The following electronic databases were searched (from database inception to May 2023): MEDLINE (OVID), Embase (Elsevier), the Cochrane Central Register of Controlled Trials (Wiley) and CINAHL (EBSCOhost). The search strategy was initially developed for MEDLINE and then adapted for each of the other databases. It included both controlled vocabulary and text-word terms, and also commercial names of devices related to CSP. No time or language limits were imposed. The search strategies are available in Supplementary Table S1.

Finally, the reference lists of all pertinent papers were examined to identify additional studies that could meet the selection criteria but were not retrieved through the electronic search in biomedical databases.

Studies were eligible for inclusion if they fulfilled the following criteria: a.

Type of study: Randomized controlled trials (RCTs) and full economic evaluations (EE) were included. If RCTs were not available, non-randomized controlled trials (nRCTs) were considered. In the absence of these, observational studies with a control group were considered. Depending on the quality and quantity of the EE identified, cost-consequences analysis and partial economic evaluations for Spain were also considered for inclusion.

Two reviewers independently screened the titles and abstracts of the references identified by the electronic search. The full texts of studies meeting the predefined selection criteria were thoroughly examined and evaluated for inclusion. Any uncertainties or discrepancies between reviewers were resolved through discussion or consultation with a third reviewer, ultimately reaching a consensus.

Data extraction and the assessment of risk of bias were carried out independently and simultaneously by two reviewers. Any discrepancies were resolved through consultation with a third reviewer. An Excel-based data extraction form was prepared by the authors, subjected to pilot testing on two studies, and refined accordingly. The extracted data encompassed various aspects, including general study characteristics (e.g., first author, publication year, country, funding, and conflicts of interest), design and methodology details (e.g., objective, number of centers and duration of follow-up), sample characteristics (e.g., age, sex, and pathology), intervention and comparator details, outcomes, and results.

The Cochrane risk of bias tool for ECAs (RoB 2) was applied to evaluate the risk of bias in RCT (18). The Risk Of Bias In Non-randomised Studies—of Interventions (ROBINS-I) (19) and—of Exposures (ROBINS-E) (20) were used to assess the risk of bias in nRCT and observational studies.

The Drummond checklist (21) and the Critical appraisal tools (FLC 3.0) (22) for Spanish studies were used to evaluate the methodological quality of EE.

Following the Cochrane Collaboration recommendations (16), the presence of publication bias was assessed by computing the Egger's test, with the statistical significance level set at 0.05. The analysis used the metabias commands in the Stata Statistical Software (STATA 17, StataCorp. 2021. Stata Statistical Software: Release 17. College Station, TX: StataCorp LLC).

The quantitative synthesis of results was conducted through meta-analysis using the Review Manager software (RevMan, version 5.4.1. Copenhagen: The Nordic Cochrane Center, The Cochrane Collaboration, 2020). The Mantel-Haenszel method was applied to estimate the pooled risk ratio (RR) for each dichotomous variable, with continuity correction applied for studies with zero events in one or both groups, the generic inverse variance method was used, along with the mean difference (MD) or standardized mean difference (SMD) to combine continuous variables (23).

Heterogeneity was assessed using Higgins’ I² statistic. In cases of heterogeneity (I² ≥ 50% or p < 0.1), meta-analyses were conducted using a random-effects model. A sensitivity analysis was performed by systematically omitting each study individually to assess the stability of the overall effect estimate. In the absence of both clinical and statistical heterogeneity, a fixed-effect model was used. The pacing modality (HBP or LBBP) was extracted for each study to explore potential differences in outcomes, as HBP procedures are generally associated with longer procedural and fluoroscopy times compared to LBBP. When not explicitly reported, we inferred this information from the study descriptions.

The certainty of evidence for all outcomes was assessed using the GRADE approach, which considered key outcomes across the following domains: risk of bias, inconsistency, indirectness, imprecision, and publication bias (24). The GRADEpro app was used to assign ratings to the evidence and generate the GRADE evidence profile. The certainty of evidence was categorized as high (indicating a high level of confidence that the true effect aligns closely with the estimated effect), moderate, low, or very low (indicating minimal confidence in the estimated effect).

Data on resource use and costs were provided by the manufacturer Abbott Laboratories, based on Ensite X EAM.

In both alternatives, CSP implantation requires the same resources (pulse generator, introducers, electrodes, measurement cables, and implant tool) and equipment (ultrasound, catheter and polygraph). Therefore, the main difference between the alternatives lies in the use of the EAM vs. fluoroscopy. These common costs were not included in the analysis, as their consideration would not impact the incremental cost. The remaining costs are expressed in euros (excluding VAT) for the year 2023.

The same procedural cost of a single-chamber pacemaker was considered for both strategies, with additional costs for the EnSite X patches and technical support for implantation with the EAM. Given that the EAM is currently used for other types of interventions, it was assumed in the base case that the acquisition of the equipment is not necessary.

Regarding fluoroscopy, although the technique based on the EAM aims to minimize its use, the administration of a small dose may be necessary, so this cost is included in both the comparator and the evaluated strategy. The SR showed significant differences in administration time between both technologies, so the cost per minute of fluoroscopy used was estimated. The cost of acquiring the x-ray tube necessary to administer fluoroscopy was annualized to estimate a value distributed over time, by calculating the equivalent annual cost using the following formulas (21):K−S(1+r)nE=A(n,r)¯A(n,r)=1−(1+r)−nrWhere S is the resale value, n the years of useful life of the equipment, r the interest rate (3%, according to the recommendations of López-Bastida et al. (27), A(n, r) is the annuity factor (n years at an interest rate r), K the acquisition price or initial disbursement and E the equivalent annual cost (see Supplementary Table S2.1).

The equivalent annual cost was divided by the total time of the interventions in a year to calculate the cost per minute. The average time of implantation with fluoroscopy (obtained from the SR of effectiveness and safety) was used, given that there are no significant differences in total duration between the two types of interventions. The authors used the total time of the intervention because we assumed the x-ray could not be used for other interventions during this whole time.

Finally, the cost per minute was multiplied by the administration time of fluoroscopy in each strategy.

A one-way deterministic sensitivity analysis was performed, varying the key parameters by ±20% of the mean and adjusting the target population by ±50%. The assumption that fluoroscopy was not used in EAM was also analyzed.

Additionally, a scenario analysis was designed in which an EAM is acquired. In this case, the equivalent annual cost of acquiring the equipment was calculated and divided by the number of patients treated in a year (Supplementary Table S2.2), to determine the equivalent cost per patient.

Furthermore, according to the Abbott Laboratories company, the equipment has an annual maintenance cost of €45,000, which was applied from the second year (after the 1-year warranty provided by the company expired) until the end of the machine's useful life (5 years). In order to calculate the maintenance cost per patient, the annual maintenance costs over the equipment's useful life were summed and divided by 1,035, corresponding to the number of patients treated during this period (=207 patients/year × 5 years of useful life).

Details of the included studies are summarized in Table 1. All studies were comparative observational studies, most with prospective recruitment and follow-up ranging from 1 to 11 months. These studies were published between 2019 and 2023 and spanned various geographic locations, including China (28, 32), the United States (14, 30), Germany (15), India (29), and Italy (31). The EAMs used were KODEX-EPD (28, 31, 32), ENSITE (15, 30) and CARTO 3 (14); one study did not report the EAM used (29).

Reported outcome measures included successful procedure, procedural details, pacing parameters, left ventricular ejection fraction, and complications. A total of 231 patients were enrolled across the seven studies, with an average sample size of 32 patients, ranging from 17 to 54 participants in the individual studies. Of the total recruits, 80 were women (35%), with a mean age of 58.5 years (SD = 11). Among the recruited patients, 48.9% received pacemakers for AVB and 30.4% for SND.

The results of the risk of bias assessment in the included studies are shown in Figure 2. Five of the seven studies showed a serious risk of bias in the domain addressing confounding variables, while the remaining two demonstrated a critical risk of bias. In the other domains, most studies showed a low risk of bias, except for those conducted by Gupta et al. (29), Jiménez et al. (30), and Scarà et al. (31).

Gupta et al. (29) raised concerns about a serious risk of bias in the domain related to the classification of interventions and lacked clear information to evaluate domains 4, 5, and 7 (30) presented a moderate risk of bias in the participant selection domain, and Scarà et al. (31) showed a moderate risk of bias due to the selection of reported results.

A funnel plot analysis and Egger's test could not be performed because the minimum number of studies required to assess the publication bias for any of the outcomes was not reached (n = 10).

The results of the meta-analysis conducted are available in the Supplementary Table S4.

Successful procedure. No statistically significant differences were observed between EAM and fluoroscopy (RR = 0.98; 95% CI: 0.92–1.05; p = 0.59; k = 5; I² = 0%) (see Figure 3).

Procedural time [minutes (min.)]. No statistically significant differences were observed between EAM and fluoroscopy (MD = −2.66 min.; 95% CI: −16.13–10.81; p = 0.70; k = 7; I² = 77%). In the subgroup analysis, no statistically significant differences were observed between HBP, LBPP and HBP/LBPP subgroups (test for subgroup differences: Chi² = 2.16; df = 2; p = 0.34; I² = 7.6%) (See Figure 4). In the sensitivity analysis, heterogeneity was reduced to 12% when the study by Sharma et al. (14) was excluded, and the result became statistically significant. A reduction in procedural time was observed in favor of the EAM (MD = −8.37 min.; 95% CI: −15.19 to −1.55; p = 0.02; k = 6, I² = 12%) and no statistically significant differences were observed between HBP, LBPP and HBP/LBPP subgroups (test for subgroup differences: Chi² = 0.98; df = 2; p = 0.6; I² = 0%).

Total fluoroscopy time (min.). Statistically significant differences were observed in favor of EAM. Fluoroscopy time was shorter in the EAM group (MD = −9.87 min.; 95% CI: −14.20 to −5.53; p < 0.001; k = 7; I² = 95%). In the subgroup analysis, no statistically significant differences were observed between HBP, LBPP and HBP/LBPP subgroups (test for subgroup differences: Chi² = 5.02; df = 2; p = 0.08; I² = 60.1%) (see Figure 5). In the sensitivity analysis, none of the studies appeared to be responsible for the heterogeneity; study-by-study exclusion did not reduce heterogeneity.

His lead fluoroscopy time (min.). A reduction was observed in favor of the EAM group (MD = −8.08 min.; 95% CI: −10.36 to −5.81; p < 0.001; k = 2; I² = 0%) (see Supplementary Figure S4.1).

Total fluoroscopy dose (mGy). No statistically significant differences were observed between the groups (MD = −55.28 mGy; 95% CI: −118.67 to 8.12; p = 0.09; k = 3; I² = 97%) (see Supplementary Figure S4.2). In the sensitivity analysis, heterogeneity was reduced to 18% when the study by (28) was excluded, and the result became statistically significant (MD = −90.23 mGy; 95% CI: −124.90 to −55.55; p < 0.001; k = 2; I² = 18%). Similarly, heterogeneity was reduced to 0% when the study by (32) was excluded (MD = −22.35 mGy; 95% CI: −29.28 to −15.41; p < 0.001; k = 2; I² = 0%).

His lead fluoroscopy dose (mGy). A reduction of 17 mGy was observed in favor of the EAM group (MD = −17.21 mGy; 95% CI: −24.08 to −10.34; k = 1; p < 0.001) (see Supplementary Figure S4.3).

Paced QRS duration [time in milliseconds (ms) from the beginning of the Q or R wave to the end of the R or S wave]. No significant differences were observed between EAM and fluoroscopy (MD = −3.92 ms; 95% CI: −9.43 to 1.60; p = 0.16; k = 4; I² = 0%) (see Supplementary Figure S4.4).

His-ventricular interval duration (period between the onset of the His bundle spike and the earliest ventricular activation) (ms). No significant differences were observed between EAM and fluoroscopy (MD = −1.04 ms; 95% CI: −7.19 to 5.12; p = 0.74; k = 2; I² = 0%) (see Supplementary Figure S4.5).

Capture threshold [the minimum amount of energy from an electrical impulse that is required to produce myocardial contraction (V/1 ms) during implantation]. No statistically significant differences were observed between EAM and fluoroscopy (MD = −0.02 V/1 ms; 95% CI: −0.15 to 0.12; p = 0.82; k = 6; I² = 63%) (see Supplementary Figure S4.6).

Impedance [assessing the integrity of pacemaker pacing and sensing leads (Ω)]. No statistically significant differences were observed between EAM and fluoroscopy (MD = 16.51 Ω; 95% CI: −15.28 to 48.30; p = 0.31; k = 5; I² = 16%) (see Supplementary Figure S4.7).

R-wave amplitude [the amplitude of the intracardiac signal detected by the device (mV)]. No statistically significant differences were observed between EAM and fluoroscopy (MD = 0.46 mV; 95% CI: −0.15 to 1.07; p = 0.14; k = 6; I² = 0%) (see Supplementary Figure S4.8).

Capture threshold (follow-up ranged from 1 to 6 months). No statistically significant differences were observed between EAM and fluoroscopy (MD = −0.08 V/1 ms; 95% CI: −0.23 to 0.07; p = 0.32; k = 4; I² = 24%) (see Supplementary Figure S4.9).

Impedance (follow-up ranged from 3 to 6 months). No statistically significant differences were observed between EAM and fluoroscopy (MD = −7.31 Ω; 95% CI: −50.17 to 35.55; p = 0.74; k = 2; I² = 0%) (see Supplementary Figure S4.10).

R-wave amplitude (follow-up ranged from 1 to 6 months). No statistically significant differences were observed between EAM and fluoroscopy (MD = −0.26 mV; 95% CI: −0.16 to 0.65; p = 0.58; k = 6; I² = 0%) (see Supplementary Figure S4.11).

Left ventricular ejection fraction (LVEF) (%) at follow-up (3–6 months). No statistically significant differences were observed between EAM and fluoroscopy (MD = 0.44%; 95% CI: −3.85 to 4.74; p = 0.84; k = 2; I² = 57%) (see Supplementary Figure S4.12).

Five of the included studies reported immediate procedure-related complications, with no complications observed in the patients (see Table 2). Regarding other complications recorded during follow-up (ranging from 1 to 6 months), only one case of lead displacement and one case of increased capture threshold were observed in the fluoroscopy group.

The certainty of evidence for all key outcomes was considered very low due to the serious risk of bias in the included studies (see evidence profile in Supplementary Table S5).

Parameters used in the cost analysis can be consulted in Table 3.

The results of the analysis show that the implantation of CSP using an Ensite X EAM would entail a cost of €4,248.86 per patient, while the cost of using the fluoroscopy technique is €2,851.06, resulting in an increase of €1,397.81 per patient (see Table 4).

The sensitivity analysis shows that the incremental cost per patient can vary between €548.21 and €2,247.40 (Table 5). This variation is mainly due to the differences in the cost of the procedure with the EAM technique, specifically the costs associated with Ensite patches and technical support, which are the differential resources between the two compared techniques.

Finally, if a hospital needed to acquire the necessary equipment for EAM, the incremental cost would increase to €1,738.74 per patient (Table 6).