Comparative Outcomes of Drug-Eluting Stents Versus Drug-Eluting Balloons in Patients with Chronic Kidney Disease: A Systematic Review of Current Evidence

Eligibility criteria

Study designs

We included Randomized Controlled Trials (RCTs) and non-randomized compara-tive studies (prospective or retrospective cohort studies) that directly compared drug-eluting balloons (DEB) with drug-eluting stents (DES) in the target population. RCTs were prioritized for inclusion due to their superior methodological rigor in minimizing selection and performance bias. Non-randomized studies were included to maximize clinical relevance and capture real-world evidence, particularly for populations or sce-narios underrepresented in RCT data. We excluded case reports, case series, single-arm studies, editorials, commentaries, and conference abstracts without subsequent peer-reviewed publication to ensure methodological quality and minimize reporting bias.

Participants

We included studies enrolling adults (≥18 years) with chronic kidney disease (CKD) undergoing percutaneous coronary intervention (PCI). CKD was defined according to the classification used in individual studies, with particular focus on the most common definition of estimated glomerular filtration rate (eGFR) <60 mL/min/1.73 m² and/or dialysis dependence. Studies with mixed or heterogeneous populations were included only if separate outcome data for the CKD subgroup were reported or if ≥80% of participants met CKD criteria. We excluded studies that did not provide specific CKD data or where CKD patients represented <20% of the study population, as these would not adequately address our research questions specific to this vulnerable population.

Interventions and comparators

The intervention of interest was any drug-eluting balloon (DEB), regardless of drug agent (e.g., paclitaxel, sirolimus), coating technology, or balloon brand. The comparator was drug-eluting stents (DES) of any generation (first-generation sirolimus-eluting, paclitaxel-eluting; second-generation everolimus-eluting, zotarolimus-eluting) or newer generations. We excluded studies comparing DEB with bare-metal stents (BMS) alone, as BMS is no longer standard of care and would not reflect contemporary clinical practice.

Outcomes

Primary outcomes included: (1) all-cause mortality (combining cardiovascular and non-cardiovascular deaths); (2) major adverse cardiovascular events (MACE), defined as the composite of death (all-cause or cardiovascular), myocardial infarction (MI), and target lesion/vessel revascularization (TLR/TVR).

Secondary outcomes included: (1) cardiovascular-specific mortality; (2) target lesion revascularization (TLR), defined as any repeat revascularization procedure of the index lesion due to symptoms, objective evidence of ischemia, or restenosis ≥50%; (3) target vessel failure (TVF), defined as cardiac death, MI, or TVR in the target vessel; (4) stent thrombosis (for DES arm); (5) adverse events including clinically significant bleeding, stroke, and infection.

Data were collected for the longest available follow-up period reported in each study, as well as at standardized intervals (6, 12, 24, and 36 months) when available to enable subgroup temporal analyses.

Exclusion criteria

We excluded studies meeting any of the following criteria: (1) non-English language publications (due to resource constraints); (2) no specific comparison between DEB and DES; (3) absence of outcome data relevant to our primary or secondary outcomes; (4) studies with <5 participants per arm, as these lack adequate statistical power; (5) studies with no CKD-specific data or in populations without CKD; (6) publications that were review articles, conference proceedings without full-text availability, or gray literature without peer-review. This approach ensures methodological quality while being pragmatic about available resources.

Information sources

Database searching

We conducted a comprehensive systematic search of five major electronic databases from inception through September 2025: PubMed/MEDLINE, Scopus, Web of Science, and ScienceDirect. These databases were selected to capture the breadth of published literature across medical and biomedical databases, ensuring comprehensive coverage of indexed publications. No language restrictions were applied during the initial search to maximize sensitivity, though we subsequently limited inclusion to English-language publications due to resource constraints. The initial search yielded 4,093 records (PubMed n=530, Scopus n=2,245, Web of Science n=909, ScienceDirect n=409), demonstrating the substantial literature landscape on this topic.

To minimize publication bias and capture completed but unpublished studies, we systematically searched grey literature and other sources:

This comprehensive grey literature approach helps reduce potential bias toward published studies with statistically significant findings.

Search strategy

Search string development

We developed a comprehensive search strategy in collaboration with a medical librarian using controlled vocabulary (Medical Subject Headings—MeSH terms) combined with free-text terms to capture variations in terminology and nomenclature. The search strategy incorporated three conceptual domains:

Truncation and boolean operators were applied systematically to ensure sensitivity. The complete search strategies for each database, including MeSH terms specific to each platform, are detailed in Supplementary Material 1. Notably, search strategies were adapted to accommodate database-specific syntaxes and indexing practices (e.g., PubMed uses MeSH subject headings, while Scopus uses Emtree descriptors).

Search strategy validation

To ensure search strategy sensitivity and specificity, we performed preliminary searches and validated the strategy by confirming that known relevant studies (identified through preliminary scoping) were retrieved. This validation process helps ensure the search was comprehensive without excessive retrieval of irrelevant citations.

Selection process

Screening and deduplication phase

All retrieved records were imported into rayyan.ai (a web-based systematic review management platform) for automated and manual deduplication. Title and Abstract Screening: Four independent reviewers (BSW, MS, SII, ANF), trained using a stand-ardized screening manual, independently reviewed titles and abstracts of all articles using rayyan.ai. A broad inclusion approach was used at this stage to maximize sensitivity and minimize the risk of excluding relevant studies. Any article deemed potentially relevant by any reviewer proceeded to full-text review. Full text review conducted ac-cording to prespecified inclusion and exclusion criteria. Authors documented reasons for exclusion such as wrong population, wrong intervention/comparison, wrong outcome measure, no comparative data, wrong publication type, unclear intervention, and full-text unavailable despite contact attempts. This transparent documentation of exclusion reasons facilitates assessment of review bias and allows future researchers to identify studies excluded for potentially remediable reasons. All studies which met all eligibility criteria were included in the qualitative synthesis.

Agreement and conflict resolution

Disagreements between reviewers at the full-text stage were resolved through discussion and consensus. When consensus could not be reached, a senior adjudicator (the primary investigator or senior methodologist) made the final decision. This hierarchical resolution process balances reviewer autonomy with the need for consistent application of eligibility criteria. All disagreements and their resolutions were documented in a standardized form to provide transparency regarding how methodological decisions were made and to identify patterns in reviewer agreement for future investigator training.

Data collection process

Data extraction

Data extraction was performed independently and in duplicate by four trained reviewers using a pre-piloted, standardized data extraction form programmed into Rayyan.ai. Before extracting data from included studies, reviewers pilot-tested the form on 2-3 representative studies to ensure consistency of interpretation and completeness of the form.

Type of Data Extracted

Study-Level Characteristics:

Participant Characteristics:

Intervention Characteristics:

Outcome Data:

Missing data management

When outcome data were incomplete or unclear, reviewers contacted corresponding authors via email with specific requests for unpublished or supplementary data. Requests included patient-level data when available, event counts by follow-up period, and clarification of outcome definitions. We allowed 4 weeks for author responses before proceeding with available data. The frequency and nature of missing data from author contact attempts were documented.

Data extraction agreement

Discrepancies between independent extractions were identified through comparison and resolved through discussion and consensus. Any unresolved discrepancies were adjudicated by a senior reviewer. The percentage agreement and specific categories with high disagreement rates were documented to inform quality assurance.

Data items

Extracted data were organized into two main evidence tables:

Table 1 about study Characteristics consist of summarizes study design, setting, participant demographics, CKD criteria, and comorbidity profiles. This table enables readers to assess population applicability and identify sources of heterogeneity.

Table 2 about intervention and Outcome Data: Presents specific details of the DEB and DES interventions (type, drug, dosing), clinical outcomes with event counts and effect estimates (HR/RR/OR with 95% CI), and follow-up durations. This facilitates outcome synthesis and identification of heterogeneity sources.

Study risk of bias assessment

Assessment tools and approach

All included studies (n=5) underwent independent duplicate risk-of-bias (RoB) assessments by four trained reviewers using validated Cochrane tools. For the RCT (n=1), Cochrane RoB 2 evaluated five domains—randomization process, deviations from intended interventions, missing data, outcome measurement, and selective reporting—judged as low risk, some concerns, or high risk; overall risk categorized accordingly (low if all low; some concerns/high otherwise). For non-randomized studies (n=4), ROBINS-I assessed seven domains—confounding, selection, intervention classification, deviations from intended interventions, missing data, outcome measurement, and selective reporting—rated low/moderate/serious/critical risk, with overall risk determined algorithmically from domain judgments.

Disagreement resolution

Disagreements in bias assessment between reviewers were resolved through discussion and consensus. Persistent disagreements were adjudicated by a senior investigator (primary investigator or experienced methodologist).

Effect measures

Not applicable. Due to substantial heterogeneity in study design, CKD definitions, intervention types, and outcome measurement timing (see Section 2.9), quantitative meta-analysis was deemed inappropriate. Therefore, no effect measures were pre-specified for meta-analysis. However, we systematically extracted and presented effect estimates (Hazard Ratios, Risk Ratios, Odds Ratios with 95% confidence intervals) from individual studies in evidence tables to facilitate narrative synthesis.

Synthesis methods

We conducted a narrative synthesis instead of quantitative meta-analysis due to substantial clinical, methodological, statistical heterogeneity, and limited included studies (n=5), which precluded valid pooled estimates, subgroup analyses, or meta-regression. Clinical heterogeneity arose from variations in CKD definitions (e.g., eGFR cutoffs vs. dialysis status), severity/comorbidity profiles, target lesion characteristics (vessel size, complexity, location), and interventions (DES generations, DEB agents). Methodological differences included mixed designs (RCTs vs. observational cohorts), inconsistent confounder adjustment, and unharmonized risks of selection bias/confounding, while statistical heterogeneity involved divergent outcome definitions/timing, follow-up durations (6 months to >3 years), and composite outcome reporting.

The narrative synthesis employed a structured, pre-specified, reproducible approach per the registered protocol, integrating evidence tables, descriptive comparisons, bias assessments, consistency evaluations, and meta-narrative synthesis to address heterogeneity constraints. Evidence tables ( Tables 1-2) systematically organized study characteristics (designs, populations, interventions) and outcomes for visual pattern/anomaly identification. Descriptive comparisons qualitatively assessed outcome findings, effect directions/magnitudes with 95% CIs, while integrating RoB assessments via sensitivity analyses (low- vs. high-bias studies) and discussions of bias influences. Consistency evaluations examined directional agreement, effect magnitude overlap in CIs, and outliers; meta-narrative elements explained heterogeneity sources, clinical implications, and contextualized findings against existing literature/guidelines.

Study selection

A total of 4,093 records were identified through database searching (PubMed 530, Scopus 2,245, Web of Science 909, ScienceDirect 409) as showed in Figure 1. After removal of 2,914 duplicates, 1,179 unique records underwent title and abstract screening. Of these, 1,103 records were excluded as clearly not meeting the eligibility criteria. Seventy six full text articles were retrieved and assessed for eligibility, resulting in exclusion of 71 reports for the following reasons: wrong population (n = 30), wrong comparison (n = 22), wrong publication type (n = 2), wrong intervention (n = 1), wrong outcome (n = 4), no comparison group (n = 9), and no full text available (n = 3). Five studies met all inclusion criteria and were included in the final qualitative synthesis.

Figure 1. PRISMA flowchart.

Study and patient characteristics

Five comparative studies (four retrospective cohort studies and one randomized controlled trial) comprising CKD or ESRD populations undergoing PCI with DEB versus DES were included (Table 1). The studies were conducted in Japan, South Korea, Poland (multicenter), and Switzerland (multicenter), reflecting predominantly East Asian and European healthcare settings. Total sample sizes ranged from 400 to 1,616 participants, with mean ages between 64.2 and 69.1 years where reported, and a consistently high proportion of male patients (68.2%–78.4%).

One study exclusively enrolled ESRD patients on dialysis (Funayama et al., 2023), whereas others included broader CKD populations defined by reduced eGFR or reported as “renal failure.” In Lee et al. (2024), CKD was explicitly defined as eGFR < 60 ml/min per 1.73 m² based on the four component MDRD equation, with CKD present in 28.3% of the cohort. The proportion of patients with CKD or ESRD ranged from 6.7% in the largest South Korean cohort (Joh et al., 2024) to 100% in the dialysis only Japanese cohort, with intermediate values of 14.9% and 18.9% reported in the Swiss and Polish multicenter studies, respectively.

Intervention characteristics

Across all five studies (Table 2), the comparator arm used contemporary new generation or second generation DES, while the intervention arm used paclitaxel coated balloons as the DEB strategy. New generation thin strut DES or specific everolimus or zotarolimus eluting platforms were most commonly employed, reflecting current PCI practice patterns. DEB therapy was uniformly based on paclitaxel coated balloons, with device details (manufacturer and drug coating) reported consistently but with some variation in stent platforms between studies.

The relative allocation of patients to DES and DEB arms was unbalanced in the retrospective cohorts, with DES used more frequently than DEB (for example, 312 vs. 88 patients in Funayama et al., 409 vs. 219 in Lee et al., 381 vs. 465 in Wańha et al., and 1,478 vs. 138 in Joh et al.). In contrast, the randomized non inferiority trial by Jeger et al. (2018) enrolled similar numbers of patients in each arm (376 DES vs. 382 DEB).

Clinical outcomes

Because of clinical and methodological heterogeneity and limited number of studies, a formal quantitative meta analysis, subgroup analysis, sensitivity analysis, and publication bias assessment were not performed; instead, outcomes are presented descriptively as in Table 2.

Target lesion revascularization and target vessel failure

Table 2 showed that Target lesion revascularization (TLR) rates were generally comparable between DEB and DES in studies focusing on dialysis or small vessel disease, but some cohorts suggested differences in target lesion or vessel failure. In the dialysis only cohort by Funayama et al., TLR occurred in 14.1% of DES treated lesions versus 14.7% of DEB treated lesions, with no statistically significant difference (P = 0.864). In the Polish multicenter cohort (Wańha et al., 2024), adjusted analyses favored DES over DEB for TLR, with an adjusted hazard ratio of 0.20 (95% CI 0.06–0.69), indicating a lower risk of repeat revascularization with thin strut new generation DES in that setting.

Target lesion or target vessel failure (TLF/TVF) outcomes showed variable patterns across studies. In Lee et al. (2024), TLF occurred in 13.7% of DES treated patients versus 26.9% of DEB treated patients, corresponding to an adjusted hazard ratio of 0.60 (95% CI 0.27–1.33), suggesting a numerically lower but not statistically significant risk with DES. Conversely, in the cohort by Joh et al. (2024), two year TVF was reported as 27.8% in the DES group compared with 14.3% in the DEB group (HR 2.5, 95% CI 0.29–20), though the wide confidence interval indicates substantial imprecision.

Major Adverse Cardiovascular Events (MACE)

Two studies explicitly reported MACE as a composite endpoint. In the Japanese dialysis cohort (Funayama et al., 2023), MACE rates were identical between groups (17% in DES vs. 17% in DEB), indicating no apparent difference in composite clinical outcomes in this high risk ESRD population. Similarly, the randomized trial by Jeger et al. (2018) found MACE rates of 10% in both DES and DEB arms, with a hazard ratio of 0.99 (95% CI 0.38–2.57), supporting non inferiority of DEB compared with second generation DES in a mixed coronary population that included patients with renal impairment ( Table 2).

Thrombosis and other safety outcomes

Data on stent or lesion thrombosis were sparsely reported but showed low event rates overall. In Funayama et al., target lesion thrombosis occurred in 2.2% of DES treated patients and in 0% of DEB treated patients, corresponding to seven events in the DES arm and none in the DEB arm, though the study was not powered to detect differences in rare thrombotic events. Other studies primarily reported composite endpoints such as TLF or TVF and did not consistently separate stent thrombosis or detailed bleeding outcomes in CKD subgroups ( Table 2).

Summary of risk of bias (Qualitative)

Risk of bias was heterogeneous across the included studies, with four retrospective cohort studies evaluated using ROBINS I and one randomized trial evaluated with RoB 2 (Table 1). The non randomized cohorts were generally at moderate to serious risk of bias due to confounding, selection of participants, and incomplete adjustment for baseline differences between DEB and DES groups. The randomized trial exhibited low to some concerns of bias in most domains, providing more robust comparative evidence but with limited CKD specific subgroup detail.

Limitations of evidence included in the review

Several important limitations affect the strength of conclusions from this systematic review. First, methodological heterogeneity is substantial: four studies employed retrospective cohort designs with inherent selection bias and confounding, while only one randomized trial with explicit non-inferiority endpoints was included.²⁰ Only the Jeger et al. (2018) study employed randomization; the remaining studies lacked propensity-score matching or multivariable adjustment for baseline differences, limiting causal inference.¹⁸

Second, CKD/ESRD definition and prevalence varied markedly across studies. Funayama et al. (2023) enrolled exclusively dialysis-dependent ESRD patients (100%), while Joh et al. (2024) included only 6.7% CKD prevalence, substantially limiting the ability to draw CKD-specific conclusions from the largest cohort.^16,17 CKD definitions ranged from explicitly defined eGFR <60 mL/min/1.73m² (Lee et al., 2024, using the four-component MDRD equation) to undefined “renal failure” terminology,^19,20 precluding severity-based subgroup analysis.

Third, outcome definitions and follow-up durations differed substantively among studies. Target lesion revascularization, target vessel failure, and MACE definitions varied, with heterogeneous follow-up durations limiting meta-analytic synthesis. The absence of standardized outcome definitions prevented quantitative pooling of results.

Fourth, four cohorts were explicitly at “moderate” risk of bias using ROBINS-I assessment, primarily due to confounding and selection bias from non-randomized designs. The unbalanced allocation of patients to DEB versus DES in retrospective cohorts (particularly extreme in Joh et al. with 1,478 DES vs. 138 DEB) introduces allocation bias and confounding by indication, as operators may have selected devices based on lesion characteristics or patient factors not uniformly reported.

Fifth, stent platform heterogeneity confounds interpretation. While all DEB arms used paclitaxel-coated balloons, DES platforms included both everolimus-eluting and zotarolimus-eluting second-generation devices (Lee et al., 2024; Wańha et al., 2024; Joh et al., 2024) and paclitaxel-eluting stents (Jeger et al., 2018; Funayama et al., 2023), introducing device-specific efficacy variation.^16–20 The single RCT (Jeger et al., 2018) compared paclitaxel-coated balloons to everolimus-eluting or paclitaxel-eluting second-generation DES, while retrospective cohorts may have included diverse or older-generation stent platforms not specified in baseline characteristics.²⁰

Sixth, insufficient CKD-subgroup analysis within studies limits interpretation. While all studies enrolled CKD/ESRD patients, none performed prespecified subgroup analysis stratified by CKD stage (eGFR thresholds), dialysis dependence, or baseline renal function, precluding assessment of whether DEB or DES advantage/disadvantage varies with renal dysfunction severity.

Seventh, procedural variables potentially influencing outcomes were incompletely reported. Critical DEB deployment parameters (transit time, balloon-to-artery ratio, inflation pressure) were not consistently documented in the studies, limiting assessment of whether technical factors—known from Villar-Matamoros et al. (2022) to substantially affect paclitaxel delivery—explained outcome heterogeneity.²⁵ Similarly, predilation adequacy, intravascular ultrasound guidance, and rotational atherectomy use for calcific lesions were not uniformly reported, yet substantially influence restenosis risk in calcified CKD vessels based on Murphy et al. (2023)'s findings of 5.2–34.5% dissection rates depending on technique.¹⁵

Limitations of review processes

The review methodology employed PRISMA 2020 guidelines but faced inherent limitations. Formal quantitative meta-analysis was explicitly not performed due to clinical and methodological heterogeneity, instead employing qualitative synthesis with descriptive outcome tabulation. This conservative approach prevents meta-analytic bias but sacrifices statistical power to detect true treatment effects. Publication bias assessment was not undertaken, limiting confidence that published results represent all conducted analyses. The review did not perform GRADE certainty-of-evidence assessment, which would provide transparency regarding confidence in summary estimates. Only five studies met inclusion criteria despite screening 4,093 initial records, resulting in a limited evidence base. Furthermore, no authors were contacted for unreported subgroup analyses or missing data, potentially introducing information loss. The review's geographic limitation to East Asian and European cohorts may limit generalizability to other healthcare settings with different interventional practices, stent availability, or antiplatelet therapy protocols.

Implications for practice, policy, and future research

Clinical practice implications

Current findings confirm MACE equivalence between DEB and DES in high-risk ESRD (Funayama et al.: 17% both) and renal-inclusive RCT (Jeger et al.: 10% both), corroborating Aoki et al. (2017)'s assertion of uremic pathophysiology's dominance over device choice.^16,20,23 In moderate-severe CKD/ESRD undergoing PCI, prioritize lesion morphology, vessel size, and bleeding risk over renal status alone, as both modalities yield comparable MACE and mortality. DEB noninferiority to DES in small vessels (<2.75–3 mm), with DEB advantages in myocardial infarction prevention, bleeding mitigation, and abbreviated dual antiplatelet therapy (1–3 vs. 6–12 months)—pertinent for uremic platelet dysfunction and hemodialysis anticoagulation; ESRD thrombosis disparity (Funayama: 0% DEB vs. 2.2% DES) reinforces this.¹⁶

Policy implications

The current review and literature advocate CKD/ESRD-specific PCI registries stratifying DEB versus DES outcomes with standardized metrics and staging to yield pragmatic evidence beyond scarce RCTs. Mandate operator training/certification for DEB in complex calcified CKD anatomy, given technique. Implement reimbursement parity for DEB matching DES, reflecting equivalent/superior CKD efficacy—especially bleeding/small-vessel benefits—despite current disparities.

Future research directions

Prospective RCTs enrolling CKD/ESRD patients with eGFR (<15, 15–29, 30–59 mL/min/1.73 m²), dialysis, and calcification stratification (SYNTAX score) are essential to compare contemporary DEB/DES, standardizing MACE, dual antiplatelet therapy, and hemodialysis anticoagulation while capturing bleeding/thrombosis/revascularization endpoints. Mechanistic inquiries should probe uremic endothelial recovery kinetics post-DEB/DES, neointimal histology in CKD restenosis models, and paclitaxel responses in smooth muscle cells to discern if DEB scaffold avoidance enhances healing amid 2.78–3.45-fold CKD mortality escalation.

Ex vivo/in vivo pharmacokinetic assessments of paclitaxel delivery/retention in calcified vs. non-calcified CKD vessels via intravascular ultrasound/optical coherence tomography will guide predilation/balloon sizing for DEB comparability to DES in heavy calcification. Health economic evaluations must quantify DEB vs. DES cost-effectiveness incorporating antiplatelet/bleeding/hospitalization/quality-of-life burdens, alongside registries with uniform MACE/KDIGO criteria and CKD subgroups for pragmatic insights. Technical trials optimizing DEB predilation (intravascular ultrasound), ratios/pressures/inflation in stratified CKD/calcification will mitigate heterogeneity.