National trends in cardiovascular deaths with concurrent accidental drug poisoning coding in the United States: a 22-year analysis of the CDC WONDER database (1999–2020)

Abstract

Objective

While the overdose epidemic garners significant public health attention, cardiovascular deaths with concurrent accidental drug poisoning coding remain understudied in the United States.Traditional drug mortality surveillance primarily focuses on direct overdose fatalities, with limited integration of modern CDC-aligned multi-cause coding frameworks and atherosclerotic cardiovascular disease (ASCVD) cardiotoxicity mechanisms.We examined national trends in this phenomenon from 1999 to 2020.

Methods

We analyzed CDC WONDER data, identifying deaths where cardiovascular disease (ICD-10: I00-I99) was the underlying cause and accidental drug poisoning (ICD-10: X40-X44) was documented as a contributory cause on death certificates. Age-adjusted mortality rates (AAMR) were calculated, and trends were assessed using Joinpoint regression analysis.This study is explicitly framed as an ecological population-level surveillance analysis of death certificate co-coding patterns, rather than an assessment of individual-level drug user cardiovascular mortality risk, given the inherent general-population denominator limitation of CDC WONDER aggregate data.

Results

We identified 33,536 deaths nationwide meeting these criteria. The AAMR nearly doubled from 0.500 to 0.979 per 100,000 population (annual percent change: +3.64%, P < 0.001). Three distinct phases emerged: rapid increase (1999–2003, + 9.746%,P < 0.001), plateau (2003–2014, + 0.843%,P = 0.188), and second increase (2014–2020, + 4.887%,P < 0.001). Significant disparities existed: mortality was 2.48 times higher in men than women, and highest among non-Hispanic Black individuals (AAMR = 1.451, 2.25 times that of non-Hispanic Whites). The West region had the highest regional mortality (AAMR = 1.165), with Nevada recording the highest state-level rate (AAMR = 3.341). Middle-aged adults(45–64 years) accounted for 60.02% of all deaths.

Conclusion

Cardiovascular deaths with concurrent drug poisoning coding rose significantly from 1999 to 2020 in the United States, with the age-adjusted mortality rate (AAMR) increasing from 0.50 to 0.98 per 100,000 population. The trend followed three phases: a rapid rise from 1999 to 2003, a period of relative stabilization from 2003 to 2014, and another rapid increase from 2014 to 2020. Disparities were substantial and uneven: in 2020, mortality in men was approximately 2.5 times higher than in women; the highest AAMR was among non-Hispanic Black or African American individuals, and the West region carried a significantly higher risk than other regions; the 45–64 age group had the highest mortality rate, while the 25–44 age group showed a fluctuating pattern. These findings underscore the need for strengthened cardiovascular assessment and monitoring within substance-use-disorder interventions and for tailored public health strategies targeting high-risk populations and regions. At the same time, interpretations should consider limitations of death certificate coding, including potential differences in surveillance intensity and misclassification across demographic and geographic groups.All population-level trends reflect coded mortality patterns and cannot be extrapolated to individual drug-user risk, consistent with modern CDC multi-cause surveillance standards.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12872-026-06351-1.

Introduction

Drug poisoning, particularly opioid overdose, has evolved into a devastating global public health crisis, characterized by a significant rise in morbidity and mortality that has negatively impacted life expectancy in multiple countries worldwide [1–3]. Reports from the U.S. Centers for Disease Control and Prevention (CDC) indicate that over the past two decades, drug poisoning-related deaths have increased exponentially, forming the core of the “opioid epidemic” [4, 5]. Although public and research attention has largely focused on direct respiratory depression and fatalities resulting from drug intoxication, growing evidence suggests that the cardiovascular system is another critical target of drug toxicity, with consequences that are equally lethal yet potentially underrecognized [6, 7].

The cardiovascular toxicity of various substances of abuse has been well-documented. Stimulants such as cocaine and methamphetamine can significantly increase the risk of acute myocardial infarction, stroke, and sudden cardiac death by inducing a catecholamine storm, causing coronary vasospasm, promoting thrombosis, and triggering malignant arrhythmias and hypertensive emergencies [8–11]. In contrast, opioids primarily mediate respiratory depression via central nervous system µ-receptors, but their use is also associated with significant cardiovascular complications, including hypotension, bradycardia, and secondary myocardial injury and cardiac arrest with respiratory failure and hypoxia [12–14]. Furthermore, injection drug use involving substances such as heroin and cocaine is a major risk factor for infective endocarditis, which can lead to valvular destruction, heart failure, and embolic events [15, 16]. Thus, the impact of drug poisoning on mortality is multifaceted, encompassing both direct toxic effects and indirect cardiovascular sequelae.

Contemporary disease surveillance frameworks emphasize the critical value of multi-cause death certificate coding for capturing underrecognized epidemic burden, moving beyond single-cause mortality statistics to characterize complex, comorbid fatal outcomes [17]. This updated CDC-aligned surveillance paradigm addresses longstanding limitations in traditional overdose mortality research, which systematically underestimates drug-related mortality burden by excluding contributory drug poisoning coding on cardiovascular fatalities [17].Despite understanding these pathophysiological mechanisms, most surveillance and public health interventions remain focused on all-cause drug poisoning mortality. Comprehensive analyses specifically examining long-term, nationwide trends in cardiovascular mortality with drug poisoning remain relatively scarce. Understanding these trends is essential for accurately assessing the true burden of the drug epidemic, designing targeted prevention strategies, and optimizing the allocation of healthcare resources [18, 19]. Existing studies are often limited to specific substances, cardiovascular outcomes, or geographic regions, lacking a broader perspective that can reveal evolving patterns over time [20, 21].

This study aims to conduct a comprehensive retrospective analysis of cardiovascular mortality with drug poisoning in the United States from 1999 to 2020—a 22-year period using the CDC WONDER database, a authoritative national mortality surveillance resource.Our specific objectives are: (1) to describe national trends and age-adjusted mortality rates of cardiovascular deaths with concurrent drug poisoning coding (CVM-DP) in the U.S. from 1999 to 2020; and (2) to analyze mortality disparities and temporal changes stratified by key demographic variables including sex, age, ethnicity and ethnicity, and geographic region (e.g., census region).Consistent with modern epidemiological surveillance standards, this study is explicitly designed as a population-level coding pattern analysis, not individual-level risk estimation, to resolve ambiguity between true disease burden and surveillance coding artifacts [17].We hypothesize that, against the backdrop of rising overall drug poisoning mortality, CVM-DP has also exhibited a significant upward trend, with the burden varying substantially across demographic subgroups and types of substances involved. The findings of this study are intended to provide clinicians, public health practitioners, and policymakers with critical evidence to enhance the monitoring, prevention, and management of cardiovascular complications related to substance use, thereby supporting more precise and effective measures in addressing the ongoing drug epidemic.In addition to the clinical and epidemiological motivations for this study, understanding trends in CVM-DP coding patterns is important because observed variations across demographic and geographic groups may reflect not only true epidemiological risk differences but also differential surveillance intensity in how drug involvement is detected and coded on death certificates across populations.

Methods

Study design and data sources

We conducted a repeated cross-sectional analysis using data from the United States National Mortality Database. Mortality data were obtained from the CDC WONDER Underlying Cause of Death database, covering the period from 1999 to 2020. This publicly available database contains death and population counts derived from death certificates from all 50 U.S. states and the District of Columbia.CDC WONDER public aggregate data do not contain individual-level exposure, socioeconomic, or granular clinical covariate information, representing a core inherent constraint of this national surveillance resource.As this study utilized publicly accessible, de-identified aggregate data, it was exempt from Institutional Review Board (IRB) approval.

Case identification and definition

We identified deaths meeting the following criteria: Underlying Cause of Death (UCOD) was cardiovascular disease (ICD-10 codes I00-I99), and Multiple Causes of Death (MCoD) included accidental poisoning by and exposure to drugs, medicaments, and biological substances(ICD-10 codes X40-X44).

The ICD-10 I00–I99 cardiovascular disease spectrum captured in this study fully encompasses acute and chronic atherosclerotic cardiovascular disease (ASCVD), arrhythmic disorders, and cardiomyopathic pathologies, aligning with contemporary cardiotoxicity epidemiological frameworks [22]. Specifically, this coding framework captures three distinct cardiovascular phenotype categories: (1) ischemic ASCVD outcomes including acute myocardial infarction, coronary thrombosis, and chronic atherosclerotic heart disease; (2) electrical cardiac disorders including drug-induced arrhythmias and sudden cardiac death; (3) structural myocardial and cardiomyopathic injury secondary to toxic or hypoxic stress. This spectrum further differentiates acute fatal toxic events and chronic progressive atherosclerotic disease acceleration mediated by substance use [22].

Conceptual Framework: Underlying Cause of Death and Multiple Causes of Death Coding: The U.S. death certificate employs a structured system for cause-of-death coding that is essential to understanding our case definition. The Underlying Cause of Death (UCOD) is defined as the disease, injury, or abnormal condition that initiates the sequence of events leading directly or indirectly to death. The UCOD is the primary cause used in national mortality surveillance and vital statistics reporting. Multiple Causes of Death (MCoD) are all diseases and abnormal conditions mentioned anywhere on the death certificate, recorded sequentially in Part I (the causal sequence) and Part II (other contributing conditions) sections. While these secondary codes provide important context, they do not replace the UCOD in primary mortality statistics.Our study specifically identified deaths in which: (1) cardiovascular disease (I00-I99) was coded as the underlying cause initiating the fatal sequence, AND (2) drug poisoning codes (X40-X44) appeared in the multiple causes fields. This definition captures deaths in which the cardiovascular system is identified as the primary driver of the fatal event, with concurrent documentation of drug involvement. However, the specific mechanistic relationship linking drug exposure to cardiovascular dysfunction must be understood within strict limitations.

Critical Clarification on Causality and Co-occurrence: Importantly: The simultaneous presence of drug poisoning codes and a cardiovascular disease UCOD on a death certificate does NOT establish that drug exposure directly caused the cardiovascular event. Rather, these cases represent epidemiological co-occurrence of cardiovascular disease and drug poisoning coding patterns on U.S. death certificates. Death certificates do not encode the precise temporal or mechanistic relationship between conditions.The co-occurrence of these coding patterns may reflect multiple different mechanisms and scenarios. First, direct toxic effects, where drugs directly precipitate acute cardiovascular events (e.g., cocaine-induced coronary vasospasm and myocardial infarction, or methamphetamine-induced cardiac arrhythmia). Second, indirect cardiovascular complications, where drug effects precipitate secondary cardiovascular dysfunction (e.g., opioid-induced respiratory depression→hypoxia→cardiac arrest; stimulant-induced hypertensive emergency→stroke). Third, coincidental findings, where patients have pre-existing cardiovascular disease (e.g., chronic ischemic heart disease) and experience acute intoxication during the study period, but these conditions are unrelated by causal mechanism. Fourth, complex interactions, where acute drug intoxication triggers decompensation in patients with underlying cardiovascular instability.

This study employs an ecological surveillance design with a general-population denominator, which creates an inherent denominator mismatch relative to individual drug-exposed populations. All reported AAMRs represent population-level coding prevalence in the total U.S. population, not individualized mortality risk among persons who use drugs. This mismatch introduces potential ecological fallacy and risk attenuation bias, which precludes causal or individual-level risk inference. All findings are strictly descriptive of national coding trend patterns, consistent with CDC multi-cause surveillance best practices [17].

Variables

Mortality data were stratified by the following demographic variables to examine disparities: Sex: Male, Female; Race and Ethnicity: Hispanic or Latino, Non-Hispanic Black or African American, Non-Hispanic White; Age Group: 25–44 years, 45–64 years, and 65 years and older; Urban–Rural Classification: Counties were categorized based on the 2013 U.S. Census classification into urban (large metropolitan, small/medium metropolitan) and rural (nonmetropolitan, population < 50000) areas.Region: As defined by the U.S. Census Bureau, regions included Northeast, Midwest, South, and West.

Notably, CDC WONDER public data does not provide disaggregated Hispanic subgroup data, granular socioeconomic variables (household income, educational attainment, health insurance status), or fine-scale healthcare access metrics. Standard census urban–rural classification is a coarse categorical measure that cannot capture continuous gradients of healthcare resource availability, surveillance intensity, and socioeconomic deprivation. All observed demographic disparities should be interpreted as hybrid outcomes of structural social determinants, population behavioral differences, and surveillance measurement bias, rather than biological racial or ethnic differences.

Statistical analysis

Age-standardized mortality rates per 100,000 population were calculated using the annual bridged-race population estimates provided by CDC WONDER. The age-adjusted mortality rate (AAMR) was standardized to the 2000 U.S. standard population to enable unbiased comparisons across populations with differing age structures and over time.

Temporal trends were analyzed using Joinpoint Regression software (Version 4.9.1.0, National Cancer Institute). Joinpoint regression is a method that identifies statistically significant changes in trend (called “joinpoints”) by testing whether changes in the line slopes are statistically significant. For all analyses, we used a maximum of 4 joinpoints (resulting in up to 5 linear segments), consistent with standard recommendations for surveillance data spanning 22 years. The model selection was based on permutation tests, where the software randomly permutes the data 4499 times and compares the number of joinpoints detected in the permuted data to the observed data. A joinpoint is considered statistically significant when the p-value from the permutation test is less than 0.05. This means we can be confident that the identified change in trend is not due to random variation. The annual percent change (APC) for each segment is calculated using weighted least-squares regression with weights equal to the inverse of the variance of the log-transformed death rate. The average annual percent change (AAPC) over the entire period (1999–2020) was also computed as a summary measure, representing the overall trend across the entire study period. For both APC and AAPC, we report the estimates along with their 95% confidence intervals (CI), which represent the range of values that have a 95% probability of containing the true value. A p-value less than 0.05 was considered statistically significant.

An important caveat regarding the interpretation of wide confidence intervals: Some subgroup analyses produced annual percent change estimates with notably wide confidence intervals, particularly for regional and demographic strata during early time periods (e.g., Northeast 1999–2006: APC = -2.564%, 95% CI: -16.152 to 2.160). Such wide confidence intervals indicate substantial uncertainty in the APC estimates and reflect smaller sample sizes in those particular strata and time periods. Wide confidence intervals suggest instability in those estimates and should be interpreted with considerable caution. Readers should avoid over-interpreting these trend segments, and instead focus on those with narrower confidence intervals that reflect more stable and reliable estimates. To inform readers about the stability of estimates in different strata, we have specified the total number of deaths for each major subgroup analysis and highlighted segments where sample sizes may be insufficient for robust trend estimation.

For stratified analyses (by sex, ethnicity, urbanization level, census region, and age group), we applied the same Joinpoint methodology. However, we note that certain strata had substantially smaller sample sizes than the overall cohort. For example, the Hispanic/Latino group comprised only 940 total deaths (2.8% of the cohort), averaging approximately 43 deaths per year across the 22-year study period. For such smaller strata, the assumptions underlying Joinpoint regression may be less well met, and trend estimates should be viewed as exploratory rather than definitive. All data extraction was performed using the CDC WONDER online interface. Subsequent data management and Joinpoint analyses were conducted using R software (Version 4.3.1) with the JoinPoint package.

Results

Overall trends

Between 1999 and 2020, a total of 33,536 deaths occurred in the United States in which cardiovascular disease (ICD-10: I00–I99) was coded as the underlying cause of death and drug poisoning (ICD-10: X40–X44) was documented in the multiple causes of death.The overall age-adjusted mortality rate (AAMR) showed a substantial and significant increase, rising from 0.500 per 100,000 population (95% CI: 0.467–0.532) in 1999 to 0.979 (95% CI: 0.937–1.020) in 2020. Among these deaths, 23,321 (69.54%) were male and 10,215 (30.46%) were female. Byethnicity and ethnicity, there were 940 deaths among Hispanic or Latino individuals, 8075 among non-Hispanic Black or African American individuals, and 24,512 among non-Hispanic non-Hispanic White individualsindividuals. Age distribution revealed 8470 deaths (25.26%) among young adults aged 25–44 years, 20,130 (60.02%) among middle-aged adults aged 45–64 years, and 4936 (14.71%) among older adults aged 65 years and above (Table 1, Table S1).

Table 1.

Demographic characteristics of deaths with cardiovascular mortality from drug poisoning in the United States from 1999 to 2020

Variables	Deaths	Population	AAMR(95%CI)
Overall	33,536	4,473,854,489	0.735(0.727–0.743)
Sex
Male	23,321	2,154,556,911	1.060(1.046–1.073)
Female	10,215	2,319,297,578	0.428(0.420–0.437)
Race/Ethnicity
Black or African American	8075	546,447,527	1.451(1.419–1.483)
non-Hispanic non-Hispanic White individualsindividuals	24,512	3,631,977,407	0.644(0.636–0.652)
Census Region
Northeast	3404	827,193,779	0.405(0.391–0.419)
Midwest	6489	969,567,311	0.654(0.638–0.670)
South	11,417	1,652,256,217	0.671(0.658–0.683)
West	12,226	1,024,837,182	1.165(1.144–1.185)
Urbaniztion
Metropolitan(Urban)	29,251	3,795,213,822	0.746(0.737–0.754)
Nonmetropolitan(Rural)	4285	678,634,169	0.639(0.619–0.658)
Ten-Year Age Groups
25–44 years	8470	1,851,376,757	0.457(0.448–0.467)
45–64 years	20,130	1,694,001,067	1.188(1.172–1.205)
65 + years	4936	928,476,665	0.501(0.487–0.515)

Abbreviations: AAMR age-adjusted mortality rate, CI confidence interval

Joinpoint regression analysis identified three distinct segments in this trend (Table 2, Figure S1):1999–2003: A period of significant increase, with an annual percent change (APC) of 9.746% (95% CI: 6.397 to 16.054; P < 0.001)0.2003–2014: A period of relative stabilization with modest growth, marked by an APC of 0.84% (95% CI: − 1.122 to 1.552; P = 0.188)0.2014–2020: Another phase of rapid increase, with an APC of 4.887% (95% CI: 3.366 to 8.337; P < 0.001).The average annual percent change (AAPC) over the entire period was 3.639% (95% CI: 3.252 to 4.218; P < 0.001) (Table 2; Fig. 1).

Table 2.

Annual percentage change (APC) and average annual percent change (AAPC) in cardiovascular mortality from drug poisoning in the United States 1999–2020

Variables	Trend segment	Year interval	APC (95% CI)	AAPC (95% CI)	p value
Entire Cohort	-	1999–2020	-	3.639(3.252 to 4.218)	< 0.001
	1	1999–2003	9.746(6.397 to 16.054)	-	< 0.001
	2	2003–2014	0.843(−1.122 to 1.552)	-	0.188
	3	2014–2020	4.887(3.366 to 8.337)	-	< 0.001
Sex
Male	-	1999–2020	-	3.982(3.574 to 4.846)	< 0.001
	1	1999–2004	9.267(6.226 to 14.992)	-	0.001
	2	2004–2012	−0.177(−6.074 to 1.356)	-	0.717
	3	2012–2020	5.010(3.640 to 9.341)	-	< 0.001
Female	-	1999–2020	-	1.739(1.199 to 2.358)	< 0.001
Race/Ethnicity
Black or African American	-	1999–2020	-	0.917(0.188 to 2.133)	0.016
	1	1999–2005	5.207(1.723 to 12.481)	-	0.004
	2	2005–2012	−7.569(−17.972 to−4.549)	-	0.001
	3	2012–2020	5.630(2.810 to 11.471)	-	0.001
non-Hispanic non-Hispanic White individualsindividuals	-	1999–2020	-	4.191(3.550 to 5.201)	< 0.001
	1	1999–2003	9.331(4.635 to 24.990)	-	< 0.001
	2	2003–2020	3.018(2.048 to 3.493)	-	0.026
Census Region
Northeast	-	1999–2020	-	2.305(1.233 to 3.776)	< 0.001
	1	1999–2006	−2.564(−16.152 to 2.160)	-	0.276
	2	2006–2020	4.831(3.362 to 11.609)	-	0.012
Midwest	-	1999–2020	-	2.193(1.621 to 3.057)	< 0.001
	1	1999–2006	4.227(2.088 to 10.460)	-	< 0.001
	2	2006–2014	−2.025(−8.824 to−0.163)	-	0.030
	3	2014–2020	5.645(2.931 to 13.795)	-	< 0.001
South	-	1999–2020	-	4.428(3.840 to 5.281)	< 0.001
	1	1999–2004	11.643(7.507 to 20.628)	-	< 0.001
	2	2004–2010	−2.064(−8.493 to 0.682)	-	0.147
	3	2010–2020	4.963(3.911 to 6.933)	-	< 0.001
West	-	1999–2020	-	3.314(2.482 to 4.666)	< 0.001
	1	1999–2003	12.571(5.782 to 34.209)	-	< 0.001
	2	2003–2020	1.248(0.353 to 1.917)	-	0.022
Urbaniztion
Metropolitan(Urban)	-	1999–2020	-	3.205(2.669 to 4.236)	< 0.001
	1	1999–2003	9.407(4.624 to 23.669)	-	< 0.001
	2	2003–2013	0.189(−5.032 to 1.245)	-	0.953
	3	2013–2020	4.139(2.454 to 10.490)	-	0.004
Nonmetropolitan(Rural)	-	1999–2020	-	5.419(4.767 to 6.279)	< 0.001
Age
25–44	-	1999–2020	-	0.956(0.584 to 1.490)	< 0.001
	1	1999–2005	5.221(3.558 to 8.348)	-	< 0.001
	2	2005–2008	−8.932(−11.294 to−3.590)	-	< 0.001
	3	2008–2020	1.471(0.786 to 2.708)	-	< 0.001
45–64	-	1999–2020	-	4.707(4.144 to 5.519)	< 0.001
	1	1999–2005	9.444(6.339 to 15.093)	-	0.006
	2	2005–2015	1.480(−3.207 to 2.559)	-	0.322
	3	2015–2020	5.707(3.280 to 11.981)	-	< 0.001
65+	-	1999–2020	-	−0.538(−1.263 to 0.241)	0.178
	1	1999–2006	−7.429(−15.485 to−0.997)	-	0.045
	2	2006–2014	0.863(−10.514 to 4.606)	-	0.8455
	3	2014−020	6.153(2.679 to 15.349)	-	0.020

Fig. 1.

Overall trends in age-standardized mortality rates for drug poisoning-related cardiovascular death in the United States (1999–2020)

Trends stratified by sex

Throughout the study period, persistent and significant disparities were observed between males and females. In 2020, the age-adjusted mortality rate (AAMR) for males (1.06 per million, 95% CI: 1.046–1.073) was 2.48 times that of females (0.428 per million, 95% CI: 0.420–0.437) (Table 1, Table S2). Trend analyses indicated that although mortality rates increased significantly in both sexes, male mortality increased at a considerably faster rate. From 1999 to 2020, the average annual percent change (AAPC) was 3.982% (95% CI: 3.574 to 4.846, P < 0.001) for males compared to 1.739% (95% CI: 1.199 to 2.358, P < 0.001) for females (Table 2; Fig. 2).The Joinpoint model for males reflected the overall three-phase pattern:1999–2004: A period of rapid increase, with an APC of 9.267% (95% CI: 6.226 to 14.992; P = 0.001)0.2004–2012: A plateau with a slight non-significant decline, APC of − 0.177% (95% CI: − 6.074 to 1.356; P = 0.717)0.2012–2020: A return to rapid increase, with an APC of 5.010% (95% CI: 3.640 to 9.341; P < 0.001) (Table 2, Figure S2).

Fig. 2.

Age-standardized mortality trends for drug poisoning-related cardiovascular deaths in the United States stratified by sex (1999–2020)

Trends stratified byethnicity

The age-adjusted mortality rate (AAMR) among non-Hispanic Black or African American individuals (1.451 per 100,000, 95% CI:1.419–1.483) was significantly higher than that among non-Hispanic non-Hispanic White individualsindividuals (0.644 per 100,000, 95% CI:0.636–0.652) and Hispanic or Latino individuals (0.335 per 100,000, 95% CI:0.311–0.359) (Table 2, Table S3).Mortality trends varied considerably by racial and ethnic group: Non-Hispanic White: Exhibited a consistent upward trend, with an AAPC of 4.191% (95% CI:3.550 to 5.201, p < 0.001).Non-Hispanic Black or African American: Demonstrated a more dynamic pattern: an initial increase from 1999 to 2005 (APC = 5.207%, 95% CI: 1.723 to 12.481, p = 0.004), followed by a significant decline during 2005–2012 (APC = − 7.569%, 95% CI–17.972 to − 4.549, p = 0.001), and a return to a sharp increase from 2012 to 2020 (APC = 5.630%, 95% CI: 2.810 to 11.471, p = 0.001).Hispanic or Latino: with the relatively small number of cases, joinpoint analysis was not performed; only the overall trend is described (Table 2; Fig. 3, Figure S3).

Fig. 3.

Age-standardized mortality trends for drug poisoning-related cardiovascular deaths in the United States stratified byethnicity (1999–2020)

Trends stratified by level of urbanization

Residents of nonmetropolitan areas had a lower mortality rate (AAMR = 0.639) than those in metropolitan areas (AAMR = 0.746) (Table 1, Table S4).Those in metropolitan areas exhibited a consistent upward trend in mortality (AAPC = 5.419%, 95% CI: 4.767 to 6.279, P < 0.001). In contrast, nonmetropolitan areas showed an overall increase across the study period (AAPC = 3.205%, 95% CI: 2.669 to 4.236, P < 0.001), with three distinct phases:1999–2003: A period of rapid increase (APC = 9.407%, 95% CI: 4.624 to 23.669, P < 0.001)0.2003–2013: A plateau with minimal change (APC = 0.189%, 95% CI: − 5.032 to 1.245, P = 0.953)0.2013–2020: A return to a significant upward trend (APC = 4.139%, 95% CI: 2.454 to 10.490, P = 0.004)(Table 2; Fig. 4, Figure S4).

Fig. 4.

Age-standardized mortality trends for drug poisoning-related cardiovascular deaths in the United States stratified by urbanization level (1999–2020)

Trends stratified by census region

Significant geographic disparities were observed at the regional level. The highest age-adjusted mortality rate (AAMR) was found in the West (1.165 per 100,000, 95% CI: 1.144–1.185), followed by the South (0.671 per 100,000, 95% CI: 0.658–0.683), the Midwest (0.654 per 100,000, 95% CI: 0.638–0.670), and the Northeast (0.405 per 100,000, 95% CI: 0.391–0.419) (Table 1, Table S5).Mortality trends varied by region: Northeast: Showed an overall upward trend with an AAPC of 2.305% (95% CI: 1.233 to 3.776, p < 0.001). A non-significant decline occurred from 1999 to 2006 (APC = − 2.564%, 95% CI: − 16.152 to 2.160, p = 0.276), followed by a significant increase from 2006 to 2020 (APC = 4.831%, 95% CI: 3.362 to 11.609, p = 0.012).Midwest: Exhibited an overall increase with an AAPC of 2.193% (95% CI: 1.621 to 3.057, p < 0.001). An initial increase occurred during 1999–2006 (APC = 4.227%, 95% CI: 2.088 to 10.460, p < 0.001), followed by a decline in 2006–2014 (APC = − 2.025%, 95% CI: − 8.824 to − 0.163, p = 0.030), and a return to a sharp upward trend from 2014 to 2020 (APC = 5.645%, 95% CI: 2.931 to 13.795, p < 0.001).South: Demonstrated an overall increase with an AAPC of 4.428% (95% CI: 3.840 to 5.281, p < 0.001). A rapid rise was observed from 1999 to 2004 (APC = 11.643%, 95% CI: 7.507 to 20.628, p < 0.001), followed by a non-significant decline during 2004–2010 (APC = − 2.064%, 95% CI: − 8.493 to 0.682, p = 0.147), and a significant upward trend from 2010 to 2020 (APC = 4.963%, 95% CI: 3.911 to 6.933, p < 0.001).West: Also showed an overall upward trend with an AAPC of 3.314% (95% CI: 2.482 to 4.666, p < 0.001). A sharp increase occurred from 1999 to 2003 (APC = 12.571%, 95% CI: 5.782 to 34.209, p < 0.001), followed by a more gradual but significant increase from 2003 to 2020 (APC = 1.248%, 95% CI: 0.353 to 1.917, p = 0.022) (Table 2; Fig. 5, Figure S5).

Fig. 5.

Age-standardized mortality trends for drug poisoning-related cardiovascular deaths in the United States stratified by census region (1999–2020)

Trends stratified by state

Substantial heterogeneity was observed in mortality rates at the state level. Nevada had the highest age-adjusted mortality rate (AAMR = 3.341; 95% CI: 3.158–3.524), followed by New Mexico (1.426; 95% CI: 1.285–1.566), California (1.395; 95% CI: 1.364–1.427), Oklahoma (1.282; 95% CI: 1.184–1.381), and Arizona (1.215; 95% CI: 1.141–1.288). Many of these states are located in the West region.In contrast, the lowest AAMRs were observed in Minnesota (0.179; 95% CI: 0.151–0.207), South Dakota (0.186; 95% CI: 0.115–0.284), and Wisconsin (0.201; 95% CI: 0.171–0.230) (Table S6 and Fig. 6).

Fig. 6.

Age-standardized percent mortality rate changes in drug poisoning-related cardiovascular deaths in US states from 1999 to 2020

Trends stratified by age group

Significant shifts in mortality distribution by age group were observed during the study period. From 1999 to 2020, the highest age-adjusted mortality rate (AAMR) was recorded in the 45–64 age group (1.188; 95% CI: 1.172–1.205), followed by the 65 + age group (0.663; 95% CI: 0.595–0.731), while the 25–44 age group had the lowest AAMR (0.457; 95% CI: 0.448–0.467) (Table 1,Table S6).Trend analyses revealed distinct patterns across age groups: Ages 45–64: Exhibited an overall upward trend, with an AAPC of 4.707% (95% CI: 4.144 to 5.519, p < 0.001). A sharp increase occurred from 1999 to 2005 (APC = 9.444%, 95% CI: 6.339 to 15.093, p = 0.006), followed by a slower non-significant rise during 2005–2015 (APC = 1.480%, 95% CI: − 3.207 to 2.559, p = 0.322), and a return to rapid increase from 2015 to 2020 (APC = 5.707%, 95% CI: 3.280 to 11.981, p < 0.001).Ages 25–44: Showed a modest overall increase with an AAPC of 0.956% (95% CI: 0.584 to 1.490, p < 0.001). An initial rapid rise from 1999 to 2005 (APC = 5.221%, 95% CI: 3.558 to 8.348, p < 0.001) was followed by a sharp decline during 2005–2008 (APC = − 8.932%, 95% CI: − 11.294 to − 3.590, p < 0.001), and a subsequent upward trend from 2008 to 2020 (APC = 1.471%, 95% CI: 0.786 to 2.708, p < 0.001).Ages 65+: Demonstrated an overall slight declining trend, with an AAPC of − 0.538% (95% CI: − 1.263 to 0.241, p = 0.178). A significant decrease occurred from 1999 to 2006 (APC = − 7.429%, 95% CI: − 15.485 to − 0.997, p = 0.045), followed by a plateau with minimal change during 2006–2014 (APC = 0.863%, 95% CI: − 10.514 to 4.606, p = 0.845), and a significant increase from 2014 to 2020 (APC = 6.153%, 95% CI: 2.679 to 15.349, p = 0.020) (Table 2; Fig. 7, Figure S6).

Fig. 7.

Age-standardized mortality trends for drug poisoning-related cardiovascular deaths in the United States stratified by age group (1999–2020)

Discussion

This study provides a comprehensive epidemiological overview of deaths with concurrent cardiovascular disease and drug poisoning coding(CVM-DP) in the United States from 1999 to 2020. It reveals a concerning trend: over the 22-year period, CVM-DP nearly doubled, rising from 0.500 per 100,000 population in 1999 to 0.979 in 2020, with significant disparities across demographic subgroups. These findings underscore that substance abuse poses not only a threat to the respiratory system but also a serious challenge to cardiovascular health, warranting heightened attention from clinicians and public health authorities.

Consistent with modern CDC multi-cause surveillance principles, this study exclusively describes population-level death certificate coding patterns, not individual-level disease risk or causal relationships [17]. All observed temporal trends and demographic disparities reflect co-coding prevalence in the general U.S. population, which may arise from true epidemiological shifts, changes in toxicology surveillance practices, death certification coding conventions, or combinations of these factors. No individual-level causal inference or drug-user risk estimation is warranted from these ecological aggregate data, which are constrained by a general-population denominator and lack individual exposure covariates.

It is essential to clarify an important distinction when interpreting this study’s findings. This analysis describes population-level mortality rates—specifically, the age-adjusted number of deaths per million persons in the U.S. population coded with both cardiovascular disease and drug poisoning on death certificates. These population-level rates should not be misinterpreted as estimates of individual-level risk among people who use drugs. The distinction is fundamental: our data use the entire U.S. population as the denominator, including the approximately 99% of the population who do not use drugs. Therefore, these rates describe epidemiological patterns in the overall population, not the risk experienced by drug-using individuals. For example, while we report that the overall age-adjusted mortality rate increased from 0.500 to 0.979 per 100,000 population from 1999 to 2020, this does not indicate that the risk of cardiovascular death among drug users increased by this magnitude. Rather, it indicates that the proportion of deaths in the entire U.S. population with this particular coding pattern increased. To estimate true cardiovascular risk among drug-using populations, one would need to use drug users as the denominator (e.g., the number of deaths among drug users divided by the total population of drug users), data which are not available from the CDC WONDER database used in this study.

Furthermore, the observed increases over time in these population-level rates may reflect multiple distinct phenomena: genuine increases in the incidence of cardiovascular deaths with concurrent drug involvement (reflecting true epidemiological change), changes in medical examiner practices and toxicology testing patterns, changes in death certificate coding practices, changes in the racial and ethnic composition of the population, or combinations of these factors. Our surveillance data, by definition, cannot distinguish between these alternative explanations. While temporal trends are valuable for detecting changes and generating hypotheses requiring further investigation, they do not establish the underlying causes of observed changes.

Drug poisoning contributes to cardiovascular mortality through three coherent, mechanistically distinct pathways integrating acute cardiotoxicity, indirect hypoxic injury, and chronic ASCVD progression [22], unifying previously fragmented mechanistic descriptions: Direct acute cardiotoxicity: Stimulant substances including cocaine and methamphetamine exert immediate cardiotoxic effects via catecholamine storm induction, coronary vasospasm, pro-thrombotic endothelial activation, and malignant cardiac arrhythmogenesis. These direct toxic mechanisms precipitate acute ischemic myocardial infarction, sudden cardiac death, and acute heart failure independent of pre-existing cardiac disease, representing primary drug-mediated cardiovascular fatal events [22]. Indirect hypoxic cardiovascular injury: Opioid-class substances primarily induce central respiratory depression, leading to systemic hypoxia, hypotension, and secondary myocardial ischemic injury. Progressive hypoxia exacerbates cardiac cellular dysfunction, bradycardia, and eventual cardiac arrest, representing a secondary cardiovascular fatal pathway mediated by drug-induced respiratory failure rather than direct cardiac toxicity.Chronic ASCVD acceleration and plaque destabilization: Long-term substance use promotes systemic inflammation, endothelial dysfunction, oxidative stress, and dyslipidemia, which accelerate atherosclerotic plaque formation, progression, and destabilization [22]. Pre-existing ASCVD lesions become vulnerable to rupture and thrombosis during acute drug intoxication episodes, linking chronic cardiovascular disease burden to acute drug-related fatal decompensation. This pathway explains the co-occurrence of chronic cardiovascular disease UCOD and acute drug poisoning contributory coding observed in our cohort [22].

The observed trend in population-level CVM-DP mortality rates displayed three distinct phases. The initial rapid increase from 1999 to 2003 (APC = 9.746%) occurred during a period marked by a substantial increase in prescriptions for opioid analgesics such as oxycodone and hydrocodone [23, 24].However, we emphasize that the temporal association between increased prescription opioid availability and increased CVM-DP death coding patterns does NOT establish causality at the individual or population level. The subsequent period of relative stability from 2003 to 2014 (APC = 0.84%) coincided with implementation of prescription drug monitoring programs and reported changes in prescribing practices, but again, temporal association does not establish causality [25, 26]. The reacceleration in growth from 2014 to 2020 (APC = 4.887%) was temporally aligned with the documented widespread infiltration of fentanyl and its analogs into the illicit drug market. These temporal alignments are interesting and suggest potential mechanistic links worth investigating through additional research, but they do not demonstrate that specific policy changes or market dynamics caused the observed changes in coding patterns. Alternative explanations including changes in death investigation and toxicology testing practices, changes in death certificate coding conventions, or combinations of factors may equally explain the observed trends.These highly potent synthetic opioids are known for their particularly severe toxic effects on the cardiovascular system [27, 28].

It is critical to emphasize that our findings describe epidemiological patterns in death certificate coding rather than establishing direct causal mechanisms linking drug exposure to cardiovascular death. Death certificates are fundamentally administrative and medical vital statistics documents rather than research instruments designed to encode detailed mechanistic relationships.

The simultaneous presence of both cardiovascular disease codes (UCOD) and drug poisoning codes (MCoD) does not necessarily indicate: (1) that drug exposure directly caused the cardiovascular event; (2) the temporal sequence of the medical events; or (3) whether drug exposure triggered, exacerbated, or merely coincided with the cardiovascular disease. To illustrate this limitation concretely: the following two scenarios would be coded identically and thus indistinguishable in our analysis: A 35-year-old with no prior cardiac history who uses cocaine, experiences acute coronary vasospasm, and dies of myocardial infarction; and a 75-year-old with longstanding chronic ischemic heart disease and multiple cardiovascular risk factors who becomes acutely intoxicated coincidentally and experiences myocardial infarction unrelated to the acute intoxication. Both cases appear in our dataset as deaths with cardiovascular disease (UCOD) and drug poisoning (MCoD), yet they represent fundamentally different etiological pathways with entirely different implications for prevention and intervention.

Despite these limitations, the epidemiological documentation of these co-occurring death certificate coding patterns remains important for several reasons. First, from a surveillance perspective, these patterns identify populations at elevated simultaneous risk for both conditions, warranting integrated cardiovascular and substance use assessment and intervention. Second, from a service delivery perspective, healthcare systems serving populations with high rates of substance use should implement cardiovascular screening and risk factor modification programs. Third, from a research hypothesis generation perspective, the identified demographic and geographic disparities point toward specific populations or contexts requiring further investigation.The overall trend in deaths with concurrent cardiovascular and drug poisoning coding displayed three distinct phases, and interpretation of these patterns should account for both direct toxic effects and alternative mechanisms. The first phase (1999–2003, APC = 9.746%) showed rapid increase coinciding with expansion of prescription opioid availability. During this phase, observed deaths may have involved direct cardiotoxic effects of prescription opioids in opioid-naive users, combined effects of opioids with pre-existing cardiovascular disease, or inadequate implementation of overdose prevention strategies. The second phase (2003–2014, APC = 0.84%) of relative stability likely reflects implementation of prescription drug monitoring programs and changes in prescribing practices. The relative stability during this phase does not definitively establish causality but suggests that external factors influenced the coding patterns observed in death certificates. The third phase (2014–2020, APC = 4.887%) shows reacceleration coinciding with infiltration of fentanyl and potent synthetic opioids into the illicit drug supply. While these highly potent substances are associated with severe acute toxicity including cardiovascular complications, the precise mechanisms linking synthetic opioid exposure to the cardiovascular deaths captured in our dataset remain undetermined and cannot be established from death certificate coding alone.

Significant gender and racial disparities were observed in deaths with concurrent cardiovascular and drug poisoning coding.All observed demographic disparities represent hybrid outcomes of hypothesized true epidemiological behavioral/risk differences, likely surveillance and coding measurement bias, and unmeasured structural social determinants of health. No observed racial or ethnic disparity is interpreted as reflecting inherent biological differences across population groups.These disparities likely reflect combinations of multiple factors: differences in drug exposure patterns and substance types used; underlying cardiovascular disease prevalence and management; healthcare access and cardiovascular screening disparities; and socioeconomic factors influencing both substance use risk and cardiovascular health. For example, the 2.48-fold higher mortality rate in males could reflect differences in drug use patterns (males more likely to use high-risk substances), physiological susceptibility to cardiotoxic effects, behavioral risk factors (such as injection drug use), underlying cardiovascular disease burden, or various combinations of these factors. Our data cannot distinguish among these potential mechanisms.

Notably, significant gender disparities were observed, with males consistently experiencing higher mortality rates that also increased at a faster pace. By 2020, the AAMR among males was 2.48 times that of females. This observed disparity likely reflects a combination of true epidemiological differences and potentially differential surveillance intensity [29, 30]. Among possible true differences: males may be more likely to use high-risk substances such as heroin and cocaine, whereas females more commonly use prescription opioids; males more frequently engage in higher-risk consumption behaviors such as intravenous injection and polydrug use, which may elevate cardiovascular complication risk; and physiological sex differences may result in differential biological susceptibility to cardiotoxic effects of certain drugs [31]. However, the observed gender differential may also partly reflect differences in surveillance practices: younger male decedents may be subject to more intensive toxicology screening and documentation of drug involvement compared to female decedents, or medical examiners may have differential clinical suspicion regarding drug involvement by gender [32]. Our data cannot distinguish the true epidemiological differences from potential measurement bias related to differential surveillance.

Racial and ethnic stratification analyses revealed significant observed differences in CVM-DP mortality rates. Non-Hispanic Black or African American individuals showed markedly higher mortality rates (AAMR = 1.451) compared to Non-Hispanic non-Hispanic White individualsindividuals (AAMR = 0.644), representing a 2.25-fold disparity. These observed differences likely reflect multiple intersecting factors. True epidemiological risk differences may exist due to: historical inequities in healthcare access and cardiovascular disease management; socioeconomic disadvantage and its effects on substance use risk and disease burden; and differences in substance use patterns across communities [33, 34]. However, we must acknowledge that the observed racial disparities may also partly reflect differential surveillance practices: Heterogeneity in medical examiner office resources, toxicology testing protocols, and forensic practice standards across jurisdictions with different racial/ethnic demographics; Potential implicit bias in clinical decision-making regarding whether to pursue toxicology testing and how to code death causes for decedents of different racial backgrounds; Differential geographic distribution of medical examiner offices relative to racial/ethnic community locations, affecting access to sophisticated death investigation and toxicology services. The significant decline in mortality between 2005 and 2012 and subsequent resurgence after 2012 were observed across racial/ethnic groups, suggesting that period effects (such as changes in prescribing practices and later emergence of synthetic opioids) influenced observed trends [35]. However, we cannot determine whether the magnitude of these changes differed due to true epidemiological shifts versus changes in surveillance practices. Our aggregate, coded-based data cannot decompose the contributions of true risk differences from measurement bias related to differential surveillance [36, 37].

Age-specific data indicated that the highest mortality burden occurred among middle-aged adults (45–64 years), consistent with the higher prevalence of substance use disorders within the baby boomer generation [38, 39]. Although the 25–44 age group had a relatively lower overall mortality rate, its sustained upward trend after 2008 reflects increasing use of illicit synthetic opioids among younger adults [40]. In contrast, adults aged 65 and older demonstrated an overall declining trend, though a significant rise occurred from 2014 to 2020. This late increase may be linked to changes in chronic pain management practices and the complex interplay of polypharmacy in this population [41, 42].

Geographic distribution analyses showed pronounced regional variation, with the highest mortality rates clustered in the West, particularly in Nevada (AAMR = 3.341), New Mexico, and California. These observed geographic patterns may reflect true epidemiological differences including regional drug trafficking routes, local market dynamics, and differences in access to medical and preventive resources across regions. The exceptionally high rate in Nevada warrants careful interpretation: it may indicate that this state faces the highest true burden of drug-related cardiovascular mortality, but it may also reflect Nevada’s particularly thorough death investigation practices, comprehensive toxicology testing infrastructure, and forensic pathology standards. Different states have varying medical examiner systems, resources, and protocols for toxicology screening, which could systematically influence observed rates. Disparities by urbanization level were also evident: metropolitan areas showed higher mortality (AAMR = 0.746) than nonmetropolitan areas (AAMR = 0.639) [43, 44]. This difference may reflect greater drug availability, more complex use patterns, and concentration of high-risk populations in urban settings, but may also reflect more robust toxicology laboratory infrastructure, greater medical examiner office density, and higher surveillance intensity in metropolitan areas compared to rural counties [42, 45]. Nonmetropolitan areas experienced rapid increase in mortality from 2013 to 2020, indicating either true disease emergence in rural communities, increased surveillance intensity for drug-related deaths in rural areas, or both. Our analysis cannot definitively determine whether geographic patterns primarily represent true epidemiological risk differences or partly reflect systematic differences in death investigation infrastructure and surveillance practices across regions [46, 47].

From clinical and public health perspectives, the findings of this study carry several important implications. First, they underscore the need to integrate cardiovascular risk assessment into the routine care of individuals with substance use disorders, particularly those with pre-existing cardiovascular conditions or risk factors [48]. Clinicians should be aware that drug abuse can precipitate serious cardiovascular events including myocardial infarction, arrhythmias, and heart failure even among younger patients [49, 50]. Second, emergency medicine practitioners should maintain a high index of suspicion not only for respiratory depression but also for early signs of cardiovascular complications in cases of drug intoxication [51]. Third, prevention strategies should be tailored to high-risk groups, such as middle-aged men and Black communities, while accounting for region-specific usage patterns and risk profiles [52, 53]. This study also has several limitations.First, Measurement and coding bias limitations: Death certificate-based surveillance is subject to systematic misclassification, variable toxicology testing intensity across demographic and geographic groups, and heterogeneous provider coding practices. Differential surveillance intensity creates systematic measurement bias that cannot be disentangled from true epidemiological disparities in aggregate CDC WONDER data. Drug involvement may be under-documented in specific populations or regions, leading to incomplete case ascertainment.Second, Study design and data source limitations: This is an ecological repeated cross-sectional study with no individual-level exposure, temporal sequence, or mechanistic covariate data. The general-population denominator creates inherent denominator mismatch and ecological fallacy risk, precluding individual-level drug-user risk estimation. CDC WONDER public data lacks granular socioeconomic, healthcare access, and disaggregated ethnic subgroup variables, limiting comprehensive equity analysis.Finally, Statistical and subgroup analysis limitations: Small underpowered demographic subgroups (e.g., Hispanic/Latino populations) exhibit unstable trend estimates with wide confidence intervals. Stratified Joinpoint analyses for small strata carry increased statistical uncertainty, and all subgroup trend findings are strictly exploratory rather than definitive.

In summary, this study highlights the substantial and growing burden of drug poisoning-related cardiovascular mortality in the United States, along with pronounced disparities across demographic subgroups. These findings emphasize the urgent need to integrate cardiovascular health considerations into substance abuse prevention, treatment, and policy responses. A multidisciplinary and multi-level approach is essential to effectively mitigate the far-reaching effects of drug abuse on cardiovascular health.

Abbreviations

CDC

Centers for Disease Control

WONDR

Wide-ranging Online Data for Epidemiologic Research

AAMR

Age-adjusted mortality rates

APC

Annual percent change

AAPC

Average annual percent change

Authors' contributions

Caoyang Fang and Longshan Yu wrote the main manuscript text;Jiaojiao Peng,Shouwei Zhu,Zhe Yang and Xin Yu prepared figures 1-7 and tables 1-2. All authors reviewed the manuscript.

Funding

No Funding.

Data availability

All data used in this study to formulate our results is publicly available online at the CDC WONDER database (https://wonder.cdc.gov/).

Declarations

Ethics approval and consent to participate

Because CDC WONDER contains anonymous publicly available data.Therefore, the Ethics Committee of The First Affiliated Hospital of USTC has exempted this study from ethical approval requirements.

Competing interests

The authors declare no competing interests.