We systematically searched the MEDLINE, Embase, and Cochrane Library databases for articles published until July 2025. The search strategy included terms such as "irreversible electroporation," "immunotherapy," and "locally advanced pancreatic cancer." The MeSH and input terms were adapted for the selected databases, combining terms with Boolean connectors (OR and AND) and conforming to the syntax rules in each database. Duplicate articles were manually excluded. References from all included studies, previous systematic reviews, and meta-analyses were also manually searched for additional studies, and reference manager software, Zotero^® (Version 7.0.3), was used.⁽¹²⁾ The search strategy for each database is presented in Table 1S, Supplementary Material. Our study was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines and the Cochrane Handbook for Systematic Reviews of Interventions.^(13,14)

Two reviewers independently assessed the initial search results to identify studies that met the eligibility criteria based on their titles and abstracts. The selected studies then underwent full-text assessment by the same reviewers, who made the final selection based on the inclusion and exclusion criteria. Any disagreements were resolved through discussion among all authors, with the lead reviewer making the final decision, if necessary.

Inclusion in this meta-analysis was restricted to studies that met the following eligibility criteria: (1) enrolled patients with LAPC, (2) compared IRE combined with systemic immunotherapy with IRE alone, (3) were clinical trials or observational studies, and (4) reported any of the outcomes of interest.

We excluded studies that: (1) used ablation methods other than IRE; (2) included intratumoral immunotherapy as treatment; (3) had no Control Group; (4) had overlapping patient populations, retaining only the study with the highest number of patients; (5) were reviews, case reports, editorials, correspondences, comments, or meeting abstracts; (6) used a murine animal model; (7) did not have available full texts; and (8) were written in languages other than English.

Two reviewers independently extracted relevant data from the selected studies in a standardized form. The extracted data included study characteristics and demographic data such as the first author, year of publication, study location, total sample size, sample size of the Control Group (irreversible electroporation alone arm), sample size of the Intervention Group (irreversible electroporation plus immunotherapy arm), immunotherapy drug regimen, sex and age (median and/or average) of the sample population, study design, single- or multi-center setup, follow-up time, and tumor diameter (Table 1). Any disagreement between the two reviewers was resolved by consensus with the assistance of a third reviewer.

The following raw statistics were extracted for data synthesis: hazard ratio (HR) for progression-free survival (PFS), overall survival (OS), carbohydrate antigen 19-9 (CA 19-9), and adverse events (AEs) in both groups. If Kaplan–Meier (KM) curves for PFS and OS were provided instead of HR and 95%CI, time-to-event data were extracted from the Kaplan–Meier curves using WebPlotDigitizer version 4.7 software.⁽¹⁵⁾ Subsequently, this data was used to calculate the HR and 95%CI using the R software package, "IPDfromKM."⁽¹⁶⁾

For CA 19-9, the values from Lin et al.⁽¹⁷⁾ and Lin et al.⁽¹⁸⁾ were extracted using WebPlotDigitizer, version 4.7.⁽¹⁵⁾ Data normality was first assessed using the method reported by Shi et al.⁽¹⁹⁾ Values from Lin et al.⁽¹⁸⁾ were transformed from median and interquartile range to mean and standard deviation using the method described by Wan et al.⁽²⁰⁾

Progression-free survival was defined as the period from the date of treatment initiation or baseline assessment to objective disease progression, subjective disease deterioration, or death, whichever occurred first. Overall survival was defined as the time from treatment initiation or baseline assessment to death. Progression-free survival was censored on the date of the last cancer assessment if no progression had occurred, and OS was censored at the time of the last follow-up for patient who were alive or lost to follow-up. Serum CA19-9 levels were evaluated at each follow-up using a quantitative sandwich enzyme immunoassay. Adverse events were defined as any unfavorable symptoms or diseases occurring after treatment initiation, including any new health issues or worsening of preexisting conditions, regardless of their relationship to the treatment.

We evaluated the risk of bias in the randomized controlled trials (RCTs) using the Cochrane Risk of Bias assessment tool (version 2).⁽²¹⁾ Non-randomized studies were assessed using the ROBINS-I ("Risk of Bias in Nonrandomized Studies of Interventions") tool.⁽²²⁾ Two authors independently assessed risk of bias. Disagreements were resolved by consensus after discussing the reasons for discrepancy. Testing for funnel plot asymmetry was not conducted because its power was too low to distinguish between chance and real asymmetry when fewer than 10 studies were included in the meta-analysis.⁽²³⁾

The statistical analysis was conducted using R software, particularly the "meta,""metafor," and "dmetar" packages.⁽¹⁶⁾ For the estimation of meta-analytic measures, an inverse variance estimator was used in a random-effects model. For the estimation of between-study variances (τ²) and considering the odds ratio, the Restricted Maximum Likelihood (REML) method⁽²⁴⁾ was applied. For the HR, the DerSimonian–Laird method⁽²⁵⁾ was applied. A p<0.05 was considered as the threshold of statistical significance. The results are presented as pooled estimates with 95%CI and plotted as forest plots.^(16,26)

Heterogeneity was assessed using the I2 statistic⁽²⁷⁾ and Q-test.⁽²⁶⁾ Sensitivity analysis was conducted by omitting one study from each analysis to evaluate the effect of each study on the overall result. The extracted data are summarized in tables 2 and 3.

The initial search identified 280 studies. After excluding 63 duplicates, 199 studies were excluded based on their titles and abstracts. Eighteen of the remaining studies were read in full. Among these, 14 were excluded because of population overlap, missing outcomes of interest, or inadequate Intervention or Control Groups This process resulted in four articles^{(17,18,28,29)} deemed eligible for analysis (Figure 1).

The selected studies resulted in a sample of 310 patients, with 187 in the Control Group and 123 in the Intervention Group. Among these patients, 48% were female and the mean age was 58.45 years. All the studies were conducted at different hospitals in China. The selected studies were published between 2017 and 2021. Two studies were RCTs, one was an observational retrospective study and the other was a prospective study. All studies were conducted at a single center. The median tumor diameter was 4.0cm in the Control Group and 4.2cm in the Intervention Group. The mean follow-up time was 17.5 months. The immunotherapy drug regimen was the same for the Interventional and Control Groups, but varied across studies. The characteristics of the individual studies are presented in table 1.

Pooled PFS indicated that the combination of IRE with immunotherapy effectively protected patients from disease progression compared to IRE alone with an HR of 0.56 (95%CI=0.39–0.80; p<0.01). Heterogeneity was not significant (I²=10%, p=0.33 (Figure 2).

Combination therapy of IRE and immunotherapy improved OS with a pooled HR of 0.52 (95%CI=0.37–0.73; p<0.01). No obvious heterogeneity was observed (I²=0%, p=0.64 (Figure 3).

The CA 19-9 outcome was reported in three studies. The pooled results showed significantly lower levels of CA 19-9 in patients receiving IRE and immunotherapy compared to those receiving IRE alone (mean difference [MD]: −70.18 U/L; 95%CI=-121.07 – −19.29; p<0.01). However, the heterogeneity was considerable, with I²=98% and p<0.01 (Figure 4). This heterogeneity might be explained by the different CA 19-9 testing intervals implemented in each study during the follow-up period. Lin et al.⁽¹⁸⁾ reported results for CA 19-9 on day 90 after the intervention; Lin et al.⁽¹⁷⁾ reported results on days 1, 7, and 30; and Pan et al.⁽²⁸⁾ reported CA 19-9 levels on days 1, 7, and 30. Lin et al. and Pan et al. found that CA 19-9 levels remained high on days 1 and 7, but dropped by day 30 in both groups. All three studies reported that CA 19-9 levels were lower in the IRE plus immunotherapy group than in the IRE alone group on days 30 or 90. This heterogeneity might also be attributed to differences in the baseline levels of CA 19-9 in patients before intervention in each study.

The overall analysis of AEs is presented in table 3, which shows no significant differences in AEs between groups. No significant difference in the occurrence of AEs such as nausea and vomiting (OR=1.58; 95%CI=0.71–3.49; p=0.26) (Figure 5) and gastroparesis (OR=0.88; 95%CI=0.23–3.40; p=0.85) (Figure 6) was observed between the groups. No heterogeneity was found, as indicated by I²=0% and p=0.93 for nausea and vomiting, and I²=0% and p=0.58 for gastroparesis. No deaths related to the procedure occurred during follow-up.

Quality ratings for the included studies ranged from "serious risk of bias" to "some concerns." Pan et al.⁽²⁸⁾ reported "some concerns" regarding the overall risk of bias due to potential issues with the blinding of participants, personnel, and outcome assessors. Lin et al.⁽¹⁸⁾ also reported "some concerns" related to the lack of explicit mention of allocation concealment in the randomization process, the potential for performance bias given the nature of the intervention, and the potential for detection bias from assessors aware of the intervention.⁽²¹⁾ Lin et al.⁽¹⁷⁾ and He et al.⁽²⁹⁾ presented a "serious risk of bias" owing to confounding factors, as at least one confounder was not adequately measured.⁽²²⁾ Individual appraisals of each study included in the meta-analysis are shown in (Table 2S and 3S, Supplementary Material).