Comprehensive Cardiovascular Risk Profiling in People Living with HIV: Insights from the Japanese Adverse Drug Event Report Database

Abstract

Antiretroviral therapy (ART) has dramatically improved outcomes for people living with HIV (PLWH), yet concerns about cardiovascular disease (CVD) remain, especially in aging populations. In this study, we aimed to evaluate the association between ART regimens and CVD events in Japan using a nationwide pharmacovigilance database. We retrospectively analyzed reports from the Japanese Adverse Drug Event Report Database spanning April 2004 to September 2024. After removing duplicates and records with key missing data, 796,402 reports (Population A) were used for signal detection based on the reporting odds ratio (ROR) and information component (IC). A refined subset (Population B; 2,721 reports) underwent logistic regression to identify risk factors for major adverse cardiovascular events (MACE) and total cardiovascular events (MACE plus angina). ART regimen classes (e.g., integrase strand transfer inhibitors, non-nucleoside reverse transcriptase inhibitors, and protease inhibitors) and backbone therapies [e.g., abacavir (ABC)/lamivudine] were included in the analysis. Signal detection revealed significant ABC signals in both ROR and IC analyses for MACE and total CVD events. In logistic regression, advanced age (≥70 years), ABC-containing regimens, and diabetes emerged as independent risk factors for MACE and total CVD events. Dyslipidemia and hypertension were not significant in the adjusted models. Our findings underscore a potentially heightened cardiovascular risk associated with ABC, particularly in older PLWH or those with diabetes. These results highlight the need to consider individual CVD risk profiles when selecting ART regimens and reinforce the importance of ongoing pharmacovigilance to guide safer, more personalized treatment strategies worldwide.

Introduction

Antiretroviral therapy (ART) has transformed HIV from a fatal disease into a manageable chronic condition.^1
–3 ART significantly enhances both survival and quality of life for people living with HIV (PLWH) by suppressing viral replication and restoring immune function.^4,5 However, long-term ART is often accompanied by adverse effects and comorbidities, particularly cardiovascular disease (CVD).^6,7 Chronic inflammation and persistent HIV-driven immune activation contribute to an increased CVD risk and may accelerate atherosclerosis.⁸ Furthermore, older-generation regimens—such as certain protease inhibitors (PIs) and nucleoside reverse transcriptase inhibitors (NRTIs)—have been linked to metabolic disturbances, including dyslipidemia and insulin resistance.^9
–11 Notably, abacavir (ABC) has been associated with a 60% increased risk of cardiovascular events, especially myocardial infarction.¹²

Newer agents, including integrase strand transfer inhibitors (INSTIs) and tenofovir alafenamide (TAF), offer improved metabolic profiles and reduced renal and bone toxicity compared with previous therapies.^5,12
–14 Despite these advances, residual CVD risks persist in specific subgroups, particularly among older individuals.^14,15 Current guidelines recommend ART regimens that combine a key drug [INSTIs, non-nucleoside reverse transcriptase inhibitors (NNRTIs), or PIs] with a two-NRTI backbone. Options include tenofovir disoproxil fumarate/emtricitabine (TDF/FTC), TAF/emtricitabine (TAF/FTC), ABC/lamivudine (ABC/3TC), and zidovudine/3TC (AZT/3TC).^15
–17 TAF-based regimens are preferred for patients at higher risk of renal or bone complications.^15
–17 While evidence from the DISCOVER trial—originally focused on pre-exposure prophylaxis—supports TAF’s favorable safety profile,¹⁸ ABC-based backbones remain viable after thoroughly evaluating hypersensitivity risks and cardiovascular concerns.

The risk of major adverse cardiovascular events (MACE) in PLWH is influenced by treatment choices, patient-specific factors, and regional health care practices.^19
–21 In Japan, the adoption of newer agents such as TAF is driven by efforts to mitigate renal and bone toxicity, especially in an aging population that exhibits higher prevalences of comorbidities such as diabetes and dyslipidemia. These factors, combined with challenges in managing polypharmacy and drug–drug interactions, underscore the need for tailored interventions within Japan’s diverse health care landscape.^19,21,22

The Japanese Adverse Drug Event Report Database (JADER), managed by the Pharmaceuticals and Medical Devices Agency (PMDA), offers a unique opportunity to explore potential associations between ART and MACE despite inherent limitations such as underreporting and notoriety bias.^23
–25 This study leverages JADER data from April 2004 to September 2024 to investigate the relationship between long-term ART (≥6 months) and MACE. Employing reporting odds ratio (ROR)-based signal detection and logistic regression analyses, we aim to identify key risk factors for adverse cardiovascular outcomes. Also, the study aspires to inform global strategies for optimizing ART and mitigating CVD risks among aging PLWH by assessing total cardiovascular events (i.e., MACE plus angina).^26,27

Materials and Methods

This retrospective observational study was conducted and reported in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology guidelines. The study design, setting, participant selection, data collection, and statistical analysis have been described in detail, ensuring transparency and reproducibility.

Study design and data source

This retrospective observational study evaluated the association between ART and MACE and relevant risk factors in PLWH. In this study, MACE were defined as a composite of myocardial infarction, heart failure, and stroke. Total cardiovascular events were defined as MACE plus reports of angina. Data were obtained from the JADER database, a pharmacovigilance system maintained by the PMDA. JADER is updated monthly and comprises four main data tables: (1) a case list table containing patients’ demographic details (e.g., age, sex); (2) a drug information table that includes drug names and treatment initiation dates; (3) an adverse event information table that details adverse event types and their onset dates; and (4) an original disease table listing underlying conditions. The JADER database used in this study was downloaded directly from the PMDA website (https://www.pmda.go.jp/).

Data extraction and cleaning

Initially, 914,713 reports were extracted from JADER. In the data cleaning process, we excluded duplicate records, records with three or more key variables missing (such as age, sex, drug details, treatment initiation date, or adverse event onset date), and records containing logical inconsistencies (e.g., adverse event onset occurring before treatment initiation). This process yielded 796,402 reports (Population A), which were used for signal detection analysis (Fig. 1). This population includes all adverse event reports in JADER for the study period and serves as the background for identifying statistical signals. From this dataset, a refined subset of 2,721 reports (Population B) was selected for the logistic regression analysis, as detailed in the relevant section below (Fig. 1).

FIG. 1.

Flowchart illustrating report inclusion and exclusion criteria for signal detection and logistic regression. A total of 914,713 reports were initially extracted from the JADER database, with 118,311 excluded due to duplication, missing essential variables, or logical inconsistencies. Finally, 796,402 reports were used for signal detection (Population A), and 2,721 reports were eligible for logistic regression analysis (Population B). ART, antiretroviral therapy; CVD, cardiovascular disease; IC, information component; JADER, Japanese Adverse Drug Event Report Database; ROC, receiver operating characteristic.

Signal detection analysis

Signal detection was conducted using the ROR and the information component (IC). A signal was defined as significant if either the lower limit of the 95% confidence interval (CI) for the ROR exceeded 1 or if the IC minus two standard deviations (IC−2SD) was greater than 0.^23
–25 Only drugs with at least five cardiovascular reports were included in the analysis to avoid false-positive signals from low-frequency events.²⁷ This case threshold was chosen based on simulation studies demonstrating that a multi-source integration model can substantially improve performance. For example, recent simulations have shown that integrating literature-derived embeddings with spontaneous reporting system data increases the receiver operating characteristic area under the curve (ROC AUC) by up to 20% (from approximately 0.80 to 0.96) and reduces misclassification rates from 20% to 9%.^28,29 Such improvements support the robustness of our signal detection methodology even under conditions of data sparsity.

Logistic regression analysis

A second dataset (Population B) was derived from JADER to explore risk factors for cardiovascular events. Reports in Population B, numbering 2,721, were selected based on refined inclusion and exclusion criteria. These reports indicated ART use for at least 6 months at the time of adverse event onset, included at least two antiretroviral drugs (but fewer than five to avoid excessive polypharmacy), and provided sufficient patient information. In addition to drug-related data (i.e., key and backbone regimen classifications), patient demographics (age, sex) and comorbidities (diabetes, dyslipidemia, and hypertension) were included. Although CD4⁺ counts, smoking status, and treatment adherence may be important, these variables were not consistently available in the JADER database and were therefore omitted. Crude odds ratios (ORs) were first obtained via univariate analysis, and those variables demonstrating significance were included in a multivariate logistic regression model to compute adjusted ORs (aORs) with 95% CIs using the Wald test.

Statistical analysis

All statistical analyses were performed using JMP Pro version 17.2.0 (SAS Institute Inc., Cary, NC, USA). Descriptive statistics were used to summarize the dataset; for example, demographic variables were summarized using means, medians, and proportions, as appropriate. ROR and IC values were computed for signal detection, and logistic regression models were used for risk factor analysis. A p value of <.05 was considered significant.

Ethical considerations

JADER data are fully anonymized and publicly accessible; therefore, no formal ethical approval was required. This study adhered to the principles of the Declaration of Helsinki and relevant biomedical research guidelines.

Results

Baseline demographics of study populations

Population A, consisting of 796,402 reports from the entire JADER database, represented the dataset for signal detection analysis. In contrast, Population B, a refined subset of 2,721 reports pertaining to PLWH on ART, was used for the logistic regression analysis. In Population A, the age group of 20–69 years comprised the largest proportion (48.8%), followed by those aged 70 years or older (39.8%). The sex distribution was nearly balanced, with males accounting for 49.6% and females 47.6%, while 2.7% of the reports did not include sex information.

In contrast, Population B included 2,721 reports selected based on refined inclusion and exclusion criteria. The age group of 20–69 years constituted the majority (85.3%), with only 5.6% aged 70 years or older. The sex distribution in Population B was skewed toward males (87.8%), while females accounted for 10.7%, and 1.4% lacked sex information. INSTI-based regimens were the most frequently reported key drugs in Population B (43.4%), followed by PIs (36.7%) and NNRTIs (20.6%). Regarding backbone regimens, TDF/FTC was the most common (26.2%), followed by ABC/3TC (24.1%) and TAF/FTC (18.0%). Comorbidities in Population B included diabetes (4.8%), dyslipidemia (6.2%), and hypertension (6.1%), each considered explanatory variables given their association with cardiovascular risk (Table 1).

Table 1.

Baseline Characteristics of Population A and Population B

Variable	Population A		Population B
Variable	n	(%)	n	(%)
Overall total reports	796,402	(100.0%)	2,721	(100.0%)
Age
Age <20 years	51,311	(6.4%)	54	(2.0%)
20 years ≦ age <70 years	389,017	(48.8%)	2,322	(85.3%)
70 years ≦ age	317,096	(39.8%)	153	(5.6%)
Unknown	38,978	(4.9%)	192	(7.1%)
Sex
Male	395,096	(49.6%)	2,390	(87.8%)
Female	379,429	(47.6%)	292	(10.7%)
Unknown	21,877	(2.7%)	39	(1.4%)
Key drugs
INSTIs	—	—	1,181	(43.4%)
NNRTIs	—	—	561	(20.6%)
PIs	—	—	999	(36.7%)
Backbones
AZT/3TC	—	—	245	(9.0%)
ABC/3TC	—	—	657	(24.1%)
TDF/FTC	—	—	713	(26.2%)
TAF/FTC	—	—	490	(18.0%)
Primary diseases
Diabetes	—	—	131	(4.8%)
Dyslipidemia	—	—	170	(6.2%)
Hypertension	—	—	167	(6.1%)

Population A (n = 796,402) includes reports eligible for signal detection analysis, while Population B (n = 2,721) represents the subset used for logistic regression analysis. Values are presented as absolute counts (n) and proportions (%).

ART, antiretroviral therapy; INSTIs, integrase strand transfer inhibitors; NNRTIs, non-nucleoside reverse transcriptase inhibitors; PIs, protease inhibitors; AZT, zidovudine; 3TC, lamivudine; ABC, abacavir; TDF, tenofovir disoproxil fumarate; TAF, tenofovir alafenamide; FTC, emtricitabine.

Signal detection analysis

Signal detection analysis for MACE was conducted on Population A, comprising 796,402 reports from the JADER database. The results of ROR, 95% CI, IC, IC−2SD, and p values for each drug category are summarized in Table 2. Significant signals were identified for several drugs relating to MACE. ABC demonstrated a significant signal (ROR, 1.56; 95% CI, 1.23–1.97; p < .001). Etravirine (ETV) (ROR, 3.38; 95% CI, 1.52–7.52; p < .01) and saquinavir (SQV) (ROR, 2.90; 95% CI, 1.23–6.81; p < .05) also yielded significant results. Atazanavir (ATV) showed elevated values (ROR, 1.55; 95% CI, 1.02–2.35; p = .052) but did not achieve significance.

Table 2.

Signal Detection Analysis for MACE

Variable	n	ROR	[95% CI]	IC	(IC−2SD)	p value
INSTIs	84	1.01	[0.81–1.26]	0.01	(−0.73)	.911
RAL	21	0.72	[0.47–1.12]	−0.45	(−1.92)	.178
BIC	8	1.10	[0.54–2.25]	0.13	(−2.28)	.704
DTG	47	1.20	[0.89–1.61]	0.25	(−0.75)	.222
EVG	7	1.58	[0.73–3.40]	0.62	(−1.99)	.226
NNRTIs	39	0.91	[0.66–1.26]	−0.13	(−1.22)	.637
RPV	9	1.06	[0.54–2.07]	0.08	(−2.19)	.860
EFV	19	0.70	[0.44–1.11]	−0.49	(−2.04)	.137
ETV	7	3.38	[1.52–7.52]	1.60	(−1.11)	<.01 ^*
PIs	63	0.75	[0.58–0.96]	−0.40	(−1.26)	<.05 ^**
SQV	6	2.90	[1.23–6.81]	1.41	(−1.49)	<.05 ^**
NFV	7	0.76	[0.36–1.62]	−0.37	(−2.93)	.610
LPV	24	0.92	[0.61–1.39]	−0.11	(−1.50)	.840
ATV	24	1.55	[1.02–2.35]	0.59	(−0.81)	.052
FPV	6	1.19	[0.52–2.72]	0.24	(−2.55)	.645
DRV	26	1.24	[0.83–1.84]	0.29	(−1.05)	.267
NRTI	228	0.56	[0.49–0.64]	−0.80	(−1.25)	<.001 ^***
AZT	16	0.62	[0.37–1.01]	−0.67	(−2.36)	.053
ddI	6	0.72	[0.32–1.62]	−0.46	(−3.22)	.590
3TC	109	0.89	[0.74–1.08]	−0.15	(−0.80)	.284
ABC	74	1.56	[1.23–1.97]	0.60	(−0.20)	<.001 ^***
FTC	42	0.56	[0.41–0.77]	−0.80	(−1.84)	<.001 ^***
TDF	31	0.45	[0.31–0.64]	−1.12	(−2.33)	<.001 ^***
TAF	45	0.85	[0.63–1.14]	−0.23	(−1.24)	.322

Signal detection results for MACE are shown using ROR and IC, presented with 95% CI and IC−2SD. Significant results are highlighted in bold with asterisks.

p < 0.01.

p < 0.05.

***

p < 0.001 (analysis performed on Population A).

MACE, major adverse cardiovascular events; RAL, raltegravir; BIC, bictegravir; DTG, dolutegravir; EVG, elvitegravir; RPV, rilpivirine; EFV, efavirenz; ETV, etravirine; SQV, saquinavir; NFV, nelfinavir; LPV, lopinavir; ATV, atazanavir; FPV, fosamprenavir; DRV, darunavir; ddI, didanosine; IC, information component; ROR, reporting odds ratio; CI, confidence interval.

Signal detection analysis for total cardiovascular events, including MACE and angina, was also conducted on Population A. As summarized in Table 3, ABC exhibited significant signals in both ROR and IC analyses (ROR, 1.77; 95% CI, 1.42–2.20; IC, 0.77; IC−2SD, 0.03; p < .001). In addition, significant signals were observed for ETV (ROR, 3.18; 95% CI, 1.43–7.08; p < .001), SQV (ROR, 2.72; 95% CI, 1.16–6.41; p < .05), dolutegravir (ROR, 1.36; 95% CI, 1.03–1.78; p < .05), ATV (ROR, 2.20; 95% CI, 1.55–3.13; p < .001), and darunavir (ROR, 1.50; 95% CI, 1.05–2.14; p < .05).

Table 3.

Signal Detection Analysis for Total Cardiovascular Events

Variable	n	ROR	[95% CI]	IC	(IC−2SD)	p value
INSTIs	102	1.17	[0.95–1.42]	0.21	(−0.47)	.140
RAL	28	0.92	[0.63–1.35]	−0.11	(−1.40)	.779
BIC	9	1.18	[0.60–2.30]	0.22	(−2.06)	.579
DTG	56	1.36	[1.03–1.78]	0.41	(−0.50)	<.05 ^*
EVG	8	1.71	[0.83–3.53]	0.73	(−1.73)	.154
NNRTIs	42	0.93	[0.68–1.26]	−0.10	(−1.15)	.702
RPV	9	1.00	[0.51–1.95]	0.00	(−2.28)	1.000
EFV	22	0.77	[0.50–1.17]	−0.37	(−1.81)	.247
ETV	7	3.18	[1.43–7.08]	1.52	(−1.19)	<.05 ^*
PIs	80	0.90	[0.72–1.13]	−0.14	(−0.90)	.382
SQV	6	2.72	[1.16–6.41]	1.33	(−1.57)	<.05 ^*
NFV	7	0.72	[0.34–1.53]	−0.46	(−3.02)	.507
LPV	25	0.91	[0.61–1.36]	−0.13	(−0.13)	.768
ATV	35	2.20	[1.55–3.13]	1.06	(−1.49)	<.001 ^**
FPV	7	1.32	[0.61–2.84]	0.38	(−2.22)	.501
DRV	33	1.50	[1.05–2.14]	0.55	(−0.655)	<.05 ^*
NRTI	266	0.62	[0.55–0.70]	−0.66	(−1.08)	<.001 ^**
AZT	17	0.62	[0.38–1.00]	−0.67	(−2.31)	<.05 ^*
ddI	7	0.79	[0.37–1.69]	−0.32	(−2.88)	.727
3TC	124	0.96	[0.80–1.15]	−0.05	(−0.66)	.717
ABC	88	1.77	[1.42–2.20]	0.77	(0.03)	<.001 ^**
FTC	59	0.75	[0.58–0.98]	−0.39	(−1.27)	<.05 ^*
TDF	41	0.56	[0.41–0.77]	−0.80	(−1.85)	<.01 ^***
TAF	53	0.95	[0.72–1.25]	−0.08	(−1.01)	.783

Signal detection results for total cardiovascular events, including MACE and angina, are shown using ROR and IC, presented with 95% CI and IC−2SD. Significant results are highlighted in bold with asterisks.

p < 0.05.

p < 0.001.

***

p < 0.01 (analysis performed on Population A).

Logistic regression analysis

Multivariate logistic regression analysis was conducted to identify factors associated with MACE. Explanatory variables included key drug classes (e.g., INSTIs), backbone regimens (e.g., ABC/3TC), and demographic and clinical characteristics such as age, sex, and comorbidities. Focusing on ART regimen components rather than individual drugs, we aimed to reflect real-world clinical practices and evaluate the broader implications of treatment protocols on cardiovascular outcomes.

In the crude analysis, advanced age (≥70 years) was significantly associated with MACE (OR, 2.04; 95% CI, 1.07–3.89; p < .05), while individuals aged <70 years showed no significant association. Similarly, sex demonstrated no significant associations across categories in the crude analysis. Key drug classes (INSTIs, NNRTIs, and PIs) also showed no significant associations in the crude analysis. ABC/3TC demonstrated a significant crude association with MACE (OR, 2.66; 95% CI, 1.79–3.96; p < .001) among the backbone regimens. Comorbidities such as diabetes (OR, 3.59; 95% CI, 2.02–6.40; p < .001), hypertension (OR, 2.70; 95% CI, 1.53–4.78; p < .01), and dyslipidemia (OR, 2.01; 95% CI, 1.08–3.74; p < .05) were also associated with MACE in the crude analysis (Table 4).

Table 4.

Logistic Regression Analysis for MACE

Variable	n	Crude analysis			Multivariate analysis
Variable	n	OR	[95% CI]	p value	aOR	[95% CI]	p value
Age
Age <20 years	1	0.46	[0.06–3.40]	.722	—	—	—
20 years ≦ age <70 years	88	0.89	[0.52–1.50]	.672	—	—	—
70 years ≦ age	11	2.04	[1.07–3.89]	<.05 ^*	2.01	[1.03–3.93]	<.05 ^*
Unknown	5	0.65	[0.26–1.61]	.439	—	—	—
Sex
Male	97	1.71	[0.82–3.55]	.171	—	—	—
Female	6	0.49	[0.21–1.13]	.107	—	—	—
Unknown	2	1.35	[0.32–5.69]	.661	—	—	—
Key drugs
INSTIs	54	1.40	[0.95–2.07]	.108	—	—	—
NNRTIs	24	1.15	[0.72–1.83]	.540	—	—	—
PIs	33	0.78	[0.51–1.19]	.302	—	—	—
Backbones
AZT/3TC	7	0.71	[0.33–1.55]	.488	—	—	—
ABC/3TC	47	2.66	[1.79–3.96]	<.001 ^**	2.50	[1.68–3.73]	<.001 ^**
TDF/FTC	8	0.22	[0.11–0.46]	<.001 ^**	—	—	—
TAF/FTC	22	1.22	[0.75–1.97]	.437	—	—	—
Primary diseases
Diabetes	15	3.59	[2.02–6.40]	<.001 ^**	2.59	[1.39–4.84]	<.01 ^***
Dyslipidemia	12	2.01	[1.08–3.74]	<.05 ^*	1.26	[0.65–2.47]	.495
Hypertension	15	2.70	[1.53–4.78]	<.01 ^***	1.84	[0.98–3.45]	.059

Results of the univariate (crude) and multivariate (adjusted) logistic regression analyses for MACE. OR, aOR, 95% CI, and p values are shown. Significant results are highlighted in bold with asterisks.

p < 0.05.

p < 0.001.

***

p < 0.01 (analysis performed on Population B).

OR, odds ratio; aOR, adjusted odds ratio.

In the adjusted analysis, although sex did not demonstrate significant associations in the crude analysis, it was included as an explanatory variable in the adjusted model due to its clinical relevance in ART response and its potential influence on MACE. Advanced age (≥70 years) remained significantly associated with MACE (aOR, 2.01; 95% CI, 1.03–3.93; p < .05). Similarly, ABC/3TC maintained a significant association (aOR, 2.50; 95% CI, 1.68–3.73; p < .001). Diabetes remained a significant explanatory variable after adjustment (aOR, 2.59; 95% CI, 1.39–4.84; p < .01). In contrast, dyslipidemia (aOR, 1.26; 95% CI, 0.65–2.47; p = .495) and hypertension (aOR, 1.84; 95% CI, 0.98–3.45; p = .059) were no longer significant after adjustment.

Multivariate logistic regression analysis was also conducted to identify factors associated with total cardiovascular events, including MACE and angina. The explanatory variables, analysis framework, and reporting format mirrored those used in the MACE analysis, as described earlier.

In the crude analysis, advanced age (≥70 years) was significantly associated with total cardiovascular events (OR, 2.15; 95% CI, 1.20–3.84; p < .05), while individuals aged <70 years showed no significant association. Similarly, sex demonstrated no significant associations across categories. In the crude analysis, key drug classes, including INSTIs, did not show significant associations with total cardiovascular events. Among the backbone regimens, ABC/3TC demonstrated a significant crude association with total cardiovascular events (OR, 2.54; 95% CI, 1.77–3.64; p < .001). Comorbidities such as diabetes (OR, 4.10; 95% CI, 2.45–6.85; p < .001), dyslipidemia (OR, 2.42; 95% CI, 1.42–4.13; p < .05), and hypertension (OR, 3.47; 95% CI, 2.13–5.66; p < .001) were also associated with total cardiovascular events in the crude analysis (Table 5).

Table 5.

Logistic Regression Analysis for Total Cardiovascular Events

Variable	n	Crude analysis			Multivariate analysis
Variable	n	OR	[95% CI]	p value	aOR	[95% CI]	p value
Age
Age <20 years	1	0.37	[0.05–2.73]	.517	—	—	—
20 years ≦ age <70 years	108	0.87	[0.54–1.42]	.610	—	—	—
70 years ≦ age	14	2.15	[1.20–3.84]	<.05 ^*	2.05	[1.12–3.76]	<.05 ^*
Unknown	6	0.63	[0.27–1.45]	.376	—	—	—
Sex
Male	118	1.51	[0.81–2.83]	.216	—	—	—
Female	9	0.61	[0.31–1.22]	.189	—	—	—
Unknown	2	1.09	[0.26–4.56]	.708	—	—	—
Key drugs
INSTIs	63	1.26	[0.88–1.79]	.204	—	—	—
NNRTIs	26	0.97	[0.62–1.51]	1.000	—	—	—
PIs	42	0.82	[0.57–1.20]	.350	—	—	—
Backbones
AZT/3TC	7	0.57	[0.26–1.23]	.205	—	—	—
ABC/3TC	56	2.54	[1.77–3.64]	<.001 ^***	2.34	[1.66–3.45]	<.001 ^***
TDF/FTC	17	0.41	[0.25–0.69]	<.001 ^***	—	—	—
TAF/FTC	29	1.34	[0.88–2.05]	.196	—	—	—
Primary diseases
Diabetes	20	4.10	[2.45–6.85]	<.001 ^***	2.74	[1.57–4.80]	<.001 ^***
Dyslipidemia	17	2.42	[1.42–4.13]	<.05 ^*	1.44	[0.80–2.58]	.225
Hypertension	22	3.47	[2.13–5.66]	<.001 ^***	2.28	[1.32–3.94]	<.001 ^***

Results of the univariate (crude) and multivariate (adjusted) logistic regression analyses for total cardiovascular events, including MACE and angina. OR, aOR, 95% CI, and p values are shown. Significant results are highlighted in bold with asterisks.

p < 0.05.

***

p < 0.001 (analysis performed on Population B).

In the adjusted analysis, as in the MACE analysis, sex was included as an explanatory variable due to its clinical relevance in ART response and its potential impact on total cardiovascular events despite showing no significant associations in the crude analysis. Advanced age (≥70 years) remained significantly associated with total cardiovascular events (aOR, 2.05; 95% CI, 1.12–3.76; p < .05). Similarly, ABC/3TC maintained a significant association (aOR, 2.34; 95% CI, 1.66–3.45; p < .001). Among comorbidities, both diabetes (aOR, 2.74; 95% CI, 1.57–4.80; p < .001) and hypertension (aOR, 2.28; 95% CI, 1.32–3.94; p < .001) retained significant associations after adjustment. In contrast, dyslipidemia (aOR, 1.44; 95% CI, 0.80–2.58; p = .225) was no longer significant.

Discussion

Significance of using JADER in evaluating ART-associated cardiovascular risk

The JADER database is an essential real-world resource for investigating potential associations between ART and cardiovascular events. Spontaneous reporting systems are subject to significant biases (e.g., underreporting) and lack a defined denominator,³⁰ making it nearly impossible to infer causality. However, they remain a powerful tool for hypothesis generation, allowing for the detection of potential safety signals that warrant further investigation in rigorous epidemiological studies. Epidemiological evidence indicates that HIV-positive individuals have a two-fold increased risk of CVD—with a growing event burden over recent decades.³¹ Moreover, long-term ART use has been associated with metabolic disturbances such as dyslipidemia and insulin resistance, which further exacerbate cardiovascular risk.^32,33 In Japan, where newer-generation regimens (e.g., INSTI-based and TAF/FTC-based therapies) have been implemented early, and a significant proportion of PLWH are older adults, JADER-based analyses are particularly valuable for generating region-specific risk assessments and tailoring preventive strategies.^34
–36

Interpreting the signal detection analysis

Robust signal detection is central to effective pharmacovigilance. In our study, we employed both the ROR and the IC to assess drug–event relationships. As van de Ven et al.³⁰ noted, while the ROR is sensitive to data sparsity, the Bayesian-based IC method adjusts for variability, enhancing specificity. Although we did not apply these more advanced approaches [e.g., generalized linear mixed models (GLMM), hierarchical Bayesian models] in the present analysis, prior simulation studies have demonstrated that such methods can reduce estimation bias and the root-mean-square error (RMSE) by over 50% compared with conventional models.²⁸ Integrating literature-derived embeddings with spontaneous reporting system data in a Bayesian framework can also improve the AUC by up to 20%.^28,29 These findings suggest potential future enhancements for pharmacovigilance studies, especially in heterogeneous data environments, but our current approach (ROR/IC) was deemed sufficient, given the scope and objectives of this work.

The case number threshold in signal detection

A critical methodological consideration in spontaneous reporting systems is establishing a minimum case threshold to minimize spurious signals from low-frequency events. Our analysis excluded drug–event pairs with fewer than five reported cases. This threshold is well-supported by simulation studies. Bagozzi et al.²⁹ demonstrated that employing a binary response (two-source) model increased the ROC AUC from 0.65 to 0.78 when combining data from the Social Conflict Analysis Database and the Integrated Crisis Early Warning System while reducing the misclassification rate from 20% to approximately 9%. Similarly, Mower et al.²⁸ reported that integrating literature-derived embeddings with Spontaneous Reporting System data boosted the ROC AUC from approximately 0.80 (for Disproportionality Measures-only models) to a maximum of 0.96, with the integrated model achieving a 12% improvement when more than 50% of the learning data was used. These quantitative improvements illustrate that setting a robust case threshold and employing multi-source integration significantly enhance model performance and reliability.

Interpreting risk factors from logistic regression and mechanisms of ABC-associated cardiovascular risk

Our multivariate logistic regression analyses identified that advanced age (≥70 years), diabetes, and using ABC/3TC-containing regimens were independent predictors of MACE. These results align with previous studies that have linked ABC exposure to increased cardiovascular risk via mechanisms including enhanced platelet activation, systemic inflammation, and endothelial dysfunction.^30,31,37 Raggi et al.³⁷ and Obirikorang et al.³³ have documented that ABC correlates with adverse changes in lipid profiles and impaired endothelial function. In addition, Brites-Alves et al.³⁸ demonstrated that increased levels of inflammatory biomarkers, such as Interleukin-6 and Tumor Necrosis Factor-alpha, are closely associated with immune activation in patients with HIV, further substantiating the role of chronic inflammation in promoting cardiovascular risk. More recent mechanistic studies further support these observations; for instance, Yan et al.³⁹ provided evidence that ABC-based regimens are associated with increased prothrombin conversion, contributing to a prothrombotic state, while Falcinelli et al.⁴⁰ demonstrated that ABC induces platelet hyperreactivity. Notably, while hypertension was a significant factor in the crude analysis, it did not retain significance in the adjusted model for MACE (p = 0.059). This might be attributable to confounding effects from stronger predictors in the model, such as advanced age and diabetes, or to the specific reporting patterns within the JADER database. Collectively, these findings reinforce the biological plausibility of our results and highlight the need for further research to elucidate the underlying molecular pathways.

Clinical and public health implications

The clinical implications of our findings are substantial. Given the significantly elevated cardiovascular risk observed among older adult PLWH—especially those with comorbid diabetes or receiving ABC/3TC-based regimens—clinicians need to consider alternative ART regimens, such as those based on TAF/FTC or INSTI-based therapies, which are associated with more favorable metabolic profiles.^31,35,40 Furthermore, integrating routine cardiovascular risk assessments—including evaluations of lipid levels, blood pressure, and inflammatory markers—into HIV care protocols is crucial for early detection and management of cardiovascular complications.

Recent successes in integrated care models underscore the benefits of multidisciplinary approaches. Davis et al.⁴¹ reported that optimizing the integration of hypertension services into existing HIV care—by leveraging pre-existing infrastructure and employing multidisciplinary teams—resulted in significant improvements in screening, diagnosis, and treatment outcomes in high-risk populations. Similarly, Oka et al.⁴² described how Japan’s AIDS Clinical Center implemented an integrated care model that combined ART with specialized interventions (e.g., hemostatic treatments for hemophiliac patients and comprehensive management of HIV/hepatitis C virus co-infection), leading to substantial improvements in clinical outcomes and shifts in mortality patterns away from AIDS-related causes. These successes illustrate that region-specific, integrated care programs, when combined with improved risk prediction (as evidenced by the 20% AUC improvement and over 50% reduction in misclassification reported by the integrated models^28,29) can meaningfully reduce cardiovascular events in HIV-infected populations.

Limitations

Despite the strengths of our study, several important limitations must be acknowledged. The JADER database is inherently subject to well-known issues such as underreporting and reporting biases. Crucially, the reported adverse events are not clinically adjudicated or validated against medical records, which could be a source of misclassification bias.^30,33 In addition, the lack of a clearly defined denominator for the exposed population prevents precise incidence rate calculations.^28,31 Taken together, these limitations preclude any definitive causal inferences from our findings. A primary limitation for our analysis was the absence of detailed clinical variables. For instance, crucial cardiovascular risk factors such as smoking status, along with key clinical markers for HIV management such as CD4⁺ counts, viral loads, and treatment adherence, were not available. Although advanced statistical techniques, such as GLMM and hierarchical Bayesian models, have been employed to mitigate these issues—achieving RMSE reductions of over 50% and AUC improvements of up to 20%²⁸—residual confounding remains a challenge. Furthermore, these findings are derived from a Japanese population, and their generalizability to other ethnic groups or health care systems may be limited, particularly given potential differences in the prevalence of cardiovascular risk factors and genetic backgrounds. Therefore, future research should validate these findings through cross-database comparisons (including the U.S. Food and Drug Administration Adverse Event Reporting System and EudraVigilance) and prospective cohort studies that offer more granular data. Moreover, integrating active surveillance methods and linking pharmacovigilance data with clinical registries may further improve the reliability of these signals.^29,41,43

Conclusions

This study, leveraging data from the JADER database, provides important insights into potential cardiovascular risks associated with specific antiretroviral therapies in Japan, where newer-generation regimens have been adopted relatively early, and a significant proportion of PLWH are older adults. Our findings indicate that ABC consistently produced significant signals across ROR and IC analyses and showed strong associations in the multivariate models, suggesting a heightened cardiovascular risk in certain subgroups. Clinicians should, therefore, exercise caution when prescribing ABC, particularly in patients of advanced age or those with diabetes. Ongoing pharmacovigilance, complemented by prospective investigations, is warranted to further clarify these associations and support the development of safer, more individualized ART regimens worldwide.

Footnotes

Acknowledgments

The authors have no additional non-author contributors to acknowledge.

Authors’ Contributions

S.H.: Conceptualization, formal analysis (lead), data curation, and writing—original draft. M.T.: Formal analysis (supporting) and data curation (supporting). H.K., T.T., and M.K.: Writing—review and editing. S.I.: Validation. All authors discussed the results, contributed to the interpretation of the findings, and approved the final version of the article.

Ethics Approval and Consent to Participate

This study utilized the JADER database, a publicly available, anonymized dataset from the PMDA. No formal ethical approval was required according to the Ethical Guidelines for Medical and Health Research Involving Human Subjects in Japan. All procedures followed the principles of the Declaration of Helsinki.

Availability of Data and Materials

All data generated or analyzed during this study are included in this published article.

Author Disclosure Statement

The authors declare no conflict of interest.

Funding Information

The authors received no specific financial or logistical support for this work.

References

McManus

, O’Connor

, Boyd

, et al.; Australian HIV Observational Database. Long-term survival in HIV positive patients with up to 15 years of antiretroviral therapy. PLoS One, 2012; 7(11):e48839; doi: 10.1371/journal.pone.0048839

UNAIDS. 90-90-90: An ambitious treatment target to help end the AIDS epidemic. Joint United Nations Programme on HIV/AIDS. Available from: https://www.unaids.org/sites/default/files/media_asset/90-90-90_en.pdf [Last accessed: January 7, 2025 ].

Weber

, Ruppik

, Rickenbach

, et al.; Swiss HIV Cohort Study (SHCS). Decreasing mortality and changing patterns of causes of death in the Swiss HIV Cohort Study. HIV Med, 2013; 14(4):195–207; doi: 10.1111/j.1468-1293.2012.01051.x

Eyawo

, Brockman

, Goldsmith

, et al. Risk of myocardial infarction among people living with HIV: An updated systematic review and meta-analysis. BMJ Open, 2019; 9(9):e025874; doi: 10.1136/bmjopen-2018-025874

Thet

, Siritientong

. Antiretroviral therapy-associated metabolic complications: Review of the recent studies. HIV AIDS (Auckl), 2020; 12:507–524; doi: 10.2147/HIV.S275314

Zhu

, Wang

, He

, et al. Higher cardiovascular disease risks in people living with HIV: A systematic review and meta-analysis. J Glob Health, 2024; 14:04078; doi: 10.7189/jogh.14.04078

Lundgren

, Babiker

, Gordin

, et al.; INSIGHT START Study Group. Initiation of antiretroviral therapy in early asymptomatic HIV infection. N Engl J Med, 2015; 373(9):795–807; doi: 10.1056/NEJMoa1506816

Zhang

, Jiang

, Li

, et al. Adjunct therapy for CD4+ T-cell recovery, inflammation and immune activation in people living with HIV: A systematic review and meta-analysis. Front Immunol, 2021; 12:632119; doi: 10.3389/fimmu.2021.632119

Sabin

, Worm

, Weber

, et al.; D:A:D Study Group. Use of nucleoside reverse transcriptase inhibitors and risk of myocardial infarction in HIV-infected patients enrolled in the D:A:D study: A multi-cohort collaboration. Lancet, 2008; 371(9622):1417–1426; doi: 10.1016/S0140-6736(08)60423-7

10.

Worm

, Sabin

, Weber

, et al. Risk of myocardial infarction in patients with HIV infection exposed to specific individual antiretroviral drugs from the 3 major drug classes: The data collection on adverse events of anti-HIV drugs (D:A:D) study. J Infect Dis, 2010; 201(3):318–330; doi: 10.1086/649897

11.

Rethy

, Feinstein

, Sinha

, et al. Coronary microvascular dysfunction in HIV: A review. J Am Heart Assoc, 2020; 9(1):e014018; doi: 10.1161/JAHA.119.014018

12.

Katlama

, Bisshop

, Bogner

, et al. Efficacy and safety of dolutegravir/lamivudine in virologically suppressed female participants: Week 48 data from the pooled TANGO and SALSA studies. HIV Med, 2024; 25(7):873–884; doi: 10.1111/hiv.13643

13.

Walmsley

, Smith

, Górgolas

, et al. Efficacy and safety of switching to dolutegravir/lamivudine in virologically suppressed people with HIV-1 aged ≥ 50 years: Week 48 pooled results from the TANGO and SALSA studies. AIDS Res Ther, 2024; 21(1):17; doi: 10.1186/s12981-024-00604-9

14.

Surial

, Chammartin

, Damas

, et al.; Swiss HIV Cohort Study. Impact of integrase inhibitors on cardiovascular disease events in people with human immunodeficiency virus starting antiretroviral therapy. Clin Infect Dis, 2023; 77(5):729–737; doi: 10.1093/cid/ciad286

15.

Gandhi

, Landovitz

, Sax

, et al. Antiretroviral drugs for treatment and prevention of HIV in adults: 2024 Recommendations of the International Antiviral Society-USA Panel. JAMA, 2025; 333(7):609–628; doi: 10.1001/jama.2024.24543

16.

European AIDS Clinical Society. EACS Guidelines Version 12.0. October, 2023. Available from: https://www.eacsociety.org [Last accessed: January 7, 2025 ].

17.

Panel on Antiretroviral Guidelines for Adults and Adolescents. Guidelines for the Use of Antiretroviral Agents in Adults and Adolescents With HIV. U.S. Department of Health and Human Services. Available from: https://clinicalinfo.hiv.gov/en/guidelines/adult-and-adolescent-arv [Last accessed: January 7, 2025 ].

18.

Ogbuagu

, Ruane

, Podzamczer

, et al.; DISCOVER study team. Long-term safety and efficacy of emtricitabine and tenofovir alafenamide vs emtricitabine and tenofovir disoproxil fumarate for HIV-1 pre-exposure prophylaxis: Week 96 results from a randomised, double-blind, placebo-controlled, phase 3 trial. Lancet HIV, 2021; 8(7):e397–e407; doi: 10.1016/S2352-3018(21)00071-0

19.

Ruzicka

, Imai

, Takahashi

, et al. Greater burden of chronic comorbidities and co-medications among people living with HIV versus people without HIV in Japan: A hospital claims database study. J Infect Chemother, 2019; 25(2):89–95; doi: 10.1016/j.jiac.2018.10.006

20.

Tanaka

, Onoda

, Ishii

. Understanding the actual use of anti-HIV drugs in Japan from 2016 to 2019: Demonstrating epidemiological relevance of NDB Open Data Japan for understanding Japanese Medical Care. Int J Environ Res Public Health, 2022; 19(19):12130; doi: 10.3390/ijerph191912130

21.

Choi

, Poonkasetwattana

, Phanuphak

, HIV Care Continuum & Beyond Committee . Overcoming challenges across the HIV care continuum in the Asia-Pacific region: Expert recommendations. BMJ Glob Health, 2023; 8(7):e012722; doi: 10.1136/bmjgh-2023-012722

22.

Nishiura

. Estimating the incidence and diagnosed proportion of HIV infections in Japan: A statistical modeling study. PeerJ, 2019; 7:e6275; doi: 10.7717/peerj.6275

23.

Noguchi

, Tachi

, Teramachi

. Review of statistical methodologies for detecting drug-drug interactions using spontaneous reporting systems. Front Pharmacol, 2019; 10:1319; doi: 10.3389/fphar.2019.01319

24.

Noguchi

, Yoshizawa

, Tachi

, et al. Effect of dipeptidyl peptidase-4 inhibitors vs. metformin on major cardiovascular events using spontaneous reporting system and real-world database study. J Clin Med, 2022; 11(17):4988; doi: 10.3390/jcm11174988

25.

Tian

, Yao

, He

, et al. Systematic analysis of safety profile for darunavir and its boosted agents using data mining in the FDA adverse event reporting system database. Sci Rep, 2021; 11(1):12438; doi: 10.1038/s41598-021-91549-w

26.

Napoli

, Wood

, Coumbis

, et al. No evident association between efavirenz use and suicidality was identified from a disproportionality analysis using the FAERS database. J Int AIDS Soc, 2014; 17(1):19214; doi: 10.7448/IAS.17.1.19214

27.

Lichtenstein

, Armon

, Buchacz

, et al.; HIV Outpatient Study (HOPS) Investigators. Low CD4+ T cell count is a risk factor for cardiovascular disease events in the HIV outpatient study. Clin Infect Dis, 2010; 51(4):435–447; doi: 10.1086/655144

28.

Mower

, Cohen

, Subramanian

. Complementing observational signals with literature-derived distributed representations for post-marketing drug surveillance. Drug Saf, 2020; 43(1):67–77; doi: 10.1007/s40264-019-00872-9

29.

Bagozzi

, Brandt

, Freeman

, et al. The prevalence and severity of underreporting bias in machine- and human-coded data. PSRM, 2019; 7(3):641–649; doi: 10.1017/psrm.2018.11

30.

van De Ven

, Pozniak

, Levi

, et al. Analysis of pharmacovigilance databases for Dolutegravir safety in pregnancy. Clin Infect Dis, 2020; 70(12):2599–2606; doi: 10.1093/cid/ciz684

31.

Shah

ASV

, Stelzle

, Lee

, et al. Global burden of atherosclerotic cardiovascular disease in people living with HIV: Systematic review and meta-analysis. Circulation, 2018; 138(11):1100–1112; doi: 10.1161/CIRCULATIONAHA.117.033369

32.

Dye

, Corley

, Li

, et al. Comparative DNA methylomic analyses reveal potential origins of novel epigenetic biomarkers of insulin resistance in monocytes from virally suppressed HIV-infected adults. Clin Epigenetics, 2019; 11(1):95; doi: 10.1186/s13148-019-0694-1

33.

Obirikorang

, Acheampong

, Quaye

, et al. Association of single nucleotide polymorphisms with dyslipidemia in antiretroviral exposed HIV patients in a Ghanaian population: A case-control study. PLoS One, 2020; 15(1):e0227779; doi: 10.1371/journal.pone.0227779

34.

Karanika

, Karantanos

, Carneiro

, et al. Development and validation of the HIV-CARDIO-PREDICT score to estimate the risk of cardiovascular events in HIV-infected patients. Cells, 2023; 12(4):523; doi: 10.3390/cells12040523

35.

Zicari

, Sessa

, Cotugno

, et al. Immune activation, inflammation, and non-AIDS co-morbidities in HIV-infected patients under long-term ART. Viruses, 2019; 11(3):200; doi: 10.3390/v11030200

36.

Vos

WAJW

, Vadaq

, Matzaraki

, et al. Cardiometabolic differences in people living with HIV receiving integrase strand transfer inhibitors compared to non-nucleoside reverse transcriptase inhibitors: Implications for current ART strategies. Viruses, 2024; 16(4):582; doi: 10.3390/v16040582

37.

Raggi

, Francesco

, Manicardi

, et al. Prediction of hard cardiovascular events in HIV patients. J Antimicrob Chemother, 2016; 71(12):3515–3518; doi: 10.1093/jac/dkw346

38.

Brites-Alves

, Luz

, Netto

, et al. Immune activation, proinflammatory cytokines, and conventional risks for cardiovascular disease in HIV patients: A case-control Study in Bahia, Brazil. Front Immunol, 2018; 9:1469; doi: 10.3389/fimmu.2018.01469

39.

Yan

, Huang

, van der Heijden

, et al. Abacavir use is associated with increased prothrombin conversion. Front Immunol, 2023; 14:1182942; doi: 10.3389/fimmu.2023.1182942

40.

Falcinelli

, Francisci

, Belfiori

, et al. In vivo platelet activation and platelet hyperreactivity in abacavir-treated HIV-infected patients. Thromb Haemost, 2013; 110(2):349–357; doi: 10.1160/TH12-07-0504

41.

Davis

, Perez-Guzman

, Hoyer

, et al. Association between HIV infection and hypertension: A global systematic review and meta-analysis of cross-sectional studies. BMC Med, 2021; 19(1):105; doi: 10.1186/s12916-021-01978-7

42.

Iqbal

, Salman

, Akhtar

, et al. Spectrum of dyslipidemias in treatment-naïve human immunodeficiency virus-infected patients presenting to an HIV clinic of a tertiary care hospital. Cureus, 2022; 14(2):e21972; doi: 10.7759/cureus.21972

43.

Mesquita

, Coelho

, Amancio

, et al. Severe infection increases cardiovascular risk among HIV-infected individuals. BMC Infect Dis, 2019; 19(1):319; doi: 10.1186/s12879-019-3894-6