The commercial and Medicare supplemental databases contain healthcare encounters of individuals in the United States with primary or Medicare supplemental coverage through privately insured fee-for-service, point-of-service, or capitated health plans. The University of Florida Institutional Review Board approved this study and waived the requirement for obtaining informed consent because all data used is de-identified. This study complied with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline for cohort studies.

Study population

Patients with narcolepsy were identified if they had received a diagnosis of narcolepsy (with cataplexy [NT1] or without cataplexy [NT2]) either as a primary or secondary diagnosis using International Classification of Diseases, Ninth or Tenth Revision, Clinical Modification (ICD-9-CM or ICD-10-CM) diagnosis codes (ICD-9-CM type 1: 347.01, 347.11; ICD-9-CM type 2: 347.00, 347.10; ICD-10-CM type 1: G47.411, G47.421; ICD-10-CM type 2, G47.419, G47.429). Patients were required to have at least two outpatient insurance claims within one year (narcolepsy group). At least one of these claims had to be for non-diagnostic diagnosis and could not involve diagnostic testing (i.e., multiple sleep latency test, polysomnography, or home sleep apnea test) to avoid the possibility that a narcolepsy diagnosis was recorded solely for testing purposes before results were available. (Table S1 in the Supplement). Based on age, sex, and narcolepsy first diagnosis date, we first matched patients with narcolepsy with patients without narcolepsy or hypersomnia (without narcolepsy group) at a ratio of 1:300. We excluded patients with hypersomnia to avoid inclusion of patients with potentially undiagnosed narcolepsy in the without narcolepsy control group. The index date was defined as the first narcolepsy diagnosis date for the narcolepsy group and the corresponding matched medical encounter date in the without narcolepsy group. We required patients to have at least 1 year of continuous enrollment in a health benefit plan before the index date to ensure adequate capture of baseline characteristics. Patients were excluded if they had any inpatient or outpatient claim (at any diagnosis or procedure position) for stroke, atrial fibrillation, HF, acute coronary syndrome, coronary artery bypass graft, percutaneous coronary intervention, edema, or MI during the baseline (Figure S1 and Table S2 in the Supplement).

Study outcomes

The two primary outcomes of interest were time to first occurrence of (1) CVD, including stroke, atrial fibrillation, HF, MI, or acute coronary syndrome, and (2) a major adverse cardiac event (MACE), including MI, ischemic stroke, HF, acute coronary syndrome, coronary artery bypass grafting, or percutaneous coronary intervention [8, 20]. These outcomes were defined as new onset by having any inpatient or outpatient claim of relevant diagnosis or procedure codes at the primary diagnosis (Table S2 in the Supplement). We also examined individual outcomes separately. Patients were followed up from the index date to the first occurrence of an outcome, end of continuous enrollment in the health plan, or end of the study period (December 31, 2021).

Statistical analysis

To adjust for differences in baseline characteristics between the two study groups, we used propensity score (PS) matching (1:3 ratio, without replacement) with the nearest neighbor method (0.05 caliper) [21, 22]. The PS score estimated the probability of having narcolepsy or not, conditional on baseline covariates during the 1 year prior to the index date, using multivariable logistic regression allowing for an estimation of the Average Treatment Effect in the Treated. Included covariates for PS matching were demographic variables (age, sex, type of benefit plan, and region of residence), medical condition (attention-deficit/hyperactivity disorder, anxiety, depressive disorder, diabetes, hypertension, hyperlipidemia, insomnia, kidney impairment, mood disorder, metabolic dysfunction-associated steatotic liver disease [formerly non-alcoholic fatty liver disease], obesity, periodic limb movement disorder, restless leg syndrome, sleep apnea, and substance use disorder), and medication use (benzodiazepines, Z-drugs, selective serotonin reuptake inhibitors, and serotonin and norepinephrine reuptake inhibitors). We did not include oxybate, stimulants, or wake-promoting agents in the PS models because these medications are prescribed for excessive daytime sleepiness, which is a hallmark symptom of narcolepsy, but we controlled for them in the multivariable models. Medical conditions were captured by identifying at least 1 relevant inpatient or outpatient ICD-9-CM or ICD-10-CM code, and medical procedures were determined by using ICD procedure codes, Current Procedural Terminology codes, or Healthcare Common Procedure Coding System codes (Table S3 in the Supplement). Medication use was obtained using National Drug Codes. We also assessed the baseline use of oxybate and wake-promoting agents as time-fixed covariates and stimulants (Table S4 in the Supplement) as time-fixed as well as time-varying covariates. The use of stimulants was categorized daily, with a 30-day grace period following discontinuation of the stimulant use. Differences in baseline characteristics between patient groups with or without narcolepsy were assessed with absolute standardized mean differences (ASMD), with a value of 0.10 or lower, indicating a well-balanced comparison. To ensure the robustness of CVD risk estimates for the groups with vs without narcolepsy, we conducted three Cox proportional hazards regression models with robust sandwich estimator: (1) model 1, a univariate model as all covariates were balanced after PS matching; (2) model 2, a multivariable model only including time-fixed covariates; and (3) model 3, a multivariate model including both time-fixed and time-varying covariates.

We conducted several subgroup analyses, categorized by age group, sex, narcolepsy type, and presence vs absence of selected comorbidities common among patients with narcolepsy, including sleep apnea, hyperlipidemia, hypertension, and diabetes, to evaluate heterogeneity within these groups. PS matching was performed again for each subgroup analysis. To assess the potential impact of baseline oxybate and wake-promoting agents, primarily prescribed for managing symptoms related to narcolepsy, on our findings, we conducted a sensitivity analysis excluding patients using these medications before the index date. We also reported E-values for the primary outcomes to assess the impact of unmeasured confounding. All analyses were conducted using SAS, version 9.4 (SAS Institute Inc).

Results

Study cohort and patient characteristics

We identified 34 593 patients in the narcolepsy group and 2 598 813 matched individuals in the group without narcolepsy between January 1, 2005 and December 31, 2021 (Figure S1 in the Supplement). After PS matching (1:3 ratio), we identified 34 562 patients with narcolepsy and 100 405 patients without narcolepsy. Table 1 shows demographic and clinical characteristics before and after PS matching. Following PS matching, patient demographic and clinical characteristics were well balanced between the study groups (all ASMDs ≤ 0.10) (Figure 1). After PS matching, all patients’ mean (SD) age was 39.9 (16.8) years, and 61.5% were female. The most common comorbid conditions were sleep apnea (36.8%) and mood disorders (28.7%). Nearly a quarter of the patients received selective serotonin reuptake inhibitors (24.2%), benzodiazepines (23.8%), and serotonin and norepinephrine reuptake inhibitors (20.9%) (Table 1).

Table 1.

Demographic and clinical characteristics of study population with or without narcolepsy

Characteristic	Patients, No. (%)
	Before PS matching			After PS matching
	With narcolepsy N = 34 593	Without narcolepsy N = 2 598 813	ASMD	With narcolepsy N = 34 562	Without narcolepsy N = 100 405	ASMD
Age, mean (SD), y	39.1 (16.2)	36.5 (18.2)	0.15	39.1 (16.2)	40.7 (17.3)	0.10
Sex			0.02			0.02
Male	13 023 (37.7)	1 277 614 (32.4)		13 016 (37.7)	38 958 (38.8)
Female	21 570 (62.4)	1 321 199 (50.8)		21 546 (62.3)	61 447 (61.2)
Type of health plan			0.49			0.06
Comprehensive	1800 (5.2)	166 997 (6.4)		1799 (5.2)	5714 (5.7)
EPO	316 (0.9)	8036 (0.3)		315 (0.9)	826 (0.8)
HMO	4334 (12.5)	310 670 (12.0)		4333 (12.5)	12 065 (12.0)
POS	2286 (6.6)	210 994 (8.12)		2286 (6.6)	6716 (6.7)
PPO	20,820 (60.2)	1 140 846 (43.9)		20,792 (60.2)	59 591 (59.4)
POS with capitation	231 (0.7)	11 082 (0.4)		231 (0.7)	535 (0.5)
CDHP	2856 (8.3)	529 695 (20.4)		2856 (8.3)	9277 (9.2)
HDHP	1950 (5.6)	220 194 (8.5)		1950 (5.6)	5681 (5.7)
US region			0.17			0.05
Northeast	4342 (12.6)	320 068 (12.3)		4338 (12.6)	11 883 (11.8)
North Central	10,720 (31.0)	732 260 (28.2)		10,706 (31.0)	31 657 (31.5)
South	15,169 (43.9)	1 170 827 (45.1)		15,159 (43.9)	45 131 (45.0)
West	3990 (11.5)	363 589 (14.0)		3989 (11.5)	10 814 (10.8)
Comorbid conditions
ADHD	2897 (8.4)	141 665 (5.5)	0.12	2892 (8.4)	8411 (8.4)	<0.01
Anxiety	7571 (21.9)	426 228 (16.4)	0.14	7564 (21.9)	24 067 (24.0)	0.05
Depressive disorder	7802 (22.6)	361 393 (13.9)	0.23	7785 (22.5)	24 300 (24.2)	0.04
Diabetes	3213 (9.3)	224 379 (8.6)	0.02	3212 (9.3)	10 594 (10.6)	0.04
Hypertension	6976 (20.2)	547 375 (21.1)	0.02	6975 (20.2)	23 318 (23.2)	0.07
Hyperlipidemia	7444 (21.5)	573 671 (22.1)	0.01	7440 (21.5)	24 742 (24.6)	0.07
Insomnia	3670 (10.6)	96 234 (3.7)	0.27	3658 (10.5)	11 042 (11.0)	0.01
Kidney impairment	431 (1.3)	53 047 (2.0)	0.06	431 (1.2)	1474 (1.5)	0.02
MASLD	571 (1.7)	53 947 (2.1)	0.03	571 (1.7)	1881 (1.9)	0.02
Mood disorder	9621 (27.8)	377 466 (14.5)	0.33	9601 (27.8)	29 177 (29.1)	0.03
Obesity	4282 (12.4)	309 872 (11.9)	0.01	4281 (12.4)	14 709 (14.7)	0.07
Periodic limb movement disorder	917 (2.7)	3222 (0.1)	0.26	888 (2.6)	1741 (1.7)	0.06
Restless legs syndrome	1272 (3.7)	12 052 (0.5)	0.23	1247 (3.6)	2870 (2.9)	0.04
Sleep apnea	12,981 (37.5)	117 413 (4.5)	0.89	12,950 (37.5)	36 696 (32.4)	0.02
Substance use Disorder	1190 (3.4)	96 006 (3.7)	0.01	1190 (3.4)	3729 (3.7)	0.02
Medications
SSRI	7614 (22.0)	374 542 (14.4)	0.19	7608 (22.0)	24 996 (24.9)	0.07
SNRI	6663 (19.3)	283 578 (10.9)	0.24	6656 (19.3)	21 526 (21.4)	0.05
Benzodiazepine	7586 (21.9)	412 355 (15.9)	0.16	7577 (21.9)	24 563 (24.5)	0.06
Z-drug	2879 (8.3)	127 435 (4.9)	0.14	2876 (8.3)	9311 (9.3)	0.03
Oxybate∗	64 (10.3)	1115 (8.2)	0.07	61 (0.18)	168 (0.17)	<0.01
Stimulant∗	8502 (24.6)	157 439 (6.1)	0.53	8490 (24.6)	9209 (9.2)	0.42
Wake-promoting agent∗	7143 (20.7)	7473 (0.3)	0.71	7131 (20.6)	971 (1.0)	0.67

Abbreviations: ADHD, attention-deficit/hyperactivity disorder; ASMD, absolute standardized mean difference; CDHP, consumer-directed health plan; EPO, exclusive provider organization; HMO, health maintenance organization; HDHP, high-deductible health plan; MASLD, metabolic dysfunction-associated steatotic liver disease; POS, point of service; PPO, preferred provider organization; PS, propensity score; SNRI, serotonin and norepinephrine reuptake inhibitors; SSRI, selective serotonin reuptake inhibitors. ∗Not included in PS matching but in multivariable regression models.

Figure 1.

Covariate balance before and after PS matching. Abbreviations: ADHD, attention deficit hyperactivity disorder; MASLD, metabolic dysfunction associated steatohepatitis; PS, propensity score; SSRI, selective serotonin reuptake inhibitors; SNRI, selective norepinephrine reuptake inhibitors.

Main analyses

The crude incidence rates of CVD and MACE were 0.13 and 0.06 per 100 person-years for the narcolepsy group and 0.07 and 0.04 per 100 person-years for the group without narcolepsy, respectively (Table 2). Similarly, crude incidence rates were higher in patients with narcolepsy compared with those without narcolepsy, including HF (0.04 vs. 0.03), stroke (0.04 vs. 0.02), and MI (0.02 vs. 0.01) per 100 person-years. The 3 different Cox proportional hazards regression models yielded consistent risk estimates. For example, in model 2 (a multivariable Cox model only including time-fixed covariates), patients in the narcolepsy group had an increased risk of CVD (adjusted hazard ratio [AHR], 1.78; 95% CI = 1.66 to 1.91) and MACE (AHR, 1.83; 95% CI = 1.65 to 2.02). In model 3 (a Cox model including both time-fixed and time-varying covariates), patients with narcolepsy had increased risk of CVD (AHR, 1.89; 95% CI = 1.71 to 2.09), MACE (AHR, 1.95; 95% CI = 1.70 to 2.23), HF (AHR, 1.90; 95% CI = 1.61 to 2.24), any stroke (AHR, 2.06; 95% CI = 1.73 to 2.45), and MI (AHR, 1.93; 95% CI = 1.48 to 2.51) compared with patients without narcolepsy (Table 2 and Table S5 in the Supplement). The findings from other individual components of primary outcomes were also consistent with the composite outcome analysis. The high E-values for CVD (3.19) and MACE (3.31) suggest that the association between narcolepsy and cardiovascular events is unlikely to be fully explained by a single unmeasured confounder unless it has a very strong effect (Table S5, Supplement). This supports the robustness of the findings to unmeasured confounding.

Table 2.

Risk of cardiovascular events in patients with versus without narcolepsy in PS–matched analyses

Outcome	Person-years	Events, No.	Incidence per 100 person-years	Hazard ratio (95% CI)
				Model 1∗	Model 2†	Model 3‡
CVD
Narcolepsy	95 099	1431	0.13	1.77 (1.65–1.89)	1.78 (1.66–1.91)	1.89 (1.71–2.09)
Without narcolepsy	266 161	2275	0.07	1 [Reference]
MACE
Narcolepsy	96 858	754	0.06	1.82 (1.66–1.99)	1.83 (1.65–2.02)	1.95 (1.70–2.23)
Without narcolepsy	268 487	1153	0.04	1 [Reference]
Stroke, edema, or atrial fibrillation
Narcolepsy	93 104	2050	0.18	2.12 (2.00–2.24)	2.05 (1.93–2.19)	2.02 (1.85–2.20)
Without narcolepsy	264 401	2752	0.09	1 [Reference]
HF
Narcolepsy	97 415	520	0.04	1.64 (1.47–1.83)	1.66 (1.48–1.86)	1.90 (1.61–2.24)
Without narcolepsy	268 884	878	0.03	1 [Reference]
Stroke
Narcolepsy	97 407	506	0.04	2.04 (1.82–2.29)	2.02 (1.79–2.29)	2.06 (1.73–2.45)
Without narcolepsy	269 191	689	0.02	1 [Reference]
Atrial fibrillation
Narcolepsy	97 538	435	0.04	1.58 (1.40–1.77)	1.61 (1.43–1.82)	1.66 (1.39–1.99)
Without narcolepsy	269 106	763	0.02	1 [Reference]
Ischemic stroke
Narcolepsy	97 488	464	0.04	2.07 (1.82–2.36)	2.07 (1.87–2.32)	2.11 (1.76–2.52)
Without narcolepsy	269 289	622	0.02	1 [Reference]
MI
Narcolepsy	98 144	194	0.02	1.64 (1.37–1.96)	1.68 (1.39–2.03)	1.93 (1.48–2.51)
Without narcolepsy	269 939	326	0.01	1 [Reference]

Abbreviations: CVD, cardiovascular disease; MACE, major adverse cardiac event.

^∗Time-fixed univariable Cox proportional hazards model.

^†Time-fixed multivariable Cox proportional hazards model adjusted for baseline stimulant, wake-promoting agent, and oxybate use.

^‡Time-varying multivariable Cox regression model adjusted for baseline stimulant, wake-promoting agent, and oxybate use and time-varying stimulant use.

Subgroup and sensitivity analyses

The subgroup analyses based on age group, sex, and the presence or absence of sleep apnea, hyperlipidemia, diabetes, and hypertension yielded findings consistent with the main analysis regarding risks of CVD (Figure 2), MACE (Figure 3), and stroke (Figure 4). Among patients with narcolepsy, NT2 was more common than NT1, accounting for 83.9% and 16.1% of cases, respectively. The subgroup analyses based on type of narcolepsy aligned with the main analysis i.e. patients with either type of narcolepsy had increased risk of developing CVD, MACE, and stroke. (Figures 2–4).

Figure 2.

Risk of CVD among patients with versus without narcolepsy in PS–matched subgroup analyses Abbreviations: CI, confidence interval.

Figure 3.

Risk of major cardiovascular adverse events among patients with versus without narcolepsy in PS–matched subgroup analyses. Abbreviations: CI, confidence interval.

Figure 4.

Risk of stroke among patients with versus without narcolepsy in PS–matched subgroup analyses. Abbreviations: CI, confidence interval.

The results of our sensitivity analysis, excluding patients on oxybate or wake-promoting agents before the index date, were consistent with those of the main analysis. (Table 3).

Table 3.

Risk of CVD, MACE, and stroke among patients with vs without narcolepsy: sensitivity analysis excluding patients using oxybate or wake-promoting agents during baseline period

Outcome	Patients, No.	Person-years	Events, No.	Incidence per 100 person-years	Adjusted hazard ratio (95% CI)
CVD
Narcolepsy	27,763	75,296	1061	2.08	1.76 (1.87–2.32)
Without narcolepsy	79,166	207,612	1578	0.76	1 [Reference]
MACE
Narcolepsy	27,763	76,589	561	0.73	2.22 (1.91–2.58)
Without narcolepsy	79,166	209,240	786	0.38	1 [Reference]
Stroke
Narcolepsy	27,763	76,984	380	0.49	2.38 (1.97–2.88)
Without narcolepsy	79,166	209,751	466	0.22	1 [Reference]

Abbreviations: CI, confidence interval; CVD, cardiovascular disease; MACE, major adverse cardiac event, No., number.

Discussion

This retrospective cohort study consists of the largest cohort of patients with narcolepsy, to our knowledge, that provides “real-world” clinical evidence on the association of narcolepsy with the incidence of CVD. Our comprehensive risk assessments included PS-matching and adjusted for time-varying use of stimulants following the diagnosis of narcolepsy in addition to controlling for the use of stimulants, oxybate, and wake-promoting agents during the baseline period. The results consistently indicated that patients with narcolepsy demonstrated a significantly greater risk of CVD and MACE compared to patients without narcolepsy.

The association between narcolepsy and CVD has garnered substantial interest in recent years—in part due to the high prevalence of cardiovascular risk factors among people with narcolepsy (e.g. obesity, obstructive sleep apnea, sleep fragmentation, and absence of nocturnal blood pressure dipping [23, 24]), as well as a potential association between orexin deficiency in NT1 cardiovascular dysregulation, further increasing CVD risk [25–29]. While some studies have examined cardiovascular physiological changes in narcolepsy (e.g. heart rate variability, blood pressure control) [7, 25], only a few population-based studies have examined this association. In a cross-sectional study, Ohayon found that patients with narcolepsy were at 2 times increased odds of having heart disease and 32% increased odds of having hypertension, among other chronic conditions, compared with a general population sample [30]. In their study using data from the UK Biobank, Tao et al.found that a greater frequency of self-reported narcolepsy-type symptoms (e.g. unintentionally “dozing off” or falling asleep during daytime activities) was an independent risk factor for CVD after adjusting for confounders, among other sleep characteristics [3].

The CV-BOND study, using the same MarketScan database as ours, aligned with our findings and showed increased risks for several CVD outcomes, including the narcolepsy type. However, the CVD risks observed in our study (e.g. AHR, 1.89) were consistently higher than those reported in the CV-BOND study (e.g. AHR, 1.30) [8]. The differences in the point estimates of CVD risks may be attributed to several differing aspects of study design between the CV-BOND study and ours. First, although CV-BOND adjusted for clinical characteristics in the final models, the two groups were not balanced in terms of these characteristics (e.g. sleep apnea, hyperlipidemia, anxiety, mood disorders, and hypertension). By contrast, we ensured a well-balanced cohort between patients with or without narcolepsy through a rigorous two-step matching process, which included PS matching. This approach enabled us to achieve a well-balanced baseline with regard to risk factors known to be associated with increased risk of CVD and more commonly occurring in people with narcolepsy [31]. Additionally, we controlled for medication use at baseline and stimulant use as a time-varying factor. Our study spanned a broader time frame (from 2005 to 2021) than CV-BOND. We also included individuals without an age limit, thus encompassing patients who developed narcolepsy during adolescence, which is common and excluded all baseline CVD to ensure the examination of incident CVD following a diagnosis of narcolepsy. The higher NT2 proportion in our study (83% vs. 17%) than in CV-BOND (77% vs. 23%) may stem from differences in study populations and periods. However, these methodological and inclusion criteria differences further emphasize the robustness and generalizability of our findings concerning the association between narcolepsy and CVD in real-world clinical practice.

The underlying mechanisms linking narcolepsy to CVD are not yet understood, but several factors may contribute to this association. Orexin knock-out and orexin-neuron deficient mouse models [32] as well as patients with NT1, have a blunted sleep-related decrease in blood pressure [ 33], which has been traced down to insufficient decrease of cardiovascular sympathetic activity during sleep in orexin knock-out mice [34]. However, orexin deficiency does not fully explain the elevated CVD risk in narcolepsy, as we observed no differences in NT1 and NT2 [35]. Sleep fragmentation, a characteristic of narcolepsy, disrupts sleep quality and leads to sympathetic nervous system activation, potentially contributing to hypertension and other cardiovascular complications in both types of narcolepsy [28, 29, 36]. Furthermore, excessive daytime sleepiness, a unifying feature of both narcolepsy types, has been found to be more strongly associated with CVD risk than the apnea-hypopnea index in obstructive sleep apnea [37]. This potentially highlights that daytime sleepiness is most relevant to CVD risk associated with sleep disorders.

Additionally, the presence of comorbidities such as obesity, diabetes, and sleep apnea, which are often associated with narcolepsy, can further exacerbate the risk of CVD [8, 38–44]. However, our present analyses using causal inference methods suggested an association between narcolepsy and CVD even after controlling for traditional CVD risk factors [45, 46]. A recent Mendelian randomization study further supports the notion of a causal relationship between narcolepsy and CVD [5].

The relationship between sleep disturbances and cardiovascular risk is multifaceted, involving mechanisms such as autonomic dysregulation (e.g., sympathoexcitation), oxidative stress, glucose and endothelial dysfunction, sedentary behavior, and inflammation [36, 47–56]. While it is conceivable that medications used to treat narcolepsy could influence this risk, our analysis indicates that the observed association between narcolepsy and CVD is independent of medication use, particularly stimulants. This finding is significant, given that amphetamine-derived agents and sodium oxybate have been linked to increased cardiovascular risk through distinct mechanisms—sympathomimetic stimulation for amphetamines and association with hypertension for sodium oxybate [14, 57, 58].

Strengths and limitations

Our study has several notable strengths. First, we used a large administrative database covering 17 years (2005 to 2021), providing a large sample of representative patients with narcolepsy and enabling us to perform numerous subgroup and sensitivity analyses. Our large narcolepsy sample represented a significant portion of nearly 0.05% of the population affected by narcolepsy in the United States [59]. Second, our study addressed methodological issues in prior epidemiological investigations, including confounding and selection bias, by employing rigorous causal inference methods, such as PS matching and time-varying Cox regression modeling. This combined approach of matching and time-dependent modeling strengthened the comparison between narcolepsy patients and matched individuals without narcolepsy, strengthening the evidence that observed differences in incident cardiovascular outcomes were attributable to narcolepsy, rather than confounding health or demographic factors between the groups.

Several limitations of this analysis should be noted. First, we lacked specific laboratory values to confirm narcolepsy diagnoses or distinguish between narcolepsy types. However, we used a combination of diagnostic and test codes to identify patients with narcolepsy (NT1 or NT2), as was done in prior studies [7]. Despite these efforts, it is possible that some individuals classified as being without narcolepsy may have had undiagnosed narcolepsy. To address this possibility, we excluded patients with hypersomnia from the group without narcolepsy to reduce selection bias. Additionally, we conducted a sensitivity analysis excluding patients using oxybate, a narcolepsy-specific drug, or wake-promoting agents before the index date. This was done to account for possibility, though minimal (<0.2%), that underdiagnosed narcolepsy patients may have been included in the control group. While our study benefited from a large sample size that enabled multiple subgroup analyses, the smaller sample size of NT1 patients compared to NT2 may have limited the statistical power to detect significant differences between these two groups. There exists a potential of overestimating diagnoses of NT2 in younger patients (<18 years) due to the difficulties in identifying cataplexy and the delayed onset of cataplexy in comparison to the onset of excessive daytime sleepiness. Second, although we extensively adjusted for known covariates associated with cardiovascular outcomes, potential unmeasured or unreported confounders (e.g. race and ethnicity, smoking status) within the administrative data could still have influenced the results. Third, our study utilized data from 2005 to 2021, during which diagnostic technologies and criteria for CVD have advanced over time. Nevertheless, since the narcolepsy group was matched with the control group by encounter date, these changes are unlikely to have differentially impacted the observed CVD outcomes between the two groups. Finally, it is plausible that incomplete, missing, or miscoded claims influenced the study findings since we lacked laboratory results to confirm CVD-related diagnoses; however, we used validated and previously used algorithms to identify study outcomes.

This cohort study provided evidence that narcolepsy was associated with notably higher increased risks of CVD and MACE than those among patients without narcolepsy, associations that remained significant in subgroup and sensitivity analyses. The analyses indicated that individuals with narcolepsy experienced a higher incidence of new-onset cardiovascular events compared with PS-matched individuals without narcolepsy, even after adjusting for medication use, specifically stimulants. Therefore, study findings suggest that patients with narcolepsy should undergo routine screening for CVD and receive monitoring as measures to mitigate the potential risk of CVD.

Supplementary Material

Supplementary material is available at SLEEP online.

Statement of significance

This study enhances our understanding of the association between narcolepsy and cardiovascular events by employing a robust causal inference framework and adjusting for medications used to manage narcolepsy symptoms, which are known to be associated with cardiovascular risk. The large population-based cohort study included 34 562 patients with narcolepsy and 100 405 propensity score-matched patients without narcolepsy. Patients with narcolepsy had an increased risk of developing cardiovascular events compared to without narcolepsy controls, even after adjusting for medication use, including stimulants. These findings highlight the importance of routine cardiovascular screening and monitoring for patients with narcolepsy to mitigate potential risks.

Data sharing statement

MarketScan Commercial and Medicare Supplemental databases are available through a data use agreement with IBM Watson Health (now Merative). For further inquiries or to request access to the data, please contact IBM Watson Health directly by visiting their website at https://www.ibm.com/watson/health/resources/ipv-opv/.